> YAMAMOTO Takashi <[email protected]> wrote: > >> ioflush (syncer) is not a kernel thread responsible for freeing memory. >> pagedaemon is. >> >> i don't think syncer alone can deadlock with puffs. > > Indeed it does not: there is at least a third thread involved. In a > typical case, ioflush sleeps awaiting for perfused/glusterfsd to finish > fsync, and perfused/glusterfsd is stuck in the kernel awaiting memory > allocation.
my bet is that ioflush is merely a victim of the deadlock. YAMAMOTO Takashi > > -- > Emmanuel Dreyfus > http://hcpnet.free.fr/pubz > [email protected]
