QEMU native compile issue
Friday I ran into an issue compiling qemu-native. I tracked the issue to the the fact that the most recent kernel creates a new definition for SIOCGSTAMP in sockios.h. The prior definition was in sys/socket.h. I have pushed a change request to the QEMU code base. It's a one-liner:
linux-user/syscall.c:
#include <linux/random.h>
+#include <linux/sockios.h>
Signature · 6 lines
Johnathan Mantey Senior Software Engineer *azad te**chnology partners* Contributing to Technology Innovation since 1992 Phone: (503) 712-6764 Email: [email protected] <mailto:[email protected]>
Hi Johnathan,
Quoted 4 lines · Johnathan Mantey
Friday I ran into an issue compiling qemu-native. I tracked the issue to the the fact that the most recent kernel creates a new definition for SIOCGSTAMP in
sockios.h. The prior definition was insys/socket.h. I have pushed a change request to the QEMU code base.
Can you provide a link to the change request that you pushed?
Cheers,
Andrew
I've reviewed the QEMU commits:
6d5d5dde9a linux-user: fix to handle variably sized SIOCGSTAMP with new kernels
I didn't push the commit to somewhere useful.
Quoted 12 lines · Andrew Jeffery
Hi Johnathan,
Quoted 4 lines · Johnathan Mantey
Friday I ran into an issue compiling qemu-native. I tracked the issue to the the fact that the most recent kernel creates a new definition for SIOCGSTAMP in
sockios.h. The prior definition was insys/socket.h. I have pushed a change request to the QEMU code base.Can you provide a link to the change request that you pushed?
Cheers,
Andrew
Signature · 6 lines
Johnathan Mantey Senior Software Engineer *azad te**chnology partners* Contributing to Technology Innovation since 1992 Phone: (503) 712-6764 Email: [email protected] <mailto:[email protected]>
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 5 lines · Johnathan Mantey
Not really. I'm not sure where the code went.
I cloned the OpenBMC QEMU, made my change, and then pushed the change
to "upstream". I have SHA "650af0e183Fix compile error when using kernel 5.2" as the text.
As you noted in your follow-up email you found the fix upstream. Further, the fix is contained in the QEMU v4.1 tag.
Upstream open-embedded has already bumped QEMU to v4.1 - I guess we should backport that change to openbmc/openbmc and this will be resolved. I've Cc'ed Brad so he's across it.
As an aside, while we have the openbmc/qemu repo on GitHub, it's not integrated into openbmc/openbmc, openbmc/qemu is a bit less volatile than Cedric's legoater/qemu aspeed-4.x branches (as it's used for OpenBMC CI purposes), but has more models implemented than what upstream typically has (I intend to update openbmc/qemu to 4.1 with some of the patches from legoater/qemu aspeed-4.2 on top in the near future). The QEMU that's built as part of the OpenBMC build process is the upstream release.
Hope that helps.
Andrew
Quoted 7 lines · Andrew Jeffery
Quoted 5 lines · Johnathan Mantey
Not really. I'm not sure where the code went.
I cloned the OpenBMC QEMU, made my change, and then pushed the change
to "upstream". I have SHA "650af0e183Fix compile error when using kernel 5.2" as the text.
Quoted 6 lines
As you noted in your follow-up email you found the fix upstream. Further, the fix is contained in the QEMU v4.1 tag.
Upstream open-embedded has already bumped QEMU to v4.1 - I guess we should backport that change to openbmc/openbmc and this will be resolved. I've Cc'ed Brad so he's across it.
I think this was picked up with openbmc c68388fccb8c0b5bf4d6b8efff91203796be98b2.
If that doesn’t have what we needed...I’m blocked on doing further oe-core updates because of this:
https://lists.ozlabs.org/pipermail/openbmc/2019-September/017984.html
-brad
Brad,
FYI...
Commit c68388fc.... is part of my latest pull, which is 9d227052b.
The upstream fix for QEMU native is not present in any commit 9d227052b. When you are unblocked please collect the QEMU update.
Quoted 27 lines · Brad Bishop
Quoted 7 lines · Andrew Jeffery
Quoted 5 lines · Johnathan Mantey
Not really. I'm not sure where the code went.
I cloned the OpenBMC QEMU, made my change, and then pushed the change to "upstream". I have SHA "
650af0e183Fix compile error when using kernel 5.2" as the text.Quoted 8 lines
As you noted in your follow-up email you found the fix upstream. Further, the fix
is contained in the QEMU v4.1 tag.Upstream open-embedded has already bumped QEMU to v4.1 - I guess we should backport that change to openbmc/openbmc and this will be resolved.
I've Cc'ed Brad so he's across it.I think this was picked up with openbmc
c68388fccb8c0b5bf4d6b8efff91203796be98b2.If that doesn’t have what we needed...I’m blocked on doing further oe-core updates because of this:
https://lists.ozlabs.org/pipermail/openbmc/2019-September/017984.html
-brad
Signature · 6 lines
Johnathan Mantey Senior Software Engineer *azad te**chnology partners* Contributing to Technology Innovation since 1992 Phone: (503) 712-6764 Email: [email protected] <mailto:[email protected]>
Quoted 8 lines · Johnathan Mantey
Brad,
FYI...
Commit
c68388fc.... is part of my latest pull, which is9d227052b.The upstream fix for QEMU native is not present in any commit
9d227052b. When you are unblocked please collect the QEMU update.
Thanks to Patrick we are unblocked now, and I pulled poky master HEAD yesterday. Everything looks good now?
-brad
Quoted 2 lines · Andrew Jeffery
(I intend to update openbmc/qemu to 4.1 with some of the patches from
legoater/qemu aspeed-4.2 on top in the near future).
I've now updated openbmc/qemu master to 4.1. CI appears to be chugging along fine, but please keep an eye out for qemu-related failures.
Andrew