On 26/08/14 13:01, David Gibson wrote: > On Fri, Aug 22, 2014 at 10:24:10AM +1000, Samuel Mendoza-Jonas wrote: >> If a guest reboots during a running migration, changes to the >> hash page table are not necessarily updated on the destination. >> Opening a new file descriptor to the HTAB forces the migration >> handler to resend the entire table. > > Of course, arguably this is a kernel bug - the htab fd should probably > start re-reporting things if the htab reset ioctl() is called. But I > guess we need a workaround for existing kernels anyway.
That's right, but as you say it would be nice to cover existing kernels. > > However, don't we still have the bug for TCG mode? spapr_reset_htab() > just memset()s the whole hashtable in that case, which will actually > clear all the dirty bits, whereas we need to set them. > True! I'll add a case for this, it should just be a matter of checking & setting htab_first_pass. -- Regards, Sam Mendoza-Jonas ----------- LTC Ozlabs IBM
