linux-kernel - Re: [PATCH v7 06/16] lockdep: Detect and handle hist

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: <20170713020745.GG20323@X58A-UD3R>
Date:   Thu, 13 Jul 2017 11:07:45 +0900
From:   Byungchul Park <byungchul.park@....com>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     mingo@...nel.org, tglx@...utronix.de, walken@...gle.com,
        boqun.feng@...il.com, kirill@...temov.name,
        linux-kernel@...r.kernel.org, linux-mm@...ck.org,
        akpm@...ux-foundation.org, willy@...radead.org, npiggin@...il.com,
        kernel-team@....com
Subject: Re: [PATCH v7 06/16] lockdep: Detect and handle hist_lock ring
 buffer overwrite

On Wed, Jul 12, 2017 at 09:56:17AM +0200, Peter Zijlstra wrote:
> On Wed, Jul 12, 2017 at 11:00:53AM +0900, Byungchul Park wrote:
> > On Tue, Jul 11, 2017 at 06:12:32PM +0200, Peter Zijlstra wrote:
> 
> > > Right, like I wrote in the comment; I don't think you need quite this
> > > much.
> > > 
> > > The problem only happens if you rewind more than MAX_XHLOCKS_NR;
> > > although I realize it can be an accumulative rewind, which makes it
> > > slightly more tricky.
> > > 
> > > We can either make the rewind more expensive and make xhlock_valid()
> > > false for each rewound entry; or we can keep the max_idx and account
> > 
> > Does max_idx mean the 'original position - 1'?
> 
> 	orig_idx = current->hist_idx;
> 	current->hist_idx++;
> 	if ((int)(current->hist_idx - orig_idx) > 0)
> 	  current->hist_idx_max = current->hist_idx;
> 
> 
> I've forgotten if the idx points to the most recent entry or beyond it.
> 
> Given the circular nature, and tail being one ahead of head, the max
> effectively tracks the tail (I suppose we can also do an explicit tail
> tracking, but that might end up more difficult).
> 
> This allows rewinds of less than array_size() while still maintaining a
> correct tail.
> 
> Only once we (cummulative or not) rewind past the tail -- iow, loose the
> _entire_ history, do we need to do something drastic.

I am sorry but I don't understand why we have to do the drastic work.

Does my approach have problems, rewinding to 'original idx' on exit and
deciding whether overwrite or not? I think, this way, no need to do the
drastic work. Or.. does my one get more overhead in usual case?

> 
> > > from there. If we rewind >= MAX_XHLOCKS_NR from the max_idx we need to
> > > invalidate the entire state, which we can do by invaliding
> > 
> > Could you explain what the entire state is?
> 
> All hist_lock[]. Did the above help?
> 
> > > xhlock_valid() or by re-introduction of the hist_gen_id. When we
> > 
> > What does the re-introduction of the hist_gen_id mean?
> 
> What you used to call work_id or something like that. A generation count
> for the hist_lock[].