From: Bart V. A. <bva...@ac...> - 2014-10-14 07:02:22
|
On 10/13/14 22:12, Matteo Tescione wrote: > Path: scst-trunk > URL: svn://svn.code.sf.net/p/scst/svn/trunk > Repository Root: svn://svn.code.sf.net/p/scst/svn > Repository UUID: d57e44dd-8a1f-0410-8b47-8ef2f437770f > Revision: 5704 > > drbd version: 8.4.4 (api:1/proto:86-101) from standard package, > running kernel 3.12.14 patched with scst_exec_req_fifo and > put_page_callback applied. Hello Matteo, To me this looks like a DRBD bug. If a timeout of a DRBD request occurs the DRBD timeout handler (request_timer_fn()) is called, that function grabs a spinlock and next calls kthread_create_on_node() with that spinlock held. kthread_create_on_node() can sleep which means that it is not allowed to call that function with a spinlock held. I'm not sure why this particular behavior has not yet been noticed before. Can you post the first call stack from the e-mail at the start of this thread together with the kernel version number on a DRBD-related mailing list and ask the DRBD authors for their opinion ? Thanks, Bart. |