[Ksummit-discuss] [CORE TOPIC] cleaning up kthread freezer hell, part 2

[Ksummit-discuss] [CORE TOPIC] cleaning up kthread freezer hell, part 2

[Jiri Kosina]

On last year's kernel summit, I've been talking about why I consider 
kthread freezer harmful and why it ultimately should be removed. LWN 
coverage of that session is here:

	https://lwn.net/Articles/662703/

During the past year, I've invested a bit of a time into actually looking 
deeper into the dark corners of kernel sources to see how kthread freezer 
is used throughout the codebase, with the intent to ultimately fix all the 
buggy places. While doing that, I was petrified by two facts:

- there are a *lot* of places where kthread freezer is used in a 
  completely buggy (or useless) way

- one of the obstacles fixing it are maintainers who actually don't 
  understand the purpose of the kthread freezer (the usual pattern is that 
  the main kthread loop has been copy/pasted from different code, which 
  already used freezer, and so disease spreads)

Therefore I'd propose a v2 of the last year's session; first summarizing 
the horrible experience I've done on this kthread freezer journey, and as 
a followup, try to (re-)explain the issue and the way I think it should be 
resolved.

The idea is to get as much coverage among high-profile maintainers as 
possible, in a hope that this will result in ultimate tree-wide cleanup of 
the current mess. That's why I propose this as a core topic rather than 
tech topic, although it might sound like a rather bordeline one.

Thanks,

-- 
Jiri Kosina
SUSE Labs

Re: [Ksummit-discuss] [CORE TOPIC] cleaning up kthread freezer hell, part 2

[Rafael J. Wysocki]

On Saturday, July 09, 2016 12:31:33 AM Jiri Kosina wrote:
> On last year's kernel summit, I've been talking about why I consider 
> kthread freezer harmful and why it ultimately should be removed. LWN 
> coverage of that session is here:
> 
> 	https://lwn.net/Articles/662703/
> 
> During the past year, I've invested a bit of a time into actually looking 
> deeper into the dark corners of kernel sources to see how kthread freezer 
> is used throughout the codebase, with the intent to ultimately fix all the 
> buggy places. While doing that, I was petrified by two facts:
> 
> - there are a *lot* of places where kthread freezer is used in a 
>   completely buggy (or useless) way
> 
> - one of the obstacles fixing it are maintainers who actually don't 
>   understand the purpose of the kthread freezer (the usual pattern is that 
>   the main kthread loop has been copy/pasted from different code, which 
>   already used freezer, and so disease spreads)
> 
> Therefore I'd propose a v2 of the last year's session; first summarizing 
> the horrible experience I've done on this kthread freezer journey, and as 
> a followup, try to (re-)explain the issue and the way I think it should be 
> resolved.

Yes, please.

Let me know if/how I can help.

> The idea is to get as much coverage among high-profile maintainers as 
> possible, in a hope that this will result in ultimate tree-wide cleanup of 
> the current mess. That's why I propose this as a core topic rather than 
> tech topic, although it might sound like a rather bordeline one.

I guess that first needs to be "core" so it can become "tech" later when
everybody is on the same page in general.

Thanks,
Rafael

Re: [Ksummit-discuss] [CORE TOPIC] cleaning up kthread freezer hell, part 2

[Julia Lawall]

On Sat, 9 Jul 2016, Rafael J. Wysocki wrote:

> On Saturday, July 09, 2016 12:31:33 AM Jiri Kosina wrote:
> > On last year's kernel summit, I've been talking about why I consider
> > kthread freezer harmful and why it ultimately should be removed. LWN
> > coverage of that session is here:
> >
> > 	https://lwn.net/Articles/662703/
> >
> > During the past year, I've invested a bit of a time into actually looking
> > deeper into the dark corners of kernel sources to see how kthread freezer
> > is used throughout the codebase, with the intent to ultimately fix all the
> > buggy places. While doing that, I was petrified by two facts:
> >
> > - there are a *lot* of places where kthread freezer is used in a
> >   completely buggy (or useless) way
> >
> > - one of the obstacles fixing it are maintainers who actually don't
> >   understand the purpose of the kthread freezer (the usual pattern is that
> >   the main kthread loop has been copy/pasted from different code, which
> >   already used freezer, and so disease spreads)
> >
> > Therefore I'd propose a v2 of the last year's session; first summarizing
> > the horrible experience I've done on this kthread freezer journey, and as
> > a followup, try to (re-)explain the issue and the way I think it should be
> > resolved.
>
> Yes, please.
>
> Let me know if/how I can help.

Likewise.

julia

>
> > The idea is to get as much coverage among high-profile maintainers as
> > possible, in a hope that this will result in ultimate tree-wide cleanup of
> > the current mess. That's why I propose this as a core topic rather than
> > tech topic, although it might sound like a rather bordeline one.
>
> I guess that first needs to be "core" so it can become "tech" later when
> everybody is on the same page in general.
>
> Thanks,
> Rafael
>
> _______________________________________________
> Ksummit-discuss mailing list
> Ksummit-discuss@lists.linuxfoundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/ksummit-discuss
>

Re: [Ksummit-discuss] [CORE TOPIC] cleaning up kthread freezer hell, part 2

[Shuah Khan]

On Sat, Jul 9, 2016 at 12:19 AM, Julia Lawall <julia.lawall@lip6.fr> wrote:
>
>
> On Sat, 9 Jul 2016, Rafael J. Wysocki wrote:
>
>> On Saturday, July 09, 2016 12:31:33 AM Jiri Kosina wrote:
>> > On last year's kernel summit, I've been talking about why I consider
>> > kthread freezer harmful and why it ultimately should be removed. LWN
>> > coverage of that session is here:
>> >
>> >     https://lwn.net/Articles/662703/
>> >
>> > During the past year, I've invested a bit of a time into actually looking
>> > deeper into the dark corners of kernel sources to see how kthread freezer
>> > is used throughout the codebase, with the intent to ultimately fix all the
>> > buggy places. While doing that, I was petrified by two facts:
>> >
>> > - there are a *lot* of places where kthread freezer is used in a
>> >   completely buggy (or useless) way
>> >
>> > - one of the obstacles fixing it are maintainers who actually don't
>> >   understand the purpose of the kthread freezer (the usual pattern is that
>> >   the main kthread loop has been copy/pasted from different code, which
>> >   already used freezer, and so disease spreads)
>> >
>> > Therefore I'd propose a v2 of the last year's session; first summarizing
>> > the horrible experience I've done on this kthread freezer journey, and as
>> > a followup, try to (re-)explain the issue and the way I think it should be
>> > resolved.
>>
>> Yes, please.
>>
>> Let me know if/how I can help.
>
> Likewise.
>

Yes. Count me in to help out on this.

-- Shuah

>
>>
>> > The idea is to get as much coverage among high-profile maintainers as
>> > possible, in a hope that this will result in ultimate tree-wide cleanup of
>> > the current mess. That's why I propose this as a core topic rather than
>> > tech topic, although it might sound like a rather bordeline one.
>>
>> I guess that first needs to be "core" so it can become "tech" later when
>> everybody is on the same page in general.
>>
>> Thanks,
>> Rafael
>>
>> _______________________________________________
>> Ksummit-discuss mailing list
>> Ksummit-discuss@lists.linuxfoundation.org
>> https://lists.linuxfoundation.org/mailman/listinfo/ksummit-discuss
>>
> _______________________________________________
> Ksummit-discuss mailing list
> Ksummit-discuss@lists.linuxfoundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/ksummit-discuss

Re: [Ksummit-discuss] [CORE TOPIC] cleaning up kthread freezer hell, part 2

[Theodore Ts'o]

On Sat, Jul 09, 2016 at 12:31:33AM +0200, Jiri Kosina wrote:
> 
> The idea is to get as much coverage among high-profile maintainers as 
> possible, in a hope that this will result in ultimate tree-wide cleanup of 
> the current mess. That's why I propose this as a core topic rather than 
> tech topic, although it might sound like a rather bordeline one.

It doesn't have to be either/or.  One possibility is after doing a
discussion amonst the core developers about how to do the tree-wide
cleanup, you could also do a presentation during the tech sessions
about how not to screw up new users of the freezer.

Although we probably don't have the resources to get a profession
video made of tech talk session for which the goal is broad
dissemination of information (although if some company wants to
sponsor such a thing, or if someone wants to volunteer to do the video
taping where the presenter has given consent, please contact the me or
the program committee), I suspect getting an audio recording, which
plus the slides could easily get turned into something that could be
uploaded to You Tube, wouldn't be that hard.

					- Ted