Re: [PATCH net-next v3 7/7] selftests: net: fdb_notify: Add a test for FDB notifications

[Petr Machata <petrm@nvidia.com>]

Jakub Kicinski <kuba@kernel.org> writes:

> On Mon, 11 Nov 2024 18:09:01 +0100 Petr Machata wrote:
>> Check that only one notification is produced for various FDB edit
>> operations.
>> 
>> Regarding the ip_link_add() and ip_link_master() helpers. This pattern of
>> action plus corresponding defer is bound to come up often, and a dedicated
>> vocabulary to capture it will be handy. tunnel_create() and vlan_create()
>> from forwarding/lib.sh are somewhat opaque and perhaps too kitchen-sinky,
>> so I tried to go in the opposite direction with these ones, and wrapped
>> only the bare minimum to schedule a corresponding cleanup.
>
> Looks like it fails about half of the time :(
>
> https://netdev.bots.linux.dev/flakes.html?min-flip=0&tn-needle=fdb-notify&br-cnt=200

OK, I can't reproduce this. Trying in VM, on an actual HW, debug, no
debug, no luck. But I see basically two failures:

- A "0 seen, 1 expected", which... I don't know, maybe it could just be
  a misplaced sleep. I don't see how, but it's a deterministing
  scenario, there shouldn't be anything racy here, either it emits or it
  doesn't, so some buffering issue is the only thing I can think of.

- Deadlocks. E.g. this, which looks like it deadlocked and timed out
  ("bad unlock balance detected" followed by "blocked for more than 122
  seconds" et.al.):

    https://netdev-3.bots.linux.dev/vmksft-net-dbg/results/846621/18-fdb-notify-sh/

  Like... how could this patchset even theoretically cause a deadlock?