On Fri, Nov 05, 2021 at 10:35:20 +0100, Michal Privoznik wrote:
Alright, here's the deal: to enable tb-cache one has to use
'-accel tcg,tb-size=' which then conflicts with '-machine
accel=tcg'. But sure, we can use the old -accel in this specific
case.
The above is all wrong after the refactors.
But because of how the tb-size argument is defined in QEMU
there's no way for us to have a capability check. The feature was
introduced in QEMU commit of v5.0.0-rc0~175^2~62 and is tied to
TCG only. Therefore, I think we can live without capability
check. Worst case scenario, QEMU fails to start.
Resolves:
https://gitlab.com/libvirt/libvirt/-/issues/229
Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com>
Tested-by: Kashyap Chamarthy <kchamart(a)redhat.com>
---
src/qemu/qemu_command.c | 5 +++
...efault-cpu-tcg-features.x86_64-latest.args | 40 +++++++++++++++++++
tests/qemuxml2argvtest.c | 1 +
3 files changed, 46 insertions(+)
create mode 100644
tests/qemuxml2argvdata/x86_64-default-cpu-tcg-features.x86_64-latest.args
diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
index 0257b86a0d..17ee3c5d3a 100644
--- a/src/qemu/qemu_command.c
+++ b/src/qemu/qemu_command.c
@@ -7271,6 +7271,11 @@ qemuBuildAccelCommandLine(virCommand *cmd,
switch ((virDomainVirtType)def->virtType) {
case VIR_DOMAIN_VIRT_QEMU:
virBufferAddLit(&buf, "tcg");
+
+ if (def->features[VIR_DOMAIN_FEATURE_TCG] == VIR_TRISTATE_SWITCH_ON) {
Similarly to previous commit. It's not extensible when adding new tcg
features. This would set the cache to 0 if a new feature is added and
tb-cache is not used.
+ virBufferAsprintf(&buf, ",tb-size=%llu",
+ def->tcg_features->tb_cache >> 10);
Okay, so input in libvirt is kiB, output is MiB. This means we are
truncating precision. In this case I don't think it's worth adding
checks, but at least the docs should mention what's going on.
+ }
break;