When kzalloc for pmus[i].boxes fails, we should clean up pmus
to prevent memleak. However, when kzalloc for attr_group fails,
pmus has been assigned to type->pmus, and freeing will be done
later on by the callers. The chain is: uncore_type_init ->
uncore_types_init -> uncore_pci_init -> uncore_types_exit ->
uncore_type_exit. Therefore, freeing pmus in uncore_type_init
may cause a double-free. Fix this by setting type->pmus to
NULL after kfree.

Fixes: 629eb703d3e4 ("perf/x86/intel/uncore: Fix memory leaks on allocation failures")
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
---
 arch/x86/events/intel/uncore.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/events/intel/uncore.c b/arch/x86/events/intel/uncore.c
index 01023aa5125b..d80445a24011 100644
--- a/arch/x86/events/intel/uncore.c
+++ b/arch/x86/events/intel/uncore.c
@@ -1041,6 +1041,7 @@ static int __init uncore_type_init(struct intel_uncore_type *type, bool setid)
 	for (i = 0; i < type->num_boxes; i++)
 		kfree(pmus[i].boxes);
 	kfree(pmus);
+	type->pmus = NULL;
 
 	return -ENOMEM;
 }
-- 
2.17.1

On 5/12/23 05:27, Dinghao Liu wrote:
> When kzalloc for pmus[i].boxes fails, we should clean up pmus
> to prevent memleak. However, when kzalloc for attr_group fails,
> pmus has been assigned to type->pmus, and freeing will be done
> later on by the callers. The chain is: uncore_type_init ->
> uncore_types_init -> uncore_pci_init -> uncore_types_exit ->
> uncore_type_exit. Therefore, freeing pmus in uncore_type_init
> may cause a double-free. Fix this by setting type->pmus to
> NULL after kfree.
> 
> Fixes: 629eb703d3e4 ("perf/x86/intel/uncore: Fix memory leaks on allocation failures")
> Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>

Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>

> ---
>  arch/x86/events/intel/uncore.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/x86/events/intel/uncore.c b/arch/x86/events/intel/uncore.c
> index 01023aa5125b..d80445a24011 100644
> --- a/arch/x86/events/intel/uncore.c
> +++ b/arch/x86/events/intel/uncore.c
> @@ -1041,6 +1041,7 @@ static int __init uncore_type_init(struct intel_uncore_type *type, bool setid)
>  	for (i = 0; i < type->num_boxes; i++)
>  		kfree(pmus[i].boxes);
>  	kfree(pmus);
> +	type->pmus = NULL;
>  
>  	return -ENOMEM;
>  }

On 2023.12.05 11:27:09 +0800, Dinghao Liu wrote:
> When kzalloc for pmus[i].boxes fails, we should clean up pmus
> to prevent memleak. However, when kzalloc for attr_group fails,
> pmus has been assigned to type->pmus, and freeing will be done
> later on by the callers. The chain is: uncore_type_init ->
> uncore_types_init -> uncore_pci_init -> uncore_types_exit ->
> uncore_type_exit. Therefore, freeing pmus in uncore_type_init
> may cause a double-free. Fix this by setting type->pmus to
> NULL after kfree.

Change is ok but the call trace you wrote here is reversed or malformed??
With that fixed or cleared.

Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>

> 
> Fixes: 629eb703d3e4 ("perf/x86/intel/uncore: Fix memory leaks on allocation failures")
> Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
> ---
>  arch/x86/events/intel/uncore.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/x86/events/intel/uncore.c b/arch/x86/events/intel/uncore.c
> index 01023aa5125b..d80445a24011 100644
> --- a/arch/x86/events/intel/uncore.c
> +++ b/arch/x86/events/intel/uncore.c
> @@ -1041,6 +1041,7 @@ static int __init uncore_type_init(struct intel_uncore_type *type, bool setid)
>  	for (i = 0; i < type->num_boxes; i++)
>  		kfree(pmus[i].boxes);
>  	kfree(pmus);
> +	type->pmus = NULL;
>  
>  	return -ENOMEM;
>  }
> -- 
> 2.17.1
> 

> On 2023.12.05 11:27:09 +0800, Dinghao Liu wrote:
> > When kzalloc for pmus[i].boxes fails, we should clean up pmus
> > to prevent memleak. However, when kzalloc for attr_group fails,
> > pmus has been assigned to type->pmus, and freeing will be done
> > later on by the callers. The chain is: uncore_type_init ->
> > uncore_types_init -> uncore_pci_init -> uncore_types_exit ->
> > uncore_type_exit. Therefore, freeing pmus in uncore_type_init
> > may cause a double-free. Fix this by setting type->pmus to
> > NULL after kfree.
> 
> Change is ok but the call trace you wrote here is reversed or malformed??
> With that fixed or cleared.
> 

Thanks for your advice. I will fix it and resend a new patch soon.

Regards,
Dinghao