On 12/19/14 at 12:40pm, Dave Young wrote:
> On 12/18/14 at 10:34am, WANG Chao wrote:
> > On 12/17/14 at 04:45pm, Dave Young wrote:
> > > On 12/17/14 at 04:40pm, Dave Young wrote:
> > > > On 12/16/14 at 11:23am, Prarit Bhargava wrote:
> > > > >
> > > > >
> > > > > On 11/17/2014 03:48 PM, Vivek Goyal wrote:
> > > > > > On Mon, Nov 17, 2014 at 11:13:50AM -0500, Prarit Bhargava
wrote:
> > > > > >> With the inclusion of 'panic_on_warn',
> > > > > >>
> > > > > >>
http://marc.info/?l=linux-api&m=141570937328528&w=2
> > > > > >>
> > > > > >> and which is now staged in Andrew Morton's tree, we
need to remove
> > > > > >> 'panic_on_warn' from the 2nd kernel's
cmdline. If it is included it is
> > > > > >> possible a non-fatal warning could panic the second
kernel.
> > > > > >
> > > > > > This patch looks good to me. I think we can commit it once
it show
> > > > > > up in Linus tree.
> > > > > >
> > > > > > Acked-by: Vivek Goyal <vgoyal(a)redhat.com>
> > > > >
> > > > > Dave and Vivek (who is on PTO),
> > > > >
> > > > > The panic_on_warn patch has been committed into linux.git as
> > > > > 9e3961a0979817c612b10b2da4f3045ec9faa779 ("kernel: add
panic_on_warn").
> > > >
> > > > Prarit, rethink about it. How about do something in kexec-tools c
code to remove
> > > > panic_on_warn for kdump kernel boot? Since it is for not only rh
distribution
> > > > it will save time for other people who do not notice the changes.
> > >
> > > We already have below:
> > > kexec-tools/kexec/kexec.c:
> > > 1055 if (kexec_flags & KEXEC_ON_CRASH)
> > > 1056 remove_parameter(line, "crashkernel");
> >
> > I don't think so. crashkernel=X is mandatory in kernel cmdline for
> > 1st kernel, but 2nd kernel definitely do not want it. That's why we have
> > to remove it at any case. That's why we do it in kexec.c
> >
> > I don't think it's a good idea to hard code removal of other kernel
> > parameters which won't hurt 2nd kernel, like panic_on_warn or something
> > like that. It's rare that one needs panic_on_warn on purpose in 2nd
> > kernel. But we shouldn't make the assumption that no one needs it.
>
> IMHO panic_on_warn is useless in 2nd kernel, it will cause vmcore capture failed
> due to some minor issues which can be ignored. It is harmful actually, if one
> forget to remove it in 2nd kernel cmdline he maybe has never chance to redo
> it again because bugs are not easily reproduced sometimes.
>
> You mentioned one can use it by purpose, but it can not convince me,
If removing panic_on_warn in kexec.c, user won't have a chance to put
this option to 2nd kernel. But if removing it in kexec-tools script,
user still have a chance to add it in KDUMP_CMDLINE_APPEND of
/etc/sysconfig/kdump.
Problem is if there's a case panic_on_warn is useful for 2nd kernel.
Let's see what's the opinion from other people, since this can be specified in
our kdump sysconfig file, anyway it is not a big issue.
Thanks
Dave