Vasudev Kamath <vasu...@debian.org> writes: > Salvatore Bonaccorso <car...@debian.org> writes: > >> Hi >> >> On Tue, Sep 15, 2020 at 02:00:51PM +0530, Vasudev Kamath wrote: >>> Tzafrir Cohen <tzaf...@cohens.org.il> writes: >>> >>> > Hi, >>> > >>> > A patch for fix of the regression is in 4.19.145 (commit >>> > 044be307e550b4532960eadabfb6942de96751f0 "net/mlx5e: Don't support phys >>> > switch id if not in switchdev mode"). >>> > >>> > Please enable CONFIG_NET_SWITCHDEV once this is merged to the Buster >>> > kernel tree. >>> > >>> > -- Tzafrir >>> >>> I just created another ticket as this bug is already archived [1]. And >>> mail is not getting tracked in the BTS. >>> >>> [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=970350 >> >> Please have look at >> https://salsa.debian.org/kernel-team/linux/-/merge_requests/278 >> >> Still it should be evaluated if this has no other impact as it was >> previously only enabled for armhf. > > We are already using this in our production with custom compiled kernel > based on kernel in 10.4.(one with regression issue still present but > workaround done manually) (Currently we have 50+ machines running this > and will eventually increase it). > > I will check this with current buster kernel from 10.5 where regression > issue fix is applied and update here.
I compiled 10.6 kernel 4.19.152 with your MR patch and tested on our setup. I can confirm no more regression issue is found on both network setup we use (bonding and non-bonding). As for the NET_SWITCHDEV we have this enabled in our production machine from kernel compiled in 10.3 with this patch. And its been running for 2 months or so without any issue. Please consider merging this patch and release it as part of Debian 10.7 release. Cheers, Vasudev