On Thu, Jan 23, 2025 at 02:50:16PM +0100, Cédric Le Goater wrote: > > > Hi, this is the 7th revision of the vfio-user client implementation. > > Well, among the reasons why we tend to push this at end of > the list are :
Thanks for the steers. Just a note in case you're not aware - most of these changes weren't authored by me, I'm in the position of picking up someone else's series. I hope you'll bear with me in this respect. > - the respins are spaced over time (1 a year ?) Understandable, I will certainly make time to be more responsive when I do get review comments. The last spin got none, whereas the earlier ones sent out by Oracle got a good amount of review. You mentioned on IRC that it's worth me documenting previous reviews: I can spelunk for that data, but it's worth pointing out that many of the patches have changed significantly since previous reviews. > - it's invasive in an already very complex subsystem As a qemu neophyte I would greatly appreciate suggestions on how to reduce this. I think the user-container refactoring was a big one (we now barely touch the other container types), but I'm definitely up for further suggestions. > - it's HUGE. See the diffstat below ... Unavoidable right? The start of the series are generic preparation so could in theory be merged separately. If you have other ways you'd prefer me to split this up I'd be happy to do so. FYI I tried sending out just the specification document (which reflects the already-merged server side implementation) but did not get any reviews. > I would introduce a new hw/vfio-user/ subsystem given the size. Could you give me some further guidance? Are you talking about duplicating all of the needed hw/vfio/pci.c code into hw/vfio-user/pci.c ? Or do you just mean putting the new hw/vfio/user* files into hw/vfio-user/ > The commit logs are short. Most are one liners, this is really > not much for such a big change. The ones I've written have longer changelogs and I hope are better. I'll take some time to rewrite the other authors' commit messages for the next cycle. > Any how, I hope to take a look before the end of 10.0 cycle. Thanks so much! Would you prefer me to apply your above suggestions prior to you reviewing? regards john
