Request for inclusion of fb90a6e93c0684 ("sched/debug: Don't dump sched debug info in SysRq-W")

Request for inclusion of fb90a6e93c0684 ("sched/debug: Don't dump sched debug info in SysRq-W")

[Nikolay Borisov]

Hello,

I'd like to ask for the patch in the subject to be included in stable
kernels. Without it if one wants to see only the blocked stats this
becomes impossible since the information dumped by sched_debug is so
much that it overflows the kmsg buffer and information about blocked
stats is overwritten. This is especially a problem in distro kernels
which are based on 4.4 and have enabled CONFIG_SCHED_DEBUG - e.g. ubuntu.

Regards,
Nikolay

Re: Request for inclusion of fb90a6e93c0684 ("sched/debug: Don't dump sched debug info in SysRq-W")

[Ingo Molnar]

* Nikolay Borisov <n.borisov.lkml@gmail.com> wrote:

> Hello,
> 
> I'd like to ask for the patch in the subject to be included in stable
> kernels. Without it if one wants to see only the blocked stats this
> becomes impossible since the information dumped by sched_debug is so
> much that it overflows the kmsg buffer and information about blocked
> stats is overwritten. This is especially a problem in distro kernels
> which are based on 4.4 and have enabled CONFIG_SCHED_DEBUG - e.g. ubuntu.

I'm fine with including fb90a6e93c0684 in -stable, assuming you tested it and 
such.

	Ingo

Re: Request for inclusion of fb90a6e93c0684 ("sched/debug: Don't dump sched debug info in SysRq-W")

[Peter Zijlstra]

On Wed, Feb 15, 2017 at 11:09:14AM +0200, Nikolay Borisov wrote:
> Hello,
> 
> I'd like to ask for the patch in the subject to be included in stable
> kernels. Without it if one wants to see only the blocked stats this
> becomes impossible since the information dumped by sched_debug is so
> much that it overflows the kmsg buffer and information about blocked
> stats is overwritten. This is especially a problem in distro kernels
> which are based on 4.4 and have enabled CONFIG_SCHED_DEBUG - e.g. ubuntu.

Who cares about the kmsg buffer? Also, just make it bigger or disable
SCHED_DEBUG, as the name implies it _is_ a debug feature after all.

Re: Request for inclusion of fb90a6e93c0684 ("sched/debug: Don't dump sched debug info in SysRq-W")

[Nikolay Borisov]

On 15.02.2017 11:13, Peter Zijlstra wrote:
> On Wed, Feb 15, 2017 at 11:09:14AM +0200, Nikolay Borisov wrote:
>> Hello,
>>
>> I'd like to ask for the patch in the subject to be included in stable
>> kernels. Without it if one wants to see only the blocked stats this
>> becomes impossible since the information dumped by sched_debug is so
>> much that it overflows the kmsg buffer and information about blocked
>> stats is overwritten. This is especially a problem in distro kernels
>> which are based on 4.4 and have enabled CONFIG_SCHED_DEBUG - e.g. ubuntu.
> 
> Who cares about the kmsg buffer? Also, just make it bigger or disable
> SCHED_DEBUG, as the name implies it _is_ a debug feature after all.

but what if you want to use your distro's kernel and deal with custom
recompiles. The kernel ought to come with sensible defaults.

> 

Re: Request for inclusion of fb90a6e93c0684 ("sched/debug: Don't dump sched debug info in SysRq-W")

[Greg KH]

On Wed, Feb 15, 2017 at 11:17:34AM +0200, Nikolay Borisov wrote:
> On 15.02.2017 11:13, Peter Zijlstra wrote:
> > On Wed, Feb 15, 2017 at 11:09:14AM +0200, Nikolay Borisov wrote:
> >> Hello,
> >>
> >> I'd like to ask for the patch in the subject to be included in stable
> >> kernels. Without it if one wants to see only the blocked stats this
> >> becomes impossible since the information dumped by sched_debug is so
> >> much that it overflows the kmsg buffer and information about blocked
> >> stats is overwritten. This is especially a problem in distro kernels
> >> which are based on 4.4 and have enabled CONFIG_SCHED_DEBUG - e.g. ubuntu.
> > 
> > Who cares about the kmsg buffer? Also, just make it bigger or disable
> > SCHED_DEBUG, as the name implies it _is_ a debug feature after all.
> 
> but what if you want to use your distro's kernel and deal with custom
> recompiles.

That doesn't make sense, if your distro wants this information, it gets
to pick what it wants.

> The kernel ought to come with sensible defaults.

You can turn it off if you have problems with it, I don't see this
meeting the stable kernel rules.

thanks,

greg k-h