Message-ID: <d44c633d-d1ef-4f9c-bcac-554ae6a4c661@redhat.com>
Date: Thu, 27 Jun 2024 08:20:32 -0400
From: Luiz Capitulino <luizcap@...hat.com>
To: Maciej Fijalkowski <maciej.fijalkowski@...el.com>, xmu@...hat.com
Cc: jesse.brandeburg@...el.com, anthony.l.nguyen@...el.com, poros@...hat.com,
 netdev@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>,
 magnus.karlsson@...el.com
Subject: Re: [BUG] ice driver crash on arm64

On 2024-06-26 08:53, Luiz Capitulino wrote:
> On 2024-06-26 05:19, Maciej Fijalkowski wrote:
>> On Thu, Jun 20, 2024 at 09:23:42AM -0400, Luiz Capitulino wrote:
>>> On 2024-06-20 09:19, Maciej Fijalkowski wrote:
>>>> On Tue, Jun 18, 2024 at 11:23:28AM -0400, Luiz Capitulino wrote:
>>>>> Hi,
>>>>>
>>>>> We have an Ampere Mount Snow system (which is arm64) with an Intel E810-C
>>>>> NIC plugged in. The kernel is configured with 64k pages. We're observing
>>>>> the crash below when we run iperf3 as a server in this system and load traffic
>>>>> from another system with the same configuration. The crash is reproducible
>>>>> with latest Linus tree 14d7c92f:
>>>>>
>>>>> [  225.715759] Unable to handle kernel paging request at virtual address 0075e625f68aa42c
>>>>> [  225.723669] Mem abort info:
>>>>> [  225.726487]   ESR = 0x0000000096000004
>>>>> [  225.730223]   EC = 0x25: DABT (current EL), IL = 32 bits
>>>>> [  225.735526]   SET = 0, FnV = 0
>>>>> [  225.738568]   EA = 0, S1PTW = 0
>>>>> [  225.741695]   FSC = 0x04: level 0 translation fault
>>>>> [  225.746564] Data abort info:
>>>>> [  225.749431]   ISV = 0, ISS = 0x00000004, ISS2 = 0x00000000
>>>>> [  225.754906]   CM = 0, WnR = 0, TnD = 0, TagAccess = 0
>>>>> [  225.759944]   GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
>>>>> [  225.765250] [0075e625f68aa42c] address between user and kernel address ranges
>>>>> [  225.772373] Internal error: Oops: 0000000096000004 [#1] SMP
>>>>> [  225.777932] Modules linked in: xfs(E) crct10dif_ce(E) ghash_ce(E) sha2_ce(E) sha256_arm64(E) sha1_ce(E) sbsa_gwdt(E) ice(E) nvme(E) libie(E) dimlib(E) nvme_core(E) gnss(E) nvme_auth(E) ixgbe(E) igb(E) mdio(E) i2c_algo_bit(E) i2c_designware_platform(E) xgene_hwmon(E) i2c_designware_core(E) dm_mirror(E) dm_region_hash(E) dm_log(E) dm_mod(E)
>>>>> [  225.807902] CPU: 61 PID: 7794 Comm: iperf3 Kdump: loaded Tainted: G            E      6.10.0-rc4+ #1
>>>>> [  225.817021] Hardware name: LTHPC GR2134/MP32-AR2-LT, BIOS F31j (SCP: 2.10.20220531) 08/01/2022
>>>>> [  225.825618] pstate: 00400009 (nzcv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>>>>> [  225.832566] pc : __arch_copy_to_user+0x4c/0x240
>>>>> [  225.837088] lr : _copy_to_iter+0x104/0x518
>>>>> [  225.841173] sp : ffff80010978f6e0
>>>>> [  225.844474] x29: ffff80010978f730 x28: 0000000000007388 x27: 4775e625f68aa42c
>>>>> [  225.851597] x26: 0000000000000001 x25: 00000000000005a8 x24: 00000000000005a8
>>>>> [  225.858720] x23: 0000000000007388 x22: ffff80010978fa60 x21: ffff80010978fa60
>>>>> [  225.865842] x20: 4775e625f68aa42c x19: 0000000000007388 x18: 0000000000000000
>>>>> [  225.872964] x17: 0000000000000000 x16: 0000000000000000 x15: 4775e625f68aa42c
>>>>> [  225.880087] x14: aaa03e61c262c44f x13: 5fb01a5ebded22da x12: 415feff815830f22
>>>>> [  225.887209] x11: 7411a8ffaab6d3d7 x10: 95af4645d12e6d70 x9 : ffffba83c2faddac
>>>>> [  225.894332] x8 : c1cbcc6e9552ed64 x7 : dfcefe933cdc57ae x6 : 0000fffde5aa9e80
>>>>> [  225.901454] x5 : 0000fffde5ab1208 x4 : 0000000000000004 x3 : 0000000000016180
>>>>> [  225.908576] x2 : 0000000000007384 x1 : 4775e625f68aa42c x0 : 0000fffde5aa9e80
>>>>> [  225.915699] Call trace:
>>>>> [  225.918132]  __arch_copy_to_user+0x4c/0x240
>>>>> [  225.922304]  simple_copy_to_iter+0x4c/0x78
>>>>> [  225.926389]  __skb_datagram_iter+0x18c/0x270
>>>>> [  225.930647]  skb_copy_datagram_iter+0x4c/0xe0
>>>>> [  225.934991]  tcp_recvmsg_locked+0x59c/0x9a0
>>>>> [  225.939162]  tcp_recvmsg+0x78/0x1d0
>>>>> [  225.942638]  inet6_recvmsg+0x54/0x128
>>>>> [  225.946289]  sock_recvmsg+0x78/0xd0
>>>>> [  225.949766]  sock_read_iter+0x98/0x108
>>>>> [  225.953502]  vfs_read+0x2a4/0x318
>>>>> [  225.956806]  ksys_read+0xec/0x110
>>>>> [  225.960108]  __arm64_sys_read+0x24/0x38
>>>>> [  225.963932]  invoke_syscall.constprop.0+0x80/0xe0
>>>>> [  225.968624]  do_el0_svc+0xc0/0xe0
>>>>> [  225.971926]  el0_svc+0x48/0x1b0
>>>>> [  225.975056]  el0t_64_sync_handler+0x13c/0x158
>>>>> [  225.979400]  el0t_64_sync+0x1a4/0x1a8
>>>>> [  225.983051] Code: 78402423 780008c3 910008c6 36100084 (b8404423)
>>>>> [  225.989132] SMP: stopping secondary CPUs
>>>>> [  225.995919] Starting crashdump kernel...
>>>>> [  225.999829] Bye!
>>>>>
>>>>> I was able to find out this is actually a regression introduced in 6.3-rc1
>>>>> and was able to bisect it down to commit:
>>>>>
>>>>> commit 1dc1a7e7f4108bad4af4c7c838f963d342ac0544
>>>>> Author: Maciej Fijalkowski <maciej.fijalkowski@...el.com>
>>>>> Date:   Tue Jan 31 21:44:59 2023 +0100
>>>>>
>>>>>       ice: Centrallize Rx buffer recycling
>>>>>
>>>>>       Currently calls to ice_put_rx_buf() are sprinkled through
>>>>>       ice_clean_rx_irq() - first place is for explicit flow director's
>>>>>       descriptor handling, second is after running XDP prog and the last one
>>>>>       is after taking care of skb.
>>>>>
>>>>>       1st callsite was actually only for ntc bump purpose, as Rx buffer to be
>>>>>       recycled is not even passed to a function.
>>>>>
>>>>>       It is possible to walk through Rx buffers processed in particular NAPI
>>>>>       cycle by caching ntc from beginning of the ice_clean_rx_irq().
>>>>>
>>>>>       To do so, let us store XDP verdict inside ice_rx_buf, so action we need
>>>>>       to take on will be known. For XDP prog absence, just store ICE_XDP_PASS
>>>>>       as a verdict.
>>>>>
>>>>>       Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@...el.com>
>>>>>       Signed-off-by: Daniel Borkmann <daniel@...earbox.net>
>>>>>       Reviewed-by: Alexander Lobakin <alexandr.lobakin@...el.com>
>>>>>       Link: https://lore.kernel.org/bpf/20230131204506.219292-7-maciej.fijalkowski@intel.com
>>>>>
>>>>> Some interesting/important information:
>>>>>
>>>>>    * Issue doesn't reproduce when:
>>>>>       - The kernel is configured w/ 4k pages
>>>>>       - UDP is used (ie. iperf3 -c <server> -u -b 0)
>>>>>       - legacy-rx is set
>>>>>    * The NIC firmware is version 4.30 (we haven't figured out how to update it from arm)
>>>>>
>>>>> By taking a quick look at the code, ICE_LAST_OFFSET in ice_can_reuse_rx_page() seems
>>>>> wrong since Rx buffers are 3k w/ bigger page sizes but just changing it to
>>>>> ICE_RXBUF_3072 doesn't fix the issue.
>>>>>
>>>>> Could you please help taking a look?
>>>>
>>>> Thanks for the report. I am on sick leave currently, will try to take
>>>> alook once I'm back.
>>>
>>> I'm sorry to hear that Maciej, I hope you get well soon.
>>
>> Thanks I'm back now. Can you tell us also what MTU is used for these
>> tests? Our validation is working on repro now.
> 
> Xiumei,
> 
> Can you take a look now that you got the machine back?

Strangely, I don't see Xiumei's answer on lore (although I did receive the
email). Just in case, the MTU is 1500.

- Luiz

Powered by blists - more mailing lists