On 2/17/25 2:15 AM, Jin Guojie wrote:
When running LTP's cpuset_memory_pressure program, an error can be
reproduced by the following steps:
(1) Create a cgroup, enable cpuset subsystem, set memory limit, and
then set cpuset_memory_pressure to 1
(2) In this cgroup, create a process to allocate a large amount of
memory and generate pressure counts
(3) Set cpuset_memory_pressure to 0
(4) Check cpuset.memory_pressure: LTP thinks it should be 0, but the
kernel returns a value of 1, so LTP determines it as FAIL
In the current kernel, the variable cpuset_memory_pressure_enabled is
not actually used。
That statement is not true. cpuset_memory_pressure_enabled is used to
determine if __cpuset_memory_pressure_bump() should be called in
cpuset_memory_pressure_bump().
This patch modifies fmeter_getrate() to determine whether to return 0
based on cpuset_memory_pressure_enabled.
Signed-off-by: Jin Guojie <guojie.jin@xxxxxxxxx>
Suggested-by: Michal Koutný <mkoutny@xxxxxxxx>
What Michal suggested is the approach you used in v2, but it has problem
as I mentioned previously.
Other than that,
Acked-by: Waiman Long <longman@xxxxxxxxxx>
Suggested-by: Waiman Long <longman@xxxxxxxxxx>
---
kernel/cgroup/cpuset-v1.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/cgroup/cpuset-v1.c b/kernel/cgroup/cpuset-v1.c
index 25c1d7b77e2f..14564e91e2b3 100644
--- a/kernel/cgroup/cpuset-v1.c
+++ b/kernel/cgroup/cpuset-v1.c
@@ -108,7 +108,7 @@ static int fmeter_getrate(struct fmeter *fmp)
fmeter_update(fmp);
val = fmp->val;
spin_unlock(&fmp->lock);
- return val;
+ return cpuset_memory_pressure_enabled ? val : 0;
}
/*
