linux-kernel - Re: [PATCH] perf: Fix oops when kthread execs user process

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]

Message-ID: <20190529102022.GC4485@fuggles.cambridge.arm.com>
Date:   Wed, 29 May 2019 11:20:22 +0100
From:   Will Deacon <will.deacon@....com>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     Young Xiao <92siuyang@...il.com>, linux@...linux.org.uk,
        mark.rutland@....com, mingo@...hat.com, bp@...en8.de,
        hpa@...or.com, x86@...nel.org, kan.liang@...ux.intel.com,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        ravi.bangoria@...ux.vnet.ibm.com, mpe@...erman.id.au
Subject: Re: [PATCH] perf: Fix oops when kthread execs user process

On Wed, May 29, 2019 at 12:10:42PM +0200, Peter Zijlstra wrote:
> On Wed, May 29, 2019 at 10:17:33AM +0100, Will Deacon wrote:
> > On Tue, May 28, 2019 at 07:32:28PM +0200, Peter Zijlstra wrote:
> 
> > > 'funny' thing that, perf_sample_regs_user() seems to assume that
> > > anything with current->mm is in fact a user task, and that assumption is
> > > just plain wrong, consider use_mm().
> > 
> > Right, I suppose that was attempting to handle interrupt skid from the PMU
> > overflow?
> 
> Nah, just a broken test to determine if there is userspace at all. It is
> mostly right, just not completely :-)
> 
> > > So I'm thinking the right thing to do here is something like the below;
> > > umh should get PF_KTHREAD cleared when it passes exec(). And this should
> > > also fix the power splat I'm thinking.
> > > 
> > > ---
> > > 
> > > diff --git a/kernel/events/core.c b/kernel/events/core.c
> > > index abbd4b3b96c2..9929404b6eb9 100644
> > > --- a/kernel/events/core.c
> > > +++ b/kernel/events/core.c
> > > @@ -5923,7 +5923,7 @@ static void perf_sample_regs_user(struct perf_regs *regs_user,
> > >  	if (user_mode(regs)) {
> > >  		regs_user->abi = perf_reg_abi(current);
> > >  		regs_user->regs = regs;
> > > -	} else if (current->mm) {
> > > +	} else if (!(current->flags & PF_KTHREAD) && current->mm) {
> > >  		perf_get_regs_user(regs_user, regs, regs_user_copy);
> > 
> > Makes sense, but under which circumstances would we have a NULL mm here?
> 
> Dunno; I'm paranoid, and also:
> 
>   mm/memcontrol.c:        if (in_interrupt() || !current->mm || (current->flags & PF_KTHREAD))

So this one I also don't understand...

>   mm/vmacache.c:  return current->mm == mm && !(current->flags & PF_KTHREAD);

... but this one is just about an mm mismatch, rather than a NULL mm.

Anyway, you can add my ack to your patch, but I bet we can remove that mm
check :D

Will