From: Al Viro <viro@ZenIV.linux.org.uk>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
Marc Zyngier <maz@kernel.org>
Subject: [RFC PATCH 20/27] ep_insert(): we only need tep->mtx around the insertion itself
Date: Sun, 4 Oct 2020 03:39:22 +0100 [thread overview]
Message-ID: <20201004023929.2740074-20-viro@ZenIV.linux.org.uk> (raw)
In-Reply-To: <20201004023929.2740074-1-viro@ZenIV.linux.org.uk>
From: Al Viro <viro@zeniv.linux.org.uk>
We do need ep->mtx (and we are holding it all along), but that's
the lock on the epoll we are inserting into; locking of the
epoll being inserted is not needed for most of that work -
as the matter of fact, we only need it to provide barriers
for the fastpath check (for now).
Move taking and releasing it into ep_insert(). The caller
(do_epoll_ctl()) doesn't need to bother with that at all.
Moreover, that way we kill the kludge in ep_item_poll() - now
it's always called with tep unlocked.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
fs/eventpoll.c | 28 ++++++++++------------------
1 file changed, 10 insertions(+), 18 deletions(-)
diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index c987b61701e4..39947b71f7af 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -731,8 +731,6 @@ static int ep_eventpoll_release(struct inode *inode, struct file *file)
static __poll_t ep_read_events_proc(struct eventpoll *ep, struct list_head *head,
int depth);
-static void ep_ptable_queue_proc(struct file *file, wait_queue_head_t *whead,
- poll_table *pt);
/*
* Differs from ep_eventpoll_poll() in that internal callers already have
@@ -745,7 +743,6 @@ static __poll_t ep_item_poll(const struct epitem *epi, poll_table *pt,
struct eventpoll *ep;
LIST_HEAD(txlist);
__poll_t res;
- bool locked;
pt->_key = epi->event.events;
if (!is_file_epoll(epi->ffd.file))
@@ -754,15 +751,11 @@ static __poll_t ep_item_poll(const struct epitem *epi, poll_table *pt,
ep = epi->ffd.file->private_data;
poll_wait(epi->ffd.file, &ep->poll_wait, pt);
- // kludge: ep_insert() calls us with ep->mtx already locked
- locked = pt && (pt->_qproc == ep_ptable_queue_proc);
- if (!locked)
- mutex_lock_nested(&ep->mtx, depth);
+ mutex_lock_nested(&ep->mtx, depth);
ep_start_scan(ep, &txlist);
res = ep_read_events_proc(ep, &txlist, depth + 1);
ep_done_scan(ep, &txlist);
- if (!locked)
- mutex_unlock(&ep->mtx);
+ mutex_unlock(&ep->mtx);
return res & epi->event.events;
}
@@ -1365,6 +1358,10 @@ static int ep_insert(struct eventpoll *ep, const struct epoll_event *event,
long user_watches;
struct epitem *epi;
struct ep_pqueue epq;
+ struct eventpoll *tep = NULL;
+
+ if (is_file_epoll(tfile))
+ tep = tfile->private_data;
lockdep_assert_irqs_enabled();
@@ -1394,6 +1391,8 @@ static int ep_insert(struct eventpoll *ep, const struct epoll_event *event,
atomic_long_inc(&ep->user->epoll_watches);
+ if (tep)
+ mutex_lock(&tep->mtx);
/* Add the current item to the list of active epoll hook for this file */
spin_lock(&tfile->f_lock);
list_add_tail_rcu(&epi->fllink, &tfile->f_ep_links);
@@ -1404,6 +1403,8 @@ static int ep_insert(struct eventpoll *ep, const struct epoll_event *event,
* protected by "mtx", and ep_insert() is called with "mtx" held.
*/
ep_rbtree_insert(ep, epi);
+ if (tep)
+ mutex_unlock(&tep->mtx);
/* now check if we've created too many backpaths */
if (unlikely(full_check && reverse_path_check())) {
@@ -2034,13 +2035,6 @@ int do_epoll_ctl(int epfd, int op, int fd, struct epoll_event *epds,
error = epoll_mutex_lock(&ep->mtx, 0, nonblock);
if (error)
goto error_tgt_fput;
- if (is_file_epoll(tf.file)) {
- error = epoll_mutex_lock(&tep->mtx, 1, nonblock);
- if (error) {
- mutex_unlock(&ep->mtx);
- goto error_tgt_fput;
- }
- }
}
}
@@ -2076,8 +2070,6 @@ int do_epoll_ctl(int epfd, int op, int fd, struct epoll_event *epds,
error = -ENOENT;
break;
}
- if (tep != NULL)
- mutex_unlock(&tep->mtx);
mutex_unlock(&ep->mtx);
error_tgt_fput:
--
2.11.0
next prev parent reply other threads:[~2020-10-04 2:40 UTC|newest]
Thread overview: 38+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-10-04 2:36 [RFC][PATCHSET] epoll cleanups Al Viro
2020-10-04 2:39 ` [RFC PATCH 01/27] epoll: switch epitem->pwqlist to single-linked list Al Viro
2020-10-04 2:39 ` [RFC PATCH 02/27] epoll: get rid of epitem->nwait Al Viro
2020-10-04 2:39 ` [RFC PATCH 03/27] untangling ep_call_nested(): get rid of useless arguments Al Viro
2020-10-04 2:39 ` [RFC PATCH 04/27] untangling ep_call_nested(): it's all serialized on epmutex Al Viro
2020-10-04 2:39 ` [RFC PATCH 05/27] untangling ep_call_nested(): take pushing cookie into a helper Al Viro
2020-10-04 2:39 ` [RFC PATCH 06/27] untangling ep_call_nested(): move push/pop of cookie into the callbacks Al Viro
2020-10-04 2:39 ` [RFC PATCH 07/27] untangling ep_call_nested(): and there was much rejoicing Al Viro
2020-10-04 2:39 ` [RFC PATCH 08/27] reverse_path_check_proc(): sane arguments Al Viro
2020-10-04 2:39 ` [RFC PATCH 09/27] reverse_path_check_proc(): don't bother with cookies Al Viro
2020-10-04 2:39 ` [RFC PATCH 10/27] clean reverse_path_check_proc() a bit Al Viro
2020-10-04 2:39 ` [RFC PATCH 11/27] ep_loop_check_proc(): lift pushing the cookie into callers Al Viro
2020-10-04 2:39 ` [RFC PATCH 12/27] get rid of ep_push_nested() Al Viro
2020-10-04 2:39 ` [RFC PATCH 13/27] ep_loop_check_proc(): saner calling conventions Al Viro
2020-10-04 2:39 ` [RFC PATCH 14/27] ep_scan_ready_list(): prepare to splitup Al Viro
2020-10-04 2:39 ` [RFC PATCH 15/27] lift the calls of ep_read_events_proc() into the callers Al Viro
2020-10-04 2:39 ` [RFC PATCH 16/27] lift the calls of ep_send_events_proc() " Al Viro
2020-10-04 2:39 ` [RFC PATCH 17/27] ep_send_events_proc(): fold into the caller Al Viro
2020-10-04 2:39 ` [RFC PATCH 18/27] lift locking/unlocking ep->mtx out of ep_{start,done}_scan() Al Viro
2020-10-04 2:39 ` [RFC PATCH 19/27] ep_insert(): don't open-code ep_remove() on failure exits Al Viro
2020-10-04 2:39 ` Al Viro [this message]
2020-10-04 12:56 ` [ep_insert()] 9ee1cc5666: WARNING:possible_recursive_locking_detected kernel test robot
2020-10-04 14:17 ` Al Viro
2020-10-04 14:27 ` Al Viro
2020-10-04 2:39 ` [RFC PATCH 21/27] take the common part of ep_eventpoll_poll() and ep_item_poll() into helper Al Viro
2020-10-04 2:39 ` [RFC PATCH 22/27] fold ep_read_events_proc() into the only caller Al Viro
2020-10-04 2:39 ` [RFC PATCH 23/27] ep_insert(): move creation of wakeup source past the fl_ep_links insertion Al Viro
2020-10-04 2:39 ` [RFC PATCH 24/27] convert ->f_ep_links/->fllink to hlist Al Viro
2020-10-04 2:39 ` [RFC PATCH 25/27] lift rcu_read_lock() into reverse_path_check() Al Viro
2020-10-04 2:39 ` [RFC PATCH 26/27] epoll: massage the check list insertion Al Viro
2020-10-04 2:39 ` [RFC PATCH 27/27] epoll: take epitem list out of struct file Al Viro
2020-10-05 20:37 ` Qian Cai
2020-10-05 20:49 ` Al Viro
2020-10-04 2:49 ` [RFC][PATCHSET] epoll cleanups Al Viro
2020-10-04 12:13 ` Matthew Wilcox
2020-10-04 14:15 ` Al Viro
2020-10-04 18:08 ` Linus Torvalds
2020-10-04 20:05 ` Al Viro
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20201004023929.2740074-20-viro@ZenIV.linux.org.uk \
--to=viro@zeniv.linux.org.uk \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=maz@kernel.org \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).