[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <39856d06-c114-47c5-92a6-4bacbdd1fc49@davidwei.uk>
Date: Mon, 10 Jun 2024 15:02:04 -0700
From: David Wei <dw@...idwei.uk>
To: Maciek Machnikowski <maciek@...hnikowski.net>
Cc: Jakub Kicinski <kuba@...nel.org>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [BUG] [PATCH net-next v3 0/2] netdevsim: add NAPI support
On 2024-06-10 15:00, Maciek Machnikowski wrote:
>
>
>> On 6 Jun 2024, at 16:31, David Wei <dw@...idwei.uk> wrote:
>>
>> On 2024-06-05 12:38, Maciek Machnikowski wrote:
>>>
>>>
>>> On 24/04/2024 04:36, David Wei wrote:
>>>> Add NAPI support to netdevsim and register its Rx queues with NAPI
>>>> instances. Then add a selftest using the new netdev Python selftest
>>>> infra to exercise the existing Netdev Netlink API, specifically the
>>>> queue-get API.
>>>>
>>>> This expands test coverage and further fleshes out netdevsim as a test
>>>> device. It's still my goal to make it useful for testing things like
>>>> flow steering and ZC Rx.
>>>>
>>>> -----
>>>> Changes since v2:
>>>> * Fix null-ptr-deref on cleanup path if netdevsim is init as VF
>>>> * Handle selftest failure if real netdev fails to change queues
>>>> * Selftest addremove_queue test case:
>>>> * Skip if queues == 1
>>>> * Changes either combined or rx queue depending on how the netdev is
>>>> configured
>>>>
>>>> Changes since v1:
>>>> * Use sk_buff_head instead of a list for per-rq skb queue
>>>> * Drop napi_schedule() if skb queue is not empty in napi poll
>>>> * Remove netif_carrier_on() in open()
>>>> * Remove unused page pool ptr in struct netdevsim
>>>> * Up the netdev in NetDrvEnv automatically
>>>> * Pass Netdev Netlink as a param instead of using globals
>>>> * Remove unused Python imports in selftest
>>>
>>> Hi!
>>>
>>> This change breaks netdevsim on my setup.
>>> Tested on Parallels ARM VM running on Mac with Fedora 40.
>>>
>>> When using netdevsim from the latest 6.10-rc2 (and -rc1) I can't pass
>>> any traffic (not completing any pings) nor complete
>>> tools/testing/selftests/drivers/net/netdevsim/peer.sh test (the test
>>> hangs at socat step trying to send anything through).
>>
>> Hi Maciek, I'm trying to reproduce the issue.
>>
>> Can you please share how you're setting up netdevsim to pass traffic?
>
> I modified peer.sh to stop after creating and linking two netdevsim adapters
> and then ran the ping from one namespace to another. The same script
> and procedure worked fine when running 6.9.3, but failed with any 6.10-rc
> releases. And the only delta between these two netdevsims is this patch.
Thanks. I'll try and reproduce on my end.
>
>
Powered by blists - more mailing lists