linux-kernel - Re: [RFC PATCH] perf: workaround unaligned NEON vector load

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] [day] [month] [year] [list]

Date:   Wed, 6 Mar 2019 17:14:34 -0300
From:   Arnaldo Carvalho de Melo <arnaldo.melo@...il.com>
To:     Lucas Stach <l.stach@...gutronix.de>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Jiri Olsa <jolsa@...hat.com>,
        Namhyung Kim <namhyung@...nel.org>,
        linux-kernel@...r.kernel.org, kernel@...gutronix.de,
        patchwork-lst@...gutronix.de
Subject: Re: [RFC PATCH] perf: workaround unaligned NEON vector load

Em Wed, Mar 06, 2019 at 03:01:16PM +0100, Lucas Stach escreveu:
> The mmap event buffer may end up in a location that violates the
> alignment requirements for a NEON vector load,

which are?

> which GCC generates to load consecutive values from the event
> structure. Fix this by copying the event structure into a properly
> aligned buffer.

At a minimum this would be done only for such arch (is that an arch?),
so that the rest of the world doesn't have to eat this extra cost?

What is it that perf_event_mmap_event() is doing to
mmap_event->event_id.header.size that this NEON vector load dislikes?

- Arnaldo
 
> Signed-off-by: Lucas Stach <l.stach@...gutronix.de>
> ---
>  tools/perf/util/machine.c | 29 ++++++++++++++++-------------
>  1 file changed, 16 insertions(+), 13 deletions(-)
> 
> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
> index 143f7057d581..ab5500e85173 100644
> --- a/tools/perf/util/machine.c
> +++ b/tools/perf/util/machine.c
> @@ -1565,37 +1565,40 @@ static int machine__process_kernel_mmap_event(struct machine *machine,
>  }
>  
>  int machine__process_mmap2_event(struct machine *machine,
> -				 union perf_event *event,
> +				 union perf_event *event_in,
>  				 struct perf_sample *sample)
>  {
> +	union perf_event event;
>  	struct thread *thread;
>  	struct map *map;
>  	int ret = 0;
>  
> +	memcpy(&event, event_in, sizeof(union perf_event));
> +
>  	if (dump_trace)
> -		perf_event__fprintf_mmap2(event, stdout);
> +		perf_event__fprintf_mmap2(&event, stdout);
>  
>  	if (sample->cpumode == PERF_RECORD_MISC_GUEST_KERNEL ||
>  	    sample->cpumode == PERF_RECORD_MISC_KERNEL) {
> -		ret = machine__process_kernel_mmap_event(machine, event);
> +		ret = machine__process_kernel_mmap_event(machine, &event);
>  		if (ret < 0)
>  			goto out_problem;
>  		return 0;
>  	}
>  
> -	thread = machine__findnew_thread(machine, event->mmap2.pid,
> -					event->mmap2.tid);
> +	thread = machine__findnew_thread(machine, event.mmap2.pid,
> +					event.mmap2.tid);
>  	if (thread == NULL)
>  		goto out_problem;
>  
> -	map = map__new(machine, event->mmap2.start,
> -			event->mmap2.len, event->mmap2.pgoff,
> -			event->mmap2.maj,
> -			event->mmap2.min, event->mmap2.ino,
> -			event->mmap2.ino_generation,
> -			event->mmap2.prot,
> -			event->mmap2.flags,
> -			event->mmap2.filename, thread);
> +	map = map__new(machine, event.mmap2.start,
> +			event.mmap2.len, event.mmap2.pgoff,
> +			event.mmap2.maj,
> +			event.mmap2.min, event.mmap2.ino,
> +			event.mmap2.ino_generation,
> +			event.mmap2.prot,
> +			event.mmap2.flags,
> +			event.mmap2.filename, thread);
>  
>  	if (map == NULL)
>  		goto out_problem_map;
> -- 
> 2.20.1

-- 

- Arnaldo