From: Lee Jones <lee@kernel.org>
To: lee@kernel.org, "David S. Miller" <davem@davemloft.net>,
Eric Dumazet <edumazet@google.com>,
Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
Kuniyuki Iwashima <kuniyu@amazon.com>,
Jens Axboe <axboe@kernel.dk>, Sasha Levin <sashal@kernel.org>,
Michal Luczaj <mhal@rbox.co>, Rao Shoaib <Rao.Shoaib@oracle.com>,
Simon Horman <horms@kernel.org>,
linux-kernel@vger.kernel.org, netdev@vger.kernel.org
Cc: stable@vger.kernel.org
Subject: [PATCH v6.6 02/26] af_unix: Run GC on only one CPU.
Date: Wed, 21 May 2025 14:45:10 +0000 [thread overview]
Message-ID: <20250521144803.2050504-3-lee@kernel.org> (raw)
In-Reply-To: <20250521144803.2050504-1-lee@kernel.org>
From: Kuniyuki Iwashima <kuniyu@amazon.com>
[ Upstream commit 8b90a9f819dc2a06baae4ec1a64d875e53b824ec ]
If more than 16000 inflight AF_UNIX sockets exist and the garbage
collector is not running, unix_(dgram|stream)_sendmsg() call unix_gc().
Also, they wait for unix_gc() to complete.
In unix_gc(), all inflight AF_UNIX sockets are traversed at least once,
and more if they are the GC candidate. Thus, sendmsg() significantly
slows down with too many inflight AF_UNIX sockets.
There is a small window to invoke multiple unix_gc() instances, which
will then be blocked by the same spinlock except for one.
Let's convert unix_gc() to use struct work so that it will not consume
CPUs unnecessarily.
Note WRITE_ONCE(gc_in_progress, true) is moved before running GC.
If we leave the WRITE_ONCE() as is and use the following test to
call flush_work(), a process might not call it.
CPU 0 CPU 1
--- ---
start work and call __unix_gc()
if (work_pending(&unix_gc_work) || <-- false
READ_ONCE(gc_in_progress)) <-- false
flush_work(); <-- missed!
WRITE_ONCE(gc_in_progress, true)
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/20240123170856.41348-5-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit 8b90a9f819dc2a06baae4ec1a64d875e53b824ec)
Signed-off-by: Lee Jones <lee@kernel.org>
---
net/unix/garbage.c | 54 +++++++++++++++++++++++-----------------------
1 file changed, 27 insertions(+), 27 deletions(-)
diff --git a/net/unix/garbage.c b/net/unix/garbage.c
index 38639766b9e7c..a2a8543613a52 100644
--- a/net/unix/garbage.c
+++ b/net/unix/garbage.c
@@ -86,7 +86,6 @@
/* Internal data structures and random procedures: */
static LIST_HEAD(gc_candidates);
-static DECLARE_WAIT_QUEUE_HEAD(unix_gc_wait);
static void scan_inflight(struct sock *x, void (*func)(struct unix_sock *),
struct sk_buff_head *hitlist)
@@ -182,23 +181,8 @@ static void inc_inflight_move_tail(struct unix_sock *u)
}
static bool gc_in_progress;
-#define UNIX_INFLIGHT_TRIGGER_GC 16000
-
-void wait_for_unix_gc(void)
-{
- /* If number of inflight sockets is insane,
- * force a garbage collect right now.
- * Paired with the WRITE_ONCE() in unix_inflight(),
- * unix_notinflight() and gc_in_progress().
- */
- if (READ_ONCE(unix_tot_inflight) > UNIX_INFLIGHT_TRIGGER_GC &&
- !READ_ONCE(gc_in_progress))
- unix_gc();
- wait_event(unix_gc_wait, !READ_ONCE(gc_in_progress));
-}
-/* The external entry point: unix_gc() */
-void unix_gc(void)
+static void __unix_gc(struct work_struct *work)
{
struct sk_buff *next_skb, *skb;
struct unix_sock *u;
@@ -209,13 +193,6 @@ void unix_gc(void)
spin_lock(&unix_gc_lock);
- /* Avoid a recursive GC. */
- if (gc_in_progress)
- goto out;
-
- /* Paired with READ_ONCE() in wait_for_unix_gc(). */
- WRITE_ONCE(gc_in_progress, true);
-
/* First, select candidates for garbage collection. Only
* in-flight sockets are considered, and from those only ones
* which don't have any external reference.
@@ -346,8 +323,31 @@ void unix_gc(void)
/* Paired with READ_ONCE() in wait_for_unix_gc(). */
WRITE_ONCE(gc_in_progress, false);
- wake_up(&unix_gc_wait);
-
- out:
spin_unlock(&unix_gc_lock);
}
+
+static DECLARE_WORK(unix_gc_work, __unix_gc);
+
+void unix_gc(void)
+{
+ WRITE_ONCE(gc_in_progress, true);
+ queue_work(system_unbound_wq, &unix_gc_work);
+}
+
+#define UNIX_INFLIGHT_TRIGGER_GC 16000
+
+void wait_for_unix_gc(void)
+{
+ /* If number of inflight sockets is insane,
+ * force a garbage collect right now.
+ *
+ * Paired with the WRITE_ONCE() in unix_inflight(),
+ * unix_notinflight(), and __unix_gc().
+ */
+ if (READ_ONCE(unix_tot_inflight) > UNIX_INFLIGHT_TRIGGER_GC &&
+ !READ_ONCE(gc_in_progress))
+ unix_gc();
+
+ if (READ_ONCE(gc_in_progress))
+ flush_work(&unix_gc_work);
+}
--
2.49.0.1112.g889b7c5bd8-goog
next prev parent reply other threads:[~2025-05-21 14:49 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-05-21 14:45 [PATCH v6.6 00/26] af_unix: Align with upstream to avoid a potential UAF Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 01/26] af_unix: Return struct unix_sock from unix_get_socket() Lee Jones
2025-05-21 14:45 ` Lee Jones [this message]
2025-05-21 14:45 ` [PATCH v6.6 03/26] af_unix: Try to run GC async Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 04/26] af_unix: Replace BUG_ON() with WARN_ON_ONCE() Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 05/26] af_unix: Remove io_uring code for GC Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 06/26] af_unix: Remove CONFIG_UNIX_SCM Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 07/26] af_unix: Allocate struct unix_vertex for each inflight AF_UNIX fd Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 08/26] af_unix: Allocate struct unix_edge " Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 09/26] af_unix: Link struct unix_edge when queuing skb Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 10/26] af_unix: Bulk update unix_tot_inflight/unix_inflight " Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 11/26] af_unix: Iterate all vertices by DFS Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 12/26] af_unix: Detect Strongly Connected Components Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 13/26] af_unix: Save listener for embryo socket Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 14/26] af_unix: Fix up unix_edge.successor " Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 15/26] af_unix: Save O(n) setup of Tarjan's algo Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 16/26] af_unix: Skip GC if no cycle exists Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 17/26] af_unix: Avoid Tarjan's algorithm if unnecessary Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 18/26] af_unix: Assign a unique index to SCC Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 19/26] af_unix: Detect dead SCC Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 20/26] af_unix: Replace garbage collection algorithm Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 21/26] af_unix: Remove lock dance in unix_peek_fds() Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 22/26] af_unix: Try not to hold unix_gc_lock during accept() Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 23/26] af_unix: Don't access successor in unix_del_edges() during GC Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 24/26] af_unix: Add dead flag to struct scm_fp_list Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 25/26] af_unix: Fix garbage collection of embryos carrying OOB with SCM_RIGHTS Lee Jones
2025-05-21 14:45 ` [PATCH v6.6 26/26] af_unix: Fix uninit-value in __unix_walk_scc() Lee Jones
2025-05-29 12:26 ` [PATCH v6.6 00/26] af_unix: Align with upstream to avoid a potential UAF Greg KH
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=20250521144803.2050504-3-lee@kernel.org \
--to=lee@kernel.org \
--cc=Rao.Shoaib@oracle.com \
--cc=axboe@kernel.dk \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=horms@kernel.org \
--cc=kuba@kernel.org \
--cc=kuniyu@amazon.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mhal@rbox.co \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=sashal@kernel.org \
--cc=stable@vger.kernel.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).