On 10/26/16 at 01:45pm, Dave Young wrote:
On 10/26/16 at 10:50am, Pratyush Anand wrote:
> Hi Peter,
>
> On Tuesday 25 October 2016 07:53 PM, Pratyush Anand wrote:
> > Moreover, with above additional two patches kexec-tools seems fine, but
> > still fedora-kernel-arm64/devel (with fh_defconfig+kexec enabled) did
> > not reboot on a seattle platform (Nothing after "Bye"). Will debug
and
> > come back on it.
>
> So couple of observation and so some queries:
>
> (a) Proposed kexec-tools patch with suggested additional patches works, but
> not with [1]. However, same kexec-tools did work when I used kernel from [2]
> which is based on 4.9-rc2. What is the kernel version that will be released
> in fc25? and which platform are you targeting? I can try to see the
> necessary diff which might cause the issue on mustang/seattle. Issue might
> not be related directly to kexec, and might be plat specific.
>
> (b) With current proposed kexec-tools, it takes more than 3 min to verify
> sha on seattle machine. If you want to do it quick (in a second) then you
> might need some more kexec-tools patches which are not yet approved in
> upstream. They are these [3].
Pratyush, I think for (b) we can just live with it, it maybe seattle
specific issues.
Copy the irc log here, I would prefer not to merge those patches even if
there is still issues at least it is the initial step of kexec/arm64 in
Fedora, we can fix it later.
I have some concerns about (b), maybe we can make it as default for
arm64 instead of adding an extra --enable-dcache if it happens on
all arm64 boards, anyway it should be discussed in upstream.
<panand> dyoung: and, see my reply to peter, not sure if d-cache patch
will also be needed for arm64..its 3+ min for sha verification currently
<dyoung> panand: see it
<dyoung> panand: but for mustang there's no such issues?
<panand> dyoung: for mustang also it will be slow
<dyoung> panand: ok
<panand> dyoung: I have not yet tested on mustang, but certainly it will
have same figure
<dyoung> panand: ok, so those patches are not kdump specific, kexec also
needs
<panand> dyoung: yes
<panand> dyoung: those are just to accelerate sha process
<dyoung> panand: and user should add an extra kexec argument to enable
it?
<panand> dyoung: yes --enable-dcache
<dyoung> panand: let's see how he responses, I think we should prefer
less fedora only pathces
<panand> dyoung: ok
<dyoung> panand: for those arguments, if anaconda is the user, then
ananconda will need change
<dyoung> panand: I'm not sure they will do it
Thanks
Dave
>
> >
> >
> > ~Pratyush
> >
> > [1]
https://git.fedorahosted.org/git/kernel-arm64.git : devel (or master)
> >
> > [2]
https://github.com/pratyushanand/linux/tree/upstream_arm64_devel
> >
> > [3]
https://github.com/pratyushanand/kexec-tools/commit/182706c9dd2e78d727343...
> >
https://github.com/pratyushanand/kexec-tools/commit/87d642cd983e68a2ca3c6...
> >
https://github.com/pratyushanand/kexec-tools/commit/a9473f1124d35f5144a5b...
> > _______________________________________________
> > kexec mailing list -- kexec(a)lists.fedoraproject.org
> > To unsubscribe send an email to kexec-leave(a)lists.fedoraproject.org
> _______________________________________________
> kexec mailing list -- kexec(a)lists.fedoraproject.org
> To unsubscribe send an email to kexec-leave(a)lists.fedoraproject.org