Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1752469AbcDZOmc (ORCPT <rfc822;w@1wt.eu>);
	Tue, 26 Apr 2016 10:42:32 -0400
Received: from mail.kernel.org ([198.145.29.136]:45767 "EHLO mail.kernel.org"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1752248AbcDZOma (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Tue, 26 Apr 2016 10:42:30 -0400
Date: Tue, 26 Apr 2016 23:42:25 +0900
From: Masami Hiramatsu <mhiramat@kernel.org>
To: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: linux-kernel@vger.kernel.org, Namhyung Kim <namhyung@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>, Ingo Molnar <mingo@redhat.com>,
        Hemant Kumar <hemant@linux.vnet.ibm.com>,
        Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Subject: Re: [PATCH perf/core v4 03/19] perf buildid-cache: Fall back to the
 old style build-id cache
Message-Id: <20160426234225.204ed49b76d02442fc57ec83@kernel.org>
In-Reply-To: <20160426134749.GB20998@kernel.org>
References: <20160426090200.11891.43944.stgit@devbox>
	<20160426090233.11891.70554.stgit@devbox>
	<20160426134749.GB20998@kernel.org>
X-Mailer: Sylpheed 3.4.3 (GTK+ 2.24.28; x86_64-redhat-linux-gnu)
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 4224
Lines: 123

On Tue, 26 Apr 2016 10:47:49 -0300
Arnaldo Carvalho de Melo <acme@kernel.org> wrote:

> Em Tue, Apr 26, 2016 at 06:02:33PM +0900, Masami Hiramatsu escreveu:
> > Fall back to the old style build-id cache
> > (~/.debug/<path-to-bin>/<build-id>) if possible.
> > Because there is old build-id cache under ~/.debug, perf
> > has to check if the old entry is there instead of returning
> > an error. Or, user have to remove ~/.debug completely.
> 
> So, from the description this seems to address the issues I mentioned in
> the previous patch.

yes :)

> 
> Probably it is a good idea not to introduce the problem in the first
> place? I.e. combine this and the previous patch, so that we don't have
> this bisection break?

Yeah, I just try to keep original clean for review.
I'll combine this with 02/19.

Thanks!

> 
> - Arnaldo
>  
> > This patch introduces fallback code to try using old-style
> > buildid caches. To update the cached entry, you just need to
> > add a cache. It automatically removes old cache entry and
> > make a new one in new style
> > (~/.debug/<path-to-bin>/<build-id>/[elf|vdso|kallsyms]).
> > 
> > IOW, without this patch, "perf buildid-cache --add" always
> > gets errors when adding existing binaries.
> >   ----
> >   # perf buildid-cache -a /usr/lib64/libc-2.20.so,/usr/bin/gcc
> >   Couldn't add /usr/bin/gcc: Not a directory
> >   Couldn't add /usr/lib64/libc-2.20.so: Not a directory
> >   ----
> > Moreover, --update option removes only symlink, thus the
> > build-id cache is corrupted.
> > 
> > This fixes above issues.
> > 
> > Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
> > ---
> >  tools/perf/util/build-id.c |   17 ++++++++++++++++-
> >  tools/perf/util/symbol.c   |   13 +++++++++----
> >  2 files changed, 25 insertions(+), 5 deletions(-)
> > 
> > diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c
> > index 46a8bcc..b035483 100644
> > --- a/tools/perf/util/build-id.c
> > +++ b/tools/perf/util/build-id.c
> > @@ -155,6 +155,12 @@ static char *build_id_cache__linkname(const char *sbuild_id, char *bf,
> >  	return bf;
> >  }
> >  
> > +static bool __is_regular_file(const char *pathname)
> > +{
> > +	struct stat sb;
> > +	return stat(pathname, &sb) == 0 && S_ISREG(sb.st_mode);
> > +}
> > +
> >  static const char *build_id_cache__basename(bool is_kallsyms, bool is_vdso)
> >  {
> >  	return is_kallsyms ? "kallsyms" : (is_vdso ? "vdso" : "elf");
> > @@ -177,7 +183,11 @@ char *dso__build_id_filename(const struct dso *dso, char *bf, size_t size)
> >  	if (!linkname)
> >  		return NULL;
> >  
> > -	ret = asnprintf(&bf, size, "%s/%s", linkname,
> > +	/* Check if old style build_id cache */
> > +	if (__is_regular_file(linkname))
> > +		ret = asnprintf(&bf, size, "%s", linkname);
> > +	else
> > +		ret = asnprintf(&bf, size, "%s/%s", linkname,
> >  			 build_id_cache__basename(is_kallsyms, is_vdso));
> >  	if (ret < 0 || (!alloc && size < (unsigned int)ret))
> >  		bf = NULL;
> > @@ -445,6 +455,11 @@ int build_id_cache__add_s(const char *sbuild_id, const char *name,
> >  	if (!dir_name)
> >  		goto out_free;
> >  
> > +	/* Remove old style build-id cache */
> > +	if (__is_regular_file(dir_name))
> > +		if (unlink(dir_name))
> > +			goto out_free;
> > +
> >  	if (mkdir_p(dir_name, 0755))
> >  		goto out_free;
> >  
> > diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> > index c57cb47..9463c7d 100644
> > --- a/tools/perf/util/symbol.c
> > +++ b/tools/perf/util/symbol.c
> > @@ -1687,11 +1687,16 @@ static char *dso__find_kallsyms(struct dso *dso, struct map *map)
> >  
> >  	scnprintf(path, sizeof(path), "%s/[kernel.kallsyms]/%s/kallsyms",
> >  		  buildid_dir, sbuild_id);
> > -
> > +	/* Try old style kallsyms cache */
> >  	if (access(path, F_OK)) {
> > -		pr_err("No kallsyms or vmlinux with build-id %s was found\n",
> > -		       sbuild_id);
> > -		return NULL;
> > +		scnprintf(path, sizeof(path), "%s/[kernel.kallsyms]/%s",
> > +			  buildid_dir, sbuild_id);
> > +
> > +		if (access(path, F_OK)) {
> > +			pr_err("No kallsyms or vmlinux with build-id %s was found\n",
> > +			       sbuild_id);
> > +			return NULL;
> > +		}
> >  	}
> >  
> >  	return strdup(path);


-- 
Masami Hiramatsu <mhiramat@kernel.org>