On Wed, Aug 08, 2018 at 10:30:23AM +0700, Robert Elz wrote: > This suggests to me that something is getting totally scrambled in > the buf headers when things get busy.
I tried to crash with BIOHIST enabled, here is he story about the buf_t that triggers the panic beause it as BO_DONE: 1533732322.896722 bread#23928@0: called! 1533732322.896727 biowait#53190@0: called: bp=0xffffa000057d96c0, oflags=0x0, ret_addr=0xffffffff804c2e40 1533732322.896727 biowait#53190@0: waiting bp=0xffffa000057d96c0 1533732322.897058 biodone#16699@0: called: bp=0xffffa000057d96c0, softint scheduled 1533732322.897058 biodone#16700@0: called: bp=0xffffa000057d96c0, softint scheduled 1533732322.897064 biointr#16597@0: called! 1533732322.897064 biointr#16597@0: bp=0xffffa000057d96c0 1533732322.897064 biodone2#16777@0: called: bp=0xffffa000057d96c0 1533732322.897064 biodone2#16777@0: wake-up 1533732322.897065 biointr#16597@0: bp=0xffffa000057d96c0 1533732322.897065 biodone2#16778@0: called: bp=0xffffa000057d96c0 There does not seems to be anything corrupted. the buf_t is handled twice in biointr because it has been inserted twice in ci_data.cpu_biodone by biodone. -- Emmanuel Dreyfus m...@netbsd.org
