tcg: Do not elide memory barriers for !CF_PARALLEL in system mode #236
Conversation
The virtio devices require proper memory ordering between the vcpus and the iothreads. Reviewed-by: Philippe Mathieu-Daudé <[email protected]> Signed-off-by: Richard Henderson <[email protected]> (cherry picked from commit c914d46d0a645e7c633292146f4e38c945d4f847)
|
Is this the cause of the timeouts after virtio_random being initialized by cheribsd? |
|
FYI I'm planning to update the 6.0 merge PR soon, not sure if that includes this fix. |
Can't say, but that did make me think of this as a possibility. On my weaker memory model arm64 MacBook it manifested years ago as vtnet queue inconsistencies that caused assertion failures. Don't know how it shows up on x86. Perhaps it doesn't due to sequential consistency. Regardless, it's definitely wrong as it is today.
v8.1.0-rc0 is the first tag it shows up in for GitLab, https://patchwork.kernel.org/project/qemu-devel/cover/[email protected]/ is the finalised patch series |
The virtio devices require proper memory ordering between the vcpus and the iothreads.
Reviewed-by: Philippe Mathieu-Daudé [email protected]
Signed-off-by: Richard Henderson [email protected]
(cherry picked from commit c914d46d0a645e7c633292146f4e38c945d4f847)