看板 DFBSD_bugs 關於我們 聯絡資訊
:On Wed, Oct 21, 2009 at 10:36:36AM -0700, Matthew Dillon wrote: :> fetch http://apollo.backplane.com/DFlyMisc/hammer05.patch :> :> It survived my bulk build tests so now the question is whether it :> will survive yours. : :It survived 7.5 hours before crashing in a different place. :The kernel and the vmcore are at ~y0etan1/crash/{kernel,vmcore}.10. : :panic: assertion: s <= 0 in hammer_btree_iterate :mp_lock = 00000000; cpuid = 0 The cursor is pointing at the wrong place again, this time at the wrong index of an internal node. Try this patch. The only thing I found looking at the code was a possible problem with hammer_cursor_up_locked(). If the routine fails due to a deadlock it does not properly restore the cursor's index. This is a rollup patch which contains all active uncommitted patches for HAMMER. fetch http://apollo.backplane.com/DFlyMisc/hammer06.patch -Matt