On Fri, May 23, 2014 at 02:49:13PM +0200, Alexander Graf wrote:

.../...

VFIO_EEH_RECOVER?

Then it should contain VFIO_EEH_PE_RESET, which is part of the
recovery. I don't expect the recovery to be done with one ioctl
command :-)

I personally prefer to keep VFIO_EEH_PE_CONFIUGRE, which means
the PE need configurtion/reconfiguration before error log retrieval
or after PE reset as defined in PAPR spec.

So what if user space accesses config space while the device is
broken? What if it accesses an mmap'ed BAR while the device is in
broken state and BARs haven't been recovered yet?

If the "broken" means device can't funtion properly, any config,
IO, DMA traffic would cause "frozen" device (PE).

If the "broken" means "frozen" here. When the device is put into
frozen state, the config or IO access are dropped from hardware or
software level. So those requests can't reach to hardware under the
situation.

Thanks,
Gavin