* [kvm-unit-tests PATCH] x86: Make set/clear_bit() atomic
@ 2025-02-14 17:36 Nicolas Saenz Julienne
2025-02-15 1:18 ` Sean Christopherson
2025-02-24 17:23 ` Sean Christopherson
0 siblings, 2 replies; 3+ messages in thread
From: Nicolas Saenz Julienne @ 2025-02-14 17:36 UTC (permalink / raw)
To: pbonzini; +Cc: kvm, nh-open-source, nsaenz
x86 is the only architecture that defines set/clear_bit() as non-atomic.
This makes it incompatible with arch-agnostic code that might implicitly
require atomicity. And it was observed to corrupt the 'online_cpus'
bitmap, as non BSP CPUs perform RmWs on the bitmap concurrently during
bring up. See:
ap_start64()
save_id()
set_bit(apic_id(), online_cpus)
Address this by making set/clear_bit() atomic.
Signed-off-by: Nicolas Saenz Julienne <nsaenz@amazon.com>
---
lib/x86/processor.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/lib/x86/processor.h b/lib/x86/processor.h
index da1ed662..82507787 100644
--- a/lib/x86/processor.h
+++ b/lib/x86/processor.h
@@ -843,13 +843,13 @@ static inline bool is_canonical(u64 addr)
static inline void clear_bit(int bit, u8 *addr)
{
- __asm__ __volatile__("btr %1, %0"
+ __asm__ __volatile__("lock; btr %1, %0"
: "+m" (*addr) : "Ir" (bit) : "cc", "memory");
}
static inline void set_bit(int bit, u8 *addr)
{
- __asm__ __volatile__("bts %1, %0"
+ __asm__ __volatile__("lock; bts %1, %0"
: "+m" (*addr) : "Ir" (bit) : "cc", "memory");
}
--
2.47.1
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [kvm-unit-tests PATCH] x86: Make set/clear_bit() atomic
2025-02-14 17:36 [kvm-unit-tests PATCH] x86: Make set/clear_bit() atomic Nicolas Saenz Julienne
@ 2025-02-15 1:18 ` Sean Christopherson
2025-02-24 17:23 ` Sean Christopherson
1 sibling, 0 replies; 3+ messages in thread
From: Sean Christopherson @ 2025-02-15 1:18 UTC (permalink / raw)
To: Nicolas Saenz Julienne; +Cc: pbonzini, kvm, nh-open-source
On Fri, Feb 14, 2025, Nicolas Saenz Julienne wrote:
> x86 is the only architecture that defines set/clear_bit() as non-atomic.
> This makes it incompatible with arch-agnostic code that might implicitly
> require atomicity. And it was observed to corrupt the 'online_cpus'
> bitmap, as non BSP CPUs perform RmWs on the bitmap concurrently during
> bring up. See:
>
> ap_start64()
> save_id()
> set_bit(apic_id(), online_cpus)
>
> Address this by making set/clear_bit() atomic.
OMG, this is arguaby worse than the per-CPU stack/data mess. *sigh*
I'll grab this, I'm putting together a pull request for a few things.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [kvm-unit-tests PATCH] x86: Make set/clear_bit() atomic
2025-02-14 17:36 [kvm-unit-tests PATCH] x86: Make set/clear_bit() atomic Nicolas Saenz Julienne
2025-02-15 1:18 ` Sean Christopherson
@ 2025-02-24 17:23 ` Sean Christopherson
1 sibling, 0 replies; 3+ messages in thread
From: Sean Christopherson @ 2025-02-24 17:23 UTC (permalink / raw)
To: Sean Christopherson, pbonzini, Nicolas Saenz Julienne; +Cc: kvm, nh-open-source
On Fri, 14 Feb 2025 17:36:44 +0000, Nicolas Saenz Julienne wrote:
> x86 is the only architecture that defines set/clear_bit() as non-atomic.
> This makes it incompatible with arch-agnostic code that might implicitly
> require atomicity. And it was observed to corrupt the 'online_cpus'
> bitmap, as non BSP CPUs perform RmWs on the bitmap concurrently during
> bring up. See:
>
> ap_start64()
> save_id()
> set_bit(apic_id(), online_cpus)
>
> [...]
Applied to kvm-x86 next (and now pulled by Paolo), thanks!
[1/1] x86: Make set/clear_bit() atomic
https://github.com/kvm-x86/kvm-unit-tests/commit/2f3c02862e03
--
https://github.com/kvm-x86/kvm-unit-tests/tree/next
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2025-02-24 17:25 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-02-14 17:36 [kvm-unit-tests PATCH] x86: Make set/clear_bit() atomic Nicolas Saenz Julienne
2025-02-15 1:18 ` Sean Christopherson
2025-02-24 17:23 ` Sean Christopherson
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox