This is a quite old kernel version, probably already fixed...

Begin forwarded message:

Date: Wed, 27 Jan 2021 01:48:33 +0000
From: bugzilla-dae...@bugzilla.kernel.org
To: step...@networkplumber.org
Subject: [Bug 211375] New: Memory leak about TCP slab which have too big used 
sockets


https://bugzilla.kernel.org/show_bug.cgi?id=211375

            Bug ID: 211375
           Summary: Memory leak about TCP slab which have too big used
                    sockets
           Product: Networking
           Version: 2.5
    Kernel Version: 4.18.16
          Hardware: All
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: IPV4
          Assignee: step...@networkplumber.org
          Reporter: 390231...@qq.com
        Regression: No

Memory leak occurred in linux of 4.18.16, we use nginx as a server, I found
some problems related to TCP as following:
1. slabtop a:
OBJS ACTIVE USE OBJ SIZE SLABS OBJ/SLAB CACHE SIZE NAME
14081971 13980986 99% 2.06K 938799 15 30041568K TCP

2. cat /proc/meminfo
SUnreclaim: 31405028 kB

3. cat /proc/net/sockstat
sockets: used 13976123
TCP: inuse 18 orphan 0 tw 44 alloc 18 mem 1
UDP: inuse 54 mem 45
UDPLITE: inuse 0 RAW: inuse 9
FRAG: inuse 0 memory 0

4. lsof
there are 19000 line, it seems like ok.

As above, it seems that tcp sk memory leak, "sockets: used 13976123" illustrate
that "net->core.sock_inuse" is too big, which increase in inet_create(socket
create) or sk_clone_lock(child socket create) and decrease in __sk_free, I kill
almost all of application layer program, but the "sockets: used" almostly not
reduce.

Do you have any suggestions for this problem, Thanks.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are the assignee for the bug.

Reply via email to