On 2 September 2016 at 08:59, Wang Nan <wangn...@huawei.com> wrote: > The requirement of this function is first proposed at 2015. > Please refer to > > http://lkml.iu.edu/hypermail/linux/kernel/1506.2/02372.html > http://lkml.iu.edu/hypermail/linux/kernel/1506.3/02290.html > http://lkml.iu.edu/hypermail/linux/kernel/1506.3/03512.html > > For systems which use hugetlbfs, if a sample is captured inside > hugetlbfs, perf should not resolve symbols from the file shown in > /prof/<pid>/mmap, because 'files' in hugetlbfs are constructed > during execution and not ELF files. If user knows positions of > symbols, he/she should provide /tmp/perf-<pid>.map file. > > This 3 patches makes perf recognize hugetlbfs mapping as anon mapping. > Before this 3 patches user has no chance to use his/her own .map file > to resolve symbols because perf tries to use hugetlbfs file. > > Wang Nan (3): > perf tools: Recognize hugetlb mapping as anon mapping > tools lib api fs: Add hugetlbfs filesystem detector > perf record: Mark MAP_HUGETLB during synthesizing mmap events > > tools/lib/api/fs/fs.c | 15 +++++++++++++++ > tools/lib/api/fs/fs.h | 1 + > tools/perf/util/event.c | 7 +++++++ > tools/perf/util/map.c | 8 +++++--- > 4 files changed, 28 insertions(+), 3 deletions(-) > > Signed-off-by: Wang Nan <wangn...@huawei.com> > Cc: Hou Pengyang <houpengy...@huawei.com> > Cc: He Kuang <heku...@huawei.com> > Cc: Arnaldo Carvalho de Melo <a...@redhat.com>
Reviewed-by: Nilay Vaish <nilayva...@gmail.com>
