io_uring: pass required context in as flags

We have a number of bits that decide what context to inherit. Set up
io-wq flags for these instead. This is in preparation for always having
the various members set, but not always needing them for all requests.

No intended functional changes in this patch.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
This commit is contained in:
Jens Axboe 2020-10-14 09:23:55 -06:00
parent a8b595b22d
commit 0f20376588
3 changed files with 52 additions and 64 deletions

View File

@ -448,6 +448,8 @@ static inline void io_wq_switch_blkcg(struct io_worker *worker,
struct io_wq_work *work)
{
#ifdef CONFIG_BLK_CGROUP
if (!(work->flags & IO_WQ_WORK_BLKCG))
return;
if (work->blkcg_css != worker->blkcg_css) {
kthread_associate_blkcg(work->blkcg_css);
worker->blkcg_css = work->blkcg_css;
@ -470,17 +472,17 @@ static void io_wq_switch_creds(struct io_worker *worker,
static void io_impersonate_work(struct io_worker *worker,
struct io_wq_work *work)
{
if (work->files && current->files != work->files) {
if ((work->flags & IO_WQ_WORK_FILES) && current->files != work->files) {
task_lock(current);
current->files = work->files;
current->nsproxy = work->nsproxy;
task_unlock(current);
}
if (work->fs && current->fs != work->fs)
if ((work->flags & IO_WQ_WORK_FS) && current->fs != work->fs)
current->fs = work->fs;
if (work->mm != worker->mm)
if ((work->flags & IO_WQ_WORK_MM) && work->mm != worker->mm)
io_wq_switch_mm(worker, work);
if (worker->cur_creds != work->creds)
if ((work->flags & IO_WQ_WORK_CREDS) && worker->cur_creds != work->creds)
io_wq_switch_creds(worker, work);
current->signal->rlim[RLIMIT_FSIZE].rlim_cur = work->fsize;
io_wq_switch_blkcg(worker, work);

View File

@ -10,6 +10,12 @@ enum {
IO_WQ_WORK_NO_CANCEL = 8,
IO_WQ_WORK_CONCURRENT = 16,
IO_WQ_WORK_FILES = 32,
IO_WQ_WORK_FS = 64,
IO_WQ_WORK_MM = 128,
IO_WQ_WORK_CREDS = 256,
IO_WQ_WORK_BLKCG = 512,
IO_WQ_HASH_SHIFT = 24, /* upper 8 bits are used for hash key */
};

View File

@ -729,8 +729,6 @@ struct io_submit_state {
};
struct io_op_def {
/* needs current->mm setup, does mm access */
unsigned needs_mm : 1;
/* needs req->file assigned */
unsigned needs_file : 1;
/* don't fail if file grab fails */
@ -741,10 +739,6 @@ struct io_op_def {
unsigned unbound_nonreg_file : 1;
/* opcode is not supported by this kernel */
unsigned not_supported : 1;
/* needs file table */
unsigned file_table : 1;
/* needs ->fs */
unsigned needs_fs : 1;
/* set if opcode supports polled "wait" */
unsigned pollin : 1;
unsigned pollout : 1;
@ -754,45 +748,42 @@ struct io_op_def {
unsigned needs_fsize : 1;
/* must always have async data allocated */
unsigned needs_async_data : 1;
/* needs blkcg context, issues async io potentially */
unsigned needs_blkcg : 1;
/* size of async data needed, if any */
unsigned short async_size;
unsigned work_flags;
};
static const struct io_op_def io_op_defs[] = {
[IORING_OP_NOP] = {},
[IORING_OP_READV] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.pollin = 1,
.buffer_select = 1,
.needs_async_data = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_rw),
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
},
[IORING_OP_WRITEV] = {
.needs_mm = 1,
.needs_file = 1,
.hash_reg_file = 1,
.unbound_nonreg_file = 1,
.pollout = 1,
.needs_fsize = 1,
.needs_async_data = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_rw),
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
},
[IORING_OP_FSYNC] = {
.needs_file = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_BLKCG,
},
[IORING_OP_READ_FIXED] = {
.needs_file = 1,
.unbound_nonreg_file = 1,
.pollin = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_rw),
.work_flags = IO_WQ_WORK_BLKCG,
},
[IORING_OP_WRITE_FIXED] = {
.needs_file = 1,
@ -800,8 +791,8 @@ static const struct io_op_def io_op_defs[] = {
.unbound_nonreg_file = 1,
.pollout = 1,
.needs_fsize = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_rw),
.work_flags = IO_WQ_WORK_BLKCG,
},
[IORING_OP_POLL_ADD] = {
.needs_file = 1,
@ -810,137 +801,123 @@ static const struct io_op_def io_op_defs[] = {
[IORING_OP_POLL_REMOVE] = {},
[IORING_OP_SYNC_FILE_RANGE] = {
.needs_file = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_BLKCG,
},
[IORING_OP_SENDMSG] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.needs_fs = 1,
.pollout = 1,
.needs_async_data = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_msghdr),
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG |
IO_WQ_WORK_FS,
},
[IORING_OP_RECVMSG] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.needs_fs = 1,
.pollin = 1,
.buffer_select = 1,
.needs_async_data = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_msghdr),
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG |
IO_WQ_WORK_FS,
},
[IORING_OP_TIMEOUT] = {
.needs_mm = 1,
.needs_async_data = 1,
.async_size = sizeof(struct io_timeout_data),
.work_flags = IO_WQ_WORK_MM,
},
[IORING_OP_TIMEOUT_REMOVE] = {},
[IORING_OP_ACCEPT] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.file_table = 1,
.pollin = 1,
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_FILES,
},
[IORING_OP_ASYNC_CANCEL] = {},
[IORING_OP_LINK_TIMEOUT] = {
.needs_mm = 1,
.needs_async_data = 1,
.async_size = sizeof(struct io_timeout_data),
.work_flags = IO_WQ_WORK_MM,
},
[IORING_OP_CONNECT] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.pollout = 1,
.needs_async_data = 1,
.async_size = sizeof(struct io_async_connect),
.work_flags = IO_WQ_WORK_MM,
},
[IORING_OP_FALLOCATE] = {
.needs_file = 1,
.needs_fsize = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_BLKCG,
},
[IORING_OP_OPENAT] = {
.file_table = 1,
.needs_fs = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_FILES | IO_WQ_WORK_BLKCG |
IO_WQ_WORK_FS,
},
[IORING_OP_CLOSE] = {
.needs_file = 1,
.needs_file_no_error = 1,
.file_table = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_FILES | IO_WQ_WORK_BLKCG,
},
[IORING_OP_FILES_UPDATE] = {
.needs_mm = 1,
.file_table = 1,
.work_flags = IO_WQ_WORK_FILES | IO_WQ_WORK_MM,
},
[IORING_OP_STATX] = {
.needs_mm = 1,
.needs_fs = 1,
.file_table = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_FILES | IO_WQ_WORK_MM |
IO_WQ_WORK_FS | IO_WQ_WORK_BLKCG,
},
[IORING_OP_READ] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.pollin = 1,
.buffer_select = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_rw),
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
},
[IORING_OP_WRITE] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.pollout = 1,
.needs_fsize = 1,
.needs_blkcg = 1,
.async_size = sizeof(struct io_async_rw),
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
},
[IORING_OP_FADVISE] = {
.needs_file = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_BLKCG,
},
[IORING_OP_MADVISE] = {
.needs_mm = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
},
[IORING_OP_SEND] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.pollout = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
},
[IORING_OP_RECV] = {
.needs_mm = 1,
.needs_file = 1,
.unbound_nonreg_file = 1,
.pollin = 1,
.buffer_select = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
},
[IORING_OP_OPENAT2] = {
.file_table = 1,
.needs_fs = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_FILES | IO_WQ_WORK_FS |
IO_WQ_WORK_BLKCG,
},
[IORING_OP_EPOLL_CTL] = {
.unbound_nonreg_file = 1,
.file_table = 1,
.work_flags = IO_WQ_WORK_FILES,
},
[IORING_OP_SPLICE] = {
.needs_file = 1,
.hash_reg_file = 1,
.unbound_nonreg_file = 1,
.needs_blkcg = 1,
.work_flags = IO_WQ_WORK_BLKCG,
},
[IORING_OP_PROVIDE_BUFFERS] = {},
[IORING_OP_REMOVE_BUFFERS] = {},
@ -1031,7 +1008,7 @@ static int __io_sq_thread_acquire_mm(struct io_ring_ctx *ctx)
static int io_sq_thread_acquire_mm(struct io_ring_ctx *ctx,
struct io_kiocb *req)
{
if (!io_op_defs[req->opcode].needs_mm)
if (!(io_op_defs[req->opcode].work_flags & IO_WQ_WORK_MM))
return 0;
return __io_sq_thread_acquire_mm(ctx);
}
@ -1224,7 +1201,8 @@ static void io_prep_async_work(struct io_kiocb *req)
if (def->unbound_nonreg_file)
req->work.flags |= IO_WQ_WORK_UNBOUND;
}
if (!req->work.files && io_op_defs[req->opcode].file_table &&
if (!req->work.files &&
(io_op_defs[req->opcode].work_flags & IO_WQ_WORK_FILES) &&
!(req->flags & REQ_F_NO_FILE_TABLE)) {
req->work.files = get_files_struct(current);
get_nsproxy(current->nsproxy);
@ -1235,12 +1213,12 @@ static void io_prep_async_work(struct io_kiocb *req)
list_add(&req->inflight_entry, &ctx->inflight_list);
spin_unlock_irq(&ctx->inflight_lock);
}
if (!req->work.mm && def->needs_mm) {
if (!req->work.mm && (def->work_flags & IO_WQ_WORK_MM)) {
mmgrab(current->mm);
req->work.mm = current->mm;
}
#ifdef CONFIG_BLK_CGROUP
if (!req->work.blkcg_css && def->needs_blkcg) {
if (!req->work.blkcg_css && (def->work_flags & IO_WQ_WORK_BLKCG)) {
rcu_read_lock();
req->work.blkcg_css = blkcg_css();
/*
@ -1254,7 +1232,7 @@ static void io_prep_async_work(struct io_kiocb *req)
#endif
if (!req->work.creds)
req->work.creds = get_current_cred();
if (!req->work.fs && def->needs_fs) {
if (!req->work.fs && (def->work_flags & IO_WQ_WORK_FS)) {
spin_lock(&current->fs->lock);
if (!current->fs->in_exec) {
req->work.fs = current->fs;
@ -1268,6 +1246,8 @@ static void io_prep_async_work(struct io_kiocb *req)
req->work.fsize = rlimit(RLIMIT_FSIZE);
else
req->work.fsize = RLIM_INFINITY;
req->work.flags |= def->work_flags;
}
static void io_prep_async_link(struct io_kiocb *req)