#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/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>
*/
-#define EVENTPOLLFS_MAGIC 0x03111965 /* My birthday should work for this :) */
-
#define DEBUG_EPOLL 0
#if DEBUG_EPOLL > 0
/* Maximum msec timeout value storeable in a long int */
#define EP_MAX_MSTIMEO min(1000ULL * MAX_SCHEDULE_TIMEOUT / HZ, (LONG_MAX - 999ULL) / HZ)
+#define EP_MAX_EVENTS (INT_MAX / sizeof(struct epoll_event))
+
struct epoll_filefd {
struct file *file;
/*
* Each file descriptor added to the eventpoll interface will
- * have an entry of this type linked to the hash.
+ * 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 */
/* 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 */
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 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);
+ struct epoll_event __user *events, int maxevents);
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().
static struct poll_safewake psw;
/* Slab cache used to allocate "struct epitem" */
-static kmem_cache_t *epi_cache __read_mostly;
+static struct kmem_cache *epi_cache __read_mostly;
/* Slab cache used to allocate "struct eppoll_entry" */
-static kmem_cache_t *pwq_cache __read_mostly;
-
-/* Virtual fs used to allocate inodes for eventpoll files */
-static struct vfsmount *eventpoll_mnt __read_mostly;
+static struct kmem_cache *pwq_cache __read_mostly;
/* File callbacks that implement the eventpoll file behaviour */
static const struct file_operations eventpoll_fops = {
.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 */
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)
{
}
/* 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;
}
mutex_lock(&epmutex);
while (!list_empty(lsthead)) {
- epi = list_entry(lsthead->next, struct epitem, fllink);
+ epi = list_first_entry(lsthead, struct epitem, fllink);
ep = epi->ep;
- ep_list_del(&epi->fllink);
+ list_del_init(&epi->fllink);
down_write(&ep->sem);
ep_remove(ep, epi);
up_write(&ep->sem);
*/
asmlinkage long sys_epoll_create(int size)
{
- int error, fd;
+ int error, fd = -1;
struct eventpoll *ep;
struct inode *inode;
struct file *file;
* 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);
+ error = anon_inode_getfd(&fd, &inode, &file, "[eventpoll]",
+ &eventpoll_fops, ep);
if (error)
goto eexit_2;
current, epfd, op, fd, event));
error = -EFAULT;
- if (ep_op_hash_event(op) &&
+ if (ep_op_has_event(op) &&
copy_from_user(&epds, event, sizeof(struct epoll_event)))
goto eexit_1;
down_write(&ep->sem);
- /* Try to lookup the file inside our hash table */
+ /* Try to lookup the file inside our RB tree */
epi = ep_find(ep, tfile, fd);
error = -EINVAL;
return error;
}
-#define MAX_EVENTS (INT_MAX / sizeof(struct epoll_event))
/*
* Implement the event wait interface for the eventpoll file. It is the kernel
current, epfd, events, maxevents, timeout));
/* The maximum number of event must be greater than zero */
- if (maxevents <= 0 || maxevents > MAX_EVENTS)
+ if (maxevents <= 0 || maxevents > EP_MAX_EVENTS)
return -EINVAL;
/* Verify that the area passed by the user is writeable */
}
+#ifdef TIF_RESTORE_SIGMASK
+
/*
- * Creates the file descriptor to be used by the epoll interface.
+ * Implement the event wait interface for the eventpoll file. It is the kernel
+ * part of the user space epoll_pwait(2).
*/
-static int ep_getfd(int *efd, struct inode **einode, struct file **efile,
- struct eventpoll *ep)
+asmlinkage long sys_epoll_pwait(int epfd, struct epoll_event __user *events,
+ int maxevents, int timeout, const sigset_t __user *sigmask,
+ size_t sigsetsize)
{
- struct qstr this;
- char name[32];
- struct dentry *dentry;
- struct inode *inode;
- struct file *file;
- int error, fd;
-
- /* Get an ready to use file */
- error = -ENFILE;
- file = get_empty_filp();
- if (!file)
- goto eexit_1;
+ int error;
+ sigset_t ksigmask, sigsaved;
- /* Allocates an inode from the eventpoll file system */
- inode = ep_eventpoll_inode();
- error = PTR_ERR(inode);
- if (IS_ERR(inode))
- goto eexit_2;
+ /*
+ * 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);
+ }
- /* Allocates a free descriptor to plug the file onto */
- error = get_unused_fd();
- if (error < 0)
- goto eexit_3;
- fd = error;
+ error = sys_epoll_wait(epfd, events, maxevents, timeout);
/*
- * Link the inode to a directory entry by creating a unique name
- * using the inode number.
+ * 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.
*/
- 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_vfsmnt = mntget(eventpoll_mnt);
- file->f_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;
+ if (sigmask) {
+ if (error == -EINTR) {
+ memcpy(¤t->saved_sigmask, &sigsaved,
+ sizeof(sigsaved));
+ set_thread_flag(TIF_RESTORE_SIGMASK);
+ } else
+ sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+ }
-eexit_4:
- put_unused_fd(fd);
-eexit_3:
- iput(inode);
-eexit_2:
- put_filp(file);
-eexit_1:
return error;
}
+#endif /* #ifdef TIF_RESTORE_SIGMASK */
+
static int ep_alloc(struct eventpoll **pep)
{
}
/*
- * Walks through the whole hash by freeing each "struct epitem". At this
+ * 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
* 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".
/*
- * Search the file inside the eventpoll hash. It add usage count to
+ * Search the file inside the eventpoll tree. It add usage count to
* the returned item, so the caller must call ep_release_epitem()
* after finished using the "struct epitem".
*/
struct epitem *epi = ep_item_from_epqueue(pt);
struct eppoll_entry *pwq;
- if (epi->nwait >= 0 && (pwq = kmem_cache_alloc(pwq_cache, SLAB_KERNEL))) {
+ if (epi->nwait >= 0 && (pwq = kmem_cache_alloc(pwq_cache, GFP_KERNEL))) {
init_waitqueue_func_entry(&pwq->wait, ep_poll_callback);
pwq->whead = whead;
pwq->base = epi;
struct ep_pqueue epq;
error = -ENOMEM;
- if (!(epi = kmem_cache_alloc(epi_cache, SLAB_KERNEL)))
+ if (!(epi = kmem_cache_alloc(epi_cache, GFP_KERNEL)))
goto eexit_1;
/* 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);
*/
write_lock_irqsave(&ep->lock, flags);
if (ep_is_linked(&epi->rdllink))
- ep_list_del(&epi->rdllink);
+ list_del_init(&epi->rdllink);
write_unlock_irqrestore(&ep->lock, flags);
kmem_cache_free(epi_cache, epi);
epi->event.data = event->data;
/*
- * If the item is not linked to the hash it means that it's on its
+ * If the item is not linked to the RB tree it means that it's on its
* way toward the removal. Do nothing in this case.
*/
if (ep_rb_linked(&epi->rbn)) {
if (nwait) {
while (!list_empty(lsthead)) {
- pwq = list_entry(lsthead->next, struct eppoll_entry, llink);
+ pwq = list_first_entry(lsthead, struct eppoll_entry, llink);
- ep_list_del(&pwq->llink);
+ list_del_init(&pwq->llink);
remove_wait_queue(pwq->whead, &pwq->wait);
kmem_cache_free(pwq_cache, pwq);
}
* we want to remove it from this list to avoid stale events.
*/
if (ep_is_linked(&epi->rdllink))
- ep_list_del(&epi->rdllink);
+ list_del_init(&epi->rdllink);
error = 0;
eexit_1:
/*
- * Removes a "struct epitem" from the eventpoll hash and deallocates
+ * Removes a "struct epitem" from the eventpoll RB tree and deallocates
* all the associated resources.
*/
static int ep_remove(struct eventpoll *ep, struct epitem *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);
+ list_del_init(&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 */
+ /* Really unlink the item from the RB tree */
error = ep_unlink(ep, epi);
write_unlock_irqrestore(&ep->lock, flags);
}
-/*
- * 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)
+ struct epoll_event __user *events, int maxevents)
{
- int eventcnt = 0;
+ int eventcnt, error = -EFAULT, pwake = 0;
unsigned int revents;
- struct list_head *lnk;
+ unsigned long flags;
struct epitem *epi;
+ struct list_head injlist;
+
+ INIT_LIST_HEAD(&injlist);
/*
* 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 "sem" in
+ * read.
*/
- 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);
+ prefetch(epi->rdllink.next);
/*
* Get the ready file event set. We can safely use the file
* 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
+ * "sem" in read, 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++;
}
- }
- 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);
-
- /* Unlink the current item from the transfer list */
- ep_list_del(&epi->txlink);
/*
- * 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.
+ * This is tricky. We are holding the "sem" in read, and this
+ * means that the operations that can change the "linked" status
+ * of the epoll item (epi->rbn and epi->rdllink), cannot touch
+ * them. Also, since we are "linked" from a epi->rdllink POV
+ * (the item is linked to our transmission list we just
+ * spliced), the ep_poll_callback() cannot touch us either,
+ * because of the check present in there. Another parallel
+ * epoll_wait() will not get the same result set, since we
+ * spliced the ready list before. Note that list_del() still
+ * shows the item as linked to the test in ep_poll_callback().
*/
- if (ep_rb_linked(&epi->rbn) && !(epi->event.events & EPOLLET) &&
- (epi->revents & epi->event.events) && !ep_is_linked(&epi->rdllink)) {
- list_add_tail(&epi->rdllink, &ep->rdllist);
- ricnt++;
+ list_del(&epi->rdllink);
+ if (!(epi->event.events & EPOLLET) &&
+ (revents & epi->event.events))
+ list_add_tail(&epi->rdllink, &injlist);
+ else {
+ /*
+ * Be sure the item is totally detached before re-init
+ * the list_head. After INIT_LIST_HEAD() is committed,
+ * the ep_poll_callback() can requeue the item again,
+ * but we don't care since we are already past it.
+ */
+ smp_mb();
+ INIT_LIST_HEAD(&epi->rdllink);
}
}
+ error = 0;
+
+ errxit:
+
+ /*
+ * If the re-injection list or the txlist are not empty, re-splice
+ * them to the ready list and do proper wakeups.
+ */
+ if (!list_empty(&injlist) || !list_empty(txlist)) {
+ write_lock_irqsave(&ep->lock, flags);
- if (ricnt) {
+ list_splice(txlist, &ep->rdllist);
+ list_splice(&injlist, &ep->rdllist);
/*
* Wake up ( if active ) both the eventpoll wait list and the ->poll()
* wait list.
TASK_INTERRUPTIBLE);
if (waitqueue_active(&ep->poll_wait))
pwake++;
- }
- write_unlock_irqrestore(&ep->lock, flags);
+ write_unlock_irqrestore(&ep->lock, flags);
+ }
/* We have to call this outside the lock */
if (pwake)
ep_poll_safewake(&psw, &ep->poll_wait);
+
+ return eventcnt == 0 ? error: eventcnt;
}
static int ep_events_transfer(struct eventpoll *ep,
struct epoll_event __user *events, int maxevents)
{
- int eventcnt = 0;
+ int eventcnt;
+ unsigned long flags;
struct list_head txlist;
INIT_LIST_HEAD(&txlist);
*/
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);
+ /*
+ * Steal the ready list, and re-init the original one to the
+ * empty list.
+ */
+ write_lock_irqsave(&ep->lock, flags);
+ list_splice(&ep->rdllist, &txlist);
+ INIT_LIST_HEAD(&ep->rdllist);
+ write_unlock_irqrestore(&ep->lock, flags);
- /* Reinject ready items into the ready list */
- ep_reinject_items(ep, &txlist);
- }
+ /* Build result set in userspace */
+ eventcnt = ep_send_events(ep, &txlist, events, maxevents);
up_read(&ep->sem);
return res;
}
-
-static int eventpollfs_delete_dentry(struct dentry *dentry)
-{
-
- return 1;
-}
-
-
-static struct inode *ep_eventpoll_inode(void)
-{
- int error = -ENOMEM;
- struct inode *inode = new_inode(eventpoll_mnt->mnt_sb);
-
- if (!inode)
- goto eexit_1;
-
- inode->i_fop = &eventpoll_fops;
-
- /*
- * 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.
- */
- 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);
-}
-
-
-static int
-eventpollfs_get_sb(struct file_system_type *fs_type, int flags,
- const char *dev_name, void *data, struct vfsmount *mnt)
-{
- return get_sb_pseudo(fs_type, "eventpoll:", NULL, EVENTPOLLFS_MAGIC,
- mnt);
-}
-
-
static int __init eventpoll_init(void)
{
- int error;
-
mutex_init(&epmutex);
/* Initialize the structure used to perform safe poll wait head wake ups */
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");
}
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);
}