Re: Backport patch from 4.2 to 3.18

[Sasha Levin <sasha.levin@oracle.com>]

On 03/04/2016 03:40 PM, Andrei Sharaev wrote:
> Hi Sasha,
> 
> Can you backport this patch for "inet-frag-fixes" to linux kernel 3.18 LTS?
> http://kernel.suse.com/cgit/kernel/commit/?h=v4.2-rc5&id=64b892ad2326348a5b8314167590d240e3bcc69e
> 
> I get 1-5 kernel panics in month for linux kernels 3.18.24-3.18.26 at my NAT server with big IPv4 traffic (10-15 Gbps).
> My kernel panics have similar symptoms:
>> <82>general protection fault: 0000 [#1] SMP
>> <82>Modules linked in: bonding ipt_NETFLOW(O) xt_recent configfs x86_pkg_temp_thermal ixgbe(O)
>> <86>CPU: 13 PID: 29908 Comm: kworker/13:2 Tainted: G          IO   3.18.26 #1
>> <86>Hardware name: Intel Corporation S2600WT2/S2600WT2, BIOS SE5C610.86B.01.01.0005.101720141054 10/17/2014
>> <82>Workqueue: events inet_frag_worker
>> <86>task: ffff88046cdba9a0 ti: ffff880454928000 task.ti: ffff880454928000
>> <82>RIP: 0010:[<ffffffff815b91d9>]  [<ffffffff815b91d9>] inet_evict_bucket+0x109/0x160
>> <86>RSP: 0018:ffff88045492bd38  EFLAGS: 00010286
>> <86>RAX: ffff880441d0e001 RBX: dead0000001000c0 RCX: 000000018030002e
>> <86>RDX: 000000018030002f RSI: ffff880441d0e000 RDI: dead0000001000c0
>> <86>RBP: ffff88045492bd88 R08: 0000000000000000 R09: ffff88086cc88500
>> <86>R10: ffff88046fdb5c50 R11: ffffea0011074380 R12: 0000000000000002
>> <86>R13: ffffffff81e02200 R14: 0000000000000000 R15: ffff88083f0942a0
>> <86>FS:  0000000000000000(0000) GS:ffff88046fda0000(0000) knlGS:0000000000000000
>> <86>CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>> <86>CR2: 00007fab8f466000 CR3: 000000085f66d000 CR4: 00000000001407e0
>> <86>Stack:
>> <82> ffffffff81e05a78 ffffffff81e05a70 ffff88046cdba9a0 ffff88083f0942e0
>> <82> ffff88046f808c00 0000000000000079 ffffffff81e02200 ffffffff81e06200
>> <82> 0000000000000388 0000000000000007 ffff88045492bdf8 ffffffff815b928a
>> <86>Call Trace:
>> <82> [<ffffffff815b928a>] inet_frag_worker+0x5a/0x230
>> <82> [<ffffffff8105808d>] process_one_work+0x12d/0x330
>> <82> [<ffffffff8105892b>] worker_thread+0x4b/0x450
>> <82> [<ffffffff810588e0>] ? cancel_delayed_work_sync+0x10/0x10
>> <82> [<ffffffff8105cd34>] kthread+0xc4/0xe0
>> <82> [<ffffffff81060c59>] ? finish_task_switch+0x49/0xc0
>> <82> [<ffffffff8105cc70>] ? kthread_create_on_node+0x170/0x170
>> <82> [<ffffffff81603d88>] ret_from_fork+0x58/0x90
>> <82> [<ffffffff8105cc70>] ? kthread_create_on_node+0x170/0x170
>> <82>Code: f6 0f 85 73 ff ff ff 48 8b 45 b8 80 40 08 01 48 8b 7d c8 48 85 ff 74 23 48 83 ef 40 75 0d eb 1b 66 90 48 83 eb 40 48 89 df 74 10 <48> 8b 5f 40 41 ff 95 70 40 00 00 48 85 db 75 e7 48 83 c4 28 44
>> <22>RIP  [<ffffffff815b91d9>] inet_evict_bucket+0x109/0x160
>> <82> RSP <ffff88045492bd38>
> 

Hey Andrei,

Thanks for the report.

Usually David Miller (Cc'ed) handles backporting network commits. In this case, I see
that he has elected not to backport it into 4.1 or 3.18, so I don't want to do it without
getting an ack from him first.

David, is it ok to backport these commits back to 3.18 (and probably 4.1)?


Thanks,
Sasha