From 86438186a7b372728c74055d441549e17915cddb Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 2 Feb 2021 14:22:51 -0500 Subject: NFSv4: Simplify nfs4_retry_setlk() Simplify the code that retries locks based on notification events. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 54 +++++++++++++++++++++++------------------------------- 1 file changed, 23 insertions(+), 31 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index c65c4b41e2c1..bca1726bcd55 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -7250,22 +7250,22 @@ nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd, #ifdef CONFIG_NFS_V4_1 struct nfs4_lock_waiter { - struct task_struct *task; struct inode *inode; - struct nfs_lowner *owner; + struct nfs_lowner owner; + wait_queue_entry_t wait; }; static int nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key) { - int ret; - struct nfs4_lock_waiter *waiter = wait->private; + struct nfs4_lock_waiter *waiter = + container_of(wait, struct nfs4_lock_waiter, wait); /* NULL key means to wake up everyone */ if (key) { struct cb_notify_lock_args *cbnl = key; struct nfs_lowner *lowner = &cbnl->cbnl_owner, - *wowner = waiter->owner; + *wowner = &waiter->owner; /* Only wake if the callback was for the same owner. */ if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev) @@ -7276,53 +7276,45 @@ nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, vo return 0; } - /* override "private" so we can use default_wake_function */ - wait->private = waiter->task; - ret = woken_wake_function(wait, mode, flags, key); - if (ret) - list_del_init(&wait->entry); - wait->private = waiter; - return ret; + return woken_wake_function(wait, mode, flags, key); } static int nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request) { - int status = -ERESTARTSYS; struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner; struct nfs_server *server = NFS_SERVER(state->inode); struct nfs_client *clp = server->nfs_client; wait_queue_head_t *q = &clp->cl_lock_waitq; - struct nfs_lowner owner = { .clientid = clp->cl_clientid, - .id = lsp->ls_seqid.owner_id, - .s_dev = server->s_dev }; - struct nfs4_lock_waiter waiter = { .task = current, - .inode = state->inode, - .owner = &owner}; - wait_queue_entry_t wait; + struct nfs4_lock_waiter waiter = { + .inode = state->inode, + .owner = { .clientid = clp->cl_clientid, + .id = lsp->ls_seqid.owner_id, + .s_dev = server->s_dev }, + }; + int status; /* Don't bother with waitqueue if we don't expect a callback */ if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags)) return nfs4_retry_setlk_simple(state, cmd, request); - init_wait(&wait); - wait.private = &waiter; - wait.func = nfs4_wake_lock_waiter; + init_wait(&waiter.wait); + waiter.wait.func = nfs4_wake_lock_waiter; + add_wait_queue(q, &waiter.wait); - while(!signalled()) { - add_wait_queue(q, &wait); + do { status = nfs4_proc_setlk(state, cmd, request); - if ((status != -EAGAIN) || IS_SETLK(cmd)) { - finish_wait(q, &wait); + if (status != -EAGAIN || IS_SETLK(cmd)) break; - } status = -ERESTARTSYS; freezer_do_not_count(); - wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT); + wait_woken(&waiter.wait, TASK_INTERRUPTIBLE, + NFS4_LOCK_MAXTIMEOUT); freezer_count(); - finish_wait(q, &wait); - } + } while (!signalled()); + + remove_wait_queue(q, &waiter.wait); return status; } -- cgit v1.2.3 From 6453bcd0d8bd67fce9e0f566250caf864b5032f6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 4 Mar 2021 20:29:50 -0500 Subject: NFS: Fix up incorrect documentation Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 8 ++++---- fs/nfs/io.c | 2 +- fs/nfs/nfs4state.c | 2 +- fs/nfs/pagelist.c | 4 ++-- 4 files changed, 8 insertions(+), 8 deletions(-) (limited to 'fs') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 04bf8066980c..6a29de964268 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -114,7 +114,7 @@ nfs4_do_check_delegation(struct inode *inode, fmode_t flags, bool mark) return ret; } /** - * nfs_have_delegation - check if inode has a delegation, mark it + * nfs4_have_delegation - check if inode has a delegation, mark it * NFS_DELEGATION_REFERENCED if there is one. * @inode: inode to check * @flags: delegation types to check for @@ -674,7 +674,7 @@ void nfs_inode_evict_delegation(struct inode *inode) } /** - * nfs_inode_return_delegation - synchronously return a delegation + * nfs4_inode_return_delegation - synchronously return a delegation * @inode: inode to process * * This routine will always flush any dirty data to disk on the @@ -697,7 +697,7 @@ int nfs4_inode_return_delegation(struct inode *inode) } /** - * nfs_inode_return_delegation_on_close - asynchronously return a delegation + * nfs4_inode_return_delegation_on_close - asynchronously return a delegation * @inode: inode to process * * This routine is called on file close in order to determine if the @@ -811,7 +811,7 @@ void nfs_expire_all_delegations(struct nfs_client *clp) } /** - * nfs_super_return_all_delegations - return delegations for one superblock + * nfs_server_return_all_delegations - return delegations for one superblock * @server: pointer to nfs_server to process * */ diff --git a/fs/nfs/io.c b/fs/nfs/io.c index 5088fda9b453..b5551ed8f648 100644 --- a/fs/nfs/io.c +++ b/fs/nfs/io.c @@ -104,7 +104,7 @@ static void nfs_block_buffered(struct nfs_inode *nfsi, struct inode *inode) } /** - * nfs_end_io_direct - declare the file is being used for direct i/o + * nfs_start_io_direct - declare the file is being used for direct i/o * @inode: file inode * * Declare that a direct I/O operation is about to start, and ensure diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 3a51351bdc6a..9db9b11acb2a 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -645,7 +645,7 @@ void nfs4_purge_state_owners(struct nfs_server *server, struct list_head *head) } /** - * nfs4_purge_state_owners - Release all cached state owners + * nfs4_free_state_owners - Release all cached state owners * @head: resulting list of state owners * * Frees a list of state owners that was generated by diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 78c9c4bdef2b..6c20b28d9d7c 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -577,7 +577,7 @@ static void nfs_clear_request(struct nfs_page *req) } /** - * nfs_release_request - Release the count on an NFS read/write request + * nfs_free_request - Release the count on an NFS read/write request * @req: request to release * * Note: Should never be called with the spinlock held! @@ -1152,7 +1152,7 @@ nfs_pageio_cleanup_request(struct nfs_pageio_descriptor *desc, } /** - * nfs_pageio_add_request - Attempt to coalesce a request into a page list. + * __nfs_pageio_add_request - Attempt to coalesce a request into a page list. * @desc: destination io descriptor * @req: request * -- cgit v1.2.3 From ee3707ae2c1f1327ad5188836b7ab62ed2c93b28 Mon Sep 17 00:00:00 2001 From: Nagendra S Tomar Date: Tue, 16 Mar 2021 10:25:14 +0000 Subject: nfs: Subsequent READDIR calls should carry non-zero cookieverifier If the loop in nfs_readdir_xdr_to_array() runs more than once, subsequent READDIR RPCs may wrongly carry a zero cookie verifier and non-zero cookie. Make sure subsequent calls to READDIR carry the cookie verifier returned by the first call. Signed-off-by: Nagendra S Tomar Fixes: b593c09f83a2 ("NFS: Improve handling of directory verifiers") Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index fc4f490f2d78..08a1e2e31d0b 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -866,6 +866,8 @@ static int nfs_readdir_xdr_to_array(struct nfs_readdir_descriptor *desc, break; } + verf_arg = verf_res; + status = nfs_readdir_page_filler(desc, entry, pages, pglen, arrays, narrays); } while (!status && nfs_readdir_page_needs_filling(page)); -- cgit v1.2.3 From 13884ff2bef01df37c450c6dd09122f92333dccc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 16 Mar 2021 07:57:40 -0400 Subject: NFS: Fix handling of cookie verifier in uncached_readdir() If we're doing uncached readdir(), then the readdir cookie could be different from the one cached in the nfs_inode. We should therefore ensure that we save that one in the struct nfs_open_dir_context. Fixes: 35df59d3ef69 ("NFS: Reduce number of RPC calls when doing uncached readdir") Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 08a1e2e31d0b..2cf2a7d92faf 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -976,10 +976,10 @@ static int readdir_search_pagecache(struct nfs_readdir_descriptor *desc) /* * Once we've found the start of the dirent within a page: fill 'er up... */ -static void nfs_do_filldir(struct nfs_readdir_descriptor *desc) +static void nfs_do_filldir(struct nfs_readdir_descriptor *desc, + const __be32 *verf) { struct file *file = desc->file; - struct nfs_inode *nfsi = NFS_I(file_inode(file)); struct nfs_cache_array *array; unsigned int i = 0; @@ -993,7 +993,7 @@ static void nfs_do_filldir(struct nfs_readdir_descriptor *desc) desc->eof = true; break; } - memcpy(desc->verf, nfsi->cookieverf, sizeof(desc->verf)); + memcpy(desc->verf, verf, sizeof(desc->verf)); if (i < (array->size-1)) desc->dir_cookie = array->array[i+1].cookie; else @@ -1050,7 +1050,7 @@ static int uncached_readdir(struct nfs_readdir_descriptor *desc) for (i = 0; !desc->eof && i < sz && arrays[i]; i++) { desc->page = arrays[i]; - nfs_do_filldir(desc); + nfs_do_filldir(desc, verf); } desc->page = NULL; @@ -1071,6 +1071,7 @@ static int nfs_readdir(struct file *file, struct dir_context *ctx) { struct dentry *dentry = file_dentry(file); struct inode *inode = d_inode(dentry); + struct nfs_inode *nfsi = NFS_I(inode); struct nfs_open_dir_context *dir_ctx = file->private_data; struct nfs_readdir_descriptor *desc; int res; @@ -1124,7 +1125,7 @@ static int nfs_readdir(struct file *file, struct dir_context *ctx) break; } if (res == -ETOOSMALL && desc->plus) { - clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags); + clear_bit(NFS_INO_ADVISE_RDPLUS, &nfsi->flags); nfs_zap_caches(inode); desc->page_index = 0; desc->plus = false; @@ -1134,7 +1135,7 @@ static int nfs_readdir(struct file *file, struct dir_context *ctx) if (res < 0) break; - nfs_do_filldir(desc); + nfs_do_filldir(desc, nfsi->cookieverf); nfs_readdir_page_unlock_and_put_cached(desc); } while (!desc->eof); -- cgit v1.2.3 From f892c41c14e0fa3d78ce37de1d5c8161ed13bf08 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 17 Mar 2021 08:46:19 -0400 Subject: NFS: Only change the cookie verifier if the directory page cache is empty The cached NFSv3/v4 readdir cookies are associated with a verifier, which is checked by the server on subsequent calls to readdir, and is only expected to change when the cookies (and hence also the page cache contents) are considered invalid. We therefore do have to store the verifier, but only when the page cache is empty. Fixes: b593c09f83a2 ("NFS: Improve handling of directory verifiers") Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 2cf2a7d92faf..0cd7c59a6601 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -929,7 +929,12 @@ static int find_and_lock_cache_page(struct nfs_readdir_descriptor *desc) } return res; } - memcpy(nfsi->cookieverf, verf, sizeof(nfsi->cookieverf)); + /* + * Set the cookie verifier if the page cache was empty + */ + if (desc->page_index == 0) + memcpy(nfsi->cookieverf, verf, + sizeof(nfsi->cookieverf)); } res = nfs_readdir_search_array(desc); if (res == 0) { -- cgit v1.2.3 From c09f11ef35955785f92369e25819bf0629df2e59 Mon Sep 17 00:00:00 2001 From: Randy Dunlap Date: Mon, 1 Mar 2021 16:19:30 -0800 Subject: NFS: fs_context: validate UDP retrans to prevent shift out-of-bounds Fix shift out-of-bounds in xprt_calc_majortimeo(). This is caused by a garbage timeout (retrans) mount option being passed to nfs mount, in this case from syzkaller. If the protocol is XPRT_TRANSPORT_UDP, then 'retrans' is a shift value for a 64-bit long integer, so 'retrans' cannot be >= 64. If it is >= 64, fail the mount and return an error. Fixes: 9954bf92c0cd ("NFS: Move mount parameterisation bits into their own file") Reported-by: syzbot+ba2e91df8f74809417fa@syzkaller.appspotmail.com Reported-by: syzbot+f3a0fa110fd630ab56c8@syzkaller.appspotmail.com Signed-off-by: Randy Dunlap Cc: Trond Myklebust Cc: Anna Schumaker Cc: linux-nfs@vger.kernel.org Cc: David Howells Cc: Al Viro Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- fs/nfs/fs_context.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'fs') diff --git a/fs/nfs/fs_context.c b/fs/nfs/fs_context.c index 971a9251c1d9..902db1262d2b 100644 --- a/fs/nfs/fs_context.c +++ b/fs/nfs/fs_context.c @@ -973,6 +973,15 @@ static int nfs23_parse_monolithic(struct fs_context *fc, memset(mntfh->data + mntfh->size, 0, sizeof(mntfh->data) - mntfh->size); + /* + * for proto == XPRT_TRANSPORT_UDP, which is what uses + * to_exponential, implying shift: limit the shift value + * to BITS_PER_LONG (majortimeo is unsigned long) + */ + if (!(data->flags & NFS_MOUNT_TCP)) /* this will be UDP */ + if (data->retrans >= 64) /* shift value is too large */ + goto out_invalid_data; + /* * Translate to nfs_fs_context, which nfs_fill_super * can deal with. @@ -1073,6 +1082,9 @@ out_no_address: out_invalid_fh: return nfs_invalf(fc, "NFS: invalid root filehandle"); + +out_invalid_data: + return nfs_invalf(fc, "NFS: invalid binary mount data"); } #if IS_ENABLED(CONFIG_NFS_V4) -- cgit v1.2.3 From c2508730d6bb62196783b3de28a09782c97d4365 Mon Sep 17 00:00:00 2001 From: Bhaskar Chowdhury Date: Mon, 15 Mar 2021 06:54:10 +0530 Subject: nfs: Fix a typo in the file nfs42xattr.c s/attribues/attributes/ Signed-off-by: Bhaskar Chowdhury Acked-by: Randy Dunlap Signed-off-by: Trond Myklebust --- fs/nfs/nfs42xattr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/nfs42xattr.c b/fs/nfs/nfs42xattr.c index 6c2ce799150f..1c4d2a05b401 100644 --- a/fs/nfs/nfs42xattr.c +++ b/fs/nfs/nfs42xattr.c @@ -168,7 +168,7 @@ nfs4_xattr_entry_lru_del(struct nfs4_xattr_entry *entry) * make it easier to copy the value after an RPC, even if * the value will not be passed up to application (e.g. * for a 'query' getxattr with NULL buffer). - * @len: Length of the value. Can be 0 for zero-length attribues. + * @len: Length of the value. Can be 0 for zero-length attributes. * @value and @pages will be NULL if @len is 0. */ static struct nfs4_xattr_entry * -- cgit v1.2.3 From 90ff57bf4df2825ace0f61befdc3f87c79838ec2 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 30 Mar 2021 09:01:50 -0400 Subject: NFS: Fix up the support for CONFIG_NFS_DISABLE_UDP_SUPPORT Rather than removing the support in nfs_init_timeout_values(), we should just fix up the validation checks in the mount option parsers. Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 2 -- fs/nfs/fs_context.c | 54 ++++++++++++++++++++++++++++++++++------------------- 2 files changed, 35 insertions(+), 21 deletions(-) (limited to 'fs') diff --git a/fs/nfs/client.c b/fs/nfs/client.c index ff5c4d0d6d13..399a8eb15397 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -476,7 +476,6 @@ void nfs_init_timeout_values(struct rpc_timeout *to, int proto, to->to_maxval = to->to_initval; to->to_exponential = 0; break; -#ifndef CONFIG_NFS_DISABLE_UDP_SUPPORT case XPRT_TRANSPORT_UDP: if (retrans == NFS_UNSPEC_RETRANS) to->to_retries = NFS_DEF_UDP_RETRANS; @@ -487,7 +486,6 @@ void nfs_init_timeout_values(struct rpc_timeout *to, int proto, to->to_maxval = NFS_MAX_UDP_TIMEOUT; to->to_exponential = 1; break; -#endif default: BUG(); } diff --git a/fs/nfs/fs_context.c b/fs/nfs/fs_context.c index 902db1262d2b..cdf32b9a6c35 100644 --- a/fs/nfs/fs_context.c +++ b/fs/nfs/fs_context.c @@ -283,20 +283,40 @@ static int nfs_verify_server_address(struct sockaddr *addr) return 0; } +#ifdef CONFIG_NFS_DISABLE_UDP_SUPPORT +static bool nfs_server_transport_udp_invalid(const struct nfs_fs_context *ctx) +{ + return true; +} +#else +static bool nfs_server_transport_udp_invalid(const struct nfs_fs_context *ctx) +{ + if (ctx->version == 4) + return true; + return false; +} +#endif + /* * Sanity check the NFS transport protocol. - * */ -static void nfs_validate_transport_protocol(struct nfs_fs_context *ctx) +static int nfs_validate_transport_protocol(struct fs_context *fc, + struct nfs_fs_context *ctx) { switch (ctx->nfs_server.protocol) { case XPRT_TRANSPORT_UDP: + if (nfs_server_transport_udp_invalid(ctx)) + goto out_invalid_transport_udp; + break; case XPRT_TRANSPORT_TCP: case XPRT_TRANSPORT_RDMA: break; default: ctx->nfs_server.protocol = XPRT_TRANSPORT_TCP; } + return 0; +out_invalid_transport_udp: + return nfs_invalf(fc, "NFS: Unsupported transport protocol udp"); } /* @@ -305,8 +325,6 @@ static void nfs_validate_transport_protocol(struct nfs_fs_context *ctx) */ static void nfs_set_mount_transport_protocol(struct nfs_fs_context *ctx) { - nfs_validate_transport_protocol(ctx); - if (ctx->mount_server.protocol == XPRT_TRANSPORT_UDP || ctx->mount_server.protocol == XPRT_TRANSPORT_TCP) return; @@ -929,6 +947,7 @@ static int nfs23_parse_monolithic(struct fs_context *fc, struct nfs_fh *mntfh = ctx->mntfh; struct sockaddr *sap = (struct sockaddr *)&ctx->nfs_server.address; int extra_flags = NFS_MOUNT_LEGACY_INTERFACE; + int ret; if (data == NULL) goto out_no_data; @@ -1054,6 +1073,10 @@ static int nfs23_parse_monolithic(struct fs_context *fc, goto generic; } + ret = nfs_validate_transport_protocol(fc, ctx); + if (ret) + return ret; + ctx->skip_reconfig_option_check = true; return 0; @@ -1155,6 +1178,7 @@ static int nfs4_parse_monolithic(struct fs_context *fc, { struct nfs_fs_context *ctx = nfs_fc2context(fc); struct sockaddr *sap = (struct sockaddr *)&ctx->nfs_server.address; + int ret; char *c; if (!data) { @@ -1227,9 +1251,9 @@ static int nfs4_parse_monolithic(struct fs_context *fc, ctx->acdirmin = data->acdirmin; ctx->acdirmax = data->acdirmax; ctx->nfs_server.protocol = data->proto; - nfs_validate_transport_protocol(ctx); - if (ctx->nfs_server.protocol == XPRT_TRANSPORT_UDP) - goto out_invalid_transport_udp; + ret = nfs_validate_transport_protocol(fc, ctx); + if (ret) + return ret; done: ctx->skip_reconfig_option_check = true; return 0; @@ -1240,9 +1264,6 @@ out_inval_auth: out_no_address: return nfs_invalf(fc, "NFS4: mount program didn't pass remote address"); - -out_invalid_transport_udp: - return nfs_invalf(fc, "NFS: Unsupported transport protocol udp"); } #endif @@ -1307,6 +1328,10 @@ static int nfs_fs_context_validate(struct fs_context *fc) if (!nfs_verify_server_address(sap)) goto out_no_address; + ret = nfs_validate_transport_protocol(fc, ctx); + if (ret) + return ret; + if (ctx->version == 4) { if (IS_ENABLED(CONFIG_NFS_V4)) { if (ctx->nfs_server.protocol == XPRT_TRANSPORT_RDMA) @@ -1315,9 +1340,6 @@ static int nfs_fs_context_validate(struct fs_context *fc) port = NFS_PORT; max_namelen = NFS4_MAXNAMLEN; max_pathlen = NFS4_MAXPATHLEN; - nfs_validate_transport_protocol(ctx); - if (ctx->nfs_server.protocol == XPRT_TRANSPORT_UDP) - goto out_invalid_transport_udp; ctx->flags &= ~(NFS_MOUNT_NONLM | NFS_MOUNT_NOACL | NFS_MOUNT_VER3 | NFS_MOUNT_LOCAL_FLOCK | NFS_MOUNT_LOCAL_FCNTL); @@ -1326,10 +1348,6 @@ static int nfs_fs_context_validate(struct fs_context *fc) } } else { nfs_set_mount_transport_protocol(ctx); -#ifdef CONFIG_NFS_DISABLE_UDP_SUPPORT - if (ctx->nfs_server.protocol == XPRT_TRANSPORT_UDP) - goto out_invalid_transport_udp; -#endif if (ctx->nfs_server.protocol == XPRT_TRANSPORT_RDMA) port = NFS_RDMA_PORT; } @@ -1363,8 +1381,6 @@ out_no_device_name: out_v4_not_compiled: nfs_errorf(fc, "NFS: NFSv4 is not compiled into kernel"); return -EPROTONOSUPPORT; -out_invalid_transport_udp: - return nfs_invalf(fc, "NFS: Unsupported transport protocol udp"); out_no_address: return nfs_invalf(fc, "NFS: mount program didn't pass remote address"); out_mountproto_mismatch: -- cgit v1.2.3 From c9301cb35b59ad7d733a7332f3aefd4da1382468 Mon Sep 17 00:00:00 2001 From: Eryu Guan Date: Tue, 23 Mar 2021 10:57:13 +0800 Subject: nfs: hornor timeo and retrans option when mounting NFSv3 Mounting NFSv3 uses default timeout parameters specified by underlying sunrpc transport, and mount options like 'timeo' and 'retrans', unlike NFSv4, are not honored. But sometimes we want to set non-default timeout value when mounting NFSv3, so pass 'timeo' and 'retrans' to nfs_mount() and fill the 'timeout' field of struct rpc_create_args before creating RPC connection. This is also consistent with NFSv4 behavior. Note that this only sets the timeout value of rpc connection to mountd, but the timeout of rpcbind connection should be set as well. A later patch will fix the rpcbind part. Signed-off-by: Eryu Guan Signed-off-by: Trond Myklebust --- fs/nfs/internal.h | 2 +- fs/nfs/mount_clnt.c | 14 +++++++++----- fs/nfs/super.c | 2 +- 3 files changed, 11 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 7b644d6c09e4..cf0d7db24d44 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -180,7 +180,7 @@ struct nfs_mount_request { struct net *net; }; -extern int nfs_mount(struct nfs_mount_request *info); +extern int nfs_mount(struct nfs_mount_request *info, int timeo, int retrans); extern void nfs_umount(const struct nfs_mount_request *info); /* client.c */ diff --git a/fs/nfs/mount_clnt.c b/fs/nfs/mount_clnt.c index dda5c3e65d8d..c5e3b6b3366a 100644 --- a/fs/nfs/mount_clnt.c +++ b/fs/nfs/mount_clnt.c @@ -136,14 +136,16 @@ struct mnt_fhstatus { /** * nfs_mount - Obtain an NFS file handle for the given host and path * @info: pointer to mount request arguments + * @timeo: deciseconds the mount waits for a response before it retries + * @retrans: number of times the mount retries a request * - * Uses default timeout parameters specified by underlying transport. On - * successful return, the auth_flavs list and auth_flav_len will be populated - * with the list from the server or a faked-up list if the server didn't - * provide one. + * Uses timeout parameters specified by caller. On successful return, the + * auth_flavs list and auth_flav_len will be populated with the list from the + * server or a faked-up list if the server didn't provide one. */ -int nfs_mount(struct nfs_mount_request *info) +int nfs_mount(struct nfs_mount_request *info, int timeo, int retrans) { + struct rpc_timeout mnt_timeout; struct mountres result = { .fh = info->fh, .auth_count = info->auth_flav_len, @@ -158,6 +160,7 @@ int nfs_mount(struct nfs_mount_request *info) .protocol = info->protocol, .address = info->sap, .addrsize = info->salen, + .timeout = &mnt_timeout, .servername = info->hostname, .program = &mnt_program, .version = info->version, @@ -177,6 +180,7 @@ int nfs_mount(struct nfs_mount_request *info) if (info->noresvport) args.flags |= RPC_CLNT_CREATE_NONPRIVPORT; + nfs_init_timeout_values(&mnt_timeout, info->protocol, timeo, retrans); mnt_clnt = rpc_create(&args); if (IS_ERR(mnt_clnt)) goto out_clnt_err; diff --git a/fs/nfs/super.c b/fs/nfs/super.c index 94885c6f8f54..13a650750f04 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -867,7 +867,7 @@ static int nfs_request_mount(struct fs_context *fc, * Now ask the mount server to map our export path * to a file handle. */ - status = nfs_mount(&request); + status = nfs_mount(&request, ctx->timeo, ctx->retrans); if (status != 0) { dfprintk(MOUNT, "NFS: unable to mount server %s, error %d\n", request.hostname, status); -- cgit v1.2.3 From b876d708316bf9b6b9678eb2beb289b93cfe6369 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 24 Mar 2021 15:32:21 -0400 Subject: NFS: fix nfs_fetch_iversion() The change attribute is always set by all NFS client versions so get rid of the open-coded version. Fixes: 3cc55f4434b4 ("nfs: use change attribute for NFS re-exports") Signed-off-by: Trond Myklebust --- fs/nfs/export.c | 15 ++++----------- 1 file changed, 4 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/nfs/export.c b/fs/nfs/export.c index f2b34cfe286c..b347e3ce0cc8 100644 --- a/fs/nfs/export.c +++ b/fs/nfs/export.c @@ -171,17 +171,10 @@ static u64 nfs_fetch_iversion(struct inode *inode) { struct nfs_server *server = NFS_SERVER(inode); - /* Is this the right call?: */ - nfs_revalidate_inode(server, inode); - /* - * Also, note we're ignoring any returned error. That seems to be - * the practice for cache consistency information elsewhere in - * the server, but I'm not sure why. - */ - if (server->nfs_client->rpc_ops->version >= 4) - return inode_peek_iversion_raw(inode); - else - return time_to_chattr(&inode->i_ctime); + if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_CHANGE | + NFS_INO_REVAL_PAGECACHE)) + __nfs_revalidate_inode(server, inode); + return inode_peek_iversion_raw(inode); } const struct export_operations nfs_export_ops = { -- cgit v1.2.3 From beab450d8ea93cdf4c6cb7714bdc31a9e0f34738 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 26 Mar 2021 16:23:39 -0400 Subject: NFS: Fix fscache invalidation in nfs_set_cache_invalid() Ensure that we invalidate the fscache before we strip the NFS_INO_INVALID_DATA flag. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index a7fb076a5f44..ff737be559dc 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -223,11 +223,11 @@ void nfs_set_cache_invalid(struct inode *inode, unsigned long flags) if (!nfs_has_xattr_cache(nfsi)) flags &= ~NFS_INO_INVALID_XATTR; + if (flags & NFS_INO_INVALID_DATA) + nfs_fscache_invalidate(inode); if (inode->i_mapping->nrpages == 0) flags &= ~(NFS_INO_INVALID_DATA|NFS_INO_DATA_INVAL_DEFER); nfsi->cache_validity |= flags; - if (flags & NFS_INO_INVALID_DATA) - nfs_fscache_invalidate(inode); } EXPORT_SYMBOL_GPL(nfs_set_cache_invalid); -- cgit v1.2.3 From 332d1a0373be32a3a3c152756bca45ff4f4e11b5 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 18:15:36 -0400 Subject: NFS: nfs4_bitmask_adjust() must not change the server global bitmasks As currently set, the calls to nfs4_bitmask_adjust() will end up overwriting the contents of the nfs_server cache_consistency_bitmask field. The intention here should be to modify a private copy of that mask in the close/delegreturn/write arguments. Fixes: 76bd5c016ef4 ("NFSv4: make cache consistency bitmask dynamic") Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 56 +++++++++++++++++++++++++++++++++---------------------- 1 file changed, 34 insertions(+), 22 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index bca1726bcd55..1cf98c40e3b2 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -108,9 +108,10 @@ static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *, static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *, const struct cred *, bool); #endif -static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode, - struct nfs_server *server, - struct nfs4_label *label); +static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], + const __u32 *src, struct inode *inode, + struct nfs_server *server, + struct nfs4_label *label); #ifdef CONFIG_NFS_V4_SECURITY_LABEL static inline struct nfs4_label * @@ -3591,6 +3592,7 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data) struct nfs4_closedata *calldata = data; struct nfs4_state *state = calldata->state; struct inode *inode = calldata->inode; + struct nfs_server *server = NFS_SERVER(inode); struct pnfs_layout_hdr *lo; bool is_rdonly, is_wronly, is_rdwr; int call_close = 0; @@ -3647,8 +3649,10 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data) if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) { /* Close-to-open cache consistency revalidation */ if (!nfs4_have_delegation(inode, FMODE_READ)) { - calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask; - nfs4_bitmask_adjust(calldata->arg.bitmask, inode, NFS_SERVER(inode), NULL); + nfs4_bitmask_set(calldata->arg.bitmask_store, + server->cache_consistency_bitmask, + inode, server, NULL); + calldata->arg.bitmask = calldata->arg.bitmask_store; } else calldata->arg.bitmask = NULL; } @@ -5416,19 +5420,17 @@ bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr) return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0; } -static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode, - struct nfs_server *server, - struct nfs4_label *label) +static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src, + struct inode *inode, struct nfs_server *server, + struct nfs4_label *label) { - unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity); + unsigned int i; - if ((cache_validity & NFS_INO_INVALID_DATA) || - (cache_validity & NFS_INO_REVAL_PAGECACHE) || - (cache_validity & NFS_INO_REVAL_FORCED) || - (cache_validity & NFS_INO_INVALID_OTHER)) - nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode); + memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ); + if (cache_validity & (NFS_INO_INVALID_CHANGE | NFS_INO_REVAL_PAGECACHE)) + bitmask[0] |= FATTR4_WORD0_CHANGE; if (cache_validity & NFS_INO_INVALID_ATIME) bitmask[1] |= FATTR4_WORD1_TIME_ACCESS; if (cache_validity & NFS_INO_INVALID_OTHER) @@ -5437,16 +5439,22 @@ static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode, FATTR4_WORD1_NUMLINKS; if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL) bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL; - if (cache_validity & NFS_INO_INVALID_CHANGE) - bitmask[0] |= FATTR4_WORD0_CHANGE; if (cache_validity & NFS_INO_INVALID_CTIME) bitmask[1] |= FATTR4_WORD1_TIME_METADATA; if (cache_validity & NFS_INO_INVALID_MTIME) bitmask[1] |= FATTR4_WORD1_TIME_MODIFY; - if (cache_validity & NFS_INO_INVALID_SIZE) - bitmask[0] |= FATTR4_WORD0_SIZE; if (cache_validity & NFS_INO_INVALID_BLOCKS) bitmask[1] |= FATTR4_WORD1_SPACE_USED; + + if (nfs4_have_delegation(inode, FMODE_READ) && + !(cache_validity & NFS_INO_REVAL_FORCED)) + bitmask[0] &= ~FATTR4_WORD0_SIZE; + else if (cache_validity & + (NFS_INO_INVALID_SIZE | NFS_INO_REVAL_PAGECACHE)) + bitmask[0] |= FATTR4_WORD0_SIZE; + + for (i = 0; i < NFS4_BITMASK_SZ; i++) + bitmask[i] &= server->attr_bitmask[i]; } static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr, @@ -5459,8 +5467,10 @@ static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr, hdr->args.bitmask = NULL; hdr->res.fattr = NULL; } else { - hdr->args.bitmask = server->cache_consistency_bitmask; - nfs4_bitmask_adjust(hdr->args.bitmask, hdr->inode, server, NULL); + nfs4_bitmask_set(hdr->args.bitmask_store, + server->cache_consistency_bitmask, + hdr->inode, server, NULL); + hdr->args.bitmask = hdr->args.bitmask_store; } if (!hdr->pgio_done_cb) @@ -6502,8 +6512,10 @@ static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, data->args.fhandle = &data->fh; data->args.stateid = &data->stateid; - data->args.bitmask = server->cache_consistency_bitmask; - nfs4_bitmask_adjust(data->args.bitmask, inode, server, NULL); + nfs4_bitmask_set(data->args.bitmask_store, + server->cache_consistency_bitmask, inode, server, + NULL); + data->args.bitmask = data->args.bitmask_store; nfs_copy_fh(&data->fh, NFS_FH(inode)); nfs4_stateid_copy(&data->stateid, stateid); data->res.fattr = &data->fattr; -- cgit v1.2.3 From e99812e1382f0bfb6149393262bc70645c9f537a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 28 Mar 2021 18:12:03 -0400 Subject: NFS: Fix attribute bitmask in _nfs42_proc_fallocate() We can't use nfs4_fattr_bitmap as a bitmask, because it hasn't been filtered to represent the attributes supported by the server. Instead, let's revert to using server->cache_consistency_bitmask after adding in the missing SPACE_USED attribute. Fixes: 913eca1aea87 ("NFS: Fallocate should use the nfs4_fattr_bitmap") Signed-off-by: Trond Myklebust --- fs/nfs/nfs42proc.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 094024b0aca1..597005bc8a05 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -46,11 +46,12 @@ static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep, { struct inode *inode = file_inode(filep); struct nfs_server *server = NFS_SERVER(inode); + u32 bitmask[3]; struct nfs42_falloc_args args = { .falloc_fh = NFS_FH(inode), .falloc_offset = offset, .falloc_length = len, - .falloc_bitmask = nfs4_fattr_bitmap, + .falloc_bitmask = bitmask, }; struct nfs42_falloc_res res = { .falloc_server = server, @@ -68,6 +69,10 @@ static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep, return status; } + memcpy(bitmask, server->cache_consistency_bitmask, sizeof(bitmask)); + if (server->attr_bitmask[1] & FATTR4_WORD1_SPACE_USED) + bitmask[1] |= FATTR4_WORD1_SPACE_USED; + res.falloc_fattr = nfs_alloc_fattr(); if (!res.falloc_fattr) return -ENOMEM; @@ -75,7 +80,8 @@ static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep, status = nfs4_call_sync(server->client, server, msg, &args.seq_args, &res.seq_res, 0); if (status == 0) - status = nfs_post_op_update_inode(inode, res.falloc_fattr); + status = nfs_post_op_update_inode_force_wcc(inode, + res.falloc_fattr); kfree(res.falloc_fattr); return status; -- cgit v1.2.3 From 99f23783224355e7022ceea9b8d9f62c0fd01bd8 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 28 Mar 2021 18:17:14 -0400 Subject: NFSv4.2: Always flush out writes in nfs42_proc_fallocate() Whether we're allocating or delallocating space, we should flush out the pending writes in order to avoid races with attribute updates. Fixes: 1e564d3dbd68 ("NFSv4.2: Fix a race in nfs42_proc_deallocate()") Signed-off-by: Trond Myklebust --- fs/nfs/nfs42proc.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 597005bc8a05..704a5246ccb5 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -90,7 +90,8 @@ static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep, static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep, loff_t offset, loff_t len) { - struct nfs_server *server = NFS_SERVER(file_inode(filep)); + struct inode *inode = file_inode(filep); + struct nfs_server *server = NFS_SERVER(inode); struct nfs4_exception exception = { }; struct nfs_lock_context *lock; int err; @@ -99,9 +100,13 @@ static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep, if (IS_ERR(lock)) return PTR_ERR(lock); - exception.inode = file_inode(filep); + exception.inode = inode; exception.state = lock->open_context->state; + err = nfs_sync_inode(inode); + if (err) + goto out; + do { err = _nfs42_proc_fallocate(msg, filep, lock, offset, len); if (err == -ENOTSUPP) { @@ -110,7 +115,7 @@ static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep, } err = nfs4_handle_exception(server, err, &exception); } while (exception.retry); - +out: nfs_put_lock_context(lock); return err; } @@ -148,16 +153,13 @@ int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len) return -EOPNOTSUPP; inode_lock(inode); - err = nfs_sync_inode(inode); - if (err) - goto out_unlock; err = nfs42_proc_fallocate(&msg, filep, offset, len); if (err == 0) truncate_pagecache_range(inode, offset, (offset + len) -1); if (err == -EOPNOTSUPP) NFS_SERVER(inode)->caps &= ~NFS_CAP_DEALLOCATE; -out_unlock: + inode_unlock(inode); return err; } -- cgit v1.2.3 From 9fdbfad1777cb4638f489eeb62d85432010c0031 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 29 Mar 2021 16:46:05 -0400 Subject: NFS: Deal correctly with attribute generation counter overflow We need to use unsigned long subtraction and then convert to signed in order to deal correcly with C overflow rules. Fixes: f5062003465c ("NFS: Set an attribute barrier on all updates") Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index ff737be559dc..8de5b3b9da91 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1662,10 +1662,10 @@ EXPORT_SYMBOL_GPL(_nfs_display_fhandle); */ static int nfs_inode_attrs_need_update(const struct inode *inode, const struct nfs_fattr *fattr) { - const struct nfs_inode *nfsi = NFS_I(inode); + unsigned long attr_gencount = NFS_I(inode)->attr_gencount; - return ((long)fattr->gencount - (long)nfsi->attr_gencount) > 0 || - ((long)nfsi->attr_gencount - (long)nfs_read_attr_generation_counter() > 0); + return (long)(fattr->gencount - attr_gencount) > 0 || + (long)(attr_gencount - nfs_read_attr_generation_counter()) > 0; } static int nfs_refresh_inode_locked(struct inode *inode, struct nfs_fattr *fattr) @@ -2094,7 +2094,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) nfsi->attrtimeo_timestamp = now; } /* Set the barrier to be more recent than this fattr */ - if ((long)fattr->gencount - (long)nfsi->attr_gencount > 0) + if ((long)(fattr->gencount - nfsi->attr_gencount) > 0) nfsi->attr_gencount = fattr->gencount; } -- cgit v1.2.3 From 8a27c7ccca3b0df83f13c591f58d68b057b67780 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 15:45:09 -0400 Subject: NFS: Fix up inode cache tracing Add missing enum definitions and missing entries for nfs_show_cache_validity(). Signed-off-by: Trond Myklebust --- fs/nfs/nfstrace.h | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs') diff --git a/fs/nfs/nfstrace.h b/fs/nfs/nfstrace.h index 5a59dcdce0b2..cdba6eebe3cb 100644 --- a/fs/nfs/nfstrace.h +++ b/fs/nfs/nfstrace.h @@ -45,6 +45,9 @@ TRACE_DEFINE_ENUM(NFS_INO_INVALID_CTIME); TRACE_DEFINE_ENUM(NFS_INO_INVALID_MTIME); TRACE_DEFINE_ENUM(NFS_INO_INVALID_SIZE); TRACE_DEFINE_ENUM(NFS_INO_INVALID_OTHER); +TRACE_DEFINE_ENUM(NFS_INO_DATA_INVAL_DEFER); +TRACE_DEFINE_ENUM(NFS_INO_INVALID_BLOCKS); +TRACE_DEFINE_ENUM(NFS_INO_INVALID_XATTR); #define nfs_show_cache_validity(v) \ __print_flags(v, "|", \ @@ -60,6 +63,8 @@ TRACE_DEFINE_ENUM(NFS_INO_INVALID_OTHER); { NFS_INO_INVALID_MTIME, "INVALID_MTIME" }, \ { NFS_INO_INVALID_SIZE, "INVALID_SIZE" }, \ { NFS_INO_INVALID_OTHER, "INVALID_OTHER" }, \ + { NFS_INO_DATA_INVAL_DEFER, "DATA_INVAL_DEFER" }, \ + { NFS_INO_INVALID_BLOCKS, "INVALID_BLOCKS" }, \ { NFS_INO_INVALID_XATTR, "INVALID_XATTR" }) TRACE_DEFINE_ENUM(NFS_INO_ADVISE_RDPLUS); -- cgit v1.2.3 From 4eb6a8230b589eb2bb02c4ce7418032552ef6470 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 15:59:53 -0400 Subject: NFS: Mask out unsupported attributes in nfs_getattr() We don't currently support STATX_BTIME, so don't advertise it in the return values for nfs_getattr(). Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 8de5b3b9da91..c429a5375b92 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -815,6 +815,10 @@ int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, trace_nfs_getattr_enter(inode); + request_mask &= STATX_TYPE | STATX_MODE | STATX_NLINK | STATX_UID | + STATX_GID | STATX_ATIME | STATX_MTIME | STATX_CTIME | + STATX_INO | STATX_SIZE | STATX_BLOCKS; + if ((query_flags & AT_STATX_DONT_SYNC) && !force_sync) { nfs_readdirplus_parent_cache_hit(path->dentry); goto out_no_update; -- cgit v1.2.3 From 50c7a7994dd20af56e4d47e90af10bab71b71001 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 20:46:36 -0400 Subject: NFS: NFS_INO_REVAL_PAGECACHE should mark the change attribute invalid When we're looking to revalidate the page cache, we should just ensure that we mark the change attribute invalid. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index c429a5375b92..a9b302b389a7 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -219,7 +219,8 @@ void nfs_set_cache_invalid(struct inode *inode, unsigned long flags) | NFS_INO_INVALID_SIZE | NFS_INO_REVAL_PAGECACHE | NFS_INO_INVALID_XATTR); - } + } else if (flags & NFS_INO_REVAL_PAGECACHE) + flags |= NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE; if (!nfs_has_xattr_cache(nfsi)) flags &= ~NFS_INO_INVALID_XATTR; -- cgit v1.2.3 From 4cdfeb648ac56ef0f961549528ccebd76042f1d3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 24 Mar 2021 12:29:41 -0400 Subject: NFS: Fix up revalidation of space used Ensure that when the change attribute or the size change, we also remember to revalidate the space used. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index a9b302b389a7..8b2f4a5eaa42 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -565,12 +565,13 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st nfs_set_cache_invalid(inode, NFS_INO_INVALID_XATTR); if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED) inode->i_blocks = fattr->du.nfs2.blocks; - if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) { + else if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) { /* * report the blocks in 512byte units */ inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used); - } + } else if (fattr->size != 0) + nfs_set_cache_invalid(inode, NFS_INO_INVALID_BLOCKS); if (nfsi->cache_validity != 0) nfsi->cache_validity |= NFS_INO_REVAL_FORCED; @@ -711,7 +712,8 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr, spin_lock(&inode->i_lock); NFS_I(inode)->attr_gencount = fattr->gencount; if ((attr->ia_valid & ATTR_SIZE) != 0) { - nfs_set_cache_invalid(inode, NFS_INO_INVALID_MTIME); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_MTIME | + NFS_INO_INVALID_BLOCKS); nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC); nfs_vmtruncate(inode, attr->ia_size); } @@ -1933,6 +1935,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) save_cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME | NFS_INO_INVALID_SIZE + | NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_OTHER; if (S_ISDIR(inode->i_mode)) nfs_force_lookup_revalidate(inode); @@ -1990,6 +1993,12 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) (long long)cur_isize, (long long)new_isize); } + if (new_isize == 0 && + !(fattr->valid & (NFS_ATTR_FATTR_SPACE_USED | + NFS_ATTR_FATTR_BLOCKS_USED))) { + fattr->du.nfs3.used = 0; + fattr->valid |= NFS_ATTR_FATTR_SPACE_USED; + } } else { nfsi->cache_validity |= save_cache_validity & (NFS_INO_INVALID_SIZE -- cgit v1.2.3 From e8764a6f96d3ca17be7466c7b0fddef64f4e3d14 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 24 Mar 2021 13:55:09 -0400 Subject: NFS: Don't revalidate attributes that are not being asked for If the user doesn't set STATX_UID/GID/MODE, then don't care if they are known to be stale. Ditto if we're not being asked for the file size. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 8b2f4a5eaa42..4982bc18ee26 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -857,12 +857,17 @@ int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, /* Check whether the cached attributes are stale */ do_update |= force_sync || nfs_attribute_cache_expired(inode); cache_validity = READ_ONCE(NFS_I(inode)->cache_validity); - do_update |= cache_validity & - (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_LABEL); + do_update |= cache_validity & NFS_INO_INVALID_CHANGE; if (request_mask & STATX_ATIME) do_update |= cache_validity & NFS_INO_INVALID_ATIME; - if (request_mask & (STATX_CTIME|STATX_MTIME)) - do_update |= cache_validity & NFS_INO_REVAL_PAGECACHE; + if (request_mask & STATX_CTIME) + do_update |= cache_validity & NFS_INO_INVALID_CTIME; + if (request_mask & STATX_MTIME) + do_update |= cache_validity & NFS_INO_INVALID_MTIME; + if (request_mask & STATX_SIZE) + do_update |= cache_validity & NFS_INO_INVALID_SIZE; + if (request_mask & (STATX_UID | STATX_GID | STATX_MODE | STATX_NLINK)) + do_update |= cache_validity & NFS_INO_INVALID_OTHER; if (request_mask & STATX_BLOCKS) do_update |= cache_validity & NFS_INO_INVALID_BLOCKS; if (do_update) { -- cgit v1.2.3 From 63cdd7edfd2871e63f4ca001ff6b8e1e166a74ae Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 24 Mar 2021 14:29:12 -0400 Subject: NFS: Fix up statx() results If statx has valid attributes available that weren't asked for, then return them and set the result mask appropriately. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 29 ++++++++++++++++++++++++++--- 1 file changed, 26 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 4982bc18ee26..8c2d5f333e81 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -806,6 +806,28 @@ static bool nfs_need_revalidate_inode(struct inode *inode) return false; } +static u32 nfs_get_valid_attrmask(struct inode *inode) +{ + unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity); + u32 reply_mask = STATX_INO | STATX_TYPE; + + if (!(cache_validity & NFS_INO_INVALID_ATIME)) + reply_mask |= STATX_ATIME; + if (!(cache_validity & NFS_INO_REVAL_PAGECACHE)) { + if (!(cache_validity & NFS_INO_INVALID_CTIME)) + reply_mask |= STATX_CTIME; + if (!(cache_validity & NFS_INO_INVALID_MTIME)) + reply_mask |= STATX_MTIME; + if (!(cache_validity & NFS_INO_INVALID_SIZE)) + reply_mask |= STATX_SIZE; + } + if (!(cache_validity & NFS_INO_INVALID_OTHER)) + reply_mask |= STATX_UID | STATX_GID | STATX_MODE | STATX_NLINK; + if (!(cache_validity & NFS_INO_INVALID_BLOCKS)) + reply_mask |= STATX_BLOCKS; + return reply_mask; +} + int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) { @@ -824,7 +846,7 @@ int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, if ((query_flags & AT_STATX_DONT_SYNC) && !force_sync) { nfs_readdirplus_parent_cache_hit(path->dentry); - goto out_no_update; + goto out_no_revalidate; } /* Flush out writes to the server in order to update c/mtime. */ @@ -870,6 +892,7 @@ int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, do_update |= cache_validity & NFS_INO_INVALID_OTHER; if (request_mask & STATX_BLOCKS) do_update |= cache_validity & NFS_INO_INVALID_BLOCKS; + if (do_update) { /* Update the attribute cache */ if (!(server->flags & NFS_MOUNT_NOAC)) @@ -883,8 +906,8 @@ int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, nfs_readdirplus_parent_cache_hit(path->dentry); out_no_revalidate: /* Only return attributes that were revalidated. */ - stat->result_mask &= request_mask; -out_no_update: + stat->result_mask = nfs_get_valid_attrmask(inode) | request_mask; + generic_fillattr(&init_user_ns, inode, stat); stat->ino = nfs_compat_user_ino64(NFS_FILEID(inode)); if (S_ISDIR(inode->i_mode)) -- cgit v1.2.3 From 1f9f4328155a6944903e2364f38bb6ed1e1ea9e9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 12 Apr 2021 20:08:00 -0400 Subject: NFS: nfs_setattr_update_inode() should clear the suid/sgid bits When we do a 'chown' or 'chgrp', the server will clear the suid/sgid bits. Ensure that we mirror that in nfs_setattr_update_inode(). Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 8c2d5f333e81..d34da63202cc 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -636,8 +636,7 @@ nfs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry, } /* Optimization: if the end result is no change, don't RPC */ - attr->ia_valid &= NFS_VALID_ATTRS; - if ((attr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0) + if (((attr->ia_valid & NFS_VALID_ATTRS) & ~(ATTR_FILE|ATTR_OPEN)) == 0) return 0; trace_nfs_setattr_enter(inode); @@ -719,6 +718,13 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr, } if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) { NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_CTIME; + if ((attr->ia_valid & ATTR_KILL_SUID) != 0 && + inode->i_mode & S_ISUID) + inode->i_mode &= ~S_ISUID; + if ((attr->ia_valid & ATTR_KILL_SGID) != 0 && + (inode->i_mode & (S_ISGID | S_IXGRP)) == + (S_ISGID | S_IXGRP)) + inode->i_mode &= ~S_ISGID; if ((attr->ia_valid & ATTR_MODE) != 0) { int mode = attr->ia_mode & S_IALLUGO; mode |= inode->i_mode & ~S_IALLUGO; -- cgit v1.2.3 From 1f3208b2d6975f31b9c7c6bf174b84fe9c97492f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 11:04:34 -0400 Subject: NFS: Add a cache validity flag argument to nfs_revalidate_inode() Add an argument to nfs_revalidate_inode() to allow callers to specify which attributes they need to check for validity. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 2 +- fs/nfs/export.c | 6 +----- fs/nfs/inode.c | 25 +++++++------------------ fs/nfs/nfs3acl.c | 2 +- fs/nfs/nfs4proc.c | 6 +++--- 5 files changed, 13 insertions(+), 28 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 0cd7c59a6601..e924d65c125e 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -3006,7 +3006,7 @@ out_notsup: if (mask & MAY_NOT_BLOCK) return -ECHILD; - res = nfs_revalidate_inode(NFS_SERVER(inode), inode); + res = nfs_revalidate_inode(inode, NFS_INO_INVALID_OTHER); if (res == 0) res = generic_permission(&init_user_ns, inode, mask); goto out; diff --git a/fs/nfs/export.c b/fs/nfs/export.c index b347e3ce0cc8..37a1a88df771 100644 --- a/fs/nfs/export.c +++ b/fs/nfs/export.c @@ -169,11 +169,7 @@ out: static u64 nfs_fetch_iversion(struct inode *inode) { - struct nfs_server *server = NFS_SERVER(inode); - - if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_CHANGE | - NFS_INO_REVAL_PAGECACHE)) - __nfs_revalidate_inode(server, inode); + nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE); return inode_peek_iversion_raw(inode); } diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index d34da63202cc..b9aac408f03a 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -802,16 +802,6 @@ static void nfs_readdirplus_parent_cache_hit(struct dentry *dentry) dput(parent); } -static bool nfs_need_revalidate_inode(struct inode *inode) -{ - if (NFS_I(inode)->cache_validity & - (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_LABEL)) - return true; - if (nfs_attribute_cache_expired(inode)) - return true; - return false; -} - static u32 nfs_get_valid_attrmask(struct inode *inode) { unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity); @@ -1004,7 +994,6 @@ void nfs_close_context(struct nfs_open_context *ctx, int is_sync) { struct nfs_inode *nfsi; struct inode *inode; - struct nfs_server *server; if (!(ctx->mode & FMODE_WRITE)) return; @@ -1020,10 +1009,10 @@ void nfs_close_context(struct nfs_open_context *ctx, int is_sync) return; if (!list_empty(&nfsi->open_files)) return; - server = NFS_SERVER(inode); - if (server->flags & NFS_MOUNT_NOCTO) + if (NFS_SERVER(inode)->flags & NFS_MOUNT_NOCTO) return; - nfs_revalidate_inode(server, inode); + nfs_revalidate_inode(inode, + NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE); } EXPORT_SYMBOL_GPL(nfs_close_context); @@ -1278,16 +1267,16 @@ int nfs_attribute_cache_expired(struct inode *inode) /** * nfs_revalidate_inode - Revalidate the inode attributes - * @server: pointer to nfs_server struct * @inode: pointer to inode struct + * @flags: cache flags to check * * Updates inode attribute information by retrieving the data from the server. */ -int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode) +int nfs_revalidate_inode(struct inode *inode, unsigned long flags) { - if (!nfs_need_revalidate_inode(inode)) + if (!nfs_check_cache_invalid(inode, flags)) return NFS_STALE(inode) ? -ESTALE : 0; - return __nfs_revalidate_inode(server, inode); + return __nfs_revalidate_inode(NFS_SERVER(inode), inode); } EXPORT_SYMBOL_GPL(nfs_revalidate_inode); diff --git a/fs/nfs/nfs3acl.c b/fs/nfs/nfs3acl.c index bb386a691e69..9ec560aa4a50 100644 --- a/fs/nfs/nfs3acl.c +++ b/fs/nfs/nfs3acl.c @@ -65,7 +65,7 @@ struct posix_acl *nfs3_get_acl(struct inode *inode, int type) if (!nfs_server_capable(inode, NFS_CAP_ACLS)) return ERR_PTR(-EOPNOTSUPP); - status = nfs_revalidate_inode(server, inode); + status = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE); if (status < 0) return ERR_PTR(status); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 1cf98c40e3b2..6b990fe5bc1f 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5868,7 +5868,7 @@ static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen) if (!nfs4_server_supports_acls(server)) return -EOPNOTSUPP; - ret = nfs_revalidate_inode(server, inode); + ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE); if (ret < 0) return ret; if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL) @@ -7619,7 +7619,7 @@ static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler, return -EACCES; } - ret = nfs_revalidate_inode(NFS_SERVER(inode), inode); + ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE); if (ret) return ret; @@ -7650,7 +7650,7 @@ nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len) return 0; } - ret = nfs_revalidate_inode(NFS_SERVER(inode), inode); + ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE); if (ret) return ret; -- cgit v1.2.3 From 13c0b082b6a90b1b87b5fb100983d05bcc75d9b6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 21:07:21 -0400 Subject: NFS: Replace use of NFS_INO_REVAL_PAGECACHE when checking cache validity When checking cache validity, be more specific than just 'we want to check the page cache validity'. In almost all cases, we want to check that change attribute, and possibly also the size. Signed-off-by: Trond Myklebust --- fs/nfs/file.c | 2 +- fs/nfs/inode.c | 41 ++++++++++++----------------------------- fs/nfs/nfs4proc.c | 5 ++--- fs/nfs/write.c | 2 +- 4 files changed, 16 insertions(+), 34 deletions(-) (limited to 'fs') diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 16ad5050e046..1fef107961bc 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -105,7 +105,7 @@ static int nfs_revalidate_file_size(struct inode *inode, struct file *filp) if (filp->f_flags & O_DIRECT) goto force_reval; - if (nfs_check_cache_invalid(inode, NFS_INO_REVAL_PAGECACHE)) + if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_SIZE)) goto force_reval; return 0; force_reval: diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index b9aac408f03a..aec402d048eb 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -164,34 +164,19 @@ static int nfs_attribute_timeout(struct inode *inode) return !time_in_range_open(jiffies, nfsi->read_cache_jiffies, nfsi->read_cache_jiffies + nfsi->attrtimeo); } -static bool nfs_check_cache_invalid_delegated(struct inode *inode, unsigned long flags) +static bool nfs_check_cache_flags_invalid(struct inode *inode, + unsigned long flags) { unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity); - /* Special case for the pagecache or access cache */ - if (flags == NFS_INO_REVAL_PAGECACHE && - !(cache_validity & NFS_INO_REVAL_FORCED)) - return false; return (cache_validity & flags) != 0; } -static bool nfs_check_cache_invalid_not_delegated(struct inode *inode, unsigned long flags) -{ - unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity); - - if ((cache_validity & flags) != 0) - return true; - if (nfs_attribute_timeout(inode)) - return true; - return false; -} - bool nfs_check_cache_invalid(struct inode *inode, unsigned long flags) { - if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) - return nfs_check_cache_invalid_delegated(inode, flags); - - return nfs_check_cache_invalid_not_delegated(inode, flags); + if (nfs_check_cache_flags_invalid(inode, flags)) + return true; + return nfs_attribute_cache_expired(inode); } EXPORT_SYMBOL_GPL(nfs_check_cache_invalid); @@ -809,14 +794,12 @@ static u32 nfs_get_valid_attrmask(struct inode *inode) if (!(cache_validity & NFS_INO_INVALID_ATIME)) reply_mask |= STATX_ATIME; - if (!(cache_validity & NFS_INO_REVAL_PAGECACHE)) { - if (!(cache_validity & NFS_INO_INVALID_CTIME)) - reply_mask |= STATX_CTIME; - if (!(cache_validity & NFS_INO_INVALID_MTIME)) - reply_mask |= STATX_MTIME; - if (!(cache_validity & NFS_INO_INVALID_SIZE)) - reply_mask |= STATX_SIZE; - } + if (!(cache_validity & NFS_INO_INVALID_CTIME)) + reply_mask |= STATX_CTIME; + if (!(cache_validity & NFS_INO_INVALID_MTIME)) + reply_mask |= STATX_MTIME; + if (!(cache_validity & NFS_INO_INVALID_SIZE)) + reply_mask |= STATX_SIZE; if (!(cache_validity & NFS_INO_INVALID_OTHER)) reply_mask |= STATX_UID | STATX_GID | STATX_MODE | STATX_NLINK; if (!(cache_validity & NFS_INO_INVALID_BLOCKS)) @@ -1362,7 +1345,7 @@ out: bool nfs_mapping_need_revalidate_inode(struct inode *inode) { - return nfs_check_cache_invalid(inode, NFS_INO_REVAL_PAGECACHE) || + return nfs_check_cache_invalid(inode, NFS_INO_INVALID_CHANGE) || NFS_STALE(inode); } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 6b990fe5bc1f..a21311520404 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5429,7 +5429,7 @@ static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src, memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ); - if (cache_validity & (NFS_INO_INVALID_CHANGE | NFS_INO_REVAL_PAGECACHE)) + if (cache_validity & NFS_INO_INVALID_CHANGE) bitmask[0] |= FATTR4_WORD0_CHANGE; if (cache_validity & NFS_INO_INVALID_ATIME) bitmask[1] |= FATTR4_WORD1_TIME_ACCESS; @@ -5449,8 +5449,7 @@ static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src, if (nfs4_have_delegation(inode, FMODE_READ) && !(cache_validity & NFS_INO_REVAL_FORCED)) bitmask[0] &= ~FATTR4_WORD0_SIZE; - else if (cache_validity & - (NFS_INO_INVALID_SIZE | NFS_INO_REVAL_PAGECACHE)) + else if (cache_validity & NFS_INO_INVALID_SIZE) bitmask[0] |= FATTR4_WORD0_SIZE; for (i = 0; i < NFS4_BITMASK_SZ; i++) diff --git a/fs/nfs/write.c b/fs/nfs/write.c index f05a90338a76..7a39b3d424da 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1293,7 +1293,7 @@ static bool nfs_write_pageuptodate(struct page *page, struct inode *inode, if (nfs_have_delegated_attributes(inode)) goto out; if (nfsi->cache_validity & - (NFS_INO_REVAL_PAGECACHE | NFS_INO_INVALID_SIZE)) + (NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE)) return false; smp_rmb(); if (test_bit(NFS_INO_INVALIDATING, &nfsi->flags) && pagelen != 0) -- cgit v1.2.3 From 36a9346c225270262d9f34e66c91aa1723fa903f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 21:20:32 -0400 Subject: NFS: Don't set NFS_INO_REVAL_PAGECACHE in the inode cache validity It is no longer necessary to preserve the NFS_INO_REVAL_PAGECACHE flag. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 6 ++---- fs/nfs/nfs4proc.c | 1 - 2 files changed, 2 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index aec402d048eb..3d18e66a4b8f 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -202,11 +202,12 @@ void nfs_set_cache_invalid(struct inode *inode, unsigned long flags) flags &= ~NFS_INO_INVALID_OTHER; flags &= ~(NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE - | NFS_INO_REVAL_PAGECACHE | NFS_INO_INVALID_XATTR); } else if (flags & NFS_INO_REVAL_PAGECACHE) flags |= NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE; + flags &= ~NFS_INO_REVAL_PAGECACHE; + if (!nfs_has_xattr_cache(nfsi)) flags &= ~NFS_INO_INVALID_XATTR; if (flags & NFS_INO_INVALID_DATA) @@ -1917,7 +1918,6 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ATIME | NFS_INO_REVAL_FORCED - | NFS_INO_REVAL_PAGECACHE | NFS_INO_INVALID_BLOCKS); /* Do atomic weak cache consistency updates */ @@ -1956,7 +1956,6 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) } else { nfsi->cache_validity |= save_cache_validity & (NFS_INO_INVALID_CHANGE - | NFS_INO_REVAL_PAGECACHE | NFS_INO_REVAL_FORCED); cache_revalidated = false; } @@ -2008,7 +2007,6 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) } else { nfsi->cache_validity |= save_cache_validity & (NFS_INO_INVALID_SIZE - | NFS_INO_REVAL_PAGECACHE | NFS_INO_REVAL_FORCED); cache_revalidated = false; } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a21311520404..a490f1373765 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1191,7 +1191,6 @@ nfs4_update_changeattr_locked(struct inode *inode, cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME; if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(inode)) { - nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE; nfsi->attrtimeo_timestamp = jiffies; } else { if (S_ISDIR(inode->i_mode)) { -- cgit v1.2.3 From a71029b86752e8d40301af235a6bbf4896cc1402 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 10 Apr 2021 00:23:03 -0400 Subject: NFSv4: Fix nfs4_bitmap_copy_adjust() Don't remove flags from the set retrieved from the cache_validity. We do want to retrieve all attributes that are listed as being invalid, whether or not there is a delegation set. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 33 ++++++++++++++++----------------- 1 file changed, 16 insertions(+), 17 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a490f1373765..8dc595ce40ca 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -284,7 +284,7 @@ const u32 nfs4_fs_locations_bitmap[3] = { }; static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src, - struct inode *inode) + struct inode *inode, unsigned long flags) { unsigned long cache_validity; @@ -292,22 +292,19 @@ static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src, if (!inode || !nfs4_have_delegation(inode, FMODE_READ)) return; - cache_validity = READ_ONCE(NFS_I(inode)->cache_validity); - if (!(cache_validity & NFS_INO_REVAL_FORCED)) - cache_validity &= ~(NFS_INO_INVALID_CHANGE - | NFS_INO_INVALID_SIZE); + cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags; + /* Remove the attributes over which we have full control */ + dst[1] &= ~FATTR4_WORD1_RAWDEV; if (!(cache_validity & NFS_INO_INVALID_SIZE)) dst[0] &= ~FATTR4_WORD0_SIZE; if (!(cache_validity & NFS_INO_INVALID_CHANGE)) dst[0] &= ~FATTR4_WORD0_CHANGE; -} -static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst, - const __u32 *src, struct inode *inode) -{ - nfs4_bitmap_copy_adjust(dst, src, inode); + if (!(cache_validity & NFS_INO_INVALID_OTHER)) + dst[1] &= ~(FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER | + FATTR4_WORD1_OWNER_GROUP); } static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry, @@ -3344,12 +3341,15 @@ static int nfs4_do_setattr(struct inode *inode, const struct cred *cred, .inode = inode, .stateid = &arg.stateid, }; + unsigned long adjust_flags = NFS_INO_INVALID_CHANGE; int err; + if (sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) + adjust_flags |= NFS_INO_INVALID_OTHER; + do { - nfs4_bitmap_copy_adjust_setattr(bitmask, - nfs4_bitmask(server, olabel), - inode); + nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, olabel), + inode, adjust_flags); err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx); switch (err) { @@ -4157,8 +4157,7 @@ static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, if (inode && (server->flags & NFS_MOUNT_SOFTREVAL)) task_flags |= RPC_TASK_TIMEOUT; - nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode); - + nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode, 0); nfs_fattr_init(fattr); nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0); return nfs4_do_call_sync(server->client, server, &msg, @@ -4764,8 +4763,8 @@ static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct } nfs4_inode_make_writeable(inode); - nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode); - + nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.label), inode, + NFS_INO_INVALID_CHANGE); status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1); if (!status) { nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start, -- cgit v1.2.3 From fabf2b341502e894001d70f91309dd6f3785e2dc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 13:14:42 -0400 Subject: NFS: Separate tracking of file nlinks cache validity from the mode/uid/gid Rename can cause us to revalidate the access cache, so lets track the nlinks separately from the mode/uid/gid. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 2 +- fs/nfs/inode.c | 15 ++++++++++----- fs/nfs/nfs4proc.c | 13 +++++++------ fs/nfs/nfstrace.h | 4 +++- 4 files changed, 21 insertions(+), 13 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index e924d65c125e..f748d2294261 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1711,7 +1711,7 @@ static void nfs_drop_nlink(struct inode *inode) NFS_I(inode)->attr_gencount = nfs_inc_attr_generation_counter(); nfs_set_cache_invalid( inode, NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_CTIME | - NFS_INO_INVALID_OTHER | NFS_INO_REVAL_FORCED); + NFS_INO_INVALID_NLINK | NFS_INO_REVAL_FORCED); spin_unlock(&inode->i_lock); } diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 3d18e66a4b8f..7bf9138330f2 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -538,7 +538,7 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st if (fattr->valid & NFS_ATTR_FATTR_NLINK) set_nlink(inode, fattr->nlink); else if (nfs_server_capable(inode, NFS_CAP_NLINK)) - nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_NLINK); if (fattr->valid & NFS_ATTR_FATTR_OWNER) inode->i_uid = fattr->uid; else if (nfs_server_capable(inode, NFS_CAP_OWNER)) @@ -801,8 +801,10 @@ static u32 nfs_get_valid_attrmask(struct inode *inode) reply_mask |= STATX_MTIME; if (!(cache_validity & NFS_INO_INVALID_SIZE)) reply_mask |= STATX_SIZE; + if (!(cache_validity & NFS_INO_INVALID_NLINK)) + reply_mask |= STATX_NLINK; if (!(cache_validity & NFS_INO_INVALID_OTHER)) - reply_mask |= STATX_UID | STATX_GID | STATX_MODE | STATX_NLINK; + reply_mask |= STATX_UID | STATX_GID | STATX_MODE; if (!(cache_validity & NFS_INO_INVALID_BLOCKS)) reply_mask |= STATX_BLOCKS; return reply_mask; @@ -868,7 +870,9 @@ int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, do_update |= cache_validity & NFS_INO_INVALID_MTIME; if (request_mask & STATX_SIZE) do_update |= cache_validity & NFS_INO_INVALID_SIZE; - if (request_mask & (STATX_UID | STATX_GID | STATX_MODE | STATX_NLINK)) + if (request_mask & STATX_NLINK) + do_update |= cache_validity & NFS_INO_INVALID_NLINK; + if (request_mask & (STATX_UID | STATX_GID | STATX_MODE)) do_update |= cache_validity & NFS_INO_INVALID_OTHER; if (request_mask & STATX_BLOCKS) do_update |= cache_validity & NFS_INO_INVALID_BLOCKS; @@ -1518,7 +1522,7 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat /* Has the link count changed? */ if ((fattr->valid & NFS_ATTR_FATTR_NLINK) && inode->i_nlink != fattr->nlink) - invalid |= NFS_INO_INVALID_OTHER; + invalid |= NFS_INO_INVALID_NLINK; ts = inode->i_atime; if ((fattr->valid & NFS_ATTR_FATTR_ATIME) && !timespec64_equal(&ts, &fattr->atime)) @@ -1942,6 +1946,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) | NFS_INO_INVALID_MTIME | NFS_INO_INVALID_SIZE | NFS_INO_INVALID_BLOCKS + | NFS_INO_INVALID_NLINK | NFS_INO_INVALID_OTHER; if (S_ISDIR(inode->i_mode)) nfs_force_lookup_revalidate(inode); @@ -2074,7 +2079,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) } } else if (server->caps & NFS_CAP_NLINK) { nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_OTHER + (NFS_INO_INVALID_NLINK | NFS_INO_REVAL_FORCED); cache_revalidated = false; } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 8dc595ce40ca..a74c1c3c4192 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1167,14 +1167,14 @@ int nfs4_call_sync(struct rpc_clnt *clnt, static void nfs4_inc_nlink_locked(struct inode *inode) { - nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_NLINK); inc_nlink(inode); } static void nfs4_dec_nlink_locked(struct inode *inode) { - nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_NLINK); drop_nlink(inode); } @@ -4717,11 +4717,11 @@ static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir, /* Note: If we moved a directory, nlink will change */ nfs4_update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start, - NFS_INO_INVALID_OTHER | + NFS_INO_INVALID_NLINK | NFS_INO_INVALID_DATA); nfs4_update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start, - NFS_INO_INVALID_OTHER | + NFS_INO_INVALID_NLINK | NFS_INO_INVALID_DATA); } else nfs4_update_changeattr(old_dir, &res->old_cinfo, @@ -5433,8 +5433,9 @@ static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src, bitmask[1] |= FATTR4_WORD1_TIME_ACCESS; if (cache_validity & NFS_INO_INVALID_OTHER) bitmask[1] |= FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER | - FATTR4_WORD1_OWNER_GROUP | - FATTR4_WORD1_NUMLINKS; + FATTR4_WORD1_OWNER_GROUP; + if (cache_validity & NFS_INO_INVALID_NLINK) + bitmask[1] |= FATTR4_WORD1_NUMLINKS; if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL) bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL; if (cache_validity & NFS_INO_INVALID_CTIME) diff --git a/fs/nfs/nfstrace.h b/fs/nfs/nfstrace.h index cdba6eebe3cb..a0ebc53160dd 100644 --- a/fs/nfs/nfstrace.h +++ b/fs/nfs/nfstrace.h @@ -48,6 +48,7 @@ TRACE_DEFINE_ENUM(NFS_INO_INVALID_OTHER); TRACE_DEFINE_ENUM(NFS_INO_DATA_INVAL_DEFER); TRACE_DEFINE_ENUM(NFS_INO_INVALID_BLOCKS); TRACE_DEFINE_ENUM(NFS_INO_INVALID_XATTR); +TRACE_DEFINE_ENUM(NFS_INO_INVALID_NLINK); #define nfs_show_cache_validity(v) \ __print_flags(v, "|", \ @@ -65,7 +66,8 @@ TRACE_DEFINE_ENUM(NFS_INO_INVALID_XATTR); { NFS_INO_INVALID_OTHER, "INVALID_OTHER" }, \ { NFS_INO_DATA_INVAL_DEFER, "DATA_INVAL_DEFER" }, \ { NFS_INO_INVALID_BLOCKS, "INVALID_BLOCKS" }, \ - { NFS_INO_INVALID_XATTR, "INVALID_XATTR" }) + { NFS_INO_INVALID_XATTR, "INVALID_XATTR" }, \ + { NFS_INO_INVALID_NLINK, "INVALID_NLINK" }) TRACE_DEFINE_ENUM(NFS_INO_ADVISE_RDPLUS); TRACE_DEFINE_ENUM(NFS_INO_STALE); -- cgit v1.2.3 From 720869eb19f3161980d6d4631d3df7e8c5355993 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 13 Apr 2021 09:41:16 -0400 Subject: NFS: Separate tracking of file mode cache validity from the uid/gid chown()/chgrp() and chmod() are separate operations, and in addition, there are mode operations that are performed automatically by the server. So let's track mode validity separately from the file ownership validity. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 5 +++-- fs/nfs/inode.c | 18 ++++++++++++------ fs/nfs/nfs4proc.c | 14 +++++++++----- fs/nfs/nfstrace.h | 4 +++- fs/nfs/write.c | 2 +- 5 files changed, 28 insertions(+), 15 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index f748d2294261..d2835d211a73 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -2948,7 +2948,7 @@ static int nfs_execute_ok(struct inode *inode, int mask) if (S_ISDIR(inode->i_mode)) return 0; - if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_OTHER)) { + if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_MODE)) { if (mask & MAY_NOT_BLOCK) return -ECHILD; ret = __nfs_revalidate_inode(server, inode); @@ -3006,7 +3006,8 @@ out_notsup: if (mask & MAY_NOT_BLOCK) return -ECHILD; - res = nfs_revalidate_inode(inode, NFS_INO_INVALID_OTHER); + res = nfs_revalidate_inode(inode, NFS_INO_INVALID_MODE | + NFS_INO_INVALID_OTHER); if (res == 0) res = generic_permission(&init_user_ns, inode, mask); goto out; diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 7bf9138330f2..81e3e140e923 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -199,7 +199,8 @@ void nfs_set_cache_invalid(struct inode *inode, unsigned long flags) if (have_delegation) { if (!(flags & NFS_INO_REVAL_FORCED)) - flags &= ~NFS_INO_INVALID_OTHER; + flags &= ~(NFS_INO_INVALID_MODE | + NFS_INO_INVALID_OTHER); flags &= ~(NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE | NFS_INO_INVALID_XATTR); @@ -472,7 +473,7 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st nfsi->cache_validity = 0; if ((fattr->valid & NFS_ATTR_FATTR_MODE) == 0 && nfs_server_capable(inode, NFS_CAP_MODE)) - nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_MODE); /* Why so? Because we want revalidate for devices/FIFOs, and * that's precisely what we have in nfs_file_inode_operations. */ @@ -803,8 +804,10 @@ static u32 nfs_get_valid_attrmask(struct inode *inode) reply_mask |= STATX_SIZE; if (!(cache_validity & NFS_INO_INVALID_NLINK)) reply_mask |= STATX_NLINK; + if (!(cache_validity & NFS_INO_INVALID_MODE)) + reply_mask |= STATX_MODE; if (!(cache_validity & NFS_INO_INVALID_OTHER)) - reply_mask |= STATX_UID | STATX_GID | STATX_MODE; + reply_mask |= STATX_UID | STATX_GID; if (!(cache_validity & NFS_INO_INVALID_BLOCKS)) reply_mask |= STATX_BLOCKS; return reply_mask; @@ -872,7 +875,9 @@ int nfs_getattr(struct user_namespace *mnt_userns, const struct path *path, do_update |= cache_validity & NFS_INO_INVALID_SIZE; if (request_mask & STATX_NLINK) do_update |= cache_validity & NFS_INO_INVALID_NLINK; - if (request_mask & (STATX_UID | STATX_GID | STATX_MODE)) + if (request_mask & STATX_MODE) + do_update |= cache_validity & NFS_INO_INVALID_MODE; + if (request_mask & (STATX_UID | STATX_GID)) do_update |= cache_validity & NFS_INO_INVALID_OTHER; if (request_mask & STATX_BLOCKS) do_update |= cache_validity & NFS_INO_INVALID_BLOCKS; @@ -1510,7 +1515,7 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat if ((fattr->valid & NFS_ATTR_FATTR_MODE) && (inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)) invalid |= NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL - | NFS_INO_INVALID_OTHER; + | NFS_INO_INVALID_MODE; if ((fattr->valid & NFS_ATTR_FATTR_OWNER) && !uid_eq(inode->i_uid, fattr->uid)) invalid |= NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL @@ -1947,6 +1952,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) | NFS_INO_INVALID_SIZE | NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK + | NFS_INO_INVALID_MODE | NFS_INO_INVALID_OTHER; if (S_ISDIR(inode->i_mode)) nfs_force_lookup_revalidate(inode); @@ -2037,7 +2043,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) } } else if (server->caps & NFS_CAP_MODE) { nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_OTHER + (NFS_INO_INVALID_MODE | NFS_INO_REVAL_FORCED); cache_revalidated = false; } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a74c1c3c4192..bc90f2a12d5d 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -302,9 +302,10 @@ static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src, if (!(cache_validity & NFS_INO_INVALID_CHANGE)) dst[0] &= ~FATTR4_WORD0_CHANGE; + if (!(cache_validity & NFS_INO_INVALID_MODE)) + dst[1] &= ~FATTR4_WORD1_MODE; if (!(cache_validity & NFS_INO_INVALID_OTHER)) - dst[1] &= ~(FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER | - FATTR4_WORD1_OWNER_GROUP); + dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP); } static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry, @@ -3344,7 +3345,9 @@ static int nfs4_do_setattr(struct inode *inode, const struct cred *cred, unsigned long adjust_flags = NFS_INO_INVALID_CHANGE; int err; - if (sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) + if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID)) + adjust_flags |= NFS_INO_INVALID_MODE; + if (sattr->ia_valid & (ATTR_UID | ATTR_GID)) adjust_flags |= NFS_INO_INVALID_OTHER; do { @@ -5431,9 +5434,10 @@ static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src, bitmask[0] |= FATTR4_WORD0_CHANGE; if (cache_validity & NFS_INO_INVALID_ATIME) bitmask[1] |= FATTR4_WORD1_TIME_ACCESS; + if (cache_validity & NFS_INO_INVALID_MODE) + bitmask[1] |= FATTR4_WORD1_MODE; if (cache_validity & NFS_INO_INVALID_OTHER) - bitmask[1] |= FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER | - FATTR4_WORD1_OWNER_GROUP; + bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP; if (cache_validity & NFS_INO_INVALID_NLINK) bitmask[1] |= FATTR4_WORD1_NUMLINKS; if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL) diff --git a/fs/nfs/nfstrace.h b/fs/nfs/nfstrace.h index a0ebc53160dd..41a161cd31f6 100644 --- a/fs/nfs/nfstrace.h +++ b/fs/nfs/nfstrace.h @@ -49,6 +49,7 @@ TRACE_DEFINE_ENUM(NFS_INO_DATA_INVAL_DEFER); TRACE_DEFINE_ENUM(NFS_INO_INVALID_BLOCKS); TRACE_DEFINE_ENUM(NFS_INO_INVALID_XATTR); TRACE_DEFINE_ENUM(NFS_INO_INVALID_NLINK); +TRACE_DEFINE_ENUM(NFS_INO_INVALID_MODE); #define nfs_show_cache_validity(v) \ __print_flags(v, "|", \ @@ -67,7 +68,8 @@ TRACE_DEFINE_ENUM(NFS_INO_INVALID_NLINK); { NFS_INO_DATA_INVAL_DEFER, "DATA_INVAL_DEFER" }, \ { NFS_INO_INVALID_BLOCKS, "INVALID_BLOCKS" }, \ { NFS_INO_INVALID_XATTR, "INVALID_XATTR" }, \ - { NFS_INO_INVALID_NLINK, "INVALID_NLINK" }) + { NFS_INO_INVALID_NLINK, "INVALID_NLINK" }, \ + { NFS_INO_INVALID_MODE, "INVALID_MODE" }) TRACE_DEFINE_ENUM(NFS_INO_ADVISE_RDPLUS); TRACE_DEFINE_ENUM(NFS_INO_STALE); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 7a39b3d424da..61d1174935b6 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1604,7 +1604,7 @@ static int nfs_writeback_done(struct rpc_task *task, /* Deal with the suid/sgid bit corner case */ if (nfs_should_remove_suid(inode)) { spin_lock(&inode->i_lock); - nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_MODE); spin_unlock(&inode->i_lock); } return 0; -- cgit v1.2.3 From 709fa5769914b377af87962bbe4ff81ffb019b2d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 14:51:36 -0400 Subject: NFS: Fix up handling of outstanding layoutcommit in nfs_update_inode() If there is an outstanding layoutcommit, then the list of attributes whose values are expected to change is not the full set. So let's be explicit about the full list. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 81e3e140e923..18c7277d17a8 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1933,7 +1933,11 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) nfs_wcc_update_inode(inode, fattr); if (pnfs_layoutcommit_outstanding(inode)) { - nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_ATTR; + nfsi->cache_validity |= + save_cache_validity & + (NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_CTIME | + NFS_INO_INVALID_MTIME | NFS_INO_INVALID_SIZE | + NFS_INO_REVAL_FORCED); cache_revalidated = false; } -- cgit v1.2.3 From c88c696c59f5775d8781d632f7d10823563027eb Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 14:53:48 -0400 Subject: NFS: Remove a line of code that has no effect in nfs_update_inode() Commit 0b467264d0db ("NFS: Fix attribute revalidation") changed the way we populate the 'invalid' attribute, and made the line that strips away the NFS_INO_INVALID_ATTR bits redundant. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 18c7277d17a8..e4333b6ab2d4 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -2110,7 +2110,6 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) /* Update attrtimeo value if we're out of the unstable period */ if (attr_changed) { - invalid &= ~NFS_INO_INVALID_ATTR; nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE); nfsi->attrtimeo = NFS_MINATTRTIMEO(inode); nfsi->attrtimeo_timestamp = now; -- cgit v1.2.3 From 04c63498b6d4ea445cd2cb02599467b48687244b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Mar 2021 18:13:36 -0400 Subject: NFS: Simplify cache consistency in nfs_check_inode_attributes() We should not be invalidating the access or acl caches in nfs_check_inode_attributes(), since the point is we're unsure about whether the contents of the struct nfs_fattr are fully up to date. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 18 +++++------------- 1 file changed, 5 insertions(+), 13 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index e4333b6ab2d4..fc8e92794636 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1491,8 +1491,7 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat if (!nfs_file_has_buffered_writers(nfsi)) { /* Verify a few of the more important attributes */ if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 && !inode_eq_iversion_raw(inode, fattr->change_attr)) - invalid |= NFS_INO_INVALID_CHANGE - | NFS_INO_REVAL_PAGECACHE; + invalid |= NFS_INO_INVALID_CHANGE; ts = inode->i_mtime; if ((fattr->valid & NFS_ATTR_FATTR_MTIME) && !timespec64_equal(&ts, &fattr->mtime)) @@ -1506,24 +1505,17 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat cur_size = i_size_read(inode); new_isize = nfs_size_to_loff_t(fattr->size); if (cur_size != new_isize) - invalid |= NFS_INO_INVALID_SIZE - | NFS_INO_REVAL_PAGECACHE; + invalid |= NFS_INO_INVALID_SIZE; } } /* Have any file permissions changed? */ if ((fattr->valid & NFS_ATTR_FATTR_MODE) && (inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)) - invalid |= NFS_INO_INVALID_ACCESS - | NFS_INO_INVALID_ACL - | NFS_INO_INVALID_MODE; + invalid |= NFS_INO_INVALID_MODE; if ((fattr->valid & NFS_ATTR_FATTR_OWNER) && !uid_eq(inode->i_uid, fattr->uid)) - invalid |= NFS_INO_INVALID_ACCESS - | NFS_INO_INVALID_ACL - | NFS_INO_INVALID_OTHER; + invalid |= NFS_INO_INVALID_OTHER; if ((fattr->valid & NFS_ATTR_FATTR_GROUP) && !gid_eq(inode->i_gid, fattr->gid)) - invalid |= NFS_INO_INVALID_ACCESS - | NFS_INO_INVALID_ACL - | NFS_INO_INVALID_OTHER; + invalid |= NFS_INO_INVALID_OTHER; /* Has the link count changed? */ if ((fattr->valid & NFS_ATTR_FATTR_NLINK) && inode->i_nlink != fattr->nlink) -- cgit v1.2.3 From 57a789a1dee39e1625956d7dea4f7637f9160f21 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 26 Mar 2021 10:17:25 -0400 Subject: NFSv4: Fix value of decode_fsinfo_maxsz At least two extra fields have been added to fsinfo since this was last updated. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4xdr.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index ac6b79ee9355..d8a1911dd39e 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -144,7 +144,16 @@ static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req, * layout types will be returned. */ #define decode_fsinfo_maxsz (op_decode_hdr_maxsz + \ - nfs4_fattr_bitmap_maxsz + 4 + 8 + 5) + nfs4_fattr_bitmap_maxsz + 1 + \ + 1 /* lease time */ + \ + 2 /* max filesize */ + \ + 2 /* max read */ + \ + 2 /* max write */ + \ + nfstime4_maxsz /* time delta */ + \ + 5 /* fs layout types */ + \ + 1 /* layout blksize */ + \ + 1 /* clone blksize */ + \ + 1 /* xattr support */) #define encode_renew_maxsz (op_encode_hdr_maxsz + 3) #define decode_renew_maxsz (op_decode_hdr_maxsz) #define encode_setclientid_maxsz \ -- cgit v1.2.3 From 993e2d4bd9efc41f0943de39e9374ffdfde62e87 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 12 Apr 2021 12:15:50 -0400 Subject: NFSv4: Don't modify the change attribute cached in the inode When the client is caching data and a write delegation is held, then the server may send a CB_GETATTR to query the attributes. When this happens, the client is supposed to bump the change attribute value that it returns if it holds cached data. However that process uses a value that is stored in the delegation. We do not want to bump the change attribute held in the inode. Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 61d1174935b6..3bf82178166a 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -764,9 +764,6 @@ static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req) * with invalidate/truncate. */ spin_lock(&mapping->private_lock); - if (!nfs_have_writebacks(inode) && - NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE)) - inode_inc_iversion_raw(inode); if (likely(!PageSwapCache(req->wb_page))) { set_bit(PG_MAPPED, &req->wb_flags); SetPagePrivate(req->wb_page); -- cgit v1.2.3 From 7f08a3359a3c1e39c2a118fbbe583d8c8db14ace Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 26 Mar 2021 09:50:19 -0400 Subject: NFSv4: Add support for the NFSv4.2 "change_attr_type" attribute The change_attr_type allows the server to provide a description of how the change attribute will behave. This again will allow the client to optimise its behaviour w.r.t. attribute revalidation. Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 3 +++ fs/nfs/nfs3xdr.c | 1 + fs/nfs/nfs4proc.c | 1 + fs/nfs/nfs4xdr.c | 32 ++++++++++++++++++++++++++++++++ fs/nfs/proc.c | 1 + 5 files changed, 38 insertions(+) (limited to 'fs') diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 399a8eb15397..2aeb4e52a4f1 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -792,6 +792,7 @@ static void nfs_server_set_fsinfo(struct nfs_server *server, server->maxfilesize = fsinfo->maxfilesize; server->time_delta = fsinfo->time_delta; + server->change_attr_type = fsinfo->change_attr_type; server->clone_blksize = fsinfo->clone_blksize; /* We're airborne Set socket buffersize */ @@ -933,6 +934,8 @@ struct nfs_server *nfs_alloc_server(void) return NULL; } + server->change_attr_type = NFS4_CHANGE_TYPE_IS_UNDEFINED; + ida_init(&server->openowner_id); ida_init(&server->lockowner_id); pnfs_init_server(server); diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index ed1c83738c30..83ad62c81fc7 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -2227,6 +2227,7 @@ static int decode_fsinfo3resok(struct xdr_stream *xdr, /* ignore properties */ result->lease_time = 0; + result->change_attr_type = NFS4_CHANGE_TYPE_IS_TIME_METADATA; return 0; } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index bc90f2a12d5d..6992c88a25e7 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -264,6 +264,7 @@ const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE | FATTR4_WORD1_FS_LAYOUT_TYPES, FATTR4_WORD2_LAYOUT_BLKSIZE | FATTR4_WORD2_CLONE_BLKSIZE + | FATTR4_WORD2_CHANGE_ATTR_TYPE | FATTR4_WORD2_XATTR_SUPPORT }; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index d8a1911dd39e..edac4718dec1 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -153,6 +153,7 @@ static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req, 5 /* fs layout types */ + \ 1 /* layout blksize */ + \ 1 /* clone blksize */ + \ + 1 /* change attr type */ + \ 1 /* xattr support */) #define encode_renew_maxsz (op_encode_hdr_maxsz + 3) #define decode_renew_maxsz (op_decode_hdr_maxsz) @@ -4846,6 +4847,32 @@ static int decode_attr_clone_blksize(struct xdr_stream *xdr, uint32_t *bitmap, return 0; } +static int decode_attr_change_attr_type(struct xdr_stream *xdr, + uint32_t *bitmap, + enum nfs4_change_attr_type *res) +{ + u32 tmp = NFS4_CHANGE_TYPE_IS_UNDEFINED; + + dprintk("%s: bitmap is %x\n", __func__, bitmap[2]); + if (bitmap[2] & FATTR4_WORD2_CHANGE_ATTR_TYPE) { + if (xdr_stream_decode_u32(xdr, &tmp)) + return -EIO; + bitmap[2] &= ~FATTR4_WORD2_CHANGE_ATTR_TYPE; + } + + switch(tmp) { + case NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR: + case NFS4_CHANGE_TYPE_IS_VERSION_COUNTER: + case NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS: + case NFS4_CHANGE_TYPE_IS_TIME_METADATA: + *res = tmp; + break; + default: + *res = NFS4_CHANGE_TYPE_IS_UNDEFINED; + } + return 0; +} + static int decode_fsinfo(struct xdr_stream *xdr, struct nfs_fsinfo *fsinfo) { unsigned int savep; @@ -4894,6 +4921,11 @@ static int decode_fsinfo(struct xdr_stream *xdr, struct nfs_fsinfo *fsinfo) if (status) goto xdr_error; + status = decode_attr_change_attr_type(xdr, bitmap, + &fsinfo->change_attr_type); + if (status) + goto xdr_error; + status = decode_attr_xattrsupport(xdr, bitmap, &fsinfo->xattr_support); if (status) diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index 73ab7c59d3a7..ea19dbf12301 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -91,6 +91,7 @@ nfs_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle, info->dtpref = fsinfo.tsize; info->maxfilesize = 0x7FFFFFFF; info->lease_time = 0; + info->change_attr_type = NFS4_CHANGE_TYPE_IS_TIME_METADATA; return 0; } -- cgit v1.2.3 From 6f9be83d07615e6af8838a1d489080b399f42a08 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 26 Mar 2021 11:01:19 -0400 Subject: NFS: Use information about the change attribute to optimise updates If the NFSv4.2 server supports the 'change_attr_type' attribute, then allow the client to optimise its attribute cache update strategy. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 111 +++++++++++++++++++++++++++++++++++++++++++++--------- fs/nfs/nfs4proc.c | 20 ++++++++-- 2 files changed, 110 insertions(+), 21 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index fc8e92794636..d218d164414f 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1657,25 +1657,20 @@ EXPORT_SYMBOL_GPL(_nfs_display_fhandle); #endif /** - * nfs_inode_attrs_need_update - check if the inode attributes need updating - * @inode: pointer to inode + * nfs_inode_attrs_cmp_generic - compare attributes * @fattr: attributes + * @inode: pointer to inode * * Attempt to divine whether or not an RPC call reply carrying stale * attributes got scheduled after another call carrying updated ones. - * - * To do so, the function first assumes that a more recent ctime means - * that the attributes in fattr are newer, however it also attempt to - * catch the case where ctime either didn't change, or went backwards - * (if someone reset the clock on the server) by looking at whether - * or not this RPC call was started after the inode was last updated. * Note also the check for wraparound of 'attr_gencount' * - * The function returns 'true' if it thinks the attributes in 'fattr' are - * more recent than the ones cached in the inode. - * + * The function returns '1' if it thinks the attributes in @fattr are + * more recent than the ones cached in @inode. Otherwise it returns + * the value '0'. */ -static int nfs_inode_attrs_need_update(const struct inode *inode, const struct nfs_fattr *fattr) +static int nfs_inode_attrs_cmp_generic(const struct nfs_fattr *fattr, + const struct inode *inode) { unsigned long attr_gencount = NFS_I(inode)->attr_gencount; @@ -1683,15 +1678,93 @@ static int nfs_inode_attrs_need_update(const struct inode *inode, const struct n (long)(attr_gencount - nfs_read_attr_generation_counter()) > 0; } -static int nfs_refresh_inode_locked(struct inode *inode, struct nfs_fattr *fattr) +/** + * nfs_inode_attrs_cmp_monotonic - compare attributes + * @fattr: attributes + * @inode: pointer to inode + * + * Attempt to divine whether or not an RPC call reply carrying stale + * attributes got scheduled after another call carrying updated ones. + * + * We assume that the server observes monotonic semantics for + * the change attribute, so a larger value means that the attributes in + * @fattr are more recent, in which case the function returns the + * value '1'. + * A return value of '0' indicates no measurable change + * A return value of '-1' means that the attributes in @inode are + * more recent. + */ +static int nfs_inode_attrs_cmp_monotonic(const struct nfs_fattr *fattr, + const struct inode *inode) { - int ret; + s64 diff = fattr->change_attr - inode_peek_iversion_raw(inode); + if (diff > 0) + return 1; + return diff == 0 ? 0 : -1; +} + +/** + * nfs_inode_attrs_cmp_strict_monotonic - compare attributes + * @fattr: attributes + * @inode: pointer to inode + * + * Attempt to divine whether or not an RPC call reply carrying stale + * attributes got scheduled after another call carrying updated ones. + * + * We assume that the server observes strictly monotonic semantics for + * the change attribute, so a larger value means that the attributes in + * @fattr are more recent, in which case the function returns the + * value '1'. + * A return value of '-1' means that the attributes in @inode are + * more recent or unchanged. + */ +static int nfs_inode_attrs_cmp_strict_monotonic(const struct nfs_fattr *fattr, + const struct inode *inode) +{ + return nfs_inode_attrs_cmp_monotonic(fattr, inode) > 0 ? 1 : -1; +} + +/** + * nfs_inode_attrs_cmp - compare attributes + * @fattr: attributes + * @inode: pointer to inode + * + * This function returns '1' if it thinks the attributes in @fattr are + * more recent than the ones cached in @inode. It returns '-1' if + * the attributes in @inode are more recent than the ones in @fattr, + * and it returns 0 if not sure. + */ +static int nfs_inode_attrs_cmp(const struct nfs_fattr *fattr, + const struct inode *inode) +{ + if (nfs_inode_attrs_cmp_generic(fattr, inode) > 0) + return 1; + switch (NFS_SERVER(inode)->change_attr_type) { + case NFS4_CHANGE_TYPE_IS_UNDEFINED: + break; + case NFS4_CHANGE_TYPE_IS_TIME_METADATA: + if (!(fattr->valid & NFS_ATTR_FATTR_CHANGE)) + break; + return nfs_inode_attrs_cmp_monotonic(fattr, inode); + default: + if (!(fattr->valid & NFS_ATTR_FATTR_CHANGE)) + break; + return nfs_inode_attrs_cmp_strict_monotonic(fattr, inode); + } + return 0; +} + +static int nfs_refresh_inode_locked(struct inode *inode, + struct nfs_fattr *fattr) +{ + int attr_cmp = nfs_inode_attrs_cmp(fattr, inode); + int ret = 0; trace_nfs_refresh_inode_enter(inode); - if (nfs_inode_attrs_need_update(inode, fattr)) + if (attr_cmp > 0) ret = nfs_update_inode(inode, fattr); - else + else if (attr_cmp == 0) ret = nfs_check_inode_attributes(inode, fattr); trace_nfs_refresh_inode_exit(inode, ret); @@ -1776,11 +1849,13 @@ EXPORT_SYMBOL_GPL(nfs_post_op_update_inode); */ int nfs_post_op_update_inode_force_wcc_locked(struct inode *inode, struct nfs_fattr *fattr) { + int attr_cmp = nfs_inode_attrs_cmp(fattr, inode); int status; /* Don't do a WCC update if these attributes are already stale */ - if ((fattr->valid & NFS_ATTR_FATTR) == 0 || - !nfs_inode_attrs_need_update(inode, fattr)) { + if (attr_cmp < 0) + return 0; + if ((fattr->valid & NFS_ATTR_FATTR) == 0 || !attr_cmp) { fattr->valid &= ~(NFS_ATTR_FATTR_PRECHANGE | NFS_ATTR_FATTR_PRESIZE | NFS_ATTR_FATTR_PREMTIME diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 6992c88a25e7..0b0a48d78299 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1186,10 +1186,23 @@ nfs4_update_changeattr_locked(struct inode *inode, unsigned long timestamp, unsigned long cache_validity) { struct nfs_inode *nfsi = NFS_I(inode); + u64 change_attr = inode_peek_iversion_raw(inode); cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME; - if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(inode)) { + switch (NFS_SERVER(inode)->change_attr_type) { + case NFS4_CHANGE_TYPE_IS_UNDEFINED: + break; + case NFS4_CHANGE_TYPE_IS_TIME_METADATA: + if ((s64)(change_attr - cinfo->after) > 0) + goto out; + break; + default: + if ((s64)(change_attr - cinfo->after) >= 0) + goto out; + } + + if (cinfo->atomic && cinfo->before == change_attr) { nfsi->attrtimeo_timestamp = jiffies; } else { if (S_ISDIR(inode->i_mode)) { @@ -1201,7 +1214,7 @@ nfs4_update_changeattr_locked(struct inode *inode, cache_validity |= NFS_INO_REVAL_PAGECACHE; } - if (cinfo->before != inode_peek_iversion_raw(inode)) + if (cinfo->before != change_attr) cache_validity |= NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL | NFS_INO_INVALID_XATTR; @@ -1209,8 +1222,9 @@ nfs4_update_changeattr_locked(struct inode *inode, inode_set_iversion_raw(inode, cinfo->after); nfsi->read_cache_jiffies = timestamp; nfsi->attr_gencount = nfs_inc_attr_generation_counter(); - nfs_set_cache_invalid(inode, cache_validity); nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE; +out: + nfs_set_cache_invalid(inode, cache_validity); } void -- cgit v1.2.3 From 7b24dacf0840056f9c51107eddc5bd7efbd20133 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 9 Apr 2021 09:30:21 -0400 Subject: NFS: Another inode revalidation improvement If we're trying to update the inode because a previous update left the cache in a partially unrevalidated state, then allow the update if the change attrs match. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 30 +++++++++++++++++++++++++++++- 1 file changed, 29 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index d218d164414f..b88e9dc72eec 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1754,6 +1754,34 @@ static int nfs_inode_attrs_cmp(const struct nfs_fattr *fattr, return 0; } +/** + * nfs_inode_finish_partial_attr_update - complete a previous inode update + * @fattr: attributes + * @inode: pointer to inode + * + * Returns '1' if the last attribute update left the inode cached + * attributes in a partially unrevalidated state, and @fattr + * matches the change attribute of that partial update. + * Otherwise returns '0'. + */ +static int nfs_inode_finish_partial_attr_update(const struct nfs_fattr *fattr, + const struct inode *inode) +{ + const unsigned long check_valid = + NFS_INO_INVALID_ATIME | NFS_INO_INVALID_CTIME | + NFS_INO_INVALID_MTIME | NFS_INO_INVALID_SIZE | + NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_OTHER | + NFS_INO_INVALID_NLINK; + unsigned long cache_validity = NFS_I(inode)->cache_validity; + + if (!(cache_validity & NFS_INO_INVALID_CHANGE) && + (cache_validity & check_valid) != 0 && + (fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 && + nfs_inode_attrs_cmp_monotonic(fattr, inode) == 0) + return 1; + return 0; +} + static int nfs_refresh_inode_locked(struct inode *inode, struct nfs_fattr *fattr) { @@ -1762,7 +1790,7 @@ static int nfs_refresh_inode_locked(struct inode *inode, trace_nfs_refresh_inode_enter(inode); - if (attr_cmp > 0) + if (attr_cmp > 0 || nfs_inode_finish_partial_attr_update(fattr, inode)) ret = nfs_update_inode(inode, fattr); else if (attr_cmp == 0) ret = nfs_check_inode_attributes(inode, fattr); -- cgit v1.2.3 From 82eae5a432cb2d50d7ea51428dcad7894e4165fe Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 1 Apr 2021 14:59:59 -0400 Subject: NFSv4: nfs4_inc/dec_nlink_locked should also invalidate ctime If the nlink changes, then so will the ctime. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 0b0a48d78299..d05f4ca5d9c0 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1169,14 +1169,18 @@ int nfs4_call_sync(struct rpc_clnt *clnt, static void nfs4_inc_nlink_locked(struct inode *inode) { - nfs_set_cache_invalid(inode, NFS_INO_INVALID_NLINK); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE | + NFS_INO_INVALID_CTIME | + NFS_INO_INVALID_NLINK); inc_nlink(inode); } static void nfs4_dec_nlink_locked(struct inode *inode) { - nfs_set_cache_invalid(inode, NFS_INO_INVALID_NLINK); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE | + NFS_INO_INVALID_CTIME | + NFS_INO_INVALID_NLINK); drop_nlink(inode); } @@ -4602,11 +4606,11 @@ _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype) status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1); if (status == 0) { spin_lock(&dir->i_lock); - nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp, - NFS_INO_INVALID_DATA); /* Removing a directory decrements nlink in the parent */ if (ftype == NF4DIR && dir->i_nlink > 2) nfs4_dec_nlink_locked(dir); + nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp, + NFS_INO_INVALID_DATA); spin_unlock(&dir->i_lock); } return status; @@ -4864,12 +4868,12 @@ static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_ &data->arg.seq_args, &data->res.seq_res, 1); if (status == 0) { spin_lock(&dir->i_lock); - nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo, - data->res.fattr->time_start, - NFS_INO_INVALID_DATA); /* Creating a directory bumps nlink in the parent */ if (data->arg.ftype == NF4DIR) nfs4_inc_nlink_locked(dir); + nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo, + data->res.fattr->time_start, + NFS_INO_INVALID_DATA); spin_unlock(&dir->i_lock); status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label); } -- cgit v1.2.3 From 1301e421b75b90b1a6101961b3aca2d91a9a0599 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 1 Apr 2021 14:57:56 -0400 Subject: NFSv4: link must update the inode nlink. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 2 +- fs/nfs/nfs4proc.c | 9 +++++++++ 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index d2835d211a73..5c25c8cc037a 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1711,7 +1711,7 @@ static void nfs_drop_nlink(struct inode *inode) NFS_I(inode)->attr_gencount = nfs_inc_attr_generation_counter(); nfs_set_cache_invalid( inode, NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_CTIME | - NFS_INO_INVALID_NLINK | NFS_INO_REVAL_FORCED); + NFS_INO_INVALID_NLINK); spin_unlock(&inode->i_lock); } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index d05f4ca5d9c0..2215f20e0e78 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1175,6 +1175,14 @@ nfs4_inc_nlink_locked(struct inode *inode) inc_nlink(inode); } +static void +nfs4_inc_nlink(struct inode *inode) +{ + spin_lock(&inode->i_lock); + nfs4_inc_nlink_locked(inode); + spin_unlock(&inode->i_lock); +} + static void nfs4_dec_nlink_locked(struct inode *inode) { @@ -4791,6 +4799,7 @@ static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct if (!status) { nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start, NFS_INO_INVALID_DATA); + nfs4_inc_nlink(inode); status = nfs_post_op_update_inode(inode, res.fattr); if (!status) nfs_setsecurity(inode, res.fattr, res.label); -- cgit v1.2.3 From cc7f2dae63bca9579d65a46f3bf2c48a15961e19 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 11 Apr 2021 14:31:24 -0400 Subject: NFS: Don't store NFS_INO_REVAL_FORCED NFS_INO_REVAL_FORCED is intended to tell us that the cache needs revalidation despite the fact that we hold a delegation. We shouldn't need to store it anymore, though. Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 21 +++++++++++++---- fs/nfs/delegation.h | 3 +-- fs/nfs/inode.c | 66 ++++++++++++++++++++--------------------------------- fs/nfs/nfs4proc.c | 5 +--- 4 files changed, 43 insertions(+), 52 deletions(-) (limited to 'fs') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 6a29de964268..e6ec6f09ac6e 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -481,6 +481,22 @@ int nfs_inode_set_delegation(struct inode *inode, const struct cred *cred, if (freeme == NULL) goto out; add_new: + /* + * If we didn't revalidate the change attribute before setting + * the delegation, then pre-emptively ask for a full attribute + * cache revalidation. + */ + spin_lock(&inode->i_lock); + if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_CHANGE) + nfs_set_cache_invalid(inode, + NFS_INO_INVALID_ATIME | NFS_INO_INVALID_CTIME | + NFS_INO_INVALID_MTIME | NFS_INO_INVALID_SIZE | + NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK | + NFS_INO_INVALID_OTHER | NFS_INO_INVALID_DATA | + NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL | + NFS_INO_INVALID_XATTR); + spin_unlock(&inode->i_lock); + list_add_tail_rcu(&delegation->super_list, &server->delegations); rcu_assign_pointer(nfsi->delegation, delegation); delegation = NULL; @@ -488,11 +504,6 @@ add_new: atomic_long_inc(&nfs_active_delegations); trace_nfs4_set_delegation(inode, type); - - spin_lock(&inode->i_lock); - if (NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME)) - NFS_I(inode)->cache_validity |= NFS_INO_REVAL_FORCED; - spin_unlock(&inode->i_lock); out: spin_unlock(&clp->cl_lock); if (delegation != NULL) diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index 9b00a0b7f832..c19b4fd20781 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -84,8 +84,7 @@ int nfs4_inode_make_writeable(struct inode *inode); static inline int nfs_have_delegated_attributes(struct inode *inode) { - return NFS_PROTO(inode)->have_delegation(inode, FMODE_READ) && - !(NFS_I(inode)->cache_validity & NFS_INO_REVAL_FORCED); + return NFS_PROTO(inode)->have_delegation(inode, FMODE_READ); } #endif diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index b88e9dc72eec..7fa914e24fc4 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -200,21 +200,19 @@ void nfs_set_cache_invalid(struct inode *inode, unsigned long flags) if (have_delegation) { if (!(flags & NFS_INO_REVAL_FORCED)) flags &= ~(NFS_INO_INVALID_MODE | - NFS_INO_INVALID_OTHER); - flags &= ~(NFS_INO_INVALID_CHANGE - | NFS_INO_INVALID_SIZE - | NFS_INO_INVALID_XATTR); + NFS_INO_INVALID_OTHER | + NFS_INO_INVALID_XATTR); + flags &= ~(NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE); } else if (flags & NFS_INO_REVAL_PAGECACHE) flags |= NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_SIZE; - flags &= ~NFS_INO_REVAL_PAGECACHE; - if (!nfs_has_xattr_cache(nfsi)) flags &= ~NFS_INO_INVALID_XATTR; if (flags & NFS_INO_INVALID_DATA) nfs_fscache_invalidate(inode); if (inode->i_mapping->nrpages == 0) flags &= ~(NFS_INO_INVALID_DATA|NFS_INO_DATA_INVAL_DEFER); + flags &= ~(NFS_INO_REVAL_PAGECACHE | NFS_INO_REVAL_FORCED); nfsi->cache_validity |= flags; } EXPORT_SYMBOL_GPL(nfs_set_cache_invalid); @@ -560,9 +558,6 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st } else if (fattr->size != 0) nfs_set_cache_invalid(inode, NFS_INO_INVALID_BLOCKS); - if (nfsi->cache_validity != 0) - nfsi->cache_validity |= NFS_INO_REVAL_FORCED; - nfs_setsecurity(inode, fattr, label); nfsi->attrtimeo = NFS_MINATTRTIMEO(inode); @@ -2032,7 +2027,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) save_cache_validity & (NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME | NFS_INO_INVALID_SIZE | - NFS_INO_REVAL_FORCED); + NFS_INO_INVALID_BLOCKS); cache_revalidated = false; } @@ -2064,27 +2059,24 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) attr_changed = true; } } else { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_CHANGE - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_CHANGE; cache_revalidated = false; } if (fattr->valid & NFS_ATTR_FATTR_MTIME) { inode->i_mtime = fattr->mtime; } else if (server->caps & NFS_CAP_MTIME) { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_MTIME - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_MTIME; cache_revalidated = false; } if (fattr->valid & NFS_ATTR_FATTR_CTIME) { inode->i_ctime = fattr->ctime; } else if (server->caps & NFS_CAP_CTIME) { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_CTIME - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_CTIME; cache_revalidated = false; } @@ -2115,19 +2107,16 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) fattr->valid |= NFS_ATTR_FATTR_SPACE_USED; } } else { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_SIZE - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_SIZE; cache_revalidated = false; } - if (fattr->valid & NFS_ATTR_FATTR_ATIME) inode->i_atime = fattr->atime; else if (server->caps & NFS_CAP_ATIME) { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_ATIME - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_ATIME; cache_revalidated = false; } @@ -2141,9 +2130,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) attr_changed = true; } } else if (server->caps & NFS_CAP_MODE) { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_MODE - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_MODE; cache_revalidated = false; } @@ -2155,9 +2143,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) attr_changed = true; } } else if (server->caps & NFS_CAP_OWNER) { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_OTHER - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_OTHER; cache_revalidated = false; } @@ -2169,9 +2156,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) attr_changed = true; } } else if (server->caps & NFS_CAP_OWNER_GROUP) { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_OTHER - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_OTHER; cache_revalidated = false; } @@ -2183,9 +2169,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) attr_changed = true; } } else if (server->caps & NFS_CAP_NLINK) { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_NLINK - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_NLINK; cache_revalidated = false; } @@ -2197,9 +2182,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) } else if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED) inode->i_blocks = fattr->du.nfs2.blocks; else { - nfsi->cache_validity |= save_cache_validity & - (NFS_INO_INVALID_BLOCKS - | NFS_INO_REVAL_FORCED); + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_BLOCKS; cache_revalidated = false; } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 2215f20e0e78..bcbb057d5529 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5477,10 +5477,7 @@ static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src, if (cache_validity & NFS_INO_INVALID_BLOCKS) bitmask[1] |= FATTR4_WORD1_SPACE_USED; - if (nfs4_have_delegation(inode, FMODE_READ) && - !(cache_validity & NFS_INO_REVAL_FORCED)) - bitmask[0] &= ~FATTR4_WORD0_SIZE; - else if (cache_validity & NFS_INO_INVALID_SIZE) + if (cache_validity & NFS_INO_INVALID_SIZE) bitmask[0] |= FATTR4_WORD0_SIZE; for (i = 0; i < NFS4_BITMASK_SZ; i++) -- cgit v1.2.3 From ce62b114bbad9346641d16853c528ba01513e1b0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 5 Mar 2018 15:01:18 -0500 Subject: NFS: Split attribute support out from the server capabilities There are lots of attributes, and they are crowding out the bit space. Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 15 ++++++++++++--- fs/nfs/inode.c | 51 +++++++++++++++++++++++++++++++-------------------- fs/nfs/nfs4proc.c | 49 +++++++++++++++++++++++++------------------------ 3 files changed, 68 insertions(+), 47 deletions(-) (limited to 'fs') diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 2aeb4e52a4f1..cfeaadf56bf0 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -696,9 +696,18 @@ static int nfs_init_server(struct nfs_server *server, /* Initialise the client representation from the mount data */ server->flags = ctx->flags; server->options = ctx->options; - server->caps |= NFS_CAP_HARDLINKS|NFS_CAP_SYMLINKS|NFS_CAP_FILEID| - NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|NFS_CAP_OWNER_GROUP| - NFS_CAP_ATIME|NFS_CAP_CTIME|NFS_CAP_MTIME; + server->caps |= NFS_CAP_HARDLINKS | NFS_CAP_SYMLINKS; + + switch (clp->rpc_ops->version) { + case 2: + server->fattr_valid = NFS_ATTR_FATTR_V2; + break; + case 3: + server->fattr_valid = NFS_ATTR_FATTR_V3; + break; + default: + server->fattr_valid = NFS_ATTR_FATTR_V4; + } if (ctx->rsize) server->rsize = nfs_block_size(ctx->rsize, NULL); diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 7fa914e24fc4..6d04ebb4f084 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -438,6 +438,7 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st .fattr = fattr }; struct inode *inode = ERR_PTR(-ENOENT); + u64 fattr_supported = NFS_SB(sb)->fattr_valid; unsigned long hash; nfs_attr_check_mountpoint(sb, fattr); @@ -470,7 +471,7 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st inode->i_mode = fattr->mode; nfsi->cache_validity = 0; if ((fattr->valid & NFS_ATTR_FATTR_MODE) == 0 - && nfs_server_capable(inode, NFS_CAP_MODE)) + && (fattr_supported & NFS_ATTR_FATTR_MODE)) nfs_set_cache_invalid(inode, NFS_INO_INVALID_MODE); /* Why so? Because we want revalidate for devices/FIFOs, and * that's precisely what we have in nfs_file_inode_operations. @@ -516,15 +517,15 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st nfsi->attr_gencount = fattr->gencount; if (fattr->valid & NFS_ATTR_FATTR_ATIME) inode->i_atime = fattr->atime; - else if (nfs_server_capable(inode, NFS_CAP_ATIME)) + else if (fattr_supported & NFS_ATTR_FATTR_ATIME) nfs_set_cache_invalid(inode, NFS_INO_INVALID_ATIME); if (fattr->valid & NFS_ATTR_FATTR_MTIME) inode->i_mtime = fattr->mtime; - else if (nfs_server_capable(inode, NFS_CAP_MTIME)) + else if (fattr_supported & NFS_ATTR_FATTR_MTIME) nfs_set_cache_invalid(inode, NFS_INO_INVALID_MTIME); if (fattr->valid & NFS_ATTR_FATTR_CTIME) inode->i_ctime = fattr->ctime; - else if (nfs_server_capable(inode, NFS_CAP_CTIME)) + else if (fattr_supported & NFS_ATTR_FATTR_CTIME) nfs_set_cache_invalid(inode, NFS_INO_INVALID_CTIME); if (fattr->valid & NFS_ATTR_FATTR_CHANGE) inode_set_iversion_raw(inode, fattr->change_attr); @@ -536,26 +537,30 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, st nfs_set_cache_invalid(inode, NFS_INO_INVALID_SIZE); if (fattr->valid & NFS_ATTR_FATTR_NLINK) set_nlink(inode, fattr->nlink); - else if (nfs_server_capable(inode, NFS_CAP_NLINK)) + else if (fattr_supported & NFS_ATTR_FATTR_NLINK) nfs_set_cache_invalid(inode, NFS_INO_INVALID_NLINK); if (fattr->valid & NFS_ATTR_FATTR_OWNER) inode->i_uid = fattr->uid; - else if (nfs_server_capable(inode, NFS_CAP_OWNER)) + else if (fattr_supported & NFS_ATTR_FATTR_OWNER) nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER); if (fattr->valid & NFS_ATTR_FATTR_GROUP) inode->i_gid = fattr->gid; - else if (nfs_server_capable(inode, NFS_CAP_OWNER_GROUP)) + else if (fattr_supported & NFS_ATTR_FATTR_GROUP) nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER); if (nfs_server_capable(inode, NFS_CAP_XATTR)) nfs_set_cache_invalid(inode, NFS_INO_INVALID_XATTR); if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED) inode->i_blocks = fattr->du.nfs2.blocks; - else if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) { + else if (fattr_supported & NFS_ATTR_FATTR_BLOCKS_USED && + fattr->size != 0) + nfs_set_cache_invalid(inode, NFS_INO_INVALID_BLOCKS); + if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) { /* * report the blocks in 512byte units */ inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used); - } else if (fattr->size != 0) + } else if (fattr_supported & NFS_ATTR_FATTR_SPACE_USED && + fattr->size != 0) nfs_set_cache_invalid(inode, NFS_INO_INVALID_BLOCKS); nfs_setsecurity(inode, fattr, label); @@ -1952,9 +1957,10 @@ EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc); */ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) { - struct nfs_server *server; + struct nfs_server *server = NFS_SERVER(inode); struct nfs_inode *nfsi = NFS_I(inode); loff_t cur_isize, new_isize; + u64 fattr_supported = server->fattr_valid; unsigned long invalid = 0; unsigned long now = jiffies; unsigned long save_cache_validity; @@ -1998,7 +2004,6 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) goto out_err; } - server = NFS_SERVER(inode); /* Update the fsid? */ if (S_ISDIR(inode->i_mode) && (fattr->valid & NFS_ATTR_FATTR_FSID) && !nfs_fsid_equal(&server->fsid, &fattr->fsid) && @@ -2066,7 +2071,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) if (fattr->valid & NFS_ATTR_FATTR_MTIME) { inode->i_mtime = fattr->mtime; - } else if (server->caps & NFS_CAP_MTIME) { + } else if (fattr_supported & NFS_ATTR_FATTR_MTIME) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_MTIME; cache_revalidated = false; @@ -2074,7 +2079,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) if (fattr->valid & NFS_ATTR_FATTR_CTIME) { inode->i_ctime = fattr->ctime; - } else if (server->caps & NFS_CAP_CTIME) { + } else if (fattr_supported & NFS_ATTR_FATTR_CTIME) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_CTIME; cache_revalidated = false; @@ -2114,7 +2119,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) if (fattr->valid & NFS_ATTR_FATTR_ATIME) inode->i_atime = fattr->atime; - else if (server->caps & NFS_CAP_ATIME) { + else if (fattr_supported & NFS_ATTR_FATTR_ATIME) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_ATIME; cache_revalidated = false; @@ -2129,7 +2134,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) | NFS_INO_INVALID_ACL; attr_changed = true; } - } else if (server->caps & NFS_CAP_MODE) { + } else if (fattr_supported & NFS_ATTR_FATTR_MODE) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_MODE; cache_revalidated = false; @@ -2142,7 +2147,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) inode->i_uid = fattr->uid; attr_changed = true; } - } else if (server->caps & NFS_CAP_OWNER) { + } else if (fattr_supported & NFS_ATTR_FATTR_OWNER) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_OTHER; cache_revalidated = false; @@ -2155,7 +2160,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) inode->i_gid = fattr->gid; attr_changed = true; } - } else if (server->caps & NFS_CAP_OWNER_GROUP) { + } else if (fattr_supported & NFS_ATTR_FATTR_GROUP) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_OTHER; cache_revalidated = false; @@ -2168,7 +2173,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) set_nlink(inode, fattr->nlink); attr_changed = true; } - } else if (server->caps & NFS_CAP_NLINK) { + } else if (fattr_supported & NFS_ATTR_FATTR_NLINK) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_NLINK; cache_revalidated = false; @@ -2179,9 +2184,15 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) * report the blocks in 512byte units */ inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used); - } else if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED) + } else if (fattr_supported & NFS_ATTR_FATTR_SPACE_USED) { + nfsi->cache_validity |= + save_cache_validity & NFS_INO_INVALID_BLOCKS; + cache_revalidated = false; + } + + if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED) { inode->i_blocks = fattr->du.nfs2.blocks; - else { + } else if (fattr_supported & NFS_ATTR_FATTR_BLOCKS_USED) { nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_BLOCKS; cache_revalidated = false; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index bcbb057d5529..21c31aebb116 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3868,12 +3868,9 @@ static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *f res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK; } memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask)); - server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS| - NFS_CAP_SYMLINKS|NFS_CAP_FILEID| - NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER| - NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME| - NFS_CAP_CTIME|NFS_CAP_MTIME| - NFS_CAP_SECURITY_LABEL); + server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS | + NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL); + server->fattr_valid = NFS_ATTR_FATTR_V4; if (res.attr_bitmask[0] & FATTR4_WORD0_ACL && res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL) server->caps |= NFS_CAP_ACLS; @@ -3881,25 +3878,29 @@ static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *f server->caps |= NFS_CAP_HARDLINKS; if (res.has_symlinks != 0) server->caps |= NFS_CAP_SYMLINKS; - if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID) - server->caps |= NFS_CAP_FILEID; - if (res.attr_bitmask[1] & FATTR4_WORD1_MODE) - server->caps |= NFS_CAP_MODE; - if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS) - server->caps |= NFS_CAP_NLINK; - if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER) - server->caps |= NFS_CAP_OWNER; - if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP) - server->caps |= NFS_CAP_OWNER_GROUP; - if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS) - server->caps |= NFS_CAP_ATIME; - if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA) - server->caps |= NFS_CAP_CTIME; - if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY) - server->caps |= NFS_CAP_MTIME; + if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID)) + server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID; + if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE)) + server->fattr_valid &= ~NFS_ATTR_FATTR_MODE; + if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)) + server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK; + if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER)) + server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER | + NFS_ATTR_FATTR_OWNER_NAME); + if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)) + server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP | + NFS_ATTR_FATTR_GROUP_NAME); + if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED)) + server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED; + if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)) + server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME; + if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)) + server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME; + if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)) + server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME; #ifdef CONFIG_NFS_V4_SECURITY_LABEL - if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL) - server->caps |= NFS_CAP_SECURITY_LABEL; + if (!(res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)) + server->fattr_valid &= ~NFS_ATTR_FATTR_V4_SECURITY_LABEL; #endif memcpy(server->attr_bitmask_nl, res.attr_bitmask, sizeof(server->attr_bitmask)); -- cgit v1.2.3 From da934ae0a8aa20f5ac6bfa04bdfe8bd8c5b438e7 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 1 Apr 2021 14:38:13 -0400 Subject: NFSv4: Add tracing for COMPOUND errors When the server returns a different operation than we expected, then trace that. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4trace.h | 35 +++++++++++++++++++++++++++++++++++ fs/nfs/nfs4xdr.c | 4 +--- 2 files changed, 36 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h index 48d761e593fb..3c505bed52a7 100644 --- a/fs/nfs/nfs4trace.h +++ b/fs/nfs/nfs4trace.h @@ -666,6 +666,41 @@ TRACE_EVENT(nfs4_state_mgr_failed, ) ) +TRACE_EVENT(nfs4_xdr_bad_operation, + TP_PROTO( + const struct xdr_stream *xdr, + u32 op, + u32 expected + ), + + TP_ARGS(xdr, op, expected), + + TP_STRUCT__entry( + __field(unsigned int, task_id) + __field(unsigned int, client_id) + __field(u32, xid) + __field(u32, op) + __field(u32, expected) + ), + + TP_fast_assign( + const struct rpc_rqst *rqstp = xdr->rqst; + const struct rpc_task *task = rqstp->rq_task; + + __entry->task_id = task->tk_pid; + __entry->client_id = task->tk_client->cl_clid; + __entry->xid = be32_to_cpu(rqstp->rq_xid); + __entry->op = op; + __entry->expected = expected; + ), + + TP_printk( + "task:%u@%d xid=0x%08x operation=%u, expected=%u", + __entry->task_id, __entry->client_id, __entry->xid, + __entry->op, __entry->expected + ) +); + TRACE_EVENT(nfs4_xdr_status, TP_PROTO( const struct xdr_stream *xdr, diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index edac4718dec1..ee7a7a132acc 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -3210,9 +3210,7 @@ out_status: *nfs_retval = nfs4_stat_to_errno(nfserr); return true; out_bad_operation: - dprintk("nfs: Server returned operation" - " %d but we issued a request for %d\n", - opnum, expected); + trace_nfs4_xdr_bad_operation(xdr, opnum, expected); *nfs_retval = -EREMOTEIO; return false; out_overflow: -- cgit v1.2.3 From 3d66bae156a652be18e278f3c88bc3e069ae824b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 1 Apr 2021 13:54:19 -0400 Subject: NFSv4: Convert nfs_xdr_status tracepoint to an event class We would like the ability to record other XDR errors, particularly those that are due to server bugs. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4trace.h | 11 ++++++++++- fs/nfs/nfstrace.h | 10 +++++++++- 2 files changed, 19 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h index 3c505bed52a7..114983a706e9 100644 --- a/fs/nfs/nfs4trace.h +++ b/fs/nfs/nfs4trace.h @@ -701,7 +701,7 @@ TRACE_EVENT(nfs4_xdr_bad_operation, ) ); -TRACE_EVENT(nfs4_xdr_status, +DECLARE_EVENT_CLASS(nfs4_xdr_event, TP_PROTO( const struct xdr_stream *xdr, u32 op, @@ -736,6 +736,15 @@ TRACE_EVENT(nfs4_xdr_status, __entry->op ) ); +#define DEFINE_NFS4_XDR_EVENT(name) \ + DEFINE_EVENT(nfs4_xdr_event, name, \ + TP_PROTO( \ + const struct xdr_stream *xdr, \ + u32 op, \ + u32 error \ + ), \ + TP_ARGS(xdr, op, error)) +DEFINE_NFS4_XDR_EVENT(nfs4_xdr_status); DECLARE_EVENT_CLASS(nfs4_cb_error_class, TP_PROTO( diff --git a/fs/nfs/nfstrace.h b/fs/nfs/nfstrace.h index 41a161cd31f6..45eda4db42fd 100644 --- a/fs/nfs/nfstrace.h +++ b/fs/nfs/nfstrace.h @@ -1401,7 +1401,7 @@ TRACE_DEFINE_ENUM(NFSERR_JUKEBOX); { NFSERR_BADTYPE, "BADTYPE" }, \ { NFSERR_JUKEBOX, "JUKEBOX" }) -TRACE_EVENT(nfs_xdr_status, +DECLARE_EVENT_CLASS(nfs_xdr_event, TP_PROTO( const struct xdr_stream *xdr, int error @@ -1443,6 +1443,14 @@ TRACE_EVENT(nfs_xdr_status, nfs_show_status(__entry->error) ) ); +#define DEFINE_NFS_XDR_EVENT(name) \ + DEFINE_EVENT(nfs_xdr_event, name, \ + TP_PROTO( \ + const struct xdr_stream *xdr, \ + int error \ + ), \ + TP_ARGS(xdr, error)) +DEFINE_NFS_XDR_EVENT(nfs_xdr_status); #endif /* _TRACE_NFS_H */ -- cgit v1.2.3 From eb3d58c68e39fad68d8054e0324eb06d82dcedbb Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 1 Apr 2021 14:03:26 -0400 Subject: NFSv4: Catch and trace server filehandle encoding errors If the server returns a filehandle with an invalid length, then trace that, and return an EREMOTEIO error. Signed-off-by: Trond Myklebust --- fs/nfs/nfs3xdr.c | 4 ++-- fs/nfs/nfs4trace.h | 1 + fs/nfs/nfs4xdr.c | 13 +++++++++---- fs/nfs/nfstrace.c | 1 + fs/nfs/nfstrace.h | 1 + 5 files changed, 14 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index 83ad62c81fc7..e6eca1d7481b 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -433,7 +433,7 @@ static int decode_nfs_fh3(struct xdr_stream *xdr, struct nfs_fh *fh) if (unlikely(!p)) return -EIO; length = be32_to_cpup(p++); - if (unlikely(length > NFS3_FHSIZE)) + if (unlikely(length > NFS3_FHSIZE || length == 0)) goto out_toobig; p = xdr_inline_decode(xdr, length); if (unlikely(!p)) @@ -442,7 +442,7 @@ static int decode_nfs_fh3(struct xdr_stream *xdr, struct nfs_fh *fh) memcpy(fh->data, p, length); return 0; out_toobig: - dprintk("NFS: file handle size (%u) too big\n", length); + trace_nfs_xdr_bad_filehandle(xdr, NFSERR_BADHANDLE); return -E2BIG; } diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h index 114983a706e9..2ef75caad6da 100644 --- a/fs/nfs/nfs4trace.h +++ b/fs/nfs/nfs4trace.h @@ -745,6 +745,7 @@ DECLARE_EVENT_CLASS(nfs4_xdr_event, ), \ TP_ARGS(xdr, op, error)) DEFINE_NFS4_XDR_EVENT(nfs4_xdr_status); +DEFINE_NFS4_XDR_EVENT(nfs4_xdr_bad_filehandle); DECLARE_EVENT_CLASS(nfs4_cb_error_class, TP_PROTO( diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index ee7a7a132acc..3d7e10c26597 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -3495,8 +3495,11 @@ static int decode_attr_filehandle(struct xdr_stream *xdr, uint32_t *bitmap, stru if (unlikely(!p)) return -EIO; len = be32_to_cpup(p); - if (len > NFS4_FHSIZE) - return -EIO; + if (len > NFS4_FHSIZE || len == 0) { + trace_nfs4_xdr_bad_filehandle(xdr, OP_READDIR, + NFS4ERR_BADHANDLE); + return -EREMOTEIO; + } p = xdr_inline_decode(xdr, len); if (unlikely(!p)) return -EIO; @@ -4952,8 +4955,10 @@ static int decode_getfh(struct xdr_stream *xdr, struct nfs_fh *fh) if (unlikely(!p)) return -EIO; len = be32_to_cpup(p); - if (len > NFS4_FHSIZE) - return -EIO; + if (len > NFS4_FHSIZE || len == 0) { + trace_nfs4_xdr_bad_filehandle(xdr, OP_GETFH, NFS4ERR_BADHANDLE); + return -EREMOTEIO; + } fh->size = len; p = xdr_inline_decode(xdr, len); if (unlikely(!p)) diff --git a/fs/nfs/nfstrace.c b/fs/nfs/nfstrace.c index a90b363500c2..5d1bfccbb4da 100644 --- a/fs/nfs/nfstrace.c +++ b/fs/nfs/nfstrace.c @@ -12,3 +12,4 @@ EXPORT_TRACEPOINT_SYMBOL_GPL(nfs_fsync_enter); EXPORT_TRACEPOINT_SYMBOL_GPL(nfs_fsync_exit); EXPORT_TRACEPOINT_SYMBOL_GPL(nfs_xdr_status); +EXPORT_TRACEPOINT_SYMBOL_GPL(nfs_xdr_bad_filehandle); diff --git a/fs/nfs/nfstrace.h b/fs/nfs/nfstrace.h index 45eda4db42fd..eb1ef3462e84 100644 --- a/fs/nfs/nfstrace.h +++ b/fs/nfs/nfstrace.h @@ -1451,6 +1451,7 @@ DECLARE_EVENT_CLASS(nfs_xdr_event, ), \ TP_ARGS(xdr, error)) DEFINE_NFS_XDR_EVENT(nfs_xdr_status); +DEFINE_NFS_XDR_EVENT(nfs_xdr_bad_filehandle); #endif /* _TRACE_NFS_H */ -- cgit v1.2.3 From ed34695e15aba74f45247f1ee2cf7e09d449f925 Mon Sep 17 00:00:00 2001 From: Nikola Livic Date: Mon, 29 Mar 2021 11:56:49 +0300 Subject: pNFS/flexfiles: fix incorrect size check in decode_nfs_fh() We (adam zabrocki, alexander matrosov, alexander tereshkin, maksym bazalii) observed the check: if (fh->size > sizeof(struct nfs_fh)) should not use the size of the nfs_fh struct which includes an extra two bytes from the size field. struct nfs_fh { unsigned short size; unsigned char data[NFS_MAXFHSIZE]; } but should determine the size from data[NFS_MAXFHSIZE] so the memcpy will not write 2 bytes beyond destination. The proposed fix is to compare against the NFS_MAXFHSIZE directly, as is done elsewhere in fs code base. Fixes: d67ae825a59d ("pnfs/flexfiles: Add the FlexFile Layout Driver") Signed-off-by: Nikola Livic Signed-off-by: Dan Carpenter Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 872112bffcab..d383de00d486 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -106,7 +106,7 @@ static int decode_nfs_fh(struct xdr_stream *xdr, struct nfs_fh *fh) if (unlikely(!p)) return -ENOBUFS; fh->size = be32_to_cpup(p++); - if (fh->size > sizeof(struct nfs_fh)) { + if (fh->size > NFS_MAXFHSIZE) { printk(KERN_ERR "NFS flexfiles: Too big fh received %d\n", fh->size); return -EOVERFLOW; -- cgit v1.2.3 From 73f5c88f521a630ea1628beb9c2d48a2e777a419 Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Wed, 31 Mar 2021 15:30:25 -0400 Subject: NFSv4.2 fix handling of sr_eof in SEEK's reply Currently the client ignores the value of the sr_eof of the SEEK operation. According to the spec, if the server didn't find the requested extent and reached the end of the file, the server would return sr_eof=true. In case the request for DATA and no data was found (ie in the middle of the hole), then the lseek expects that ENXIO would be returned. Fixes: 1c6dcbe5ceff8 ("NFS: Implement SEEK") Signed-off-by: Olga Kornievskaia Signed-off-by: Trond Myklebust --- fs/nfs/nfs42proc.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 704a5246ccb5..8d64eb953347 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -667,7 +667,10 @@ static loff_t _nfs42_proc_llseek(struct file *filep, if (status) return status; - return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes); + if (whence == SEEK_DATA && res.sr_eof) + return -NFS4ERR_NXIO; + else + return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes); } loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence) -- cgit v1.2.3 From 94d202d5ca39d0eb757d16ef2624b013fb64f64d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 14 Apr 2021 10:10:09 -0400 Subject: NFSv42: Copy offload should update the file size when appropriate If the result of a copy offload or clone operation is to grow the destination file size, then we should update it. The reason is that when a client holds a delegation, it is authoritative for the file size. Fixes: 16abd2a0c124 ("NFSv4.2: fix client's attribute cache management for copy_file_range") Signed-off-by: Trond Myklebust --- fs/nfs/nfs42proc.c | 41 ++++++++++++++++++++++++++++++++--------- 1 file changed, 32 insertions(+), 9 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 8d64eb953347..3875120ef3ef 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -269,6 +269,33 @@ out: return status; } +/** + * nfs42_copy_dest_done - perform inode cache updates after clone/copy offload + * @inode: pointer to destination inode + * @pos: destination offset + * @len: copy length + * + * Punch a hole in the inode page cache, so that the NFS client will + * know to retrieve new data. + * Update the file size if necessary, and then mark the inode as having + * invalid cached values for change attribute, ctime, mtime and space used. + */ +static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len) +{ + loff_t newsize = pos + len; + loff_t end = newsize - 1; + + truncate_pagecache_range(inode, pos, end); + spin_lock(&inode->i_lock); + if (newsize > i_size_read(inode)) + i_size_write(inode, newsize); + nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE | + NFS_INO_INVALID_CTIME | + NFS_INO_INVALID_MTIME | + NFS_INO_INVALID_BLOCKS); + spin_unlock(&inode->i_lock); +} + static ssize_t _nfs42_proc_copy(struct file *src, struct nfs_lock_context *src_lock, struct file *dst, @@ -362,14 +389,8 @@ static ssize_t _nfs42_proc_copy(struct file *src, goto out; } - truncate_pagecache_range(dst_inode, pos_dst, - pos_dst + res->write_res.count); - spin_lock(&dst_inode->i_lock); - nfs_set_cache_invalid( - dst_inode, NFS_INO_REVAL_PAGECACHE | NFS_INO_REVAL_FORCED | - NFS_INO_INVALID_SIZE | NFS_INO_INVALID_ATTR | - NFS_INO_INVALID_DATA); - spin_unlock(&dst_inode->i_lock); + nfs42_copy_dest_done(dst_inode, pos_dst, res->write_res.count); + spin_lock(&src_inode->i_lock); nfs_set_cache_invalid(src_inode, NFS_INO_REVAL_PAGECACHE | NFS_INO_REVAL_FORCED | @@ -1055,8 +1076,10 @@ static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f, status = nfs4_call_sync(server->client, server, msg, &args.seq_args, &res.seq_res, 0); - if (status == 0) + if (status == 0) { + nfs42_copy_dest_done(dst_inode, dst_offset, count); status = nfs_post_op_update_inode(dst_inode, res.dst_fattr); + } kfree(res.dst_fattr); return status; -- cgit v1.2.3 From febfeaaefefb6e3a42e4ca279270a740014ce227 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 14 Apr 2021 10:27:39 -0400 Subject: NFSv42: Don't force attribute revalidation of the copy offload source When a copy offload is performed, we do not expect the source file to change other than perhaps to see the atime be updated. Signed-off-by: Trond Myklebust --- fs/nfs/nfs42proc.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 3875120ef3ef..a24349512ffe 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -390,12 +390,7 @@ static ssize_t _nfs42_proc_copy(struct file *src, } nfs42_copy_dest_done(dst_inode, pos_dst, res->write_res.count); - - spin_lock(&src_inode->i_lock); - nfs_set_cache_invalid(src_inode, NFS_INO_REVAL_PAGECACHE | - NFS_INO_REVAL_FORCED | - NFS_INO_INVALID_ATIME); - spin_unlock(&src_inode->i_lock); + nfs_invalidate_atime(src_inode); status = res->write_res.count; out: if (args->sync) -- cgit v1.2.3 From 8926cc8302819be9e67f70409ed001ecb2c924a9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 15 Apr 2021 15:09:41 -0400 Subject: NFSv4.x: Don't return NFS4ERR_NOMATCHING_LAYOUT if we're unmounting If the NFS super block is being unmounted, then we currently may end up telling the server that we've forgotten the layout while it is actually still in use by the client. In that case, just assume that the client will soon return the layout anyway, and so return NFS4ERR_DELAY in response to the layout recall. Fixes: 58ac3e59235f ("NFSv4/pnfs: Clean up nfs_layout_find_inode()") Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'fs') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index f7786e00a6a7..ed9d580826f5 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -137,12 +137,12 @@ static struct inode *nfs_layout_find_inode_by_stateid(struct nfs_client *clp, list_for_each_entry_rcu(lo, &server->layouts, plh_layouts) { if (!pnfs_layout_is_valid(lo)) continue; - if (stateid != NULL && - !nfs4_stateid_match_other(stateid, &lo->plh_stateid)) + if (!nfs4_stateid_match_other(stateid, &lo->plh_stateid)) continue; - if (!nfs_sb_active(server->super)) - continue; - inode = igrab(lo->plh_inode); + if (nfs_sb_active(server->super)) + inode = igrab(lo->plh_inode); + else + inode = ERR_PTR(-EAGAIN); rcu_read_unlock(); if (inode) return inode; @@ -176,9 +176,10 @@ static struct inode *nfs_layout_find_inode_by_fh(struct nfs_client *clp, continue; if (nfsi->layout != lo) continue; - if (!nfs_sb_active(server->super)) - continue; - inode = igrab(lo->plh_inode); + if (nfs_sb_active(server->super)) + inode = igrab(lo->plh_inode); + else + inode = ERR_PTR(-EAGAIN); rcu_read_unlock(); if (inode) return inode; -- cgit v1.2.3 From 39fd01863616964f009599e50ca5c6ea9ebf88d6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 15 Apr 2021 15:41:57 -0400 Subject: NFS: Don't discard pNFS layout segments that are marked for return If the pNFS layout segment is marked with the NFS_LSEG_LAYOUTRETURN flag, then the assumption is that it has some reporting requirement to perform through a layoutreturn (e.g. flexfiles layout stats or error information). Fixes: e0b7d420f72a ("pNFS: Don't discard layout segments that are marked for return") Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 102b66e0bdef..33574f47601f 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2468,6 +2468,9 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, assert_spin_locked(&lo->plh_inode->i_lock); + if (test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags)) + tmp_list = &lo->plh_return_segs; + list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list) if (pnfs_match_lseg_recall(lseg, return_range, seq)) { dprintk("%s: marking lseg %p iomode %d " @@ -2475,6 +2478,8 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, lseg, lseg->pls_range.iomode, lseg->pls_range.offset, lseg->pls_range.length); + if (test_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags)) + tmp_list = &lo->plh_return_segs; if (mark_lseg_invalid(lseg, tmp_list)) continue; remaining++; -- cgit v1.2.3 From de144ff4234f935bd2150108019b5d87a90a8a96 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 18 Apr 2021 15:00:45 -0400 Subject: NFSv4: Don't discard segments marked for return in _pnfs_return_layout() If the pNFS layout segment is marked with the NFS_LSEG_LAYOUTRETURN flag, then the assumption is that it has some reporting requirement to perform through a layoutreturn (e.g. flexfiles layout stats or error information). Fixes: 6d597e175012 ("pnfs: only tear down lsegs that precede seqid in LAYOUTRETURN args") Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 33574f47601f..f726f8b12b7e 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1344,7 +1344,7 @@ _pnfs_return_layout(struct inode *ino) } valid_layout = pnfs_layout_is_valid(lo); pnfs_clear_layoutcommit(ino, &tmp_list); - pnfs_mark_matching_lsegs_invalid(lo, &tmp_list, NULL, 0); + pnfs_mark_matching_lsegs_return(lo, &tmp_list, NULL, 0); if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) { struct pnfs_layout_range range = { -- cgit v1.2.3 From fb700ef026766c95578aafc0db1b208946e7ad4f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 15 Apr 2021 17:33:07 -0400 Subject: NFSv4.1: Simplify layout return in pnfs_layout_process() If the server hands us a layout that does not match the one we currently hold, then have pnfs_mark_matching_lsegs_return() just ditch the old layout if NFS_LSEG_LAYOUTRETURN is not set. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index f726f8b12b7e..03e0b34c4a64 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2410,9 +2410,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) .iomode = IOMODE_ANY, .length = NFS4_MAX_UINT64, }; - pnfs_set_plh_return_info(lo, IOMODE_ANY, 0); - pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs, - &range, 0); + pnfs_mark_matching_lsegs_return(lo, &free_me, &range, 0); goto out_forget; } else { /* We have a completely new layout */ -- cgit v1.2.3 From d9092b4bb2109502eb8972021a3f74febc931a63 Mon Sep 17 00:00:00 2001 From: Dai Ngo Date: Thu, 22 Apr 2021 03:37:49 -0400 Subject: NFSv4.2: Remove ifdef CONFIG_NFSD from NFSv4.2 client SSC code. The client SSC code should not depend on any of the CONFIG_NFSD config. This patch removes all CONFIG_NFSD from NFSv4.2 client SSC code and simplifies the config of CONFIG_NFS_V4_2_SSC_HELPER, NFSD_V4_2_INTER_SSC. Signed-off-by: Dai Ngo Signed-off-by: Trond Myklebust --- fs/Kconfig | 4 ++-- fs/nfs/nfs4file.c | 4 ---- fs/nfs/super.c | 4 ---- fs/nfsd/Kconfig | 2 +- 4 files changed, 3 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/Kconfig b/fs/Kconfig index a55bda4233bb..afa585e62332 100644 --- a/fs/Kconfig +++ b/fs/Kconfig @@ -334,8 +334,8 @@ config NFS_COMMON default y config NFS_V4_2_SSC_HELPER - tristate - default y if NFS_V4=y || NFS_FS=y + bool + default y if NFS_V4_2 source "net/sunrpc/Kconfig" source "fs/ceph/Kconfig" diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c index 441a2fa073c8..57b3821d975a 100644 --- a/fs/nfs/nfs4file.c +++ b/fs/nfs/nfs4file.c @@ -420,9 +420,7 @@ static const struct nfs4_ssc_client_ops nfs4_ssc_clnt_ops_tbl = { */ void nfs42_ssc_register_ops(void) { -#ifdef CONFIG_NFSD_V4 nfs42_ssc_register(&nfs4_ssc_clnt_ops_tbl); -#endif } /** @@ -433,9 +431,7 @@ void nfs42_ssc_register_ops(void) */ void nfs42_ssc_unregister_ops(void) { -#ifdef CONFIG_NFSD_V4 nfs42_ssc_unregister(&nfs4_ssc_clnt_ops_tbl); -#endif } #endif /* CONFIG_NFS_V4_2 */ diff --git a/fs/nfs/super.c b/fs/nfs/super.c index 13a650750f04..5bfcc2557187 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -116,16 +116,12 @@ static void unregister_nfs4_fs(void) #ifdef CONFIG_NFS_V4_2 static void nfs_ssc_register_ops(void) { -#ifdef CONFIG_NFSD_V4 nfs_ssc_register(&nfs_ssc_clnt_ops_tbl); -#endif } static void nfs_ssc_unregister_ops(void) { -#ifdef CONFIG_NFSD_V4 nfs_ssc_unregister(&nfs_ssc_clnt_ops_tbl); -#endif } #endif /* CONFIG_NFS_V4_2 */ diff --git a/fs/nfsd/Kconfig b/fs/nfsd/Kconfig index d6cff5fbe705..c5346196c46f 100644 --- a/fs/nfsd/Kconfig +++ b/fs/nfsd/Kconfig @@ -138,7 +138,7 @@ config NFSD_FLEXFILELAYOUT config NFSD_V4_2_INTER_SSC bool "NFSv4.2 inter server to server COPY" - depends on NFSD_V4 && NFS_V4_1 && NFS_V4_2 + depends on NFSD_V4 && NFS_V4_2 help This option enables support for NFSv4.2 inter server to server copy where the destination server calls the NFSv4.2 -- cgit v1.2.3