perf session: Fix perf_session__peek_event()
authorAdrian Hunter <adrian.hunter@intel.com>
Tue, 19 May 2015 13:05:45 +0000 (16:05 +0300)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Wed, 27 May 2015 15:21:43 +0000 (12:21 -0300)
perf_session__peek_event() generally leverages there being a single mmap
of the perf.data file, however on 32-bit platforms when there is more
that 32MiB of data, then there are multiple mmaps, so
perf_session__peek_event() reads from the file.

In that case a couple of bugs were exposed (note how the seg. fault
appears with >32M of data):

   $ perf record --per-thread -e intel_bts// ../rtit-tests/loopy 1000000
   [ perf record: Woken up 13 times to write data ]
   [ perf record: Captured and wrote 24.568 MB perf.data ]
   $ perf script > /dev/null
   $ perf record --per-thread -e intel_bts// ../rtit-tests/loopy 10000000
   [ perf record: Woken up 136 times to write data ]
   [ perf record: Captured and wrote 270.794 MB perf.data ]
   $ perf script > /dev/null
   Segmentation fault (core dumped)

The wrong address was being passed to the readn() function and the
buffer size was not being checked.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Link: http://lkml.kernel.org/r/1432040746-1755-5-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/util/session.c

index e722107f932ac8383e570c83bd94ee4a014ff398..39fe09d5a87efbbc1f15b622625a4dc62df2551f 100644 (file)
@@ -1182,7 +1182,7 @@ int perf_session__peek_event(struct perf_session *session, off_t file_offset,
                return -1;
 
        if (lseek(fd, file_offset, SEEK_SET) == (off_t)-1 ||
-           readn(fd, &buf, hdr_sz) != (ssize_t)hdr_sz)
+           readn(fd, buf, hdr_sz) != (ssize_t)hdr_sz)
                return -1;
 
        event = (union perf_event *)buf;
@@ -1190,12 +1190,12 @@ int perf_session__peek_event(struct perf_session *session, off_t file_offset,
        if (session->header.needs_swap)
                perf_event_header__bswap(&event->header);
 
-       if (event->header.size < hdr_sz)
+       if (event->header.size < hdr_sz || event->header.size > buf_sz)
                return -1;
 
        rest = event->header.size - hdr_sz;
 
-       if (readn(fd, &buf, rest) != (ssize_t)rest)
+       if (readn(fd, buf, rest) != (ssize_t)rest)
                return -1;
 
        if (session->header.needs_swap)