Currently the user can set up isolated cpus via cpuset and nohz_full in
such a way that leaves no housekeeping CPU (i.e. no CPU that is neither
domain isolated nor nohz full). This can be a problem for other
subsystems (e.g. the timer wheel imgration).

Prevent this configuration by blocking any assignation that would cause
the union of domain isolated cpus and nohz_full to covers all CPUs.

[longman: Remove isolated_cpus_should_update() and rewrite the checking
 in update_prstate() and update_parent_effective_cpumask()]

Originally-by: Gabriele Monaco <gmonaco@redhat.com>
Signed-off-by: Waiman Long <longman@redhat.com>
---
 kernel/cgroup/cpuset.c | 74 +++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 73 insertions(+), 1 deletion(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index da770dac955e..99622e90991a 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -1393,6 +1393,45 @@ static bool partition_xcpus_del(int old_prs, struct cpuset *parent,
 	return isolcpus_updated;
 }
 
+/*
+ * isolated_cpus_can_update - check for isolated & nohz_full conflicts
+ * @add_cpus: cpu mask for cpus that are going to be isolated
+ * @del_cpus: cpu mask for cpus that are no longer isolated, can be NULL
+ * Return: false if there is conflict, true otherwise
+ *
+ * If nohz_full is enabled and we have isolated CPUs, their combination must
+ * still leave housekeeping CPUs.
+ *
+ * TBD: Should consider merging this function into
+ *      prstate_housekeeping_conflict().
+ */
+static bool isolated_cpus_can_update(struct cpumask *add_cpus,
+				     struct cpumask *del_cpus)
+{
+	cpumask_var_t full_hk_cpus;
+	int res = true;
+
+	if (!housekeeping_enabled(HK_TYPE_KERNEL_NOISE))
+		return true;
+
+	if (del_cpus && cpumask_weight_and(del_cpus,
+			housekeeping_cpumask(HK_TYPE_KERNEL_NOISE)))
+		return true;
+
+	if (!alloc_cpumask_var(&full_hk_cpus, GFP_KERNEL))
+		return false;
+
+	cpumask_and(full_hk_cpus, housekeeping_cpumask(HK_TYPE_KERNEL_NOISE),
+		    housekeeping_cpumask(HK_TYPE_DOMAIN));
+	cpumask_andnot(full_hk_cpus, full_hk_cpus, isolated_cpus);
+	cpumask_and(full_hk_cpus, full_hk_cpus, cpu_active_mask);
+	if (!cpumask_weight_andnot(full_hk_cpus, add_cpus))
+		res = false;
+
+	free_cpumask_var(full_hk_cpus);
+	return res;
+}
+
 static void update_isolation_cpumasks(bool isolcpus_updated)
 {
 	int ret;
@@ -1551,6 +1590,9 @@ static int remote_partition_enable(struct cpuset *cs, int new_prs,
 	if (!cpumask_intersects(tmp->new_cpus, cpu_active_mask) ||
 	    cpumask_subset(top_cpuset.effective_cpus, tmp->new_cpus))
 		return PERR_INVCPUS;
+	if ((new_prs == PRS_ISOLATED) &&
+	    !isolated_cpus_can_update(tmp->new_cpus, NULL))
+		return PERR_HKEEPING;
 
 	spin_lock_irq(&callback_lock);
 	isolcpus_updated = partition_xcpus_add(new_prs, NULL, tmp->new_cpus);
@@ -1650,6 +1692,9 @@ static void remote_cpus_update(struct cpuset *cs, struct cpumask *xcpus,
 		else if (cpumask_intersects(tmp->addmask, subpartitions_cpus) ||
 			 cpumask_subset(top_cpuset.effective_cpus, tmp->addmask))
 			cs->prs_err = PERR_NOCPUS;
+		else if ((prs == PRS_ISOLATED) &&
+			 !isolated_cpus_can_update(tmp->addmask, tmp->delmask))
+			cs->prs_err = PERR_HKEEPING;
 		if (cs->prs_err)
 			goto invalidate;
 	}
@@ -1750,6 +1795,7 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
 	int part_error = PERR_NONE;	/* Partition error? */
 	int isolcpus_updated = 0;
 	struct cpumask *xcpus = user_xcpus(cs);
+	int parent_prs = parent->partition_root_state;
 	bool nocpu;
 
 	lockdep_assert_held(&cpuset_mutex);
@@ -1813,6 +1859,10 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
 		if (prstate_housekeeping_conflict(new_prs, xcpus))
 			return PERR_HKEEPING;
 
+		if ((new_prs == PRS_ISOLATED) && (new_prs != parent_prs) &&
+		    !isolated_cpus_can_update(xcpus, NULL))
+			return PERR_HKEEPING;
+
 		if (tasks_nocpu_error(parent, cs, xcpus))
 			return PERR_NOCPUS;
 
@@ -1866,6 +1916,7 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
 		 *
 		 * For invalid partition:
 		 *   delmask = newmask & parent->effective_xcpus
+		 *   The partition may become valid soon.
 		 */
 		if (is_partition_invalid(cs)) {
 			adding = false;
@@ -1880,6 +1931,23 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
 			deleting = cpumask_and(tmp->delmask, tmp->delmask,
 					       parent->effective_xcpus);
 		}
+
+		/*
+		 * TBD: Invalidate a currently valid child root partition may
+		 * still break isolated_cpus_can_update() rule if parent is an
+		 * isolated partition.
+		 */
+		if (is_partition_valid(cs) && (old_prs != parent_prs)) {
+			if ((parent_prs == PRS_ROOT) &&
+			    /* Adding to parent means removing isolated CPUs */
+			    !isolated_cpus_can_update(tmp->delmask, tmp->addmask))
+				part_error = PERR_HKEEPING;
+			if ((parent_prs == PRS_ISOLATED) &&
+			    /* Adding to parent means adding isolated CPUs */
+			    !isolated_cpus_can_update(tmp->addmask, tmp->delmask))
+				part_error = PERR_HKEEPING;
+		}
+
 		/*
 		 * The new CPUs to be removed from parent's effective CPUs
 		 * must be present.
@@ -2994,7 +3062,11 @@ static int update_prstate(struct cpuset *cs, int new_prs)
 		 * A change in load balance state only, no change in cpumasks.
 		 * Need to update isolated_cpus.
 		 */
-		isolcpus_updated = true;
+		if ((new_prs == PRS_ISOLATED) &&
+		    !isolated_cpus_can_update(cs->effective_xcpus, NULL))
+			err = PERR_HKEEPING;
+		else
+			isolcpus_updated = true;
 	} else {
 		/*
 		 * Switching back to member is always allowed even if it
-- 
2.51.1

On 2025/11/5 12:38, Waiman Long wrote:
> Currently the user can set up isolated cpus via cpuset and nohz_full in
> such a way that leaves no housekeeping CPU (i.e. no CPU that is neither
> domain isolated nor nohz full). This can be a problem for other
> subsystems (e.g. the timer wheel imgration).
> 
> Prevent this configuration by blocking any assignation that would cause
> the union of domain isolated cpus and nohz_full to covers all CPUs.
> 
> [longman: Remove isolated_cpus_should_update() and rewrite the checking
>  in update_prstate() and update_parent_effective_cpumask()]
> 
> Originally-by: Gabriele Monaco <gmonaco@redhat.com>
> Signed-off-by: Waiman Long <longman@redhat.com>
> ---
>  kernel/cgroup/cpuset.c | 74 +++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 73 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
> index da770dac955e..99622e90991a 100644
> --- a/kernel/cgroup/cpuset.c
> +++ b/kernel/cgroup/cpuset.c
> @@ -1393,6 +1393,45 @@ static bool partition_xcpus_del(int old_prs, struct cpuset *parent,
>  	return isolcpus_updated;
>  }
>  
> +/*
> + * isolated_cpus_can_update - check for isolated & nohz_full conflicts
> + * @add_cpus: cpu mask for cpus that are going to be isolated
> + * @del_cpus: cpu mask for cpus that are no longer isolated, can be NULL
> + * Return: false if there is conflict, true otherwise
> + *
> + * If nohz_full is enabled and we have isolated CPUs, their combination must
> + * still leave housekeeping CPUs.
> + *
> + * TBD: Should consider merging this function into
> + *      prstate_housekeeping_conflict().
> + */
> +static bool isolated_cpus_can_update(struct cpumask *add_cpus,
> +				     struct cpumask *del_cpus)
> +{
> +	cpumask_var_t full_hk_cpus;
> +	int res = true;
> +
> +	if (!housekeeping_enabled(HK_TYPE_KERNEL_NOISE))
> +		return true;
> +
> +	if (del_cpus && cpumask_weight_and(del_cpus,
> +			housekeeping_cpumask(HK_TYPE_KERNEL_NOISE)))
> +		return true;
> +
> +	if (!alloc_cpumask_var(&full_hk_cpus, GFP_KERNEL))
> +		return false;
> +
> +	cpumask_and(full_hk_cpus, housekeeping_cpumask(HK_TYPE_KERNEL_NOISE),
> +		    housekeeping_cpumask(HK_TYPE_DOMAIN));
> +	cpumask_andnot(full_hk_cpus, full_hk_cpus, isolated_cpus);
> +	cpumask_and(full_hk_cpus, full_hk_cpus, cpu_active_mask);
> +	if (!cpumask_weight_andnot(full_hk_cpus, add_cpus))
> +		res = false;
> +
> +	free_cpumask_var(full_hk_cpus);
> +	return res;
> +}
> +
>  static void update_isolation_cpumasks(bool isolcpus_updated)
>  {
>  	int ret;
> @@ -1551,6 +1590,9 @@ static int remote_partition_enable(struct cpuset *cs, int new_prs,
>  	if (!cpumask_intersects(tmp->new_cpus, cpu_active_mask) ||
>  	    cpumask_subset(top_cpuset.effective_cpus, tmp->new_cpus))
>  		return PERR_INVCPUS;
> +	if ((new_prs == PRS_ISOLATED) &&
> +	    !isolated_cpus_can_update(tmp->new_cpus, NULL))
> +		return PERR_HKEEPING;
>  
>  	spin_lock_irq(&callback_lock);
>  	isolcpus_updated = partition_xcpus_add(new_prs, NULL, tmp->new_cpus);
> @@ -1650,6 +1692,9 @@ static void remote_cpus_update(struct cpuset *cs, struct cpumask *xcpus,
>  		else if (cpumask_intersects(tmp->addmask, subpartitions_cpus) ||
>  			 cpumask_subset(top_cpuset.effective_cpus, tmp->addmask))
>  			cs->prs_err = PERR_NOCPUS;
> +		else if ((prs == PRS_ISOLATED) &&
> +			 !isolated_cpus_can_update(tmp->addmask, tmp->delmask))
> +			cs->prs_err = PERR_HKEEPING;
>  		if (cs->prs_err)
>  			goto invalidate;
>  	}
> @@ -1750,6 +1795,7 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
>  	int part_error = PERR_NONE;	/* Partition error? */
>  	int isolcpus_updated = 0;
>  	struct cpumask *xcpus = user_xcpus(cs);
> +	int parent_prs = parent->partition_root_state;
>  	bool nocpu;
>  
>  	lockdep_assert_held(&cpuset_mutex);
> @@ -1813,6 +1859,10 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
>  		if (prstate_housekeeping_conflict(new_prs, xcpus))
>  			return PERR_HKEEPING;
>  
> +		if ((new_prs == PRS_ISOLATED) && (new_prs != parent_prs) &&
> +		    !isolated_cpus_can_update(xcpus, NULL))
> +			return PERR_HKEEPING;
> +
>  		if (tasks_nocpu_error(parent, cs, xcpus))
>  			return PERR_NOCPUS;
>  
> @@ -1866,6 +1916,7 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
>  		 *
>  		 * For invalid partition:
>  		 *   delmask = newmask & parent->effective_xcpus
> +		 *   The partition may become valid soon.
>  		 */
>  		if (is_partition_invalid(cs)) {
>  			adding = false;
> @@ -1880,6 +1931,23 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
>  			deleting = cpumask_and(tmp->delmask, tmp->delmask,
>  					       parent->effective_xcpus);
>  		}
> +
> +		/*
> +		 * TBD: Invalidate a currently valid child root partition may
> +		 * still break isolated_cpus_can_update() rule if parent is an
> +		 * isolated partition.
> +		 */
> +		if (is_partition_valid(cs) && (old_prs != parent_prs)) {
> +			if ((parent_prs == PRS_ROOT) &&
> +			    /* Adding to parent means removing isolated CPUs */
> +			    !isolated_cpus_can_update(tmp->delmask, tmp->addmask))
> +				part_error = PERR_HKEEPING;
> +			if ((parent_prs == PRS_ISOLATED) &&
> +			    /* Adding to parent means adding isolated CPUs */
> +			    !isolated_cpus_can_update(tmp->addmask, tmp->delmask))
> +				part_error = PERR_HKEEPING;
> +		}
> +
>  		/*
>  		 * The new CPUs to be removed from parent's effective CPUs
>  		 * must be present.
> @@ -2994,7 +3062,11 @@ static int update_prstate(struct cpuset *cs, int new_prs)
>  		 * A change in load balance state only, no change in cpumasks.
>  		 * Need to update isolated_cpus.
>  		 */
> -		isolcpus_updated = true;
> +		if ((new_prs == PRS_ISOLATED) &&
> +		    !isolated_cpus_can_update(cs->effective_xcpus, NULL))
> +			err = PERR_HKEEPING;
> +		else
> +			isolcpus_updated = true;
>  	} else {
>  		/*
>  		 * Switching back to member is always allowed even if it

Reviewed-by: Chen Ridong <chenridong@huawei.com>

-- 
Best regards,
Ridong