Re: [PATCH 1/1] cxl/edac: Fix potential memory leak issues

[Li Ming <ming.li@zohomail.com>]

On 6/11/2025 9:16 PM, Jonathan Cameron wrote:
> On Wed, 11 Jun 2025 11:35:42 +0800
> Li Ming <ming.li@zohomail.com> wrote:
>
>> In cxl_store_rec_gen_media() and cxl_store_rec_dram(), use kmemdup() to
>> duplicate a cxl gen_media/dram event to store the event in a xarray by
>> xa_store(). The cxl gen_media/dram event allocated by kmemdup() should
>> be freed in the case that the xa_store() fails.
>>
>> Fixes: 0b5ccb0de1e2 ("cxl/edac: Support for finding memory operation attributes from the current boot")
>> Signed-off-by: Li Ming <ming.li@zohomail.com>
> Good fine but I'm not sure this is the best fix.
>
>> ---
>> base-commit: 87b42c114cdda76c8ad3002f2096699ad5146cb3 cxl/fixes
>> ---
>>  drivers/cxl/core/edac.c | 10 ++++++----
>>  1 file changed, 6 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/cxl/core/edac.c b/drivers/cxl/core/edac.c
>> index 2cbc664e5d62..b4c5c23a45d4 100644
>> --- a/drivers/cxl/core/edac.c
>> +++ b/drivers/cxl/core/edac.c
>> @@ -1086,13 +1086,13 @@ static void cxl_del_overflow_old_recs(struct xarray *rec_xarray)
>>  int cxl_store_rec_gen_media(struct cxl_memdev *cxlmd, union cxl_event *evt)
>>  {
>>  	struct cxl_mem_err_rec *array_rec = cxlmd->err_rec_array;
>> -	struct cxl_event_gen_media *rec;
>>  	void *old_rec;
>>  
>>  	if (!IS_ENABLED(CONFIG_CXL_EDAC_MEM_REPAIR) || !array_rec)
>>  		return 0;
>>  
>> -	rec = kmemdup(&evt->gen_media, sizeof(*rec), GFP_KERNEL);
>> +	struct cxl_event_gen_media *rec __free(kfree) =
>> +		kmemdup(&evt->gen_media, sizeof(*rec), GFP_KERNEL);
>>  	if (!rec)
>>  		return -ENOMEM;
>>  
>> @@ -1106,6 +1106,7 @@ int cxl_store_rec_gen_media(struct cxl_memdev *cxlmd, union cxl_event *evt)
>>  
>>  	cxl_del_expired_gmedia_recs(&array_rec->rec_gen_media, rec);
>>  	cxl_del_overflow_old_recs(&array_rec->rec_gen_media);
>> +	retain_and_null_ptr(rec);
>>  
>>  	return 0;
>>  }
>> @@ -1114,13 +1115,13 @@ EXPORT_SYMBOL_NS_GPL(cxl_store_rec_gen_media, "CXL");
>>  int cxl_store_rec_dram(struct cxl_memdev *cxlmd, union cxl_event *evt)
>>  {
>>  	struct cxl_mem_err_rec *array_rec = cxlmd->err_rec_array;
>> -	struct cxl_event_dram *rec;
>>  	void *old_rec;
>>  
>>  	if (!IS_ENABLED(CONFIG_CXL_EDAC_MEM_REPAIR) || !array_rec)
>>  		return 0;
>>  
>> -	rec = kmemdup(&evt->dram, sizeof(*rec), GFP_KERNEL);
>> +	struct cxl_event_dram *rec __free(kfree) =
>> +		kmemdup(&evt->dram, sizeof(*rec), GFP_KERNEL);
>>  	if (!rec)
>>  		return -ENOMEM;
>>  
>> @@ -1134,6 +1135,7 @@ int cxl_store_rec_dram(struct cxl_memdev *cxlmd, union cxl_event *evt)
>>  
>>  	cxl_del_expired_dram_recs(&array_rec->rec_dram, rec);
>>  	cxl_del_overflow_old_recs(&array_rec->rec_dram);
>> +	retain_and_null_ptr(rec);
> I'd move this up to immediately after we hand it over to the xa successfully.
>
> Actually I'm not sure this is a good use of __free given we have a single
> error path to handle it in and have to manually release it in a different
> call to where it is passed to the xa_store() function.  Maybe better
> to just kfree in the error path.

Sure, I can do that, using kfree directly instead of the __free(kfree). Thanks for reviewing.


Ming

>
>
>>  
>>  	return 0;
>>  }