Re: [PATCH 2/2] tools/migration: Fix potential overflow in send_checkpoint_dirty_pfn_list()

[Olaf Hering <olaf@xxxxxxxxx>]

Am Tue, 6 Jul 2021 14:19:21 +0100
schrieb Andrew Cooper <andrew.cooper3@xxxxxxxxxx>:

> > 20: faults= 0 dirty= 80  
> 
> What is this showing, other than (unsurprisingly) faults doesn't work
> for an HVM guest?

The dirty count goes down after a while for a domU that constantly touches as 
many pages as it can.
But yes, the current code uses stats.dirty_count in a number of places.
It seems all of the usage is just for reporting, so if these code paths would 
get wrong input nothing bad happens.
precopy_policy may indicate more or less iterations, this is not critical 
either.

Olaf

Attachment: pgpRfuRGq39v7.pgp
Description: Digitale Signatur von OpenPGP