Re: [PATCH v2] perf session: fix decompression of PERF_RECORD_COMPRESSED records

[Jiri Olsa <jolsa@redhat.com>]

On Mon, Nov 18, 2019 at 02:30:04PM +0300, Alexey Budankov wrote:
> 
> Avoid termination of trace loading in case the last record in
> the decompressed buffer partly resides in the following
> mmaped PERF_RECORD_COMPRESSED record. In this case NULL value
> returned by fetch_mmaped_event() means to proceed to the next
> mmaped record then decompress it and load compressed events.
> 
> The issue can be reproduced like this:
> 
>   $ perf record -z -- some_long_running_workload
>   $ perf report --stdio -vv
>   decomp (B): 44519 to 163000
>   decomp (B): 48119 to 174800
>   decomp (B): 65527 to 131072
>   fetch_mmaped_event: head=0x1ffe0 event->header_size=0x28, mmap_size=0x20000: fuzzed perf.data?
>   Error:
>   failed to process sample
>   ...
> 
> Testing:
> 71: Zstd perf.data compression/decompression              : Ok
> 
> Fixes: 57fc032ad643 ("perf session: Avoid infinite loop when seeing invalid header.size")
> Link: https://marc.info/?l=linux-kernel&m=156580812427554&w=2
> Co-developed-by: Jiri Olsa <jolsa@kernel.org>
> Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
> ---
> Changes in v2:
> - avoided static declaration of prefetch_event();
> - renamed 'ret' to 'error' for returning in case of split compressed 
>   or overlapping records;
> - passed only needs_swap quality into fetch_*_event() instead of 
>   the whole session object;
> ---
>  tools/perf/util/session.c | 44 ++++++++++++++++++++++++---------------
>  1 file changed, 27 insertions(+), 17 deletions(-)
> 
> diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
> index f07b8ecb91bc..c2b0703d6587 100644
> --- a/tools/perf/util/session.c
> +++ b/tools/perf/util/session.c
> @@ -1958,8 +1958,8 @@ static int __perf_session__process_pipe_events(struct perf_session *session)
>  }
>  
>  static union perf_event *
> -fetch_mmaped_event(struct perf_session *session,
> -		   u64 head, size_t mmap_size, char *buf)
> +prefetch_event(char *buf, u64 head, size_t mmap_size,
> +	       bool needs_swap, union perf_event *error)
>  {
>  	union perf_event *event;
>  
> @@ -1971,20 +1971,32 @@ fetch_mmaped_event(struct perf_session *session,
>  		return NULL;
>  
>  	event = (union perf_event *)(buf + head);
> +	if (needs_swap)
> +		perf_event_header__bswap(&event->header);
>  
> -	if (session->header.needs_swap)
> +	if (head + event->header.size <= mmap_size)
> +		return event;
> +
> +	/* We're not fetching the event so swap back again */
> +	if (needs_swap)
>  		perf_event_header__bswap(&event->header);
>  
> -	if (head + event->header.size > mmap_size) {
> -		/* We're not fetching the event so swap back again */
> -		if (session->header.needs_swap)
> -			perf_event_header__bswap(&event->header);
> -		pr_debug("%s: head=%#" PRIx64 " event->header_size=%#x, mmap_size=%#zx: fuzzed perf.data?\n",
> -			 __func__, head, event->header.size, mmap_size);
> -		return ERR_PTR(-EINVAL);
> -	}
> +	pr_debug("%s: head=%#" PRIx64 " event->header_size=%#x, mmap_size=%#zx\n",
> +		__func__, head, event->header.size, mmap_size);

you're missign the 'fuzzed perf.data?' in here

I think we should keep it just change to: 'fuzzed or compressed perf.data?'

thanks,
jirka