Re: [libvirt PATCH 0/6] cleanup and fixes of the mapped-ram feature

On Fri, Mar 21, 2025 at 08:29:30AM +0000, Daniel P. Berrangé via Devel wrote: Well, restore should figure out the format and save still has --image-format without which parallel save will fail (unless you have it set as default in a config). So I think this series is fine, at least it fixes the unit tests. Reviewed-by: Martin Kletzander <mkletzan(a)redhat.com&gt; But there's more to it than meets the eye. There is also a check that the number of parallel channels is not lower than 1, for saving. Restoring happily takes parallel.channels=0 and (at least for non-sparse images) fails weirdly with: error: Failed to restore domain from test.img error: Failed to open file 'test.img': No such file or directory and the daemon reports: 2025-03-21 10:00:08.490+0000: 4076349: error : virFileIsSharedFSType:3598 : Invalid relative path 'test.img': Invalid argument 2025-03-21 10:00:08.490+0000: 4076349: error : virQEMUFileOpenAs:10448 : Failed to open file 'test.img': No such file or directory even though the file exists. And it works without --parallel-channels, this only happens with --parallel-channels 0. Last, but not least, our CI is broken with the patches here. And that is because now one cannot do save-image-edit, with both the new and the old format with: error: failed to write header to domain save file '/home/nert/test.img': Bad file descriptor And that's about what I've found out. I'll spend some time on this, trying to fix it up, but if anyone has a fix ready, then even better. What I'd like to know is whether we can