From 72d78717c6d06adf65d2e3dccc96d9e9dc978593 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 11 Dec 2020 12:26:14 -0500 Subject: [PATCH 1/5] nfsd: Fixes for nfsd4_encode_read_plus_data() Ensure that we encode the data payload + padding, and that we truncate the preallocated buffer to the actual read size. Fixes: 528b84934eb9 ("NFSD: Add READ_PLUS data support") Signed-off-by: Trond Myklebust Signed-off-by: Chuck Lever --- fs/nfsd/nfs4xdr.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c index 45ee6b12ce5b..5dacc673ef17 100644 --- a/fs/nfsd/nfs4xdr.c +++ b/fs/nfsd/nfs4xdr.c @@ -4756,6 +4756,7 @@ nfsd4_encode_read_plus_data(struct nfsd4_compoundres *resp, resp->rqstp->rq_vec, read->rd_vlen, maxcount, eof); if (nfserr) return nfserr; + xdr_truncate_encode(xdr, starting_len + 16 + xdr_align_size(*maxcount)); tmp = htonl(NFS4_CONTENT_DATA); write_bytes_to_xdr_buf(xdr->buf, starting_len, &tmp, 4); @@ -4763,6 +4764,10 @@ nfsd4_encode_read_plus_data(struct nfsd4_compoundres *resp, write_bytes_to_xdr_buf(xdr->buf, starting_len + 4, &tmp64, 8); tmp = htonl(*maxcount); write_bytes_to_xdr_buf(xdr->buf, starting_len + 12, &tmp, 4); + + tmp = xdr_zero; + write_bytes_to_xdr_buf(xdr->buf, starting_len + 16 + *maxcount, &tmp, + xdr_pad_size(*maxcount)); return nfs_ok; } From b68f0cbd3f95f2df81e525c310a41fc73c2ed0d3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 11 Dec 2020 12:26:15 -0500 Subject: [PATCH 2/5] nfsd: Don't set eof on a truncated READ_PLUS If the READ_PLUS operation was truncated due to an error, then ensure we clear the 'eof' flag. Fixes: 9f0b5792f07d ("NFSD: Encode a full READ_PLUS reply") Signed-off-by: Trond Myklebust Signed-off-by: Chuck Lever --- fs/nfsd/nfs4xdr.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c index 5dacc673ef17..20178f60f612 100644 --- a/fs/nfsd/nfs4xdr.c +++ b/fs/nfsd/nfs4xdr.c @@ -4860,14 +4860,15 @@ out: if (nfserr && segments == 0) xdr_truncate_encode(xdr, starting_len); else { + if (nfserr) { + xdr_truncate_encode(xdr, last_segment); + nfserr = nfs_ok; + eof = 0; + } tmp = htonl(eof); write_bytes_to_xdr_buf(xdr->buf, starting_len, &tmp, 4); tmp = htonl(segments); write_bytes_to_xdr_buf(xdr->buf, starting_len + 4, &tmp, 4); - if (nfserr) { - xdr_truncate_encode(xdr, last_segment); - nfserr = nfs_ok; - } } return nfserr; From d6c9e4368cc6a61bf25c9c72437ced509c854563 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Fri, 18 Dec 2020 12:28:23 -0500 Subject: [PATCH 3/5] NFSD: Fix sparse warning in nfssvc.c fs/nfsd/nfssvc.c:36:6: warning: symbol 'inter_copy_offload_enable' was not declared. Should it be static? The parameter was added by commit ce0887ac96d3 ("NFSD add nfs4 inter ssc to nfsd4_copy"). Relocate it into the source file that uses it, and make it static. This approach is similar to the nfs4_disable_idmapping, cltrack_prog, and cltrack_legacy_disable module parameters. Signed-off-by: Chuck Lever --- fs/nfsd/nfs4proc.c | 5 +++++ fs/nfsd/nfssvc.c | 6 ------ fs/nfsd/xdr4.h | 1 - 3 files changed, 5 insertions(+), 7 deletions(-) diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c index 4727b7f03c5b..8d6d2678abad 100644 --- a/fs/nfsd/nfs4proc.c +++ b/fs/nfsd/nfs4proc.c @@ -50,6 +50,11 @@ #include "pnfs.h" #include "trace.h" +static bool inter_copy_offload_enable; +module_param(inter_copy_offload_enable, bool, 0644); +MODULE_PARM_DESC(inter_copy_offload_enable, + "Enable inter server to server copy offload. Default: false"); + #ifdef CONFIG_NFSD_V4_SECURITY_LABEL #include diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c index 00384c332f9b..f9c9f4c63cc7 100644 --- a/fs/nfsd/nfssvc.c +++ b/fs/nfsd/nfssvc.c @@ -33,12 +33,6 @@ #define NFSDDBG_FACILITY NFSDDBG_SVC -bool inter_copy_offload_enable; -EXPORT_SYMBOL_GPL(inter_copy_offload_enable); -module_param(inter_copy_offload_enable, bool, 0644); -MODULE_PARM_DESC(inter_copy_offload_enable, - "Enable inter server to server copy offload. Default: false"); - extern struct svc_program nfsd_program; static int nfsd(void *vrqstp); #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) diff --git a/fs/nfsd/xdr4.h b/fs/nfsd/xdr4.h index a60ff5ce1a37..c300885ae75d 100644 --- a/fs/nfsd/xdr4.h +++ b/fs/nfsd/xdr4.h @@ -568,7 +568,6 @@ struct nfsd4_copy { struct nfs_fh c_fh; nfs4_stateid stateid; }; -extern bool inter_copy_offload_enable; struct nfsd4_seek { /* request */ From 4a85a6a3320b4a622315d2e0ea91a1d2b013bce4 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Fri, 18 Dec 2020 12:28:41 -0500 Subject: [PATCH 4/5] SUNRPC: Handle TCP socket sends with kernel_sendpage() again Daire Byrne reports a ~50% aggregrate throughput regression on his Linux NFS server after commit da1661b93bf4 ("SUNRPC: Teach server to use xprt_sock_sendmsg for socket sends"), which replaced kernel_send_page() calls in NFSD's socket send path with calls to sock_sendmsg() using iov_iter. Investigation showed that tcp_sendmsg() was not using zero-copy to send the xdr_buf's bvec pages, but instead was relying on memcpy. This means copying every byte of a large NFS READ payload. It looks like TLS sockets do indeed support a ->sendpage method, so it's really not necessary to use xprt_sock_sendmsg() to support TLS fully on the server. A mechanical reversion of da1661b93bf4 is not possible at this point, but we can re-implement the server's TCP socket sendmsg path using kernel_sendpage(). Reported-by: Daire Byrne BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=209439 Signed-off-by: Chuck Lever --- net/sunrpc/svcsock.c | 86 +++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 85 insertions(+), 1 deletion(-) diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c index b248f2349437..c9766d07eb81 100644 --- a/net/sunrpc/svcsock.c +++ b/net/sunrpc/svcsock.c @@ -1062,6 +1062,90 @@ err_noclose: return 0; /* record not complete */ } +static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec, + int flags) +{ + return kernel_sendpage(sock, virt_to_page(vec->iov_base), + offset_in_page(vec->iov_base), + vec->iov_len, flags); +} + +/* + * kernel_sendpage() is used exclusively to reduce the number of + * copy operations in this path. Therefore the caller must ensure + * that the pages backing @xdr are unchanging. + * + * In addition, the logic assumes that * .bv_len is never larger + * than PAGE_SIZE. + */ +static int svc_tcp_sendmsg(struct socket *sock, struct msghdr *msg, + struct xdr_buf *xdr, rpc_fraghdr marker, + unsigned int *sentp) +{ + const struct kvec *head = xdr->head; + const struct kvec *tail = xdr->tail; + struct kvec rm = { + .iov_base = &marker, + .iov_len = sizeof(marker), + }; + int flags, ret; + + *sentp = 0; + xdr_alloc_bvec(xdr, GFP_KERNEL); + + msg->msg_flags = MSG_MORE; + ret = kernel_sendmsg(sock, msg, &rm, 1, rm.iov_len); + if (ret < 0) + return ret; + *sentp += ret; + if (ret != rm.iov_len) + return -EAGAIN; + + flags = head->iov_len < xdr->len ? MSG_MORE | MSG_SENDPAGE_NOTLAST : 0; + ret = svc_tcp_send_kvec(sock, head, flags); + if (ret < 0) + return ret; + *sentp += ret; + if (ret != head->iov_len) + goto out; + + if (xdr->page_len) { + unsigned int offset, len, remaining; + struct bio_vec *bvec; + + bvec = xdr->bvec; + offset = xdr->page_base; + remaining = xdr->page_len; + flags = MSG_MORE | MSG_SENDPAGE_NOTLAST; + while (remaining > 0) { + if (remaining <= PAGE_SIZE && tail->iov_len == 0) + flags = 0; + len = min(remaining, bvec->bv_len); + ret = kernel_sendpage(sock, bvec->bv_page, + bvec->bv_offset + offset, + len, flags); + if (ret < 0) + return ret; + *sentp += ret; + if (ret != len) + goto out; + remaining -= len; + offset = 0; + bvec++; + } + } + + if (tail->iov_len) { + ret = svc_tcp_send_kvec(sock, tail, 0); + if (ret < 0) + return ret; + *sentp += ret; + } + +out: + return 0; +} + /** * svc_tcp_sendto - Send out a reply on a TCP socket * @rqstp: completed svc_rqst @@ -1089,7 +1173,7 @@ static int svc_tcp_sendto(struct svc_rqst *rqstp) mutex_lock(&xprt->xpt_mutex); if (svc_xprt_is_dead(xprt)) goto out_notconn; - err = xprt_sock_sendmsg(svsk->sk_sock, &msg, xdr, 0, marker, &sent); + err = svc_tcp_sendmsg(svsk->sk_sock, &msg, xdr, marker, &sent); xdr_free_bvec(xdr); trace_svcsock_tcp_send(xprt, err < 0 ? err : sent); if (err < 0 || sent != (xdr->len + sizeof(marker))) From 7b723008f9c95624c848fad661c01b06e47b20da Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Fri, 18 Dec 2020 12:28:58 -0500 Subject: [PATCH 5/5] NFSD: Restore NFSv4 decoding's SAVEMEM functionality While converting the NFSv4 decoder to use xdr_stream-based XDR processing, I removed the old SAVEMEM() macro. This macro wrapped a bit of logic that avoided a memory allocation by recognizing when the decoded item resides in a linear section of the Receive buffer. In that case, it returned a pointer into that buffer instead of allocating a bounce buffer. The bounce buffer is necessary only when xdr_inline_decode() has placed the decoded item in the xdr_stream's scratch buffer, which disappears the next time xdr_inline_decode() is called with that xdr_stream. That happens only if the data item crosses a page boundary in the receive buffer, an exceedingly rare occurrence. Allocating a bounce buffer every time results in a minor performance regression that was introduced by the recent NFSv4 decoder overhaul. Let's restore the previous behavior. On average, it saves about 1.5 kmalloc() calls per COMPOUND. Signed-off-by: Chuck Lever --- fs/nfsd/nfs4xdr.c | 42 ++++++++++++++++++++++++++---------------- 1 file changed, 26 insertions(+), 16 deletions(-) diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c index 20178f60f612..eaaa1605b5b5 100644 --- a/fs/nfsd/nfs4xdr.c +++ b/fs/nfsd/nfs4xdr.c @@ -147,6 +147,25 @@ svcxdr_dupstr(struct nfsd4_compoundargs *argp, void *buf, u32 len) return p; } +static void * +svcxdr_savemem(struct nfsd4_compoundargs *argp, __be32 *p, u32 len) +{ + __be32 *tmp; + + /* + * The location of the decoded data item is stable, + * so @p is OK to use. This is the common case. + */ + if (p != argp->xdr->scratch.iov_base) + return p; + + tmp = svcxdr_tmpalloc(argp, len); + if (!tmp) + return NULL; + memcpy(tmp, p, len); + return tmp; +} + /* * NFSv4 basic data type decoders */ @@ -183,11 +202,10 @@ nfsd4_decode_opaque(struct nfsd4_compoundargs *argp, struct xdr_netobj *o) p = xdr_inline_decode(argp->xdr, len); if (!p) return nfserr_bad_xdr; - o->data = svcxdr_tmpalloc(argp, len); + o->data = svcxdr_savemem(argp, p, len); if (!o->data) return nfserr_jukebox; o->len = len; - memcpy(o->data, p, len); return nfs_ok; } @@ -205,10 +223,9 @@ nfsd4_decode_component4(struct nfsd4_compoundargs *argp, char **namp, u32 *lenp) status = check_filename((char *)p, *lenp); if (status) return status; - *namp = svcxdr_tmpalloc(argp, *lenp); + *namp = svcxdr_savemem(argp, p, *lenp); if (!*namp) return nfserr_jukebox; - memcpy(*namp, p, *lenp); return nfs_ok; } @@ -1200,10 +1217,9 @@ nfsd4_decode_putfh(struct nfsd4_compoundargs *argp, struct nfsd4_putfh *putfh) p = xdr_inline_decode(argp->xdr, putfh->pf_fhlen); if (!p) return nfserr_bad_xdr; - putfh->pf_fhval = svcxdr_tmpalloc(argp, putfh->pf_fhlen); + putfh->pf_fhval = svcxdr_savemem(argp, p, putfh->pf_fhlen); if (!putfh->pf_fhval) return nfserr_jukebox; - memcpy(putfh->pf_fhval, p, putfh->pf_fhlen); return nfs_ok; } @@ -1318,24 +1334,20 @@ nfsd4_decode_setclientid(struct nfsd4_compoundargs *argp, struct nfsd4_setclient p = xdr_inline_decode(argp->xdr, setclientid->se_callback_netid_len); if (!p) return nfserr_bad_xdr; - setclientid->se_callback_netid_val = svcxdr_tmpalloc(argp, + setclientid->se_callback_netid_val = svcxdr_savemem(argp, p, setclientid->se_callback_netid_len); if (!setclientid->se_callback_netid_val) return nfserr_jukebox; - memcpy(setclientid->se_callback_netid_val, p, - setclientid->se_callback_netid_len); if (xdr_stream_decode_u32(argp->xdr, &setclientid->se_callback_addr_len) < 0) return nfserr_bad_xdr; p = xdr_inline_decode(argp->xdr, setclientid->se_callback_addr_len); if (!p) return nfserr_bad_xdr; - setclientid->se_callback_addr_val = svcxdr_tmpalloc(argp, + setclientid->se_callback_addr_val = svcxdr_savemem(argp, p, setclientid->se_callback_addr_len); if (!setclientid->se_callback_addr_val) return nfserr_jukebox; - memcpy(setclientid->se_callback_addr_val, p, - setclientid->se_callback_addr_len); if (xdr_stream_decode_u32(argp->xdr, &setclientid->se_callback_ident) < 0) return nfserr_bad_xdr; @@ -1375,10 +1387,9 @@ nfsd4_decode_verify(struct nfsd4_compoundargs *argp, struct nfsd4_verify *verify p = xdr_inline_decode(argp->xdr, verify->ve_attrlen); if (!p) return nfserr_bad_xdr; - verify->ve_attrval = svcxdr_tmpalloc(argp, verify->ve_attrlen); + verify->ve_attrval = svcxdr_savemem(argp, p, verify->ve_attrlen); if (!verify->ve_attrval) return nfserr_jukebox; - memcpy(verify->ve_attrval, p, verify->ve_attrlen); return nfs_ok; } @@ -2333,10 +2344,9 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp) p = xdr_inline_decode(argp->xdr, argp->taglen); if (!p) return 0; - argp->tag = svcxdr_tmpalloc(argp, argp->taglen); + argp->tag = svcxdr_savemem(argp, p, argp->taglen); if (!argp->tag) return 0; - memcpy(argp->tag, p, argp->taglen); max_reply += xdr_align_size(argp->taglen); }