On 11/06/2011 07:25 PM, Paolo Bonzini wrote: > On 11/06/2011 03:27 PM, Avi Kivity wrote: >> On 10/20/2011 01:16 PM, Paolo Bonzini wrote: >>> This does the first part of the conversion to coroutines, by >>> wrapping bdrv_read implementations to take the read side of the >>> rwlock. >>> >>> Drivers that implement bdrv_read rather than bdrv_co_readv can >>> then benefit from asynchronous operation (at least if the underlying >>> protocol supports it, which is not the case for raw-win32), even >>> though they still operate with a bounce buffer. >>> >>> raw-win32 does not need the lock, because it cannot yield. >>> nbd also doesn't probably, but better be safe. >> >> This patch (2914caa088e3fbbd) breaks autotest when a guest reboots after >> install; instead of rebooting, the guest is stuck in the bootloader or >> kernel. > > Are any of these formats used by autotest? >
It's configurable; in my case, qcow2. > block/bochs.c | 13 ++++++++++++- > block/cloop.c | 13 ++++++++++++- > block/cow.c | 13 ++++++++++++- > block/dmg.c | 13 ++++++++++++- > block/nbd.c | 13 ++++++++++++- > block/parallels.c | 13 ++++++++++++- > block/vmdk.c | 13 ++++++++++++- > block/vpc.c | 13 ++++++++++++- > block/vvfat.c | 13 ++++++++++++- > 9 files changed, 108 insertions(+), 9 deletions(-) > So no. > Perhaps the failure is only reproduced 80-90% of the time and this > screws up the bisection. I thought I checked the before/after commit, but looking at the diffstat, that's doesn't make sense. On a related note, booting with -cdrom http://blah seems broken. -- error compiling committee.c: too many arguments to function