On Thu, Jul 7, 2016 at 2:55 PM, Tomohiro Kusumi <[email protected]> wrote: > I've changed SLIST to TAILQ at 3cd1dc08 a while ago, but it seems the > location of insertion comes from NetBSD.
The insertion before the init is highly suspicious, since he got a NULL de-reference panic. However, I don't know why the I/O happens simultaneously w/ the init. > > 2016-07-07 12:16 GMT+09:00 Sepherosa Ziehau <[email protected]>: >> On Thu, Jul 7, 2016 at 10:57 AM, Sepherosa Ziehau <[email protected]> >> wrote: >>> On Wed, Jul 6, 2016 at 5:54 AM, Stefan Unterweger >>> <[email protected]> wrote: >>>> Yet another problem on the same virtualised host as in the other thread >>>> a few weeks back. >>>> >>>> As before, this is a presumably KVM virtualised host. With the >>>> exception of a small boot partition, all storage volumes are encrypted >>>> via cryptsetup‘s LUKS mode. >>>> >>>> It looks like a race condition of some weird sort, and there’s no doubt >>>> that the virtualisation is again playing tricks with me. If I type in >>>> the password incorrectly, then it will most likely crash violently right >>>> there. If typed in correctly, then in one of five it -might- boot. The >>>> kernel panic is attached below. >>> >>> It looks like a race between the dm init and the I/O path. i.e. the >>> read happens even before the dm init is completed. Not sure about >>> why... >> >> Try this patch: >> https://leaf.dragonflybsd.org/~sephe/dfly_dm_order.diff >> >> Thanks, >> sephe >> >> -- >> Tomorrow Will Never Die -- Tomorrow Will Never Die
