linux-kernel - Re: [RFC PATCH] sched/wait: Introduce new, more compact wait

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 8 Mar 2017 18:16:19 +0100
From:   Ingo Molnar <mingo@...nel.org>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Peter Zijlstra <a.p.zijlstra@...llo.nl>,
        Thomas Gleixner <tglx@...utronix.de>,
        Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [RFC PATCH] sched/wait: Introduce new, more compact
 wait_event*() primitives


* Linus Torvalds <torvalds@...ux-foundation.org> wrote:

> On Wed, Mar 8, 2017 at 12:37 AM, Ingo Molnar <mingo@...nel.org> wrote:
> >
> > The idea is to allow call sites to supply the 'condition' function as free-form C
> > code, while pushing everything else into non-macro form: there's a 'struct
> > wait_event_state' on stack, and a state machine. The waiting logic is converted
> > from procedural form to a state machine, because we have to call out into the
> > 'condition' code in different circumstances.
> 
> Ok, I think the concept is fine, but you don't actually fix the
> problem with the locked version that needs to unlock (with irq
> versions etc) around the schedule.

Indeed it doesn't, yet.

> And using "bool" in a struct is disgusting and wrong, and hides the
> fact that the compiler will just turn it into "char" (or even "int"
> for platforms where "char'" is slow, like alpha).
> 
> So it would be better with a "state" variable that just has fields, I suspect.
> 
> .. and as mentioned, it doesn't actually fix the case that hit the
> signal_pending() problem.
> 
> Honestly, I think my "pass in a waiter function" model was both less
> subtle and indirect, and more generic.

True!

> And we can actually *fix* the problem with it for 4.11, instead of
> adding the stupid header file includes.

Ok - I'm perfectly fine with your patch too, if you think it's v4.11 material!

Thanks,

	Ingo