Matt Mackall suggested converting epoll's ep_lock to a bitlock as a way of
saving space in struct file. This patch makes that change.
Signed-off-by: Jonathan Corbet <corbet@xxxxxxx>
---
fs/eventpoll.c | 13 ++++++++-----
fs/fcntl.c | 2 +-
include/asm-generic/fcntl.h | 4 ++++
include/linux/eventpoll.h | 2 +-
include/linux/fs.h | 1 -
5 files changed, 14 insertions(+), 8 deletions(-)
diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index ba2f9ec..95fae01 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -25,6 +25,7 @@
#include <linux/list.h>
#include <linux/hash.h>
#include <linux/spinlock.h>
+#include <linux/bit_spinlock.h>
#include <linux/syscalls.h>
#include <linux/rbtree.h>
#include <linux/wait.h>
@@ -427,10 +428,10 @@ static int ep_remove(struct eventpoll *ep, struct epitem *epi)
ep_unregister_pollwait(ep, epi);
/* Remove the current item from the list of epoll hooks */
- spin_lock(&file->f_ep_lock);
+ bit_spin_lock(NR_EPOLL_FILE_LOCK, &file->f_flags);
if (ep_is_linked(&epi->fllink))
list_del_init(&epi->fllink);
- spin_unlock(&file->f_ep_lock);
+ bit_spin_unlock(NR_EPOLL_FILE_LOCK, &file->f_flags);
rb_erase(&epi->rbn, &ep->rbr);
@@ -549,7 +550,7 @@ void eventpoll_release_file(struct file *file)
struct epitem *epi;
/*
- * We don't want to get "file->f_ep_lock" because it is not
+ * We don't want to get EPOLL_FILE_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
@@ -558,6 +559,8 @@ void eventpoll_release_file(struct file *file)
* 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().
+ *
+ * Besides, ep_remove() takes the lock.
*/
mutex_lock(&epmutex);
@@ -800,9 +803,9 @@ static int ep_insert(struct eventpoll *ep, struct epoll_event *event,
goto error_unregister;
/* Add the current item to the list of active epoll hook for this file */
- spin_lock(&tfile->f_ep_lock);
+ bit_spin_lock(NR_EPOLL_FILE_LOCK, &tfile->f_flags);
list_add_tail(&epi->fllink, &tfile->f_ep_links);
- spin_unlock(&tfile->f_ep_lock);
+ bit_spin_unlock(NR_EPOLL_FILE_LOCK, &tfile->f_flags);
/*
* Add the current item to the RB tree. All RB tree operations are
diff --git a/fs/fcntl.c b/fs/fcntl.c
index 3affd3e..756119e 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -299,7 +299,7 @@ static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
set_close_on_exec(fd, arg & FD_CLOEXEC);
break;
case F_GETFL:
- err = filp->f_flags;
+ err = filp->f_flags & ~EPOLL_FILE_LOCK;
break;
case F_SETFL:
err = setfl(fd, filp, arg);
diff --git a/include/asm-generic/fcntl.h b/include/asm-generic/fcntl.h
index 41e9f88..d56a513 100644
--- a/include/asm-generic/fcntl.h
+++ b/include/asm-generic/fcntl.h
@@ -62,6 +62,10 @@
#define NR_O_NDELAY NR_O_NONBLOCK
#endif
+/* Use bit 31 for epoll locking */
+#define NR_EPOLL_FILE_LOCK 31
+#define EPOLL_FILE_LOCK (1 << NR_EPOLL_FILE_LOCK)
+
#define F_DUPFD 0 /* dup */
#define F_GETFD 1 /* get close_on_exec */
#define F_SETFD 2 /* set/clear close_on_exec */
diff --git a/include/linux/eventpoll.h b/include/linux/eventpoll.h
index f1e1d3c..7b5c58a 100644
--- a/include/linux/eventpoll.h
+++ b/include/linux/eventpoll.h
@@ -61,7 +61,7 @@ struct file;
static inline void eventpoll_init_file(struct file *file)
{
INIT_LIST_HEAD(&file->f_ep_links);
- spin_lock_init(&file->f_ep_lock);
+ clear_bit(NR_EPOLL_FILE_LOCK, &file->f_flags);
}
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 9d7cb0e..ecd9a5b 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -865,7 +865,6 @@ struct file {
#ifdef CONFIG_EPOLL
/* Used by fs/eventpoll.c to link all the hooks to this file */
struct list_head f_ep_links;
- spinlock_t f_ep_lock;
#endif /* #ifdef CONFIG_EPOLL */
struct address_space *f_mapping;
#ifdef CONFIG_DEBUG_WRITECOUNT
--
1.6.1.1
--
To unsubscribe from this list: send the line "unsubscribe linux-api" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
