From 0b3a551fa58b4da941efeb209b3770868e2eddd7 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 5 Jan 2023 14:55:56 -0500 Subject: nfsd: fix handling of cached open files in nfsd4_open codepath Commit fb70bf124b05 ("NFSD: Instantiate a struct file when creating a regular NFSv4 file") added the ability to cache an open fd over a compound. There are a couple of problems with the way this currently works: It's racy, as a newly-created nfsd_file can end up with its PENDING bit cleared while the nf is hashed, and the nf_file pointer is still zeroed out. Other tasks can find it in this state and they expect to see a valid nf_file, and can oops if nf_file is NULL. Also, there is no guarantee that we'll end up creating a new nfsd_file if one is already in the hash. If an extant entry is in the hash with a valid nf_file, nfs4_get_vfs_file will clobber its nf_file pointer with the value of op_file and the old nf_file will leak. Fix both issues by making a new nfsd_file_acquirei_opened variant that takes an optional file pointer. If one is present when this is called, we'll take a new reference to it instead of trying to open the file. If the nfsd_file already has a valid nf_file, we'll just ignore the optional file and pass the nfsd_file back as-is. Also rework the tracepoints a bit to allow for an "opened" variant and don't try to avoid counting acquisitions in the case where we already have a cached open file. Fixes: fb70bf124b05 ("NFSD: Instantiate a struct file when creating a regular NFSv4 file") Cc: Trond Myklebust Reported-by: Stanislav Saner Reported-and-Tested-by: Ruben Vestergaard Reported-and-Tested-by: Torkil Svensgaard Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 40 ++++++++++++++++++++++------------------ fs/nfsd/filecache.h | 5 +++-- fs/nfsd/nfs4state.c | 16 ++++------------ fs/nfsd/trace.h | 52 +++++++++++++--------------------------------------- 4 files changed, 42 insertions(+), 71 deletions(-) (limited to 'fs') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 45b2c9e3f636..0ef070349014 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -1071,8 +1071,8 @@ nfsd_file_is_cached(struct inode *inode) static __be32 nfsd_file_do_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp, - unsigned int may_flags, struct nfsd_file **pnf, - bool open, bool want_gc) + unsigned int may_flags, struct file *file, + struct nfsd_file **pnf, bool want_gc) { struct nfsd_file_lookup_key key = { .type = NFSD_FILE_KEY_FULL, @@ -1147,8 +1147,7 @@ wait_for_construction: status = nfserrno(nfsd_open_break_lease(file_inode(nf->nf_file), may_flags)); out: if (status == nfs_ok) { - if (open) - this_cpu_inc(nfsd_file_acquisitions); + this_cpu_inc(nfsd_file_acquisitions); *pnf = nf; } else { if (refcount_dec_and_test(&nf->nf_ref)) @@ -1158,20 +1157,23 @@ out: out_status: put_cred(key.cred); - if (open) - trace_nfsd_file_acquire(rqstp, key.inode, may_flags, nf, status); + trace_nfsd_file_acquire(rqstp, key.inode, may_flags, nf, status); return status; open_file: trace_nfsd_file_alloc(nf); nf->nf_mark = nfsd_file_mark_find_or_create(nf, key.inode); if (nf->nf_mark) { - if (open) { + if (file) { + get_file(file); + nf->nf_file = file; + status = nfs_ok; + trace_nfsd_file_opened(nf, status); + } else { status = nfsd_open_verified(rqstp, fhp, may_flags, &nf->nf_file); trace_nfsd_file_open(nf, status); - } else - status = nfs_ok; + } } else status = nfserr_jukebox; /* @@ -1207,7 +1209,7 @@ __be32 nfsd_file_acquire_gc(struct svc_rqst *rqstp, struct svc_fh *fhp, unsigned int may_flags, struct nfsd_file **pnf) { - return nfsd_file_do_acquire(rqstp, fhp, may_flags, pnf, true, true); + return nfsd_file_do_acquire(rqstp, fhp, may_flags, NULL, pnf, true); } /** @@ -1228,28 +1230,30 @@ __be32 nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp, unsigned int may_flags, struct nfsd_file **pnf) { - return nfsd_file_do_acquire(rqstp, fhp, may_flags, pnf, true, false); + return nfsd_file_do_acquire(rqstp, fhp, may_flags, NULL, pnf, false); } /** - * nfsd_file_create - Get a struct nfsd_file, do not open + * nfsd_file_acquire_opened - Get a struct nfsd_file using existing open file * @rqstp: the RPC transaction being executed * @fhp: the NFS filehandle of the file just created * @may_flags: NFSD_MAY_ settings for the file + * @file: cached, already-open file (may be NULL) * @pnf: OUT: new or found "struct nfsd_file" object * - * The nfsd_file_object returned by this API is reference-counted - * but not garbage-collected. The object is released immediately - * one RCU grace period after the final nfsd_file_put(). + * Acquire a nfsd_file object that is not GC'ed. If one doesn't already exist, + * and @file is non-NULL, use it to instantiate a new nfsd_file instead of + * opening a new one. * * Returns nfs_ok and sets @pnf on success; otherwise an nfsstat in * network byte order is returned. */ __be32 -nfsd_file_create(struct svc_rqst *rqstp, struct svc_fh *fhp, - unsigned int may_flags, struct nfsd_file **pnf) +nfsd_file_acquire_opened(struct svc_rqst *rqstp, struct svc_fh *fhp, + unsigned int may_flags, struct file *file, + struct nfsd_file **pnf) { - return nfsd_file_do_acquire(rqstp, fhp, may_flags, pnf, false, false); + return nfsd_file_do_acquire(rqstp, fhp, may_flags, file, pnf, false); } /* diff --git a/fs/nfsd/filecache.h b/fs/nfsd/filecache.h index b7efb2c3ddb1..41516a4263ea 100644 --- a/fs/nfsd/filecache.h +++ b/fs/nfsd/filecache.h @@ -60,7 +60,8 @@ __be32 nfsd_file_acquire_gc(struct svc_rqst *rqstp, struct svc_fh *fhp, unsigned int may_flags, struct nfsd_file **nfp); __be32 nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp, unsigned int may_flags, struct nfsd_file **nfp); -__be32 nfsd_file_create(struct svc_rqst *rqstp, struct svc_fh *fhp, - unsigned int may_flags, struct nfsd_file **nfp); +__be32 nfsd_file_acquire_opened(struct svc_rqst *rqstp, struct svc_fh *fhp, + unsigned int may_flags, struct file *file, + struct nfsd_file **nfp); int nfsd_file_cache_stats_show(struct seq_file *m, void *v); #endif /* _FS_NFSD_FILECACHE_H */ diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c index e1e85c21f12b..313f666d5357 100644 --- a/fs/nfsd/nfs4state.c +++ b/fs/nfsd/nfs4state.c @@ -5262,18 +5262,10 @@ static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp, if (!fp->fi_fds[oflag]) { spin_unlock(&fp->fi_lock); - if (!open->op_filp) { - status = nfsd_file_acquire(rqstp, cur_fh, access, &nf); - if (status != nfs_ok) - goto out_put_access; - } else { - status = nfsd_file_create(rqstp, cur_fh, access, &nf); - if (status != nfs_ok) - goto out_put_access; - nf->nf_file = open->op_filp; - open->op_filp = NULL; - trace_nfsd_file_create(rqstp, access, nf); - } + status = nfsd_file_acquire_opened(rqstp, cur_fh, access, + open->op_filp, &nf); + if (status != nfs_ok) + goto out_put_access; spin_lock(&fp->fi_lock); if (!fp->fi_fds[oflag]) { diff --git a/fs/nfsd/trace.h b/fs/nfsd/trace.h index c852ae8eaf37..8f9c82d9e075 100644 --- a/fs/nfsd/trace.h +++ b/fs/nfsd/trace.h @@ -981,43 +981,6 @@ TRACE_EVENT(nfsd_file_acquire, ) ); -TRACE_EVENT(nfsd_file_create, - TP_PROTO( - const struct svc_rqst *rqstp, - unsigned int may_flags, - const struct nfsd_file *nf - ), - - TP_ARGS(rqstp, may_flags, nf), - - TP_STRUCT__entry( - __field(const void *, nf_inode) - __field(const void *, nf_file) - __field(unsigned long, may_flags) - __field(unsigned long, nf_flags) - __field(unsigned long, nf_may) - __field(unsigned int, nf_ref) - __field(u32, xid) - ), - - TP_fast_assign( - __entry->nf_inode = nf->nf_inode; - __entry->nf_file = nf->nf_file; - __entry->may_flags = may_flags; - __entry->nf_flags = nf->nf_flags; - __entry->nf_may = nf->nf_may; - __entry->nf_ref = refcount_read(&nf->nf_ref); - __entry->xid = be32_to_cpu(rqstp->rq_xid); - ), - - TP_printk("xid=0x%x inode=%p may_flags=%s ref=%u nf_flags=%s nf_may=%s nf_file=%p", - __entry->xid, __entry->nf_inode, - show_nfsd_may_flags(__entry->may_flags), - __entry->nf_ref, show_nf_flags(__entry->nf_flags), - show_nfsd_may_flags(__entry->nf_may), __entry->nf_file - ) -); - TRACE_EVENT(nfsd_file_insert_err, TP_PROTO( const struct svc_rqst *rqstp, @@ -1079,8 +1042,8 @@ TRACE_EVENT(nfsd_file_cons_err, ) ); -TRACE_EVENT(nfsd_file_open, - TP_PROTO(struct nfsd_file *nf, __be32 status), +DECLARE_EVENT_CLASS(nfsd_file_open_class, + TP_PROTO(const struct nfsd_file *nf, __be32 status), TP_ARGS(nf, status), TP_STRUCT__entry( __field(void *, nf_inode) /* cannot be dereferenced */ @@ -1104,6 +1067,17 @@ TRACE_EVENT(nfsd_file_open, __entry->nf_file) ) +#define DEFINE_NFSD_FILE_OPEN_EVENT(name) \ +DEFINE_EVENT(nfsd_file_open_class, name, \ + TP_PROTO( \ + const struct nfsd_file *nf, \ + __be32 status \ + ), \ + TP_ARGS(nf, status)) + +DEFINE_NFSD_FILE_OPEN_EVENT(nfsd_file_open); +DEFINE_NFSD_FILE_OPEN_EVENT(nfsd_file_opened); + TRACE_EVENT(nfsd_file_is_cached, TP_PROTO( const struct inode *inode, -- cgit v1.2.3 From 7827c81f0248e3c2f40d438b020f3d222f002171 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Fri, 6 Jan 2023 12:43:37 -0500 Subject: Revert "SUNRPC: Use RMW bitops in single-threaded hot paths" The premise that "Once an svc thread is scheduled and executing an RPC, no other processes will touch svc_rqst::rq_flags" is false. svc_xprt_enqueue() examines the RQ_BUSY flag in scheduled nfsd threads when determining which thread to wake up next. Found via KCSAN. Fixes: 28df0988815f ("SUNRPC: Use RMW bitops in single-threaded hot paths") Reviewed-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/nfs4proc.c | 7 +++---- fs/nfsd/nfs4xdr.c | 2 +- net/sunrpc/auth_gss/svcauth_gss.c | 4 ++-- net/sunrpc/svc.c | 6 +++--- net/sunrpc/svc_xprt.c | 2 +- net/sunrpc/svcsock.c | 8 ++++---- net/sunrpc/xprtrdma/svc_rdma_transport.c | 2 +- 7 files changed, 15 insertions(+), 16 deletions(-) (limited to 'fs') diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c index bd880d55f565..9b81d012666e 100644 --- a/fs/nfsd/nfs4proc.c +++ b/fs/nfsd/nfs4proc.c @@ -937,7 +937,7 @@ nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, * the client wants us to do more in this compound: */ if (!nfsd4_last_compound_op(rqstp)) - __clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags); + clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags); /* check stateid */ status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh, @@ -2607,12 +2607,11 @@ nfsd4_proc_compound(struct svc_rqst *rqstp) cstate->minorversion = args->minorversion; fh_init(current_fh, NFS4_FHSIZE); fh_init(save_fh, NFS4_FHSIZE); - /* * Don't use the deferral mechanism for NFSv4; compounds make it * too hard to avoid non-idempotency problems. */ - __clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags); + clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags); /* * According to RFC3010, this takes precedence over all other errors. @@ -2734,7 +2733,7 @@ encode_op: out: cstate->status = status; /* Reset deferral mechanism for RPC deferrals */ - __set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags); + set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags); return rpc_success; } diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c index ebb4d02a42ce..97edb32be77f 100644 --- a/fs/nfsd/nfs4xdr.c +++ b/fs/nfsd/nfs4xdr.c @@ -2523,7 +2523,7 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp) argp->rqstp->rq_cachetype = cachethis ? RC_REPLBUFF : RC_NOCACHE; if (readcount > 1 || max_reply > PAGE_SIZE - auth_slack) - __clear_bit(RQ_SPLICE_OK, &argp->rqstp->rq_flags); + clear_bit(RQ_SPLICE_OK, &argp->rqstp->rq_flags); return true; } diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c index 148bb0a7fa5b..acb822b23af1 100644 --- a/net/sunrpc/auth_gss/svcauth_gss.c +++ b/net/sunrpc/auth_gss/svcauth_gss.c @@ -923,7 +923,7 @@ unwrap_integ_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct g * rejecting the server-computed MIC in this somewhat rare case, * do not use splice with the GSS integrity service. */ - __clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags); + clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags); /* Did we already verify the signature on the original pass through? */ if (rqstp->rq_deferred) @@ -990,7 +990,7 @@ unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gs int pad, remaining_len, offset; u32 rseqno; - __clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags); + clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags); priv_len = svc_getnl(&buf->head[0]); if (rqstp->rq_deferred) { diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c index 8f1b596db33f..b1da586c4476 100644 --- a/net/sunrpc/svc.c +++ b/net/sunrpc/svc.c @@ -1243,10 +1243,10 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv) goto err_short_len; /* Will be turned off by GSS integrity and privacy services */ - __set_bit(RQ_SPLICE_OK, &rqstp->rq_flags); + set_bit(RQ_SPLICE_OK, &rqstp->rq_flags); /* Will be turned off only when NFSv4 Sessions are used */ - __set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags); - __clear_bit(RQ_DROPME, &rqstp->rq_flags); + set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags); + clear_bit(RQ_DROPME, &rqstp->rq_flags); svc_putu32(resv, rqstp->rq_xid); diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c index 2106003645a7..c2ce12538008 100644 --- a/net/sunrpc/svc_xprt.c +++ b/net/sunrpc/svc_xprt.c @@ -1238,7 +1238,7 @@ static struct cache_deferred_req *svc_defer(struct cache_req *req) trace_svc_defer(rqstp); svc_xprt_get(rqstp->rq_xprt); dr->xprt = rqstp->rq_xprt; - __set_bit(RQ_DROPME, &rqstp->rq_flags); + set_bit(RQ_DROPME, &rqstp->rq_flags); dr->handle.revisit = svc_revisit; return &dr->handle; diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c index 2fc98fea59b4..e833103f4629 100644 --- a/net/sunrpc/svcsock.c +++ b/net/sunrpc/svcsock.c @@ -298,9 +298,9 @@ static void svc_sock_setbufsize(struct svc_sock *svsk, unsigned int nreqs) static void svc_sock_secure_port(struct svc_rqst *rqstp) { if (svc_port_is_privileged(svc_addr(rqstp))) - __set_bit(RQ_SECURE, &rqstp->rq_flags); + set_bit(RQ_SECURE, &rqstp->rq_flags); else - __clear_bit(RQ_SECURE, &rqstp->rq_flags); + clear_bit(RQ_SECURE, &rqstp->rq_flags); } /* @@ -1008,9 +1008,9 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp) rqstp->rq_xprt_ctxt = NULL; rqstp->rq_prot = IPPROTO_TCP; if (test_bit(XPT_LOCAL, &svsk->sk_xprt.xpt_flags)) - __set_bit(RQ_LOCAL, &rqstp->rq_flags); + set_bit(RQ_LOCAL, &rqstp->rq_flags); else - __clear_bit(RQ_LOCAL, &rqstp->rq_flags); + clear_bit(RQ_LOCAL, &rqstp->rq_flags); p = (__be32 *)rqstp->rq_arg.head[0].iov_base; calldir = p[1]; diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c index 199fa012f18a..94b20fb47135 100644 --- a/net/sunrpc/xprtrdma/svc_rdma_transport.c +++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c @@ -602,7 +602,7 @@ static int svc_rdma_has_wspace(struct svc_xprt *xprt) static void svc_rdma_secure_port(struct svc_rqst *rqstp) { - __set_bit(RQ_SECURE, &rqstp->rq_flags); + set_bit(RQ_SECURE, &rqstp->rq_flags); } static void svc_rdma_kill_temp_xprt(struct svc_xprt *xprt) -- cgit v1.2.3 From 5304930dbae82d259bcf7e5611db7c81e7a42eff Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sat, 7 Jan 2023 10:15:35 -0500 Subject: NFSD: Use set_bit(RQ_DROPME) The premise that "Once an svc thread is scheduled and executing an RPC, no other processes will touch svc_rqst::rq_flags" is false. svc_xprt_enqueue() examines the RQ_BUSY flag in scheduled nfsd threads when determining which thread to wake up next. Fixes: 9315564747cb ("NFSD: Use only RQ_DROPME to signal the need to drop a reply") Signed-off-by: Chuck Lever --- fs/nfsd/nfsproc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfsd/nfsproc.c b/fs/nfsd/nfsproc.c index a5570cf75f3f..9744443c3965 100644 --- a/fs/nfsd/nfsproc.c +++ b/fs/nfsd/nfsproc.c @@ -211,7 +211,7 @@ nfsd_proc_read(struct svc_rqst *rqstp) if (resp->status == nfs_ok) resp->status = fh_getattr(&resp->fh, &resp->stat); else if (resp->status == nfserr_jukebox) - __set_bit(RQ_DROPME, &rqstp->rq_flags); + set_bit(RQ_DROPME, &rqstp->rq_flags); return rpc_success; } @@ -246,7 +246,7 @@ nfsd_proc_write(struct svc_rqst *rqstp) if (resp->status == nfs_ok) resp->status = fh_getattr(&resp->fh, &resp->stat); else if (resp->status == nfserr_jukebox) - __set_bit(RQ_DROPME, &rqstp->rq_flags); + set_bit(RQ_DROPME, &rqstp->rq_flags); return rpc_success; } -- cgit v1.2.3