On 26/10/26 15:10, QingboZhang wrote:
Hi Philippe,
Thanks for the review! Brief replies inline.
On 26/10/2026 11:57, Philippe Mathieu-Daudé wrote:
Fix this by patching on the current vCPU via async_safe_run_on_cpu() [...]
Maybe worth citing similar TCG fix in commit cb58a6d3611b.
Cited in v2.
+static void do_vapic_nop_patch(CPUState *cs, run_on_cpu_data data)
Rename as do_patch_trigger_nop? And mention this runs in exclusive section?
Done in v2 (renamed, and noted it runs in an exclusive section).
+ target_ulong addr = (target_ulong)data.target_ptr;
Pre-existing, patch_byte() should take a vaddr, then this cast is not needed.
Agreed. I'll address it in a separate follow-up cleanup patch.
For non-x86 experts: /* 2-byte NOP */
Added in v2.
+ async_safe_run_on_cpu(current_cpu, do_vapic_nop_patch,
+ RUN_ON_CPU_TARGET_PTR(ip));
Why change on current_cpu, can't we simply use CPU(cpu)? Otherwise:
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Indeed, switched to CPU(cpu) in v2. I'll carry your Reviewed-by then;
please tell me if you would rather I drop the tag. I'll post v2 shortly.
With the comment addressed, keep in in v2 :) Thanks for the quick update!
