2023-10-25 18:27:24

by Waiman Long

[permalink] [raw]
Subject: [PATCH v2 0/4] cgroup/cpuset: Improve CPU isolation in isolated partitions

v2:
- Add 2 read-only workqueue sysfs files to expose the user requested
cpumask as well as the isolated CPUs to be excluded from
wq_unbound_cpumask.
- Ensure that caller of the new workqueue_unbound_exclude_cpumask()
hold cpus_read_lock.
- Update the cpuset code to make sure the cpus_read_lock is held
whenever workqueue_unbound_exclude_cpumask() may be called.

Isolated cpuset partition can currently be created to contain an
exclusive set of CPUs not used in other cgroups and with load balancing
disabled to reduce interference from the scheduler.

The main purpose of this isolated partition type is to dynamically
emulate what can be done via the "isolcpus" boot command line option,
specifically the default domain flag. One effect of the "isolcpus" option
is to remove the isolated CPUs from the cpumasks of unbound workqueues
since running work functions in an isolated CPU can be a major source
of interference. Changing the unbound workqueue cpumasks can be done at
run time by writing an appropriate cpumask without the isolated CPUs to
/sys/devices/virtual/workqueue/cpumask. So one can set up an isolated
cpuset partition and then write to the cpumask sysfs file to achieve
similar level of CPU isolation. However, this manual process can be
error prone.

This patch series implements automatic exclusion of isolated CPUs from
unbound workqueue cpumasks when an isolated cpuset partition is created
and then adds those CPUs back when the isolated partition is destroyed.

There are also other places in the kernel that look at the HK_FLAG_DOMAIN
cpumask or other HK_FLAG_* cpumasks and exclude the isolated CPUs from
certain actions to further reduce interference. CPUs in an isolated
cpuset partition will not be able to avoid those interferences yet. That
may change in the future as the need arises.

Waiman Long (4):
workqueue: Add workqueue_unbound_exclude_cpumask() to exclude CPUs
from wq_unbound_cpumask
selftests/cgroup: Minor code cleanup and reorganization of
test_cpuset_prs.sh
cgroup/cpuset: Keep track of CPUs in isolated partitions
cgroup/cpuset: Take isolated CPUs out of workqueue unbound cpumask

Documentation/admin-guide/cgroup-v2.rst | 10 +-
include/linux/workqueue.h | 2 +-
kernel/cgroup/cpuset.c | 286 +++++++++++++-----
kernel/workqueue.c | 91 +++++-
.../selftests/cgroup/test_cpuset_prs.sh | 216 ++++++++-----
5 files changed, 438 insertions(+), 167 deletions(-)

--
2.39.3


2023-10-25 18:27:27

by Waiman Long

[permalink] [raw]
Subject: [PATCH v2 4/4] cgroup/cpuset: Take isolated CPUs out of workqueue unbound cpumask

To make CPUs in isolated cpuset partition closer in isolation to
the boot time isolated CPUs specified in the "isolcpus" boot command
line option, we need to take those CPUs out of the workqueue unbound
cpumask so that work functions from the unbound workqueues won't run
on those CPUs. Otherwise, they will interfere the user tasks running
on those isolated CPUs.

With the introduction of the workqueue_unbound_exclude_cpumask() helper
function in an earlier commit, those isolated CPUs can now be taken
out from the workqueue unbound cpumask.

This patch also updates cgroup-v2.rst to mention that isolated
CPUs will be excluded from unbound workqueue cpumask as well as
updating test_cpuset_prs.sh to verify the correctness of the new
*cpuset.cpus.isolated file, if available via cgroup_debug option.

Signed-off-by: Waiman Long <[email protected]>
---
Documentation/admin-guide/cgroup-v2.rst | 10 +-
kernel/cgroup/cpuset.c | 116 +++++++++++++++---
.../selftests/cgroup/test_cpuset_prs.sh | 74 +++++++++--
3 files changed, 166 insertions(+), 34 deletions(-)

diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst
index e440aee4fe94..9db311448a21 100644
--- a/Documentation/admin-guide/cgroup-v2.rst
+++ b/Documentation/admin-guide/cgroup-v2.rst
@@ -2318,11 +2318,11 @@ Cpuset Interface Files
partition or scheduling domain. The set of exclusive CPUs is
determined by the value of its "cpuset.cpus.exclusive.effective".

- When set to "isolated", the CPUs in that partition will
- be in an isolated state without any load balancing from the
- scheduler. Tasks placed in such a partition with multiple
- CPUs should be carefully distributed and bound to each of the
- individual CPUs for optimal performance.
+ When set to "isolated", the CPUs in that partition will be in
+ an isolated state without any load balancing from the scheduler
+ and excluded from the unbound workqueues. Tasks placed in such
+ a partition with multiple CPUs should be carefully distributed
+ and bound to each of the individual CPUs for optimal performance.

A partition root ("root" or "isolated") can be in one of the
two possible states - valid or invalid. An invalid partition
diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index 19c8779798fd..1bad4007ff4b 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -25,6 +25,7 @@
#include <linux/cpu.h>
#include <linux/cpumask.h>
#include <linux/cpuset.h>
+#include <linux/delay.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/kernel.h>
@@ -43,6 +44,7 @@
#include <linux/sched/isolation.h>
#include <linux/cgroup.h>
#include <linux/wait.h>
+#include <linux/workqueue.h>

DEFINE_STATIC_KEY_FALSE(cpusets_pre_enable_key);
DEFINE_STATIC_KEY_FALSE(cpusets_enabled_key);
@@ -1444,25 +1446,31 @@ static void partition_xcpus_newstate(int old_prs, int new_prs, struct cpumask *x
* @new_prs: new partition_root_state
* @parent: parent cpuset
* @xcpus: exclusive CPUs to be added
+ * Return: true if isolated_cpus modified, false otherwise
*
* Remote partition if parent == NULL
*/
-static void partition_xcpus_add(int new_prs, struct cpuset *parent,
+static bool partition_xcpus_add(int new_prs, struct cpuset *parent,
struct cpumask *xcpus)
{
+ bool isolcpus_updated;
+
WARN_ON_ONCE(new_prs < 0);
lockdep_assert_held(&callback_lock);
if (!parent)
parent = &top_cpuset;

+
if (parent == &top_cpuset)
cpumask_or(subpartitions_cpus, subpartitions_cpus, xcpus);

- if (new_prs != parent->partition_root_state)
+ isolcpus_updated = (new_prs != parent->partition_root_state);
+ if (isolcpus_updated)
partition_xcpus_newstate(parent->partition_root_state, new_prs,
xcpus);

cpumask_andnot(parent->effective_cpus, parent->effective_cpus, xcpus);
+ return isolcpus_updated;
}

/*
@@ -1470,12 +1478,15 @@ static void partition_xcpus_add(int new_prs, struct cpuset *parent,
* @old_prs: old partition_root_state
* @parent: parent cpuset
* @xcpus: exclusive CPUs to be removed
+ * Return: true if isolated_cpus modified, false otherwise
*
* Remote partition if parent == NULL
*/
-static void partition_xcpus_del(int old_prs, struct cpuset *parent,
+static bool partition_xcpus_del(int old_prs, struct cpuset *parent,
struct cpumask *xcpus)
{
+ bool isolcpus_updated;
+
WARN_ON_ONCE(old_prs < 0);
lockdep_assert_held(&callback_lock);
if (!parent)
@@ -1484,12 +1495,27 @@ static void partition_xcpus_del(int old_prs, struct cpuset *parent,
if (parent == &top_cpuset)
cpumask_andnot(subpartitions_cpus, subpartitions_cpus, xcpus);

- if (old_prs != parent->partition_root_state)
+ isolcpus_updated = (old_prs != parent->partition_root_state);
+ if (isolcpus_updated)
partition_xcpus_newstate(old_prs, parent->partition_root_state,
xcpus);

cpumask_and(xcpus, xcpus, cpu_active_mask);
cpumask_or(parent->effective_cpus, parent->effective_cpus, xcpus);
+ return isolcpus_updated;
+}
+
+static void update_unbound_workqueue_cpumask(bool isolcpus_updated)
+{
+ int ret;
+
+ lockdep_assert_cpus_held();
+
+ if (!isolcpus_updated)
+ return;
+
+ ret = workqueue_unbound_exclude_cpumask(isolated_cpus);
+ WARN_ON_ONCE(ret < 0);
}

/*
@@ -1540,6 +1566,8 @@ static inline bool is_local_partition(struct cpuset *cs)
static int remote_partition_enable(struct cpuset *cs, int new_prs,
struct tmpmasks *tmp)
{
+ bool isolcpus_updated;
+
/*
* The user must have sysadmin privilege.
*/
@@ -1561,7 +1589,7 @@ static int remote_partition_enable(struct cpuset *cs, int new_prs,
return 0;

spin_lock_irq(&callback_lock);
- partition_xcpus_add(new_prs, NULL, tmp->new_cpus);
+ isolcpus_updated = partition_xcpus_add(new_prs, NULL, tmp->new_cpus);
list_add(&cs->remote_sibling, &remote_children);
if (cs->use_parent_ecpus) {
struct cpuset *parent = parent_cs(cs);
@@ -1570,13 +1598,13 @@ static int remote_partition_enable(struct cpuset *cs, int new_prs,
parent->child_ecpus_count--;
}
spin_unlock_irq(&callback_lock);
+ update_unbound_workqueue_cpumask(isolcpus_updated);

/*
* Proprogate changes in top_cpuset's effective_cpus down the hierarchy.
*/
update_tasks_cpumask(&top_cpuset, tmp->new_cpus);
update_sibling_cpumasks(&top_cpuset, NULL, tmp);
-
return 1;
}

@@ -1591,18 +1619,22 @@ static int remote_partition_enable(struct cpuset *cs, int new_prs,
*/
static void remote_partition_disable(struct cpuset *cs, struct tmpmasks *tmp)
{
+ bool isolcpus_updated;
+
compute_effective_exclusive_cpumask(cs, tmp->new_cpus);
WARN_ON_ONCE(!is_remote_partition(cs));
WARN_ON_ONCE(!cpumask_subset(tmp->new_cpus, subpartitions_cpus));

spin_lock_irq(&callback_lock);
list_del_init(&cs->remote_sibling);
- partition_xcpus_del(cs->partition_root_state, NULL, tmp->new_cpus);
+ isolcpus_updated = partition_xcpus_del(cs->partition_root_state,
+ NULL, tmp->new_cpus);
cs->partition_root_state = -cs->partition_root_state;
if (!cs->prs_err)
cs->prs_err = PERR_INVCPUS;
reset_partition_data(cs);
spin_unlock_irq(&callback_lock);
+ update_unbound_workqueue_cpumask(isolcpus_updated);

/*
* Proprogate changes in top_cpuset's effective_cpus down the hierarchy.
@@ -1625,6 +1657,7 @@ static void remote_cpus_update(struct cpuset *cs, struct cpumask *newmask,
{
bool adding, deleting;
int prs = cs->partition_root_state;
+ int isolcpus_updated = 0;

if (WARN_ON_ONCE(!is_remote_partition(cs)))
return;
@@ -1649,10 +1682,11 @@ static void remote_cpus_update(struct cpuset *cs, struct cpumask *newmask,

spin_lock_irq(&callback_lock);
if (adding)
- partition_xcpus_add(prs, NULL, tmp->addmask);
+ isolcpus_updated += partition_xcpus_add(prs, NULL, tmp->addmask);
if (deleting)
- partition_xcpus_del(prs, NULL, tmp->delmask);
+ isolcpus_updated += partition_xcpus_del(prs, NULL, tmp->delmask);
spin_unlock_irq(&callback_lock);
+ update_unbound_workqueue_cpumask(isolcpus_updated);

/*
* Proprogate changes in top_cpuset's effective_cpus down the hierarchy.
@@ -1774,6 +1808,7 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
int part_error = PERR_NONE; /* Partition error? */
int subparts_delta = 0;
struct cpumask *xcpus; /* cs effective_xcpus */
+ int isolcpus_updated = 0;
bool nocpu;

lockdep_assert_held(&cpuset_mutex);
@@ -2010,15 +2045,18 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
* and vice versa.
*/
if (adding)
- partition_xcpus_del(old_prs, parent, tmp->addmask);
+ isolcpus_updated += partition_xcpus_del(old_prs, parent,
+ tmp->addmask);
if (deleting)
- partition_xcpus_add(new_prs, parent, tmp->delmask);
+ isolcpus_updated += partition_xcpus_add(new_prs, parent,
+ tmp->delmask);

if (is_partition_valid(parent)) {
parent->nr_subparts += subparts_delta;
WARN_ON_ONCE(parent->nr_subparts < 0);
}
spin_unlock_irq(&callback_lock);
+ update_unbound_workqueue_cpumask(isolcpus_updated);

if ((old_prs != new_prs) && (cmd == partcmd_update))
update_partition_exclusive(cs, new_prs);
@@ -3082,6 +3120,7 @@ static int update_prstate(struct cpuset *cs, int new_prs)
else if (new_xcpus_state)
partition_xcpus_newstate(old_prs, new_prs, cs->effective_xcpus);
spin_unlock_irq(&callback_lock);
+ update_unbound_workqueue_cpumask(new_xcpus_state);

/* Force update if switching back to member */
update_cpumasks_hier(cs, &tmpmask, !new_prs ? HIER_CHECKALL : 0);
@@ -4370,6 +4409,30 @@ void cpuset_force_rebuild(void)
force_rebuild = true;
}

+/*
+ * Attempt to acquire a cpus_read_lock while a hotplug operation may be in
+ * progress.
+ * Return: true if successful, false otherwise
+ *
+ * To avoid circular lock dependency between cpuset_mutex and cpus_read_lock,
+ * cpus_read_trylock() is used here to acquire the lock.
+ */
+static bool cpuset_hotplug_cpus_read_trylock(void)
+{
+ int retries = 0;
+
+ while (!cpus_read_trylock()) {
+ /*
+ * CPU hotplug still in progress. Retry 5 times
+ * with a 10ms wait before bailing out.
+ */
+ if (++retries > 5)
+ return false;
+ msleep(10);
+ }
+ return true;
+}
+
/**
* cpuset_hotplug_update_tasks - update tasks in a cpuset for hotunplug
* @cs: cpuset in interest
@@ -4386,6 +4449,7 @@ static void cpuset_hotplug_update_tasks(struct cpuset *cs, struct tmpmasks *tmp)
bool cpus_updated;
bool mems_updated;
bool remote;
+ int partcmd = -1;
struct cpuset *parent;
retry:
wait_event(cpuset_attach_wq, cs->attach_in_progress == 0);
@@ -4417,11 +4481,13 @@ static void cpuset_hotplug_update_tasks(struct cpuset *cs, struct tmpmasks *tmp)
compute_partition_effective_cpumask(cs, &new_cpus);

if (remote && cpumask_empty(&new_cpus) &&
- partition_is_populated(cs, NULL)) {
+ partition_is_populated(cs, NULL) &&
+ cpuset_hotplug_cpus_read_trylock()) {
remote_partition_disable(cs, tmp);
compute_effective_cpumask(&new_cpus, cs, parent);
remote = false;
cpuset_force_rebuild();
+ cpus_read_unlock();
}

/*
@@ -4432,18 +4498,28 @@ static void cpuset_hotplug_update_tasks(struct cpuset *cs, struct tmpmasks *tmp)
* partitions.
*/
if (is_local_partition(cs) && (!is_partition_valid(parent) ||
- tasks_nocpu_error(parent, cs, &new_cpus))) {
- update_parent_effective_cpumask(cs, partcmd_invalidate, NULL, tmp);
- compute_effective_cpumask(&new_cpus, cs, parent);
- cpuset_force_rebuild();
- }
+ tasks_nocpu_error(parent, cs, &new_cpus)))
+ partcmd = partcmd_invalidate;
/*
* On the other hand, an invalid partition root may be transitioned
* back to a regular one.
*/
- else if (is_partition_valid(parent) && is_partition_invalid(cs)) {
- update_parent_effective_cpumask(cs, partcmd_update, NULL, tmp);
- if (is_partition_valid(cs)) {
+ else if (is_partition_valid(parent) && is_partition_invalid(cs))
+ partcmd = partcmd_update;
+
+ /*
+ * cpus_read_lock needs to be held before calling
+ * update_parent_effective_cpumask(). To avoid circular lock
+ * dependency between cpuset_mutex and cpus_read_lock,
+ * cpus_read_trylock() is used here to acquire the lock.
+ */
+ if (partcmd >= 0) {
+ if (!cpuset_hotplug_cpus_read_trylock())
+ goto update_tasks;
+
+ update_parent_effective_cpumask(cs, partcmd, NULL, tmp);
+ cpus_read_unlock();
+ if ((partcmd == partcmd_invalidate) || is_partition_valid(cs)) {
compute_partition_effective_cpumask(cs, &new_cpus);
cpuset_force_rebuild();
}
diff --git a/tools/testing/selftests/cgroup/test_cpuset_prs.sh b/tools/testing/selftests/cgroup/test_cpuset_prs.sh
index 2b825019f806..7b7c4c2b6d85 100755
--- a/tools/testing/selftests/cgroup/test_cpuset_prs.sh
+++ b/tools/testing/selftests/cgroup/test_cpuset_prs.sh
@@ -232,11 +232,11 @@ TEST_MATRIX=(
" C0-3:S+ C1-3:S+ C2-3 C4-5 X2-3 X2-3:P1 P2 P1 0 A1:0-1,A2:,A3:2-3,B1:4-5 \
A1:P0,A2:P1,A3:P2,B1:P1 2-3"
" C0-3:S+ C1-3:S+ C2-3 C4 X2-3 X2-3:P1 P2 P1 0 A1:0-1,A2:,A3:2-3,B1:4 \
- A1:P0,A2:P1,A3:P2,B1:P1 2-4"
+ A1:P0,A2:P1,A3:P2,B1:P1 2-4,2-3"
" C0-3:S+ C1-3:S+ C3 C4 X2-3 X2-3:P1 P2 P1 0 A1:0-1,A2:2,A3:3,B1:4 \
- A1:P0,A2:P1,A3:P2,B1:P1 2-4"
+ A1:P0,A2:P1,A3:P2,B1:P1 2-4,3"
" C0-4:S+ C1-4:S+ C2-4 . X2-4 X2-4:P2 X4:P1 . 0 A1:0-1,A2:2-3,A3:4 \
- A1:P0,A2:P2,A3:P1 2-4"
+ A1:P0,A2:P2,A3:P1 2-4,2-3"
" C0-4:X2-4:S+ C1-4:X2-4:S+:P2 C2-4:X4:P1 \
. . X5 . . 0 A1:0-4,A2:1-4,A3:2-4 \
A1:P0,A2:P-2,A3:P-1"
@@ -248,7 +248,7 @@ TEST_MATRIX=(
" C0-3:S+ C1-3:S+ C2-3 . X2-3 X2-3 X2-3:P2:O2=0 . 0 A1:0-1,A2:1,A3:3 A1:P0,A3:P2 2-3"
" C0-3:S+ C1-3:S+ C2-3 . X2-3 X2-3 X2-3:P2:O2=0 O2=1 0 A1:0-1,A2:1,A3:2-3 A1:P0,A3:P2 2-3"
" C0-3:S+ C1-3:S+ C3 . X2-3 X2-3 P2:O3=0 . 0 A1:0-2,A2:1-2,A3: A1:P0,A3:P2 3"
- " C0-3:S+ C1-3:S+ C3 . X2-3 X2-3 T:P2:O3=0 . 0 A1:0-2,A2:1-2,A3:1-2 A1:P0,A3:P-2 3"
+ " C0-3:S+ C1-3:S+ C3 . X2-3 X2-3 T:P2:O3=0 . 0 A1:0-2,A2:1-2,A3:1-2 A1:P0,A3:P-2 3,"

# An invalidated remote partition cannot self-recover from hotplug
" C0-3:S+ C1-3:S+ C2 . X2-3 X2-3 T:P2:O2=0 O2=1 0 A1:0-3,A2:1-3,A3:2 A1:P0,A3:P-2"
@@ -376,7 +376,7 @@ write_cpu_online()
}
fi
echo $VAL > $CPUFILE
- pause 0.01
+ pause 0.05
}

#
@@ -508,12 +508,14 @@ dump_states()
XECPUS=$DIR/cpuset.cpus.exclusive.effective
PRS=$DIR/cpuset.cpus.partition
PCPUS=$DIR/.__DEBUG__.cpuset.cpus.subpartitions
+ ISCPUS=$DIR/.__DEBUG__.cpuset.cpus.isolated
[[ -e $CPUS ]] && echo "$CPUS: $(cat $CPUS)"
[[ -e $XCPUS ]] && echo "$XCPUS: $(cat $XCPUS)"
[[ -e $ECPUS ]] && echo "$ECPUS: $(cat $ECPUS)"
[[ -e $XECPUS ]] && echo "$XECPUS: $(cat $XECPUS)"
[[ -e $PRS ]] && echo "$PRS: $(cat $PRS)"
[[ -e $PCPUS ]] && echo "$PCPUS: $(cat $PCPUS)"
+ [[ -e $ISCPUS ]] && echo "$ISCPUS: $(cat $ISCPUS)"
done
}

@@ -591,11 +593,17 @@ check_cgroup_states()

#
# Get isolated (including offline) CPUs by looking at
-# /sys/kernel/debug/sched/domains and compare that with the expected value.
+# /sys/kernel/debug/sched/domains and *cpuset.cpus.isolated control file,
+# if available, and compare that with the expected value.
#
-# Note that a sched domain of just 1 CPU will be considered isolated.
+# Note that isolated CPUs from the sched/domains context include offline
+# CPUs as well as CPUs in non-isolated 1-CPU partition. Those CPUs may
+# not be included in the *cpuset.cpus.isolated control file which contains
+# only CPUs in isolated partitions.
#
-# $1 - expected isolated cpu list
+# $1 - expected isolated cpu list(s) <isolcpus1>{,<isolcpus2>}
+# <isolcpus1> - expected sched/domains value
+# <isolcpus2> - *cpuset.cpus.isolated value = <isolcpus1> if not defined
#
check_isolcpus()
{
@@ -603,8 +611,38 @@ check_isolcpus()
ISOLCPUS=
LASTISOLCPU=
SCHED_DOMAINS=/sys/kernel/debug/sched/domains
+ ISCPUS=${CGROUP2}/.__DEBUG__.cpuset.cpus.isolated
+ if [[ $EXPECT_VAL = . ]]
+ then
+ EXPECT_VAL=
+ EXPECT_VAL2=
+ elif [[ $(expr $EXPECT_VAL : ".*,.*") > 0 ]]
+ then
+ set -- $(echo $EXPECT_VAL | sed -e "s/,/ /g")
+ EXPECT_VAL=$1
+ EXPECT_VAL2=$2
+ else
+ EXPECT_VAL2=$EXPECT_VAL
+ fi
+
+ #
+ # Check the debug isolated cpumask, if present
+ #
+ [[ -f $ISCPUS ]] && {
+ ISOLCPUS=$(cat $ISCPUS)
+ [[ "$EXPECT_VAL2" != "$ISOLCPUS" ]] && {
+ # Take a 50ms pause and try again
+ pause 0.05
+ ISOLCPUS=$(cat $ISCPUS)
+ }
+ [[ "$EXPECT_VAL2" != "$ISOLCPUS" ]] && return 1
+ ISOLCPUS=
+ }
+
+ #
+ # Use the sched domain in debugfs to check isolated CPUs, if available
+ #
[[ -d $SCHED_DOMAINS ]] || return 0
- [[ $EXPECT_VAL = . ]] && EXPECT_VAL=

for ((CPU=0; CPU < $NR_CPUS; CPU++))
do
@@ -648,6 +686,22 @@ test_fail()
exit 1
}

+#
+# Check to see if there are unexpected isolated CPUs left
+#
+null_isolcpus_check()
+{
+ [[ $VERBOSE -gt 0 ]] || return 0
+ pause 0.02
+ check_isolcpus "."
+ if [[ $? -ne 0 ]]
+ then
+ echo "Unexpected isolated CPUs: $ISOLCPUS"
+ dump_states
+ exit 1
+ fi
+}
+
#
# Run cpuset state transition test
# $1 - test matrix name
@@ -733,6 +787,7 @@ run_state_test()
echo "Effective cpus changed to $NEWLIST after test $I!"
exit 1
}
+ null_isolcpus_check
[[ $VERBOSE -gt 0 ]] && echo "Test $I done."
((I++))
done
@@ -802,6 +857,7 @@ test_isolated()
console_msg "Cleaning up"
echo $$ > $CGROUP2/cgroup.procs
[[ -d A1 ]] && rmdir A1
+ null_isolcpus_check
}

#
--
2.39.3

2023-10-25 18:27:37

by Waiman Long

[permalink] [raw]
Subject: [PATCH v2 1/4] workqueue: Add workqueue_unbound_exclude_cpumask() to exclude CPUs from wq_unbound_cpumask

When the "isolcpus" boot command line option is used to add a set
of isolated CPUs, those CPUs will be excluded automatically from
wq_unbound_cpumask to avoid running work functions from unbound
workqueues.

Recently cpuset has been extended to allow the creation of partitions
of isolated CPUs dynamically. To make it closer to the "isolcpus"
in functionality, the CPUs in those isolated cpuset partitions should be
excluded from wq_unbound_cpumask as well. This can be done currently by
explicitly writing to the workqueue's cpumask sysfs file after creating
the isolated partitions. However, this process can be error prone.

Ideally, the cpuset code should be allowed to request the workqueue code
to exclude those isolated CPUs from wq_unbound_cpumask so that this
operation can be done automatically and the isolated CPUs will be returned
back to wq_unbound_cpumask after the destructions of the isolated
cpuset partitions.

This patch adds a new workqueue_unbound_exclude_cpumask() function to
enable that. This new function will exclude the specified isolated
CPUs from wq_unbound_cpumask. To be able to restore those isolated
CPUs back after the destruction of isolated cpuset partitions, a new
wq_requested_unbound_cpumask is added to store the user provided unbound
cpumask either from the boot command line options or from writing to
the cpumask sysfs file. This new cpumask provides the basis for CPU
exclusion.

To enable users to understand how the wq_unbound_cpumask is being
modified internally, this patch also exposes the newly introduced
wq_requested_unbound_cpumask as well as a wq_isolated_cpumask to
store the cpumask to be excluded from wq_unbound_cpumask as read-only
sysfs files.

Signed-off-by: Waiman Long <[email protected]>
---
include/linux/workqueue.h | 2 +-
kernel/workqueue.c | 91 +++++++++++++++++++++++++++++++++++----
2 files changed, 84 insertions(+), 9 deletions(-)

diff --git a/include/linux/workqueue.h b/include/linux/workqueue.h
index 1c1d06804d45..a936460ccc7e 100644
--- a/include/linux/workqueue.h
+++ b/include/linux/workqueue.h
@@ -483,7 +483,7 @@ struct workqueue_attrs *alloc_workqueue_attrs(void);
void free_workqueue_attrs(struct workqueue_attrs *attrs);
int apply_workqueue_attrs(struct workqueue_struct *wq,
const struct workqueue_attrs *attrs);
-int workqueue_set_unbound_cpumask(cpumask_var_t cpumask);
+extern int workqueue_unbound_exclude_cpumask(cpumask_var_t cpumask);

extern bool queue_work_on(int cpu, struct workqueue_struct *wq,
struct work_struct *work);
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 010b674b02a7..ce20080751bd 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -381,6 +381,12 @@ static bool workqueue_freezing; /* PL: have wqs started freezing? */
/* PL&A: allowable cpus for unbound wqs and work items */
static cpumask_var_t wq_unbound_cpumask;

+/* PL: user requested unbound cpumask via sysfs */
+static cpumask_var_t wq_requested_unbound_cpumask;
+
+/* PL: isolated cpumask to be excluded from unbound cpumask */
+static cpumask_var_t wq_isolated_cpumask;
+
/* for further constrain wq_unbound_cpumask by cmdline parameter*/
static struct cpumask wq_cmdline_cpumask __initdata;

@@ -5825,7 +5831,7 @@ static int workqueue_apply_unbound_cpumask(const cpumask_var_t unbound_cpumask)
* -EINVAL - Invalid @cpumask
* -ENOMEM - Failed to allocate memory for attrs or pwqs.
*/
-int workqueue_set_unbound_cpumask(cpumask_var_t cpumask)
+static int workqueue_set_unbound_cpumask(cpumask_var_t cpumask)
{
int ret = -EINVAL;

@@ -5836,6 +5842,7 @@ int workqueue_set_unbound_cpumask(cpumask_var_t cpumask)
cpumask_and(cpumask, cpumask, cpu_possible_mask);
if (!cpumask_empty(cpumask)) {
apply_wqattrs_lock();
+ cpumask_copy(wq_requested_unbound_cpumask, cpumask);
if (cpumask_equal(cpumask, wq_unbound_cpumask)) {
ret = 0;
goto out_unlock;
@@ -5850,6 +5857,44 @@ int workqueue_set_unbound_cpumask(cpumask_var_t cpumask)
return ret;
}

+/**
+ * workqueue_unbound_exclude_cpumask - Exclude given CPUs from unbound cpumask
+ * @exclude_cpumask: the cpumask to be excluded from wq_unbound_cpumask
+ *
+ * This function can be called from cpuset code to provide a set of isolated
+ * CPUs that should be excluded from wq_unbound_cpumask. The caller must hold
+ * either cpus_read_lock or cpus_write_lock.
+ */
+int workqueue_unbound_exclude_cpumask(cpumask_var_t exclude_cpumask)
+{
+ cpumask_var_t cpumask;
+ int ret = 0;
+
+ if (!zalloc_cpumask_var(&cpumask, GFP_KERNEL))
+ return -ENOMEM;
+
+ lockdep_assert_cpus_held();
+ mutex_lock(&wq_pool_mutex);
+
+ /* Save the current isolated cpumask & export it via sysfs */
+ cpumask_copy(wq_isolated_cpumask, exclude_cpumask);
+
+ /*
+ * If the operation fails, it will fall back to
+ * wq_requested_unbound_cpumask which is initially set to
+ * (HK_TYPE_WQ ∩ HK_TYPE_DOMAIN) house keeping mask and rewritten
+ * by any subsequent write to workqueue/cpumask sysfs file.
+ */
+ if (!cpumask_andnot(cpumask, wq_requested_unbound_cpumask, exclude_cpumask))
+ cpumask_copy(cpumask, wq_requested_unbound_cpumask);
+ if (!cpumask_equal(cpumask, wq_unbound_cpumask))
+ ret = workqueue_apply_unbound_cpumask(cpumask);
+
+ mutex_unlock(&wq_pool_mutex);
+ free_cpumask_var(cpumask);
+ return ret;
+}
+
static int parse_affn_scope(const char *val)
{
int i;
@@ -6144,19 +6189,36 @@ static struct bus_type wq_subsys = {
.dev_groups = wq_sysfs_groups,
};

-static ssize_t wq_unbound_cpumask_show(struct device *dev,
- struct device_attribute *attr, char *buf)
+static ssize_t __wq_cpumask_show(struct device *dev,
+ struct device_attribute *attr, char *buf, cpumask_var_t mask)
{
int written;

mutex_lock(&wq_pool_mutex);
- written = scnprintf(buf, PAGE_SIZE, "%*pb\n",
- cpumask_pr_args(wq_unbound_cpumask));
+ written = scnprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask));
mutex_unlock(&wq_pool_mutex);

return written;
}

+static ssize_t wq_unbound_cpumask_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ return __wq_cpumask_show(dev, attr, buf, wq_unbound_cpumask);
+}
+
+static ssize_t wq_requested_cpumask_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ return __wq_cpumask_show(dev, attr, buf, wq_requested_unbound_cpumask);
+}
+
+static ssize_t wq_isolated_cpumask_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ return __wq_cpumask_show(dev, attr, buf, wq_isolated_cpumask);
+}
+
static ssize_t wq_unbound_cpumask_store(struct device *dev,
struct device_attribute *attr, const char *buf, size_t count)
{
@@ -6174,9 +6236,13 @@ static ssize_t wq_unbound_cpumask_store(struct device *dev,
return ret ? ret : count;
}

-static struct device_attribute wq_sysfs_cpumask_attr =
+static struct device_attribute wq_sysfs_cpumask_attrs[] = {
__ATTR(cpumask, 0644, wq_unbound_cpumask_show,
- wq_unbound_cpumask_store);
+ wq_unbound_cpumask_store),
+ __ATTR(cpumask_requested, 0444, wq_requested_cpumask_show, NULL),
+ __ATTR(cpumask_isolated, 0444, wq_isolated_cpumask_show, NULL),
+ __ATTR_NULL,
+};

static int __init wq_sysfs_init(void)
{
@@ -6189,7 +6255,13 @@ static int __init wq_sysfs_init(void)

dev_root = bus_get_dev_root(&wq_subsys);
if (dev_root) {
- err = device_create_file(dev_root, &wq_sysfs_cpumask_attr);
+ struct device_attribute *attr;
+
+ for (attr = wq_sysfs_cpumask_attrs; attr->attr.name; attr++) {
+ err = device_create_file(dev_root, attr);
+ if (err)
+ break;
+ }
put_device(dev_root);
}
return err;
@@ -6520,11 +6592,14 @@ void __init workqueue_init_early(void)
BUILD_BUG_ON(__alignof__(struct pool_workqueue) < __alignof__(long long));

BUG_ON(!alloc_cpumask_var(&wq_unbound_cpumask, GFP_KERNEL));
+ BUG_ON(!alloc_cpumask_var(&wq_requested_unbound_cpumask, GFP_KERNEL));
+ BUG_ON(!zalloc_cpumask_var(&wq_isolated_cpumask, GFP_KERNEL));
cpumask_copy(wq_unbound_cpumask, housekeeping_cpumask(HK_TYPE_WQ));
cpumask_and(wq_unbound_cpumask, wq_unbound_cpumask, housekeeping_cpumask(HK_TYPE_DOMAIN));

if (!cpumask_empty(&wq_cmdline_cpumask))
cpumask_and(wq_unbound_cpumask, wq_unbound_cpumask, &wq_cmdline_cpumask);
+ cpumask_copy(wq_requested_unbound_cpumask, wq_unbound_cpumask);

pwq_cache = KMEM_CACHE(pool_workqueue, SLAB_PANIC);

--
2.39.3