From b8bea9f6cdd7236c7c2238d022145e9b2f8aac22 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 5 Jan 2023 07:15:09 -0500 Subject: nfsd: don't open-code clear_and_wake_up_bit Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 6e8712bd7c99..5b5d39ec7b01 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -1174,9 +1174,7 @@ open_file: status = nfserr_jukebox; if (status != nfs_ok) nfsd_file_unhash(nf); - clear_bit_unlock(NFSD_FILE_PENDING, &nf->nf_flags); - smp_mb__after_atomic(); - wake_up_bit(&nf->nf_flags, NFSD_FILE_PENDING); + clear_and_wake_up_bit(NFSD_FILE_PENDING, &nf->nf_flags); goto out; } -- cgit v1.2.3 From 6c31e4c98853a4ba47355ea151b36a77c42b7734 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Fri, 6 Jan 2023 10:39:00 -0500 Subject: nfsd: NFSD_FILE_KEY_INODE only needs to find GC'ed entries Since v4 files are expected to be long-lived, there's little value in closing them out of the cache when there is conflicting access. Change the comparator to also match the gc value in the key. Change both of the current users of that key to set the gc value in the key to "true". Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 5b5d39ec7b01..c36e3032d438 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -175,6 +175,8 @@ static int nfsd_file_obj_cmpfn(struct rhashtable_compare_arg *arg, switch (key->type) { case NFSD_FILE_KEY_INODE: + if (test_bit(NFSD_FILE_GC, &nf->nf_flags) != key->gc) + return 1; if (nf->nf_inode != key->inode) return 1; break; @@ -695,6 +697,7 @@ nfsd_file_queue_for_close(struct inode *inode, struct list_head *dispose) struct nfsd_file_lookup_key key = { .type = NFSD_FILE_KEY_INODE, .inode = inode, + .gc = true, }; struct nfsd_file *nf; @@ -1049,6 +1052,7 @@ nfsd_file_is_cached(struct inode *inode) struct nfsd_file_lookup_key key = { .type = NFSD_FILE_KEY_INODE, .inode = inode, + .gc = true, }; bool ret = false; -- cgit v1.2.3 From d69b8dbfd0866abc5ec84652cc1c10fc3d4d91ef Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Fri, 6 Jan 2023 10:39:01 -0500 Subject: nfsd: simplify test_bit return in NFSD_FILE_KEY_FULL comparator test_bit returns bool, so we can just compare the result of that to the key->gc value without the "!!". Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index c36e3032d438..568963b8a477 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -189,7 +189,7 @@ static int nfsd_file_obj_cmpfn(struct rhashtable_compare_arg *arg, return 1; if (!nfsd_match_cred(nf->nf_cred, key->cred)) return 1; - if (!!test_bit(NFSD_FILE_GC, &nf->nf_flags) != key->gc) + if (test_bit(NFSD_FILE_GC, &nf->nf_flags) != key->gc) return 1; if (test_bit(NFSD_FILE_HASHED, &nf->nf_flags) == 0) return 1; -- cgit v1.2.3 From c6593366c0bf222be9c7561354dfb921c611745e Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 5 Jan 2023 07:15:11 -0500 Subject: nfsd: don't kill nfsd_files because of lease break error An error from break_lease is non-fatal, so we needn't destroy the nfsd_file in that case. Just put the reference like we normally would and return the error. Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 29 +++++++++++++++-------------- 1 file changed, 15 insertions(+), 14 deletions(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 568963b8a477..ab37b85b7207 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -1102,7 +1102,7 @@ retry: nf = nfsd_file_alloc(&key, may_flags); if (!nf) { status = nfserr_jukebox; - goto out_status; + goto out; } ret = rhashtable_lookup_insert_key(&nfsd_file_rhash_tbl, @@ -1111,13 +1111,11 @@ retry: if (likely(ret == 0)) goto open_file; - nfsd_file_slab_free(&nf->nf_rcu); - nf = NULL; if (ret == -EEXIST) goto retry; trace_nfsd_file_insert_err(rqstp, key.inode, may_flags, ret); status = nfserr_jukebox; - goto out_status; + goto construction_err; wait_for_construction: wait_on_bit(&nf->nf_flags, NFSD_FILE_PENDING, TASK_UNINTERRUPTIBLE); @@ -1127,29 +1125,25 @@ wait_for_construction: trace_nfsd_file_cons_err(rqstp, key.inode, may_flags, nf); if (!open_retry) { status = nfserr_jukebox; - goto out; + goto construction_err; } open_retry = false; - if (refcount_dec_and_test(&nf->nf_ref)) - nfsd_file_free(nf); goto retry; } - this_cpu_inc(nfsd_file_cache_hits); status = nfserrno(nfsd_open_break_lease(file_inode(nf->nf_file), may_flags)); + if (status != nfs_ok) { + nfsd_file_put(nf); + nf = NULL; + } + out: if (status == nfs_ok) { this_cpu_inc(nfsd_file_acquisitions); nfsd_file_check_write_error(nf); *pnf = nf; - } else { - if (refcount_dec_and_test(&nf->nf_ref)) - nfsd_file_free(nf); - nf = NULL; } - -out_status: put_cred(key.cred); trace_nfsd_file_acquire(rqstp, key.inode, may_flags, nf, status); return status; @@ -1179,6 +1173,13 @@ open_file: if (status != nfs_ok) nfsd_file_unhash(nf); clear_and_wake_up_bit(NFSD_FILE_PENDING, &nf->nf_flags); + if (status == nfs_ok) + goto out; + +construction_err: + if (refcount_dec_and_test(&nf->nf_ref)) + nfsd_file_free(nf); + nf = NULL; goto out; } -- cgit v1.2.3 From b680cb9b737331aad271feebbedafb865504e234 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 5 Jan 2023 07:15:12 -0500 Subject: nfsd: add some comments to nfsd_file_do_acquire David Howells mentioned that he found this bit of code confusing, so sprinkle in some comments to clarify. Reported-by: David Howells Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index ab37b85b7207..50349449a4e5 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -1094,6 +1094,11 @@ retry: rcu_read_unlock(); if (nf) { + /* + * If the nf is on the LRU then it holds an extra reference + * that must be put if it's removed. It had better not be + * the last one however, since we should hold another. + */ if (nfsd_file_lru_remove(nf)) WARN_ON_ONCE(refcount_dec_and_test(&nf->nf_ref)); goto wait_for_construction; -- cgit v1.2.3 From b2ff1bd71db2a1b193a6dde0845adcd69cbcf75e Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Wed, 18 Jan 2023 12:31:37 -0500 Subject: nfsd: don't take/put an extra reference when putting a file The last thing that filp_close does is an fput, so don't bother taking and putting the extra reference. Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 50349449a4e5..51e2947c21a7 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -382,10 +382,8 @@ nfsd_file_free(struct nfsd_file *nf) if (nf->nf_mark) nfsd_file_mark_put(nf->nf_mark); if (nf->nf_file) { - get_file(nf->nf_file); - filp_close(nf->nf_file, NULL); nfsd_file_check_write_error(nf); - fput(nf->nf_file); + filp_close(nf->nf_file, NULL); } /* -- cgit v1.2.3 From 972cc0e0924598cb293b919d39c848dc038b2c28 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 26 Jan 2023 12:21:16 -0500 Subject: nfsd: update comment over __nfsd_file_cache_purge Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 51e2947c21a7..9b7082fdd211 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -907,7 +907,8 @@ out_err: * @net: net-namespace to shut down the cache (may be NULL) * * Walk the nfsd_file cache and close out any that match @net. If @net is NULL, - * then close out everything. Called when an nfsd instance is being shut down. + * then close out everything. Called when an nfsd instance is being shut down, + * and when the exports table is flushed. */ static void __nfsd_file_cache_purge(struct net *net) -- cgit v1.2.3 From dcb779fcd4ed5984ad15991d574943d12a8693d1 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Wed, 15 Feb 2023 06:53:54 -0500 Subject: nfsd: allow reaping files still under writeback On most filesystems, there is no reason to delay reaping an nfsd_file just because its underlying inode is still under writeback. nfsd just relies on client activity or the local flusher threads to do writeback. The main exception is NFS, which flushes all of its dirty data on last close. Add a new EXPORT_OP_FLUSH_ON_CLOSE flag to allow filesystems to signal that they do this, and only skip closing files under writeback on such filesystems. Also, remove a redundant NULL file pointer check in nfsd_file_check_writeback, and clean up nfs's export op flag definitions. Signed-off-by: Jeff Layton Acked-by: Anna Schumaker Signed-off-by: Chuck Lever --- fs/nfs/export.c | 9 ++++++--- fs/nfsd/filecache.c | 12 +++++++++++- include/linux/exportfs.h | 1 + 3 files changed, 18 insertions(+), 4 deletions(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfs/export.c b/fs/nfs/export.c index d6a6d1ebb8fd..be686b8e0c54 100644 --- a/fs/nfs/export.c +++ b/fs/nfs/export.c @@ -149,7 +149,10 @@ const struct export_operations nfs_export_ops = { .encode_fh = nfs_encode_fh, .fh_to_dentry = nfs_fh_to_dentry, .get_parent = nfs_get_parent, - .flags = EXPORT_OP_NOWCC|EXPORT_OP_NOSUBTREECHK| - EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS| - EXPORT_OP_NOATOMIC_ATTR, + .flags = EXPORT_OP_NOWCC | + EXPORT_OP_NOSUBTREECHK | + EXPORT_OP_CLOSE_BEFORE_UNLINK | + EXPORT_OP_REMOTE_FS | + EXPORT_OP_NOATOMIC_ATTR | + EXPORT_OP_FLUSH_ON_CLOSE, }; diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 9b7082fdd211..a6fa6e980277 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -402,13 +402,23 @@ nfsd_file_check_writeback(struct nfsd_file *nf) struct file *file = nf->nf_file; struct address_space *mapping; - if (!file || !(file->f_mode & FMODE_WRITE)) + /* File not open for write? */ + if (!(file->f_mode & FMODE_WRITE)) return false; + + /* + * Some filesystems (e.g. NFS) flush all dirty data on close. + * On others, there is no need to wait for writeback. + */ + if (!(file_inode(file)->i_sb->s_export_op->flags & EXPORT_OP_FLUSH_ON_CLOSE)) + return false; + mapping = file->f_mapping; return mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) || mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK); } + static bool nfsd_file_lru_add(struct nfsd_file *nf) { set_bit(NFSD_FILE_REFERENCED, &nf->nf_flags); diff --git a/include/linux/exportfs.h b/include/linux/exportfs.h index 601700fedc91..9edb29101ec8 100644 --- a/include/linux/exportfs.h +++ b/include/linux/exportfs.h @@ -220,6 +220,7 @@ struct export_operations { #define EXPORT_OP_NOATOMIC_ATTR (0x10) /* Filesystem cannot supply atomic attribute updates */ +#define EXPORT_OP_FLUSH_ON_CLOSE (0x20) /* fs flushes file data on close */ unsigned long flags; }; -- cgit v1.2.3 From c4c649ab413ba6a785b25f0edbb12f617c87db2a Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Thu, 24 Nov 2022 15:09:04 -0500 Subject: NFSD: Convert filecache to rhltable While we were converting the nfs4_file hashtable to use the kernel's resizable hashtable data structure, Neil Brown observed that the list variant (rhltable) would be better for managing nfsd_file items as well. The nfsd_file hash table will contain multiple entries for the same inode -- these should be kept together on a list. And, it could be possible for exotic or malicious client behavior to cause the hash table to resize itself on every insertion. A nice simplification is that rhltable_lookup() can return a list that contains only nfsd_file items that match a given inode, which enables us to eliminate specialized hash table helper functions and use the default functions provided by the rhashtable implementation). Since we are now storing nfsd_file items for the same inode on a single list, that effectively reduces the number of hash entries that have to be tracked in the hash table. The mininum bucket count is therefore lowered. Light testing with fstests generic/531 show no regressions. Suggested-by: Neil Brown Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 311 +++++++++++++++++++++------------------------------- fs/nfsd/filecache.h | 9 +- 2 files changed, 133 insertions(+), 187 deletions(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index a6fa6e980277..2f0b2d964cbb 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -74,70 +74,9 @@ static struct list_lru nfsd_file_lru; static unsigned long nfsd_file_flags; static struct fsnotify_group *nfsd_file_fsnotify_group; static struct delayed_work nfsd_filecache_laundrette; -static struct rhashtable nfsd_file_rhash_tbl +static struct rhltable nfsd_file_rhltable ____cacheline_aligned_in_smp; -enum nfsd_file_lookup_type { - NFSD_FILE_KEY_INODE, - NFSD_FILE_KEY_FULL, -}; - -struct nfsd_file_lookup_key { - struct inode *inode; - struct net *net; - const struct cred *cred; - unsigned char need; - bool gc; - enum nfsd_file_lookup_type type; -}; - -/* - * The returned hash value is based solely on the address of an in-code - * inode, a pointer to a slab-allocated object. The entropy in such a - * pointer is concentrated in its middle bits. - */ -static u32 nfsd_file_inode_hash(const struct inode *inode, u32 seed) -{ - unsigned long ptr = (unsigned long)inode; - u32 k; - - k = ptr >> L1_CACHE_SHIFT; - k &= 0x00ffffff; - return jhash2(&k, 1, seed); -} - -/** - * nfsd_file_key_hashfn - Compute the hash value of a lookup key - * @data: key on which to compute the hash value - * @len: rhash table's key_len parameter (unused) - * @seed: rhash table's random seed of the day - * - * Return value: - * Computed 32-bit hash value - */ -static u32 nfsd_file_key_hashfn(const void *data, u32 len, u32 seed) -{ - const struct nfsd_file_lookup_key *key = data; - - return nfsd_file_inode_hash(key->inode, seed); -} - -/** - * nfsd_file_obj_hashfn - Compute the hash value of an nfsd_file - * @data: object on which to compute the hash value - * @len: rhash table's key_len parameter (unused) - * @seed: rhash table's random seed of the day - * - * Return value: - * Computed 32-bit hash value - */ -static u32 nfsd_file_obj_hashfn(const void *data, u32 len, u32 seed) -{ - const struct nfsd_file *nf = data; - - return nfsd_file_inode_hash(nf->nf_inode, seed); -} - static bool nfsd_match_cred(const struct cred *c1, const struct cred *c2) { @@ -158,55 +97,16 @@ nfsd_match_cred(const struct cred *c1, const struct cred *c2) return true; } -/** - * nfsd_file_obj_cmpfn - Match a cache item against search criteria - * @arg: search criteria - * @ptr: cache item to check - * - * Return values: - * %0 - Item matches search criteria - * %1 - Item does not match search criteria - */ -static int nfsd_file_obj_cmpfn(struct rhashtable_compare_arg *arg, - const void *ptr) -{ - const struct nfsd_file_lookup_key *key = arg->key; - const struct nfsd_file *nf = ptr; - - switch (key->type) { - case NFSD_FILE_KEY_INODE: - if (test_bit(NFSD_FILE_GC, &nf->nf_flags) != key->gc) - return 1; - if (nf->nf_inode != key->inode) - return 1; - break; - case NFSD_FILE_KEY_FULL: - if (nf->nf_inode != key->inode) - return 1; - if (nf->nf_may != key->need) - return 1; - if (nf->nf_net != key->net) - return 1; - if (!nfsd_match_cred(nf->nf_cred, key->cred)) - return 1; - if (test_bit(NFSD_FILE_GC, &nf->nf_flags) != key->gc) - return 1; - if (test_bit(NFSD_FILE_HASHED, &nf->nf_flags) == 0) - return 1; - break; - } - return 0; -} - static const struct rhashtable_params nfsd_file_rhash_params = { .key_len = sizeof_field(struct nfsd_file, nf_inode), .key_offset = offsetof(struct nfsd_file, nf_inode), - .head_offset = offsetof(struct nfsd_file, nf_rhash), - .hashfn = nfsd_file_key_hashfn, - .obj_hashfn = nfsd_file_obj_hashfn, - .obj_cmpfn = nfsd_file_obj_cmpfn, - /* Reduce resizing churn on light workloads */ - .min_size = 512, /* buckets */ + .head_offset = offsetof(struct nfsd_file, nf_rlist), + + /* + * Start with a single page hash table to reduce resizing churn + * on light workloads. + */ + .min_size = 256, .automatic_shrinking = true, }; @@ -309,27 +209,27 @@ nfsd_file_mark_find_or_create(struct nfsd_file *nf, struct inode *inode) } static struct nfsd_file * -nfsd_file_alloc(struct nfsd_file_lookup_key *key, unsigned int may) +nfsd_file_alloc(struct net *net, struct inode *inode, unsigned char need, + bool want_gc) { struct nfsd_file *nf; nf = kmem_cache_alloc(nfsd_file_slab, GFP_KERNEL); - if (nf) { - INIT_LIST_HEAD(&nf->nf_lru); - nf->nf_birthtime = ktime_get(); - nf->nf_file = NULL; - nf->nf_cred = get_current_cred(); - nf->nf_net = key->net; - nf->nf_flags = 0; - __set_bit(NFSD_FILE_HASHED, &nf->nf_flags); - __set_bit(NFSD_FILE_PENDING, &nf->nf_flags); - if (key->gc) - __set_bit(NFSD_FILE_GC, &nf->nf_flags); - nf->nf_inode = key->inode; - refcount_set(&nf->nf_ref, 1); - nf->nf_may = key->need; - nf->nf_mark = NULL; - } + if (unlikely(!nf)) + return NULL; + + INIT_LIST_HEAD(&nf->nf_lru); + nf->nf_birthtime = ktime_get(); + nf->nf_file = NULL; + nf->nf_cred = get_current_cred(); + nf->nf_net = net; + nf->nf_flags = want_gc ? + BIT(NFSD_FILE_HASHED) | BIT(NFSD_FILE_PENDING) | BIT(NFSD_FILE_GC) : + BIT(NFSD_FILE_HASHED) | BIT(NFSD_FILE_PENDING); + nf->nf_inode = inode; + refcount_set(&nf->nf_ref, 1); + nf->nf_may = need; + nf->nf_mark = NULL; return nf; } @@ -354,8 +254,8 @@ static void nfsd_file_hash_remove(struct nfsd_file *nf) { trace_nfsd_file_unhash(nf); - rhashtable_remove_fast(&nfsd_file_rhash_tbl, &nf->nf_rhash, - nfsd_file_rhash_params); + rhltable_remove(&nfsd_file_rhltable, &nf->nf_rlist, + nfsd_file_rhash_params); } static bool @@ -688,8 +588,8 @@ nfsd_file_cond_queue(struct nfsd_file *nf, struct list_head *dispose) * @inode: inode on which to close out nfsd_files * @dispose: list on which to gather nfsd_files to close out * - * An nfsd_file represents a struct file being held open on behalf of nfsd. An - * open file however can block other activity (such as leases), or cause + * An nfsd_file represents a struct file being held open on behalf of nfsd. + * An open file however can block other activity (such as leases), or cause * undesirable behavior (e.g. spurious silly-renames when reexporting NFS). * * This function is intended to find open nfsd_files when this sort of @@ -702,21 +602,17 @@ nfsd_file_cond_queue(struct nfsd_file *nf, struct list_head *dispose) static void nfsd_file_queue_for_close(struct inode *inode, struct list_head *dispose) { - struct nfsd_file_lookup_key key = { - .type = NFSD_FILE_KEY_INODE, - .inode = inode, - .gc = true, - }; + struct rhlist_head *tmp, *list; struct nfsd_file *nf; rcu_read_lock(); - do { - nf = rhashtable_lookup(&nfsd_file_rhash_tbl, &key, - nfsd_file_rhash_params); - if (!nf) - break; + list = rhltable_lookup(&nfsd_file_rhltable, &inode, + nfsd_file_rhash_params); + rhl_for_each_entry_rcu(nf, tmp, list, nf_rlist) { + if (!test_bit(NFSD_FILE_GC, &nf->nf_flags)) + continue; nfsd_file_cond_queue(nf, dispose); - } while (1); + } rcu_read_unlock(); } @@ -840,7 +736,7 @@ nfsd_file_cache_init(void) if (test_and_set_bit(NFSD_FILE_CACHE_UP, &nfsd_file_flags) == 1) return 0; - ret = rhashtable_init(&nfsd_file_rhash_tbl, &nfsd_file_rhash_params); + ret = rhltable_init(&nfsd_file_rhltable, &nfsd_file_rhash_params); if (ret) return ret; @@ -908,7 +804,7 @@ out_err: nfsd_file_mark_slab = NULL; destroy_workqueue(nfsd_filecache_wq); nfsd_filecache_wq = NULL; - rhashtable_destroy(&nfsd_file_rhash_tbl); + rhltable_destroy(&nfsd_file_rhltable); goto out; } @@ -927,7 +823,7 @@ __nfsd_file_cache_purge(struct net *net) struct nfsd_file *nf; LIST_HEAD(dispose); - rhashtable_walk_enter(&nfsd_file_rhash_tbl, &iter); + rhltable_walk_enter(&nfsd_file_rhltable, &iter); do { rhashtable_walk_start(&iter); @@ -1033,7 +929,7 @@ nfsd_file_cache_shutdown(void) nfsd_file_mark_slab = NULL; destroy_workqueue(nfsd_filecache_wq); nfsd_filecache_wq = NULL; - rhashtable_destroy(&nfsd_file_rhash_tbl); + rhltable_destroy(&nfsd_file_rhltable); for_each_possible_cpu(i) { per_cpu(nfsd_file_cache_hits, i) = 0; @@ -1044,6 +940,35 @@ nfsd_file_cache_shutdown(void) } } +static struct nfsd_file * +nfsd_file_lookup_locked(const struct net *net, const struct cred *cred, + struct inode *inode, unsigned char need, + bool want_gc) +{ + struct rhlist_head *tmp, *list; + struct nfsd_file *nf; + + list = rhltable_lookup(&nfsd_file_rhltable, &inode, + nfsd_file_rhash_params); + rhl_for_each_entry_rcu(nf, tmp, list, nf_rlist) { + if (nf->nf_may != need) + continue; + if (nf->nf_net != net) + continue; + if (!nfsd_match_cred(nf->nf_cred, cred)) + continue; + if (test_bit(NFSD_FILE_GC, &nf->nf_flags) != want_gc) + continue; + if (test_bit(NFSD_FILE_HASHED, &nf->nf_flags) == 0) + continue; + + if (!nfsd_file_get(nf)) + continue; + return nf; + } + return NULL; +} + /** * nfsd_file_is_cached - are there any cached open files for this inode? * @inode: inode to check @@ -1058,16 +983,20 @@ nfsd_file_cache_shutdown(void) bool nfsd_file_is_cached(struct inode *inode) { - struct nfsd_file_lookup_key key = { - .type = NFSD_FILE_KEY_INODE, - .inode = inode, - .gc = true, - }; + struct rhlist_head *tmp, *list; + struct nfsd_file *nf; bool ret = false; - if (rhashtable_lookup_fast(&nfsd_file_rhash_tbl, &key, - nfsd_file_rhash_params) != NULL) - ret = true; + rcu_read_lock(); + list = rhltable_lookup(&nfsd_file_rhltable, &inode, + nfsd_file_rhash_params); + rhl_for_each_entry_rcu(nf, tmp, list, nf_rlist) + if (test_bit(NFSD_FILE_GC, &nf->nf_flags)) { + ret = true; + break; + } + rcu_read_unlock(); + trace_nfsd_file_is_cached(inode, (int)ret); return ret; } @@ -1077,14 +1006,12 @@ nfsd_file_do_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp, unsigned int may_flags, struct file *file, struct nfsd_file **pnf, bool want_gc) { - struct nfsd_file_lookup_key key = { - .type = NFSD_FILE_KEY_FULL, - .need = may_flags & NFSD_FILE_MAY_MASK, - .net = SVC_NET(rqstp), - .gc = want_gc, - }; + unsigned char need = may_flags & NFSD_FILE_MAY_MASK; + struct net *net = SVC_NET(rqstp); + struct nfsd_file *new, *nf; + const struct cred *cred; bool open_retry = true; - struct nfsd_file *nf; + struct inode *inode; __be32 status; int ret; @@ -1092,14 +1019,12 @@ nfsd_file_do_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp, may_flags|NFSD_MAY_OWNER_OVERRIDE); if (status != nfs_ok) return status; - key.inode = d_inode(fhp->fh_dentry); - key.cred = get_current_cred(); + inode = d_inode(fhp->fh_dentry); + cred = get_current_cred(); retry: rcu_read_lock(); - nf = rhashtable_lookup(&nfsd_file_rhash_tbl, &key, - nfsd_file_rhash_params); - nf = nfsd_file_get(nf); + nf = nfsd_file_lookup_locked(net, cred, inode, need, want_gc); rcu_read_unlock(); if (nf) { @@ -1113,21 +1038,32 @@ retry: goto wait_for_construction; } - nf = nfsd_file_alloc(&key, may_flags); - if (!nf) { + new = nfsd_file_alloc(net, inode, need, want_gc); + if (!new) { status = nfserr_jukebox; goto out; } - ret = rhashtable_lookup_insert_key(&nfsd_file_rhash_tbl, - &key, &nf->nf_rhash, - nfsd_file_rhash_params); + rcu_read_lock(); + spin_lock(&inode->i_lock); + nf = nfsd_file_lookup_locked(net, cred, inode, need, want_gc); + if (unlikely(nf)) { + spin_unlock(&inode->i_lock); + rcu_read_unlock(); + nfsd_file_slab_free(&new->nf_rcu); + goto wait_for_construction; + } + nf = new; + ret = rhltable_insert(&nfsd_file_rhltable, &nf->nf_rlist, + nfsd_file_rhash_params); + spin_unlock(&inode->i_lock); + rcu_read_unlock(); if (likely(ret == 0)) goto open_file; if (ret == -EEXIST) goto retry; - trace_nfsd_file_insert_err(rqstp, key.inode, may_flags, ret); + trace_nfsd_file_insert_err(rqstp, inode, may_flags, ret); status = nfserr_jukebox; goto construction_err; @@ -1136,7 +1072,7 @@ wait_for_construction: /* Did construction of this file fail? */ if (!test_bit(NFSD_FILE_HASHED, &nf->nf_flags)) { - trace_nfsd_file_cons_err(rqstp, key.inode, may_flags, nf); + trace_nfsd_file_cons_err(rqstp, inode, may_flags, nf); if (!open_retry) { status = nfserr_jukebox; goto construction_err; @@ -1158,13 +1094,13 @@ out: nfsd_file_check_write_error(nf); *pnf = nf; } - put_cred(key.cred); - trace_nfsd_file_acquire(rqstp, key.inode, may_flags, nf, status); + put_cred(cred); + trace_nfsd_file_acquire(rqstp, inode, may_flags, nf, status); return status; open_file: trace_nfsd_file_alloc(nf); - nf->nf_mark = nfsd_file_mark_find_or_create(nf, key.inode); + nf->nf_mark = nfsd_file_mark_find_or_create(nf, inode); if (nf->nf_mark) { if (file) { get_file(file); @@ -1182,7 +1118,7 @@ open_file: * If construction failed, or we raced with a call to unlink() * then unhash. */ - if (status == nfs_ok && key.inode->i_nlink == 0) + if (status != nfs_ok || inode->i_nlink == 0) status = nfserr_jukebox; if (status != nfs_ok) nfsd_file_unhash(nf); @@ -1209,8 +1145,11 @@ construction_err: * seconds after the final nfsd_file_put() in case the caller * wants to re-use it. * - * Returns nfs_ok and sets @pnf on success; otherwise an nfsstat in - * network byte order is returned. + * Return values: + * %nfs_ok - @pnf points to an nfsd_file with its reference + * count boosted. + * + * On error, an nfsstat value in network byte order is returned. */ __be32 nfsd_file_acquire_gc(struct svc_rqst *rqstp, struct svc_fh *fhp, @@ -1230,8 +1169,11 @@ nfsd_file_acquire_gc(struct svc_rqst *rqstp, struct svc_fh *fhp, * but not garbage-collected. The object is unhashed after the * final nfsd_file_put(). * - * Returns nfs_ok and sets @pnf on success; otherwise an nfsstat in - * network byte order is returned. + * Return values: + * %nfs_ok - @pnf points to an nfsd_file with its reference + * count boosted. + * + * On error, an nfsstat value in network byte order is returned. */ __be32 nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp, @@ -1252,8 +1194,11 @@ nfsd_file_acquire(struct svc_rqst *rqstp, struct svc_fh *fhp, * and @file is non-NULL, use it to instantiate a new nfsd_file instead of * opening a new one. * - * Returns nfs_ok and sets @pnf on success; otherwise an nfsstat in - * network byte order is returned. + * Return values: + * %nfs_ok - @pnf points to an nfsd_file with its reference + * count boosted. + * + * On error, an nfsstat value in network byte order is returned. */ __be32 nfsd_file_acquire_opened(struct svc_rqst *rqstp, struct svc_fh *fhp, @@ -1284,7 +1229,7 @@ int nfsd_file_cache_stats_show(struct seq_file *m, void *v) lru = list_lru_count(&nfsd_file_lru); rcu_read_lock(); - ht = &nfsd_file_rhash_tbl; + ht = &nfsd_file_rhltable.ht; count = atomic_read(&ht->nelems); tbl = rht_dereference_rcu(ht->tbl, ht); buckets = tbl->size; @@ -1300,7 +1245,7 @@ int nfsd_file_cache_stats_show(struct seq_file *m, void *v) evictions += per_cpu(nfsd_file_evictions, i); } - seq_printf(m, "total entries: %u\n", count); + seq_printf(m, "total inodes: %u\n", count); seq_printf(m, "hash buckets: %u\n", buckets); seq_printf(m, "lru entries: %lu\n", lru); seq_printf(m, "cache hits: %lu\n", hits); diff --git a/fs/nfsd/filecache.h b/fs/nfsd/filecache.h index 41516a4263ea..e54165a3224f 100644 --- a/fs/nfsd/filecache.h +++ b/fs/nfsd/filecache.h @@ -29,9 +29,8 @@ struct nfsd_file_mark { * never be dereferenced, only used for comparison. */ struct nfsd_file { - struct rhash_head nf_rhash; - struct list_head nf_lru; - struct rcu_head nf_rcu; + struct rhlist_head nf_rlist; + void *nf_inode; struct file *nf_file; const struct cred *nf_cred; struct net *nf_net; @@ -40,10 +39,12 @@ struct nfsd_file { #define NFSD_FILE_REFERENCED (2) #define NFSD_FILE_GC (3) unsigned long nf_flags; - struct inode *nf_inode; /* don't deref */ refcount_t nf_ref; unsigned char nf_may; + struct nfsd_file_mark *nf_mark; + struct list_head nf_lru; + struct rcu_head nf_rcu; ktime_t nf_birthtime; }; -- cgit v1.2.3 From 92e4a6733f922f0fef1d0995f7b2d0eaff86c7ea Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Fri, 14 Apr 2023 17:31:44 -0400 Subject: nfsd: simplify the delayed disposal list code When queueing a dispose list to the appropriate "freeme" lists, it pointlessly queues the objects one at a time to an intermediate list. Remove a few helpers and just open code a list_move to make it more clear and efficient. Better document the resulting functions with kerneldoc comments. Signed-off-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 64 ++++++++++++++++++----------------------------------- 1 file changed, 22 insertions(+), 42 deletions(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index 2f0b2d964cbb..f40d8f3b35a4 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -402,49 +402,26 @@ nfsd_file_dispose_list(struct list_head *dispose) } } -static void -nfsd_file_list_remove_disposal(struct list_head *dst, - struct nfsd_fcache_disposal *l) -{ - spin_lock(&l->lock); - list_splice_init(&l->freeme, dst); - spin_unlock(&l->lock); -} - -static void -nfsd_file_list_add_disposal(struct list_head *files, struct net *net) -{ - struct nfsd_net *nn = net_generic(net, nfsd_net_id); - struct nfsd_fcache_disposal *l = nn->fcache_disposal; - - spin_lock(&l->lock); - list_splice_tail_init(files, &l->freeme); - spin_unlock(&l->lock); - queue_work(nfsd_filecache_wq, &l->work); -} - -static void -nfsd_file_list_add_pernet(struct list_head *dst, struct list_head *src, - struct net *net) -{ - struct nfsd_file *nf, *tmp; - - list_for_each_entry_safe(nf, tmp, src, nf_lru) { - if (nf->nf_net == net) - list_move_tail(&nf->nf_lru, dst); - } -} - +/** + * nfsd_file_dispose_list_delayed - move list of dead files to net's freeme list + * @dispose: list of nfsd_files to be disposed + * + * Transfers each file to the "freeme" list for its nfsd_net, to eventually + * be disposed of by the per-net garbage collector. + */ static void nfsd_file_dispose_list_delayed(struct list_head *dispose) { - LIST_HEAD(list); - struct nfsd_file *nf; - while(!list_empty(dispose)) { - nf = list_first_entry(dispose, struct nfsd_file, nf_lru); - nfsd_file_list_add_pernet(&list, dispose, nf->nf_net); - nfsd_file_list_add_disposal(&list, nf->nf_net); + struct nfsd_file *nf = list_first_entry(dispose, + struct nfsd_file, nf_lru); + struct nfsd_net *nn = net_generic(nf->nf_net, nfsd_net_id); + struct nfsd_fcache_disposal *l = nn->fcache_disposal; + + spin_lock(&l->lock); + list_move_tail(&nf->nf_lru, &l->freeme); + spin_unlock(&l->lock); + queue_work(nfsd_filecache_wq, &l->work); } } @@ -665,8 +642,8 @@ nfsd_file_close_inode_sync(struct inode *inode) * nfsd_file_delayed_close - close unused nfsd_files * @work: dummy * - * Walk the LRU list and destroy any entries that have not been used since - * the last scan. + * Scrape the freeme list for this nfsd_net, and then dispose of them + * all. */ static void nfsd_file_delayed_close(struct work_struct *work) @@ -675,7 +652,10 @@ nfsd_file_delayed_close(struct work_struct *work) struct nfsd_fcache_disposal *l = container_of(work, struct nfsd_fcache_disposal, work); - nfsd_file_list_remove_disposal(&head, l); + spin_lock(&l->lock); + list_splice_init(&l->freeme, &head); + spin_unlock(&l->lock); + nfsd_file_dispose_list(&head); } -- cgit v1.2.3 From 147abcacee33781e75588869e944ddb07528a897 Mon Sep 17 00:00:00 2001 From: Dai Ngo Date: Wed, 19 Apr 2023 10:53:18 -0700 Subject: NFSD: Fix problem of COMMIT and NFS4ERR_DELAY in infinite loop The following request sequence to the same file causes the NFS client and server getting into an infinite loop with COMMIT and NFS4ERR_DELAY: OPEN REMOVE WRITE COMMIT Problem reported by recall11, recall12, recall14, recall20, recall22, recall40, recall42, recall48, recall50 of nfstest suite. This patch restores the handling of race condition in nfsd_file_do_acquire with unlink to that prior of the regression. Fixes: ac3a2585f018 ("nfsd: rework refcounting in filecache") Signed-off-by: Dai Ngo Reviewed-by: Jeff Layton Signed-off-by: Chuck Lever --- fs/nfsd/filecache.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'fs/nfsd/filecache.c') diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c index f40d8f3b35a4..ee9c923192e0 100644 --- a/fs/nfsd/filecache.c +++ b/fs/nfsd/filecache.c @@ -1099,8 +1099,6 @@ open_file: * then unhash. */ if (status != nfs_ok || inode->i_nlink == 0) - status = nfserr_jukebox; - if (status != nfs_ok) nfsd_file_unhash(nf); clear_and_wake_up_bit(NFSD_FILE_PENDING, &nf->nf_flags); if (status == nfs_ok) -- cgit v1.2.3