| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20250307134933.1033872-8-kent.overstreet@linux.dev>
Date: Fri, 7 Mar 2025 08:49:31 -0500
From: Kent Overstreet <kent.overstreet@...ux.dev>
To: linux-bcachefs@...r.kernel.org,
linux-kernel@...r.kernel.org
Cc: Kent Overstreet <kent.overstreet@...ux.dev>
Subject: [PATCH 7/7] bcachefs: Implement freeze/thaw
This fills out our blk_holders_ops with freeze and thaw callbacks, for
shutting down IO (generally during a system suspend).
This is implemented completely differently as on other filesystems
since we have a low level synchronization object which conveniently
works well for us - bch_dev.io_ref, normally used for guarding against a
device being offlined while in use.
bch2_dev_get_ioref() now checks if a freeze is in progress if it fails
to get ca->io_ref, and sleeps until complete and ca->io_ref is alive.
We also need a bit of synchronization for freeze/suspend vs. device
online/offline, done with the new bch_dev.io_ref_statechange_lock.
Signed-off-by: Kent Overstreet <kent.overstreet@...ux.dev>
---
fs/bcachefs/bcachefs.h | 3 ++
fs/bcachefs/journal_io.c | 2 +-
fs/bcachefs/sb-members.c | 49 ++++++++++++++++++++++
fs/bcachefs/sb-members.h | 20 +--------
fs/bcachefs/super.c | 87 +++++++++++++++++++++++++++++++++++++---
5 files changed, 136 insertions(+), 25 deletions(-)
diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h
index d2c3f59a668f..d03aa62907ad 100644
--- a/fs/bcachefs/bcachefs.h
+++ b/fs/bcachefs/bcachefs.h
@@ -526,6 +526,9 @@ struct bch_dev {
struct completion ref_completion;
struct percpu_ref io_ref;
struct completion io_ref_completion;
+ struct mutex io_ref_statechange_lock;
+ unsigned frozen;
+ wait_queue_head_t frozen_wait;
struct bch_fs *fs;
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index a510755a8364..6979fef5c128 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -1769,7 +1769,7 @@ static CLOSURE_CALLBACK(journal_write_submit)
struct bch_dev *ca = bch2_dev_get_ioref(c, ptr->dev, WRITE);
if (!ca) {
/* XXX: fix this */
- bch_err(c, "missing device for journal write\n");
+ bch_err(c, "missing device for journal write");
continue;
}
diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c
index 116131f95815..2363367cb32d 100644
--- a/fs/bcachefs/sb-members.c
+++ b/fs/bcachefs/sb-members.c
@@ -9,6 +9,55 @@
#include "sb-members.h"
#include "super-io.h"
+/*
+ * Use of bch2_dev_get_ioref() is subject to deadlocks if used incorrectly, and
+ * we cannot write asserts for correct usage, so: pay attention, because this is
+ * where we implement freeze.
+ *
+ * Waiting on an outstanding freeze to complete will indirectly wait on all
+ * other outstanding io_refs to be released. That means:
+ *
+ * - Don't use bch2_dev_get_ioref() if you already have an io_ref, use
+ * percpu_ref_get(). Since dev_get_ioref() has tryget() semantics, that's what
+ * you should be doing anyways.
+ *
+ * - All io_refs must be released without blocking on locks that might be held
+ * while calling dev_get_ioref(). This is easy to obey since we generally
+ * release io_refs from endio functions.
+ *
+ */
+struct bch_dev *bch2_dev_get_ioref(struct bch_fs *c, unsigned dev, int rw)
+{
+ might_sleep();
+again:
+ rcu_read_lock();
+ struct bch_dev *ca = bch2_dev_rcu(c, dev);
+ if (likely(ca)) {
+ if (unlikely(!percpu_ref_tryget(&ca->io_ref))) {
+ smp_mb();
+ if (ca->frozen) {
+ bch2_dev_get(ca);
+ rcu_read_unlock();
+
+ wait_event(ca->frozen_wait, !ca->frozen);
+ bch2_dev_put(ca);
+ goto again;
+ }
+ ca = NULL;
+ }
+ }
+ rcu_read_unlock();
+
+ if (ca &&
+ (ca->mi.state == BCH_MEMBER_STATE_rw ||
+ (ca->mi.state == BCH_MEMBER_STATE_ro && rw == READ)))
+ return ca;
+
+ if (ca)
+ percpu_ref_put(&ca->io_ref);
+ return NULL;
+}
+
void bch2_dev_missing(struct bch_fs *c, unsigned dev)
{
if (dev != BCH_SB_MEMBER_INVALID)
diff --git a/fs/bcachefs/sb-members.h b/fs/bcachefs/sb-members.h
index df91b02ce575..b3359ee63b0e 100644
--- a/fs/bcachefs/sb-members.h
+++ b/fs/bcachefs/sb-members.h
@@ -281,25 +281,7 @@ static inline struct bch_dev *bch2_dev_iterate(struct bch_fs *c, struct bch_dev
return bch2_dev_tryget(c, dev_idx);
}
-static inline struct bch_dev *bch2_dev_get_ioref(struct bch_fs *c, unsigned dev, int rw)
-{
- might_sleep();
-
- rcu_read_lock();
- struct bch_dev *ca = bch2_dev_rcu(c, dev);
- if (ca && !percpu_ref_tryget(&ca->io_ref))
- ca = NULL;
- rcu_read_unlock();
-
- if (ca &&
- (ca->mi.state == BCH_MEMBER_STATE_rw ||
- (ca->mi.state == BCH_MEMBER_STATE_ro && rw == READ)))
- return ca;
-
- if (ca)
- percpu_ref_put(&ca->io_ref);
- return NULL;
-}
+struct bch_dev *bch2_dev_get_ioref(struct bch_fs *, unsigned, int);
/* XXX kill, move to struct bch_fs */
static inline struct bch_devs_mask bch2_online_devs(struct bch_fs *c)
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 05a2dc5ef513..dfdeab7d847c 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -1236,6 +1236,22 @@ static void bch2_dev_free(struct bch_dev *ca)
kobject_put(&ca->kobj);
}
+static void bch2_dev_io_ref_stop(struct bch_dev *ca)
+{
+ lockdep_assert_held(&ca->io_ref_statechange_lock);
+
+ reinit_completion(&ca->io_ref_completion);
+ percpu_ref_kill(&ca->io_ref);
+ wait_for_completion(&ca->io_ref_completion);
+}
+
+static void bch2_dev_io_ref_start(struct bch_dev *ca)
+{
+ lockdep_assert_held(&ca->io_ref_statechange_lock);
+
+ percpu_ref_reinit(&ca->io_ref);
+}
+
static void __bch2_dev_offline(struct bch_fs *c, struct bch_dev *ca)
{
@@ -1246,13 +1262,14 @@ static void __bch2_dev_offline(struct bch_fs *c, struct bch_dev *ca)
__bch2_dev_read_only(c, ca);
- reinit_completion(&ca->io_ref_completion);
- percpu_ref_kill(&ca->io_ref);
- wait_for_completion(&ca->io_ref_completion);
-
bch2_dev_unlink(ca);
+ mutex_lock(&ca->io_ref_statechange_lock);
+ bch2_dev_io_ref_stop(ca);
+
bch2_free_super(&ca->disk_sb);
+ mutex_unlock(&ca->io_ref_statechange_lock);
+
bch2_dev_journal_exit(ca);
}
@@ -1334,6 +1351,8 @@ static struct bch_dev *__bch2_dev_alloc(struct bch_fs *c,
kobject_init(&ca->kobj, &bch2_dev_ktype);
init_completion(&ca->ref_completion);
init_completion(&ca->io_ref_completion);
+ mutex_init(&ca->io_ref_statechange_lock);
+ init_waitqueue_head(&ca->frozen_wait);
INIT_WORK(&ca->io_error_work, bch2_io_error_work);
@@ -1428,6 +1447,8 @@ static int __bch2_dev_attach_bdev(struct bch_dev *ca, struct bch_sb_handle *sb)
if (ret)
return ret;
+ mutex_lock(&ca->io_ref_statechange_lock);
+
/* Commit: */
ca->disk_sb = *sb;
memset(sb, 0, sizeof(*sb));
@@ -1441,7 +1462,9 @@ static int __bch2_dev_attach_bdev(struct bch_dev *ca, struct bch_sb_handle *sb)
ca->dev = ca->disk_sb.bdev->bd_dev;
- percpu_ref_reinit(&ca->io_ref);
+ if (!ca->frozen)
+ bch2_dev_io_ref_start(ca);
+ mutex_unlock(&ca->io_ref_statechange_lock);
return 0;
}
@@ -2115,9 +2138,63 @@ static void bch2_fs_bdev_sync(struct block_device *bdev)
bch2_ro_ref_put(c);
}
+static int bch2_fs_bdev_freeze(struct block_device *bdev)
+{
+ int ret = -EINVAL;
+ struct bch_fs *c = bdev_get_fs(bdev);
+ if (!c)
+ return ret;
+
+ struct bch_dev *ca = bdev_to_bch_dev(c, bdev);
+ if (!ca)
+ goto err;
+
+ mutex_lock(&ca->io_ref_statechange_lock);
+ ca->frozen++;
+ smp_mb();
+ bch2_dev_io_ref_stop(ca);
+ mutex_unlock(&ca->io_ref_statechange_lock);
+
+ ret = sync_blockdev(bdev);
+
+ bch2_dev_put(ca);
+err:
+ bch2_ro_ref_put(c);
+ return ret;
+}
+
+static int bch2_fs_bdev_thaw(struct block_device *bdev)
+{
+ int ret = -EINVAL;
+ struct bch_fs *c = bdev_get_fs(bdev);
+ if (!c)
+ return ret;
+
+ struct bch_dev *ca = bdev_to_bch_dev(c, bdev);
+ if (!ca)
+ goto err;
+
+ mutex_lock(&ca->io_ref_statechange_lock);
+ if (ca->disk_sb.bdev &&
+ ca->frozen == 1)
+ bch2_dev_io_ref_start(ca);
+ --ca->frozen;
+ wake_up(&ca->frozen_wait);
+ mutex_unlock(&ca->io_ref_statechange_lock);
+
+ ret = 0;
+
+ bch2_dev_put(ca);
+err:
+ bch2_ro_ref_put(c);
+ return ret;
+}
+
const struct blk_holder_ops bch2_sb_handle_bdev_ops = {
.mark_dead = bch2_fs_bdev_mark_dead,
.sync = bch2_fs_bdev_sync,
+ .freeze = bch2_fs_bdev_freeze,
+ .thaw = bch2_fs_bdev_thaw,
};
/* Filesystem open: */
--
2.47.2
Powered by blists - more mailing lists