https://bugs.kde.org/show_bug.cgi?id=511717

--- Comment #35 from Philippe Waroquiers <[email protected]> ---
(In reply to Libor Peltan from comment #34)
> Happy New Year Philippe,
Happy new year for you too.

> thanks much for your analysis!
> 
> I confirm that with the above patch to coregrind/m_syswrap/syswrap-generic.c
> , the perceived issue disappears.
Thanks for the quick feedback.
> 
> Anyway, is there anything (e.g. compile options, tiny code changes...) we
> can do in our software (Knot DNS) to prevent glibc from this new behavior in
> order to workaround the issue on our side, before a solution in Valgrind is
> found?
I have not seen in glibc code any flag or tunable to force glibc to use
mprotect.
I believe the only way to do that from the application side is to have the
program allocating the thread stack
(using e.g. mmap) and using mprotect+pthread_attr_setstack 
> 
> My expectation is that with the new Ubuntu LTS on April, the newer glibc
> will be used more widely by everyone.
Effectively with next Ubuntu LTS (and with other distributions switching to a
newer glibc),
this problem will become much more frequent.

IMO, we should put a fix in valgrind in the coming months.

If finally we believe that what needs to be done is to make valgrind address
space manager
more intelligent; we might still in the meantime commit the patch you have
tried 
as a temporary measure.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to