linux-kernel - Re: [PATCH] perf-probe: Failback to symbol-base probe for probes on module

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: <20210223163653.e7cf141d207ba07576d95f59@kernel.org>
Date:   Tue, 23 Feb 2021 16:36:53 +0900
From:   Masami Hiramatsu <mhiramat@...nel.org>
To:     Masami Hiramatsu <mhiramat@...nel.org>
Cc:     Arnaldo Carvalho de Melo <acme@...nel.org>,
        Josh Poimboeuf <jpoimboe@...hat.com>,
        Evgenii Shatokhin <eshatokhin@...tuozzo.com>,
        Kristen Carlson Accardi <kristen@...ux.intel.com>,
        live-patching@...r.kernel.org,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>, linux-kernel@...r.kernel.org,
        Konstantin Khorenko <khorenko@...tuozzo.com>
Subject: Re: [PATCH] perf-probe: Failback to symbol-base probe for probes on
 module

Please ignore this. I will send a better fix.

Thanks,

On Tue, 23 Feb 2021 10:48:30 +0900
Masami Hiramatsu <mhiramat@...nel.org> wrote:

> If an error occurs on post processing (this converts probe point to
> _text relative address for identifying non-unique symbols) for the
> probes on module, failback to symbol-base probe.
> 
> There are many non-unique name symbols (local static functions can
> be in the different name spaces) in the kernel. If perf-probe uses
> a symbol-based probe definition, it can be put on an unintended
> symbol. To avoid such mistake, perf-probe tries to convert the
> address to the _text relative address expression.
> 
> However, such symbol duplication is rare for only one module. Thus
> even if the conversion fails, perf probe can failback to the symbol
> based probe.
> 
> Signed-off-by: Masami Hiramatsu <mhiramat@...nel.org>
> ---
>  tools/perf/util/probe-event.c |   12 ++++++------
>  1 file changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c
> index a9cff3a50ddf..af946f68e32e 100644
> --- a/tools/perf/util/probe-event.c
> +++ b/tools/perf/util/probe-event.c
> @@ -779,16 +779,16 @@ post_process_module_probe_trace_events(struct probe_trace_event *tevs,
>  
>  	map = get_target_map(module, NULL, false);
>  	if (!map || debuginfo__get_text_offset(dinfo, &text_offs, true) < 0) {
> -		pr_warning("Failed to get ELF symbols for %s\n", module);
> -		return -EINVAL;
> +		pr_info("NOTE: Failed to get ELF symbols for %s. Use symbol based probe.\n", module);
> +		return 0;
>  	}
>  
>  	mod_name = find_module_name(module);
>  	for (i = 0; i < ntevs; i++) {
> -		ret = post_process_probe_trace_point(&tevs[i].point,
> -						map, (unsigned long)text_offs);
> -		if (ret < 0)
> -			break;
> +		if (post_process_probe_trace_point(&tevs[i].point,
> +				map, (unsigned long)text_offs) < 0)
> +			pr_info("NOTE: %s is not in the symbol map. Use symbol based probe.\n",
> +				   tevs[i].point.symbol);
>  		tevs[i].point.module =
>  			strdup(mod_name ? mod_name : module);
>  		if (!tevs[i].point.module) {
> 


-- 
Masami Hiramatsu <mhiramat@...nel.org>