From c5b1f0d92c36851aca09ac6c7c0c4f9690ac14f3 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Wed, 27 Oct 2010 15:46:08 +0200 Subject: locks/nfsd: allocate file lock outside of spinlock As suggested by Christoph Hellwig, this moves allocation of new file locks out of generic_setlease into the callers, nfs4_open_delegation and fcntl_setlease in order to allow GFP_KERNEL allocations when lock_flocks has become a spinlock. Signed-off-by: Arnd Bergmann Acked-by: J. Bruce Fields --- fs/locks.c | 36 ++++++++++++------------------------ 1 file changed, 12 insertions(+), 24 deletions(-) (limited to 'fs/locks.c') diff --git a/fs/locks.c b/fs/locks.c index 8b2b6ad56a0..0391d2ff5a4 100644 --- a/fs/locks.c +++ b/fs/locks.c @@ -162,10 +162,11 @@ EXPORT_SYMBOL_GPL(unlock_flocks); static struct kmem_cache *filelock_cache __read_mostly; /* Allocate an empty lock structure. */ -static struct file_lock *locks_alloc_lock(void) +struct file_lock *locks_alloc_lock(void) { return kmem_cache_alloc(filelock_cache, GFP_KERNEL); } +EXPORT_SYMBOL_GPL(locks_alloc_lock); void locks_release_private(struct file_lock *fl) { @@ -1365,7 +1366,6 @@ int fcntl_getlease(struct file *filp) int generic_setlease(struct file *filp, long arg, struct file_lock **flp) { struct file_lock *fl, **before, **my_before = NULL, *lease; - struct file_lock *new_fl = NULL; struct dentry *dentry = filp->f_path.dentry; struct inode *inode = dentry->d_inode; int error, rdlease_count = 0, wrlease_count = 0; @@ -1385,11 +1385,6 @@ int generic_setlease(struct file *filp, long arg, struct file_lock **flp) lease = *flp; if (arg != F_UNLCK) { - error = -ENOMEM; - new_fl = locks_alloc_lock(); - if (new_fl == NULL) - goto out; - error = -EAGAIN; if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) goto out; @@ -1434,7 +1429,6 @@ int generic_setlease(struct file *filp, long arg, struct file_lock **flp) goto out; } - error = 0; if (arg == F_UNLCK) goto out; @@ -1442,15 +1436,11 @@ int generic_setlease(struct file *filp, long arg, struct file_lock **flp) if (!leases_enable) goto out; - locks_copy_lock(new_fl, lease); - locks_insert_lock(before, new_fl); - - *flp = new_fl; + locks_insert_lock(before, lease); return 0; out: - if (new_fl != NULL) - locks_free_lock(new_fl); + locks_free_lock(lease); return error; } EXPORT_SYMBOL(generic_setlease); @@ -1514,26 +1504,24 @@ EXPORT_SYMBOL_GPL(vfs_setlease); */ int fcntl_setlease(unsigned int fd, struct file *filp, long arg) { - struct file_lock fl, *flp = &fl; + struct file_lock *fl; struct inode *inode = filp->f_path.dentry->d_inode; int error; - locks_init_lock(&fl); - error = lease_init(filp, arg, &fl); - if (error) - return error; + fl = lease_alloc(filp, arg); + if (IS_ERR(fl)) + return PTR_ERR(fl); lock_flocks(); - - error = __vfs_setlease(filp, arg, &flp); + error = __vfs_setlease(filp, arg, &fl); if (error || arg == F_UNLCK) goto out_unlock; - error = fasync_helper(fd, filp, 1, &flp->fl_fasync); + error = fasync_helper(fd, filp, 1, &fl->fl_fasync); if (error < 0) { /* remove lease just inserted by setlease */ - flp->fl_type = F_UNLCK | F_INPROGRESS; - flp->fl_break_time = jiffies - 10; + fl->fl_type = F_UNLCK | F_INPROGRESS; + fl->fl_break_time = jiffies - 10; time_out_leases(inode); goto out_unlock; } -- cgit v1.2.3 From f7347ce4ee7c65415f84be915c018473e7076f31 Mon Sep 17 00:00:00 2001 From: Linus Torvalds Date: Wed, 27 Oct 2010 12:38:12 -0400 Subject: fasync: re-organize fasync entry insertion to allow it under a spinlock You currently cannot use "fasync_helper()" in an atomic environment to insert a new fasync entry, because it will need to allocate the new "struct fasync_struct". Yet fcntl_setlease() wants to call this under lock_flocks(), which is in the process of being converted from the BKL to a spinlock. In order to fix this, this abstracts out the actual fasync list insertion and the fasync allocations into functions of their own, and teaches fs/locks.c to pre-allocate the fasync_struct entry. That way the actual list insertion can happen while holding the required spinlock. Signed-off-by: Linus Torvalds [bfields@redhat.com: rebase on top of my changes to Arnd's patch] Tested-by: J. Bruce Fields Signed-off-by: Arnd Bergmann --- fs/fcntl.c | 66 +++++++++++++++++++++++++++++++++++++++++------------- fs/locks.c | 18 ++++++++++++++- include/linux/fs.h | 5 +++++ 3 files changed, 72 insertions(+), 17 deletions(-) (limited to 'fs/locks.c') diff --git a/fs/fcntl.c b/fs/fcntl.c index f8cc34f542c..dcdbc6f5c33 100644 --- a/fs/fcntl.c +++ b/fs/fcntl.c @@ -640,7 +640,7 @@ static void fasync_free_rcu(struct rcu_head *head) * match the state "is the filp on a fasync list". * */ -static int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp) +int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp) { struct fasync_struct *fa, **fp; int result = 0; @@ -666,21 +666,28 @@ static int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp) return result; } +struct fasync_struct *fasync_alloc(void) +{ + return kmem_cache_alloc(fasync_cache, GFP_KERNEL); +} + /* - * Add a fasync entry. Return negative on error, positive if - * added, and zero if did nothing but change an existing one. - * - * NOTE! It is very important that the FASYNC flag always - * match the state "is the filp on a fasync list". + * NOTE! This can be used only for unused fasync entries: + * entries that actually got inserted on the fasync list + * need to be released by rcu - see fasync_remove_entry. */ -static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp) +void fasync_free(struct fasync_struct *new) { - struct fasync_struct *new, *fa, **fp; - int result = 0; + kmem_cache_free(fasync_cache, new); +} - new = kmem_cache_alloc(fasync_cache, GFP_KERNEL); - if (!new) - return -ENOMEM; +/* + * Insert a new entry into the fasync list. Return the pointer to the + * old one if we didn't use the new one. + */ +struct fasync_struct *fasync_insert_entry(int fd, struct file *filp, struct fasync_struct **fapp, struct fasync_struct *new) +{ + struct fasync_struct *fa, **fp; spin_lock(&filp->f_lock); spin_lock(&fasync_lock); @@ -691,8 +698,6 @@ static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fa spin_lock_irq(&fa->fa_lock); fa->fa_fd = fd; spin_unlock_irq(&fa->fa_lock); - - kmem_cache_free(fasync_cache, new); goto out; } @@ -702,13 +707,42 @@ static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fa new->fa_fd = fd; new->fa_next = *fapp; rcu_assign_pointer(*fapp, new); - result = 1; filp->f_flags |= FASYNC; out: spin_unlock(&fasync_lock); spin_unlock(&filp->f_lock); - return result; + return fa; +} + +/* + * Add a fasync entry. Return negative on error, positive if + * added, and zero if did nothing but change an existing one. + * + * NOTE! It is very important that the FASYNC flag always + * match the state "is the filp on a fasync list". + */ +static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp) +{ + struct fasync_struct *new; + + new = fasync_alloc(); + if (!new) + return -ENOMEM; + + /* + * fasync_insert_entry() returns the old (update) entry if + * it existed. + * + * So free the (unused) new entry and return 0 to let the + * caller know that we didn't add any new fasync entries. + */ + if (fasync_insert_entry(fd, filp, fapp, new)) { + fasync_free(new); + return 0; + } + + return 1; } /* diff --git a/fs/locks.c b/fs/locks.c index 0391d2ff5a4..85fd9ce1aba 100644 --- a/fs/locks.c +++ b/fs/locks.c @@ -1505,6 +1505,7 @@ EXPORT_SYMBOL_GPL(vfs_setlease); int fcntl_setlease(unsigned int fd, struct file *filp, long arg) { struct file_lock *fl; + struct fasync_struct *new; struct inode *inode = filp->f_path.dentry->d_inode; int error; @@ -1512,12 +1513,25 @@ int fcntl_setlease(unsigned int fd, struct file *filp, long arg) if (IS_ERR(fl)) return PTR_ERR(fl); + new = fasync_alloc(); + if (!new) { + locks_free_lock(fl); + return -ENOMEM; + } lock_flocks(); error = __vfs_setlease(filp, arg, &fl); if (error || arg == F_UNLCK) goto out_unlock; - error = fasync_helper(fd, filp, 1, &fl->fl_fasync); + /* + * fasync_insert_entry() returns the old entry if any. + * If there was no old entry, then it used 'new' and + * inserted it into the fasync list. Clear new so that + * we don't release it here. + */ + if (!fasync_insert_entry(fd, filp, &fl->fl_fasync, new)) + new = NULL; + if (error < 0) { /* remove lease just inserted by setlease */ fl->fl_type = F_UNLCK | F_INPROGRESS; @@ -1529,6 +1543,8 @@ int fcntl_setlease(unsigned int fd, struct file *filp, long arg) error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); out_unlock: unlock_flocks(); + if (new) + fasync_free(new); return error; } diff --git a/include/linux/fs.h b/include/linux/fs.h index 8d7de08ab54..56285e5e1de 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -1302,6 +1302,11 @@ struct fasync_struct { /* SMP safe fasync helpers: */ extern int fasync_helper(int, struct file *, int, struct fasync_struct **); +extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *); +extern int fasync_remove_entry(struct file *, struct fasync_struct **); +extern struct fasync_struct *fasync_alloc(void); +extern void fasync_free(struct fasync_struct *); + /* can be called from interrupts */ extern void kill_fasync(struct fasync_struct **, int, int); -- cgit v1.2.3 From 72f98e72551fad573c6cace8e8551ef094f482dd Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Wed, 27 Oct 2010 21:39:58 +0200 Subject: locks: turn lock_flocks into a spinlock Nothing depends on lock_flocks using the BKL any more, so we can do the switch over to a private spinlock. Signed-off-by: Arnd Bergmann --- fs/Kconfig | 1 - fs/locks.c | 5 +++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/locks.c') diff --git a/fs/Kconfig b/fs/Kconfig index 65781de44fc..3d185308ec8 100644 --- a/fs/Kconfig +++ b/fs/Kconfig @@ -50,7 +50,6 @@ endif # BLOCK config FILE_LOCKING bool "Enable POSIX file locking API" if EMBEDDED default y - select BKL # while lockd still uses it. help This option enables standard file locking support, required for filesystems like NFS and for the flock() system diff --git a/fs/locks.c b/fs/locks.c index 85fd9ce1aba..74c3df99c0e 100644 --- a/fs/locks.c +++ b/fs/locks.c @@ -142,6 +142,7 @@ int lease_break_time = 45; static LIST_HEAD(file_lock_list); static LIST_HEAD(blocked_list); +static DEFINE_SPINLOCK(file_lock_lock); /* * Protects the two list heads above, plus the inode->i_flock list @@ -149,13 +150,13 @@ static LIST_HEAD(blocked_list); */ void lock_flocks(void) { - lock_kernel(); + spin_lock(&file_lock_lock); } EXPORT_SYMBOL_GPL(lock_flocks); void unlock_flocks(void) { - unlock_kernel(); + spin_unlock(&file_lock_lock); } EXPORT_SYMBOL_GPL(unlock_flocks); -- cgit v1.2.3