https://bugzilla.novell.com/show_bug.cgi?id=253968
User npiggin@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=253968#c13
Nick Piggin changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |andrea@kernel.org
--- Comment #13 from Nick Piggin 2008-01-10 22:19:12 MST ---
It's interesting with this bug: some things (including) md_raid thread
are stuck in mempool_alloc -> page reclaim.
I guess the system may not be stuck, just livelocked. And this could be
a contributing factor for why systems sometimes go unresponsive when
they are near OOM.
I think mempool allocation code could be enhanced to do a fair (eg. FIFO)
distribution of pool elements to allocators. I also think it should probably
not do much work in page reclaim, especially if it is having problems freeing
stuff.
Perhaps we don't even want it to enter page reclaim at all (but kick kswapd if
needed). On the other hand, it is nice to get some parallelism in page reclaim,
and we don't necessarily want to leave that work up to the non-mempool allocs.
I don't really know, but it is something I'm thinking about...
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.