linux-kernel - Re: [PATCH] tracing: Create generic syscall TRACE

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

Message-ID: <20090819135011.GA8674@Krystal>
Date:	Wed, 19 Aug 2009 09:50:12 -0400
From:	Mathieu Desnoyers <mathieu.desnoyers@...ymtl.ca>
To:	Ingo Molnar <mingo@...e.hu>
Cc:	Josh Stone <jistone@...hat.com>, Li Zefan <lizf@...fujitsu.com>,
	linux-kernel@...r.kernel.org, fweisbec@...il.com,
	laijs@...fujitsu.com, rostedt@...dmis.org, peterz@...radead.org,
	jiayingz@...gle.com, mbligh@...gle.com,
	Jason Baron <jbaron@...hat.com>
Subject: Re: [PATCH] tracing: Create generic syscall TRACE_EVENTs

* Ingo Molnar (mingo@...e.hu) wrote:
> 
> * Josh Stone <jistone@...hat.com> wrote:
> 
> > On 08/18/2009 06:32 PM, Li Zefan wrote:
> > > It would be much better to see a sample output of these TRACE_EVENTs,
> > > so we'll see the what's the output if a syscall has less than 6 args.
> > > 
> > > And I guess it should be "(%lx, %lx, ..., %lx)"? Otherwise I think we'll
> > > see some negative values.
> > 
> > Thanks, you're right -- the args should be unsigned long.  Whether they
> > are hex or decimal doesn't matter to me, but I guess that does make
> > pointers easier to spot.
> > 
> > Here's some output that I get on x86_64:
> > 
> > > # tail trace
> > >             tail-1106  [000] 113218.474657: syscall_enter: NR 2 (7f2d34dd38f0, 0, 7f2d35009010, 5, 4, ffffffffffffffb0)
> > > 
> > >             tail-1106  [000] 113218.474809: syscall_enter: NR 5 (3, 7f2d35009040, 7f2d35009040, 5, 4, ffffffffffffffb0)
> > > 
> > >             tail-1106  [000] 113218.474855: syscall_enter: NR 9 (0, 50d3ad0, 1, 2, 3, 0)
> > > 
> > >             tail-1106  [000] 113218.474946: syscall_enter: NR 3 (3, 7f2d2f9a0000, 5e22, 2, 3, 0)
> > > 
> > >             tail-1106  [000] 113218.475673: syscall_enter: NR 2 (7fff5e42a654, 0, 0, 7fff5e429600, 80, 3)
> > 
> > It's very raw indeed, but that's what Ingo asked for. :) The 
> > extra args show up as whatever happened to be in those registers.
> 
> If the syscalls are handled as a single tracepoint, then they 
> cannot really have per syscall properties.
> 
> The per syscall tracepoints on the other hand give very finegrained 
> information.
> 

Hi Ingo,

Aiming to identify the event payload type with the event ID alone, I
would recommend either:

- one tracepoint per syscall argument set types (e.g. same syscall
  prototype)
or
- one tracepoint per syscall (seems to be more straightforward)

Otherwise getting to fit a struct sockaddr into an integer might be
cumbersome.

Thanks,

Mathieu

> Note that via the use of filters one can limit the generic 
> tracepoint to only output certain syscall NRs, and only with 
> certain parameters.
> 
> 	Ingo

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/