http://bugzilla.suse.com/show_bug.cgi?id=998850
http://bugzilla.suse.com/show_bug.cgi?id=998850#c7
--- Comment #7 from Michal Hocko ---
And some more data. Many allocations seem to be just fine because the same
address is returned again and again. So it must have been freed in the
meantime. We have
$ zgrep kmem_cache_alloc: kmemlog.gz | sed 's@.*ptr=\([0-9a-f]*\) .*@\1@' |
sort | uniq -c | grep -w 1 | wc -l
649
some returned just once. This alone doesn't mean much. It might be in use or
just not reused yet. But if we look at which of the call sites sees most of
those then
$ zgrep "call_site=ffffffff811903b2" kmemlog.gz | sed 's@.*ptr=\([0-9a-f]*\)
.*@\1@' | sort | uniq -c | grep -w 1 | wc -l
774
rffffffff811903b2 really stands out.
So I would be really curious who is that guy.
Btw. the call site might be an offset into a function is it is quite possible
that a pure grep on the file will not give you anything. So either attach your
kallsyms file here or find the last function with address smaller than the
given one.
--
You are receiving this mail because:
You are on the CC list for the bug.