[<prev] [next>] [day] [month] [year] [list]
Message-ID: <48ef184c-0160-4a50-aeb5-013e60fc7b40@dev-mail.net>
Date: Tue, 15 Jul 2025 18:03:13 -0400
From: pgnd <pgnd@...-mail.net>
To: linux-kernel@...r.kernel.org
Cc: jpoimboe@...nel.org
Subject: [REGRESSION]: fwd: regression in 6.15.5: KVM guest launch FAILSs with
missing CPU feature error (sbpb, ibpb-brtype)
(already submitted @ Fedora; cc'd to regressions@.
advised to post to 'upstream' as well.
not stable@, but here, i'm just told ... my bad)
-------- Forwarded Message --------
Subject: regression in 6.15.5: KVM guest launch FAILSs with missing CPU feature error (sbpb, ibpb-brtype)
Date: Sun, 13 Jul 2025 17:37:57 -0400
From: pgnd <pgnd@...-mail.net>
Reply-To: pgnd@...-mail.net
To: kernel@...ts.fedoraproject.org
i'm seeing a regression on Fedora 42 kernel 6.15.4 -> 6.15.5 on AMD Ryzen 5 5600G host (x86_64)
kvm guests that launch ok under kernels 6.15.[3,4] fail with the following error when attempting to autostart under 6.15.5:
internal error: Failed to autostart VM: operation failed: guest CPU doesn't match specification: missing features: sbpb,ibpb-brtype
no changes made to libvirt, qemu, or VM defs between kernel versions.
re-booting to old kernel versions restores expected behavior.
maybe related (?) to recent changes in AMD CPU feature exposure / mitigation handling in kernel 6.15.5?
i've opened a bug:
https://bugzilla.redhat.com/show_bug.cgi?id=2379784
Powered by blists - more mailing lists