Fix a possible memory leak in update_exclusive_cpumask() by moving the
alloc_cpumasks() down after the validate_change() check which can fail
and still before the temporary cpumasks are needed.
Fixes: e2ffe502ba45 ("cgroup/cpuset: Add cpuset.cpus.exclusive for v2")
Reported-by: Mirsad Todorovac <[email protected]>
Closes: https://lore.kernel.org/lkml/[email protected]
Signed-off-by: Waiman Long <[email protected]>
---
kernel/cgroup/cpuset.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index ba36c073304a..7260f095802a 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -2598,9 +2598,6 @@ static int update_exclusive_cpumask(struct cpuset *cs, struct cpuset *trialcs,
if (cpumask_equal(cs->exclusive_cpus, trialcs->exclusive_cpus))
return 0;
- if (alloc_cpumasks(NULL, &tmp))
- return -ENOMEM;
-
if (*buf)
compute_effective_exclusive_cpumask(trialcs, NULL);
@@ -2615,6 +2612,9 @@ static int update_exclusive_cpumask(struct cpuset *cs, struct cpuset *trialcs,
if (retval)
return retval;
+ if (alloc_cpumasks(NULL, &tmp))
+ return -ENOMEM;
+
if (old_prs) {
if (cpumask_empty(trialcs->effective_xcpus)) {
invalidate = true;
--
2.39.3
Hi,
On 2/28/2024 1:58 AM, Waiman Long wrote:
> Fix a possible memory leak in update_exclusive_cpumask() by moving the
> alloc_cpumasks() down after the validate_change() check which can fail
> and still before the temporary cpumasks are needed.
>
> Fixes: e2ffe502ba45 ("cgroup/cpuset: Add cpuset.cpus.exclusive for v2")
> Reported-by: Mirsad Todorovac <[email protected]>
> Closes: https://lore.kernel.org/lkml/[email protected]
> Signed-off-by: Waiman Long <[email protected]>
> ---
> kernel/cgroup/cpuset.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
> index ba36c073304a..7260f095802a 100644
> --- a/kernel/cgroup/cpuset.c
> +++ b/kernel/cgroup/cpuset.c
> @@ -2598,9 +2598,6 @@ static int update_exclusive_cpumask(struct cpuset *cs, struct cpuset *trialcs,
> if (cpumask_equal(cs->exclusive_cpus, trialcs->exclusive_cpus))
> return 0;
>
> - if (alloc_cpumasks(NULL, &tmp))
> - return -ENOMEM;
> -
> if (*buf)
> compute_effective_exclusive_cpumask(trialcs, NULL);
>
> @@ -2615,6 +2612,9 @@ static int update_exclusive_cpumask(struct cpuset *cs, struct cpuset *trialcs,
> if (retval)
> return retval;
>
> + if (alloc_cpumasks(NULL, &tmp))
> + return -ENOMEM;
> +
> if (old_prs) {
> if (cpumask_empty(trialcs->effective_xcpus)) {
> invalidate = true;Great work at such a short notice! (I failed to find the culprit myself.)
The leak is no longer present.
Please feel free to add:
Tested-by: Mirsad Todorovac <[email protected]>
Best regards, Mirsad Todorovac
P.S. APologies for the previous email accidentally sent as HTML.
On 2/28/24 08:56, Mirsad Todorovac wrote:
> Hi,
>
> On 2/28/2024 1:58 AM, Waiman Long wrote:
>> Fix a possible memory leak in update_exclusive_cpumask() by moving the
>> alloc_cpumasks() down after the validate_change() check which can fail
>> and still before the temporary cpumasks are needed.
>>
>> Fixes: e2ffe502ba45 ("cgroup/cpuset: Add cpuset.cpus.exclusive for v2")
>> Reported-by: Mirsad Todorovac <[email protected]>
>> Closes:
>> https://lore.kernel.org/lkml/[email protected]
>> Signed-off-by: Waiman Long <[email protected]>
>> ---
>> kernel/cgroup/cpuset.c | 6 +++---
>> 1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
>> index ba36c073304a..7260f095802a 100644
>> --- a/kernel/cgroup/cpuset.c
>> +++ b/kernel/cgroup/cpuset.c
>> @@ -2598,9 +2598,6 @@ static int update_exclusive_cpumask(struct
>> cpuset *cs, struct cpuset *trialcs,
>> if (cpumask_equal(cs->exclusive_cpus, trialcs->exclusive_cpus))
>> return 0;
>> - if (alloc_cpumasks(NULL, &tmp))
>> - return -ENOMEM;
>> -
>> if (*buf)
>> compute_effective_exclusive_cpumask(trialcs, NULL);
>> @@ -2615,6 +2612,9 @@ static int update_exclusive_cpumask(struct
>> cpuset *cs, struct cpuset *trialcs,
>> if (retval)
>> return retval;
>> + if (alloc_cpumasks(NULL, &tmp))
>> + return -ENOMEM;
>> +
>> if (old_prs) {
>> if (cpumask_empty(trialcs->effective_xcpus)) {
>> invalidate = true;Great work at such a short notice! (I
>> failed to find the culprit myself.)
>
> The leak is no longer present.
>
> Please feel free to add:
>
> Tested-by: Mirsad Todorovac <[email protected]>
>
> Best regards, Mirsad Todorovac
>
> P.S. APologies for the previous email accidentally sent as HTML.
Thanks for verifying the fix.
Cheers,
Longman
Hello,
On Tue, Feb 27, 2024 at 07:58:01PM -0500, Waiman Long wrote:
> Fix a possible memory leak in update_exclusive_cpumask() by moving the
> alloc_cpumasks() down after the validate_change() check which can fail
> and still before the temporary cpumasks are needed.
>
> Fixes: e2ffe502ba45 ("cgroup/cpuset: Add cpuset.cpus.exclusive for v2")
> Reported-by: Mirsad Todorovac <[email protected]>
> Closes: https://lore.kernel.org/lkml/[email protected]
> Signed-off-by: Waiman Long <[email protected]>
Applied to cgroup/for-6.8-fixes with stable cc'd. Depending on when v6.9
gets cut, this might go in together with for-6.9 during rc1.
Thanks.
--
tejun