Merge ~paelzer/ubuntu/+source/qemu:lp-1829380-vhost-1828288-amd-xenial into ubuntu/+source/qemu:ubuntu/xenial-devel

[Christian Ehrhardt ]

[Christian Ehrhardt  (paelzer)]

PPA: https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/3729

Adding original Authors as reviewers -> Dan/Rafael

[Christian Ehrhardt  (paelzer)]

It seems 1:2.5+dfsg-5ubuntu10.39 wasn't imported correctly yet which makes this look odd in LP at the moment.
For now, if you clone the git branch you should get a better view I'd think.

[Christian Ehrhardt  (paelzer)]

I rebased to a new git fetch, now things look more reasonable

[Rafael David Tinoco (rafaeldtinoco)]

@Christian,

Just discovered that might be something else going on blocking my AMD QEMU test. This time I was able to boot my guest with versions 10.39 (-updates) and yours (ppa) - and I'm not able to boot nested KVM.

So, for my patch, only difference was esthetic since it did not show the unsupported CPUID features in libvirt qemu log. You may keep it in the SRU or remove, if you judge its not worth. The patch is small and already tested, might be good to keep it, just wanted you to know that it is not fixing boot issue in AMD, just the feature check.

Thanks

Rafael

[Christian Ehrhardt  (paelzer)]

@Rafael - thanks for the test - Lets fix the aesthetics now that things seem ready and prepared, but throw anything on the "other" issue blocking your boot into a new bug then please.

[Rafael David Tinoco (rafaeldtinoco)]

@Christian,

Finished testing nested AMD KVM (AMD FX(tm)-8350). Cosmic is okay, all the others are "broken" for nested KVM. I just discovered it seems to be a kernel issue since I installed 4.19 in Xenial (1st level) running on top of Debian Sid (4.19) and it was able to boot my nested KVM guest.

I might have to bisect kernel for this :\, maybe add this to server-next somehow ? (Unsure because of possible LTS openstack releases that might be affected by this).

Cheers o/

[Dan Streetman (ddstreet)]

Since this is the same code content as my pkg in the upload queue, I approve of the patch for lp #1829380, that i tested with:
https://launchpad.net/~ddstreet/+archive/ubuntu/lp1829380/+packages

I will defer to tinoco to approve his patch.

review: Approve

[Rafael David Tinoco (rafaeldtinoco)]

Thanks Dan!

Yep, It's good for me, since my AMD issue does not seem to have anything to do with previous QEMU version, or this one. If Christian is okay with my SRU being aesthetics... I can confirm package worked for me!!

Thanks a lot!

[Rafael David Tinoco (rafaeldtinoco)]

review: Approve (tested package and reviewed code.)

[Christian Ehrhardt  (paelzer)]

Thanks you two.
I have verified that the code in -unapproved by Dan is (other than changelog wording and patch naming) identical.
So we can leave that upload there as is without replacing it.

I'm currently wrestling with my test setup, once working I'll comment on the bugs as well.

[Christian Ehrhardt  (paelzer)]

Since this will be uploaded by ddstreet (in unapproved already) we can set this one to "rejected"

Unmerged commits

c2fd05a... by Christian Ehrhardt  on 2019-05-28

changelog: Fix issues with CPUID_EXT2_AMD_ALIASES allowing guests using cpu passthrough to boot. (LP: #1828288)

Signed-off-by: Christian Ehrhardt <email address hidden>

d09e09b... by Christian Ehrhardt  on 2019-05-28

Fix issues with CPUID_EXT2_AMD_ALIASES allowing guests using cpu passthrough to boot. (LP: #1828288)

Signed-off-by: Christian Ehrhardt <email address hidden>

0b310e6... by Christian Ehrhardt  on 2019-05-28

changelog: Re-enable patches for (LP: 1823458) with a fixup for the former regression added (LP: #1829380)

Signed-off-by: Christian Ehrhardt <email address hidden>

17f438f... by Christian Ehrhardt  on 2019-05-28

Re-enable patches for (LP: 1823458) with a fixup for the former regression added (LP: #1829380)

Signed-off-by: Christian Ehrhardt <email address hidden>

diff --git a/debian/changelog b/debian/changelog
index 5342ca8..edc5b6d 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -1,3 +1,19 @@
+qemu (1:2.5+dfsg-5ubuntu10.40) xenial; urgency=medium
+
+  [ Dan Streetman ]
+  * Re-enable patches for (LP: 1823458) with a fixup for the former
+    regression added (LP: #1829380).
+    - d/p/lp1823458/add-VirtIONet-vhost_stopped-flag-to-prevent-multiple.patch
+    - d/p/lp1823458/do-not-call-vhost_net_cleanup-on-running-net-from-ch.patch
+    - d/p/lp1823458/lp1829380.patch
+
+  [ Rafael David Tinoco ]
+  * d/p/lp1828288/target-i386-Set-AMD-alias-bits-after-filtering-CPUID.patch
+    - Fix issues with CPUID_EXT2_AMD_ALIASES allowing guests using
+      cpu passthrough to boot. (LP: #1828288)
+
+ -- Christian Ehrhardt <[e-mail]>  Tue, 28 May 2019 09:35:55 +0200
+
 qemu (1:2.5+dfsg-5ubuntu10.39) xenial-security; urgency=medium
 
   * Disable patches from 1:2.5+dfsg-5ubuntu10.37 to prevent regression
diff --git a/debian/patches/lp1823458/lp1829380.patch b/debian/patches/lp1823458/lp1829380.patch
 new file mode 100644
index 0000000..06b3f62
--- /dev/null
+++ b/debian/patches/lp1823458/lp1829380.patch
@@ -0,0 +1,20 @@
+From: Dan Streetman <[e-mail]>
+Subject: Fix regression of failing qemu guest networking
+Bug-Ubuntu: https://launchpad.net/bugs/1829380
+
+This fixes the regression detailed in LP: #1829245
+
+The original patch that caused the regression is from LP: #1823458
+
+This fix will be uploaded to Xenial in LP: #1829380
+
+--- qemu-2.5+dfsg.orig/hw/net/virtio-net.c
++++ qemu-2.5+dfsg/hw/net/virtio-net.c
+@@ -140,6 +140,7 @@ static void virtio_net_vhost_status(Virt
+             qemu_net_queue_purge(qnc->incoming_queue, qnc->peer);
+         }
+ 
++        atomic_set(&n->vhost_stopped, 0);
+         n->vhost_started = 1;
+         r = vhost_net_start(vdev, n->nic->ncs, queues);
+         if (r < 0) {
diff --git a/debian/patches/lp1828288/target-i386-Set-AMD-alias-bits-after-filtering-CPUID.patch b/debian/patches/lp1828288/target-i386-Set-AMD-alias-bits-after-filtering-CPUID.patch
 new file mode 100644
index 0000000..bb74c77
--- /dev/null
+++ b/debian/patches/lp1828288/target-i386-Set-AMD-alias-bits-after-filtering-CPUID.patch
@@ -0,0 +1,64 @@
+Description: target-i386: Set AMD alias bits after filtering CPUID data
+
+QEMU complains about -cpu host on an AMD machine:
+  warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 0]
+For bits 0,1,3,4,5,6,7,8,9,12,13,14,15,16,17,23,24.
+
+KVM_GET_SUPPORTED_CPUID and and x86_cpu_get_migratable_flags()
+don't handle the AMD CPUID aliases bits, making
+x86_cpu_filter_features() print warnings and clear those CPUID
+bits incorrectly.
+
+To avoid hacking x86_cpu_get_migratable_flags() to handle
+CPUID_EXT2_AMD_ALIASES (just like the existing hack inside
+kvm_arch_get_supported_cpuid()), simply move the
+CPUID_EXT2_AMD_ALIASES code in x86_cpu_realizefn() after the
+x86_cpu_filter_features() call.
+
+This will probably make the CPUID_EXT2_AMD_ALIASES hack in
+kvm_arch_get_supported_cpuid() unnecessary, too. The hack will be
+removed in a follow-up patch after v2.6.0.
+
+Reported-by: Radim Krčmář <[e-mail]>
+Tested-by: Radim Krčmář <[e-mail]>
+Signed-off-by: Eduardo Habkost <[e-mail]>
+---
+Author: Eduardo Habkost <[e-mail]>
+Origin: upstream, commit: 9997cf7bdac056aeed246613639675c5a9f8fdc2
+Bug: http://lists.nongnu.org/archive/html/qemu-devel/2016-04/msg02597.html
+Bug-Ubuntu: https://launchpad.net/bugs/1828288
+Reviewed-By: Rafael David Tinoco <[e-mail]>
+Last-Update: <2019-05-08>
+
+--- qemu-2.5+dfsg.orig/target-i386/cpu.c
++++ qemu-2.5+dfsg/target-i386/cpu.c
+@@ -3156,6 +3156,14 @@ static void x86_cpu_realizefn(DeviceStat
+         env->cpuid_level = 7;
+     }
+ 
++    if (x86_cpu_filter_features(cpu) && cpu->enforce_cpuid) {
++        error_setg(&local_err,
++                   kvm_enabled() ?
++                       "Host doesn't support requested features" :
++                       "TCG doesn't support requested features");
++        goto out;
++    }
++
+     /* On AMD CPUs, some CPUID[8000_0001].EDX bits must match the bits on
+      * CPUID[1].EDX.
+      */
+@@ -3166,14 +3174,6 @@ static void x86_cpu_realizefn(DeviceStat
+     }
+ 
+ 
+-    if (x86_cpu_filter_features(cpu) && cpu->enforce_cpuid) {
+-        error_setg(&local_err,
+-                   kvm_enabled() ?
+-                       "Host doesn't support requested features" :
+-                       "TCG doesn't support requested features");
+-        goto out;
+-    }
+-
+ #ifndef CONFIG_USER_ONLY
+     qemu_register_reset(x86_cpu_machine_reset_cb, cpu);
+ 
diff --git a/debian/patches/series b/debian/patches/series
index 4b12a37..0b7e586 100644
--- a/debian/patches/series
+++ b/debian/patches/series
@@ -278,8 +278,10 @@ lp1765364/0024-ppc-spapr_caps-Don-t-disable-cap_cfpc-on-POWER8-by-d.patch
 lp1765364/0025-missing-header.patch
 CVE-2018-19489.patch
 CVE-2019-6778.patch
-#lp1823458/add-VirtIONet-vhost_stopped-flag-to-prevent-multiple.patch
-#lp1823458/do-not-call-vhost_net_cleanup-on-running-net-from-ch.patch
+lp1823458/add-VirtIONet-vhost_stopped-flag-to-prevent-multiple.patch
+lp1823458/do-not-call-vhost_net_cleanup-on-running-net-from-ch.patch
+lp1823458/lp1829380.patch
 ubuntu/enable-md-clear.patch
 CVE-2018-20815.patch
 CVE-2019-9824.patch
+lp1828288/target-i386-Set-AMD-alias-bits-after-filtering-CPUID.patch