[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <DM8PR11MB562191CF011875AC0E707096AB099@DM8PR11MB5621.namprd11.prod.outlook.com>
Date: Fri, 18 Nov 2022 14:30:09 +0000
From: "Jankowski, Konrad0" <konrad0.jankowski@...el.com>
To: ivecera <ivecera@...hat.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
CC: SlawomirX Laba <slawomirx.laba@...el.com>,
Eric Dumazet <edumazet@...gle.com>,
"moderated list:INTEL ETHERNET DRIVERS"
<intel-wired-lan@...ts.osuosl.org>,
open list <linux-kernel@...r.kernel.org>,
"Piotrowski, Patryk" <patryk.piotrowski@...el.com>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
"David S. Miller" <davem@...emloft.net>,
"sassmann@...hat.com" <sassmann@...hat.com>
Subject: RE: [Intel-wired-lan] [PATCH net] iavf: Do not restart Tx queues
after reset task failure
> -----Original Message-----
> From: Intel-wired-lan <intel-wired-lan-bounces@...osl.org> On Behalf Of Ivan
> Vecera
> Sent: Tuesday, November 8, 2022 11:25 AM
> To: netdev@...r.kernel.org
> Cc: SlawomirX Laba <slawomirx.laba@...el.com>; Eric Dumazet
> <edumazet@...gle.com>; moderated list:INTEL ETHERNET DRIVERS <intel-
> wired-lan@...ts.osuosl.org>; open list <linux-kernel@...r.kernel.org>;
> Piotrowski, Patryk <patryk.piotrowski@...el.com>; Jakub Kicinski
> <kuba@...nel.org>; Paolo Abeni <pabeni@...hat.com>; David S. Miller
> <davem@...emloft.net>; sassmann@...hat.com
> Subject: [Intel-wired-lan] [PATCH net] iavf: Do not restart Tx queues after reset
> task failure
>
> After commit aa626da947e9 ("iavf: Detach device during reset task") the device
> is detached during reset task and re-attached at its end.
> The problem occurs when reset task fails because Tx queues are restarted during
> device re-attach and this leads later to a crash.
>
> To resolve this issue properly close the net device in cause of failure in reset task
> to avoid restarting of tx queues at the end.
> Also replace the hacky manipulation with IFF_UP flag by device close that clears
> properly both IFF_UP and __LINK_STATE_START flags.
> In these case iavf_close() does not do anything because the adapter state is
> already __IAVF_DOWN.
>
> Reproducer:
> 1) Run some Tx traffic (e.g. iperf3) over iavf interface
> 2) Set VF trusted / untrusted in loop
>
> [root@...t ~]# cat repro.sh
> #!/bin/sh
>
> PF=enp65s0f0
> IF=${PF}v0
>
> ip link set up $IF
> ip addr add 192.168.0.2/24 dev $IF
> sleep 1
>
> iperf3 -c 192.168.0.1 -t 600 --logfile /dev/null & sleep 2
>
> while :; do
> ip link set $PF vf 0 trust on
> ip link set $PF vf 0 trust off
> done
> [root@...t ~]# ./repro.sh
>
> Result:
> [ 2006.650969] iavf 0000:41:01.0: Failed to init adminq: -53 [ 2006.675662] ice
> 0000:41:00.0: VF 0 is now trusted [ 2006.689997] iavf 0000:41:01.0: Reset task
> did not complete, VF disabled [ 2006.696611] iavf 0000:41:01.0: failed to allocate
> resources during reinit [ 2006.703209] ice 0000:41:00.0: VF 0 is now untrusted [
> 2006.737011] ice 0000:41:00.0: VF 0 is now trusted [ 2006.764536] ice
> 0000:41:00.0: VF 0 is now untrusted [ 2006.768919] BUG: kernel NULL pointer
> dereference, address: 0000000000000b4a [ 2006.776358] #PF: supervisor read
> access in kernel mode [ 2006.781488] #PF: error_code(0x0000) - not-present
> page [ 2006.786620] PGD 0 P4D 0 [ 2006.789152] Oops: 0000 [#1] PREEMPT SMP
> NOPTI [ 2006.792903] ice 0000:41:00.0: VF 0 is now trusted [ 2006.793501] CPU:
> 4 PID: 0 Comm: swapper/4 Kdump: loaded Not tainted 6.1.0-rc3+ #2 [
> 2006.805668] Hardware name: Abacus electric, s.r.o. - servis@...cus.cz Super
> Server/H12SSW-iN, BIOS 2.4 04/13/2022 [ 2006.815915] RIP:
> 0010:iavf_xmit_frame_ring+0x96/0xf70 [iavf] [ 2006.821028] ice 0000:41:00.0:
> VF 0 is now untrusted [ 2006.821572] Code: 48 83 c1 04 48 c1 e1 04 48 01 f9 48
> 83 c0 10 6b 50 f8 55 c1 ea 14 45 8d 64 14 01 48 39 c8 75 eb 41 83 fc 07 0f 8f e9 08
> 00 00 <0f> b7 45 4a 0f b7 55 48 41 8d 74 24 05 31 c9 66 39 d0 0f 86 da 00 [
> 2006.845181] RSP: 0018:ffffb253004bc9e8 EFLAGS: 00010293 [ 2006.850397]
> RAX: ffff9d154de45b00 RBX: ffff9d15497d52e8 RCX: ffff9d154de45b00 [
> 2006.856327] ice 0000:41:00.0: VF 0 is now trusted [ 2006.857523] RDX:
> 0000000000000000 RSI: 00000000000005a8 RDI: ffff9d154de45ac0 [
> 2006.857525] RBP: 0000000000000b00 R08: ffff9d159cb010ac R09:
> 0000000000000001 [ 2006.857526] R10: ffff9d154de45940 R11:
> 0000000000000000 R12: 0000000000000002 [ 2006.883600] R13:
> ffff9d1770838dc0 R14: 0000000000000000 R15: ffffffffc07b8380 [ 2006.885840]
> ice 0000:41:00.0: VF 0 is now untrusted [ 2006.890725] FS:
> 0000000000000000(0000) GS:ffff9d248e900000(0000) knlGS:0000000000000000
> [ 2006.890727] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [
> 2006.909419] CR2: 0000000000000b4a CR3: 0000000c39c10002 CR4:
> 0000000000770ee0 [ 2006.916543] PKRU: 55555554 [ 2006.918254] ice
> 0000:41:00.0: VF 0 is now trusted [ 2006.919248] Call Trace:
> [ 2006.919250] <IRQ>
> [ 2006.919252] dev_hard_start_xmit+0x9e/0x1f0 [ 2006.932587]
> sch_direct_xmit+0xa0/0x370 [ 2006.936424] __dev_queue_xmit+0x7af/0xd00 [
> 2006.940429] ip_finish_output2+0x26c/0x540 [ 2006.944519]
> ip_output+0x71/0x110 [ 2006.947831] ? __ip_finish_output+0x2b0/0x2b0 [
> 2006.952180] __ip_queue_xmit+0x16d/0x400 [ 2006.952721] ice 0000:41:00.0:
> VF 0 is now untrusted [ 2006.956098] __tcp_transmit_skb+0xa96/0xbf0 [
> 2006.965148] __tcp_retransmit_skb+0x174/0x860 [ 2006.969499] ?
> cubictcp_cwnd_event+0x40/0x40 [ 2006.973769]
> tcp_retransmit_skb+0x14/0xb0 ...
>
> Fixes: aa626da947e9 ("iavf: Detach device during reset task")
> Cc: Jacob Keller <jacob.e.keller@...el.com>
> Cc: Patryk Piotrowski <patryk.piotrowski@...el.com>
> Cc: SlawomirX Laba <slawomirx.laba@...el.com>
> Signed-off-by: Ivan Vecera <ivecera@...hat.com>
> ---
> drivers/net/ethernet/intel/iavf/iavf_main.c | 16 +++++++++++++++-
> 1 file changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c
> b/drivers/net/ethernet/intel/iavf/iavf_main.c
> index 5abcd66e7c7a..b66f8fa1d83b 100644
> --- a/drivers/net/ethernet/intel/iavf/iavf_main.c
> +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c
Tested-by: Konrad Jankowski <konrad0.jankowski@...el.com>
Powered by blists - more mailing lists