On 21/01/2026 2:28 pm, Alejandro Vallejo wrote:
> Configure the Bus Lock intercept when supported by the host.
"which is available on Zen4 and later".
(I think ?)
> diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
> index 5d23603fc1..abda5a9063 100644
> --- a/xen/arch/x86/hvm/svm/svm.c
> +++ b/xen/arch/x86/hvm/svm/svm.c
> @@ -2524,6 +2524,7 @@ const struct hvm_function_table * __init start_svm(void)
> P(cpu_has_tsc_ratio, "TSC Rate MSR");
> P(cpu_has_svm_sss, "NPT Supervisor Shadow Stack");
> P(cpu_has_svm_spec_ctrl, "MSR_SPEC_CTRL virtualisation");
> + P(cpu_has_svm_bus_lock, "BusLock-Intercept Filter");
"Bus Lock Filter". The Intercept part isn't terribly useful.
> #undef P
>
> if ( !printed )
> @@ -3087,6 +3088,11 @@ void asmlinkage svm_vmexit_handler(void)
> break;
> }
>
> + case VMEXIT_BUS_LOCK:
> + perfc_incr(buslock);
> + vmcb->bus_lock_count = 1;
> + break;
This needs an explanation of the behaviour.
/* This is a fault and blocked the Bus Lock inducing action. We're only
interested in rate limiting the guest, so credit it one "lock" in order
to re-execute the instruction. */
> +
> default:
> unexpected_exit_type:
> gprintk(XENLOG_ERR, "Unexpected vmexit: reason %#"PRIx64", "
> diff --git a/xen/arch/x86/hvm/svm/vmcb.c b/xen/arch/x86/hvm/svm/vmcb.c
> index cbee10d046..15223a693e 100644
> --- a/xen/arch/x86/hvm/svm/vmcb.c
> +++ b/xen/arch/x86/hvm/svm/vmcb.c
> @@ -66,6 +66,9 @@ static int construct_vmcb(struct vcpu *v)
> GENERAL2_INTERCEPT_XSETBV | GENERAL2_INTERCEPT_ICEBP |
> GENERAL2_INTERCEPT_RDPRU;
>
> + if ( cpu_has_svm_bus_lock )
> + vmcb->_general3_intercepts |= GENERAL3_INTERCEPT_BUS_LOCK;
> +
This wants a justification for why it's unconditional. Something like:
/* Well behaved logic shouldn't ever Bus Lock, but we care about rate
limiting buggy/malicious cases. */
> /* Intercept all debug-register writes. */
> vmcb->_dr_intercepts = ~0u;
>
> diff --git a/xen/arch/x86/hvm/svm/vmcb.h b/xen/arch/x86/hvm/svm/vmcb.h
> index 231f9b1b06..68cf5eaf0b 100644
> --- a/xen/arch/x86/hvm/svm/vmcb.h
> +++ b/xen/arch/x86/hvm/svm/vmcb.h
> @@ -68,7 +68,7 @@ enum GenericIntercept2bits
> /* general 3 intercepts */
> enum GenericIntercept3bits
> {
> - GENERAL3_INTERCEPT_BUS_LOCK_THRESH = 1 << 5,
> + GENERAL3_INTERCEPT_BUS_LOCK = 1 << 5,
> };
>
> /* control register intercepts */
> @@ -497,7 +497,7 @@ struct vmcb_struct {
> u8 guest_ins_len; /* offset 0xD0 */
> u8 guest_ins[15]; /* offset 0xD1 */
> u64 res10a[8]; /* offset 0xE0 */
> - u16 bus_lock_thresh; /* offset 0x120 */
> + u16 bus_lock_count; /* offset 0x120 */
> u16 res10b[3]; /* offset 0x122 */
> u64 res10c[91]; /* offset 0x128 pad to save area */
>
Both of these hunks want moving into the prior patch, which resolves one
of my concerns there.
All can be fixed on commit.
~Andrew
