The patch titled
Subject: eventfd: don't take the spinlock in eventfd_poll
has been removed from the -mm tree. Its filename was
eventfd-dont-take-the-spinlock-in-eventfd_poll.patch
This patch was dropped because it was merged into mainline or a subsystem tree
------------------------------------------------------
From: Chris Mason <clm@xxxxxx>
Subject: eventfd: don't take the spinlock in eventfd_poll
The spinlock in eventfd_poll is trying to protect the count of events so
it can decide if it should return POLLIN, POLLERR, or POLLOUT. But,
because of the way we drop the lock after calling poll_wait, and drop it
again before returning, we have the same pile of races with the lock as we
do with a single read of ctx->count().
This replaces the lock with a read barrier and single read.
eventfd_write does a single bump of ctx->count, so this should not add new
races with adding events. eventfd_read is similar, it will do a single
decrement with the lock held, and so we're making the race with concurrent
readers slightly larger.
This spinlock is the top CPU user in kernel code during one of our
workloads. Removing it gives us a ~2% boost.
[arnd@xxxxxxxx: avoid unused variable warning]
[dan.carpenter@xxxxxxxxxx: type bug in eventfd_poll()]
Signed-off-by: Chris Mason <clm@xxxxxx>
Cc: Davide Libenzi <davidel@xxxxxxxxxxxxxxx>
Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/eventfd.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff -puN fs/eventfd.c~eventfd-dont-take-the-spinlock-in-eventfd_poll fs/eventfd.c
--- a/fs/eventfd.c~eventfd-dont-take-the-spinlock-in-eventfd_poll
+++ a/fs/eventfd.c
@@ -118,18 +118,18 @@ static unsigned int eventfd_poll(struct
{
struct eventfd_ctx *ctx = file->private_data;
unsigned int events = 0;
- unsigned long flags;
+ u64 count;
poll_wait(file, &ctx->wqh, wait);
+ smp_rmb();
+ count = ctx->count;
- spin_lock_irqsave(&ctx->wqh.lock, flags);
- if (ctx->count > 0)
+ if (count > 0)
events |= POLLIN;
- if (ctx->count == ULLONG_MAX)
+ if (count == ULLONG_MAX)
events |= POLLERR;
- if (ULLONG_MAX - 1 > ctx->count)
+ if (ULLONG_MAX - 1 > count)
events |= POLLOUT;
- spin_unlock_irqrestore(&ctx->wqh.lock, flags);
return events;
}
_
Patches currently in -mm which might be from clm@xxxxxx are
origin.patch
linux-next.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
