Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 1 May 2019 10:28:05 +0300
From:      Slava Shwartsman <slavash@FreeBSD.org>
To:        John Baldwin <jhb@FreeBSD.org>, Hans Petter Selasky <hps@selasky.org>, "Andrey V. Elsukov" <bu7cher@yandex.ru>, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   Re: svn commit: r341586 - head/sys/dev/mlx5/mlx5_en
Message-ID:  <737bfa62-eb3e-7c48-6356-0e0b76bb0593@FreeBSD.org>
In-Reply-To: <0e80cb1d-20b9-5b8c-330e-5f64a603ff10@FreeBSD.org>
References:  <201812051425.wB5EP38T004562@repo.freebsd.org> <a1ff0879-abae-25c5-9350-809186d2cf85@yandex.ru> <bc601976-7d6b-f849-0305-66feb11a7130@selasky.org> <7cea5305-4136-a0b6-487b-51307b1c6db9@FreeBSD.org> <123654bf-59f7-1db8-55ce-36306bdac43d@FreeBSD.org> <67b638e0-8e6f-f960-b263-62eed8ce28b7@FreeBSD.org> <0e80cb1d-20b9-5b8c-330e-5f64a603ff10@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help


On 01-May-19 10:09, Slava Shwartsman wrote:
> 
> 
> On 30-Apr-19 00:14, John Baldwin wrote:
>> On 4/25/19 12:10 AM, Slava Shwartsman wrote:
>>>
>>>
>>> On 17-Apr-19 00:28, John Baldwin wrote:
>>>> On 4/16/19 8:32 AM, Hans Petter Selasky wrote:
>>>>> On 4/16/19 4:39 PM, Andrey V. Elsukov wrote:
>>>>>> On 05.12.2018 17:25, Slava Shwartsman wrote:
>>>>>>> Author: slavash
>>>>>>> Date: Wed Dec  5 14:25:03 2018
>>>>>>> New Revision: 341586
>>>>>>> URL: https://svnweb.freebsd.org/changeset/base/341586
>>>>>>>
>>>>>>> Log:
>>>>>>>      mlx5en: Implement backpressure indication.
>>>>>>>      The backpressure indication is implemented using an 
>>>>>>> unlimited rate type of
>>>>>>>      mbuf send tag. When the upper layers typically the socket 
>>>>>>> layer has obtained such
>>>>>>>      a tag, it can then query the destination driver queue for 
>>>>>>> the current
>>>>>>>      amount of space available in the send queue.
>>>>>>>      A single mbuf send tag may be referenced multiple times and 
>>>>>>> a refcount has been added
>>>>>>>      to the mlx5e_priv structure to track its usage. Because the 
>>>>>>> send tag resides
>>>>>>>      in the mlx5e_channel structure, there is no need to wait for 
>>>>>>> refcounts to reach
>>>>>>>      zero until the mlx4en(4) driver is detached. The channels 
>>>>>>> structure is persistant
>>>>>>>      during the lifetime of the mlx5en(4) driver it belongs to 
>>>>>>> and can so be accessed
>>>>>>>      without any need of synchronization.
>>>>>>>      The mlx5e_snd_tag structure was extended to contain a type 
>>>>>>> field, because there are now
>>>>>>>      two different tag types which end up in the driver which 
>>>>>>> need to be distinguished.
>>>>>>>      Submitted by:   hselasky@
>>>>>>>      Approved by:    hselasky (mentor)
>>>>>>>      MFC after:      1 week
>>>>>>>      Sponsored by:   Mellanox Technologies
>>>>>>> @@ -587,27 +609,33 @@ mlx5e_xmit(struct ifnet *ifp, struct mbuf *mb)
>>>>>>>         struct mlx5e_sq *sq;
>>>>>>>         int ret;
>>>>>>> -    sq = mlx5e_select_queue(ifp, mb);
>>>>>>> -    if (unlikely(sq == NULL)) {
>>>>>>> -#ifdef RATELIMIT
>>>>>>> -        /* Check for route change */
>>>>>>> -        if (mb->m_pkthdr.snd_tag != NULL &&
>>>>>>> -            mb->m_pkthdr.snd_tag->ifp != ifp) {
>>>>>>> +    if (mb->m_pkthdr.snd_tag != NULL) {
>>>>>>> +        sq = mlx5e_select_queue_by_send_tag(ifp, mb);
>>>>>>> +        if (unlikely(sq == NULL)) {
>>>>>>> +            /* Check for route change */
>>>>>>> +            if (mb->m_pkthdr.snd_tag->ifp != ifp) {
>>>>>>> +                /* Free mbuf */
>>>>>>> +                m_freem(mb);
>>>>>>> +
>>>>>>> +                /*
>>>>>>> +                 * Tell upper layers about route
>>>>>>> +                 * change and to re-transmit this
>>>>>>> +                 * packet:
>>>>>>> +                 */
>>>>>>> +                return (EAGAIN);
>>>>>>> +            }
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> I just discovered something strange and found that this commit is the
>>>>>> cause.
>>>>>> The test system has mlx5en 100G interface. It has two vlans: 
>>>>>> vlan500 and
>>>>>> vlan100.
>>>>>> Via vlan500 it receives some packets flows. Then it routes these 
>>>>>> packets
>>>>>> into vlan100.
>>>>>> But packets are dropped in mlx5e_xmit() with EAGAIN error code.
>>>>>>
>>>>>> # dtrace -n 'fbt::ip6_output:return {printf("%d", arg1);}'
>>>>>> dtrace: description 'fbt::ip6_output:return ' matched 1 probe
>>>>>> CPU     ID                    FUNCTION:NAME
>>>>>>     23  54338                ip6_output:return 35
>>>>>>     16  54338                ip6_output:return 35
>>>>>>     21  54338                ip6_output:return 35
>>>>>>     22  54338                ip6_output:return 35
>>>>>>     24  54338                ip6_output:return 35
>>>>>>     23  54338                ip6_output:return 35
>>>>>>     14  54338                ip6_output:return 35
>>>>>> ^C
>>>>>>
>>>>>> # dtrace -n 'fbt::mlx5e_xmit:return {printf("%d", arg1);}'
>>>>>> dtrace: description 'fbt::mlx5e_xmit:return ' matched 1 probe
>>>>>> CPU     ID                    FUNCTION:NAME
>>>>>>     16  69030                mlx5e_xmit:return 35
>>>>>>     23  69030                mlx5e_xmit:return 35
>>>>>>     26  69030                mlx5e_xmit:return 35
>>>>>>     25  69030                mlx5e_xmit:return 35
>>>>>>     24  69030                mlx5e_xmit:return 35
>>>>>>     21  69030                mlx5e_xmit:return 35
>>>>>>     26  69030                mlx5e_xmit:return 35
>>>>>> ^C
>>>>>>
>>>>>> The kernel config is GENERIC.
>>>>>> 13.0-CURRENT #9 r345758+82f3d57(svn_head)-dirty
>>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>> This might be a case where rcvif in the mbuf's pktheader is not 
>>>>> cleared
>>>>> before the packet is fed back on the wire.
>>>>>
>>>>> John Baldwin is working on the send tags implementation, to eliminate
>>>>> the EAGAIN handling in the network drivers.
>>>>
>>>> I will try to push this branch sooner then since it affects more 
>>>> than just
>>>> TLS.  Part of the change includes a new flag we can use to assert 
>>>> that we
>>> Thanks John!
>>>> aren't just getting a stale rcvif (though there are also now 
>>>> assertions in
>>>> ip_output that should catch this case I think).
>>>>
>>>
>>>
>>>
>>> Hi Andrey,
>>>
>>> Yes, we were able to reproduce this issue in house. If you don't mind, I
>>> prefer to wait for John's update - where he eliminates the EAGAIN
>>> handling in the network drivers.
>>
>> I have rebased the branch for this, but for now it will just panic sooner
>> I believe by tripping an assertion.  Can you grab the diff (or just 
>> the branch)
>> from the 'send_tags' branch at github/bsdjhb/freebsd and reproduce 
>> under a
>> kernel with INVARIANTS?  I think we will have to explicitly clear the 
>> 'rcvif'
>> pointer somewhere, but I want to see what the stack trace looks like 
>> so I can
>> think about the "right" place to clear it.
>>
> 
> Hi John,
> 
> I grabbed your branch (which doesn't build BTW due to libbe(3): Fix 
> mis-application of patch (SHLIBDIR) so I just reverted it).
> 
> The kernel doesn't panic in this scenario - it just that the packets are 
> being dropped. So I added a kdb_backtrace right before the return 
> (EAGAIN) in mlx5e_xmit:
> 
> KDB: stack backtrace:
> db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 
> 0xfffffe0000547d90
> mlx5e_xmit() at mlx5e_xmit+0x3d/frame 0xfffffe0000548160
> vlan_transmit() at vlan_transmit+0xdc/frame 0xfffffe00005481d0
> ether_output_frame() at ether_output_frame+0xa2/frame 0xfffffe0000548200
> ether_output() at ether_output+0x689/frame 0xfffffe00005482a0
> ip_output() at ip_output+0x13a4/frame 0xfffffe00005483f0
> ip_forward() at ip_forward+0x344/frame 0xfffffe00005484b0
> ip_input() at ip_input+0x7f5/frame 0xfffffe0000548560
> netisr_dispatch_src() at netisr_dispatch_src+0xa2/frame 0xfffffe00005485d0
> ether_demux() at ether_demux+0x147/frame 0xfffffe0000548600
> ether_nh_input() at ether_nh_input+0x403/frame 0xfffffe0000548660
> netisr_dispatch_src() at netisr_dispatch_src+0xa2/frame 0xfffffe00005486d0
> ether_input() at ether_input+0x73/frame 0xfffffe0000548700
> vlan_input() at vlan_input+0x1e7/frame 0xfffffe0000548750
> ether_demux() at ether_demux+0x12d/frame 0xfffffe0000548780
> ether_nh_input() at ether_nh_input+0x403/frame 0xfffffe00005487e0
> netisr_dispatch_src() at netisr_dispatch_src+0xa2/frame 0xfffffe0000548850
> ether_input() at ether_input+0x73/frame 0xfffffe0000548880
> mlx5e_rx_cq_comp() at mlx5e_rx_cq_comp+0x8b4/frame 0xfffffe00005489a0
> mlx5_cq_completion() at mlx5_cq_completion+0x5e/frame 0xfffffe00005489d0
> mlx5_msix_handler() at mlx5_msix_handler+0x1ba/frame 0xfffffe0000548a10
> ithread_loop() at ithread_loop+0x187/frame 0xfffffe0000548a70
> fork_exit() at fork_exit+0x84/frame 0xfffffe0000548ab0
> fork_trampoline() at fork_trampoline+0xe/frame 0xfffffe0000548ab0
> --- trap 0, rip = 0, rsp = 0, rbp = 0 ---
> 
> 
> Please ping me if you want me to try anything else.
> 
> 
> Slava

My bad - tested with master. Re-testing now.


Slava



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?737bfa62-eb3e-7c48-6356-0e0b76bb0593>