diff options
Diffstat (limited to 'fs/eventpoll.c')
| -rw-r--r-- | fs/eventpoll.c | 1615 |
1 files changed, 619 insertions, 996 deletions
diff --git a/fs/eventpoll.c b/fs/eventpoll.c index 3ae644e7e860..0b73cd45a06d 100644 --- a/fs/eventpoll.c +++ b/fs/eventpoll.c @@ -1,6 +1,6 @@ /* - * fs/eventpoll.c ( Efficent event polling implementation ) - * Copyright (C) 2001,...,2006 Davide Libenzi + * fs/eventpoll.c (Efficent event polling implementation) + * Copyright (C) 2001,...,2007 Davide Libenzi * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -11,7 +11,6 @@ * */ -#include <linux/module.h> #include <linux/init.h> #include <linux/kernel.h> #include <linux/sched.h> @@ -22,34 +21,31 @@ #include <linux/mm.h> #include <linux/slab.h> #include <linux/poll.h> -#include <linux/smp_lock.h> #include <linux/string.h> #include <linux/list.h> #include <linux/hash.h> #include <linux/spinlock.h> #include <linux/syscalls.h> -#include <linux/rwsem.h> #include <linux/rbtree.h> #include <linux/wait.h> #include <linux/eventpoll.h> #include <linux/mount.h> #include <linux/bitops.h> #include <linux/mutex.h> +#include <linux/anon_inodes.h> #include <asm/uaccess.h> #include <asm/system.h> #include <asm/io.h> #include <asm/mman.h> #include <asm/atomic.h> -#include <asm/semaphore.h> - /* * LOCKING: * There are three level of locking required by epoll : * * 1) epmutex (mutex) - * 2) ep->sem (rw_semaphore) - * 3) ep->lock (rw_lock) + * 2) ep->mtx (mutex) + * 3) ep->lock (spinlock) * * The acquire order is the one listed above, from 1 to 3. * We need a spinlock (ep->lock) because we manipulate objects @@ -59,25 +55,22 @@ * a spinlock. During the event transfer loop (from kernel to * user space) we could end up sleeping due a copy_to_user(), so * we need a lock that will allow us to sleep. This lock is a - * read-write semaphore (ep->sem). It is acquired on read during - * the event transfer loop and in write during epoll_ctl(EPOLL_CTL_DEL) - * and during eventpoll_release_file(). Then we also need a global - * semaphore to serialize eventpoll_release_file() and ep_free(). - * This semaphore is acquired by ep_free() during the epoll file + * mutex (ep->mtx). It is acquired during the event transfer loop, + * during epoll_ctl(EPOLL_CTL_DEL) and during eventpoll_release_file(). + * Then we also need a global mutex to serialize eventpoll_release_file() + * and ep_free(). + * This mutex is acquired by ep_free() during the epoll file * cleanup path and it is also acquired by eventpoll_release_file() * if a file has been pushed inside an epoll set and it is then * close()d without a previous call toepoll_ctl(EPOLL_CTL_DEL). - * It is possible to drop the "ep->sem" and to use the global - * semaphore "epmutex" (together with "ep->lock") to have it working, - * but having "ep->sem" will make the interface more scalable. + * It is possible to drop the "ep->mtx" and to use the global + * mutex "epmutex" (together with "ep->lock") to have it working, + * but having "ep->mtx" will make the interface more scalable. * Events that require holding "epmutex" are very rare, while for - * normal operations the epoll private "ep->sem" will guarantee - * a greater scalability. + * normal operations the epoll private "ep->mtx" will guarantee + * a better scalability. */ - -#define EVENTPOLLFS_MAGIC 0x03111965 /* My birthday should work for this :) */ - #define DEBUG_EPOLL 0 #if DEBUG_EPOLL > 0 @@ -107,6 +100,7 @@ #define EP_MAX_EVENTS (INT_MAX / sizeof(struct epoll_event)) +#define EP_UNACTIVE_PTR ((void *) -1L) struct epoll_filefd { struct file *file; @@ -117,7 +111,7 @@ struct epoll_filefd { * Node that is linked into the "wake_task_list" member of the "struct poll_safewake". * It is used to keep track on all tasks that are currently inside the wake_up() code * to 1) short-circuit the one coming from the same task and same wait queue head - * ( loop ) 2) allow a maximum number of epoll descriptors inclusion nesting + * (loop) 2) allow a maximum number of epoll descriptors inclusion nesting * 3) let go the ones coming from other tasks. */ struct wake_task_node { @@ -136,21 +130,57 @@ struct poll_safewake { }; /* + * Each file descriptor added to the eventpoll interface will + * have an entry of this type linked to the "rbr" RB tree. + */ +struct epitem { + /* RB tree node used to link this structure to the eventpoll RB tree */ + struct rb_node rbn; + + /* List header used to link this structure to the eventpoll ready list */ + struct list_head rdllink; + + /* + * Works together "struct eventpoll"->ovflist in keeping the + * single linked chain of items. + */ + struct epitem *next; + + /* The file descriptor information this item refers to */ + struct epoll_filefd ffd; + + /* Number of active wait queue attached to poll operations */ + int nwait; + + /* List containing poll wait queues */ + struct list_head pwqlist; + + /* The "container" of this item */ + struct eventpoll *ep; + + /* List header used to link this item to the "struct file" items list */ + struct list_head fllink; + + /* The structure that describe the interested events and the source fd */ + struct epoll_event event; +}; + +/* * This structure is stored inside the "private_data" member of the file * structure and rapresent the main data sructure for the eventpoll * interface. */ struct eventpoll { /* Protect the this structure access */ - rwlock_t lock; + spinlock_t lock; /* - * This semaphore is used to ensure that files are not removed - * while epoll is using them. This is read-held during the event - * collection loop and it is write-held during the file cleanup - * path, the epoll file exit code and the ctl operations. + * This mutex is used to ensure that files are not removed + * while epoll is using them. This is held during the event + * collection loop, the file cleanup path, the epoll file exit + * code and the ctl operations. */ - struct rw_semaphore sem; + struct mutex mtx; /* Wait queue used by sys_epoll_wait() */ wait_queue_head_t wq; @@ -161,8 +191,15 @@ struct eventpoll { /* List of ready file descriptors */ struct list_head rdllist; - /* RB-Tree root used to store monitored fd structs */ + /* RB tree root used to store monitored fd structs */ struct rb_root rbr; + + /* + * This is a single linked list that chains all the "struct epitem" that + * happened while transfering ready events to userspace w/out + * holding ->lock. + */ + struct epitem *ovflist; }; /* Wait structure used by the poll hooks */ @@ -183,99 +220,14 @@ struct eppoll_entry { wait_queue_head_t *whead; }; -/* - * Each file descriptor added to the eventpoll interface will - * have an entry of this type linked to the hash. - */ -struct epitem { - /* RB-Tree node used to link this structure to the eventpoll rb-tree */ - struct rb_node rbn; - - /* List header used to link this structure to the eventpoll ready list */ - struct list_head rdllink; - - /* The file descriptor information this item refers to */ - struct epoll_filefd ffd; - - /* Number of active wait queue attached to poll operations */ - int nwait; - - /* List containing poll wait queues */ - struct list_head pwqlist; - - /* The "container" of this item */ - struct eventpoll *ep; - - /* The structure that describe the interested events and the source fd */ - struct epoll_event event; - - /* - * Used to keep track of the usage count of the structure. This avoids - * that the structure will desappear from underneath our processing. - */ - atomic_t usecnt; - - /* List header used to link this item to the "struct file" items list */ - struct list_head fllink; - - /* List header used to link the item to the transfer list */ - struct list_head txlink; - - /* - * This is used during the collection/transfer of events to userspace - * to pin items empty events set. - */ - unsigned int revents; -}; - /* Wrapper struct used by poll queueing */ struct ep_pqueue { poll_table pt; struct epitem *epi; }; - - -static void ep_poll_safewake_init(struct poll_safewake *psw); -static void ep_poll_safewake(struct poll_safewake *psw, wait_queue_head_t *wq); -static int ep_getfd(int *efd, struct inode **einode, struct file **efile, - struct eventpoll *ep); -static int ep_alloc(struct eventpoll **pep); -static void ep_free(struct eventpoll *ep); -static struct epitem *ep_find(struct eventpoll *ep, struct file *file, int fd); -static void ep_use_epitem(struct epitem *epi); -static void ep_release_epitem(struct epitem *epi); -static void ep_ptable_queue_proc(struct file *file, wait_queue_head_t *whead, - poll_table *pt); -static void ep_rbtree_insert(struct eventpoll *ep, struct epitem *epi); -static int ep_insert(struct eventpoll *ep, struct epoll_event *event, - struct file *tfile, int fd); -static int ep_modify(struct eventpoll *ep, struct epitem *epi, - struct epoll_event *event); -static void ep_unregister_pollwait(struct eventpoll *ep, struct epitem *epi); -static int ep_unlink(struct eventpoll *ep, struct epitem *epi); -static int ep_remove(struct eventpoll *ep, struct epitem *epi); -static int ep_poll_callback(wait_queue_t *wait, unsigned mode, int sync, void *key); -static int ep_eventpoll_close(struct inode *inode, struct file *file); -static unsigned int ep_eventpoll_poll(struct file *file, poll_table *wait); -static int ep_collect_ready_items(struct eventpoll *ep, - struct list_head *txlist, int maxevents); -static int ep_send_events(struct eventpoll *ep, struct list_head *txlist, - struct epoll_event __user *events); -static void ep_reinject_items(struct eventpoll *ep, struct list_head *txlist); -static int ep_events_transfer(struct eventpoll *ep, - struct epoll_event __user *events, - int maxevents); -static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events, - int maxevents, long timeout); -static int eventpollfs_delete_dentry(struct dentry *dentry); -static struct inode *ep_eventpoll_inode(void); -static int eventpollfs_get_sb(struct file_system_type *fs_type, - int flags, const char *dev_name, - void *data, struct vfsmount *mnt); - /* - * This semaphore is used to serialize ep_free() and eventpoll_release_file(). + * This mutex is used to serialize ep_free() and eventpoll_release_file(). */ static struct mutex epmutex; @@ -288,39 +240,8 @@ static struct kmem_cache *epi_cache __read_mostly; /* Slab cache used to allocate "struct eppoll_entry" */ static struct kmem_cache *pwq_cache __read_mostly; -/* Virtual fs used to allocate inodes for eventpoll files */ -static struct vfsmount *eventpoll_mnt __read_mostly; - -/* File callbacks that implement the eventpoll file behaviour */ -static const struct file_operations eventpoll_fops = { - .release = ep_eventpoll_close, - .poll = ep_eventpoll_poll -}; - -/* - * This is used to register the virtual file system from where - * eventpoll inodes are allocated. - */ -static struct file_system_type eventpoll_fs_type = { - .name = "eventpollfs", - .get_sb = eventpollfs_get_sb, - .kill_sb = kill_anon_super, -}; - -/* Very basic directory entry operations for the eventpoll virtual file system */ -static struct dentry_operations eventpollfs_dentry_operations = { - .d_delete = eventpollfs_delete_dentry, -}; - - -/* Fast test to see if the file is an evenpoll file */ -static inline int is_file_epoll(struct file *f) -{ - return f->f_op == &eventpoll_fops; -} - -/* Setup the structure that is used as key for the rb-tree */ +/* Setup the structure that is used as key for the RB tree */ static inline void ep_set_ffd(struct epoll_filefd *ffd, struct file *file, int fd) { @@ -328,7 +249,7 @@ static inline void ep_set_ffd(struct epoll_filefd *ffd, ffd->fd = fd; } -/* Compare rb-tree keys */ +/* Compare RB tree keys */ static inline int ep_cmp_ffd(struct epoll_filefd *p1, struct epoll_filefd *p2) { @@ -336,36 +257,25 @@ static inline int ep_cmp_ffd(struct epoll_filefd *p1, (p1->file < p2->file ? -1 : p1->fd - p2->fd)); } -/* Special initialization for the rb-tree node to detect linkage */ +/* Special initialization for the RB tree node to detect linkage */ static inline void ep_rb_initnode(struct rb_node *n) { rb_set_parent(n, n); } -/* Removes a node from the rb-tree and marks it for a fast is-linked check */ +/* Removes a node from the RB tree and marks it for a fast is-linked check */ static inline void ep_rb_erase(struct rb_node *n, struct rb_root *r) { rb_erase(n, r); rb_set_parent(n, n); } -/* Fast check to verify that the item is linked to the main rb-tree */ +/* Fast check to verify that the item is linked to the main RB tree */ static inline int ep_rb_linked(struct rb_node *n) { return rb_parent(n) != n; } -/* - * Remove the item from the list and perform its initialization. - * This is useful for us because we can test if the item is linked - * using "ep_is_linked(p)". - */ -static inline void ep_list_del(struct list_head *p) -{ - list_del(p); - INIT_LIST_HEAD(p); -} - /* Tells us if the item is currently linked */ static inline int ep_is_linked(struct list_head *p) { @@ -385,7 +295,7 @@ static inline struct epitem * ep_item_from_epqueue(poll_table *p) } /* Tells if the epoll_ctl(2) operation needs an event copy from userspace */ -static inline int ep_op_hash_event(int op) +static inline int ep_op_has_event(int op) { return op != EPOLL_CTL_DEL; } @@ -398,7 +308,6 @@ static void ep_poll_safewake_init(struct poll_safewake *psw) spin_lock_init(&psw->lock); } - /* * Perform a safe wake up of the poll wait list. The problem is that * with the new callback'd wake up system, it is possible that the @@ -453,378 +362,195 @@ static void ep_poll_safewake(struct poll_safewake *psw, wait_queue_head_t *wq) spin_unlock_irqrestore(&psw->lock, flags); } - /* - * This is called from eventpoll_release() to unlink files from the eventpoll - * interface. We need to have this facility to cleanup correctly files that are - * closed without being removed from the eventpoll interface. + * This function unregister poll callbacks from the associated file descriptor. + * Since this must be called without holding "ep->lock" the atomic exchange trick + * will protect us from multiple unregister. */ -void eventpoll_release_file(struct file *file) +static void ep_unregister_pollwait(struct eventpoll *ep, struct epitem *epi) { - struct list_head *lsthead = &file->f_ep_links; - struct eventpoll *ep; - struct epitem *epi; + int nwait; + struct list_head *lsthead = &epi->pwqlist; + struct eppoll_entry *pwq; - /* - * We don't want to get "file->f_ep_lock" because it is not - * necessary. It is not necessary because we're in the "struct file" - * cleanup path, and this means that noone is using this file anymore. - * The only hit might come from ep_free() but by holding the semaphore - * will correctly serialize the operation. We do need to acquire - * "ep->sem" after "epmutex" because ep_remove() requires it when called - * from anywhere but ep_free(). - */ - mutex_lock(&epmutex); + /* This is called without locks, so we need the atomic exchange */ + nwait = xchg(&epi->nwait, 0); - while (!list_empty(lsthead)) { - epi = list_entry(lsthead->next, struct epitem, fllink); + if (nwait) { + while (!list_empty(lsthead)) { + pwq = list_first_entry(lsthead, struct eppoll_entry, llink); - ep = epi->ep; - ep_list_del(&epi->fllink); - down_write(&ep->sem); - ep_remove(ep, epi); - up_write(&ep->sem); + list_del_init(&pwq->llink); + remove_wait_queue(pwq->whead, &pwq->wait); + kmem_cache_free(pwq_cache, pwq); + } } - - mutex_unlock(&epmutex); } - /* - * It opens an eventpoll file descriptor by suggesting a storage of "size" - * file descriptors. The size parameter is just an hint about how to size - * data structures. It won't prevent the user to store more than "size" - * file descriptors inside the epoll interface. It is the kernel part of - * the userspace epoll_create(2). + * Removes a "struct epitem" from the eventpoll RB tree and deallocates + * all the associated resources. Must be called with "mtx" held. */ -asmlinkage long sys_epoll_create(int size) +static int ep_remove(struct eventpoll *ep, struct epitem *epi) { - int error, fd = -1; - struct eventpoll *ep; - struct inode *inode; - struct file *file; - - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_create(%d)\n", - current, size)); + unsigned long flags; + struct file *file = epi->ffd.file; /* - * Sanity check on the size parameter, and create the internal data - * structure ( "struct eventpoll" ). + * Removes poll wait queue hooks. We _have_ to do this without holding + * the "ep->lock" otherwise a deadlock might occur. This because of the + * sequence of the lock acquisition. Here we do "ep->lock" then the wait + * queue head lock when unregistering the wait queue. The wakeup callback + * will run by holding the wait queue head lock and will call our callback + * that will try to get "ep->lock". */ - error = -EINVAL; - if (size <= 0 || (error = ep_alloc(&ep)) != 0) - goto eexit_1; + ep_unregister_pollwait(ep, epi); - /* - * Creates all the items needed to setup an eventpoll file. That is, - * a file structure, and inode and a free file descriptor. - */ - error = ep_getfd(&fd, &inode, &file, ep); - if (error) - goto eexit_2; + /* Remove the current item from the list of epoll hooks */ + spin_lock(&file->f_ep_lock); + if (ep_is_linked(&epi->fllink)) + list_del_init(&epi->fllink); + spin_unlock(&file->f_ep_lock); - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_create(%d) = %d\n", - current, size, fd)); + if (ep_rb_linked(&epi->rbn)) + ep_rb_erase(&epi->rbn, &ep->rbr); - return fd; + spin_lock_irqsave(&ep->lock, flags); + if (ep_is_linked(&epi->rdllink)) + list_del_init(&epi->rdllink); + spin_unlock_irqrestore(&ep->lock, flags); -eexit_2: - ep_free(ep); - kfree(ep); -eexit_1: - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_create(%d) = %d\n", - current, size, error)); - return error; -} + /* At this point it is safe to free the eventpoll item */ + kmem_cache_free(epi_cache, epi); + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: ep_remove(%p, %p)\n", + current, ep, file)); -/* - * The following function implements the controller interface for - * the eventpoll file that enables the insertion/removal/change of - * file descriptors inside the interest set. It represents - * the kernel part of the user space epoll_ctl(2). - */ -asmlinkage long -sys_epoll_ctl(int epfd, int op, int fd, struct epoll_event __user *event) + return 0; +} + +static void ep_free(struct eventpoll *ep) { - int error; - struct file *file, *tfile; - struct eventpoll *ep; + struct rb_node *rbp; struct epitem *epi; - struct epoll_event epds; - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_ctl(%d, %d, %d, %p)\n", - current, epfd, op, fd, event)); - - error = -EFAULT; - if (ep_op_hash_event(op) && - copy_from_user(&epds, event, sizeof(struct epoll_event))) - goto eexit_1; - - /* Get the "struct file *" for the eventpoll file */ - error = -EBADF; - file = fget(epfd); - if (!file) - goto eexit_1; - - /* Get the "struct file *" for the target file */ - tfile = fget(fd); - if (!tfile) - goto eexit_2; - - /* The target file descriptor must support poll */ - error = -EPERM; - if (!tfile->f_op || !tfile->f_op->poll) - goto eexit_3; + /* We need to release all tasks waiting for these file */ + if (waitqueue_active(&ep->poll_wait)) + ep_poll_safewake(&psw, &ep->poll_wait); /* - * We have to check that the file structure underneath the file descriptor - * the user passed to us _is_ an eventpoll file. And also we do not permit - * adding an epoll file descriptor inside itself. + * We need to lock this because we could be hit by + * eventpoll_release_file() while we're freeing the "struct eventpoll". + * We do not need to hold "ep->mtx" here because the epoll file + * is on the way to be removed and no one has references to it + * anymore. The only hit might come from eventpoll_release_file() but + * holding "epmutex" is sufficent here. */ - error = -EINVAL; - if (file == tfile || !is_file_epoll(file)) - goto eexit_3; + mutex_lock(&epmutex); /* - * At this point it is safe to assume that the "private_data" contains - * our own data structure. + * Walks through the whole tree by unregistering poll callbacks. */ - ep = file->private_data; - - down_write(&ep->sem); - - /* Try to lookup the file inside our hash table */ - epi = ep_find(ep, tfile, fd); - - error = -EINVAL; - switch (op) { - case EPOLL_CTL_ADD: - if (!epi) { - epds.events |= POLLERR | POLLHUP; + for (rbp = rb_first(&ep->rbr); rbp; rbp = rb_next(rbp)) { + epi = rb_entry(rbp, struct epitem, rbn); - error = ep_insert(ep, &epds, tfile, fd); - } else - error = -EEXIST; - break; - case EPOLL_CTL_DEL: - if (epi) - error = ep_remove(ep, epi); - else - error = -ENOENT; - break; - case EPOLL_CTL_MOD: - if (epi) { - epds.events |= POLLERR | POLLHUP; - error = ep_modify(ep, epi, &epds); - } else - error = -ENOENT; - break; + ep_unregister_pollwait(ep, epi); } /* - * The function ep_find() increments the usage count of the structure - * so, if this is not NULL, we need to release it. + * Walks through the whole tree by freeing each "struct epitem". At this + * point we are sure no poll callbacks will be lingering around, and also by + * holding "epmutex" we can be sure that no file cleanup code will hit + * us during this operation. So we can avoid the lock on "ep->lock". */ - if (epi) - ep_release_epitem(epi); - - up_write(&ep->sem); - -eexit_3: - fput(tfile); -eexit_2: - fput(file); -eexit_1: - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_ctl(%d, %d, %d, %p) = %d\n", - current, epfd, op, fd, event, error)); + while ((rbp = rb_first(&ep->rbr)) != 0) { + epi = rb_entry(rbp, struct epitem, rbn); + ep_remove(ep, epi); + } - return error; + mutex_unlock(&epmutex); + mutex_destroy(&ep->mtx); + kfree(ep); } - -/* - * Implement the event wait interface for the eventpoll file. It is the kernel - * part of the user space epoll_wait(2). - */ -asmlinkage long sys_epoll_wait(int epfd, struct epoll_event __user *events, - int maxevents, int timeout) +static int ep_eventpoll_release(struct inode *inode, struct file *file) { - int error; - struct file *file; - struct eventpoll *ep; - - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_wait(%d, %p, %d, %d)\n", - current, epfd, events, maxevents, timeout)); - - /* The maximum number of event must be greater than zero */ - if (maxevents <= 0 || maxevents > EP_MAX_EVENTS) - return -EINVAL; - - /* Verify that the area passed by the user is writeable */ - if (!access_ok(VERIFY_WRITE, events, maxevents * sizeof(struct epoll_event))) { - error = -EFAULT; - goto eexit_1; - } + struct eventpoll *ep = file->private_data; - /* Get the "struct file *" for the eventpoll file */ - error = -EBADF; - file = fget(epfd); - if (!file) - goto eexit_1; + if (ep) + ep_free(ep); - /* - * We have to check that the file structure underneath the fd - * the user passed to us _is_ an eventpoll file. - */ - error = -EINVAL; - if (!is_file_epoll(file)) - goto eexit_2; + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: close() ep=%p\n", current, ep)); + return 0; +} - /* - * At this point it is safe to assume that the "private_data" contains - * our own data structure. - */ - ep = file->private_data; +static unsigned int ep_eventpoll_poll(struct file *file, poll_table *wait) +{ + unsigned int pollflags = 0; + unsigned long flags; + struct eventpoll *ep = file->private_data; - /* Time to fish for events ... */ - error = ep_poll(ep, events, maxevents, timeout); + /* Insert inside our poll wait queue */ + poll_wait(file, &ep->poll_wait, wait); -eexit_2: - fput(file); -eexit_1: - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_wait(%d, %p, %d, %d) = %d\n", - current, epfd, events, maxevents, timeout, error)); + /* Check our condition */ + spin_lock_irqsave(&ep->lock, flags); + if (!list_empty(&ep->rdllist)) + pollflags = POLLIN | POLLRDNORM; + spin_unlock_irqrestore(&ep->lock, flags); - return error; + return pollflags; } +/* File callbacks that implement the eventpoll file behaviour */ +static const struct file_operations eventpoll_fops = { + .release = ep_eventpoll_release, + .poll = ep_eventpoll_poll +}; -#ifdef TIF_RESTORE_SIGMASK +/* Fast test to see if the file is an evenpoll file */ +static inline int is_file_epoll(struct file *f) +{ + return f->f_op == &eventpoll_fops; +} /* - * Implement the event wait interface for the eventpoll file. It is the kernel - * part of the user space epoll_pwait(2). + * This is called from eventpoll_release() to unlink files from the eventpoll + * interface. We need to have this facility to cleanup correctly files that are + * closed without being removed from the eventpoll interface. */ -asmlinkage long sys_epoll_pwait(int epfd, struct epoll_event __user *events, - int maxevents, int timeout, const sigset_t __user *sigmask, - size_t sigsetsize) +void eventpoll_release_file(struct file *file) { - int error; - sigset_t ksigmask, sigsaved; - - /* - * If the caller wants a certain signal mask to be set during the wait, - * we apply it here. - */ - if (sigmask) { - if (sigsetsize != sizeof(sigset_t)) - return -EINVAL; - if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask))) - return -EFAULT; - sigdelsetmask(&ksigmask, sigmask(SIGKILL) | sigmask(SIGSTOP)); - sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved); - } - - error = sys_epoll_wait(epfd, events, maxevents, timeout); + struct list_head *lsthead = &file->f_ep_links; + struct eventpoll *ep; + struct epitem *epi; /* - * If we changed the signal mask, we need to restore the original one. - * In case we've got a signal while waiting, we do not restore the - * signal mask yet, and we allow do_signal() to deliver the signal on - * the way back to userspace, before the signal mask is restored. + * We don't want to get "file->f_ep_lock" because it is not + * necessary. It is not necessary because we're in the "struct file" + * cleanup path, and this means that noone is using this file anymore. + * So, for example, epoll_ctl() cannot hit here sicne if we reach this + * point, the file counter already went to zero and fget() would fail. + * The only hit might come from ep_free() but by holding the mutex + * will correctly serialize the operation. We do need to acquire + * "ep->mtx" after "epmutex" because ep_remove() requires it when called + * from anywhere but ep_free(). */ - if (sigmask) { - if (error == -EINTR) { - memcpy(¤t->saved_sigmask, &sigsaved, - sizeof(sigsaved)); - set_thread_flag(TIF_RESTORE_SIGMASK); - } else - sigprocmask(SIG_SETMASK, &sigsaved, NULL); - } - - return error; -} - -#endif /* #ifdef TIF_RESTORE_SIGMASK */ - - -/* - * Creates the file descriptor to be used by the epoll interface. - */ -static int ep_getfd(int *efd, struct inode **einode, struct file **efile, - struct eventpoll *ep) -{ - struct qstr this; - char name[32]; - struct dentry *dentry; - struct inode *inode; - struct file *file; - int error, fd; + mutex_lock(&epmutex); - /* Get an ready to use file */ - error = -ENFILE; - file = get_empty_filp(); - if (!file) - goto eexit_1; + while (!list_empty(lsthead)) { + epi = list_first_entry(lsthead, struct epitem, fllink); - /* Allocates an inode from the eventpoll file system */ - inode = ep_eventpoll_inode(); - if (IS_ERR(inode)) { - error = PTR_ERR(inode); - goto eexit_2; + ep = epi->ep; + list_del_init(&epi->fllink); + mutex_lock(&ep->mtx); + ep_remove(ep, epi); + mutex_unlock(&ep->mtx); } - /* Allocates a free descriptor to plug the file onto */ - error = get_unused_fd(); - if (error < 0) - goto eexit_3; - fd = error; - - /* - * Link the inode to a directory entry by creating a unique name - * using the inode number. - */ - error = -ENOMEM; - sprintf(name, "[%lu]", inode->i_ino); - this.name = name; - this.len = strlen(name); - this.hash = inode->i_ino; - dentry = d_alloc(eventpoll_mnt->mnt_sb->s_root, &this); - if (!dentry) - goto eexit_4; - dentry->d_op = &eventpollfs_dentry_operations; - d_add(dentry, inode); - file->f_path.mnt = mntget(eventpoll_mnt); - file->f_path.dentry = dentry; - file->f_mapping = inode->i_mapping; - - file->f_pos = 0; - file->f_flags = O_RDONLY; - file->f_op = &eventpoll_fops; - file->f_mode = FMODE_READ; - file->f_version = 0; - file->private_data = ep; - - /* Install the new setup file into the allocated fd. */ - fd_install(fd, file); - - *efd = fd; - *einode = inode; - *efile = file; - return 0; - -eexit_4: - put_unused_fd(fd); -eexit_3: - iput(inode); -eexit_2: - put_filp(file); -eexit_1: - return error; + mutex_unlock(&epmutex); } - static int ep_alloc(struct eventpoll **pep) { struct eventpoll *ep = kzalloc(sizeof(*ep), GFP_KERNEL); @@ -832,12 +558,13 @@ static int ep_alloc(struct eventpoll **pep) if (!ep) return -ENOMEM; - rwlock_init(&ep->lock); - init_rwsem(&ep->sem); + spin_lock_init(&ep->lock); + mutex_init(&ep->mtx); init_waitqueue_head(&ep->wq); init_waitqueue_head(&ep->poll_wait); INIT_LIST_HEAD(&ep->rdllist); ep->rbr = RB_ROOT; + ep->ovflist = EP_UNACTIVE_PTR; *pep = ep; @@ -846,65 +573,19 @@ static int ep_alloc(struct eventpoll **pep) return 0; } - -static void ep_free(struct eventpoll *ep) -{ - struct rb_node *rbp; - struct epitem *epi; - - /* We need to release all tasks waiting for these file */ - if (waitqueue_active(&ep->poll_wait)) - ep_poll_safewake(&psw, &ep->poll_wait); - - /* - * We need to lock this because we could be hit by - * eventpoll_release_file() while we're freeing the "struct eventpoll". - * We do not need to hold "ep->sem" here because the epoll file - * is on the way to be removed and no one has references to it - * anymore. The only hit might come from eventpoll_release_file() but - * holding "epmutex" is sufficent here. - */ - mutex_lock(&epmutex); - - /* - * Walks through the whole tree by unregistering poll callbacks. - */ - for (rbp = rb_first(&ep->rbr); rbp; rbp = rb_next(rbp)) { - epi = rb_entry(rbp, struct epitem, rbn); - - ep_unregister_pollwait(ep, epi); - } - - /* - * Walks through the whole hash by freeing each "struct epitem". At this - * point we are sure no poll callbacks will be lingering around, and also by - * write-holding "sem" we can be sure that no file cleanup code will hit - * us during this operation. So we can avoid the lock on "ep->lock". - */ - while ((rbp = rb_first(&ep->rbr)) != 0) { - epi = rb_entry(rbp, struct epitem, rbn); - ep_remove(ep, epi); - } - - mutex_unlock(&epmutex); -} - - /* - * Search the file inside the eventpoll hash. It add usage count to - * the returned item, so the caller must call ep_release_epitem() - * after finished using the "struct epitem". + * Search the file inside the eventpoll tree. The RB tree operations + * are protected by the "mtx" mutex, and ep_find() must be called with + * "mtx" held. */ static struct epitem *ep_find(struct eventpoll *ep, struct file *file, int fd) { int kcmp; - unsigned long flags; struct rb_node *rbp; struct epitem *epi, *epir = NULL; struct epoll_filefd ffd; ep_set_ffd(&ffd, file, fd); - read_lock_irqsave(&ep->lock, flags); for (rbp = ep->rbr.rb_node; rbp; ) { epi = rb_entry(rbp, struct epitem, rbn); kcmp = ep_cmp_ffd(&ffd, &epi->ffd); @@ -913,12 +594,10 @@ static struct epitem *ep_find(struct eventpoll *ep, struct file *file, int fd) else if (kcmp < 0) rbp = rbp->rb_left; else { - ep_use_epitem(epi); epir = epi; break; } } - read_unlock_irqrestore(&ep->lock, flags); DNPRINTK(3, (KERN_INFO "[%p] eventpoll: ep_find(%p) -> %p\n", current, file, epir)); @@ -926,30 +605,72 @@ static struct epitem *ep_find(struct eventpoll *ep, struct file *file, int fd) return epir; } - /* - * Increment the usage count of the "struct epitem" making it sure - * that the user will have a valid pointer to reference. + * This is the callback that is passed to the wait queue wakeup + * machanism. It is called by the stored file descriptors when they + * have events to report. */ -static void ep_use_epitem(struct epitem *epi) +static int ep_poll_callback(wait_queue_t *wait, unsigned mode, int sync, void *key) { + int pwake = 0; + unsigned long flags; + struct epitem *epi = ep_item_from_wait(wait); + struct eventpoll *ep = epi->ep; - atomic_inc(&epi->usecnt); -} + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: poll_callback(%p) epi=%p ep=%p\n", + current, epi->ffd.file, epi, ep)); + spin_lock_irqsave(&ep->lock, flags); -/* - * Decrement ( release ) the usage count by signaling that the user - * has finished using the structure. It might lead to freeing the - * structure itself if the count goes to zero. - */ -static void ep_release_epitem(struct epitem *epi) -{ + /* + * If the event mask does not contain any poll(2) event, we consider the + * descriptor to be disabled. This condition is likely the effect of the + * EPOLLONESHOT bit that disables the descriptor when an event is received, + * until the next EPOLL_CTL_MOD will be issued. + */ + if (!(epi->event.events & ~EP_PRIVATE_BITS)) + goto out_unlock; - if (atomic_dec_and_test(&epi->usecnt)) - kmem_cache_free(epi_cache, epi); -} + /* + * If we are trasfering events to userspace, we can hold no locks + * (because we're accessing user memory, and because of linux f_op->poll() + * semantics). All the events that happens during that period of time are + * chained in ep->ovflist and requeued later on. + */ + if (unlikely(ep->ovflist != EP_UNACTIVE_PTR)) { + if (epi->next == EP_UNACTIVE_PTR) { + epi->next = ep->ovflist; + ep->ovflist = epi; + } + goto out_unlock; + } + /* If this file is already in the ready list we exit soon */ + if (ep_is_linked(&epi->rdllink)) + goto is_linked; + + list_add_tail(&epi->rdllink, &ep->rdllist); + +is_linked: + /* + * Wake up ( if active ) both the eventpoll wait list and the ->poll() + * wait list. + */ + if (waitqueue_active(&ep->wq)) + __wake_up_locked(&ep->wq, TASK_UNINTERRUPTIBLE | + TASK_INTERRUPTIBLE); + if (waitqueue_active(&ep->poll_wait)) + pwake++; + +out_unlock: + spin_unlock_irqrestore(&ep->lock, flags); + + /* We have to call this outside the lock */ + if (pwake) + ep_poll_safewake(&psw, &ep->poll_wait); + + return 1; +} /* * This is the callback that is used to add our wait queue to the @@ -974,7 +695,6 @@ static void ep_ptable_queue_proc(struct file *file, wait_queue_head_t *whead, } } - static void ep_rbtree_insert(struct eventpoll *ep, struct epitem *epi) { int kcmp; @@ -994,7 +714,9 @@ static void ep_rbtree_insert(struct eventpoll *ep, struct epitem *epi) rb_insert_color(&epi->rbn, &ep->rbr); } - +/* + * Must be called with "mtx" held. + */ static int ep_insert(struct eventpoll *ep, struct epoll_event *event, struct file *tfile, int fd) { @@ -1005,19 +727,18 @@ static int ep_insert(struct eventpoll *ep, struct epoll_event *event, error = -ENOMEM; if (!(epi = kmem_cache_alloc(epi_cache, GFP_KERNEL))) - goto eexit_1; + goto error_return; /* Item initialization follow here ... */ ep_rb_initnode(&epi->rbn); INIT_LIST_HEAD(&epi->rdllink); INIT_LIST_HEAD(&epi->fllink); - INIT_LIST_HEAD(&epi->txlink); INIT_LIST_HEAD(&epi->pwqlist); epi->ep = ep; ep_set_ffd(&epi->ffd, tfile, fd); epi->event = *event; - atomic_set(&epi->usecnt, 1); epi->nwait = 0; + epi->next = EP_UNACTIVE_PTR; /* Initialize the poll table using the queue callback */ epq.epi = epi; @@ -1026,7 +747,9 @@ static int ep_insert(struct eventpoll *ep, struct epoll_event *event, /* * Attach the item to the poll hooks and get current event bits. * We can safely use the file* here because its usage count has - * been increased by the caller of this function. + * been increased by the caller of this function. Note that after + * this operation completes, the poll callback can start hitting + * the new item. */ revents = tfile->f_op->poll(tfile, &epq.pt); @@ -1036,19 +759,22 @@ static int ep_insert(struct eventpoll *ep, struct epoll_event *event, * high memory pressure. */ if (epi->nwait < 0) - goto eexit_2; + goto error_unregister; /* Add the current item to the list of active epoll hook for this file */ spin_lock(&tfile->f_ep_lock); list_add_tail(&epi->fllink, &tfile->f_ep_links); spin_unlock(&tfile->f_ep_lock); - /* We have to drop the new item inside our item list to keep track of it */ - write_lock_irqsave(&ep->lock, flags); - - /* Add the current item to the rb-tree */ + /* + * Add the current item to the RB tree. All RB tree operations are + * protected by "mtx", and ep_insert() is called with "mtx" held. + */ ep_rbtree_insert(ep, epi); + /* We have to drop the new item inside our item list to keep track of it */ + spin_lock_irqsave(&ep->lock, flags); + /* If the file is already "ready" we drop it inside the ready list */ if ((revents & event->events) && !ep_is_linked(&epi->rdllink)) { list_add_tail(&epi->rdllink, &ep->rdllist); @@ -1060,7 +786,7 @@ static int ep_insert(struct eventpoll *ep, struct epoll_event *event, pwake++; } - write_unlock_irqrestore(&ep->lock, flags); + spin_unlock_irqrestore(&ep->lock, flags); /* We have to call this outside the lock */ if (pwake) @@ -1071,27 +797,28 @@ static int ep_insert(struct eventpoll *ep, struct epoll_event *event, return 0; -eexit_2: +error_unregister: ep_unregister_pollwait(ep, epi); /* * We need to do this because an event could have been arrived on some - * allocated wait queue. + * allocated wait queue. Note that we don't care about the ep->ovflist + * list, since that is used/cleaned only inside a section bound by "mtx". + * And ep_insert() is called with "mtx" held. */ - write_lock_irqsave(&ep->lock, flags); + spin_lock_irqsave(&ep->lock, flags); if (ep_is_linked(&epi->rdllink)) - ep_list_del(&epi->rdllink); - write_unlock_irqrestore(&ep->lock, flags); + list_del_init(&epi->rdllink); + spin_unlock_irqrestore(&ep->lock, flags); kmem_cache_free(epi_cache, epi); -eexit_1: +error_return: return error; } - /* * Modify the interest event mask by dropping an event if the new mask - * has a match in the current file status. + * has a match in the current file status. Must be called with "mtx" held. */ static int ep_modify(struct eventpoll *ep, struct epitem *epi, struct epoll_event *event) { @@ -1113,36 +840,28 @@ static int ep_modify(struct eventpoll *ep, struct epitem *epi, struct epoll_even */ revents = epi->ffd.file->f_op->poll(epi->ffd.file, NULL); - write_lock_irqsave(&ep->lock, flags); + spin_lock_irqsave(&ep->lock, flags); /* Copy the data member from inside the lock */ epi->event.data = event->data; /* - * If the item is not linked to the hash it means that it's on its - * way toward the removal. Do nothing in this case. + * If the item is "hot" and it is not registered inside the ready + * list, push it inside. */ - if (ep_rb_linked(&epi->rbn)) { - /* - * If the item is "hot" and it is not registered inside the ready - * list, push it inside. If the item is not "hot" and it is currently - * registered inside the ready list, unlink it. - */ - if (revents & event->events) { - if (!ep_is_linked(&epi->rdllink)) { - list_add_tail(&epi->rdllink, &ep->rdllist); - - /* Notify waiting tasks that events are available */ - if (waitqueue_active(&ep->wq)) - __wake_up_locked(&ep->wq, TASK_UNINTERRUPTIBLE | - TASK_INTERRUPTIBLE); - if (waitqueue_active(&ep->poll_wait)) - pwake++; - } + if (revents & event->events) { + if (!ep_is_linked(&epi->rdllink)) { + list_add_tail(&epi->rdllink, &ep->rdllist); + + /* Notify waiting tasks that events are available */ + if (waitqueue_active(&ep->wq)) + __wake_up_locked(&ep->wq, TASK_UNINTERRUPTIBLE | + TASK_INTERRUPTIBLE); + if (waitqueue_active(&ep->poll_wait)) + pwake++; } } - - write_unlock_irqrestore(&ep->lock, flags); + spin_unlock_irqrestore(&ep->lock, flags); /* We have to call this outside the lock */ if (pwake) @@ -1151,350 +870,113 @@ static int ep_modify(struct eventpoll *ep, struct epitem *epi, struct epoll_even return 0; } - -/* - * This function unregister poll callbacks from the associated file descriptor. - * Since this must be called without holding "ep->lock" the atomic exchange trick - * will protect us from multiple unregister. - */ -static void ep_unregister_pollwait(struct eventpoll *ep, struct epitem *epi) -{ - int nwait; - struct list_head *lsthead = &epi->pwqlist; - struct eppoll_entry *pwq; - - /* This is called without locks, so we need the atomic exchange */ - nwait = xchg(&epi->nwait, 0); - - if (nwait) { - while (!list_empty(lsthead)) { - pwq = list_entry(lsthead->next, struct eppoll_entry, llink); - - ep_list_del(&pwq->llink); - remove_wait_queue(pwq->whead, &pwq->wait); - kmem_cache_free(pwq_cache, pwq); - } - } -} - - -/* - * Unlink the "struct epitem" from all places it might have been hooked up. - * This function must be called with write IRQ lock on "ep->lock". - */ -static int ep_unlink(struct eventpoll *ep, struct epitem *epi) +static int ep_send_events(struct eventpoll *ep, struct epoll_event __user *events, + int maxevents) { - int error; - - /* - * It can happen that this one is called for an item already unlinked. - * The check protect us from doing a double unlink ( crash ). - */ - error = -ENOENT; - if (!ep_rb_linked(&epi->rbn)) - goto eexit_1; - - /* - * Clear the event mask for the unlinked item. This will avoid item - * notifications to be sent after the unlink operation from inside - * the kernel->userspace event transfer loop. - */ - epi->event.events = 0; - - /* - * At this point is safe to do the job, unlink the item from our rb-tree. - * This operation togheter with the above check closes the door to - * double unlinks. - */ - ep_rb_erase(&epi->rbn, &ep->rbr); - - /* - * If the item we are going to remove is inside the ready file descriptors - * we want to remove it from this list to avoid stale events. - */ - if (ep_is_linked(&epi->rdllink)) - ep_list_del(&epi->rdllink); - - error = 0; -eexit_1: - - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: ep_unlink(%p, %p) = %d\n", - current, ep, epi->ffd.file, error)); - - return error; -} - - -/* - * Removes a "struct epitem" from the eventpoll hash and deallocates - * all the associated resources. - */ -static int ep_remove(struct eventpoll *ep, struct epitem *epi) -{ - int error; - unsigned long flags; - struct file *file = epi->ffd.file; - - /* - * Removes poll wait queue hooks. We _have_ to do this without holding - * the "ep->lock" otherwise a deadlock might occur. This because of the - * sequence of the lock acquisition. Here we do "ep->lock" then the wait - * queue head lock when unregistering the wait queue. The wakeup callback - * will run by holding the wait queue head lock and will call our callback - * that will try to get "ep->lock". - */ - ep_unregister_pollwait(ep, epi); - - /* Remove the current item from the list of epoll hooks */ - spin_lock(&file->f_ep_lock); - if (ep_is_linked(&epi->fllink)) - ep_list_del(&epi->fllink); - spin_unlock(&file->f_ep_lock); - - /* We need to acquire the write IRQ lock before calling ep_unlink() */ - write_lock_irqsave(&ep->lock, flags); - - /* Really unlink the item from the hash */ - error = ep_unlink(ep, epi); - - write_unlock_irqrestore(&ep->lock, flags); - - if (error) - goto eexit_1; - - /* At this point it is safe to free the eventpoll item */ - ep_release_epitem(epi); - - error = 0; -eexit_1: - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: ep_remove(%p, %p) = %d\n", - current, ep, file, error)); - - return error; -} - - -/* - * This is the callback that is passed to the wait queue wakeup - * machanism. It is called by the stored file descriptors when they - * have events to report. - */ -static int ep_poll_callback(wait_queue_t *wait, unsigned mode, int sync, void *key) -{ - int pwake = 0; + int eventcnt, error = -EFAULT, pwake = 0; + unsigned int revents; unsigned long flags; - struct epitem *epi = ep_item_from_wait(wait); - struct eventpoll *ep = epi->ep; - - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: poll_callback(%p) epi=%p ep=%p\n", - current, epi->ffd.file, epi, ep)); + struct epitem *epi, *nepi; + struct list_head txlist; - write_lock_irqsave(&ep->lock, flags); + INIT_LIST_HEAD(&txlist); /* - * If the event mask does not contain any poll(2) event, we consider the - * descriptor to be disabled. This condition is likely the effect of the - * EPOLLONESHOT bit that disables the descriptor when an event is received, - * until the next EPOLL_CTL_MOD will be issued. + * We need to lock this because we could be hit by + * eventpoll_release_file() and epoll_ctl(EPOLL_CTL_DEL). */ - if (!(epi->event.events & ~EP_PRIVATE_BITS)) - goto is_disabled; + mutex_lock(&ep->mtx); - /* If this file is already in the ready list we exit soon */ - if (ep_is_linked(&epi->rdllink)) - goto is_linked; - - list_add_tail(&epi->rdllink, &ep->rdllist); - -is_linked: /* - * Wake up ( if active ) both the eventpoll wait list and the ->poll() - * wait list. + * Steal the ready list, and re-init the original one to the + * empty list. Also, set ep->ovflist to NULL so that events + * happening while looping w/out locks, are not lost. We cannot + * have the poll callback to queue directly on ep->rdllist, + * because we are doing it in the loop below, in a lockless way. */ - if (waitqueue_active(&ep->wq)) - __wake_up_locked(&ep->wq, TASK_UNINTERRUPTIBLE | - TASK_INTERRUPTIBLE); - if (waitqueue_active(&ep->poll_wait)) - pwake++; - -is_disabled: - write_unlock_irqrestore(&ep->lock, flags); - - /* We have to call this outside the lock */ - if (pwake) - ep_poll_safewake(&psw, &ep->poll_wait); - - return 1; -} - - -static int ep_eventpoll_close(struct inode *inode, struct file *file) -{ - struct eventpoll *ep = file->private_data; - - if (ep) { - ep_free(ep); - kfree(ep); - } - - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: close() ep=%p\n", current, ep)); - return 0; -} - - -static unsigned int ep_eventpoll_poll(struct file *file, poll_table *wait) -{ - unsigned int pollflags = 0; - unsigned long flags; - struct eventpoll *ep = file->private_data; - - /* Insert inside our poll wait queue */ - poll_wait(file, &ep->poll_wait, wait); - - /* Check our condition */ - read_lock_irqsave(&ep->lock, flags); - if (!list_empty(&ep->rdllist)) - pollflags = POLLIN | POLLRDNORM; - read_unlock_irqrestore(&ep->lock, flags); - - return pollflags; -} - - -/* - * Since we have to release the lock during the __copy_to_user() operation and - * during the f_op->poll() call, we try to collect the maximum number of items - * by reducing the irqlock/irqunlock switching rate. - */ -static int ep_collect_ready_items(struct eventpoll *ep, struct list_head *txlist, int maxevents) -{ - int nepi; - unsigned long flags; - struct list_head *lsthead = &ep->rdllist, *lnk; - struct epitem *epi; - - write_lock_irqsave(&ep->lock, flags); - - for (nepi = 0, lnk = lsthead->next; lnk != lsthead && nepi < maxevents;) { - epi = list_entry(lnk, struct epitem, rdllink); - - lnk = lnk->next; - - /* If this file is already in the ready list we exit soon */ - if (!ep_is_linked(&epi->txlink)) { - /* - * This is initialized in this way so that the default - * behaviour of the reinjecting code will be to push back - * the item inside the ready list. - */ - epi->revents = epi->event.events; - - /* Link the ready item into the transfer list */ - list_add(&epi->txlink, txlist); - nepi++; - - /* - * Unlink the item from the ready list. - */ - ep_list_del(&epi->rdllink); - } - } - - write_unlock_irqrestore(&ep->lock, flags); - - return nepi; -} - - -/* - * This function is called without holding the "ep->lock" since the call to - * __copy_to_user() might sleep, and also f_op->poll() might reenable the IRQ - * because of the way poll() is traditionally implemented in Linux. - */ -static int ep_send_events(struct eventpoll *ep, struct list_head *txlist, - struct epoll_event __user *events) -{ - int eventcnt = 0; - unsigned int revents; - struct list_head *lnk; - struct epitem *epi; + spin_lock_irqsave(&ep->lock, flags); + list_splice(&ep->rdllist, &txlist); + INIT_LIST_HEAD(&ep->rdllist); + ep->ovflist = NULL; + spin_unlock_irqrestore(&ep->lock, flags); /* * We can loop without lock because this is a task private list. - * The test done during the collection loop will guarantee us that - * another task will not try to collect this file. Also, items - * cannot vanish during the loop because we are holding "sem". + * We just splice'd out the ep->rdllist in ep_collect_ready_items(). + * Items cannot vanish during the loop because we are holding "mtx". */ - list_for_each(lnk, txlist) { - epi = list_entry(lnk, struct epitem, txlink); + for (eventcnt = 0; !list_empty(&txlist) && eventcnt < maxevents;) { + epi = list_first_entry(&txlist, struct epitem, rdllink); + + list_del_init(&epi->rdllink); /* * Get the ready file event set. We can safely use the file - * because we are holding the "sem" in read and this will - * guarantee that both the file and the item will not vanish. + * because we are holding the "mtx" and this will guarantee + * that both the file and the item will not vanish. */ revents = epi->ffd.file->f_op->poll(epi->ffd.file, NULL); + revents &= epi->event.events; /* - * Set the return event set for the current file descriptor. - * Note that only the task task was successfully able to link - * the item to its "txlist" will write this field. + * Is the event mask intersect the caller-requested one, + * deliver the event to userspace. Again, we are holding + * "mtx", so no operations coming from userspace can change + * the item. */ - epi->revents = revents & epi->event.events; - - if (epi->revents) { - if (__put_user(epi->revents, + if (revents) { + if (__put_user(revents, &events[eventcnt].events) || __put_user(epi->event.data, &events[eventcnt].data)) - return -EFAULT; + goto errxit; if (epi->event.events & EPOLLONESHOT) epi->event.events &= EP_PRIVATE_BITS; eventcnt++; } + /* + * At this point, noone can insert into ep->rdllist besides + * us. The epoll_ctl() callers are locked out by us holding + * "mtx" and the poll callback will queue them in ep->ovflist. + */ + if (!(epi->event.events & EPOLLET) && + (revents & epi->event.events)) + list_add_tail(&epi->rdllink, &ep->rdllist); } - return eventcnt; -} - - -/* - * Walk through the transfer list we collected with ep_collect_ready_items() - * and, if 1) the item is still "alive" 2) its event set is not empty 3) it's - * not already linked, links it to the ready list. Same as above, we are holding - * "sem" so items cannot vanish underneath our nose. - */ -static void ep_reinject_items(struct eventpoll *ep, struct list_head *txlist) -{ - int ricnt = 0, pwake = 0; - unsigned long flags; - struct epitem *epi; - - write_lock_irqsave(&ep->lock, flags); - - while (!list_empty(txlist)) { - epi = list_entry(txlist->next, struct epitem, txlink); + error = 0; - /* Unlink the current item from the transfer list */ - ep_list_del(&epi->txlink); +errxit: - /* - * If the item is no more linked to the interest set, we don't - * have to push it inside the ready list because the following - * ep_release_epitem() is going to drop it. Also, if the current - * item is set to have an Edge Triggered behaviour, we don't have - * to push it back either. - */ - if (ep_rb_linked(&epi->rbn) && !(epi->event.events & EPOLLET) && - (epi->revents & epi->event.events) && !ep_is_linked(&epi->rdllink)) { + spin_lock_irqsave(&ep->lock, flags); + /* + * During the time we spent in the loop above, some other events + * might have been queued by the poll callback. We re-insert them + * here (in case they are not already queued, or they're one-shot). + */ + for (nepi = ep->ovflist; (epi = nepi) != NULL; + nepi = epi->next, epi->next = EP_UNACTIVE_PTR) { + if (!ep_is_linked(&epi->rdllink) && + (epi->event.events & ~EP_PRIVATE_BITS)) list_add_tail(&epi->rdllink, &ep->rdllist); - ricnt++; - } } + /* + * We need to set back ep->ovflist to EP_UNACTIVE_PTR, so that after + * releasing the lock, events will be queued in the normal way inside + * ep->rdllist. + */ + ep->ovflist = EP_UNACTIVE_PTR; - if (ricnt) { + /* + * In case of error in the event-send loop, or in case the number of + * ready events exceeds the userspace limit, we need to splice the + * "txlist" back inside ep->rdllist. + */ + list_splice(&txlist, &ep->rdllist); + + if (!list_empty(&ep->rdllist)) { /* - * Wake up ( if active ) both the eventpoll wait list and the ->poll() - * wait list. + * Wake up (if active) both the eventpoll wait list and the ->poll() + * wait list (delayed after we release the lock). */ if (waitqueue_active(&ep->wq)) __wake_up_locked(&ep->wq, TASK_UNINTERRUPTIBLE | @@ -1502,47 +984,17 @@ static void ep_reinject_items(struct eventpoll *ep, struct list_head *txlist) if (waitqueue_active(&ep->poll_wait)) pwake++; } + spin_unlock_irqrestore(&ep->lock, flags); - write_unlock_irqrestore(&ep->lock, flags); + mutex_unlock(&ep->mtx); /* We have to call this outside the lock */ if (pwake) ep_poll_safewake(&psw, &ep->poll_wait); -} - -/* - * Perform the transfer of events to user space. - */ -static int ep_events_transfer(struct eventpoll *ep, - struct epoll_event __user *events, int maxevents) -{ - int eventcnt = 0; - struct list_head txlist; - - INIT_LIST_HEAD(&txlist); - - /* - * We need to lock this because we could be hit by - * eventpoll_release_file() and epoll_ctl(EPOLL_CTL_DEL). - */ - down_read(&ep->sem); - - /* Collect/extract ready items */ - if (ep_collect_ready_items(ep, &txlist, maxevents) > 0) { - /* Build result set in userspace */ - eventcnt = ep_send_events(ep, &txlist, events); - - /* Reinject ready items into the ready list */ - ep_reinject_items(ep, &txlist); - } - - up_read(&ep->sem); - - return eventcnt; + return eventcnt == 0 ? error: eventcnt; } - static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events, int maxevents, long timeout) { @@ -1560,7 +1012,7 @@ static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events, MAX_SCHEDULE_TIMEOUT : (timeout * HZ + 999) / 1000; retry: - write_lock_irqsave(&ep->lock, flags); + spin_lock_irqsave(&ep->lock, flags); res = 0; if (list_empty(&ep->rdllist)) { @@ -1570,6 +1022,7 @@ retry: * ep_poll_callback() when events will become available. */ init_waitqueue_entry(&wait, current); + wait.flags |= WQ_FLAG_EXCLUSIVE; __add_wait_queue(&ep->wq, &wait); for (;;) { @@ -1586,9 +1039,9 @@ retry: break; } - write_unlock_irqrestore(&ep->lock, flags); + spin_unlock_irqrestore(&ep->lock, flags); jtimeout = schedule_timeout(jtimeout); - write_lock_irqsave(&ep->lock, flags); + spin_lock_irqsave(&ep->lock, flags); } __remove_wait_queue(&ep->wq, &wait); @@ -1598,7 +1051,7 @@ retry: /* Is it worth to try to dig for events ? */ eavail = !list_empty(&ep->rdllist); - write_unlock_irqrestore(&ep->lock, flags); + spin_unlock_irqrestore(&ep->lock, flags); /* * Try to transfer events to user space. In case we get 0 events and @@ -1606,61 +1059,263 @@ retry: * more luck. */ if (!res && eavail && - !(res = ep_events_transfer(ep, events, maxevents)) && jtimeout) + !(res = ep_send_events(ep, events, maxevents)) && jtimeout) goto retry; return res; } - -static int eventpollfs_delete_dentry(struct dentry *dentry) +/* + * It opens an eventpoll file descriptor. The "size" parameter is there + * for historical reasons, when epoll was using an hash instead of an + * RB tree. With the current implementation, the "size" parameter is ignored + * (besides sanity checks). + */ +asmlinkage long sys_epoll_create(int size) { + int error, fd = -1; + struct eventpoll *ep; + struct inode *inode; + struct file *file; - return 1; + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_create(%d)\n", + current, size)); + + /* + * Sanity check on the size parameter, and create the internal data + * structure ( "struct eventpoll" ). + */ + error = -EINVAL; + if (size <= 0 || (error = ep_alloc(&ep)) != 0) + goto error_return; + + /* + * Creates all the items needed to setup an eventpoll file. That is, + * a file structure, and inode and a free file descriptor. + */ + error = anon_inode_getfd(&fd, &inode, &file, "[eventpoll]", + &eventpoll_fops, ep); + if (error) + goto error_free; + + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_create(%d) = %d\n", + current, size, fd)); + + return fd; + +error_free: + ep_free(ep); +error_return: + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_create(%d) = %d\n", + current, size, error)); + return error; } +/* + * The following function implements the controller interface for + * the eventpoll file that enables the insertion/removal/change of + * file descriptors inside the interest set. + */ +asmlinkage long sys_epoll_ctl(int epfd, int op, int fd, + struct epoll_event __user *event) +{ + int error; + struct file *file, *tfile; + struct eventpoll *ep; + struct epitem *epi; + struct epoll_event epds; + + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_ctl(%d, %d, %d, %p)\n", + current, epfd, op, fd, event)); -static struct inode *ep_eventpoll_inode(void) + error = -EFAULT; + if (ep_op_has_event(op) && + copy_from_user(&epds, event, sizeof(struct epoll_event))) + goto error_return; + + /* Get the "struct file *" for the eventpoll file */ + error = -EBADF; + file = fget(epfd); + if (!file) + goto error_return; + + /* Get the "struct file *" for the target file */ + tfile = fget(fd); + if (!tfile) + goto error_fput; + + /* The target file descriptor must support poll */ + error = -EPERM; + if (!tfile->f_op || !tfile->f_op->poll) + goto error_tgt_fput; + + /* + * We have to check that the file structure underneath the file descriptor + * the user passed to us _is_ an eventpoll file. And also we do not permit + * adding an epoll file descriptor inside itself. + */ + error = -EINVAL; + if (file == tfile || !is_file_epoll(file)) + goto error_tgt_fput; + + /* + * At this point it is safe to assume that the "private_data" contains + * our own data structure. + */ + ep = file->private_data; + + mutex_lock(&ep->mtx); + + /* + * Try to lookup the file inside our RB tree, Since we grabbed "mtx" + * above, we can be sure to be able to use the item looked up by + * ep_find() till we release the mutex. + */ + epi = ep_find(ep, tfile, fd); + + error = -EINVAL; + switch (op) { + case EPOLL_CTL_ADD: + if (!epi) { + epds.events |= POLLERR | POLLHUP; + + error = ep_insert(ep, &epds, tfile, fd); + } else + error = -EEXIST; + break; + case EPOLL_CTL_DEL: + if (epi) + error = ep_remove(ep, epi); + else + error = -ENOENT; + break; + case EPOLL_CTL_MOD: + if (epi) { + epds.events |= POLLERR | POLLHUP; + error = ep_modify(ep, epi, &epds); + } else + error = -ENOENT; + break; + } + mutex_unlock(&ep->mtx); + +error_tgt_fput: + fput(tfile); +error_fput: + fput(file); +error_return: + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_ctl(%d, %d, %d, %p) = %d\n", + current, epfd, op, fd, event, error)); + + return error; +} + +/* + * Implement the event wait interface for the eventpoll file. It is the kernel + * part of the user space epoll_wait(2). + */ +asmlinkage long sys_epoll_wait(int epfd, struct epoll_event __user *events, + int maxevents, int timeout) { - int error = -ENOMEM; - struct inode *inode = new_inode(eventpoll_mnt->mnt_sb); + int error; + struct file *file; + struct eventpoll *ep; + + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_wait(%d, %p, %d, %d)\n", + current, epfd, events, maxevents, timeout)); + + /* The maximum number of event must be greater than zero */ + if (maxevents <= 0 || maxevents > EP_MAX_EVENTS) + return -EINVAL; - if (!inode) - goto eexit_1; + /* Verify that the area passed by the user is writeable */ + if (!access_ok(VERIFY_WRITE, events, maxevents * sizeof(struct epoll_event))) { + error = -EFAULT; + goto error_return; + } + + /* Get the "struct file *" for the eventpoll file */ + error = -EBADF; + file = fget(epfd); + if (!file) + goto error_return; - inode->i_fop = &eventpoll_fops; + /* + * We have to check that the file structure underneath the fd + * the user passed to us _is_ an eventpoll file. + */ + error = -EINVAL; + if (!is_file_epoll(file)) + goto error_fput; /* - * Mark the inode dirty from the very beginning, - * that way it will never be moved to the dirty - * list because mark_inode_dirty() will think - * that it already _is_ on the dirty list. + * At this point it is safe to assume that the "private_data" contains + * our own data structure. */ - inode->i_state = I_DIRTY; - inode->i_mode = S_IRUSR | S_IWUSR; - inode->i_uid = current->fsuid; - inode->i_gid = current->fsgid; - inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; - return inode; - -eexit_1: - return ERR_PTR(error); + ep = file->private_data; + + /* Time to fish for events ... */ + error = ep_poll(ep, events, maxevents, timeout); + +error_fput: + fput(file); +error_return: + DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_wait(%d, %p, %d, %d) = %d\n", + current, epfd, events, maxevents, timeout, error)); + + return error; } +#ifdef TIF_RESTORE_SIGMASK -static int -eventpollfs_get_sb(struct file_system_type *fs_type, int flags, - const char *dev_name, void *data, struct vfsmount *mnt) +/* + * Implement the event wait interface for the eventpoll file. It is the kernel + * part of the user space epoll_pwait(2). + */ +asmlinkage long sys_epoll_pwait(int epfd, struct epoll_event __user *events, + int maxevents, int timeout, const sigset_t __user *sigmask, + size_t sigsetsize) { - return get_sb_pseudo(fs_type, "eventpoll:", NULL, EVENTPOLLFS_MAGIC, - mnt); + int error; + sigset_t ksigmask, sigsaved; + + /* + * If the caller wants a certain signal mask to be set during the wait, + * we apply it here. + */ + if (sigmask) { + if (sigsetsize != sizeof(sigset_t)) + return -EINVAL; + if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask))) + return -EFAULT; + sigdelsetmask(&ksigmask, sigmask(SIGKILL) | sigmask(SIGSTOP)); + sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved); + } + + error = sys_epoll_wait(epfd, events, maxevents, timeout); + + /* + * If we changed the signal mask, we need to restore the original one. + * In case we've got a signal while waiting, we do not restore the + * signal mask yet, and we allow do_signal() to deliver the signal on + * the way back to userspace, before the signal mask is restored. + */ + if (sigmask) { + if (error == -EINTR) { + memcpy(¤t->saved_sigmask, &sigsaved, + sizeof(sigsaved)); + set_thread_flag(TIF_RESTORE_SIGMASK); + } else + sigprocmask(SIG_SETMASK, &sigsaved, NULL); + } + + return error; } +#endif /* #ifdef TIF_RESTORE_SIGMASK */ static int __init eventpoll_init(void) { - int error; - mutex_init(&epmutex); /* Initialize the structure used to perform safe poll wait head wake ups */ @@ -1676,39 +1331,7 @@ static int __init eventpoll_init(void) sizeof(struct eppoll_entry), 0, EPI_SLAB_DEBUG|SLAB_PANIC, NULL, NULL); - /* - * Register the virtual file system that will be the source of inodes - * for the eventpoll files - */ - error = register_filesystem(&eventpoll_fs_type); - if (error) - goto epanic; - - /* Mount the above commented virtual file system */ - eventpoll_mnt = kern_mount(&eventpoll_fs_type); - error = PTR_ERR(eventpoll_mnt); - if (IS_ERR(eventpoll_mnt)) - goto epanic; - - DNPRINTK(3, (KERN_INFO "[%p] eventpoll: successfully initialized.\n", - current)); return 0; - -epanic: - panic("eventpoll_init() failed\n"); } +fs_initcall(eventpoll_init); - -static void __exit eventpoll_exit(void) -{ - /* Undo all operations done inside eventpoll_init() */ - unregister_filesystem(&eventpoll_fs_type); - mntput(eventpoll_mnt); - kmem_cache_destroy(pwq_cache); - kmem_cache_destroy(epi_cache); -} - -module_init(eventpoll_init); -module_exit(eventpoll_exit); - -MODULE_LICENSE("GPL"); |