> Does ''xm save / xm restore'' work with this patch (at least as well as it > currently does ;-)?xm save/restore still doesn''t work for me, neither with the sparse tree nor with the linux-2.6 repository, so I can''t try. I can''t see any reason why it should become worse with that patch though. I''ve resynced the blkback threading patch with the latest sparse tree, here we are. Changes: * One thread per blkif. The I/O scheduler can do a better job that way, also you can use ionice on the blkback threads to adjust the block I/O priorities for the domain. * Various stuff has been moved from global variables into blkif_t. * The scary allocation ring for pending_req''s is gone and has been replaced by a free list. * made dispatch_rw_block_io() reentrant. * general linux coding style cleanup, at least for the code I''ve touched anyway. * number of outstanding requests is runtime-configurable now. * made the ia64 #ifdefs smaller and dropped one. It should still work on ia64 in theory, but would be great if the ia64 folks can have a look ... cheers, Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> I''ve resynced the blkback threading patch with the latest sparse tree, > here we are. Changes: > > * One thread per blkif. The I/O scheduler can do a better job that > way, also you can use ionice on the blkback threads to adjust the > block I/O priorities for the domain. > * Various stuff has been moved from global variables into blkif_t. > * The scary allocation ring for pending_req''s is gone and has been > replaced by a free list. > * made dispatch_rw_block_io() reentrant. > * general linux coding style cleanup, at least for the code I''ve > touched anyway. > * number of outstanding requests is runtime-configurable now. > * made the ia64 #ifdefs smaller and dropped one. It should still > work on ia64 in theory, but would be great if the ia64 folks > can have a look ...Next version of that patch, with those additional changes: * re-added the xen_init() which got lost by mistake (pointed out by the ia64 guys). * adapted to the driver architecture changes merged recently. cheers, Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Hi folks,
New version of the patch, adapted to apply cleanly against latest
unstable, also debug output is runtime-switchable now, no major changes
from last version. I''m running my xen machine all day with that patch,
without any issues showing up, including working save/restore. Any
chance to get it merged?
Full list changes:
* One thread per blkif. The I/O scheduler can do a better job that
way, also you can use ionice on the blkback threads to adjust the
block I/O priorities for the domain.
* Various stuff has been moved from global variables into blkif_t.
* The scary allocation ring for pending_req''s is gone and has been
replaced by a free list.
* made dispatch_rw_block_io() reentrant.
* general linux coding style cleanup, at least for the code I''ve
touched anyway.
* number of outstanding requests is runtime-configurable now.
* made the ia64 #ifdefs smaller and dropped one. It should still
work on ia64 in theory, but would be great if the ia64 folks
can have a look ...
* re-added the xen_init() which got lost by mistake (pointed
out by the ia64 guys).
* runtime-switchable stats and debug output.
cheers,
Gerd
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
On 24 Nov 2005, at 11:31, Gerd Knorr wrote:> New version of the patch, adapted to apply cleanly against latest > unstable, also debug output is runtime-switchable now, no major > changes from last version. I''m running my xen machine all day with > that patch, without any issues showing up, including working > save/restore. Any chance to get it merged?The patch is looking good, but we really want to limit patches to bug fixes right now, to rein in 3.0.0. The risk of small regressions and bugs creeping, and time required to thoroughly review the patch, are things we''d like to avoid for the next week or so. For example, I see that one of the recent fixes (initialisation of pending_grant_handles array) isn''t included in the current patch. We''ll be much more receptive after 3.0.0 is out the door. Improved QoS control is one of the targets for 3.0.x. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Gerd Knorr wrote:> Hi folks, > > New version of the patch, adapted to apply cleanly against latestOne more version, this time against 3.0-final ;) cheers, Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel