Re: [PATCH v3 03/13] fs/resctrl: Avoid "may be used uninitialized" warning

[Reinette Chatre <reinette.chatre@intel.com>]

Hi Boris,

Thank you very much for taking a look.

On 4/7/26 11:55 AM, Borislav Petkov wrote:
> On Tue, Apr 07, 2026 at 09:02:00AM -0700, Reinette Chatre wrote:
>> Building resctrl with extra checks ("W=12") produces the following warning:
>>  .../include/linux/ucopysize.h:22:17: warning: ‘buf’ may be used uninitialized [-Wmaybe-uninitialized]
>>    22 |                 __check_object_size(ptr, n, to_user);
>>       |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>  .../include/linux/ucopysize.h: In function ‘pseudo_lock_measure_trigger’:
>>  .../include/linux/ucopysize.h:10:13: note: by argument 1 of type ‘const void *’ to ‘__check_object_size’ declared here
>>    10 | extern void __check_object_size(const void *ptr, unsigned long n,
>>       |             ^~~~~~~~~~~~~~~~~~~
>>  .../fs/resctrl/pseudo_lock.c:754:14: note: ‘buf’ declared here
>>   754 |         char buf[32];
>>       |              ^~~
>>
>> __check_object_size() ensures the provided buffer is within a valid location
>> but does not read from the uninitialized buffer. Even so, initialize the
>> buffer to silence the warning to help resctrl have a cleaner build.
>>
>> Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
>> ---
>>  fs/resctrl/pseudo_lock.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/resctrl/pseudo_lock.c b/fs/resctrl/pseudo_lock.c
>> index fa3687d69ebd..e1e9134474f4 100644
>> --- a/fs/resctrl/pseudo_lock.c
>> +++ b/fs/resctrl/pseudo_lock.c
>> @@ -750,8 +750,8 @@ static ssize_t pseudo_lock_measure_trigger(struct file *file,
>>  					   size_t count, loff_t *ppos)
>>  {
>>  	struct rdtgroup *rdtgrp = file->private_data;
>> +	char buf[32] = {};
> 
> AFAIU, you're not leaking any uninitialized stack data from that buffer,
> right?

Right. From what I can tell __check_object_size() just checks that the address
is from a valid region and does not read from the buffer.

> 
> If so, why do you care about some silly build warning and are willing to waste
> a 32-byte memset on every function entry?

I care because I am including a W=12 build as part of checking all resctrl patches
and having this be the one and only warning that always shows up is distracting.
Removing it to accomplish a clean W=12 build does not seem impactful to me.

Of course you are right that this wastes a memset. This function is not on a
hot path though and I do not believe it would be called frequently since it is
intended to demonstrate how successful the setup of a pseudo-lock region was and
successive calls should produce consistent results.

> 
> There's a reason those warnings are behind W= ...

Indeed. This patch is for developer and maintenance convenience and not required for
resctrl health. No problem from me if you find it not to be appropriate. I will just
continue to parse the warning out of the logs.

Reinette