[PATCH v3 0/1] accel/kvm: Fix segmentation fault

[PATCH v3 0/1] accel/kvm: Fix segmentation fault

[Masato Imai]

Changes from v2:
- avoid segfault in kvm/accel instead of migration/dirtyrate

v2: https://lore.kernel.org/qemu-devel/20240423091306.754432-1-mii@sfc.wide.ad.jp

Masato Imai (1):
  accel/kvm: Fix segmentation fault

 accel/kvm/kvm-all.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

-- 
2.34.1

[PATCH v3 1/1] accel/kvm: Fix segmentation fault

[Masato Imai]

When the KVM acceleration parameter is not set, executing calc_dirty_rate
with the -r or -b option results in a segmentation fault due to accessing
a null kvm_state pointer in the kvm_dirty_ring_enabled function. This
commit adds a null check for kvm_status to prevent segmentation faults.

Signed-off-by: Masato Imai <mii@sfc.wide.ad.jp>
---
 accel/kvm/kvm-all.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index c0be9f5eed..544293be8a 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -2329,7 +2329,7 @@ bool kvm_vcpu_id_is_valid(int vcpu_id)
 
 bool kvm_dirty_ring_enabled(void)
 {
-    return kvm_state->kvm_dirty_ring_size ? true : false;
+    return kvm_state && kvm_state->kvm_dirty_ring_size;
 }
 
 static void query_stats_cb(StatsResultList **result, StatsTarget target,
-- 
2.34.1

Re: [PATCH v3 1/1] accel/kvm: Fix segmentation fault

[Zhijian Li (Fujitsu)]

on 5/7/2024 10:50 AM, Masato Imai wrote:
> When the KVM acceleration parameter is not set, executing calc_dirty_rate
> with the -r or -b option results in a segmentation fault due to accessing
> a null kvm_state pointer in the kvm_dirty_ring_enabled function. This
> commit adds a null check for kvm_status to prevent segmentation faults.
>
> Signed-off-by: Masato Imai <mii@sfc.wide.ad.jp>

LGTM,
Tested-by: Li Zhijian <lizhijian@fujitsu.com>


> ---
>   accel/kvm/kvm-all.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
> index c0be9f5eed..544293be8a 100644
> --- a/accel/kvm/kvm-all.c
> +++ b/accel/kvm/kvm-all.c
> @@ -2329,7 +2329,7 @@ bool kvm_vcpu_id_is_valid(int vcpu_id)
>   
>   bool kvm_dirty_ring_enabled(void)
>   {
> -    return kvm_state->kvm_dirty_ring_size ? true : false;
> +    return kvm_state && kvm_state->kvm_dirty_ring_size;
>   }
>   
>   static void query_stats_cb(StatsResultList **result, StatsTarget target,

Re: [PATCH v3 1/1] accel/kvm: Fix segmentation fault

[Zhijian Li (Fujitsu) via]

on 5/7/2024 10:50 AM, Masato Imai wrote:
> When the KVM acceleration parameter is not set, executing calc_dirty_rate
> with the -r or -b option results in a segmentation fault due to accessing
> a null kvm_state pointer in the kvm_dirty_ring_enabled function. This
> commit adds a null check for kvm_status to prevent segmentation faults.
>
> Signed-off-by: Masato Imai <mii@sfc.wide.ad.jp>

LGTM,
Tested-by: Li Zhijian <lizhijian@fujitsu.com>


> ---
>   accel/kvm/kvm-all.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
> index c0be9f5eed..544293be8a 100644
> --- a/accel/kvm/kvm-all.c
> +++ b/accel/kvm/kvm-all.c
> @@ -2329,7 +2329,7 @@ bool kvm_vcpu_id_is_valid(int vcpu_id)
>   
>   bool kvm_dirty_ring_enabled(void)
>   {
> -    return kvm_state->kvm_dirty_ring_size ? true : false;
> +    return kvm_state && kvm_state->kvm_dirty_ring_size;
>   }
>   
>   static void query_stats_cb(StatsResultList **result, StatsTarget target,

Re: [PATCH v3 1/1] accel/kvm: Fix segmentation fault

[Philippe Mathieu-Daudé]

Hi Masato,

On 7/5/24 04:50, Masato Imai wrote:
> When the KVM acceleration parameter is not set, executing calc_dirty_rate
> with the -r or -b option results in a segmentation fault due to accessing
> a null kvm_state pointer in the kvm_dirty_ring_enabled function. This
> commit adds a null check for kvm_status to prevent segmentation faults.
> 
> Signed-off-by: Masato Imai <mii@sfc.wide.ad.jp>
> ---
>   accel/kvm/kvm-all.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
> index c0be9f5eed..544293be8a 100644
> --- a/accel/kvm/kvm-all.c
> +++ b/accel/kvm/kvm-all.c
> @@ -2329,7 +2329,7 @@ bool kvm_vcpu_id_is_valid(int vcpu_id)
>   
>   bool kvm_dirty_ring_enabled(void)
>   {
> -    return kvm_state->kvm_dirty_ring_size ? true : false;
> +    return kvm_state && kvm_state->kvm_dirty_ring_size;

I missed the previous iterations of this patch. I disagree
with this approach, we shouldn't call kvm_dirty_ring_enabled()
if kvm_state is NULL, this is a bad API usage. So I'd rather
assert(kvm_state) here and force the callers to check for
kvm_enabled() before calling.

>   }
>   
>   static void query_stats_cb(StatsResultList **result, StatsTarget target,