linux-kernel - Re: [PATCH trace-cmd 2/2] trace-recorder: better error handling during copy

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: <20160323070216.GN28183@pxdev.xzpeter.org>
Date:	Wed, 23 Mar 2016 15:02:16 +0800
From:	Peter Xu <peterx@...hat.com>
To:	Steven Rostedt <rostedt@...dmis.org>
Cc:	linux-kernel@...r.kernel.org
Subject: Re: [PATCH trace-cmd 2/2] trace-recorder: better error handling
 during copy

On Tue, Mar 22, 2016 at 09:19:56AM -0400, Steven Rostedt wrote:
> On Tue, 22 Mar 2016 09:14:33 -0400
> Peter Xu <peterx@...hat.com> wrote:
> 
> > Currently we have two ways to copy data, one is splice, one is read +
> > write. For both, dump more information when we got errors during the
> > copy. Also, when we update_fd(), we should make sure all bytes written,
> > and update written bytes only.
> > 
> > These information might be important to better diagnose when the copy
> > got wrong, like no space error, or connection error when copying data to
> > remote sockets. In the past, we just got silence errors without notice.
> > 
> > Signed-off-by: Peter Xu <peterx@...hat.com>
> > ---
> >  trace-recorder.c | 34 ++++++++++++++++++++++++----------
> >  1 file changed, 24 insertions(+), 10 deletions(-)
> > 
> > diff --git a/trace-recorder.c b/trace-recorder.c
> > index 49b04ea..7d6feb0 100644
> > --- a/trace-recorder.c
> > +++ b/trace-recorder.c
> > @@ -334,13 +334,14 @@ static inline void update_fd(struct tracecmd_recorder *recorder, int size)
> >   */
> >  static long splice_data(struct tracecmd_recorder *recorder)
> >  {
> > -	long ret;
> > +	long ret, written;
> >  
> >  	ret = splice(recorder->trace_fd, NULL, recorder->brass[1], NULL,
> >  		     recorder->page_size, 1 /* SPLICE_F_MOVE */);
> >  	if (ret < 0) {
> >  		if (errno != EAGAIN && errno != EINTR) {
> > -			warning("recorder error in splice input");
> > +			warning("recorder error in splice input: %s",
> > +				strerror(errno));
> 
> I'm wondering if we should add a "pwarning()" helper function that will
> do the stderror(error) for us.

Ah, I just found that __vwarning() will dump errno string, and
warning() is calling that. So it explained why I got one more line
when got errors... ;)

How about remove all the strerror() things, and just keep capturing
the written size? Let me send a v2 for this patch.

Thanks.

-- peterx