To provide virtio-fs on s390x as well, the following changes are needed: - add a virtio-ccw proxy device (analogous to the virtio-pci device) - allow fd sharing to be configured on targets that do not support numa This is to reach feature parity between architectures.
------- Comment From PASIC.com 2020-05-18 06:35 EDT------- I have patches for: and was taken care of by Igor Mammedov. The outstanding issue that prevents virtio-fs form working is libvhost-user does not do endianness properly. I spoke with Marc about this some time ago, and he showed some willingness to tackle this problem if I remember correctly. I will coordinate myself with Marc.
I'm afraid your comment did not come through correctly -- I assume you were saying that you have patches for adding the virtio-ccw proxy and that the fd sharing was taken care of by Igor? The libvhost-user endianness problem is interesting; fixing this would probably be the last missing part as far as we know?
(In reply to Cornelia Huck from comment #4) > I'm afraid your comment did not come through correctly -- I assume you were > saying that you have patches for adding the virtio-ccw proxy and that the fd > sharing was taken care of by Igor? > . ... yes, comment #3 has to read: " I have patches for: - add a virtio-ccw proxy device (analogous to the virtio-pci device) and - allow fd sharing to be configured on targets that do not support numa was taken care of by Igor Mammedov. The outstanding issue that prevents virtio-fs form working is libvhost-user does not do endianness properly. I spoke with Marc about this some time ago, and he showed some willingness to tackle this problem if I remember correctly. I will coordinate myself with Marc. " . ... sorry for the noise ...
------- Comment From PASIC.com 2020-05-19 07:20 EDT------- (In reply to comment #6) Yes. after s/^-/*/ """ * add a virtio-ccw proxy device (analogous to the virtio-pci device) * allow fd sharing to be configured on targets that do not support numa """ > The libvhost-user endianness problem is interesting; fixing this would > probably be the last missing part as far as we know? Yes. Marc started working on it yesterday. Bluntly all virtioXX_to_cpu() type conversions are simply missing. > > (In reply to Cornelia Huck from comment #4) > > I'm afraid your comment did not come through correctly -- I assume you were > > saying that you have patches for adding the virtio-ccw proxy and that the fd > > sharing was taken care of by Igor? > > > . > ... yes, comment #3 has to read: > " > I have patches for: > and > was taken care of by Igor Mammedov. > > The outstanding issue that prevents virtio-fs form working is libvhost-user > does not do endianness properly. I spoke with Marc about this some time ago, > and he showed some willingness to tackle this problem if I remember > correctly. I will coordinate myself with Marc. > " > . > ... sorry for the noise ...
oops ... the previous comment has to read: " (In reply to comment #6) > ------- Comment From cohuck 2020-05-19 09:35:16 EDT------- > I'm afraid your comment did not come through correctly -- I assume you were > saying that you have patches for adding the virtio-ccw proxy and that the fd > sharing was taken care of by Igor? > Yes. after s/^-/*/ """ I have patches for: * add a virtio-ccw proxy device (analogous to the virtio-pci device) and * allow fd sharing to be configured on targets that do not support numa was taken care of by Igor Mammedov. The outstanding issue that prevents virtio-fs form working is libvhost-user does not do endianness properly. I spoke with Marc about this some time ago, and he showed some willingness to tackle this problem if I remember correctly. I will coordinate myself with Marc. """ > The libvhost-user endianness problem is interesting; fixing this would > probably be the last missing part as far as we know? Yes. Marc started working on it yesterday. Bluntly all virtioXX_to_cpu() type conversions are simply missing. > > ------- Comment From hannsj_uhl.com 2020-05-19 09:37:04 EDT------- > (In reply to Cornelia Huck from comment #4) > > I'm afraid your comment did not come through correctly -- I assume you were > > saying that you have patches for adding the virtio-ccw proxy and that the fd > > sharing was taken care of by Igor? > > > . > ... yes, comment #3 has to read: > " > I have patches for: > and > was taken care of by Igor Mammedov. > > The outstanding issue that prevents virtio-fs form working is libvhost-user > does not do endianness properly. I spoke with Marc about this some time ago, > and he showed some willingness to tackle this problem if I remember > correctly. I will coordinate myself with Marc. > " > . > ... sorry for the noise ... " ...
------- Comment From MHartmay.com 2020-06-25 06:20 EDT------- I sent an RFC to the QEMU mailing list: https://lists.gnu.org/archive/html/qemu-devel/2020-06/msg08133.html
We'll likely get this feature via a rebase in RHEL-AV, so I'm closing this ticket now.