On 6/3/21 2:29 PM, Claudio Fontana wrote:
> v1 -> v2:
> * moved hyperv realizefn call before cpu expansion (Vitaly)
> * added more comments (Eduardo)
> * fixed references to commit ids (Eduardo)
>
> The combination of Commits:
> f5cc5a5c ("i386: split cpu accelerators from cpu.c"...)
> 30565f10 ("cpu: call AccelCPUClass::cpu_realizefn in"...)
>
> introduced two bugs that break cpu max and host in the refactoring,
> by running initializations in the wrong order.
>
> This small series of two patches is an attempt to correct the situation.
>
> Please provide your test results and feedback, thanks!
>
> Claudio
>
> Claudio Fontana (2):
> i386: reorder call to cpu_exec_realizefn in x86_cpu_realizefn
> i386: run accel_cpu_instance_init as instance_post_init
>
> target/i386/cpu.c | 89 +++++++++++++++++++++++++--------------
> target/i386/kvm/kvm-cpu.c | 12 +++++-
> 2 files changed, 68 insertions(+), 33 deletions(-)
>
Btw, CI/CD is all green, but as mentioned, it does not seem to catch these kind of issues.
Hi Claudio,
Not familiar with the specifics of this bug, but can it be caught by attempting to boot an image other than Windows? If so, we can consider adding a test along the lines of tests/acceptance/boot_linux_console.py.
Thanks,
- Cleber.
https://gitlab.com/hw-claudio/qemu/-/pipelines/314286751
C.