[PATCH OLK-6.6 V1] sched/fair: limit burst to zero when cfs bandwidth is toggled off
From: Zhao Wenhui <zhaowenhui8@huawei.com> hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I9PR8C CVE: NA -------------------------------- sched/fair: limit burst to zero when cfs bandwidth is toggled off When the quota value in CFS bandwidth is set to -1, that imples the cfs bandwidth is toggled off. So the burst feature is supposed to be disable as well. Currently, when quota is -1, burst will not be check, so that it can be set to almost arbitery value. Examples: mkdir /sys/fs/cgroup/cpu/test echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us echo 10000000000000000 > /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us Moreover, after the burst modified by this way, quota can't be set to any value: echo 100000 > cpu.cfs_quota_us -bash: echo: write error: Invalid argument This patch can ensure the burst value being zero and unalterable, when quota is set to -1. Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS controller") Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com> Signed-off-by: Cheng Yu <serein.chengyu@huawei.com> --- kernel/sched/core.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 9e7a99947de6..5441afe18953 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -11051,6 +11051,12 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota, burst + quota > max_cfs_runtime)) return -EINVAL; + /* + * Ensure burst equals to zero when quota is -1. + */ + if (quota == RUNTIME_INF && burst) + return -EINVAL; + /* * Prevent race between setting of cfs_rq->runtime_enabled and * unthrottle_offline_cfs_rqs(). @@ -11110,8 +11116,10 @@ static int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us) period = ktime_to_ns(tg->cfs_bandwidth.period); burst = tg->cfs_bandwidth.burst; - if (cfs_quota_us < 0) + if (cfs_quota_us < 0) { quota = RUNTIME_INF; + burst = 0; + } else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC) quota = (u64)cfs_quota_us * NSEC_PER_USEC; else -- 2.25.1
反馈： 您发送到kernel@openeuler.org的补丁/补丁集，已成功转换为PR！ PR链接地址： https://gitee.com/openeuler/kernel/pulls/7418 邮件列表地址：https://mailweb.openeuler.org/hyperkitty/list/kernel@openeuler.org/message/5... FeedBack: The patch(es) which you have sent to kernel@openeuler.org mailing list has been converted to a pull request successfully! Pull request link: https://gitee.com/openeuler/kernel/pulls/7418 Mailing list address: https://mailweb.openeuler.org/hyperkitty/list/kernel@openeuler.org/message/5...
On 2024/5/16 14:27, Cheng Yu wrote:
From: Zhao Wenhui <zhaowenhui8@huawei.com>
hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I9PR8C CVE: NA
--------------------------------
sched/fair: limit burst to zero when cfs bandwidth is toggled off
When the quota value in CFS bandwidth is set to -1, that imples the cfs bandwidth is toggled off. So the burst feature is supposed to be disable as well.
Currently, when quota is -1, burst will not be check, so that it can be set to almost arbitery value. Examples: mkdir /sys/fs/cgroup/cpu/test echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us echo 10000000000000000 > /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us
Moreover, after the burst modified by this way, quota can't be set to any value: echo 100000 > cpu.cfs_quota_us -bash: echo: write error: Invalid argument
This patch can ensure the burst value being zero and unalterable, when quota is set to -1.
Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS controller") Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com> Signed-off-by: Cheng Yu <serein.chengyu@huawei.com> --- kernel/sched/core.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 9e7a99947de6..5441afe18953 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -11051,6 +11051,12 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota, burst + quota > max_cfs_runtime)) return -EINVAL;
+ /* + * Ensure burst equals to zero when quota is -1. + */ + if (quota == RUNTIME_INF && burst) + return -EINVAL; +
How about put this forward， before the first quota check? + /* + * Ensure burst equals to zero when quota is -1. + */ + if (quota == RUNTIME_INF && burst) + return -EINVAL; + /* * Ensure we have at some amount of bandwidth every period. This is * to prevent reaching a state of large arrears when throttled via * entity_tick() resulting in prolonged exit starvation. */ if (quota < min_cfs_quota_period || period < min_cfs_quota_period) return -EINVAL; Besides，burst already been set to zero when cfs_quota_us < 0, the check of non-zone burst is meaningless.
/* * Prevent race between setting of cfs_rq->runtime_enabled and * unthrottle_offline_cfs_rqs(). @@ -11110,8 +11116,10 @@ static int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
period = ktime_to_ns(tg->cfs_bandwidth.period); burst = tg->cfs_bandwidth.burst; - if (cfs_quota_us < 0) + if (cfs_quota_us < 0) { quota = RUNTIME_INF; + burst = 0; + } else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC) quota = (u64)cfs_quota_us * NSEC_PER_USEC; else
*}* and *else* on same line.
participants (3)
-
Cheng Yu -
patchwork bot -
Zhao Mengmeng