linux-kernel - Re: [PATCH] x86/stacktrace: Do not access user space memory unnecessarily

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:   Tue, 2 Jul 2019 11:33:55 -0400
From:   Steven Rostedt <rostedt@...dmis.org>
To:     Thomas Gleixner <tglx@...utronix.de>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Eiichi Tsukata <devel@...ukata.com>, edwintorok@...il.com,
        mingo@...hat.com, bp@...en8.de, hpa@...or.com, x86@...nel.org,
        linux-kernel@...r.kernel.org, Josh Poimboeuf <jpoimboe@...hat.com>
Subject: Re: [PATCH] x86/stacktrace: Do not access user space memory
 unnecessarily

On Tue, 2 Jul 2019 16:14:05 +0200 (CEST)
Thomas Gleixner <tglx@...utronix.de> wrote:

> On Tue, 2 Jul 2019, Peter Zijlstra wrote:
> 
> > On Tue, Jul 02, 2019 at 02:31:51PM +0900, Eiichi Tsukata wrote:  
> > > Put the boundary check before it accesses user space to prevent unnecessary
> > > access which might crash the machine.
> > > 
> > > Especially, ftrace preemptirq/irq_disable event with user stack trace
> > > option can trigger SEGV in pid 1 which leads to panic.  

Note, I'm only able to trigger this crash with the irq_disable event.
The irq_enable and preempt_disable/enable events work just fine. This
leads me to believe that the TRACE_IRQS_OFF macro (which uses a thunk
trampoline) may have some issues and is probably the place to look at.

-- Steve

> 
> It triggers segfaults in random user processes which is bad enough.
> 
> And even with that 'fix' applied I can see random segfaults just less
> frequent.
> 
> > >   RIP: 0033:0x55be7ad1c89f
> > >   Code: Bad RIP value.  
> > 
> > ^^^ that's weird, no amount of unwinding should affect regs->ip.  
> 
>