Hi Saeed,
On 01/22/21 at 05:14pm, Saeed Mirzamohammadi wrote:
> Hi,
>
> > On Jan 21, 2021, at 7:12 PM, Dave Young <[email protected]> wrote:
> >
> > On 01/22/21 at 09:22am, Dave Young wrote:
> >> Hi John,
> >>
> >> On 01/21/21 at 09:32am, [email protected] wrote:
> >>> On 11/22/20 9:47 PM, Dave Young wrote:
> >>>> Hi Guilherme,
> >>>> On 11/22/20 at 12:32pm, Guilherme Piccoli wrote:
> >>>>> Hi Dave and Kairui, thanks for your responses! OK, if that makes sense
> >>>>> to you I'm fine with it. I'd just recommend to test recent kernels in
> >>>>> multiple distros with the minimum "range" to see if 64M is enough for
> >>>>> crashkernel, maybe we'd need to bump that.
> >>>>
> >>>> Giving the different kernel configs and the different userspace
> >>>> initramfs setup it is hard to get an uniform value for all distributions,
> >>>> but we can have an interface/kconfig-option for them to provide a value like this patch
> >>>> is doing. And it could be improved like Kairui said about some known
> >>>> kernel added extra values later, probably some more improvements if
> >>>> doable.
> >>>>
> >>>> Thanks
> >>>> Dave
> >>>>
> >>>
> >>> Hi.
> >>>
> >>> Are we going to move forward with implementing this for X86 and Arm ?
> >>>
> >>> If other platform maintainers want to include this CONFIG option in their
> >>> configuration settings they have a starting point.
> >>
> >> I would expect this become arch independent.
> >
> > Clarify a bit, it can be a general config option under arch/Kconfig and
> > just put the code in general arch independent part.
>
> Does this mean that we need to add the option to def_configs in all archs as well?
>
I think we do not need to add defconfig, something like this will just work?
BTW, it should depend on CRASH_CORE instead of CRASH_DUMP, the logic of
parsing crashkernel is in kernel/crash_core.c
diff --git a/arch/Kconfig b/arch/Kconfig
index af14a567b493..fa6efeb52dc5 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -14,6 +14,11 @@ menu "General architecture-dependent options"
config CRASH_CORE
bool
+config CRASH_AUTO_STR
+ depends on CRASH_CORE
+ string "Memory reserved for crash kernel"
+ default "1G-:128M"
+ ... help text [snip] ...
+
config KEXEC_CORE
select CRASH_CORE
bool
[...]
> Thanks,
> Saeed
>
> >
> >>
> >> Saeed, Kairui, would any of you like to update the patch?
> >>
> >>>
> >>> Thank you,
> >>>
> >>> John.
> >>>
> >>> ( I am not currently on many of the included dist lists in this email, so
> >>> hopefully key contributors are included in this exchange )
> >>>
> >>
> >> Thanks
> >> Dave
>
Thanks
Dave
On 01/23/21 at 11:51am, Dave Young wrote:
> Hi Saeed,
> On 01/22/21 at 05:14pm, Saeed Mirzamohammadi wrote:
> > Hi,
> >
> > > On Jan 21, 2021, at 7:12 PM, Dave Young <[email protected]> wrote:
> > >
> > > On 01/22/21 at 09:22am, Dave Young wrote:
> > >> Hi John,
> > >>
> > >> On 01/21/21 at 09:32am, [email protected] wrote:
> > >>> On 11/22/20 9:47 PM, Dave Young wrote:
> > >>>> Hi Guilherme,
> > >>>> On 11/22/20 at 12:32pm, Guilherme Piccoli wrote:
> > >>>>> Hi Dave and Kairui, thanks for your responses! OK, if that makes sense
> > >>>>> to you I'm fine with it. I'd just recommend to test recent kernels in
> > >>>>> multiple distros with the minimum "range" to see if 64M is enough for
> > >>>>> crashkernel, maybe we'd need to bump that.
> > >>>>
> > >>>> Giving the different kernel configs and the different userspace
> > >>>> initramfs setup it is hard to get an uniform value for all distributions,
> > >>>> but we can have an interface/kconfig-option for them to provide a value like this patch
> > >>>> is doing. And it could be improved like Kairui said about some known
> > >>>> kernel added extra values later, probably some more improvements if
> > >>>> doable.
> > >>>>
> > >>>> Thanks
> > >>>> Dave
> > >>>>
> > >>>
> > >>> Hi.
> > >>>
> > >>> Are we going to move forward with implementing this for X86 and Arm ?
> > >>>
> > >>> If other platform maintainers want to include this CONFIG option in their
> > >>> configuration settings they have a starting point.
> > >>
> > >> I would expect this become arch independent.
> > >
> > > Clarify a bit, it can be a general config option under arch/Kconfig and
> > > just put the code in general arch independent part.
> >
> > Does this mean that we need to add the option to def_configs in all archs as well?
> >
>
> I think we do not need to add defconfig, something like this will just work?
>
> BTW, it should depend on CRASH_CORE instead of CRASH_DUMP, the logic of
> parsing crashkernel is in kernel/crash_core.c
>
> diff --git a/arch/Kconfig b/arch/Kconfig
> index af14a567b493..fa6efeb52dc5 100644
> --- a/arch/Kconfig
> +++ b/arch/Kconfig
> @@ -14,6 +14,11 @@ menu "General architecture-dependent options"
> config CRASH_CORE
> bool
>
> +config CRASH_AUTO_STR
> + depends on CRASH_CORE
> + string "Memory reserved for crash kernel"
> + default "1G-:128M"
People do not want to see the default value if they do not need kdump
so it would be better to add another kconfig option as a switch which is
set default as off in bool state.
> + ... help text [snip] ...
> +
> config KEXEC_CORE
> select CRASH_CORE
> bool
>
> [...]
>
> > Thanks,
> > Saeed
> >
> > >
> > >>
> > >> Saeed, Kairui, would any of you like to update the patch?
> > >>
> > >>>
> > >>> Thank you,
> > >>>
> > >>> John.
> > >>>
> > >>> ( I am not currently on many of the included dist lists in this email, so
> > >>> hopefully key contributors are included in this exchange )
> > >>>
> > >>
> > >> Thanks
> > >> Dave
> >
>
> Thanks
> Dave