On 08/22/2011 02:09 PM, Dan Carpenter wrote: > On Mon, Aug 22, 2011 at 01:47:49PM -0500, Seth Jennings wrote: >> zcache_new_pool() calls kmalloc() with GFP_KERNEL which has >> __GFP_WAIT set. However, zcache_new_pool() gets called on >> a stack that holds the swap_lock spinlock, leading to a >> possible sleep-with-lock situation. The lock is obtained >> in enable_swap_info(). >> >> The patch replaces GFP_KERNEL with GFP_IOFS, which is >> GFP_KERNEL & ~__GFP_WAIT. >> > > You should use GFP_ATOMIC. We don't want to do IO with the locks > held. The only reason GFP_IOFS exists is so that we can turn off > io during suspend and resume. > > regards, > dan carpenter > I guess I was looking to change it as little as possible and didn't know what allocations should be allowed to use the "emergency pool" of pages. I'll update and resend. Thanks -- Seth _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
