linux/fs/afs/yfsclient.c

2205 lines
54 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-or-later
/* YFS File Server client stubs
*
* Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*/
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/sched.h>
#include <linux/circ_buf.h>
#include <linux/iversion.h>
#include "internal.h"
#include "afs_fs.h"
#include "xdr_fs.h"
#include "protocol_yfs.h"
static const struct afs_fid afs_zero_fid;
static inline void afs_use_fs_server(struct afs_call *call, struct afs_cb_interest *cbi)
{
call->cbi = afs_get_cb_interest(cbi);
}
#define xdr_size(x) (sizeof(*x) / sizeof(__be32))
static void xdr_decode_YFSFid(const __be32 **_bp, struct afs_fid *fid)
{
const struct yfs_xdr_YFSFid *x = (const void *)*_bp;
fid->vid = xdr_to_u64(x->volume);
fid->vnode = xdr_to_u64(x->vnode.lo);
fid->vnode_hi = ntohl(x->vnode.hi);
fid->unique = ntohl(x->vnode.unique);
*_bp += xdr_size(x);
}
static __be32 *xdr_encode_u32(__be32 *bp, u32 n)
{
*bp++ = htonl(n);
return bp;
}
static __be32 *xdr_encode_u64(__be32 *bp, u64 n)
{
struct yfs_xdr_u64 *x = (void *)bp;
*x = u64_to_xdr(n);
return bp + xdr_size(x);
}
static __be32 *xdr_encode_YFSFid(__be32 *bp, struct afs_fid *fid)
{
struct yfs_xdr_YFSFid *x = (void *)bp;
x->volume = u64_to_xdr(fid->vid);
x->vnode.lo = u64_to_xdr(fid->vnode);
x->vnode.hi = htonl(fid->vnode_hi);
x->vnode.unique = htonl(fid->unique);
return bp + xdr_size(x);
}
static size_t xdr_strlen(unsigned int len)
{
return sizeof(__be32) + round_up(len, sizeof(__be32));
}
static __be32 *xdr_encode_string(__be32 *bp, const char *p, unsigned int len)
{
bp = xdr_encode_u32(bp, len);
bp = memcpy(bp, p, len);
if (len & 3) {
unsigned int pad = 4 - (len & 3);
memset((u8 *)bp + len, 0, pad);
len += pad;
}
return bp + len / sizeof(__be32);
}
static s64 linux_to_yfs_time(const struct timespec64 *t)
{
/* Convert to 100ns intervals. */
return (u64)t->tv_sec * 10000000 + t->tv_nsec/100;
}
static __be32 *xdr_encode_YFSStoreStatus_mode(__be32 *bp, mode_t mode)
{
struct yfs_xdr_YFSStoreStatus *x = (void *)bp;
x->mask = htonl(AFS_SET_MODE);
x->mode = htonl(mode & S_IALLUGO);
x->mtime_client = u64_to_xdr(0);
x->owner = u64_to_xdr(0);
x->group = u64_to_xdr(0);
return bp + xdr_size(x);
}
static __be32 *xdr_encode_YFSStoreStatus_mtime(__be32 *bp, const struct timespec64 *t)
{
struct yfs_xdr_YFSStoreStatus *x = (void *)bp;
s64 mtime = linux_to_yfs_time(t);
x->mask = htonl(AFS_SET_MTIME);
x->mode = htonl(0);
x->mtime_client = u64_to_xdr(mtime);
x->owner = u64_to_xdr(0);
x->group = u64_to_xdr(0);
return bp + xdr_size(x);
}
/*
* Convert a signed 100ns-resolution 64-bit time into a timespec.
*/
static struct timespec64 yfs_time_to_linux(s64 t)
{
struct timespec64 ts;
u64 abs_t;
/*
* Unfortunately can not use normal 64 bit division on 32 bit arch, but
* the alternative, do_div, does not work with negative numbers so have
* to special case them
*/
if (t < 0) {
abs_t = -t;
ts.tv_nsec = (time64_t)(do_div(abs_t, 10000000) * 100);
ts.tv_nsec = -ts.tv_nsec;
ts.tv_sec = -abs_t;
} else {
abs_t = t;
ts.tv_nsec = (time64_t)do_div(abs_t, 10000000) * 100;
ts.tv_sec = abs_t;
}
return ts;
}
static struct timespec64 xdr_to_time(const struct yfs_xdr_u64 xdr)
{
s64 t = xdr_to_u64(xdr);
return yfs_time_to_linux(t);
}
static void yfs_check_req(struct afs_call *call, __be32 *bp)
{
size_t len = (void *)bp - call->request;
if (len > call->request_size)
pr_err("kAFS: %s: Request buffer overflow (%zu>%u)\n",
call->type->name, len, call->request_size);
else if (len < call->request_size)
pr_warn("kAFS: %s: Request buffer underflow (%zu<%u)\n",
call->type->name, len, call->request_size);
}
/*
* Dump a bad file status record.
*/
static void xdr_dump_bad(const __be32 *bp)
{
__be32 x[4];
int i;
pr_notice("YFS XDR: Bad status record\n");
for (i = 0; i < 6 * 4 * 4; i += 16) {
memcpy(x, bp, 16);
bp += 4;
pr_notice("%03x: %08x %08x %08x %08x\n",
i, ntohl(x[0]), ntohl(x[1]), ntohl(x[2]), ntohl(x[3]));
}
memcpy(x, bp, 8);
pr_notice("0x60: %08x %08x\n", ntohl(x[0]), ntohl(x[1]));
}
/*
* Decode a YFSFetchStatus block
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
static int xdr_decode_YFSFetchStatus(const __be32 **_bp,
struct afs_call *call,
struct afs_status_cb *scb)
{
const struct yfs_xdr_YFSFetchStatus *xdr = (const void *)*_bp;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_file_status *status = &scb->status;
u32 type;
int ret;
status->abort_code = ntohl(xdr->abort_code);
if (status->abort_code != 0) {
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
if (status->abort_code == VNOVNODE)
status->nlink = 0;
scb->have_error = true;
goto good;
}
type = ntohl(xdr->type);
switch (type) {
case AFS_FTYPE_FILE:
case AFS_FTYPE_DIR:
case AFS_FTYPE_SYMLINK:
status->type = type;
break;
default:
goto bad;
}
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
status->nlink = ntohl(xdr->nlink);
status->author = xdr_to_u64(xdr->author);
status->owner = xdr_to_u64(xdr->owner);
status->caller_access = ntohl(xdr->caller_access); /* Ticket dependent */
status->anon_access = ntohl(xdr->anon_access);
status->mode = ntohl(xdr->mode) & S_IALLUGO;
status->group = xdr_to_u64(xdr->group);
status->lock_count = ntohl(xdr->lock_count);
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
status->mtime_client = xdr_to_time(xdr->mtime_client);
status->mtime_server = xdr_to_time(xdr->mtime_server);
status->size = xdr_to_u64(xdr->size);
status->data_version = xdr_to_u64(xdr->data_version);
scb->have_status = true;
good:
ret = 0;
advance:
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
*_bp += xdr_size(xdr);
return ret;
bad:
xdr_dump_bad(*_bp);
ret = afs_protocol_error(call, -EBADMSG, afs_eproto_bad_status);
goto advance;
}
/*
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
* Decode a YFSCallBack block
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
static void xdr_decode_YFSCallBack(const __be32 **_bp,
struct afs_call *call,
struct afs_status_cb *scb)
{
struct yfs_xdr_YFSCallBack *x = (void *)*_bp;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_callback *cb = &scb->callback;
ktime_t cb_expiry;
cb_expiry = call->reply_time;
cb_expiry = ktime_add(cb_expiry, xdr_to_u64(x->expiration_time) * 100);
cb->expires_at = ktime_divns(cb_expiry, NSEC_PER_SEC);
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
scb->have_cb = true;
*_bp += xdr_size(x);
}
/*
* Decode a YFSVolSync block
*/
static void xdr_decode_YFSVolSync(const __be32 **_bp,
struct afs_volsync *volsync)
{
struct yfs_xdr_YFSVolSync *x = (void *)*_bp;
u64 creation;
if (volsync) {
creation = xdr_to_u64(x->vol_creation_date);
do_div(creation, 10 * 1000 * 1000);
volsync->creation = creation;
}
*_bp += xdr_size(x);
}
/*
* Encode the requested attributes into a YFSStoreStatus block
*/
static __be32 *xdr_encode_YFS_StoreStatus(__be32 *bp, struct iattr *attr)
{
struct yfs_xdr_YFSStoreStatus *x = (void *)bp;
s64 mtime = 0, owner = 0, group = 0;
u32 mask = 0, mode = 0;
mask = 0;
if (attr->ia_valid & ATTR_MTIME) {
mask |= AFS_SET_MTIME;
mtime = linux_to_yfs_time(&attr->ia_mtime);
}
if (attr->ia_valid & ATTR_UID) {
mask |= AFS_SET_OWNER;
owner = from_kuid(&init_user_ns, attr->ia_uid);
}
if (attr->ia_valid & ATTR_GID) {
mask |= AFS_SET_GROUP;
group = from_kgid(&init_user_ns, attr->ia_gid);
}
if (attr->ia_valid & ATTR_MODE) {
mask |= AFS_SET_MODE;
mode = attr->ia_mode & S_IALLUGO;
}
x->mask = htonl(mask);
x->mode = htonl(mode);
x->mtime_client = u64_to_xdr(mtime);
x->owner = u64_to_xdr(owner);
x->group = u64_to_xdr(group);
return bp + xdr_size(x);
}
/*
* Decode a YFSFetchVolumeStatus block.
*/
static void xdr_decode_YFSFetchVolumeStatus(const __be32 **_bp,
struct afs_volume_status *vs)
{
const struct yfs_xdr_YFSFetchVolumeStatus *x = (const void *)*_bp;
u32 flags;
vs->vid = xdr_to_u64(x->vid);
vs->parent_id = xdr_to_u64(x->parent_id);
flags = ntohl(x->flags);
vs->online = flags & yfs_FVSOnline;
vs->in_service = flags & yfs_FVSInservice;
vs->blessed = flags & yfs_FVSBlessed;
vs->needs_salvage = flags & yfs_FVSNeedsSalvage;
vs->type = ntohl(x->type);
vs->min_quota = 0;
vs->max_quota = xdr_to_u64(x->max_quota);
vs->blocks_in_use = xdr_to_u64(x->blocks_in_use);
vs->part_blocks_avail = xdr_to_u64(x->part_blocks_avail);
vs->part_max_blocks = xdr_to_u64(x->part_max_blocks);
vs->vol_copy_date = xdr_to_u64(x->vol_copy_date);
vs->vol_backup_date = xdr_to_u64(x->vol_backup_date);
*_bp += sizeof(*x) / sizeof(__be32);
}
/*
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
* Deliver a reply that's a status, callback and volsync.
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
static int yfs_deliver_fs_status_cb_and_volsync(struct afs_call *call)
{
const __be32 *bp;
int ret;
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
/* unmarshall the reply once we've received all of it */
bp = call->buffer;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
xdr_decode_YFSCallBack(&bp, call, call->out_scb);
xdr_decode_YFSVolSync(&bp, call->out_volsync);
_leave(" = 0 [done]");
return 0;
}
/*
* Deliver reply data to operations that just return a file status and a volume
* sync record.
*/
static int yfs_deliver_status_and_volsync(struct afs_call *call)
{
const __be32 *bp;
int ret;
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
bp = call->buffer;
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
xdr_decode_YFSVolSync(&bp, call->out_volsync);
_leave(" = 0 [done]");
return 0;
}
/*
* YFS.FetchStatus operation type
*/
static const struct afs_call_type yfs_RXYFSFetchStatus_vnode = {
.name = "YFS.FetchStatus(vnode)",
.op = yfs_FS_FetchStatus,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
.deliver = yfs_deliver_fs_status_cb_and_volsync,
.destructor = afs_flat_call_destructor,
};
/*
* Fetch the status information for a file.
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
int yfs_fs_fetch_file_status(struct afs_fs_cursor *fc, struct afs_status_cb *scb,
struct afs_volsync *volsync)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter(",%x,{%llx:%llu},,",
key_serial(fc->key), vnode->fid.vid, vnode->fid.vnode);
call = afs_alloc_flat_call(net, &yfs_RXYFSFetchStatus_vnode,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSCallBack) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call) {
fc->ac.error = -ENOMEM;
return -ENOMEM;
}
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
call->out_volsync = volsync;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSFETCHSTATUS);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to an YFS.FetchData64.
*/
static int yfs_deliver_fs_fetch_data64(struct afs_call *call)
{
struct afs_read *req = call->read_request;
const __be32 *bp;
unsigned int size;
int ret;
_enter("{%u,%zu/%llu}",
call->unmarshall, iov_iter_count(call->iter), req->actual_len);
switch (call->unmarshall) {
case 0:
req->actual_len = 0;
req->index = 0;
req->offset = req->pos & (PAGE_SIZE - 1);
afs_extract_to_tmp64(call);
call->unmarshall++;
/* Fall through */
/* extract the returned data length */
case 1:
_debug("extract data length");
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
req->actual_len = be64_to_cpu(call->tmp64);
_debug("DATA length: %llu", req->actual_len);
req->remain = min(req->len, req->actual_len);
if (req->remain == 0)
goto no_more_data;
call->unmarshall++;
begin_page:
ASSERTCMP(req->index, <, req->nr_pages);
if (req->remain > PAGE_SIZE - req->offset)
size = PAGE_SIZE - req->offset;
else
size = req->remain;
call->bvec[0].bv_len = size;
call->bvec[0].bv_offset = req->offset;
call->bvec[0].bv_page = req->pages[req->index];
iov_iter_bvec(&call->def_iter, READ, call->bvec, 1, size);
ASSERTCMP(size, <=, PAGE_SIZE);
/* Fall through */
/* extract the returned data */
case 2:
_debug("extract data %zu/%llu",
iov_iter_count(call->iter), req->remain);
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
req->remain -= call->bvec[0].bv_len;
req->offset += call->bvec[0].bv_len;
ASSERTCMP(req->offset, <=, PAGE_SIZE);
if (req->offset == PAGE_SIZE) {
req->offset = 0;
req->index++;
if (req->remain > 0)
goto begin_page;
}
ASSERTCMP(req->remain, ==, 0);
if (req->actual_len <= req->len)
goto no_more_data;
/* Discard any excess data the server gave us */
afs_extract_discard(call, req->actual_len - req->len);
call->unmarshall = 3;
/* Fall through */
case 3:
_debug("extract discard %zu/%llu",
iov_iter_count(call->iter), req->actual_len - req->len);
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
no_more_data:
call->unmarshall = 4;
afs_extract_to_buf(call,
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSCallBack) +
sizeof(struct yfs_xdr_YFSVolSync));
/* Fall through */
/* extract the metadata */
case 4:
ret = afs_extract_data(call, false);
if (ret < 0)
return ret;
bp = call->buffer;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSCallBack(&bp, call, call->out_scb);
xdr_decode_YFSVolSync(&bp, call->out_volsync);
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
req->data_version = call->out_scb->status.data_version;
req->file_size = call->out_scb->status.size;
call->unmarshall++;
/* Fall through */
case 5:
break;
}
for (; req->index < req->nr_pages; req->index++) {
if (req->offset < PAGE_SIZE)
zero_user_segment(req->pages[req->index],
req->offset, PAGE_SIZE);
req->offset = 0;
}
afs: Don't unlock fetched data pages until the op completes successfully Don't call req->page_done() on each page as we finish filling it with the data coming from the network. Whilst this might speed up the application a bit, it's a problem if there's a network failure and the operation has to be reissued. If this happens, an oops occurs because afs_readpages_page_done() clears the pointer to each page it unlocks and when a retry happens, the pointers to the pages it wants to fill are now NULL (and the pages have been unlocked anyway). Instead, wait till the operation completes successfully and only then release all the pages after clearing any terminal gap (the server can give us less data than we requested as we're allowed to ask for more than is available). KASAN produces a bug like the following, and even without KASAN, it can oops and panic. BUG: KASAN: wild-memory-access in _copy_to_iter+0x323/0x5f4 Write of size 1404 at addr 0005088000000000 by task md5sum/5235 CPU: 0 PID: 5235 Comm: md5sum Not tainted 5.7.0-rc3-fscache+ #250 Hardware name: ASUS All Series/H97-PLUS, BIOS 2306 10/09/2014 Call Trace: memcpy+0x39/0x58 _copy_to_iter+0x323/0x5f4 __skb_datagram_iter+0x89/0x2a6 skb_copy_datagram_iter+0x129/0x135 rxrpc_recvmsg_data.isra.0+0x615/0xd42 rxrpc_kernel_recv_data+0x1e9/0x3ae afs_extract_data+0x139/0x33a yfs_deliver_fs_fetch_data64+0x47a/0x91b afs_deliver_to_call+0x304/0x709 afs_wait_for_call_to_complete+0x1cc/0x4ad yfs_fs_fetch_data+0x279/0x288 afs_fetch_data+0x1e1/0x38d afs_readpages+0x593/0x72e read_pages+0xf5/0x21e __do_page_cache_readahead+0x128/0x23f ondemand_readahead+0x36e/0x37f generic_file_buffered_read+0x234/0x680 new_sync_read+0x109/0x17e vfs_read+0xe6/0x138 ksys_read+0xd8/0x14d do_syscall_64+0x6e/0x8a entry_SYSCALL_64_after_hwframe+0x49/0xb3 Fixes: 196ee9cd2d04 ("afs: Make afs_fs_fetch_data() take a list of pages") Fixes: 30062bd13e36 ("afs: Implement YFS support in the fs client") Signed-off-by: David Howells <dhowells@redhat.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-05-18 04:21:05 +08:00
if (req->page_done)
for (req->index = 0; req->index < req->nr_pages; req->index++)
req->page_done(req);
_leave(" = 0 [done]");
return 0;
}
static void yfs_fetch_data_destructor(struct afs_call *call)
{
afs_put_read(call->read_request);
afs_flat_call_destructor(call);
}
/*
* YFS.FetchData64 operation type
*/
static const struct afs_call_type yfs_RXYFSFetchData64 = {
.name = "YFS.FetchData64",
.op = yfs_FS_FetchData64,
.deliver = yfs_deliver_fs_fetch_data64,
.destructor = yfs_fetch_data_destructor,
};
/*
* Fetch data from a file.
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
int yfs_fs_fetch_data(struct afs_fs_cursor *fc, struct afs_status_cb *scb,
struct afs_read *req)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter(",%x,{%llx:%llu},%llx,%llx",
key_serial(fc->key), vnode->fid.vid, vnode->fid.vnode,
req->pos, req->len);
call = afs_alloc_flat_call(net, &yfs_RXYFSFetchData64,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_u64) * 2,
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSCallBack) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
call->out_volsync = NULL;
call->read_request = afs_get_read(req);
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSFETCHDATA64);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
bp = xdr_encode_u64(bp, req->pos);
bp = xdr_encode_u64(bp, req->len);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data for YFS.CreateFile or YFS.MakeDir.
*/
static int yfs_deliver_fs_create_vnode(struct afs_call *call)
{
const __be32 *bp;
int ret;
_enter("{%u}", call->unmarshall);
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
/* unmarshall the reply once we've received all of it */
bp = call->buffer;
xdr_decode_YFSFid(&bp, call->out_fid);
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_dir_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSCallBack(&bp, call, call->out_scb);
xdr_decode_YFSVolSync(&bp, call->out_volsync);
_leave(" = 0 [done]");
return 0;
}
/*
* FS.CreateFile and FS.MakeDir operation type
*/
static const struct afs_call_type afs_RXFSCreateFile = {
.name = "YFS.CreateFile",
.op = yfs_FS_CreateFile,
.deliver = yfs_deliver_fs_create_vnode,
.destructor = afs_flat_call_destructor,
};
/*
* Create a file.
*/
int yfs_fs_create_file(struct afs_fs_cursor *fc,
const char *name,
umode_t mode,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *dvnode_scb,
struct afs_fid *newfid,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *new_scb)
{
struct afs_vnode *dvnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(dvnode);
size_t namesz, reqsz, rplsz;
__be32 *bp;
_enter("");
namesz = strlen(name);
reqsz = (sizeof(__be32) +
sizeof(__be32) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(namesz) +
sizeof(struct yfs_xdr_YFSStoreStatus) +
sizeof(__be32));
rplsz = (sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSCallBack) +
sizeof(struct yfs_xdr_YFSVolSync));
call = afs_alloc_flat_call(net, &afs_RXFSCreateFile, reqsz, rplsz);
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_dir_scb = dvnode_scb;
call->out_fid = newfid;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = new_scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSCREATEFILE);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &dvnode->fid);
bp = xdr_encode_string(bp, name, namesz);
bp = xdr_encode_YFSStoreStatus_mode(bp, mode);
bp = xdr_encode_u32(bp, yfs_LockNone); /* ViceLockType */
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call1(call, &dvnode->fid, name);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
static const struct afs_call_type yfs_RXFSMakeDir = {
.name = "YFS.MakeDir",
.op = yfs_FS_MakeDir,
.deliver = yfs_deliver_fs_create_vnode,
.destructor = afs_flat_call_destructor,
};
/*
* Make a directory.
*/
int yfs_fs_make_dir(struct afs_fs_cursor *fc,
const char *name,
umode_t mode,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *dvnode_scb,
struct afs_fid *newfid,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *new_scb)
{
struct afs_vnode *dvnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(dvnode);
size_t namesz, reqsz, rplsz;
__be32 *bp;
_enter("");
namesz = strlen(name);
reqsz = (sizeof(__be32) +
sizeof(struct yfs_xdr_RPCFlags) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(namesz) +
sizeof(struct yfs_xdr_YFSStoreStatus));
rplsz = (sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSCallBack) +
sizeof(struct yfs_xdr_YFSVolSync));
call = afs_alloc_flat_call(net, &yfs_RXFSMakeDir, reqsz, rplsz);
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_dir_scb = dvnode_scb;
call->out_fid = newfid;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = new_scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSMAKEDIR);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &dvnode->fid);
bp = xdr_encode_string(bp, name, namesz);
bp = xdr_encode_YFSStoreStatus_mode(bp, mode);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call1(call, &dvnode->fid, name);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to a YFS.RemoveFile2 operation.
*/
static int yfs_deliver_fs_remove_file2(struct afs_call *call)
{
struct afs_fid fid;
const __be32 *bp;
int ret;
_enter("{%u}", call->unmarshall);
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
bp = call->buffer;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_dir_scb);
if (ret < 0)
return ret;
xdr_decode_YFSFid(&bp, &fid);
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
/* Was deleted if vnode->status.abort_code == VNOVNODE. */
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSVolSync(&bp, call->out_volsync);
return 0;
}
/*
* YFS.RemoveFile2 operation type.
*/
static const struct afs_call_type yfs_RXYFSRemoveFile2 = {
.name = "YFS.RemoveFile2",
.op = yfs_FS_RemoveFile2,
.deliver = yfs_deliver_fs_remove_file2,
.destructor = afs_flat_call_destructor,
};
/*
* Remove a file and retrieve new file status.
*/
int yfs_fs_remove_file2(struct afs_fs_cursor *fc, struct afs_vnode *vnode,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
const char *name, struct afs_status_cb *dvnode_scb,
struct afs_status_cb *vnode_scb)
{
struct afs_vnode *dvnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(dvnode);
size_t namesz;
__be32 *bp;
_enter("");
namesz = strlen(name);
call = afs_alloc_flat_call(net, &yfs_RXYFSRemoveFile2,
sizeof(__be32) +
sizeof(struct yfs_xdr_RPCFlags) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(namesz),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_dir_scb = dvnode_scb;
call->out_scb = vnode_scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSREMOVEFILE2);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &dvnode->fid);
bp = xdr_encode_string(bp, name, namesz);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call1(call, &dvnode->fid, name);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to a YFS.RemoveFile or YFS.RemoveDir operation.
*/
static int yfs_deliver_fs_remove(struct afs_call *call)
{
const __be32 *bp;
int ret;
_enter("{%u}", call->unmarshall);
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
bp = call->buffer;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_dir_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSVolSync(&bp, call->out_volsync);
return 0;
}
/*
* FS.RemoveDir and FS.RemoveFile operation types.
*/
static const struct afs_call_type yfs_RXYFSRemoveFile = {
.name = "YFS.RemoveFile",
.op = yfs_FS_RemoveFile,
.deliver = yfs_deliver_fs_remove,
.destructor = afs_flat_call_destructor,
};
static const struct afs_call_type yfs_RXYFSRemoveDir = {
.name = "YFS.RemoveDir",
.op = yfs_FS_RemoveDir,
.deliver = yfs_deliver_fs_remove,
.destructor = afs_flat_call_destructor,
};
/*
* remove a file or directory
*/
int yfs_fs_remove(struct afs_fs_cursor *fc, struct afs_vnode *vnode,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
const char *name, bool isdir,
struct afs_status_cb *dvnode_scb)
{
struct afs_vnode *dvnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(dvnode);
size_t namesz;
__be32 *bp;
_enter("");
namesz = strlen(name);
call = afs_alloc_flat_call(
net, isdir ? &yfs_RXYFSRemoveDir : &yfs_RXYFSRemoveFile,
sizeof(__be32) +
sizeof(struct yfs_xdr_RPCFlags) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(namesz),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_dir_scb = dvnode_scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, isdir ? YFSREMOVEDIR : YFSREMOVEFILE);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &dvnode->fid);
bp = xdr_encode_string(bp, name, namesz);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call1(call, &dvnode->fid, name);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to a YFS.Link operation.
*/
static int yfs_deliver_fs_link(struct afs_call *call)
{
const __be32 *bp;
int ret;
_enter("{%u}", call->unmarshall);
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
bp = call->buffer;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_dir_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSVolSync(&bp, call->out_volsync);
_leave(" = 0 [done]");
return 0;
}
/*
* YFS.Link operation type.
*/
static const struct afs_call_type yfs_RXYFSLink = {
.name = "YFS.Link",
.op = yfs_FS_Link,
.deliver = yfs_deliver_fs_link,
.destructor = afs_flat_call_destructor,
};
/*
* Make a hard link.
*/
int yfs_fs_link(struct afs_fs_cursor *fc, struct afs_vnode *vnode,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
const char *name,
struct afs_status_cb *dvnode_scb,
struct afs_status_cb *vnode_scb)
{
struct afs_vnode *dvnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
size_t namesz;
__be32 *bp;
_enter("");
namesz = strlen(name);
call = afs_alloc_flat_call(net, &yfs_RXYFSLink,
sizeof(__be32) +
sizeof(struct yfs_xdr_RPCFlags) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(namesz) +
sizeof(struct yfs_xdr_YFSFid),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_dir_scb = dvnode_scb;
call->out_scb = vnode_scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSLINK);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &dvnode->fid);
bp = xdr_encode_string(bp, name, namesz);
bp = xdr_encode_YFSFid(bp, &vnode->fid);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call1(call, &vnode->fid, name);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to a YFS.Symlink operation.
*/
static int yfs_deliver_fs_symlink(struct afs_call *call)
{
const __be32 *bp;
int ret;
_enter("{%u}", call->unmarshall);
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
/* unmarshall the reply once we've received all of it */
bp = call->buffer;
xdr_decode_YFSFid(&bp, call->out_fid);
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_dir_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSVolSync(&bp, call->out_volsync);
_leave(" = 0 [done]");
return 0;
}
/*
* YFS.Symlink operation type
*/
static const struct afs_call_type yfs_RXYFSSymlink = {
.name = "YFS.Symlink",
.op = yfs_FS_Symlink,
.deliver = yfs_deliver_fs_symlink,
.destructor = afs_flat_call_destructor,
};
/*
* Create a symbolic link.
*/
int yfs_fs_symlink(struct afs_fs_cursor *fc,
const char *name,
const char *contents,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *dvnode_scb,
struct afs_fid *newfid,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *vnode_scb)
{
struct afs_vnode *dvnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(dvnode);
size_t namesz, contents_sz;
__be32 *bp;
_enter("");
namesz = strlen(name);
contents_sz = strlen(contents);
call = afs_alloc_flat_call(net, &yfs_RXYFSSymlink,
sizeof(__be32) +
sizeof(struct yfs_xdr_RPCFlags) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(namesz) +
xdr_strlen(contents_sz) +
sizeof(struct yfs_xdr_YFSStoreStatus),
sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_dir_scb = dvnode_scb;
call->out_fid = newfid;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = vnode_scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSSYMLINK);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &dvnode->fid);
bp = xdr_encode_string(bp, name, namesz);
bp = xdr_encode_string(bp, contents, contents_sz);
bp = xdr_encode_YFSStoreStatus_mode(bp, S_IRWXUGO);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call1(call, &dvnode->fid, name);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to a YFS.Rename operation.
*/
static int yfs_deliver_fs_rename(struct afs_call *call)
{
const __be32 *bp;
int ret;
_enter("{%u}", call->unmarshall);
ret = afs_transfer_reply(call);
if (ret < 0)
return ret;
bp = call->buffer;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_dir_scb);
if (ret < 0)
return ret;
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSVolSync(&bp, call->out_volsync);
_leave(" = 0 [done]");
return 0;
}
/*
* YFS.Rename operation type
*/
static const struct afs_call_type yfs_RXYFSRename = {
.name = "FS.Rename",
.op = yfs_FS_Rename,
.deliver = yfs_deliver_fs_rename,
.destructor = afs_flat_call_destructor,
};
/*
* Rename a file or directory.
*/
int yfs_fs_rename(struct afs_fs_cursor *fc,
const char *orig_name,
struct afs_vnode *new_dvnode,
const char *new_name,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *orig_dvnode_scb,
struct afs_status_cb *new_dvnode_scb)
{
struct afs_vnode *orig_dvnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(orig_dvnode);
size_t o_namesz, n_namesz;
__be32 *bp;
_enter("");
o_namesz = strlen(orig_name);
n_namesz = strlen(new_name);
call = afs_alloc_flat_call(net, &yfs_RXYFSRename,
sizeof(__be32) +
sizeof(struct yfs_xdr_RPCFlags) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(o_namesz) +
sizeof(struct yfs_xdr_YFSFid) +
xdr_strlen(n_namesz),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_dir_scb = orig_dvnode_scb;
call->out_scb = new_dvnode_scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSRENAME);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &orig_dvnode->fid);
bp = xdr_encode_string(bp, orig_name, o_namesz);
bp = xdr_encode_YFSFid(bp, &new_dvnode->fid);
bp = xdr_encode_string(bp, new_name, n_namesz);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call2(call, &orig_dvnode->fid, orig_name, new_name);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* YFS.StoreData64 operation type.
*/
static const struct afs_call_type yfs_RXYFSStoreData64 = {
.name = "YFS.StoreData64",
.op = yfs_FS_StoreData64,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
.deliver = yfs_deliver_status_and_volsync,
.destructor = afs_flat_call_destructor,
};
/*
* Store a set of pages to a large file.
*/
int yfs_fs_store_data(struct afs_fs_cursor *fc, struct address_space *mapping,
pgoff_t first, pgoff_t last,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
unsigned offset, unsigned to,
struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
loff_t size, pos, i_size;
__be32 *bp;
_enter(",%x,{%llx:%llu},,",
key_serial(fc->key), vnode->fid.vid, vnode->fid.vnode);
size = (loff_t)to - (loff_t)offset;
if (first != last)
size += (loff_t)(last - first) << PAGE_SHIFT;
pos = (loff_t)first << PAGE_SHIFT;
pos += offset;
i_size = i_size_read(&vnode->vfs_inode);
if (pos + size > i_size)
i_size = size + pos;
_debug("size %llx, at %llx, i_size %llx",
(unsigned long long)size, (unsigned long long)pos,
(unsigned long long)i_size);
call = afs_alloc_flat_call(net, &yfs_RXYFSStoreData64,
sizeof(__be32) +
sizeof(__be32) +
sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_YFSStoreStatus) +
sizeof(struct yfs_xdr_u64) * 3,
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
call->mapping = mapping;
call->first = first;
call->last = last;
call->first_offset = offset;
call->last_to = to;
call->send_pages = true;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSSTOREDATA64);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
bp = xdr_encode_YFSStoreStatus_mtime(bp, &vnode->vfs_inode.i_mtime);
bp = xdr_encode_u64(bp, pos);
bp = xdr_encode_u64(bp, size);
bp = xdr_encode_u64(bp, i_size);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* YFS.StoreStatus operation type
*/
static const struct afs_call_type yfs_RXYFSStoreStatus = {
.name = "YFS.StoreStatus",
.op = yfs_FS_StoreStatus,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
.deliver = yfs_deliver_status_and_volsync,
.destructor = afs_flat_call_destructor,
};
static const struct afs_call_type yfs_RXYFSStoreData64_as_Status = {
.name = "YFS.StoreData64",
.op = yfs_FS_StoreData64,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
.deliver = yfs_deliver_status_and_volsync,
.destructor = afs_flat_call_destructor,
};
/*
* Set the attributes on a file, using YFS.StoreData64 rather than
* YFS.StoreStatus so as to alter the file size also.
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
static int yfs_fs_setattr_size(struct afs_fs_cursor *fc, struct iattr *attr,
struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter(",%x,{%llx:%llu},,",
key_serial(fc->key), vnode->fid.vid, vnode->fid.vnode);
call = afs_alloc_flat_call(net, &yfs_RXYFSStoreData64_as_Status,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_YFSStoreStatus) +
sizeof(struct yfs_xdr_u64) * 3,
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSSTOREDATA64);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
bp = xdr_encode_YFS_StoreStatus(bp, attr);
afs: Fix StoreData op marshalling The marshalling of AFS.StoreData, AFS.StoreData64 and YFS.StoreData64 calls generated by ->setattr() ops for the purpose of expanding a file is incorrect due to older documentation incorrectly describing the way the RPC 'FileLength' parameter is meant to work. The older documentation says that this is the length the file is meant to end up at the end of the operation; however, it was never implemented this way in any of the servers, but rather the file is truncated down to this before the write operation is effected, and never expanded to it (and, indeed, it was renamed to 'TruncPos' in 2014). Fix this by setting the position parameter to the new file length and doing a zero-lengh write there. The bug causes Xwayland to SIGBUS due to unexpected non-expansion of a file it then mmaps. This can be tested by giving the following test program a filename in an AFS directory: #include <stdio.h> #include <stdlib.h> #include <unistd.h> #include <fcntl.h> #include <sys/mman.h> int main(int argc, char *argv[]) { char *p; int fd; if (argc != 2) { fprintf(stderr, "Format: test-trunc-mmap <file>\n"); exit(2); } fd = open(argv[1], O_RDWR | O_CREAT | O_TRUNC); if (fd < 0) { perror(argv[1]); exit(1); } if (ftruncate(fd, 0x140008) == -1) { perror("ftruncate"); exit(1); } p = mmap(NULL, 4096, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0); if (p == MAP_FAILED) { perror("mmap"); exit(1); } p[0] = 'a'; if (munmap(p, 4096) < 0) { perror("munmap"); exit(1); } if (close(fd) < 0) { perror("close"); exit(1); } exit(0); } Fixes: 31143d5d515e ("AFS: implement basic file write support") Reported-by: Jonathan Billings <jsbillin@umich.edu> Tested-by: Jonathan Billings <jsbillin@umich.edu> Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-03-28 06:48:02 +08:00
bp = xdr_encode_u64(bp, attr->ia_size); /* position of start of write */
bp = xdr_encode_u64(bp, 0); /* size of write */
bp = xdr_encode_u64(bp, attr->ia_size); /* new file length */
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Set the attributes on a file, using YFS.StoreData64 if there's a change in
* file size, and YFS.StoreStatus otherwise.
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
int yfs_fs_setattr(struct afs_fs_cursor *fc, struct iattr *attr,
struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
if (attr->ia_valid & ATTR_SIZE)
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
return yfs_fs_setattr_size(fc, attr, scb);
_enter(",%x,{%llx:%llu},,",
key_serial(fc->key), vnode->fid.vid, vnode->fid.vnode);
call = afs_alloc_flat_call(net, &yfs_RXYFSStoreStatus,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid) +
sizeof(struct yfs_xdr_YFSStoreStatus),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSSTORESTATUS);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
bp = xdr_encode_YFS_StoreStatus(bp, attr);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to a YFS.GetVolumeStatus operation.
*/
static int yfs_deliver_fs_get_volume_status(struct afs_call *call)
{
const __be32 *bp;
char *p;
u32 size;
int ret;
_enter("{%u}", call->unmarshall);
switch (call->unmarshall) {
case 0:
call->unmarshall++;
afs_extract_to_buf(call, sizeof(struct yfs_xdr_YFSFetchVolumeStatus));
/* Fall through */
/* extract the returned status record */
case 1:
_debug("extract status");
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
bp = call->buffer;
xdr_decode_YFSFetchVolumeStatus(&bp, call->out_volstatus);
call->unmarshall++;
afs_extract_to_tmp(call);
/* Fall through */
/* extract the volume name length */
case 2:
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
call->count = ntohl(call->tmp);
_debug("volname length: %u", call->count);
if (call->count >= AFSNAMEMAX)
return afs_protocol_error(call, -EBADMSG,
afs_eproto_volname_len);
size = (call->count + 3) & ~3; /* It's padded */
afs_extract_to_buf(call, size);
call->unmarshall++;
/* Fall through */
/* extract the volume name */
case 3:
_debug("extract volname");
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
p = call->buffer;
p[call->count] = 0;
_debug("volname '%s'", p);
afs_extract_to_tmp(call);
call->unmarshall++;
/* Fall through */
/* extract the offline message length */
case 4:
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
call->count = ntohl(call->tmp);
_debug("offline msg length: %u", call->count);
if (call->count >= AFSNAMEMAX)
return afs_protocol_error(call, -EBADMSG,
afs_eproto_offline_msg_len);
size = (call->count + 3) & ~3; /* It's padded */
afs_extract_to_buf(call, size);
call->unmarshall++;
/* Fall through */
/* extract the offline message */
case 5:
_debug("extract offline");
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
p = call->buffer;
p[call->count] = 0;
_debug("offline '%s'", p);
afs_extract_to_tmp(call);
call->unmarshall++;
/* Fall through */
/* extract the message of the day length */
case 6:
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
call->count = ntohl(call->tmp);
_debug("motd length: %u", call->count);
if (call->count >= AFSNAMEMAX)
return afs_protocol_error(call, -EBADMSG,
afs_eproto_motd_len);
size = (call->count + 3) & ~3; /* It's padded */
afs_extract_to_buf(call, size);
call->unmarshall++;
/* Fall through */
/* extract the message of the day */
case 7:
_debug("extract motd");
ret = afs_extract_data(call, false);
if (ret < 0)
return ret;
p = call->buffer;
p[call->count] = 0;
_debug("motd '%s'", p);
call->unmarshall++;
/* Fall through */
case 8:
break;
}
_leave(" = 0 [done]");
return 0;
}
/*
* YFS.GetVolumeStatus operation type
*/
static const struct afs_call_type yfs_RXYFSGetVolumeStatus = {
.name = "YFS.GetVolumeStatus",
.op = yfs_FS_GetVolumeStatus,
.deliver = yfs_deliver_fs_get_volume_status,
.destructor = afs_flat_call_destructor,
};
/*
* fetch the status of a volume
*/
int yfs_fs_get_volume_status(struct afs_fs_cursor *fc,
struct afs_volume_status *vs)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter("");
call = afs_alloc_flat_call(net, &yfs_RXYFSGetVolumeStatus,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_u64),
max_t(size_t,
sizeof(struct yfs_xdr_YFSFetchVolumeStatus) +
sizeof(__be32),
AFSOPAQUEMAX + 1));
if (!call)
return -ENOMEM;
call->key = fc->key;
call->out_volstatus = vs;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSGETVOLUMESTATUS);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_u64(bp, vnode->fid.vid);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* YFS.SetLock operation type
*/
static const struct afs_call_type yfs_RXYFSSetLock = {
.name = "YFS.SetLock",
.op = yfs_FS_SetLock,
.deliver = yfs_deliver_status_and_volsync,
.done = afs_lock_op_done,
.destructor = afs_flat_call_destructor,
};
/*
* YFS.ExtendLock operation type
*/
static const struct afs_call_type yfs_RXYFSExtendLock = {
.name = "YFS.ExtendLock",
.op = yfs_FS_ExtendLock,
.deliver = yfs_deliver_status_and_volsync,
.done = afs_lock_op_done,
.destructor = afs_flat_call_destructor,
};
/*
* YFS.ReleaseLock operation type
*/
static const struct afs_call_type yfs_RXYFSReleaseLock = {
.name = "YFS.ReleaseLock",
.op = yfs_FS_ReleaseLock,
.deliver = yfs_deliver_status_and_volsync,
.destructor = afs_flat_call_destructor,
};
/*
* Set a lock on a file
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
int yfs_fs_set_lock(struct afs_fs_cursor *fc, afs_lock_type_t type,
struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter("");
call = afs_alloc_flat_call(net, &yfs_RXYFSSetLock,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid) +
sizeof(__be32),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->lvnode = vnode;
call->out_scb = scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSSETLOCK);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
bp = xdr_encode_u32(bp, type);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_calli(call, &vnode->fid, type);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* extend a lock on a file
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
int yfs_fs_extend_lock(struct afs_fs_cursor *fc, struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter("");
call = afs_alloc_flat_call(net, &yfs_RXYFSExtendLock,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->lvnode = vnode;
call->out_scb = scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSEXTENDLOCK);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* release a lock on a file
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
int yfs_fs_release_lock(struct afs_fs_cursor *fc, struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter("");
call = afs_alloc_flat_call(net, &yfs_RXYFSReleaseLock,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call)
return -ENOMEM;
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->lvnode = vnode;
call->out_scb = scb;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSRELEASELOCK);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* YFS.FetchStatus operation type
*/
static const struct afs_call_type yfs_RXYFSFetchStatus = {
.name = "YFS.FetchStatus",
.op = yfs_FS_FetchStatus,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
.deliver = yfs_deliver_fs_status_cb_and_volsync,
.destructor = afs_flat_call_destructor,
};
/*
* Fetch the status information for a fid without needing a vnode handle.
*/
int yfs_fs_fetch_status(struct afs_fs_cursor *fc,
struct afs_net *net,
struct afs_fid *fid,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct afs_status_cb *scb,
struct afs_volsync *volsync)
{
struct afs_call *call;
__be32 *bp;
_enter(",%x,{%llx:%llu},,",
key_serial(fc->key), fid->vid, fid->vnode);
call = afs_alloc_flat_call(net, &yfs_RXYFSFetchStatus,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid),
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSCallBack) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call) {
fc->ac.error = -ENOMEM;
return -ENOMEM;
}
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
call->out_volsync = volsync;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSFETCHSTATUS);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, fid);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, fid);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to an YFS.InlineBulkStatus call
*/
static int yfs_deliver_fs_inline_bulk_status(struct afs_call *call)
{
struct afs_status_cb *scb;
const __be32 *bp;
u32 tmp;
int ret;
_enter("{%u}", call->unmarshall);
switch (call->unmarshall) {
case 0:
afs_extract_to_tmp(call);
call->unmarshall++;
/* Fall through */
/* Extract the file status count and array in two steps */
case 1:
_debug("extract status count");
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
tmp = ntohl(call->tmp);
_debug("status count: %u/%u", tmp, call->count2);
if (tmp != call->count2)
return afs_protocol_error(call, -EBADMSG,
afs_eproto_ibulkst_count);
call->count = 0;
call->unmarshall++;
more_counts:
afs_extract_to_buf(call, sizeof(struct yfs_xdr_YFSFetchStatus));
/* Fall through */
case 2:
_debug("extract status array %u", call->count);
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
bp = call->buffer;
scb = &call->out_scb[call->count];
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, scb);
if (ret < 0)
return ret;
call->count++;
if (call->count < call->count2)
goto more_counts;
call->count = 0;
call->unmarshall++;
afs_extract_to_tmp(call);
/* Fall through */
/* Extract the callback count and array in two steps */
case 3:
_debug("extract CB count");
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
tmp = ntohl(call->tmp);
_debug("CB count: %u", tmp);
if (tmp != call->count2)
return afs_protocol_error(call, -EBADMSG,
afs_eproto_ibulkst_cb_count);
call->count = 0;
call->unmarshall++;
more_cbs:
afs_extract_to_buf(call, sizeof(struct yfs_xdr_YFSCallBack));
/* Fall through */
case 4:
_debug("extract CB array");
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
_debug("unmarshall CB array");
bp = call->buffer;
scb = &call->out_scb[call->count];
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSCallBack(&bp, call, scb);
call->count++;
if (call->count < call->count2)
goto more_cbs;
afs_extract_to_buf(call, sizeof(struct yfs_xdr_YFSVolSync));
call->unmarshall++;
/* Fall through */
case 5:
ret = afs_extract_data(call, false);
if (ret < 0)
return ret;
bp = call->buffer;
xdr_decode_YFSVolSync(&bp, call->out_volsync);
call->unmarshall++;
/* Fall through */
case 6:
break;
}
_leave(" = 0 [done]");
return 0;
}
/*
* FS.InlineBulkStatus operation type
*/
static const struct afs_call_type yfs_RXYFSInlineBulkStatus = {
.name = "YFS.InlineBulkStatus",
.op = yfs_FS_InlineBulkStatus,
.deliver = yfs_deliver_fs_inline_bulk_status,
.destructor = afs_flat_call_destructor,
};
/*
* Fetch the status information for up to 1024 files
*/
int yfs_fs_inline_bulk_status(struct afs_fs_cursor *fc,
struct afs_net *net,
struct afs_fid *fids,
struct afs_status_cb *statuses,
unsigned int nr_fids,
struct afs_volsync *volsync)
{
struct afs_call *call;
__be32 *bp;
int i;
_enter(",%x,{%llx:%llu},%u",
key_serial(fc->key), fids[0].vid, fids[1].vnode, nr_fids);
call = afs_alloc_flat_call(net, &yfs_RXYFSInlineBulkStatus,
sizeof(__be32) +
sizeof(__be32) +
sizeof(__be32) +
sizeof(struct yfs_xdr_YFSFid) * nr_fids,
sizeof(struct yfs_xdr_YFSFetchStatus));
if (!call) {
fc->ac.error = -ENOMEM;
return -ENOMEM;
}
call->key = fc->key;
call->out_scb = statuses;
call->out_volsync = volsync;
call->count2 = nr_fids;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSINLINEBULKSTATUS);
bp = xdr_encode_u32(bp, 0); /* RPCFlags */
bp = xdr_encode_u32(bp, nr_fids);
for (i = 0; i < nr_fids; i++)
bp = xdr_encode_YFSFid(bp, &fids[i]);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &fids[0]);
afs: Make some RPC operations non-interruptible Make certain RPC operations non-interruptible, including: (*) Set attributes (*) Store data We don't want to get interrupted during a flush on close, flush on unlock, writeback or an inode update, leaving us in a state where we still need to do the writeback or update. (*) Extend lock (*) Release lock We don't want to get lock extension interrupted as the file locks on the server are time-limited. Interruption during lock release is less of an issue since the lock is time-limited, but it's better to complete the release to avoid a several-minute wait to recover it. *Setting* the lock isn't a problem if it's interrupted since we can just return to the user and tell them they were interrupted - at which point they can elect to retry. (*) Silly unlink We want to remove silly unlink files if we can, rather than leaving them for the salvager to clear up. Note that whilst these calls are no longer interruptible, they do have timeouts on them, so if the server stops responding the call will fail with something like ETIME or ECONNRESET. Without this, the following: kAFS: Unexpected error from FS.StoreData -512 appears in dmesg when a pending store data gets interrupted and some processes may just hang. Additionally, make the code that checks/updates the server record ignore failure due to interruption if the main call is uninterruptible and if the server has an address list. The next op will check it again since the expiration time on the old list has past. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Reported-by: Jonathan Billings <jsbillings@jsbillings.org> Reported-by: Marc Dionne <marc.dionne@auristor.com> Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-08 23:16:31 +08:00
afs_set_fc_call(call, fc);
afs_make_call(&fc->ac, call, GFP_NOFS);
return afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* Deliver reply data to an YFS.FetchOpaqueACL.
*/
static int yfs_deliver_fs_fetch_opaque_acl(struct afs_call *call)
{
struct yfs_acl *yacl = call->out_yacl;
struct afs_acl *acl;
const __be32 *bp;
unsigned int size;
int ret;
_enter("{%u}", call->unmarshall);
switch (call->unmarshall) {
case 0:
afs_extract_to_tmp(call);
call->unmarshall++;
/* Fall through */
/* Extract the file ACL length */
case 1:
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
size = call->count2 = ntohl(call->tmp);
size = round_up(size, 4);
if (yacl->flags & YFS_ACL_WANT_ACL) {
acl = kmalloc(struct_size(acl, data, size), GFP_KERNEL);
if (!acl)
return -ENOMEM;
yacl->acl = acl;
acl->size = call->count2;
afs_extract_begin(call, acl->data, size);
} else {
afs_extract_discard(call, size);
}
call->unmarshall++;
/* Fall through */
/* Extract the file ACL */
case 2:
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
afs_extract_to_tmp(call);
call->unmarshall++;
/* Fall through */
/* Extract the volume ACL length */
case 3:
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
size = call->count2 = ntohl(call->tmp);
size = round_up(size, 4);
if (yacl->flags & YFS_ACL_WANT_VOL_ACL) {
acl = kmalloc(struct_size(acl, data, size), GFP_KERNEL);
if (!acl)
return -ENOMEM;
yacl->vol_acl = acl;
acl->size = call->count2;
afs_extract_begin(call, acl->data, size);
} else {
afs_extract_discard(call, size);
}
call->unmarshall++;
/* Fall through */
/* Extract the volume ACL */
case 4:
ret = afs_extract_data(call, true);
if (ret < 0)
return ret;
afs_extract_to_buf(call,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
call->unmarshall++;
/* Fall through */
/* extract the metadata */
case 5:
ret = afs_extract_data(call, false);
if (ret < 0)
return ret;
bp = call->buffer;
yacl->inherit_flag = ntohl(*bp++);
yacl->num_cleaned = ntohl(*bp++);
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
ret = xdr_decode_YFSFetchStatus(&bp, call, call->out_scb);
if (ret < 0)
return ret;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
xdr_decode_YFSVolSync(&bp, call->out_volsync);
call->unmarshall++;
/* Fall through */
case 6:
break;
}
_leave(" = 0 [done]");
return 0;
}
void yfs_free_opaque_acl(struct yfs_acl *yacl)
{
if (yacl) {
kfree(yacl->acl);
kfree(yacl->vol_acl);
kfree(yacl);
}
}
/*
* YFS.FetchOpaqueACL operation type
*/
static const struct afs_call_type yfs_RXYFSFetchOpaqueACL = {
.name = "YFS.FetchOpaqueACL",
.op = yfs_FS_FetchOpaqueACL,
.deliver = yfs_deliver_fs_fetch_opaque_acl,
.destructor = afs_flat_call_destructor,
};
/*
* Fetch the YFS advanced ACLs for a file.
*/
struct yfs_acl *yfs_fs_fetch_opaque_acl(struct afs_fs_cursor *fc,
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
struct yfs_acl *yacl,
struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
__be32 *bp;
_enter(",%x,{%llx:%llu},,",
key_serial(fc->key), vnode->fid.vid, vnode->fid.vnode);
call = afs_alloc_flat_call(net, &yfs_RXYFSFetchOpaqueACL,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid),
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call) {
fc->ac.error = -ENOMEM;
return ERR_PTR(-ENOMEM);
}
call->key = fc->key;
call->out_yacl = yacl;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
call->out_volsync = NULL;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSFETCHOPAQUEACL);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
yfs_check_req(call, bp);
afs_use_fs_server(call, fc->cbi);
trace_afs_make_fs_call(call, &vnode->fid);
afs_make_call(&fc->ac, call, GFP_KERNEL);
return (struct yfs_acl *)afs_wait_for_call_to_complete(call, &fc->ac);
}
/*
* YFS.StoreOpaqueACL2 operation type
*/
static const struct afs_call_type yfs_RXYFSStoreOpaqueACL2 = {
.name = "YFS.StoreOpaqueACL2",
.op = yfs_FS_StoreOpaqueACL2,
.deliver = yfs_deliver_status_and_volsync,
.destructor = afs_flat_call_destructor,
};
/*
* Fetch the YFS ACL for a file.
*/
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
int yfs_fs_store_opaque_acl2(struct afs_fs_cursor *fc, const struct afs_acl *acl,
struct afs_status_cb *scb)
{
struct afs_vnode *vnode = fc->vnode;
struct afs_call *call;
struct afs_net *net = afs_v2net(vnode);
size_t size;
__be32 *bp;
_enter(",%x,{%llx:%llu},,",
key_serial(fc->key), vnode->fid.vid, vnode->fid.vnode);
size = round_up(acl->size, 4);
call = afs_alloc_flat_call(net, &yfs_RXYFSStoreOpaqueACL2,
sizeof(__be32) * 2 +
sizeof(struct yfs_xdr_YFSFid) +
sizeof(__be32) + size,
sizeof(struct yfs_xdr_YFSFetchStatus) +
sizeof(struct yfs_xdr_YFSVolSync));
if (!call) {
fc->ac.error = -ENOMEM;
return -ENOMEM;
}
call->key = fc->key;
afs: Fix application of status and callback to be under same lock When applying the status and callback in the response of an operation, apply them in the same critical section so that there's no race between checking the callback state and checking status-dependent state (such as the data version). Fix this by: (1) Allocating a joint {status,callback} record (afs_status_cb) before calling the RPC function for each vnode for which the RPC reply contains a status or a status plus a callback. A flag is set in the record to indicate if a callback was actually received. (2) These records are passed into the RPC functions to be filled in. The afs_decode_status() and yfs_decode_status() functions are removed and the cb_lock is no longer taken. (3) xdr_decode_AFSFetchStatus() and xdr_decode_YFSFetchStatus() no longer update the vnode. (4) xdr_decode_AFSCallBack() and xdr_decode_YFSCallBack() no longer update the vnode. (5) vnodes, expected data-version numbers and callback break counters (cb_break) no longer need to be passed to the reply delivery functions. Note that, for the moment, the file locking functions still need access to both the call and the vnode at the same time. (6) afs_vnode_commit_status() is now given the cb_break value and the expected data_version and the task of applying the status and the callback to the vnode are now done here. This is done under a single taking of vnode->cb_lock. (7) afs_pages_written_back() is now called by afs_store_data() rather than by the reply delivery function. afs_pages_written_back() has been moved to before the call point and is now given the first and last page numbers rather than a pointer to the call. (8) The indicator from YFS.RemoveFile2 as to whether the target file actually got removed (status.abort_code == VNOVNODE) rather than merely dropping a link is now checked in afs_unlink rather than in xdr_decode_YFSFetchStatus(). Supplementary fixes: (*) afs_cache_permit() now gets the caller_access mask from the afs_status_cb object rather than picking it out of the vnode's status record. afs_fetch_status() returns caller_access through its argument list for this purpose also. (*) afs_inode_init_from_status() now uses a write lock on cb_lock rather than a read lock and now sets the callback inside the same critical section. Fixes: c435ee34551e ("afs: Overhaul the callback handling") Signed-off-by: David Howells <dhowells@redhat.com>
2019-05-09 22:16:10 +08:00
call->out_scb = scb;
call->out_volsync = NULL;
/* marshall the parameters */
bp = call->request;
bp = xdr_encode_u32(bp, YFSSTOREOPAQUEACL2);
bp = xdr_encode_u32(bp, 0); /* RPC flags */
bp = xdr_encode_YFSFid(bp, &vnode->fid);
bp = xdr_encode_u32(bp, acl->size);
memcpy(bp, acl->data, acl->size);
if (acl->size != size)
memset((void *)bp + acl->size, 0, size - acl->size);
yfs_check_req(call, bp);
trace_afs_make_fs_call(call, &vnode->fid);
afs_make_call(&fc->ac, call, GFP_KERNEL);
return afs_wait_for_call_to_complete(call, &fc->ac);
}