From: Yu Kuai <[email protected]>
Changes in v8:
- fix that another unrelated patch is merged into this patchset
- update commit message for patch 5
- rebase on the top of another patch from me: [PATCH v3] md: fix duplicate filename for rdev
Changes in v7:
- add missing changes for md-cluster
- rebase for the latest md-next branch
Changes in v6:
- remove first patch from v5, and use rcu_asign_pointer() directly from
caller.
- always use rcu_read_lock/unlock to protect mddev_set_timeout().
Changes in v5:
- use rcu_dereference_protected() instead of rcu_access_pointer() where
rcu_read_lock/unlock is not required.
- add patch 4,5 to handle that bitmap timeout is set multiple times.
Changes in v4:
- remove patch 2 from v3
- fix sparse errors and warnings from v3, in order to do that, all access
to md_thread need to be modified, patch 2-4 is splited to avoid a huge
patch.
Changes in v3:
- remove patch 3 from v2
- use rcu instead of a new lock
Changes in v2:
- fix a compile error for md-cluster in patch 2
- replace spin_lock/unlock with spin_lock/unlock_irq in patch 5
- don't wake up inside the new lock in md wakeup_thread in patch 5
Yu Kuai (5):
md: factor out a helper to wake up md_thread directly
dm-raid: remove useless checking in raid_message()
md/bitmap: always wake up md_thread in timeout_store
md/bitmap: factor out a helper to set timeout
md: protect md_thread with rcu
drivers/md/dm-raid.c | 4 +-
drivers/md/md-bitmap.c | 43 +++++++++++--------
drivers/md/md-cluster.c | 17 +++++---
drivers/md/md-multipath.c | 4 +-
drivers/md/md.c | 88 +++++++++++++++++++++------------------
drivers/md/md.h | 8 ++--
drivers/md/raid1.c | 7 ++--
drivers/md/raid1.h | 2 +-
drivers/md/raid10.c | 20 +++++----
drivers/md/raid10.h | 2 +-
drivers/md/raid5-cache.c | 22 ++++++----
drivers/md/raid5.c | 15 +++----
drivers/md/raid5.h | 2 +-
13 files changed, 132 insertions(+), 102 deletions(-)
--
2.39.2
From: Yu Kuai <[email protected]>
md_wakeup_thread() can't wakeup md_thread->tsk if md_thread->run is
still in progress, and in some cases md_thread->tsk need to be woke up
directly, like md_set_readonly() and do_md_stop().
Commit 9dfbdafda3b3 ("md: unlock mddev before reap sync_thread in
action_store") introduce a new scenario where unregister sync_thread is
not protected by 'reconfig_mutex', this can cause null-ptr-deference in
theroy:
t1: md_set_readonly t2: action_store
md_unregister_thread
// 'reconfig_mutex' is not held
// 'reconfig_mutex' is held by caller
if (mddev->sync_thread)
thread = *threadp
*threadp = NULL
wake_up_process(mddev->sync_thread->tsk)
// null-ptr-deference
Fix this problem by factoring out a helper to wake up md_thread directly,
so that 'sync_thread' won't be accessed multiple times from the reader
side. This helper also prepare to protect md_thread with rcu.
Noted that later patches is going to fix that unregister sync_thread is
not protected by 'reconfig_mutex' from action_store().
Signed-off-by: Yu Kuai <[email protected]>
---
drivers/md/md.c | 27 +++++++++++++++++++--------
1 file changed, 19 insertions(+), 8 deletions(-)
diff --git a/drivers/md/md.c b/drivers/md/md.c
index a0a2be43aabf..66a7a17510ce 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -92,6 +92,7 @@ static int remove_and_add_spares(struct mddev *mddev,
struct md_rdev *this);
static void mddev_detach(struct mddev *mddev);
static void export_rdev(struct md_rdev *rdev);
+static void md_wakeup_thread_directly(struct md_thread *thread);
/*
* Default number of read corrections we'll attempt on an rdev
@@ -6278,10 +6279,12 @@ static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
}
if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
set_bit(MD_RECOVERY_INTR, &mddev->recovery);
- if (mddev->sync_thread)
- /* Thread might be blocked waiting for metadata update
- * which will now never happen */
- wake_up_process(mddev->sync_thread->tsk);
+
+ /*
+ * Thread might be blocked waiting for metadata update which will now
+ * never happen
+ */
+ md_wakeup_thread_directly(mddev->sync_thread);
if (mddev->external && test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
return -EBUSY;
@@ -6342,10 +6345,12 @@ static int do_md_stop(struct mddev *mddev, int mode,
}
if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
set_bit(MD_RECOVERY_INTR, &mddev->recovery);
- if (mddev->sync_thread)
- /* Thread might be blocked waiting for metadata update
- * which will now never happen */
- wake_up_process(mddev->sync_thread->tsk);
+
+ /*
+ * Thread might be blocked waiting for metadata update which will now
+ * never happen
+ */
+ md_wakeup_thread_directly(mddev->sync_thread);
mddev_unlock(mddev);
wait_event(resync_wait, (mddev->sync_thread == NULL &&
@@ -7892,6 +7897,12 @@ static int md_thread(void *arg)
return 0;
}
+static void md_wakeup_thread_directly(struct md_thread *thread)
+{
+ if (thread)
+ wake_up_process(thread->tsk);
+}
+
void md_wakeup_thread(struct md_thread *thread)
{
if (thread) {
--
2.39.2
From: Yu Kuai <[email protected]>
Register/unregister 'mddev->thread' are both under 'reconfig_mutex',
however, some context didn't hold the mutex to access mddev->thread,
which can cause null-ptr-deference:
1) md_bitmap_daemon_work() can be called from md_check_recovery() where
'reconfig_mutex' is not held, deference 'mddev->thread' might cause
null-ptr-deference, because md_unregister_thread() reset the pointer
before stopping the thread.
2) timeout_store() access 'mddev->thread' multiple times,
null-ptr-deference can be triggered if 'mddev->thread' is reset in the
middle.
This patch factor out a helper to set timeout, the new helper always
check if 'mddev->thread' is null first, so that problem 1 can be fixed.
Now that this helper only access 'mddev->thread' once, but it's possible
that 'mddev->thread' can be freed while this helper is still in progress,
hence the problem is not fixed yet. Follow up patches will fix this by
protecting md_thread with rcu.
Signed-off-by: Yu Kuai <[email protected]>
---
drivers/md/md-bitmap.c | 35 +++++++++++++++++++----------------
1 file changed, 19 insertions(+), 16 deletions(-)
diff --git a/drivers/md/md-bitmap.c b/drivers/md/md-bitmap.c
index 4b5ba81c53be..23522df41ca5 100644
--- a/drivers/md/md-bitmap.c
+++ b/drivers/md/md-bitmap.c
@@ -1234,11 +1234,22 @@ static bitmap_counter_t *md_bitmap_get_counter(struct bitmap_counts *bitmap,
sector_t offset, sector_t *blocks,
int create);
+static void mddev_set_timeout(struct mddev *mddev, unsigned long timeout,
+ bool force)
+{
+ struct md_thread *thread = mddev->thread;
+
+ if (!thread)
+ return;
+
+ if (force || thread->timeout < MAX_SCHEDULE_TIMEOUT)
+ thread->timeout = timeout;
+}
+
/*
* bitmap daemon -- periodically wakes up to clean bits and flush pages
* out to disk
*/
-
void md_bitmap_daemon_work(struct mddev *mddev)
{
struct bitmap *bitmap;
@@ -1262,7 +1273,7 @@ void md_bitmap_daemon_work(struct mddev *mddev)
bitmap->daemon_lastrun = jiffies;
if (bitmap->allclean) {
- mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
+ mddev_set_timeout(mddev, MAX_SCHEDULE_TIMEOUT, true);
goto done;
}
bitmap->allclean = 1;
@@ -1359,8 +1370,7 @@ void md_bitmap_daemon_work(struct mddev *mddev)
done:
if (bitmap->allclean == 0)
- mddev->thread->timeout =
- mddev->bitmap_info.daemon_sleep;
+ mddev_set_timeout(mddev, mddev->bitmap_info.daemon_sleep, true);
mutex_unlock(&mddev->bitmap_info.mutex);
}
@@ -1821,8 +1831,7 @@ void md_bitmap_destroy(struct mddev *mddev)
mddev->bitmap = NULL; /* disconnect from the md device */
spin_unlock(&mddev->lock);
mutex_unlock(&mddev->bitmap_info.mutex);
- if (mddev->thread)
- mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
+ mddev_set_timeout(mddev, MAX_SCHEDULE_TIMEOUT, true);
md_bitmap_free(bitmap);
}
@@ -1965,7 +1974,7 @@ int md_bitmap_load(struct mddev *mddev)
/* Kick recovery in case any bits were set */
set_bit(MD_RECOVERY_NEEDED, &bitmap->mddev->recovery);
- mddev->thread->timeout = mddev->bitmap_info.daemon_sleep;
+ mddev_set_timeout(mddev, mddev->bitmap_info.daemon_sleep, true);
md_wakeup_thread(mddev->thread);
md_bitmap_update_sb(bitmap);
@@ -2470,17 +2479,11 @@ timeout_store(struct mddev *mddev, const char *buf, size_t len)
timeout = MAX_SCHEDULE_TIMEOUT-1;
if (timeout < 1)
timeout = 1;
- mddev->bitmap_info.daemon_sleep = timeout;
- if (mddev->thread) {
- /* if thread->timeout is MAX_SCHEDULE_TIMEOUT, then
- * the bitmap is all clean and we don't need to
- * adjust the timeout right now
- */
- if (mddev->thread->timeout < MAX_SCHEDULE_TIMEOUT)
- mddev->thread->timeout = timeout;
- }
+ mddev->bitmap_info.daemon_sleep = timeout;
+ mddev_set_timeout(mddev, timeout, false);
md_wakeup_thread(mddev->thread);
+
return len;
}
--
2.39.2
From: Yu Kuai <[email protected]>
Currently, there are many places that md_thread can be accessed without
protection, following are known scenarios that can cause
null-ptr-dereference or uaf:
1) sync_thread that is allocated and started from md_start_sync()
2) mddev->thread can be accessed directly from timeout_store() and
md_bitmap_daemon_work()
3) md_unregister_thread() from action_store().
Currently, a global spinlock 'pers_lock' is borrowed to protect
'mddev->thread' in some places, this problem can be fixed likewise,
however, use a global lock for all the cases is not good.
Fix this problem by protecting all md_thread with rcu.
Signed-off-by: Yu Kuai <[email protected]>
---
drivers/md/md-bitmap.c | 10 ++++--
drivers/md/md-cluster.c | 17 ++++++----
drivers/md/md-multipath.c | 4 +--
drivers/md/md.c | 69 ++++++++++++++++++---------------------
drivers/md/md.h | 8 ++---
drivers/md/raid1.c | 7 ++--
drivers/md/raid1.h | 2 +-
drivers/md/raid10.c | 20 +++++++-----
drivers/md/raid10.h | 2 +-
drivers/md/raid5-cache.c | 22 ++++++++-----
drivers/md/raid5.c | 15 +++++----
drivers/md/raid5.h | 2 +-
12 files changed, 97 insertions(+), 81 deletions(-)
diff --git a/drivers/md/md-bitmap.c b/drivers/md/md-bitmap.c
index 23522df41ca5..ad5a3456cd8a 100644
--- a/drivers/md/md-bitmap.c
+++ b/drivers/md/md-bitmap.c
@@ -1237,13 +1237,19 @@ static bitmap_counter_t *md_bitmap_get_counter(struct bitmap_counts *bitmap,
static void mddev_set_timeout(struct mddev *mddev, unsigned long timeout,
bool force)
{
- struct md_thread *thread = mddev->thread;
+ struct md_thread *thread;
+
+ rcu_read_lock();
+ thread = rcu_dereference(mddev->thread);
if (!thread)
- return;
+ goto out;
if (force || thread->timeout < MAX_SCHEDULE_TIMEOUT)
thread->timeout = timeout;
+
+out:
+ rcu_read_unlock();
}
/*
diff --git a/drivers/md/md-cluster.c b/drivers/md/md-cluster.c
index 10e0c5381d01..3d9fd74233df 100644
--- a/drivers/md/md-cluster.c
+++ b/drivers/md/md-cluster.c
@@ -75,14 +75,14 @@ struct md_cluster_info {
sector_t suspend_hi;
int suspend_from; /* the slot which broadcast suspend_lo/hi */
- struct md_thread *recovery_thread;
+ struct md_thread __rcu *recovery_thread;
unsigned long recovery_map;
/* communication loc resources */
struct dlm_lock_resource *ack_lockres;
struct dlm_lock_resource *message_lockres;
struct dlm_lock_resource *token_lockres;
struct dlm_lock_resource *no_new_dev_lockres;
- struct md_thread *recv_thread;
+ struct md_thread __rcu *recv_thread;
struct completion newdisk_completion;
wait_queue_head_t wait;
unsigned long state;
@@ -362,8 +362,8 @@ static void __recover_slot(struct mddev *mddev, int slot)
set_bit(slot, &cinfo->recovery_map);
if (!cinfo->recovery_thread) {
- cinfo->recovery_thread = md_register_thread(recover_bitmaps,
- mddev, "recover");
+ rcu_assign_pointer(cinfo->recovery_thread,
+ md_register_thread(recover_bitmaps, mddev, "recover"));
if (!cinfo->recovery_thread) {
pr_warn("md-cluster: Could not create recovery thread\n");
return;
@@ -526,11 +526,15 @@ static void process_add_new_disk(struct mddev *mddev, struct cluster_msg *cmsg)
static void process_metadata_update(struct mddev *mddev, struct cluster_msg *msg)
{
int got_lock = 0;
+ struct md_thread *thread;
struct md_cluster_info *cinfo = mddev->cluster_info;
mddev->good_device_nr = le32_to_cpu(msg->raid_slot);
dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
- wait_event(mddev->thread->wqueue,
+
+ /* daemaon thread must exist */
+ thread = rcu_dereference_protected(mddev->thread, true);
+ wait_event(thread->wqueue,
(got_lock = mddev_trylock(mddev)) ||
test_bit(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state));
md_reload_sb(mddev, mddev->good_device_nr);
@@ -889,7 +893,8 @@ static int join(struct mddev *mddev, int nodes)
}
/* Initiate the communication resources */
ret = -ENOMEM;
- cinfo->recv_thread = md_register_thread(recv_daemon, mddev, "cluster_recv");
+ rcu_assign_pointer(cinfo->recv_thread,
+ md_register_thread(recv_daemon, mddev, "cluster_recv"));
if (!cinfo->recv_thread) {
pr_err("md-cluster: cannot allocate memory for recv_thread!\n");
goto err;
diff --git a/drivers/md/md-multipath.c b/drivers/md/md-multipath.c
index 66edf5e72bd6..92c45be203d7 100644
--- a/drivers/md/md-multipath.c
+++ b/drivers/md/md-multipath.c
@@ -400,8 +400,8 @@ static int multipath_run (struct mddev *mddev)
if (ret)
goto out_free_conf;
- mddev->thread = md_register_thread(multipathd, mddev,
- "multipath");
+ rcu_assign_pointer(mddev->thread,
+ md_register_thread(multipathd, mddev, "multipath"));
if (!mddev->thread)
goto out_free_conf;
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 66a7a17510ce..50da6b671c49 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -70,11 +70,7 @@
#include "md-bitmap.h"
#include "md-cluster.h"
-/* pers_list is a list of registered personalities protected
- * by pers_lock.
- * pers_lock does extra service to protect accesses to
- * mddev->thread when the mutex cannot be held.
- */
+/* pers_list is a list of registered personalities protected by pers_lock. */
static LIST_HEAD(pers_list);
static DEFINE_SPINLOCK(pers_lock);
@@ -92,7 +88,7 @@ static int remove_and_add_spares(struct mddev *mddev,
struct md_rdev *this);
static void mddev_detach(struct mddev *mddev);
static void export_rdev(struct md_rdev *rdev);
-static void md_wakeup_thread_directly(struct md_thread *thread);
+static void md_wakeup_thread_directly(struct md_thread __rcu *thread);
/*
* Default number of read corrections we'll attempt on an rdev
@@ -442,8 +438,10 @@ static void md_submit_bio(struct bio *bio)
*/
void mddev_suspend(struct mddev *mddev)
{
- WARN_ON_ONCE(mddev->thread && current == mddev->thread->tsk);
- lockdep_assert_held(&mddev->reconfig_mutex);
+ struct md_thread *thread = rcu_dereference_protected(mddev->thread,
+ lockdep_is_held(&mddev->reconfig_mutex));
+
+ WARN_ON_ONCE(thread && current == thread->tsk);
if (mddev->suspended++)
return;
wake_up(&mddev->sb_wait);
@@ -810,13 +808,8 @@ void mddev_unlock(struct mddev *mddev)
md_free_rdev(mddev);
- /* As we've dropped the mutex we need a spinlock to
- * make sure the thread doesn't disappear
- */
- spin_lock(&pers_lock);
md_wakeup_thread(mddev->thread);
wake_up(&mddev->sb_wait);
- spin_unlock(&pers_lock);
}
EXPORT_SYMBOL_GPL(mddev_unlock);
@@ -7897,19 +7890,29 @@ static int md_thread(void *arg)
return 0;
}
-static void md_wakeup_thread_directly(struct md_thread *thread)
+static void md_wakeup_thread_directly(struct md_thread __rcu *thread)
{
- if (thread)
- wake_up_process(thread->tsk);
+ struct md_thread *t;
+
+ rcu_read_lock();
+ t = rcu_dereference(thread);
+ if (t)
+ wake_up_process(t->tsk);
+ rcu_read_unlock();
}
-void md_wakeup_thread(struct md_thread *thread)
+void md_wakeup_thread(struct md_thread __rcu *thread)
{
- if (thread) {
- pr_debug("md: waking up MD thread %s.\n", thread->tsk->comm);
- set_bit(THREAD_WAKEUP, &thread->flags);
- wake_up(&thread->wqueue);
+ struct md_thread *t;
+
+ rcu_read_lock();
+ t = rcu_dereference(thread);
+ if (t) {
+ pr_debug("md: waking up MD thread %s.\n", t->tsk->comm);
+ set_bit(THREAD_WAKEUP, &t->flags);
+ wake_up(&t->wqueue);
}
+ rcu_read_unlock();
}
EXPORT_SYMBOL(md_wakeup_thread);
@@ -7939,22 +7942,15 @@ struct md_thread *md_register_thread(void (*run) (struct md_thread *),
}
EXPORT_SYMBOL(md_register_thread);
-void md_unregister_thread(struct md_thread **threadp)
+void md_unregister_thread(struct md_thread __rcu **threadp)
{
- struct md_thread *thread;
+ struct md_thread *thread = rcu_dereference_protected(*threadp, true);
- /*
- * Locking ensures that mddev_unlock does not wake_up a
- * non-existent thread
- */
- spin_lock(&pers_lock);
- thread = *threadp;
- if (!thread) {
- spin_unlock(&pers_lock);
+ if (!thread)
return;
- }
- *threadp = NULL;
- spin_unlock(&pers_lock);
+
+ rcu_assign_pointer(*threadp, NULL);
+ synchronize_rcu();
pr_debug("interrupting MD-thread pid %d\n", task_pid_nr(thread->tsk));
kthread_stop(thread->tsk);
@@ -9220,9 +9216,8 @@ static void md_start_sync(struct work_struct *ws)
{
struct mddev *mddev = container_of(ws, struct mddev, del_work);
- mddev->sync_thread = md_register_thread(md_do_sync,
- mddev,
- "resync");
+ rcu_assign_pointer(mddev->sync_thread,
+ md_register_thread(md_do_sync, mddev, "resync"));
if (!mddev->sync_thread) {
pr_warn("%s: could not start resync thread...\n",
mdname(mddev));
diff --git a/drivers/md/md.h b/drivers/md/md.h
index 4d191db831da..15d82517572e 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -367,8 +367,8 @@ struct mddev {
int new_chunk_sectors;
int reshape_backwards;
- struct md_thread *thread; /* management thread */
- struct md_thread *sync_thread; /* doing resync or reconstruct */
+ struct md_thread __rcu *thread; /* management thread */
+ struct md_thread __rcu *sync_thread; /* doing resync or reconstruct */
/* 'last_sync_action' is initialized to "none". It is set when a
* sync operation (i.e "data-check", "requested-resync", "resync",
@@ -760,8 +760,8 @@ extern struct md_thread *md_register_thread(
void (*run)(struct md_thread *thread),
struct mddev *mddev,
const char *name);
-extern void md_unregister_thread(struct md_thread **threadp);
-extern void md_wakeup_thread(struct md_thread *thread);
+extern void md_unregister_thread(struct md_thread __rcu **threadp);
+extern void md_wakeup_thread(struct md_thread __rcu *thread);
extern void md_check_recovery(struct mddev *mddev);
extern void md_reap_sync_thread(struct mddev *mddev);
extern int mddev_init_writes_pending(struct mddev *mddev);
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index 68a9e2d9985b..2f1011ffdf09 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -3084,7 +3084,8 @@ static struct r1conf *setup_conf(struct mddev *mddev)
}
err = -ENOMEM;
- conf->thread = md_register_thread(raid1d, mddev, "raid1");
+ rcu_assign_pointer(conf->thread,
+ md_register_thread(raid1d, mddev, "raid1"));
if (!conf->thread)
goto abort;
@@ -3177,8 +3178,8 @@ static int raid1_run(struct mddev *mddev)
/*
* Ok, everything is just fine now
*/
- mddev->thread = conf->thread;
- conf->thread = NULL;
+ rcu_assign_pointer(mddev->thread, conf->thread);
+ rcu_assign_pointer(conf->thread, NULL);
mddev->private = conf;
set_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
diff --git a/drivers/md/raid1.h b/drivers/md/raid1.h
index ebb6788820e7..468f189da7a0 100644
--- a/drivers/md/raid1.h
+++ b/drivers/md/raid1.h
@@ -130,7 +130,7 @@ struct r1conf {
/* When taking over an array from a different personality, we store
* the new thread here until we fully activate the array.
*/
- struct md_thread *thread;
+ struct md_thread __rcu *thread;
/* Keep track of cluster resync window to send to other
* nodes.
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 4fcfcb350d2b..32fb4ff0acdb 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -982,6 +982,7 @@ static void lower_barrier(struct r10conf *conf)
static bool stop_waiting_barrier(struct r10conf *conf)
{
struct bio_list *bio_list = current->bio_list;
+ struct md_thread *thread;
/* barrier is dropped */
if (!conf->barrier)
@@ -997,12 +998,14 @@ static bool stop_waiting_barrier(struct r10conf *conf)
(!bio_list_empty(&bio_list[0]) || !bio_list_empty(&bio_list[1])))
return true;
+ /* daemon thread must exist while handling io */
+ thread = rcu_dereference_protected(conf->mddev->thread, true);
/*
* move on if io is issued from raid10d(), nr_pending is not released
* from original io(see handle_read_error()). All raise barrier is
* blocked until this io is done.
*/
- if (conf->mddev->thread->tsk == current) {
+ if (thread->tsk == current) {
WARN_ON_ONCE(atomic_read(&conf->nr_pending) == 0);
return true;
}
@@ -4107,7 +4110,8 @@ static struct r10conf *setup_conf(struct mddev *mddev)
atomic_set(&conf->nr_pending, 0);
err = -ENOMEM;
- conf->thread = md_register_thread(raid10d, mddev, "raid10");
+ rcu_assign_pointer(conf->thread,
+ md_register_thread(raid10d, mddev, "raid10"));
if (!conf->thread)
goto out;
@@ -4152,8 +4156,8 @@ static int raid10_run(struct mddev *mddev)
if (!conf)
goto out;
- mddev->thread = conf->thread;
- conf->thread = NULL;
+ rcu_assign_pointer(mddev->thread, conf->thread);
+ rcu_assign_pointer(conf->thread, NULL);
if (mddev_is_clustered(conf->mddev)) {
int fc, fo;
@@ -4296,8 +4300,8 @@ static int raid10_run(struct mddev *mddev)
clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
- mddev->sync_thread = md_register_thread(md_do_sync, mddev,
- "reshape");
+ rcu_assign_pointer(mddev->sync_thread,
+ md_register_thread(md_do_sync, mddev, "reshape"));
if (!mddev->sync_thread)
goto out_free_conf;
}
@@ -4698,8 +4702,8 @@ static int raid10_start_reshape(struct mddev *mddev)
set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
- mddev->sync_thread = md_register_thread(md_do_sync, mddev,
- "reshape");
+ rcu_assign_pointer(mddev->sync_thread,
+ md_register_thread(md_do_sync, mddev, "reshape"));
if (!mddev->sync_thread) {
ret = -EAGAIN;
goto abort;
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index 8c072ce0bc54..63e48b11b552 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -100,7 +100,7 @@ struct r10conf {
/* When taking over an array from a different personality, we store
* the new thread here until we fully activate the array.
*/
- struct md_thread *thread;
+ struct md_thread __rcu *thread;
/*
* Keep track of cluster resync window to send to other nodes.
diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c
index 46182b955aef..040f5d6e1298 100644
--- a/drivers/md/raid5-cache.c
+++ b/drivers/md/raid5-cache.c
@@ -120,7 +120,7 @@ struct r5l_log {
struct bio_set bs;
mempool_t meta_pool;
- struct md_thread *reclaim_thread;
+ struct md_thread __rcu *reclaim_thread;
unsigned long reclaim_target; /* number of space that need to be
* reclaimed. if it's 0, reclaim spaces
* used by io_units which are in
@@ -1576,17 +1576,18 @@ void r5l_wake_reclaim(struct r5l_log *log, sector_t space)
void r5l_quiesce(struct r5l_log *log, int quiesce)
{
- struct mddev *mddev;
+ struct mddev *mddev = log->rdev->mddev;
+ struct md_thread *thread = rcu_dereference_protected(
+ log->reclaim_thread, lockdep_is_held(&mddev->reconfig_mutex));
if (quiesce) {
/* make sure r5l_write_super_and_discard_space exits */
- mddev = log->rdev->mddev;
wake_up(&mddev->sb_wait);
- kthread_park(log->reclaim_thread->tsk);
+ kthread_park(thread->tsk);
r5l_wake_reclaim(log, MaxSector);
r5l_do_reclaim(log);
} else
- kthread_unpark(log->reclaim_thread->tsk);
+ kthread_unpark(thread->tsk);
}
bool r5l_log_disk_error(struct r5conf *conf)
@@ -3063,6 +3064,7 @@ void r5c_update_on_rdev_error(struct mddev *mddev, struct md_rdev *rdev)
int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
{
struct r5l_log *log;
+ struct md_thread *thread;
int ret;
pr_debug("md/raid:%s: using device %pg as journal\n",
@@ -3121,11 +3123,13 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
spin_lock_init(&log->tree_lock);
INIT_RADIX_TREE(&log->big_stripe_tree, GFP_NOWAIT | __GFP_NOWARN);
- log->reclaim_thread = md_register_thread(r5l_reclaim_thread,
- log->rdev->mddev, "reclaim");
- if (!log->reclaim_thread)
+ thread = md_register_thread(r5l_reclaim_thread, log->rdev->mddev,
+ "reclaim");
+ if (!thread)
goto reclaim_thread;
- log->reclaim_thread->timeout = R5C_RECLAIM_WAKEUP_INTERVAL;
+
+ thread->timeout = R5C_RECLAIM_WAKEUP_INTERVAL;
+ rcu_assign_pointer(log->reclaim_thread, thread);
init_waitqueue_head(&log->iounit_wait);
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 01c55f24ab09..7e2bbcfef325 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -7731,7 +7731,8 @@ static struct r5conf *setup_conf(struct mddev *mddev)
}
sprintf(pers_name, "raid%d", mddev->new_level);
- conf->thread = md_register_thread(raid5d, mddev, pers_name);
+ rcu_assign_pointer(conf->thread,
+ md_register_thread(raid5d, mddev, pers_name));
if (!conf->thread) {
pr_warn("md/raid:%s: couldn't allocate thread.\n",
mdname(mddev));
@@ -7954,8 +7955,8 @@ static int raid5_run(struct mddev *mddev)
}
conf->min_offset_diff = min_offset_diff;
- mddev->thread = conf->thread;
- conf->thread = NULL;
+ rcu_assign_pointer(mddev->thread, conf->thread);
+ rcu_assign_pointer(conf->thread, NULL);
mddev->private = conf;
for (i = 0; i < conf->raid_disks && conf->previous_raid_disks;
@@ -8052,8 +8053,8 @@ static int raid5_run(struct mddev *mddev)
clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
- mddev->sync_thread = md_register_thread(md_do_sync, mddev,
- "reshape");
+ rcu_assign_pointer(mddev->sync_thread,
+ md_register_thread(md_do_sync, mddev, "reshape"));
if (!mddev->sync_thread)
goto abort;
}
@@ -8631,8 +8632,8 @@ static int raid5_start_reshape(struct mddev *mddev)
clear_bit(MD_RECOVERY_DONE, &mddev->recovery);
set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
- mddev->sync_thread = md_register_thread(md_do_sync, mddev,
- "reshape");
+ rcu_assign_pointer(mddev->sync_thread,
+ md_register_thread(md_do_sync, mddev, "reshape"));
if (!mddev->sync_thread) {
mddev->recovery = 0;
spin_lock_irq(&conf->device_lock);
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index e873938a6125..f19707189a7b 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -679,7 +679,7 @@ struct r5conf {
/* When taking over an array from a different personality, we store
* the new thread here until we fully activate the array.
*/
- struct md_thread *thread;
+ struct md_thread __rcu *thread;
struct list_head temp_inactive_list[NR_STRIPE_HASH_LOCKS];
struct r5worker_group *worker_groups;
int group_cnt;
--
2.39.2
On Mon, May 22, 2023 at 7:13 PM Yu Kuai <[email protected]> wrote:
>
> From: Yu Kuai <[email protected]>
>
> Changes in v8:
> - fix that another unrelated patch is merged into this patchset
> - update commit message for patch 5
> - rebase on the top of another patch from me: [PATCH v3] md: fix duplicate filename for rdev
Applied to md-next.
Thanks,
Song