On 30/01/14 03:31, Eric Blake wrote:
On 01/29/2014 10:22 AM, Osier Yang wrote:
It doesn't make sense to fail if the SCSI host device is specified as "shareable" explicitly between domains (NB, it works if and only if the device is specified as "shareable" for *all* domains, otherwise it fails).
To fix the problem, this patch introduces an array for virSCSIDevice struct, which records all the names of domain which are using the device (note that the recorded domains must specifiy the device as s/specifiy/specify/
shareable). And the change on the data struct brings on many subsequent changes in the code.
Since prior to this patch, the "shareable" tag didn't work as expected, it actually work like "non-shareable". If there was a live domain using the SCSI device with "shareable" specified, then after upgrading (assuming the live domain keep running during upgrading) the older libvirt (without this patch) to newer libvirt (with this patch), it may cause some confusion when user tries to start later domains as "shareable". So this patch also added notes in formatdomain.html to declare the fact. I'm not sure I follow what you're trying to warn about here - when you upgrade libvirtd, doesn't the state reloading figure out which domains have a device in use, as part of the reload machinery? That is, while the old code mishandled shareable with regards to which domain is using the device, we don't actually store the array of domains using a device in XML, but recompute it on libvirtd restart. Thus, upgrading to a newer libvirtd (whether release 1.2.2 or to a version with this patch backported), then starting yet another guest with a shareable request, should work correctly, right?
I think both John and I were confused and forgot upgrading will triger configuration reloading. :-) Will reword the commit log a bit when pushing. Thanks. Osier