I'm following up to my earlier message, where I pointed out
that current kernels generate a RIDICULOUS number of context
switches when streaming to disk.  I've tracked this down a little
and the problem is that bdflush is being awoken sometimes 
30-40000 times per second.  this is obviously not a good thing!

I haven't figured out why, but the bdflush loop is very odd
looking, since it contains a schedule() but no wait_on.  could 
someone explain how it's supposed to be throttled?

thanks, mark hahn.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/