Re: PROBLEM: virtio_net LRO kernel panics

["Michael S. Tsirkin" <mst@redhat.com>]

On Thu, Jul 22, 2021 at 06:27:18PM -0500, Ivan wrote:
> Dear Sir,
> 
> I've been plagued with kernel panics recently. The problem is easily
> reproducible on any virtual machine that uses the virtio-net driver
> from stock Linux kernel. Simply isuse this command:
> 
> echo 1 > /proc/sys/net/ipv4/ip_forward
> ...and the kernel panics.
> 
> Is there any way we can possibly fix this?
> 
> kernel: ------------[ cut here ]------------
> kernel: netdevice: eth0: failed to disable LRO!
> kernel: WARNING: CPU: 1 PID: 424 at net/core/dev.c:1768
> dev_disable_lro+0x108/0x150
> kernel: Modules linked in: nls_iso8859_1 nls_cp437 vfat fat usbhid
> atkbd libps2 ahci libahci virtio_net ohci_pci net_failover failover
> i8042 serio lpc_ich mfd_core libata ohci_hcd ehci_pci ehci_hcd usbcore
> rng_core i2c_piix4 i2c_core virtio_pci usb_common
> virtio_pci_modern_dev virtio_ring virtio loop unix
> kernel: CPU: 1 PID: 424 Comm: bash Not tainted 5.13.4-gnu.4-NuMini #1
> kernel: Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS
> VirtualBox 12/01/2006
> kernel: RIP: 0010:dev_disable_lro+0x108/0x150
> kernel: Code: ae 88 74 14 be 25 00 00 00 48 89 df e8 f1 54 ed ff 48 85
> c0 48 0f 44 eb 4c 89 e2 48 89 ee 48 c7 c7 00 c6 ae 88 e8 7a 76 0c 00
> <0f> 0b e9 2d ff ff ff 80 3d e8 70 97 00 00 49 c7 c4 73 bb ae 88 75
> kernel: RSP: 0018:ffffb596c0237d80 EFLAGS: 00010282
> kernel: RAX: 0000000000000000 RBX: ffff9af9c1835000 RCX: ffff9af9fed17538
> kernel: RDX: 00000000ffffffd8 RSI: 0000000000000027 RDI: ffff9af9fed17530
> kernel: RBP: ffff9af9c1835000 R08: ffffffff88c96ac8 R09: 0000000000004ffb
> kernel: R10: 00000000fffff000 R11: 3fffffffffffffff R12: ffffffff88ac7c3d
> kernel: R13: 0000000000000000 R14: ffffffff88cb2748 R15: ffff9af9c12166c8
> kernel: FS:  00007fd4911b8740(0000) GS:ffff9af9fed00000(0000)
> knlGS:0000000000000000
> kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> kernel: CR2: 0000000000532008 CR3: 000000000115c000 CR4: 00000000000406e0
> kernel: Call Trace:
> kernel:  devinet_sysctl_forward+0x1ac/0x1e0
> kernel:  proc_sys_call_handler+0x127/0x230
> kernel:  new_sync_write+0x114/0x1a0
> kernel:  vfs_write+0x18c/0x220
> kernel:  ksys_write+0x5a/0xd0
> kernel:  do_syscall_64+0x45/0x80
> kernel:  entry_SYSCALL_64_after_hwframe+0x44/0xae
> kernel: RIP: 0033:0x7fd4912b79b3
> kernel: Code: 8b 15 b9 74 0d 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb
> b7 0f 1f 00 64 8b 04 25 18 00 00 00 85 c0 75 14 b8 01 00 00 00 0f 05
> <48> 3d 00 f0 ff ff 77 55 c3 0f 1f 40 00 48 83 ec 28 48 89 54 24 18
> kernel: RSP: 002b:00007ffe96fdd858 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
> kernel: RAX: ffffffffffffffda RBX: 0000000000000002 RCX: 00007fd4912b79b3
> kernel: RDX: 0000000000000002 RSI: 0000000000536810 RDI: 0000000000000001
> kernel: RBP: 0000000000536810 R08: 000000000000000a R09: 0000000000000000
> kernel: R10: 00007fd49134f040 R11: 0000000000000246 R12: 0000000000000002
> kernel: R13: 00007fd4913906c0 R14: 00007fd49138c520 R15: 00007fd49138b920
> kernel: ---[ end trace ee7985b10570603d ]---
> kernel: ------------[ cut here ]------------

So the warning is easy to reproduce.
On qemu/kvm just set ctrl_guest_offloads=off for the device.

The panic does not seem to trigger for me and you did not provide
any data about it.  What happens? Does guest just freeze?

I am guessing the issue is that dev_disable_lro does not report the
return status and inet_forward_change assumes it's successful.  We then
end up with LRO packets in unexpected places.

Cc netdev and a bunch of people who might have a better idea.

-- 
MST

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization