On Tue, 2014-12-16 at 16:51 +0100, Bart Van Assche wrote:
> On 12/15/14 19:25, Vasu Dev wrote:
> > On Mon, 2014-12-15 at 16:22 +0100, Bart Van Assche wrote:
> >> Hello Vasu,
> >>
> >> Every now and then I receive an e-mail from an FCoE user who runs an
> >> older kernel reporting an issue that was fixed by the patches that were
> >> accepted upstream in kernel 3.13 (see also
> >> http://thread.gmane.org/gmane.linux.scsi.open-fcoe.devel/12132). 
> > 
> > Hi Bart,
> > 
> > The link is to your patches series applied earlier this year and not
> > about any reported issues with older kernels. I don't know why one would
> > report issues to you and not to open-fcoe.org with older stable kernels.
> > Am I missing such reports and how can we have them on open-fcoe also ?
> 
> Hello Vasu,
> 
> If someone creates a setup with a Linux FCoE initiator and a target
> system based on SCST and something does not work as expected usually a
> bug report is sent to the scst-devel mailing list without figuring out
> first which side was causing the unexpected behavior. An example can be
> found on http://sourceforge.net/p/scst/mailman/message/33130107/.
> 

I've seen CC to open-fcoe from both target (SCST & TCM) mails lists  for
an fcoe specific issue and I think target maintainer should CC to
open-fcoe once cause is ruled out of their target core and open-fcoe not
already on the email thread by issue reporter.
 
> > However that doesn't mean one would not run into any issue fixed
> > upstream while using older stable kernel w/o that fix, so critical
> > fixes should be CCed for stable kernel proactively but up to which
> > older kernels version ? I don't recall any specific policy for that
> > on either open-fcoe or linus-scsi, if any such linux-scsi policy then
> > we could consider that as well. Do you know what is the policy on
> > linux-scsi ?
> 
> I'm not sure but I think for linux-scsi everyone who posts a patch must
> decide himself whether or not that patch should be included in stable
> kernels by adding "Cc: <[email protected]>" in the patch itself.
> 

Yes that was my understanding also beside that I have seen stable kernel
maintainer Greg KH making call to pick fixes for stable kernel/s. I
think we should continue to use same process as used for linux-scsi
since now open-fcoe kernel patches are also generated directly against
linux-scsi gits. 
 
> >> This
> >> made me wonder what the policy is for sending patches to the stable tree
> >> ? The list of patches I would like to see being sent to the stable tree
> >> is as follows (tested on top of kernel v3.10.62):
> > 
> > Looks fine to apply these patches to older 3.10 if someone really wants
> > to use that specific v3.10.62 kernel.
> > 
> > I verified all these are in latest v3.18 kernel since these patches
> > dates back in 2013 with top and bottom patches in the list as:
> > 
> > commit e0335f67a281cb8eb11868e614ee9390fbbe9b1d
> > Author: Mark Rustad <[email protected]>
> > Date:   Sat May 18 04:01:36 2013 +0000
> > libfc: Reject PLOGI from nodes with incompatible role
> > 
> > 
> > commit 55d0ac5d2839fe270cea02fad44eed13750a0efd
> > Author: Neil Horman <[email protected]>
> > Date:   Tue Oct 8 23:43:58 2013 +0000
> > fcoe: Fix missing mutex_unlock in fcoe_sysfs_fcf_add error path
> > 
> > You may also want to include patches from your link above and possibly
> > some more missing for 3.10 stable, so if possibly then better for user
> > to move to v3.18 which has all recent fcoe fixes in it.
> 
> Agreed that moving to a more recent kernel would be the best option.
> However, it is a concern to me is that RHEL 7 is based on kernel 3.10.
> Maybe the RHEL maintainers keep an eye on which patches are sent to the
> stable branch and sending the relevant patches to the stable tree helps
> to get these patches included in RHEL 7 ?

Yes RHEL maintainers sync with patches on their releases beside there
are times they have applied fcoe fixes to their maintenance updates thru
their z-stream. Intel also helps distros (RHEL, SLES ..) to identify
pending fixes and getting in sync with their releases but that is all
out side open-fcoe working primary scope. Open-fcoe is for upstream only
and likewise the case with any other Linux subsystems/drivers upstream
work.

Thanks Bart for the discussion and I hope this will help all understand
open-fcoe process better.

        Vasu   

> 
> Bart.
> 


_______________________________________________
fcoe-devel mailing list
[email protected]
http://lists.open-fcoe.org/mailman/listinfo/fcoe-devel

Reply via email to