Re: REGRESSION: RIP: 0010:skb_release_data+0xb8/0x1e0 in vhost/tun

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Mar 19, 2024 at 2:41 PM Willem de Bruijn
<willemdebruijn.kernel@xxxxxxxxx> wrote:
>
> Igor Raits wrote:
> > Hello Willem,
> >
> > On Tue, Mar 19, 2024 at 2:08 PM Willem de Bruijn
> > <willemdebruijn.kernel@xxxxxxxxx> wrote:
> > >
> > > Igor Raits wrote:
> > > > Hello,
> > > >
> > > > We have started to observe kernel crashes on 6.7.y kernels (atm we
> > > > have hit the issue 5 times on 6.7.5 and 6.7.10). On 6.6.9 where we
> > > > have nodes of cluster it looks stable. Please see stacktrace below. If
> > > > you need more information please let me know.
> > > >
> > > > We do not have a consistent reproducer but when we put some bigger
> > > > network load on a VM, the hypervisor's kernel crashes.
> > > >
> > > > Help is much appreciated! We are happy to test any patches.
> > > >
> > > > [62254.167584] stack segment: 0000 [#1] PREEMPT SMP NOPTI
> > >
> > > Did you miss the first part of the Oops?
> >
> > Actually I copied it as-is from our log system. As it is a physical
> > server, such logs are sent via netconsole to another server. This is
> > the first line I see in the log in the time segment.
> >
> > >
> > > > [62254.173450] CPU: 63 PID: 11939 Comm: vhost-11890 Tainted: G
> > > >    E      6.7.10-1.gdc.el9.x86_64 #1
> > > > [62254.183743] Hardware name: Dell Inc. PowerEdge R7525/0H3K7P, BIOS
> > > > 2.14.1 12/17/2023
> > > > [62254.192083] RIP: 0010:skb_release_data+0xb8/0x1e0
> > > > [62254.197357] Code: 48 83 c3 01 39 d8 7e 54 48 89 d8 48 c1 e0 04 41
> > > > 80 7d 7e 00 49 8b 6c 04 30 79 0f 44 89 f6 48 89 ef e8 4c e4 ff ff 84
> > > > c0 75 d0 <48> 8b 45 08 a8 01 0f 85 09 01 00 00 e9 d9 00 00 00 0f 1f 44
> > > > 00 00
> > > > [62254.217013] RSP: 0018:ffffa975a0247ba8 EFLAGS: 00010206
> > > > [62254.222692] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000785
> > > > [62254.230263] RDX: 0000000000000016 RSI: 0000000000000002 RDI: ffff989862b32b00
> > > > [62254.237878] RBP: 4f2b318c69a8b0f9 R08: 000000000001fe4d R09: 000000000000003a
> > > > [62254.245417] R10: 0000000000000000 R11: 0000000000001736 R12: ffff9880b819aec0
> > > > [62254.252963] R13: ffff989862b32b00 R14: 0000000000000000 R15: 0000000000000002
> > > > [62254.260591] FS:  00007f6cf388bf80(0000) GS:ffff98b85fbc0000(0000)
> > > > knlGS:0000000000000000
> > > > [62254.269061] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > > [62254.275170] CR2: 000000c002236020 CR3: 000000387d37a002 CR4: 0000000000770ef0
> > > > [62254.282733] PKRU: 55555554
> > > > [62254.285911] Call Trace:
> > > > [62254.288884]  <TASK>
> > > > [62254.291549]  ? die+0x33/0x90
> > > > [62254.294769]  ? do_trap+0xe0/0x110
> > > > [62254.298405]  ? do_error_trap+0x65/0x80
> > > > [62254.302471]  ? exc_stack_segment+0x35/0x50
> > > > [62254.306884]  ? asm_exc_stack_segment+0x22/0x30
> > > > [62254.311637]  ? skb_release_data+0xb8/0x1e0
> > > > [62254.316047]  kfree_skb_list_reason+0x6d/0x210
> > > > [62254.320697]  ? free_unref_page_commit+0x80/0x2f0
> > > > [62254.325700]  ? free_unref_page+0xe9/0x130
> > > > [62254.330013]  skb_release_data+0xfc/0x1e0
> > > > [62254.334261]  consume_skb+0x45/0xd0
> > > > [62254.338077]  tun_do_read+0x68/0x1f0 [tun]
> > > > [62254.342414]  tun_recvmsg+0x7e/0x160 [tun]
> > > > [62254.346696]  handle_rx+0x3ab/0x750 [vhost_net]
> > > > [62254.351488]  vhost_worker+0x42/0x70 [vhost]
> > > > [62254.355934]  vhost_task_fn+0x4b/0xb0
> > >
> > > Neither tun nor vhost_net saw significant changes between the two
> > > reported kernels.
> > >
> > >     $ git log --oneline v6.6..v6.7 -- drivers/net/tun.c drivers/vhost/net.c | wc -l
> > >     0
> > >
> > >     $ git log --oneline linux/v6.6.9..linux/v6.7.5 -- drivers/net/tun.c drivers/vhost/net.c
> > >     6438382dd9f8 tun: add missing rx stats accounting in tun_xdp_act
> > >     4efd09da0d49 tun: fix missing dropped counter in tun_xdp_act
> > >
> > > So the cause is likely in the code that generated the skb or something
> > > that modified it along the way.
> > >
> > > It could be helpful if it is possible to bisect further. Though odds
> > > are that the issue is between v6.6 and v6.7, not introduced in the
> > > stable backports after that. So it is a large target.
> >
> > Yeah, as I replied later to my original message - we actually also see
> > the issue on 6.6.9 as well but it looks slightly different.
> >
> > Actually while writing reply got 6.6.9 crashed too:
> >
> > [13330.391004] tun: unexpected GSO type: 0x4ec1c942, gso_size 20948,
> > hdr_len 3072
>
> This looks like memory corruption
>
> > > Getting the exact line in skb_release_data that causes the Oops
> > > would be helpful too, e.g.,
> > >
> > > gdb vmlinux
> > > list *(skb_release_data+0xb8)
> >
> > Unfortunately we do not collect kdumps so this is not going to be easy
> > :( We will investigate the possibility of getting the dump though.
>
> No need for a kdump. As long as you have the vmlinux of the kernel.

(gdb) list *(skb_release_data+0xb8)
0xffffffff81a36088 is in skb_release_data (./include/linux/page-flags.h:247).
242 return page_fixed_fake_head(page) != page;
243 }
244
245 static inline unsigned long _compound_head(const struct page *page)
246 {
247 unsigned long head = READ_ONCE(page->compound_head);
248
249 if (unlikely(head & 1))
250 return head - 1;
251 return (unsigned long)page_fixed_fake_head(page);





[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux