linux-kernel - Re: [PATCH] perf tools: Fix indexing for decoder packet queue

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]

Date:   Mon, 28 May 2018 10:45:08 -0600
From:   Mathieu Poirier <mathieu.poirier@...aro.org>
To:     Leo Yan <leo.yan@...aro.org>
Cc:     Arnaldo Carvalho de Melo <acme@...nel.org>,
        Robert Walker <robert.walker@....com>,
        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>,
        LAK <linux-arm-kernel@...ts.infradead.org>,
        linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] perf tools: Fix indexing for decoder packet queue

On 27 May 2018 at 21:13, Leo Yan <leo.yan@...aro.org> wrote:
> On Fri, May 25, 2018 at 05:10:54PM -0600, Mathieu Poirier wrote:
>> The tail of a queue is supposed to be pointing to the next available slot
>> in a queue.  In this implementation the tail is incremented before it is
>> used and as such points to the last used element, something that has the
>> immense advantage of centralizing tail management at a single location
>> and eliminating a lot of redundant code.
>>
>> But this needs to be taken into consideration on the dequeueing side where
>> the head also needs to be incremented before it is used, or the first
>> available element of the queue will be skipped.
>>
>> Signed-off-by: Mathieu Poirier <mathieu.poirier@...aro.org>
>> ---
>>  tools/perf/util/cs-etm-decoder/cs-etm-decoder.c | 12 ++++++++++--
>>  1 file changed, 10 insertions(+), 2 deletions(-)
>>
>> diff --git a/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c b/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
>> index c8b98fa22997..4d5fc374e730 100644
>> --- a/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
>> +++ b/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
>> @@ -96,11 +96,19 @@ int cs_etm_decoder__get_packet(struct cs_etm_decoder *decoder,
>>   /* Nothing to do, might as well just return */
>>   if (decoder->packet_count == 0)
>>   return 0;
>> + /*
>> + * The queueing process in function cs_etm_decoder__buffer_packet()
>> + * increments the tail *before* using it.  This is somewhat counter
>> + * intuitive but it has the advantage of centralizing tail management
>> + * at a single location.  Because of that we need to follow the same
>> + * heuristic with the head, i.e we increment it before using its
>> + * value.  Otherwise the first element of the packet queue is not
>> + * used.
>> + */
>> + decoder->head = (decoder->head + 1) & (MAX_BUFFER - 1);
>>
>>   *packet = decoder->packet_buffer[decoder->head];
>>
>> - decoder->head = (decoder->head + 1) & (MAX_BUFFER - 1);
>> -
>
> I tested this patch and confirmed it can work well with python
> decoding script:
>
> Tested-by: Leo Yan <leo.yan@...aro.org>
>
> Actually, I have another idea for this fixing, seems to me
> the unchanged code has right logic for decoder->head, and I think this
> issue is more related with incorrect initialization index.  So we can
> change the initialization index for decoder->head as below.  How about
> you think for this?
>
> diff --git a/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
> b/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
> index c8b98fa..b133260 100644
> --- a/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
> +++ b/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
> @@ -249,7 +249,7 @@ static void cs_etm_decoder__clear_buffer(struct
> cs_etm_decoder *decoder)
>  {
>         int i;
>
> -       decoder->head = 0;
> +       decoder->head = 1;

I flirted with that idea but thought the problem is really with the
tail and as such would have done:

        decoder->tail = -1;

But since both head and tail are declared as u32 it would have
required changing the types to int, necessitating modifications
everywhere other parts of the code deals with them.  As such I just
decided to swap the order of events in cs_etm_decoder__get_packet().

I'm not strongly opinionated on this and can send another patch if you're keen.

Thanks for the feedback,
Mathieu


>         decoder->tail = 0;
>         decoder->packet_count = 0;
>         for (i = 0; i < MAX_BUFFER; i++) {
>
> Thanks,
> Leo Yan
>
>>   decoder->packet_count--;
>>
>>   return 1;
>> --
>> 2.7.4
>>