linux-kernel - Re: [PATCH v2 2/2] fs: super.c: Add tracepoint to get name of superblock shrinker

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 for Android: free password hash cracker in your pocket

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

Message-ID: <1471526527.2581.2.camel@surriel.com>
Date:	Thu, 18 Aug 2016 09:22:07 -0400
From:	Rik van Riel <riel@...riel.com>
To:	Al Viro <viro@...IV.linux.org.uk>,
	Janani Ravichandran <janani.rvchndrn@...il.com>
Cc:	linux-mm@...ck.org, linux-kernel@...r.kernel.org,
	akpm@...ux-foundation.org, vdavydov@...tuozzo.com, mhocko@...e.com,
	vbabka@...e.cz, mgorman@...hsingularity.net,
	kirill.shutemov@...ux.intel.com, bywxiaobai@....com
Subject: Re: [PATCH v2 2/2] fs: super.c: Add tracepoint to get name of
 superblock shrinker

On Thu, 2016-08-18 at 07:32 +0100, Al Viro wrote:
> On Thu, Aug 18, 2016 at 02:09:31AM -0400, Janani Ravichandran wrote:
> 
> >  static LIST_HEAD(super_blocks);
> > @@ -64,6 +65,7 @@ static unsigned long super_cache_scan(struct
> > shrinker *shrink,
> >  	long	inodes;
> >  
> >  	sb = container_of(shrink, struct super_block, s_shrink);
> > +	trace_mm_shrinker_callback(shrink, sb->s_type->name);
> 
> IOW, we are (should that patch be accepted) obliged to keep the
> function in
> question and the guts of struct shrinker indefinitely.
> 
> NAK.  Keep your debugging patches in your tree and maintain them
> yourself.
> And if a change in the kernel data structures breaks them (and your
> userland
> code relying on those), it's your problem.
> 
> Tracepoints are very nice for local debugging/data collection/etc.
> patches.
> 

The issue is that production systems often need
debugging, and when there are performance issues
we need some way to gather all the necessary info,
without rebooting the production system into a
special debug kernel.

This is not an ABI that userspace can rely on,
and should not be considered as such. Any
performance tracing/debugging scripts can be
easily changed to match the kernel running on
the system in question.