[DRBD-user] drbd build failure due to missing REQ_OP_WRITE_SAME symbol

Reid Wahl nwahl at redhat.com
Tue Jul 26 21:31:21 CEST 2022


On Tue, Jul 26, 2022 at 2:11 AM Christoph Böhmwalder
<christoph.boehmwalder at linbit.com> wrote:
>
> Am 26.07.22 um 03:24 schrieb Reid Wahl:
> > drbd_csum_bio() in drbd_sender.c uses a constant (REQ_OP_WRITE_SAME)
> > that was removed from the kernel in February:
> > - https://github.com/LINBIT/drbd/blob/drbd-9.1/drbd/drbd_sender.c#L360-L361
> >
> > Here's where the constant was removed:
> > - https://github.com/torvalds/linux/commit/73bd66d9#diff-3b0e31d21eab4c9595b9d14730e06067f27b9f21134edcd9cb47215d23d69583
> >
> > I'm using CentOS Stream 9 with kernel-5.14.0-130.el9.x86_64. I'm aware
> > that CentOS Stream "is not one of the distributions we care about too
> > deeply." I'm not sure to what extent we care about it though. It would
> > be nice to be able to build from the latest drbd upstream using a
> > recent kernel.
> >
> > Please let me know if this is something we can get fixed.
> >
>
> Hi Reid,
>
> it's not only about caring, this is mostly an issue of time. Basically,
> we race to support the "important" kernels (i.e. the ones our paying
> customers request), but that takes up a lot of time already, so there is
> not a lot left for more "exotic" kernels.
>
> Whenever we are already compatible with the most recent "customer
> relevant kernel", I just start going through the remaining patches in
> chronological order. If they are easy to port, they get done pretty
> quickly. If they are more convoluted (like the recent bio_alloc mess),
> it might take more time.
> What I'm trying to say is that it's tough to even give a rough estimate
> on when a specific patch will get ported to out-of-tree.

I appreciate the detailed response. I completely get the need to
prioritize stable/important kernels.

>
> The only thing I can share – if it provides any solace – is that the
> patch you are referencing is currently #3 in the queue, so it will
> possibly get done in the near future. Again, unfortunately I can't make
> any promises.

It's not that big of a deal :) However one may feel about CentOS's
direction, users of Stream have to expect that behavior may not always
be smooth.

I've been updating some documentation that was written for Centos 8
Stream (c8s), and thus far I've been unable to make drbd work on c9s.
The kmod-drbd9x RPM package from ELRepo isn't compatible (and won't be
made compatible) with the c9s kernel. The latest stable drbd release
tarball hits a bio_alloc issue (and possibly others). The current git
drbd requires that I build my own coccinelle/spatch (no RPM package
for EL9), but after I do that, I hit the REQ_OP_WRITE_SAME issue from
the first message.

We're just going to have to use Alma Linux in place of c9s.

>
> ... and this is why we'll all be happy when DRBD is finally *only* in
> the upstream kernel again :)
>
> --
> Christoph Böhmwalder
> LINBIT | Keeping the Digital World Running
> DRBD HA —  Disaster Recovery — Software defined Storage
>



-- 
Regards,

Reid Wahl (He/Him)
Senior Software Engineer, Red Hat
RHEL High Availability - Pacemaker



More information about the drbd-user mailing list