Skip to contentSkip to conversation

Fixes for FS-Cache and CacheFiles

PATCH In review David Howells 2 participants 36 messages Nov 19, 2009 – Nov 20, 2009

Subsystems (8)
lore
Share & tools
lore

David Howells

author maintainer
#0 lore raw

cover Fixes for FS-Cache and CacheFiles

The associated patches provide three sets of changes:

(1) More ways to get debugging information from the slow-work, FS-Cache and CacheFiles facilities.

(2) Fixes for all facilities.

(3) Some extensions for slow-work for users other than FS-Cache/CacheFiles that aren't included in this series, but upon which other patches have been built.

The patches can also be found in:

http://people.redhat.com/~dhowells/fscache/patches/fscache-fixes.tar.bz2

The bugs fixed were found by running a simple stress tester:

http://people.redhat.com/~dhowells/slurp.c

Build it and point it at a bunch of NFS (or AFS or whatever) directory trees, and it forks and execs a copy of itself to handle each directory that it is given. Whilst scanning a directory, for each subdir of that directory a new copy forked and exec'd, and then each file in that directory is read from end to end.

This results in a large number of processes reading in parallel a large number of files. If the data set is large enough, it applies pressure to both RAM space and cache space, and can cause OOMs to occur, and can cause cachefilesd to run in parallel trying to cull the cache as well.

With these patches applied, you can see more counters in /proc/fs/fscache/stats and new sources of information exist too:

(*) /proc/slow_work_rq

The items currently being processed by or currently queued for processing by the slow-work facility.

(*) /proc/fs/fscache/objects

A list of current fscache objects and their states. This can be configured by loading a key into your session keyring before viewing the file, for instance:

keyctl add user fscache:objlist KB @s

This is described in Documentation/filesystems/caching/fscache.txt

David
---

David Howells (25):
CacheFiles: Don't log lookup/create failing with ENOBUFS
CacheFiles: Catch an overly long wait for an old active object
CacheFiles: Better showing of debugging information in active object problems
CacheFiles: Mark parent directory locks as I_MUTEX_PARENT to keep lockdep happy
CacheFiles: Handle truncate unlocking the page we're reading
CacheFiles: Don't write a full page if there's only a partial page to cache
FS-Cache: Actually requeue an object when requested
FS-Cache: Start processing an object's operations on that object's death
FS-Cache: Make sure FSCACHE_COOKIE_LOOKING_UP cleared on lookup failure
FS-Cache: Add a retirement stat counter
FS-Cache: Handle pages pending storage that get evicted under OOM conditions
FS-Cache: Handle read request vs lookup, creation or other cache failure
FS-Cache: Don't delete pending pages from the page-store tracking tree
FS-Cache: Fix lock misorder in fscache_write_op()
FS-Cache: The object-available state can't rely on the cookie to be available
FS-Cache: Permit cache retrieval ops to be interrupted in the initial wait phase
FS-Cache: Use radix tree preload correctly in tracking of pages to be stored
FS-Cache: Clear netfs pointers in cookie after detaching object, not before
FS-Cache: Add counters for entry/exit to/from cache operation functions
FS-Cache: Allow the current state of all objects to be dumped
FS-Cache: Annotate slow-work runqueue proc lines for FS-Cache work items
SLOW_WORK: Allow a requeueable work item to sleep till the thread is needed
SLOW_WORK: Allow the owner of a work item to determine if it is queued or not
SLOW_WORK: Allow the work items to be viewed through a /proc file
SLOW_WORK: Wait for outstanding work items belonging to a module to clear

Jens Axboe (3):
SLOW_WORK: Add delayed_slow_work support
SLOW_WORK: Add support for cancellation of slow work
SLOW_WORK: Make slow_work_ops ->get_ref/->put_ref optional

 Documentation/filesystems/caching/fscache.txt   |  110 +++++
 Documentation/filesystems/caching/netfs-api.txt |   21 +
 Documentation/slow-work.txt                     |  160 +++++++
 fs/9p/cache.c                                   |   14 -
 fs/afs/file.c                                   |   15 -
 fs/cachefiles/interface.c                       |   32 +
 fs/cachefiles/namei.c                           |  187 +++++++--
 fs/cachefiles/rdwr.c                            |  128 +++++-
 fs/fscache/Kconfig                              |    7 
 fs/fscache/Makefile                             |    1 
 fs/fscache/cache.c                              |    5 
 fs/fscache/cookie.c                             |   26 +
 fs/fscache/internal.h                           |   55 +++
 fs/fscache/main.c                               |    6 
 fs/fscache/object-list.c                        |  432 ++++++++++++++++++++
 fs/fscache/object.c                             |  104 ++++-
 fs/fscache/operation.c                          |  120 ++++--
 fs/fscache/page.c                               |  273 ++++++++++---
 fs/fscache/proc.c                               |   13 +
 fs/fscache/stats.c                              |   94 ++++
 fs/gfs2/main.c                                  |    4 
 fs/gfs2/recovery.c                              |    1 
 fs/nfs/fscache.c                                |   10 
 include/linux/fscache-cache.h                   |   40 ++
 include/linux/fscache.h                         |   27 +
 include/linux/slow-work.h                       |   72 +++
 init/Kconfig                                    |   10 
 kernel/Makefile                                 |    1 
 kernel/slow-work-proc.c                         |  227 +++++++++++
 kernel/slow-work.c                              |  494 +++++++++++++++++++++--
 kernel/slow-work.h                              |   72 +++
 lib/radix-tree.c                                |    5 
 32 files changed, 2502 insertions(+), 264 deletions(-)
 create mode 100644 fs/fscache/object-list.c
 create mode 100644 kernel/slow-work-proc.c
 create mode 100644 kernel/slow-work.h
author maintainer
#1 lore raw

PATCH 1/28 SLOW_WORK: Wait for outstanding work items belonging to a module to clear

Wait for outstanding slow work items belonging to a module to clear when unregistering that module as a user of the facility. This prevents the put_ref code of a work item from being taken away before it returns.

---

 Documentation/slow-work.txt |   13 +++-
 fs/fscache/main.c           |    6 +-
 fs/fscache/object.c         |    1 
 fs/fscache/operation.c      |    1 
 fs/gfs2/main.c              |    4 +
 fs/gfs2/recovery.c          |    1 
 include/linux/slow-work.h   |    8 ++-
 kernel/slow-work.c          |  132 +++++++++++++++++++++++++++++++++++++++++--
 8 files changed, 150 insertions(+), 16 deletions(-)
Documentation/slow-work.txt +10 -3
index ebc50f8..f12fda3 100644
@@ -64,9 +64,11 @@ USING SLOW WORK ITEMS
 Firstly, a module or subsystem wanting to make use of slow work items must
 register its interest:
 
-	 int ret = slow_work_register_user();
+	 int ret = slow_work_register_user(struct module *module);
 
-This will return 0 if successful, or a -ve error upon failure.
+This will return 0 if successful, or a -ve error upon failure.  The module
+pointer should be the module interested in using this facility (almost
+certainly THIS_MODULE).
 
 
 Slow work items may then be set up by:
@@ -110,7 +112,12 @@ operation.  When all a module's slow work items have been processed, and the
 module has no further interest in the facility, it should unregister its
 interest:
 
-	slow_work_unregister_user();
+	slow_work_unregister_user(struct module *module);
+
+The module pointer is used to wait for all outstanding work items for that
+module before completing the unregistration.  This prevents the put_ref() code
+from being taken away before it completes.  module should almost certainly be
+THIS_MODULE.
 
 
 ===============
fs/fscache/main.c +3 -3
index 4de41b5..add6bdb 100644
@@ -48,7 +48,7 @@ static int __init fscache_init(void)
 {
 	int ret;
 
-	ret = slow_work_register_user();
+	ret = slow_work_register_user(THIS_MODULE);
 	if (ret < 0)
 		goto error_slow_work;
 
@@ -80,7 +80,7 @@ error_kobj:
 error_cookie_jar:
 	fscache_proc_cleanup();
 error_proc:
-	slow_work_unregister_user();
+	slow_work_unregister_user(THIS_MODULE);
 error_slow_work:
 	return ret;
 }
@@ -97,7 +97,7 @@ static void __exit fscache_exit(void)
 	kobject_put(fscache_root);
 	kmem_cache_destroy(fscache_cookie_jar);
 	fscache_proc_cleanup();
-	slow_work_unregister_user();
+	slow_work_unregister_user(THIS_MODULE);
 	printk(KERN_NOTICE "FS-Cache: Unloaded\n");
 }
 
fs/fscache/object.c +1 -0
index 392a41b..d236eb1 100644
@@ -45,6 +45,7 @@ static void fscache_enqueue_dependents(struct fscache_object *);
 static void fscache_dequeue_object(struct fscache_object *);
 
 const struct slow_work_ops fscache_object_slow_work_ops = {
+	.owner		= THIS_MODULE,
 	.get_ref	= fscache_object_slow_work_get_ref,
 	.put_ref	= fscache_object_slow_work_put_ref,
 	.execute	= fscache_object_slow_work_execute,
fs/fscache/operation.c +1 -0
index e7f8d53..f1a2857 100644
@@ -453,6 +453,7 @@ static void fscache_op_execute(struct slow_work *work)
 }
 
 const struct slow_work_ops fscache_op_slow_work_ops = {
+	.owner		= THIS_MODULE,
 	.get_ref	= fscache_op_get_ref,
 	.put_ref	= fscache_op_put_ref,
 	.execute	= fscache_op_execute,
fs/gfs2/main.c +2 -2
index eacd78a..5b31f77 100644
@@ -114,7 +114,7 @@ static int __init init_gfs2_fs(void)
 	if (error)
 		goto fail_unregister;
 
-	error = slow_work_register_user();
+	error = slow_work_register_user(THIS_MODULE);
 	if (error)
 		goto fail_slow;
 
@@ -163,7 +163,7 @@ static void __exit exit_gfs2_fs(void)
 	gfs2_unregister_debugfs();
 	unregister_filesystem(&gfs2_fs_type);
 	unregister_filesystem(&gfs2meta_fs_type);
-	slow_work_unregister_user();
+	slow_work_unregister_user(THIS_MODULE);
 
 	kmem_cache_destroy(gfs2_quotad_cachep);
 	kmem_cache_destroy(gfs2_rgrpd_cachep);
fs/gfs2/recovery.c +1 -0
index 59d2695..b2bb779 100644
@@ -593,6 +593,7 @@ fail:
 }
 
 struct slow_work_ops gfs2_recover_ops = {
+	.owner	 = THIS_MODULE,
 	.get_ref = gfs2_recover_get_ref,
 	.put_ref = gfs2_recover_put_ref,
 	.execute = gfs2_recover_work,
include/linux/slow-work.h +6 -2
index b65c888..9adb2b3 100644
@@ -24,6 +24,9 @@ struct slow_work;
  * The operations used to support slow work items
  */
 struct slow_work_ops {
+	/* owner */
+	struct module *owner;
+
 	/* get a ref on a work item
 	 * - return 0 if successful, -ve if not
 	 */
@@ -42,6 +45,7 @@ struct slow_work_ops {
  *   queued
  */
 struct slow_work {
+	struct module		*owner;	/* the owning module */
 	unsigned long		flags;
 #define SLOW_WORK_PENDING	0	/* item pending (further) execution */
 #define SLOW_WORK_EXECUTING	1	/* item currently executing */
@@ -84,8 +88,8 @@ static inline void vslow_work_init(struct slow_work *work,
 }
 
 extern int slow_work_enqueue(struct slow_work *work);
-extern int slow_work_register_user(void);
-extern void slow_work_unregister_user(void);
+extern int slow_work_register_user(struct module *owner);
+extern void slow_work_unregister_user(struct module *owner);
 
 #ifdef CONFIG_SYSCTL
 extern ctl_table slow_work_sysctls[];
kernel/slow-work.c +126 -6

230 lines not on this page · raw message

#2 lore raw

Re: [PATCH 01/28] SLOW_WORK: Wait for outstanding work items belonging to a module to clear

Hi,

Quoted 7 lines · David Howells

Wait for outstanding slow work items belonging to a module to clear when unregistering that module as a user of the facility. This prevents the put_ref code of a work item from being taken away before it returns.

---

Quoted 10 lines
 Documentation/slow-work.txt |   13 +++-
 fs/fscache/main.c           |    6 +-
 fs/fscache/object.c         |    1 
 fs/fscache/operation.c      |    1 
 fs/gfs2/main.c              |    4 +
 fs/gfs2/recovery.c          |    1 
 include/linux/slow-work.h   |    8 ++-
 kernel/slow-work.c          |  132 +++++++++++++++++++++++++++++++++++++++++--
 8 files changed, 150 insertions(+), 16 deletions(-)

GFS2 bits Acked-By: Steven Whitehouse <[email protected]>

Steve.

author maintainer
#3 lore raw

PATCH 2/28 SLOW_WORK: Make slow_work_ops ->get_ref/->put_ref optional

In reply to David Howells

From: Jens Axboe <[email protected]>

Make the ability for the slow-work facility to take references on a work item optional as not everyone requires this.

Even the internal slow-work stubs them out, so those can be got rid of too.

---

 Documentation/slow-work.txt |    2 +-
 kernel/slow-work.c          |   36 ++++++++++++++++--------------------
 2 files changed, 17 insertions(+), 21 deletions(-)
Documentation/slow-work.txt +1 -1
index f12fda3..c655c51 100644
@@ -125,7 +125,7 @@ ITEM OPERATIONS
 ===============
 
 Each work item requires a table of operations of type struct slow_work_ops.
-All members are required:
+Only ->execute() is required, getting and putting of a reference are optional.
 
  (*) Get a reference on an item:
 
kernel/slow-work.c +16 -20
index dd08f37..fccf421 100644
@@ -145,6 +145,20 @@ static DECLARE_COMPLETION(slow_work_last_thread_exited);
 static int slow_work_user_count;
 static DEFINE_MUTEX(slow_work_user_lock);
 
+static inline int slow_work_get_ref(struct slow_work *work)
+{
+	if (work->ops->get_ref)
+		return work->ops->get_ref(work);
+
+	return 0;
+}
+
+static inline void slow_work_put_ref(struct slow_work *work)
+{
+	if (work->ops->put_ref)
+		work->ops->put_ref(work);
+}
+
 /*
  * Calculate the maximum number of active threads in the pool that are
  * permitted to process very slow work items.
@@ -248,7 +262,7 @@ static bool slow_work_execute(int id)
 	}
 
 	/* sort out the race between module unloading and put_ref() */
-	work->ops->put_ref(work);
+	slow_work_put_ref(work);
 
 #ifdef CONFIG_MODULES
 	module = slow_work_thread_processing[id];
@@ -309,7 +323,6 @@ int slow_work_enqueue(struct slow_work *work)
 	BUG_ON(slow_work_user_count <= 0);
 	BUG_ON(!work);
 	BUG_ON(!work->ops);
-	BUG_ON(!work->ops->get_ref);
 
 	/* when honouring an enqueue request, we only promise that we will run
 	 * the work function in the future; we do not promise to run it once
@@ -339,7 +352,7 @@ int slow_work_enqueue(struct slow_work *work)
 		if (test_bit(SLOW_WORK_EXECUTING, &work->flags)) {
 			set_bit(SLOW_WORK_ENQ_DEFERRED, &work->flags);
 		} else {
-			if (work->ops->get_ref(work) < 0)
+			if (slow_work_get_ref(work) < 0)
 				goto cant_get_ref;
 			if (test_bit(SLOW_WORK_VERY_SLOW, &work->flags))
 				list_add_tail(&work->link, &vslow_work_queue);
@@ -480,21 +493,6 @@ static void slow_work_cull_timeout(unsigned long data)
 }
 
 /*
- * Get a reference on slow work thread starter
- */
-static int slow_work_new_thread_get_ref(struct slow_work *work)
-{
-	return 0;
-}
-
-/*
- * Drop a reference on slow work thread starter
- */
-static void slow_work_new_thread_put_ref(struct slow_work *work)
-{
-}
-
-/*
  * Start a new slow work thread
  */
 static void slow_work_new_thread_execute(struct slow_work *work)
@@ -529,8 +527,6 @@ static void slow_work_new_thread_execute(struct slow_work *work)
 
 static const struct slow_work_ops slow_work_new_thread_ops = {
 	.owner		= THIS_MODULE,
-	.get_ref	= slow_work_new_thread_get_ref,
-	.put_ref	= slow_work_new_thread_put_ref,
 	.execute	= slow_work_new_thread_execute,
 };
 
author maintainer
#4 lore raw

PATCH 3/28 SLOW_WORK: Add support for cancellation of slow work

In reply to David Howells

From: Jens Axboe <[email protected]>

Add support for cancellation of queued slow work and delayed slow work items. The cancellation functions will wait for items that are pending or undergoing execution to be discarded by the slow work facility.

Attempting to enqueue work that is in the process of being cancelled will result in ECANCELED.

---

 Documentation/slow-work.txt |   12 ++++++
 include/linux/slow-work.h   |    2 +
 kernel/slow-work.c          |   81 ++++++++++++++++++++++++++++++++++++++++---
 3 files changed, 88 insertions(+), 7 deletions(-)
Documentation/slow-work.txt +11 -1
index c655c51..2e384bd 100644
@@ -108,7 +108,17 @@ on the item, 0 otherwise.
 
 
 The items are reference counted, so there ought to be no need for a flush
-operation.  When all a module's slow work items have been processed, and the
+operation.  But as the reference counting is optional, means to cancel
+existing work items are also included:
+
+	cancel_slow_work(&myitem);
+
+can be used to cancel pending work.  The above cancel function waits for
+existing work to have been executed (or prevent execution of them, depending
+on timing).
+
+
+When all a module's slow work items have been processed, and the
 module has no further interest in the facility, it should unregister its
 interest:
 
include/linux/slow-work.h +2 -0
index 9adb2b3..eef2018 100644
@@ -51,6 +51,7 @@ struct slow_work {
 #define SLOW_WORK_EXECUTING	1	/* item currently executing */
 #define SLOW_WORK_ENQ_DEFERRED	2	/* item enqueue deferred */
 #define SLOW_WORK_VERY_SLOW	3	/* item is very slow */
+#define SLOW_WORK_CANCELLING	4	/* item is being cancelled, don't enqueue */
 	const struct slow_work_ops *ops; /* operations table for this item */
 	struct list_head	link;	/* link in queue */
 };
@@ -88,6 +89,7 @@ static inline void vslow_work_init(struct slow_work *work,
 }
 
 extern int slow_work_enqueue(struct slow_work *work);
+extern void slow_work_cancel(struct slow_work *work);
 extern int slow_work_register_user(struct module *owner);
 extern void slow_work_unregister_user(struct module *owner);
 
kernel/slow-work.c +75 -6

124 lines not on this page · raw message

author maintainer
#5 lore raw

PATCH 4/28 SLOW_WORK: Add delayed_slow_work support

In reply to David Howells

From: Jens Axboe <[email protected]>

This adds support for starting slow work with a delay, similar to the functionality we have for workqueues.

---

 Documentation/slow-work.txt |   16 +++++
 include/linux/slow-work.h   |   29 ++++++++++
 kernel/slow-work.c          |  129 ++++++++++++++++++++++++++++++++++++++++++-
 3 files changed, 171 insertions(+), 3 deletions(-)
Documentation/slow-work.txt +15 -1
index 2e384bd..a9d1b0f 100644
@@ -41,6 +41,13 @@ expand files, provided the time taken to do so isn't too long.
 Operations of both types may sleep during execution, thus tying up the thread
 loaned to it.
 
+A further class of work item is available, based on the slow work item class:
+
+ (*) Delayed slow work items.
+
+These are slow work items that have a timer to defer queueing of the item for
+a while.
+
 
 THREAD-TO-CLASS ALLOCATION
 --------------------------
@@ -95,6 +102,10 @@ Slow work items may then be set up by:
 
      or:
 
+	delayed_slow_work_init(&myitem, &myitem_ops);
+
+     or:
+
 	vslow_work_init(&myitem, &myitem_ops);
 
      depending on its class.
@@ -104,7 +115,9 @@ A suitably set up work item can then be enqueued for processing:
 	int ret = slow_work_enqueue(&myitem);
 
 This will return a -ve error if the thread pool is unable to gain a reference
-on the item, 0 otherwise.
+on the item, 0 otherwise, or (for delayed work):
+
+	int ret = delayed_slow_work_enqueue(&myitem, my_jiffy_delay);
 
 
 The items are reference counted, so there ought to be no need for a flush
@@ -112,6 +125,7 @@ operation.  But as the reference counting is optional, means to cancel
 existing work items are also included:
 
 	cancel_slow_work(&myitem);
+	cancel_delayed_slow_work(&myitem);
 
 can be used to cancel pending work.  The above cancel function waits for
 existing work to have been executed (or prevent execution of them, depending
include/linux/slow-work.h +29 -0
index eef2018..b245b9a 100644
@@ -17,6 +17,7 @@
 #ifdef CONFIG_SLOW_WORK
 
 #include <linux/sysctl.h>
+#include <linux/timer.h>
 
 struct slow_work;
 
@@ -52,10 +53,16 @@ struct slow_work {
 #define SLOW_WORK_ENQ_DEFERRED	2	/* item enqueue deferred */
 #define SLOW_WORK_VERY_SLOW	3	/* item is very slow */
 #define SLOW_WORK_CANCELLING	4	/* item is being cancelled, don't enqueue */
+#define SLOW_WORK_DELAYED	5	/* item is struct delayed_slow_work with active timer */
 	const struct slow_work_ops *ops; /* operations table for this item */
 	struct list_head	link;	/* link in queue */
 };
 
+struct delayed_slow_work {
+	struct slow_work	work;
+	struct timer_list	timer;
+};
+
 /**
  * slow_work_init - Initialise a slow work item
  * @work: The work item to initialise
@@ -72,6 +79,20 @@ static inline void slow_work_init(struct slow_work *work,
 }
 
 /**
+ * slow_work_init - Initialise a delayed slow work item
+ * @work: The work item to initialise
+ * @ops: The operations to use to handle the slow work item
+ *
+ * Initialise a delayed slow work item.
+ */
+static inline void delayed_slow_work_init(struct delayed_slow_work *dwork,
+					  const struct slow_work_ops *ops)
+{
+	init_timer(&dwork->timer);
+	slow_work_init(&dwork->work, ops);
+}
+
+/**
  * vslow_work_init - Initialise a very slow work item
  * @work: The work item to initialise
  * @ops: The operations to use to handle the slow work item
@@ -93,6 +114,14 @@ extern void slow_work_cancel(struct slow_work *work);
 extern int slow_work_register_user(struct module *owner);
 extern void slow_work_unregister_user(struct module *owner);
 
+extern int delayed_slow_work_enqueue(struct delayed_slow_work *dwork,
+				     unsigned long delay);
+
+static inline void delayed_slow_work_cancel(struct delayed_slow_work *dwork)
+{
+	slow_work_cancel(&dwork->work);
+}
+
 #ifdef CONFIG_SYSCTL
 extern ctl_table slow_work_sysctls[];
 #endif
kernel/slow-work.c +127 -2

144 lines not on this page · raw message

author maintainer
#6 lore raw

PATCH 5/28 SLOW_WORK: Allow the work items to be viewed through a /proc file

In reply to David Howells

Allow the executing and queued work items to be viewed through a /proc file for debugging purposes. The contents look something like the following:

    THR PID   ITEM ADDR        FL MARK  DESC
    === ===== ================ == ===== ==========
      0  3005 ffff880023f52348  a 952ms FSC: OBJ17d3: LOOK
      1  3006 ffff880024e33668  2 160ms FSC: OBJ17e5 OP60d3b: Write1/Store fl=2
      2  3165 ffff8800296dd180  a 424ms FSC: OBJ17e4: LOOK
      3  4089 ffff8800262c8d78  a 212ms FSC: OBJ17ea: CRTN
      4  4090 ffff88002792bed8  2 388ms FSC: OBJ17e8 OP60d36: Write1/Store fl=2
      5  4092 ffff88002a0ef308  2 388ms FSC: OBJ17e7 OP60d2e: Write1/Store fl=2
      6  4094 ffff88002abaf4b8  2 132ms FSC: OBJ17e2 OP60d4e: Write1/Store fl=2
      7  4095 ffff88002bb188e0  a 388ms FSC: OBJ17e9: CRTN
    vsq     - ffff880023d99668  1 308ms FSC: OBJ17e0 OP60f91: Write1/EnQ fl=2
    vsq     - ffff8800295d1740  1 212ms FSC: OBJ16be OP4d4b6: Write1/EnQ fl=2
    vsq     - ffff880025ba3308  1 160ms FSC: OBJ179a OP58dec: Write1/EnQ fl=2
    vsq     - ffff880024ec83e0  1 160ms FSC: OBJ17ae OP599f2: Write1/EnQ fl=2
    vsq     - ffff880026618e00  1 160ms FSC: OBJ17e6 OP60d33: Write1/EnQ fl=2
    vsq     - ffff880025a2a4b8  1 132ms FSC: OBJ16a2 OP4d583: Write1/EnQ fl=2
    vsq     - ffff880023cbe6d8  9 212ms FSC: OBJ17eb: LOOK
    vsq     - ffff880024d37590  9 212ms FSC: OBJ17ec: LOOK
    vsq     - ffff880027746cb0  9 212ms FSC: OBJ17ed: LOOK
    vsq     - ffff880024d37ae8  9 212ms FSC: OBJ17ee: LOOK
    vsq     - ffff880024d37cb0  9 212ms FSC: OBJ17ef: LOOK
    vsq     - ffff880025036550  9 212ms FSC: OBJ17f0: LOOK
    vsq     - ffff8800250368e0  9 212ms FSC: OBJ17f1: LOOK
    vsq     - ffff880025036aa8  9 212ms FSC: OBJ17f2: LOOK

In the 'THR' column, executing items show the thread they're occupying and queued threads indicate which queue they're on. 'PID' shows the process ID of a slow-work thread that's executing something. 'FL' shows the work item flags. 'MARK' indicates how long since an item was queued or began executing. Lastly, the 'DESC' column permits the owner of an item to give some information.

---

 Documentation/slow-work.txt |   60 +++++++++++
 include/linux/slow-work.h   |   11 ++
 init/Kconfig                |   10 ++
 kernel/Makefile             |    1 
 kernel/slow-work-proc.c     |  227 +++++++++++++++++++++++++++++++++++++++++++
 kernel/slow-work.c          |   44 ++++++--
 kernel/slow-work.h          |   72 ++++++++++++++
 7 files changed, 413 insertions(+), 12 deletions(-)
 create mode 100644 kernel/slow-work-proc.c
 create mode 100644 kernel/slow-work.h
Documentation/slow-work.txt +59 -1
index a9d1b0f..f120238 100644
@@ -149,7 +149,8 @@ ITEM OPERATIONS
 ===============
 
 Each work item requires a table of operations of type struct slow_work_ops.
-Only ->execute() is required, getting and putting of a reference are optional.
+Only ->execute() is required; the getting and putting of a reference and the
+describing of an item are all optional.
 
  (*) Get a reference on an item:
 
@@ -179,6 +180,16 @@ Only ->execute() is required, getting and putting of a reference are optional.
      This should perform the work required of the item.  It may sleep, it may
      perform disk I/O and it may wait for locks.
 
+ (*) View an item through /proc:
+
+	void (*desc)(struct slow_work *work, struct seq_file *m);
+
+     If supplied, this should print to 'm' a small string describing the work
+     the item is to do.  This should be no more than about 40 characters, and
+     shouldn't include a newline character.
+
+     See the 'Viewing executing and queued items' section below.
+
 
 ==================
 POOL CONFIGURATION
@@ -203,3 +214,50 @@ The slow-work thread pool has a number of configurables:
      is bounded to between 1 and one fewer than the number of active threads.
      This ensures there is always at least one thread that can process very
      slow work items, and always at least one thread that won't.
+
+
+==================================
+VIEWING EXECUTING AND QUEUED ITEMS
+==================================
+
+If CONFIG_SLOW_WORK_PROC is enabled, a proc file is made available:
+
+	/proc/slow_work_rq
+
+through which the list of work items being executed and the queues of items to
+be executed may be viewed.  The owner of a work item is given the chance to
+add some information of its own.
+
+The contents look something like the following:
+
+    THR PID   ITEM ADDR        FL MARK  DESC
+    === ===== ================ == ===== ==========
+      0  3005 ffff880023f52348  a 952ms FSC: OBJ17d3: LOOK
+      1  3006 ffff880024e33668  2 160ms FSC: OBJ17e5 OP60d3b: Write1/Store fl=2
+      2  3165 ffff8800296dd180  a 424ms FSC: OBJ17e4: LOOK
+      3  4089 ffff8800262c8d78  a 212ms FSC: OBJ17ea: CRTN
+      4  4090 ffff88002792bed8  2 388ms FSC: OBJ17e8 OP60d36: Write1/Store fl=2
+      5  4092 ffff88002a0ef308  2 388ms FSC: OBJ17e7 OP60d2e: Write1/Store fl=2
+      6  4094 ffff88002abaf4b8  2 132ms FSC: OBJ17e2 OP60d4e: Write1/Store fl=2
+      7  4095 ffff88002bb188e0  a 388ms FSC: OBJ17e9: CRTN
+    vsq     - ffff880023d99668  1 308ms FSC: OBJ17e0 OP60f91: Write1/EnQ fl=2
+    vsq     - ffff8800295d1740  1 212ms FSC: OBJ16be OP4d4b6: Write1/EnQ fl=2
+    vsq     - ffff880025ba3308  1 160ms FSC: OBJ179a OP58dec: Write1/EnQ fl=2
+    vsq     - ffff880024ec83e0  1 160ms FSC: OBJ17ae OP599f2: Write1/EnQ fl=2
+    vsq     - ffff880026618e00  1 160ms FSC: OBJ17e6 OP60d33: Write1/EnQ fl=2
+    vsq     - ffff880025a2a4b8  1 132ms FSC: OBJ16a2 OP4d583: Write1/EnQ fl=2
+    vsq     - ffff880023cbe6d8  9 212ms FSC: OBJ17eb: LOOK
+    vsq     - ffff880024d37590  9 212ms FSC: OBJ17ec: LOOK
+    vsq     - ffff880027746cb0  9 212ms FSC: OBJ17ed: LOOK
+    vsq     - ffff880024d37ae8  9 212ms FSC: OBJ17ee: LOOK
+    vsq     - ffff880024d37cb0  9 212ms FSC: OBJ17ef: LOOK
+    vsq     - ffff880025036550  9 212ms FSC: OBJ17f0: LOOK
+    vsq     - ffff8800250368e0  9 212ms FSC: OBJ17f1: LOOK
+    vsq     - ffff880025036aa8  9 212ms FSC: OBJ17f2: LOOK
+
+In the 'THR' column, executing items show the thread they're occupying and
+queued threads indicate which queue they're on.  'PID' shows the process ID of
+a slow-work thread that's executing something.  'FL' shows the work item flags.
+'MARK' indicates how long since an item was queued or began executing.  Lastly,
+the 'DESC' column permits the owner of an item to give some information.
+
include/linux/slow-work.h +11 -0
index b245b9a..f414851 100644
@@ -20,6 +20,9 @@
 #include <linux/timer.h>
 
 struct slow_work;
+#ifdef CONFIG_SLOW_WORK_PROC
+struct seq_file;
+#endif
 
 /*
  * The operations used to support slow work items
@@ -38,6 +41,11 @@ struct slow_work_ops {
 
 	/* execute a work item */
 	void (*execute)(struct slow_work *work);
+
+#ifdef CONFIG_SLOW_WORK_PROC
+	/* describe a work item for /proc */
+	void (*desc)(struct slow_work *work, struct seq_file *m);
+#endif
 };
 
 /*
@@ -56,6 +64,9 @@ struct slow_work {
 #define SLOW_WORK_DELAYED	5	/* item is struct delayed_slow_work with active timer */
 	const struct slow_work_ops *ops; /* operations table for this item */
 	struct list_head	link;	/* link in queue */
+#ifdef CONFIG_SLOW_WORK_PROC
+	struct timespec		mark;	/* jiffies at which queued or exec begun */
+#endif
 };
 
 struct delayed_slow_work {
init/Kconfig +10 -0
index 9e03ef8..ab5c648 100644
@@ -1098,6 +1098,16 @@ config SLOW_WORK
 
 	  See Documentation/slow-work.txt.
 
+config SLOW_WORK_PROC
+	bool "Slow work debugging through /proc"
+	default n
+	depends on SLOW_WORK && PROC_FS
+	help
+	  Display the contents of the slow work run queue through /proc,
+	  including items currently executing.
+
+	  See Documentation/slow-work.txt.
+
 endmenu		# General setup
 
 config HAVE_GENERIC_DMA_COHERENT
kernel/Makefile +1 -0
index b8d4cd8..776ffed 100644
@@ -94,6 +94,7 @@ obj-$(CONFIG_X86_DS) += trace/
 obj-$(CONFIG_RING_BUFFER) += trace/
 obj-$(CONFIG_SMP) += sched_cpupri.o
 obj-$(CONFIG_SLOW_WORK) += slow-work.o
+obj-$(CONFIG_SLOW_WORK_PROC) += slow-work-proc.o
 obj-$(CONFIG_PERF_EVENTS) += perf_event.o
 
 ifneq ($(CONFIG_SCHED_OMIT_FRAME_POINTER),y)
kernel/slow-work-proc.c new file +227 -0

227 lines not on this page · raw message

kernel/slow-work.c +33 -11

122 lines not on this page · raw message

kernel/slow-work.h new file +72 -0

72 lines not on this page · raw message

author maintainer
#7 lore raw

PATCH 6/28 SLOW_WORK: Allow the owner of a work item to determine if it is queued or not

In reply to David Howells

Add a function (slow_work_is_queued()) to permit the owner of a work item to determine if the item is queued or not.

The work item is counted as being queued if it is actually on the queue, not just if it is pending. If it is executing and pending, then it is not on the queue, but will rather be put back on the queue when execution finishes.

This permits a caller to quickly work out if it may be able to put another, dependent work item on the queue behind it, or whether it will have to wait till that is finished.

This can be used by CacheFiles to work out whether the creation a new object can be immediately deferred when it has to wait for an old object to be deleted, or whether a wait must take place. If a wait is necessary, then the slow-work thread can otherwise get blocked, preventing the deletion from taking place.

---

 Documentation/slow-work.txt |   15 +++++++++++++++
 include/linux/slow-work.h   |   19 +++++++++++++++++++
 2 files changed, 34 insertions(+), 0 deletions(-)
Documentation/slow-work.txt +15 -0
index f120238..0169c9d 100644
@@ -144,6 +144,21 @@ from being taken away before it completes.  module should almost certainly be
 THIS_MODULE.
 
 
+================
+HELPER FUNCTIONS
+================
+
+The slow-work facility provides a function by which it can be determined
+whether or not an item is queued for later execution:
+
+	bool queued = slow_work_is_queued(struct slow_work *work);
+
+If it returns false, then the item is not on the queue (it may be executing
+with a requeue pending).  This can be used to work out whether an item on which
+another depends is on the queue, thus allowing a dependent item to be queued
+after it.
+
+
 ===============
 ITEM OPERATIONS
 ===============
include/linux/slow-work.h +19 -0
index f414851..bfd3ab4 100644
@@ -120,6 +120,25 @@ static inline void vslow_work_init(struct slow_work *work,
 	INIT_LIST_HEAD(&work->link);
 }
 
+/**
+ * slow_work_is_queued - Determine if a slow work item is on the work queue
+ * work: The work item to test
+ *
+ * Determine if the specified slow-work item is on the work queue.  This
+ * returns true if it is actually on the queue.
+ *
+ * If the item is executing and has been marked for requeue when execution
+ * finishes, then false will be returned.
+ *
+ * Anyone wishing to wait for completion of execution can wait on the
+ * SLOW_WORK_EXECUTING bit.
+ */
+static inline bool slow_work_is_queued(struct slow_work *work)
+{
+	unsigned long flags = work->flags;
+	return flags & SLOW_WORK_PENDING && !(flags & SLOW_WORK_EXECUTING);
+}
+
 extern int slow_work_enqueue(struct slow_work *work);
 extern void slow_work_cancel(struct slow_work *work);
 extern int slow_work_register_user(struct module *owner);
author maintainer
#8 lore raw

PATCH 7/28 SLOW_WORK: Allow a requeueable work item to sleep till the thread is needed

In reply to David Howells

Add a function to allow a requeueable work item to sleep till the thread processing it is needed by the slow-work facility to perform other work.

Sometimes a work item can't progress immediately, but must wait for the completion of another work item that's currently being processed by another slow-work thread.

In some circumstances, the waiting item could instead - theoretically - put itself back on the queue and yield its thread back to the slow-work facility, thus waiting till it gets processing time again before attempting to progress. This would allow other work items processing time on that thread.

However, this only works if there is something on the queue for it to queue behind - otherwise it will just get a thread again immediately, and will end up cycling between the queue and the thread, eating up valuable CPU time.

So, slow_work_sleep_till_thread_needed() is provided such that an item can put itself on a wait queue that will wake it up when the event it is actually interested in occurs, then call this function in lieu of calling schedule().

This function will then sleep until either the item's event occurs or another work item appears on the queue. If another work item is queued, but the item's event hasn't occurred, then the work item should requeue itself and yield the thread back to the slow-work facility by returning.

This can be used by CacheFiles for an object that is being created on one thread to wait for an object being deleted on another thread where there is nothing on the queue for the creation to go and wait behind. As soon as an item appears on the queue that could be given thread time instead, CacheFiles can stick the creating object back on the queue and return to the slow-work facility - assuming the object deletion didn't also complete.

---

 Documentation/slow-work.txt |   44 ++++++++++++++++++++
 include/linux/slow-work.h   |    3 +
 kernel/slow-work.c          |   94 +++++++++++++++++++++++++++++++++++++++----
 3 files changed, 132 insertions(+), 9 deletions(-)
Documentation/slow-work.txt +44 -0
index 0169c9d..52bc314 100644
@@ -158,6 +158,50 @@ with a requeue pending).  This can be used to work out whether an item on which
 another depends is on the queue, thus allowing a dependent item to be queued
 after it.
 
+If the above shows an item on which another depends not to be queued, then the
+owner of the dependent item might need to wait.  However, to avoid locking up
+the threads unnecessarily be sleeping in them, it can make sense under some
+circumstances to return the work item to the queue, thus deferring it until
+some other items have had a chance to make use of the yielded thread.
+
+To yield a thread and defer an item, the work function should simply enqueue
+the work item again and return.  However, this doesn't work if there's nothing
+actually on the queue, as the thread just vacated will jump straight back into
+the item's work function, thus busy waiting on a CPU.
+
+Instead, the item should use the thread to wait for the dependency to go away,
+but rather than using schedule() or schedule_timeout() to sleep, it should use
+the following function:
+
+	bool requeue = slow_work_sleep_till_thread_needed(
+			struct slow_work *work,
+			signed long *_timeout);
+
+This will add a second wait and then sleep, such that it will be woken up if
+either something appears on the queue that could usefully make use of the
+thread - and behind which this item can be queued, or if the event the caller
+set up to wait for happens.  True will be returned if something else appeared
+on the queue and this work function should perhaps return, of false if
+something else woke it up.  The timeout is as for schedule_timeout().
+
+For example:
+
+	wq = bit_waitqueue(&my_flags, MY_BIT);
+	init_wait(&wait);
+	requeue = false;
+	do {
+		prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE);
+		if (!test_bit(MY_BIT, &my_flags))
+			break;
+		requeue = slow_work_sleep_till_thread_needed(&my_work,
+							     &timeout);
+	} while (timeout > 0 && !requeue);
+	finish_wait(wq, &wait);
+	if (!test_bit(MY_BIT, &my_flags)
+		goto do_my_thing;
+	if (requeue)
+		return; // to slow_work
+
 
 ===============
 ITEM OPERATIONS
include/linux/slow-work.h +3 -0
index bfd3ab4..5035a26 100644
@@ -152,6 +152,9 @@ static inline void delayed_slow_work_cancel(struct delayed_slow_work *dwork)
 	slow_work_cancel(&dwork->work);
 }
 
+extern bool slow_work_sleep_till_thread_needed(struct slow_work *work,
+					       signed long *_timeout);
+
 #ifdef CONFIG_SYSCTL
 extern ctl_table slow_work_sysctls[];
 #endif
kernel/slow-work.c +85 -9

146 lines not on this page · raw message

author maintainer
#9 lore raw

PATCH 8/28 FS-Cache: Annotate slow-work runqueue proc lines for FS-Cache work items

In reply to David Howells

Annotate slow-work runqueue proc lines for FS-Cache work items. Objects include the object ID and the state. Operations include the object ID, the operation ID and the operation type and state.

---

 fs/fscache/object.c           |   41 ++++++++++++++++++++++++++++++++++++++++-
 fs/fscache/operation.c        |   29 +++++++++++++++++++++++++++++
 fs/fscache/page.c             |   27 ++++++++++++++++++++++-----
 include/linux/fscache-cache.h |   12 ++++++++++++
 4 files changed, 103 insertions(+), 6 deletions(-)
fs/fscache/object.c +40 -1
index d236eb1..615b63d 100644
@@ -14,9 +14,10 @@
 
 #define FSCACHE_DEBUG_LEVEL COOKIE
 #include <linux/module.h>
+#include <linux/seq_file.h>
 #include "internal.h"
 
-const char *fscache_object_states[] = {
+const char *fscache_object_states[FSCACHE_OBJECT__NSTATES] = {
 	[FSCACHE_OBJECT_INIT]		= "OBJECT_INIT",
 	[FSCACHE_OBJECT_LOOKING_UP]	= "OBJECT_LOOKING_UP",
 	[FSCACHE_OBJECT_CREATING]	= "OBJECT_CREATING",
@@ -33,9 +34,28 @@ const char *fscache_object_states[] = {
 };
 EXPORT_SYMBOL(fscache_object_states);
 
+static const char fscache_object_states_short[FSCACHE_OBJECT__NSTATES][5] = {
+	[FSCACHE_OBJECT_INIT]		= "INIT",
+	[FSCACHE_OBJECT_LOOKING_UP]	= "LOOK",
+	[FSCACHE_OBJECT_CREATING]	= "CRTN",
+	[FSCACHE_OBJECT_AVAILABLE]	= "AVBL",
+	[FSCACHE_OBJECT_ACTIVE]		= "ACTV",
+	[FSCACHE_OBJECT_UPDATING]	= "UPDT",
+	[FSCACHE_OBJECT_DYING]		= "DYNG",
+	[FSCACHE_OBJECT_LC_DYING]	= "LCDY",
+	[FSCACHE_OBJECT_ABORT_INIT]	= "ABTI",
+	[FSCACHE_OBJECT_RELEASING]	= "RELS",
+	[FSCACHE_OBJECT_RECYCLING]	= "RCYC",
+	[FSCACHE_OBJECT_WITHDRAWING]	= "WTHD",
+	[FSCACHE_OBJECT_DEAD]		= "DEAD",
+};
+
 static void fscache_object_slow_work_put_ref(struct slow_work *);
 static int  fscache_object_slow_work_get_ref(struct slow_work *);
 static void fscache_object_slow_work_execute(struct slow_work *);
+#ifdef CONFIG_SLOW_WORK_PROC
+static void fscache_object_slow_work_desc(struct slow_work *, struct seq_file *);
+#endif
 static void fscache_initialise_object(struct fscache_object *);
 static void fscache_lookup_object(struct fscache_object *);
 static void fscache_object_available(struct fscache_object *);
@@ -49,6 +69,9 @@ const struct slow_work_ops fscache_object_slow_work_ops = {
 	.get_ref	= fscache_object_slow_work_get_ref,
 	.put_ref	= fscache_object_slow_work_put_ref,
 	.execute	= fscache_object_slow_work_execute,
+#ifdef CONFIG_SLOW_WORK_PROC
+	.desc		= fscache_object_slow_work_desc,
+#endif
 };
 EXPORT_SYMBOL(fscache_object_slow_work_ops);
 
@@ -327,6 +350,22 @@ static void fscache_object_slow_work_execute(struct slow_work *work)
 }
 
 /*
+ * describe an object for slow-work debugging
+ */
+#ifdef CONFIG_SLOW_WORK_PROC
+static void fscache_object_slow_work_desc(struct slow_work *work,
+					  struct seq_file *m)
+{
+	struct fscache_object *object =
+		container_of(work, struct fscache_object, work);
+
+	seq_printf(m, "FSC: OBJ%x: %s",
+		   object->debug_id,
+		   fscache_object_states_short[object->state]);
+}
+#endif
+
+/*
  * initialise an object
  * - check the specified object's parent to see if we can make use of it
  *   immediately to do a creation
fs/fscache/operation.c +29 -0
index f1a2857..91bbe6f 100644
@@ -13,6 +13,7 @@
 
 #define FSCACHE_DEBUG_LEVEL OPERATION
 #include <linux/module.h>
+#include <linux/seq_file.h>
 #include "internal.h"
 
 atomic_t fscache_op_debug_id;
@@ -31,6 +32,8 @@ void fscache_enqueue_operation(struct fscache_operation *op)
 	_enter("{OBJ%x OP%x,%u}",
 	       op->object->debug_id, op->debug_id, atomic_read(&op->usage));
 
+	fscache_set_op_state(op, "EnQ");
+
 	ASSERT(op->processor != NULL);
 	ASSERTCMP(op->object->state, >=, FSCACHE_OBJECT_AVAILABLE);
 	ASSERTCMP(atomic_read(&op->usage), >, 0);
@@ -67,6 +70,8 @@ EXPORT_SYMBOL(fscache_enqueue_operation);
 static void fscache_run_op(struct fscache_object *object,
 			   struct fscache_operation *op)
 {
+	fscache_set_op_state(op, "Run");
+
 	object->n_in_progress++;
 	if (test_and_clear_bit(FSCACHE_OP_WAITING, &op->flags))
 		wake_up_bit(&op->flags, FSCACHE_OP_WAITING);
@@ -87,6 +92,8 @@ int fscache_submit_exclusive_op(struct fscache_object *object,
 
 	_enter("{OBJ%x OP%x},", object->debug_id, op->debug_id);
 
+	fscache_set_op_state(op, "SubmitX");
+
 	spin_lock(&object->lock);
 	ASSERTCMP(object->n_ops, >=, object->n_in_progress);
 	ASSERTCMP(object->n_ops, >=, object->n_exclusive);
@@ -190,6 +197,8 @@ int fscache_submit_op(struct fscache_object *object,
 
 	ASSERTCMP(atomic_read(&op->usage), >, 0);
 
+	fscache_set_op_state(op, "Submit");
+
 	spin_lock(&object->lock);
 	ASSERTCMP(object->n_ops, >=, object->n_in_progress);
 	ASSERTCMP(object->n_ops, >=, object->n_exclusive);
@@ -298,6 +307,8 @@ void fscache_put_operation(struct fscache_operation *op)
 	if (!atomic_dec_and_test(&op->usage))
 		return;
 
+	fscache_set_op_state(op, "Put");
+
 	_debug("PUT OP");
 	if (test_and_set_bit(FSCACHE_OP_DEAD, &op->flags))
 		BUG();
@@ -452,9 +463,27 @@ static void fscache_op_execute(struct slow_work *work)
 	_leave("");
 }
 
+/*
+ * describe an operation for slow-work debugging
+ */
+#ifdef CONFIG_SLOW_WORK_PROC
+static void fscache_op_desc(struct slow_work *work, struct seq_file *m)
+{
+	struct fscache_operation *op =
+		container_of(work, struct fscache_operation, slow_work);
+
+	seq_printf(m, "FSC: OBJ%x OP%x: %s/%s fl=%lx",
+		   op->object->debug_id, op->debug_id,
+		   op->name, op->state, op->flags);
+}
+#endif
+
 const struct slow_work_ops fscache_op_slow_work_ops = {
 	.owner		= THIS_MODULE,
 	.get_ref	= fscache_op_get_ref,
 	.put_ref	= fscache_op_put_ref,
 	.execute	= fscache_op_execute,
+#ifdef CONFIG_SLOW_WORK_PROC
+	.desc		= fscache_op_desc,
+#endif
 };
fs/fscache/page.c +22 -5
index 2568e0e..e8bbc39 100644
@@ -63,14 +63,19 @@ static void fscache_end_page_write(struct fscache_cookie *cookie, struct page *p
 static void fscache_attr_changed_op(struct fscache_operation *op)
 {
 	struct fscache_object *object = op->object;
+	int ret;
 
 	_enter("{OBJ%x OP%x}", object->debug_id, op->debug_id);
 
 	fscache_stat(&fscache_n_attr_changed_calls);
 
-	if (fscache_object_is_active(object) &&
-	    object->cache->ops->attr_changed(object) < 0)
-		fscache_abort_object(object);
+	if (fscache_object_is_active(object)) {
+		fscache_set_op_state(op, "CallFS");
+		ret = object->cache->ops->attr_changed(object);
+		fscache_set_op_state(op, "Done");
+		if (ret < 0)
+			fscache_abort_object(object);
+	}
 
 	_leave("");
 }
@@ -99,6 +104,7 @@ int __fscache_attr_changed(struct fscache_cookie *cookie)
 	fscache_operation_init(op, NULL);
 	fscache_operation_init_slow(op, fscache_attr_changed_op);
 	op->flags = FSCACHE_OP_SLOW | (1 << FSCACHE_OP_EXCLUSIVE);
+	fscache_set_op_name(op, "Attr");
 
 	spin_lock(&cookie->lock);
 
@@ -184,6 +190,7 @@ static struct fscache_retrieval *fscache_alloc_retrieval(
 	op->start_time	= jiffies;
 	INIT_WORK(&op->op.fast_work, fscache_retrieval_work);
 	INIT_LIST_HEAD(&op->to_do);
+	fscache_set_op_name(&op->op, "Retr");
 	return op;
 }
 
@@ -257,6 +264,7 @@ int __fscache_read_or_alloc_page(struct fscache_cookie *cookie,
 		_leave(" = -ENOMEM");
 		return -ENOMEM;
 	}
+	fscache_set_op_name(&op->op, "RetrRA1");
 
 	spin_lock(&cookie->lock);
 
@@ -369,6 +377,7 @@ int __fscache_read_or_alloc_pages(struct fscache_cookie *cookie,
 	op = fscache_alloc_retrieval(mapping, end_io_func, context);
 	if (!op)
 		return -ENOMEM;
+	fscache_set_op_name(&op->op, "RetrRAN");
 
 	spin_lock(&cookie->lock);
 
@@ -461,6 +470,7 @@ int __fscache_alloc_page(struct fscache_cookie *cookie,
 	op = fscache_alloc_retrieval(page->mapping, NULL, NULL);
 	if (!op)
 		return -ENOMEM;
+	fscache_set_op_name(&op->op, "RetrAL1");
 
 	spin_lock(&cookie->lock);
 
@@ -529,6 +539,8 @@ static void fscache_write_op(struct fscache_operation *_op)
 
 	_enter("{OP%x,%d}", op->op.debug_id, atomic_read(&op->op.usage));
 
+	fscache_set_op_state(&op->op, "GetPage");
+
 	spin_lock(&cookie->lock);
 	spin_lock(&object->lock);
 
@@ -559,13 +571,17 @@ static void fscache_write_op(struct fscache_operation *_op)
 	spin_unlock(&cookie->lock);
 
 	if (page) {
+		fscache_set_op_state(&op->op, "Store");
 		ret = object->cache->ops->write_page(op, page);
+		fscache_set_op_state(&op->op, "EndWrite");
 		fscache_end_page_write(cookie, page);
 		page_cache_release(page);
-		if (ret < 0)
+		if (ret < 0) {
+			fscache_set_op_state(&op->op, "Abort");
 			fscache_abort_object(object);
-		else
+		} else {
 			fscache_enqueue_operation(&op->op);
+		}
 	}
 
 	_leave("");
@@ -634,6 +650,7 @@ int __fscache_write_page(struct fscache_cookie *cookie,
 	fscache_operation_init(&op->op, fscache_release_write_op);
 	fscache_operation_init_slow(&op->op, fscache_write_op);
 	op->op.flags = FSCACHE_OP_SLOW | (1 << FSCACHE_OP_WAITING);
+	fscache_set_op_name(&op->op, "Write1");
 
 	ret = radix_tree_preload(gfp & ~__GFP_HIGHMEM);
 	if (ret < 0)
include/linux/fscache-cache.h +12 -0
index 84d3532..7a9847c 100644
@@ -102,6 +102,16 @@ struct fscache_operation {
 
 	/* operation releaser */
 	fscache_operation_release_t release;
+
+#ifdef CONFIG_SLOW_WORK_PROC
+	const char *name;		/* operation name */
+	const char *state;		/* operation state */
+#define fscache_set_op_name(OP, N)	do { (OP)->name  = (N); } while(0)
+#define fscache_set_op_state(OP, S)	do { (OP)->state = (S); } while(0)
+#else
+#define fscache_set_op_name(OP, N)	do { } while(0)
+#define fscache_set_op_state(OP, S)	do { } while(0)
+#endif
 };
 
 extern atomic_t fscache_op_debug_id;
@@ -125,6 +135,7 @@ static inline void fscache_operation_init(struct fscache_operation *op,
 	op->debug_id = atomic_inc_return(&fscache_op_debug_id);
 	op->release = release;
 	INIT_LIST_HEAD(&op->pend_link);
+	fscache_set_op_state(op, "Init");
 }
 
 /**
@@ -337,6 +348,7 @@ struct fscache_object {
 		FSCACHE_OBJECT_RECYCLING,	/* retiring object */
 		FSCACHE_OBJECT_WITHDRAWING,	/* withdrawing object */
 		FSCACHE_OBJECT_DEAD,		/* object is now dead */
+		FSCACHE_OBJECT__NSTATES
 	} state;
 
 	int			debug_id;	/* debugging ID */
author maintainer
#10 lore raw

PATCH 9/28 FS-Cache: Allow the current state of all objects to be dumped

In reply to David Howells

Allow the current state of all fscache objects to be dumped by doing:

cat /proc/fs/fscache/objects

By default, all objects and all fields will be shown. This can be restricted by adding a suitable key to one of the caller's keyrings (such as the session keyring):

keyctl add user fscache:objlist "<restrictions>" @s

The <restrictions> are:

K Show hexdump of object key (don't show if not given)
A Show hexdump of object aux data (don't show if not given)

And paired restrictions:

C Show objects that have a cookie
c Show objects that don't have a cookie
B Show objects that are busy
b Show objects that aren't busy
W Show objects that have pending writes
w Show objects that don't have pending writes
R Show objects that have outstanding reads
r Show objects that don't have outstanding reads
S Show objects that have slow work queued
s Show objects that don't have slow work queued

If neither side of a restriction pair is given, then both are implied. For example:

keyctl add user fscache:objlist KB @s

shows objects that are busy, and lists their object keys, but does not dump their auxiliary data. It also implies "CcWwRrSs", but as 'B' is given, 'b' is not implied.

---

 Documentation/filesystems/caching/fscache.txt |   81 +++++
 fs/cachefiles/interface.c                     |    1 
 fs/cachefiles/rdwr.c                          |    6 
 fs/fscache/Kconfig                            |    7 
 fs/fscache/Makefile                           |    1 
 fs/fscache/cache.c                            |    1 
 fs/fscache/cookie.c                           |    2 
 fs/fscache/internal.h                         |   13 +
 fs/fscache/object-list.c                      |  432 +++++++++++++++++++++++++
 fs/fscache/object.c                           |    2 
 fs/fscache/operation.c                        |    3 
 fs/fscache/page.c                             |    6 
 fs/fscache/proc.c                             |   13 +
 include/linux/fscache-cache.h                 |   13 +
 14 files changed, 578 insertions(+), 3 deletions(-)
 create mode 100644 fs/fscache/object-list.c
Documentation/filesystems/caching/fscache.txt +81 -0
index 9e94b94..cf13cd3 100644
@@ -299,6 +299,87 @@ proc files.
      jiffy range covered, and the SECS field the equivalent number of seconds.
 
 
+===========
+OBJECT LIST
+===========
+
+If CONFIG_FSCACHE_OBJECT_LIST is enabled, the FS-Cache facility will maintain a
+list of all the objects currently allocated and allow them to be viewed
+through:
+
+	/proc/fs/fscache/objects
+
+This will look something like:
+
+	[root@andromeda ~]# head /proc/fs/fscache/objects 
+	OBJECT   PARENT   STAT CHLDN OPS OOP IPR EX READS EM EV F S | NETFS_COOKIE_DEF TY FL NETFS_DATA       OBJECT_KEY, AUX_DATA
+	======== ======== ==== ===== === === === == ===== == == = = | ================ == == ================ ================
+	   17e4b        2 ACTV     0   0   0   0  0     0 7b  4 0 8 | NFS.fh           DT  0 ffff88001dd82820 010006017edcf8bbc93b43298fdfbe71e50b57b13a172c0117f38472, e567634700000000000000000000000063f2404a000000000000000000000000c9030000000000000000000063f2404a
+	   1693a        2 ACTV     0   0   0   0  0     0 7b  4 0 8 | NFS.fh           DT  0 ffff88002db23380 010006017edcf8bbc93b43298fdfbe71e50b57b1e0162c01a2df0ea6, 420ebc4a000000000000000000000000420ebc4a0000000000000000000000000e1801000000000000000000420ebc4a
+
+where the first set of columns before the '|' describe the object:
+
+	COLUMN	DESCRIPTION
+	=======	===============================================================
+	OBJECT	Object debugging ID (appears as OBJ%x in some debug messages)
+	PARENT	Debugging ID of parent object
+	STAT	Object state
+	CHLDN	Number of child objects of this object
+	OPS	Number of outstanding operations on this object
+	OOP	Number of outstanding child object management operations
+	IPR
+	EX	Number of outstanding exclusive operations
+	READS	Number of outstanding read operations
+	EM	Object's event mask
+	EV	Events raised on this object
+	F	Object flags
+	S	Object slow-work work item flags
+
+and the second set of columns describe the object's cookie, if present:
+
+	COLUMN		DESCRIPTION
+	===============	=======================================================
+	NETFS_COOKIE_DEF Name of netfs cookie definition
+	TY		Cookie type (IX - index, DT - data, hex - special)
+	FL		Cookie flags
+	NETFS_DATA	Netfs private data stored in the cookie
+	OBJECT_KEY	Object key	} 1 column, with separating comma
+	AUX_DATA	Object aux data	} presence may be configured
+
+The data shown may be filtered by attaching the a key to an appropriate keyring
+before viewing the file.  Something like:
+
+		keyctl add user fscache:objlist <restrictions> @s
+
+where <restrictions> are a selection of the following letters:
+
+	K	Show hexdump of object key (don't show if not given)
+	A	Show hexdump of object aux data (don't show if not given)
+
+and the following paired letters:
+
+	C	Show objects that have a cookie
+	c	Show objects that don't have a cookie
+	B	Show objects that are busy
+	b	Show objects that aren't busy
+	W	Show objects that have pending writes
+	w	Show objects that don't have pending writes
+	R	Show objects that have outstanding reads
+	r	Show objects that don't have outstanding reads
+	S	Show objects that have slow work queued
+	s	Show objects that don't have slow work queued
+
+If neither side of a letter pair is given, then both are implied.  For example:
+
+	keyctl add user fscache:objlist KB @s
+
+shows objects that are busy, and lists their object keys, but does not dump
+their auxiliary data.  It also implies "CcWwRrSs", but as 'B' is given, 'b' is
+not implied.
+
+By default all objects and all fields will be shown.
+
+
 =========
 DEBUGGING
 =========
fs/cachefiles/interface.c +1 -0
index 431accd..dd7f852 100644
@@ -331,6 +331,7 @@ static void cachefiles_put_object(struct fscache_object *_object)
 		}
 
 		cache = object->fscache.cache;
+		fscache_object_destroy(&object->fscache);
 		kmem_cache_free(cachefiles_object_jar, object);
 		fscache_object_destroyed(cache);
 	}
fs/cachefiles/rdwr.c +4 -2
index a69787e..3304646 100644
@@ -333,7 +333,8 @@ int cachefiles_read_or_alloc_page(struct fscache_retrieval *op,
 
 	shift = PAGE_SHIFT - inode->i_sb->s_blocksize_bits;
 
-	op->op.flags = FSCACHE_OP_FAST;
+	op->op.flags &= FSCACHE_OP_KEEP_FLAGS;
+	op->op.flags |= FSCACHE_OP_FAST;
 	op->op.processor = cachefiles_read_copier;
 
 	pagevec_init(&pagevec, 0);
@@ -639,7 +640,8 @@ int cachefiles_read_or_alloc_pages(struct fscache_retrieval *op,
 
 	pagevec_init(&pagevec, 0);
 
-	op->op.flags = FSCACHE_OP_FAST;
+	op->op.flags &= FSCACHE_OP_KEEP_FLAGS;
+	op->op.flags |= FSCACHE_OP_FAST;
 	op->op.processor = cachefiles_read_copier;
 
 	INIT_LIST_HEAD(&backpages);
fs/fscache/Kconfig +7 -0
index 9bbb8ce..864dac2 100644
@@ -54,3 +54,10 @@ config FSCACHE_DEBUG
 	  enabled by setting bits in /sys/modules/fscache/parameter/debug.
 
 	  See Documentation/filesystems/caching/fscache.txt for more information.
+
+config FSCACHE_OBJECT_LIST
+	bool "Maintain global object list for debugging purposes"
+	depends on FSCACHE && PROC_FS
+	help
+	  Maintain a global list of active fscache objects that can be
+	  retrieved through /proc/fs/fscache/objects for debugging purposes
fs/fscache/Makefile +1 -0
index 91571b9..6d56153 100644
@@ -15,5 +15,6 @@ fscache-y := \
 fscache-$(CONFIG_PROC_FS) += proc.o
 fscache-$(CONFIG_FSCACHE_STATS) += stats.o
 fscache-$(CONFIG_FSCACHE_HISTOGRAM) += histogram.o
+fscache-$(CONFIG_FSCACHE_OBJECT_LIST) += object-list.o
 
 obj-$(CONFIG_FSCACHE) := fscache.o
fs/fscache/cache.c +1 -0
index e21985b..724384e 100644
@@ -263,6 +263,7 @@ int fscache_add_cache(struct fscache_cache *cache,
 	spin_lock(&cache->object_list_lock);
 	list_add_tail(&ifsdef->cache_link, &cache->object_list);
 	spin_unlock(&cache->object_list_lock);
+	fscache_objlist_add(ifsdef);
 
 	/* add the cache's netfs definition index object to the top level index
 	 * cookie as a known backing object */
fs/fscache/cookie.c +2 -0
index 72fd18f..9b51873 100644
@@ -349,6 +349,8 @@ static int fscache_attach_object(struct fscache_cookie *cookie,
 	object->cookie = cookie;
 	atomic_inc(&cookie->usage);
 	hlist_add_head(&object->cookie_link, &cookie->backing_objects);
+
+	fscache_objlist_add(object);
 	ret = 0;
 
 cant_attach_object:
fs/fscache/internal.h +13 -0
index 1c34130..fe02973 100644
@@ -88,11 +88,24 @@ extern int fscache_wait_bit_interruptible(void *);
 /*
  * object.c
  */
+extern const char fscache_object_states_short[FSCACHE_OBJECT__NSTATES][5];
+
 extern void fscache_withdrawing_object(struct fscache_cache *,
 				       struct fscache_object *);
 extern void fscache_enqueue_object(struct fscache_object *);
 
 /*
+ * object-list.c
+ */
+#ifdef CONFIG_FSCACHE_OBJECT_LIST
+extern const struct file_operations fscache_objlist_fops;
+
+extern void fscache_objlist_add(struct fscache_object *);
+#else
+#define fscache_objlist_add(object) do {} while(0)
+#endif
+
+/*
  * operation.c
  */
 extern int fscache_submit_exclusive_op(struct fscache_object *,
fs/fscache/object-list.c new file +432 -0

432 lines not on this page · raw message

fs/fscache/object.c +1 -1
index 615b63d..ad1644f 100644
@@ -34,7 +34,7 @@ const char *fscache_object_states[FSCACHE_OBJECT__NSTATES] = {
 };
 EXPORT_SYMBOL(fscache_object_states);
 
-static const char fscache_object_states_short[FSCACHE_OBJECT__NSTATES][5] = {
+const char fscache_object_states_short[FSCACHE_OBJECT__NSTATES][5] = {
 	[FSCACHE_OBJECT_INIT]		= "INIT",
 	[FSCACHE_OBJECT_LOOKING_UP]	= "LOOK",
 	[FSCACHE_OBJECT_CREATING]	= "CRTN",
fs/fscache/operation.c +3 -0
index 91bbe6f..09e43b6 100644
@@ -322,6 +322,9 @@ void fscache_put_operation(struct fscache_operation *op)
 
 	object = op->object;
 
+	if (test_bit(FSCACHE_OP_DEC_READ_CNT, &op->flags))
+		atomic_dec(&object->n_reads);
+
 	/* now... we may get called with the object spinlock held, so we
 	 * complete the cleanup here only if we can immediately acquire the
 	 * lock, and defer it otherwise */
fs/fscache/page.c +6 -0
index e8bbc39..c5973e3 100644
@@ -275,6 +275,9 @@ int __fscache_read_or_alloc_page(struct fscache_cookie *cookie,
 
 	ASSERTCMP(object->state, >, FSCACHE_OBJECT_LOOKING_UP);
 
+	atomic_inc(&object->n_reads);
+	set_bit(FSCACHE_OP_DEC_READ_CNT, &op->op.flags);
+
 	if (fscache_submit_op(object, &op->op) < 0)
 		goto nobufs_unlock;
 	spin_unlock(&cookie->lock);
@@ -386,6 +389,9 @@ int __fscache_read_or_alloc_pages(struct fscache_cookie *cookie,
 	object = hlist_entry(cookie->backing_objects.first,
 			     struct fscache_object, cookie_link);
 
+	atomic_inc(&object->n_reads);
+	set_bit(FSCACHE_OP_DEC_READ_CNT, &op->op.flags);
+
 	if (fscache_submit_op(object, &op->op) < 0)
 		goto nobufs_unlock;
 	spin_unlock(&cookie->lock);
fs/fscache/proc.c +13 -0
index beeab44..1d9e495 100644
@@ -37,10 +37,20 @@ int __init fscache_proc_init(void)
 		goto error_histogram;
 #endif
 
+#ifdef CONFIG_FSCACHE_OBJECT_LIST
+	if (!proc_create("fs/fscache/objects", S_IFREG | 0444, NULL,
+			 &fscache_objlist_fops))
+		goto error_objects;
+#endif
+
 	_leave(" = 0");
 	return 0;
 
+#ifdef CONFIG_FSCACHE_OBJECT_LIST
+error_objects:
+#endif
 #ifdef CONFIG_FSCACHE_HISTOGRAM
+	remove_proc_entry("fs/fscache/histogram", NULL);
 error_histogram:
 #endif
 #ifdef CONFIG_FSCACHE_STATS
@@ -58,6 +68,9 @@ error_dir:
  */
 void fscache_proc_cleanup(void)
 {
+#ifdef CONFIG_FSCACHE_OBJECT_LIST
+	remove_proc_entry("fs/fscache/objects", NULL);
+#endif
 #ifdef CONFIG_FSCACHE_HISTOGRAM
 	remove_proc_entry("fs/fscache/histogram", NULL);
 #endif
include/linux/fscache-cache.h +13 -0
index 7a9847c..184cbdf 100644
@@ -91,6 +91,8 @@ struct fscache_operation {
 #define FSCACHE_OP_WAITING	4	/* cleared when op is woken */
 #define FSCACHE_OP_EXCLUSIVE	5	/* exclusive op, other ops must wait */
 #define FSCACHE_OP_DEAD		6	/* op is now dead */
+#define FSCACHE_OP_DEC_READ_CNT	7	/* decrement object->n_reads on destruction */
+#define FSCACHE_OP_KEEP_FLAGS	0xc0	/* flags to keep when repurposing an op */
 
 	atomic_t		usage;
 	unsigned		debug_id;	/* debugging ID */
@@ -357,6 +359,7 @@ struct fscache_object {
 	int			n_obj_ops;	/* number of object ops outstanding on object */
 	int			n_in_progress;	/* number of ops in progress */
 	int			n_exclusive;	/* number of exclusive ops queued */
+	atomic_t		n_reads;	/* number of read ops in progress */
 	spinlock_t		lock;		/* state and operations lock */
 
 	unsigned long		lookup_jif;	/* time at which lookup started */
@@ -370,6 +373,7 @@ struct fscache_object {
 #define FSCACHE_OBJECT_EV_RELEASE	4	/* T if netfs requested object release */
 #define FSCACHE_OBJECT_EV_RETIRE	5	/* T if netfs requested object retirement */
 #define FSCACHE_OBJECT_EV_WITHDRAW	6	/* T if cache requested object withdrawal */
+#define FSCACHE_OBJECT_EVENTS_MASK	0x7f	/* mask of all events*/
 
 	unsigned long		flags;
 #define FSCACHE_OBJECT_LOCK		0	/* T if object is busy being processed */
@@ -385,6 +389,9 @@ struct fscache_object {
 	struct list_head	dependents;	/* FIFO of dependent objects */
 	struct list_head	dep_link;	/* link in parent's dependents list */
 	struct list_head	pending_ops;	/* unstarted operations on this object */
+#ifdef CONFIG_FSCACHE_OBJECT_LIST
+	struct rb_node		objlist_link;	/* link in global object list */
+#endif
 	pgoff_t			store_limit;	/* current storage limit */
 };
 
@@ -434,6 +441,12 @@ void fscache_object_init(struct fscache_object *object,
 extern void fscache_object_lookup_negative(struct fscache_object *object);
 extern void fscache_obtained_object(struct fscache_object *object);
 
+#ifdef CONFIG_FSCACHE_OBJECT_LIST
+extern void fscache_object_destroy(struct fscache_object *object);
+#else
+#define fscache_object_destroy(object) do {} while(0)
+#endif
+
 /**
  * fscache_object_destroyed - Note destruction of an object in a cache
  * @cache: The cache from which the object came
author maintainer
#11 lore raw

PATCH 10/28 FS-Cache: Add counters for entry/exit to/from cache operation functions

In reply to David Howells

Count entries to and exits from cache operation table functions. Maintain these as a single counter that's added to or removed from as appropriate.

---

 Documentation/filesystems/caching/fscache.txt |   16 +++++++++++
 fs/fscache/cache.c                            |    4 +++
 fs/fscache/cookie.c                           |    9 +++++-
 fs/fscache/internal.h                         |   22 +++++++++++++++
 fs/fscache/object.c                           |   26 ++++++++++++++++--
 fs/fscache/page.c                             |   29 ++++++++++++++++----
 fs/fscache/stats.c                            |   37 +++++++++++++++++++++++++
 7 files changed, 134 insertions(+), 9 deletions(-)
Documentation/filesystems/caching/fscache.txt +16 -0
index cf13cd3..472398f 100644
@@ -274,6 +274,22 @@ proc files.
 		dfr=N	Number of async ops queued for deferred release
 		rel=N	Number of async ops released
 		gc=N	Number of deferred-release async ops garbage collected
+	CacheOp	alo=N	Number of in-progress alloc_object() cache ops
+		luo=N	Number of in-progress lookup_object() cache ops
+		luc=N	Number of in-progress lookup_complete() cache ops
+		gro=N	Number of in-progress grab_object() cache ops
+		upo=N	Number of in-progress update_object() cache ops
+		dro=N	Number of in-progress drop_object() cache ops
+		pto=N	Number of in-progress put_object() cache ops
+		syn=N	Number of in-progress sync_cache() cache ops
+		atc=N	Number of in-progress attr_changed() cache ops
+		rap=N	Number of in-progress read_or_alloc_page() cache ops
+		ras=N	Number of in-progress read_or_alloc_pages() cache ops
+		alp=N	Number of in-progress allocate_page() cache ops
+		als=N	Number of in-progress allocate_pages() cache ops
+		wrp=N	Number of in-progress write_page() cache ops
+		ucp=N	Number of in-progress uncache_page() cache ops
+		dsp=N	Number of in-progress dissociate_pages() cache ops
 
 
  (*) /proc/fs/fscache/histogram
fs/fscache/cache.c +4 -0
index 724384e..6a3c48a 100644
@@ -381,11 +381,15 @@ void fscache_withdraw_cache(struct fscache_cache *cache)
 
 	/* make sure all pages pinned by operations on behalf of the netfs are
 	 * written to disk */
+	fscache_stat(&fscache_n_cop_sync_cache);
 	cache->ops->sync_cache(cache);
+	fscache_stat_d(&fscache_n_cop_sync_cache);
 
 	/* dissociate all the netfs pages backed by this cache from the block
 	 * mappings in the cache */
+	fscache_stat(&fscache_n_cop_dissociate_pages);
 	cache->ops->dissociate_pages(cache);
+	fscache_stat_d(&fscache_n_cop_dissociate_pages);
 
 	/* we now have to destroy all the active objects pertaining to this
 	 * cache - which we do by passing them off to thread pool to be
fs/fscache/cookie.c +8 -1
index 9b51873..432482e 100644
@@ -249,7 +249,9 @@ static int fscache_alloc_object(struct fscache_cache *cache,
 
 	/* ask the cache to allocate an object (we may end up with duplicate
 	 * objects at this stage, but we sort that out later) */
+	fscache_stat(&fscache_n_cop_alloc_object);
 	object = cache->ops->alloc_object(cache, cookie);
+	fscache_stat_d(&fscache_n_cop_alloc_object);
 	if (IS_ERR(object)) {
 		fscache_stat(&fscache_n_object_no_alloc);
 		ret = PTR_ERR(object);
@@ -270,8 +272,11 @@ static int fscache_alloc_object(struct fscache_cache *cache,
 	/* only attach if we managed to allocate all we needed, otherwise
 	 * discard the object we just allocated and instead use the one
 	 * attached to the cookie */
-	if (fscache_attach_object(cookie, object) < 0)
+	if (fscache_attach_object(cookie, object) < 0) {
+		fscache_stat(&fscache_n_cop_put_object);
 		cache->ops->put_object(object);
+		fscache_stat_d(&fscache_n_cop_put_object);
+	}
 
 	_leave(" = 0");
 	return 0;
@@ -287,7 +292,9 @@ object_already_extant:
 	return 0;
 
 error_put:
+	fscache_stat(&fscache_n_cop_put_object);
 	cache->ops->put_object(object);
+	fscache_stat_d(&fscache_n_cop_put_object);
 error:
 	_leave(" = %d", ret);
 	return ret;
fs/fscache/internal.h +22 -0
index fe02973..b85cc89 100644
@@ -208,11 +208,33 @@ extern atomic_t fscache_n_checkaux_okay;
 extern atomic_t fscache_n_checkaux_update;
 extern atomic_t fscache_n_checkaux_obsolete;
 
+extern atomic_t fscache_n_cop_alloc_object;
+extern atomic_t fscache_n_cop_lookup_object;
+extern atomic_t fscache_n_cop_lookup_complete;
+extern atomic_t fscache_n_cop_grab_object;
+extern atomic_t fscache_n_cop_update_object;
+extern atomic_t fscache_n_cop_drop_object;
+extern atomic_t fscache_n_cop_put_object;
+extern atomic_t fscache_n_cop_sync_cache;
+extern atomic_t fscache_n_cop_attr_changed;
+extern atomic_t fscache_n_cop_read_or_alloc_page;
+extern atomic_t fscache_n_cop_read_or_alloc_pages;
+extern atomic_t fscache_n_cop_allocate_page;
+extern atomic_t fscache_n_cop_allocate_pages;
+extern atomic_t fscache_n_cop_write_page;
+extern atomic_t fscache_n_cop_uncache_page;
+extern atomic_t fscache_n_cop_dissociate_pages;
+
 static inline void fscache_stat(atomic_t *stat)
 {
 	atomic_inc(stat);
 }
 
+static inline void fscache_stat_d(atomic_t *stat)
+{
+	atomic_dec(stat);
+}
+
 extern const struct file_operations fscache_stats_fops;
 #else
 
fs/fscache/object.c +24 -2
index ad1644f..0d65c0c 100644
@@ -144,13 +144,17 @@ static void fscache_object_state_machine(struct fscache_object *object)
 	case FSCACHE_OBJECT_UPDATING:
 		clear_bit(FSCACHE_OBJECT_EV_UPDATE, &object->events);
 		fscache_stat(&fscache_n_updates_run);
+		fscache_stat(&fscache_n_cop_update_object);
 		object->cache->ops->update_object(object);
+		fscache_stat_d(&fscache_n_cop_update_object);
 		goto active_transit;
 
 		/* handle an object dying during lookup or creation */
 	case FSCACHE_OBJECT_LC_DYING:
 		object->event_mask &= ~(1 << FSCACHE_OBJECT_EV_UPDATE);
+		fscache_stat(&fscache_n_cop_lookup_complete);
 		object->cache->ops->lookup_complete(object);
+		fscache_stat_d(&fscache_n_cop_lookup_complete);
 
 		spin_lock(&object->lock);
 		object->state = FSCACHE_OBJECT_DYING;
@@ -416,7 +420,9 @@ static void fscache_initialise_object(struct fscache_object *object)
 			 * binding on to us, so we need to make sure we don't
 			 * add ourself to the list multiple times */
 			if (list_empty(&object->dep_link)) {
+				fscache_stat(&fscache_n_cop_grab_object);
 				object->cache->ops->grab_object(object);
+				fscache_stat_d(&fscache_n_cop_grab_object);
 				list_add(&object->dep_link,
 					 &parent->dependents);
 
@@ -478,7 +484,9 @@ static void fscache_lookup_object(struct fscache_object *object)
 	       object->cache->tag->name);
 
 	fscache_stat(&fscache_n_object_lookups);
+	fscache_stat(&fscache_n_cop_lookup_object);
 	object->cache->ops->lookup_object(object);
+	fscache_stat_d(&fscache_n_cop_lookup_object);
 
 	if (test_bit(FSCACHE_OBJECT_EV_ERROR, &object->events))
 		set_bit(FSCACHE_COOKIE_UNAVAILABLE, &cookie->flags);
@@ -602,7 +610,9 @@ static void fscache_object_available(struct fscache_object *object)
 	}
 	spin_unlock(&object->lock);
 
+	fscache_stat(&fscache_n_cop_lookup_complete);
 	object->cache->ops->lookup_complete(object);
+	fscache_stat_d(&fscache_n_cop_lookup_complete);
 	fscache_enqueue_dependents(object);
 
 	fscache_hist(fscache_obj_instantiate_histogram, object->lookup_jif);
@@ -625,7 +635,9 @@ static void fscache_drop_object(struct fscache_object *object)
 	list_del_init(&object->cache_link);
 	spin_unlock(&cache->object_list_lock);
 
+	fscache_stat(&fscache_n_cop_drop_object);
 	cache->ops->drop_object(object);
+	fscache_stat_d(&fscache_n_cop_drop_object);
 
 	if (parent) {
 		_debug("release parent OBJ%x {%d}",
@@ -640,7 +652,9 @@ static void fscache_drop_object(struct fscache_object *object)
 	}
 
 	/* this just shifts the object release to the slow work processor */
+	fscache_stat(&fscache_n_cop_put_object);
 	object->cache->ops->put_object(object);
+	fscache_stat_d(&fscache_n_cop_put_object);
 
 	_leave("");
 }
@@ -730,8 +744,12 @@ static int fscache_object_slow_work_get_ref(struct slow_work *work)
 {
 	struct fscache_object *object =
 		container_of(work, struct fscache_object, work);
+	int ret;
 
-	return object->cache->ops->grab_object(object) ? 0 : -EAGAIN;
+	fscache_stat(&fscache_n_cop_grab_object);
+	ret = object->cache->ops->grab_object(object) ? 0 : -EAGAIN;
+	fscache_stat_d(&fscache_n_cop_grab_object);
+	return ret;
 }
 
 /*
@@ -742,7 +760,9 @@ static void fscache_object_slow_work_put_ref(struct slow_work *work)
 	struct fscache_object *object =
 		container_of(work, struct fscache_object, work);
 
-	return object->cache->ops->put_object(object);
+	fscache_stat(&fscache_n_cop_put_object);
+	object->cache->ops->put_object(object);
+	fscache_stat_d(&fscache_n_cop_put_object);
 }
 
 /*
@@ -779,7 +799,9 @@ static void fscache_enqueue_dependents(struct fscache_object *object)
 
 		/* sort onto appropriate lists */
 		fscache_enqueue_object(dep);
+		fscache_stat(&fscache_n_cop_put_object);
 		dep->cache->ops->put_object(dep);
+		fscache_stat_d(&fscache_n_cop_put_object);
 
 		if (!list_empty(&object->dependents))
 			cond_resched_lock(&object->lock);
fs/fscache/page.c +23 -6
index c5973e3..250dfd3 100644
@@ -71,7 +71,9 @@ static void fscache_attr_changed_op(struct fscache_operation *op)
 
 	if (fscache_object_is_active(object)) {
 		fscache_set_op_state(op, "CallFS");
+		fscache_stat(&fscache_n_cop_attr_changed);
 		ret = object->cache->ops->attr_changed(object);
+		fscache_stat_d(&fscache_n_cop_attr_changed);
 		fscache_set_op_state(op, "Done");
 		if (ret < 0)
 			fscache_abort_object(object);
@@ -300,11 +302,15 @@ int __fscache_read_or_alloc_page(struct fscache_cookie *cookie,
 
 	/* ask the cache to honour the operation */
 	if (test_bit(FSCACHE_COOKIE_NO_DATA_YET, &object->cookie->flags)) {
+		fscache_stat(&fscache_n_cop_allocate_page);
 		ret = object->cache->ops->allocate_page(op, page, gfp);
+		fscache_stat_d(&fscache_n_cop_allocate_page);
 		if (ret == 0)
 			ret = -ENODATA;
 	} else {
+		fscache_stat(&fscache_n_cop_read_or_alloc_page);
 		ret = object->cache->ops->read_or_alloc_page(op, page, gfp);
+		fscache_stat_d(&fscache_n_cop_read_or_alloc_page);
 	}
 
 	if (ret == -ENOMEM)
@@ -358,7 +364,6 @@ int __fscache_read_or_alloc_pages(struct fscache_cookie *cookie,
 				  void *context,
 				  gfp_t gfp)
 {
-	fscache_pages_retrieval_func_t func;
 	struct fscache_retrieval *op;
 	struct fscache_object *object;
 	int ret;
@@ -413,11 +418,17 @@ int __fscache_read_or_alloc_pages(struct fscache_cookie *cookie,
 	}
 
 	/* ask the cache to honour the operation */
-	if (test_bit(FSCACHE_COOKIE_NO_DATA_YET, &object->cookie->flags))
-		func = object->cache->ops->allocate_pages;
-	else
-		func = object->cache->ops->read_or_alloc_pages;
-	ret = func(op, pages, nr_pages, gfp);
+	if (test_bit(FSCACHE_COOKIE_NO_DATA_YET, &object->cookie->flags)) {
+		fscache_stat(&fscache_n_cop_allocate_pages);
+		ret = object->cache->ops->allocate_pages(
+			op, pages, nr_pages, gfp);
+		fscache_stat_d(&fscache_n_cop_allocate_pages);
+	} else {
+		fscache_stat(&fscache_n_cop_read_or_alloc_pages);
+		ret = object->cache->ops->read_or_alloc_pages(
+			op, pages, nr_pages, gfp);
+		fscache_stat_d(&fscache_n_cop_read_or_alloc_pages);
+	}
 
 	if (ret == -ENOMEM)
 		fscache_stat(&fscache_n_retrievals_nomem);
@@ -500,7 +511,9 @@ int __fscache_alloc_page(struct fscache_cookie *cookie,
 	}
 
 	/* ask the cache to honour the operation */
+	fscache_stat(&fscache_n_cop_allocate_page);
 	ret = object->cache->ops->allocate_page(op, page, gfp);
+	fscache_stat_d(&fscache_n_cop_allocate_page);
 
 	if (ret < 0)
 		fscache_stat(&fscache_n_allocs_nobufs);
@@ -578,7 +591,9 @@ static void fscache_write_op(struct fscache_operation *_op)
 
 	if (page) {
 		fscache_set_op_state(&op->op, "Store");
+		fscache_stat(&fscache_n_cop_write_page);
 		ret = object->cache->ops->write_page(op, page);
+		fscache_stat_d(&fscache_n_cop_write_page);
 		fscache_set_op_state(&op->op, "EndWrite");
 		fscache_end_page_write(cookie, page);
 		page_cache_release(page);
@@ -786,7 +801,9 @@ void __fscache_uncache_page(struct fscache_cookie *cookie, struct page *page)
 	if (TestClearPageFsCache(page) &&
 	    object->cache->ops->uncache_page) {
 		/* the cache backend releases the cookie lock */
+		fscache_stat(&fscache_n_cop_uncache_page);
 		object->cache->ops->uncache_page(object, page);
+		fscache_stat_d(&fscache_n_cop_uncache_page);
 		goto done;
 	}
 
fs/fscache/stats.c +37 -0
index 65deb99..20233fb 100644
@@ -93,6 +93,23 @@ atomic_t fscache_n_checkaux_okay;
 atomic_t fscache_n_checkaux_update;
 atomic_t fscache_n_checkaux_obsolete;
 
+atomic_t fscache_n_cop_alloc_object;
+atomic_t fscache_n_cop_lookup_object;
+atomic_t fscache_n_cop_lookup_complete;
+atomic_t fscache_n_cop_grab_object;
+atomic_t fscache_n_cop_update_object;
+atomic_t fscache_n_cop_drop_object;
+atomic_t fscache_n_cop_put_object;
+atomic_t fscache_n_cop_sync_cache;
+atomic_t fscache_n_cop_attr_changed;
+atomic_t fscache_n_cop_read_or_alloc_page;
+atomic_t fscache_n_cop_read_or_alloc_pages;
+atomic_t fscache_n_cop_allocate_page;
+atomic_t fscache_n_cop_allocate_pages;
+atomic_t fscache_n_cop_write_page;
+atomic_t fscache_n_cop_uncache_page;
+atomic_t fscache_n_cop_dissociate_pages;
+
 /*
  * display the general statistics
  */
@@ -192,6 +209,26 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_op_deferred_release),
 		   atomic_read(&fscache_n_op_release),
 		   atomic_read(&fscache_n_op_gc));
+
+	seq_printf(m, "CacheOp: alo=%d luo=%d luc=%d gro=%d\n",
+		   atomic_read(&fscache_n_cop_alloc_object),
+		   atomic_read(&fscache_n_cop_lookup_object),
+		   atomic_read(&fscache_n_cop_lookup_complete),
+		   atomic_read(&fscache_n_cop_grab_object));
+	seq_printf(m, "CacheOp: upo=%d dro=%d pto=%d atc=%d syn=%d\n",
+		   atomic_read(&fscache_n_cop_update_object),
+		   atomic_read(&fscache_n_cop_drop_object),
+		   atomic_read(&fscache_n_cop_put_object),
+		   atomic_read(&fscache_n_cop_attr_changed),
+		   atomic_read(&fscache_n_cop_sync_cache));
+	seq_printf(m, "CacheOp: rap=%d ras=%d alp=%d als=%d wrp=%d ucp=%d dsp=%d\n",
+		   atomic_read(&fscache_n_cop_read_or_alloc_page),
+		   atomic_read(&fscache_n_cop_read_or_alloc_pages),
+		   atomic_read(&fscache_n_cop_allocate_page),
+		   atomic_read(&fscache_n_cop_allocate_pages),
+		   atomic_read(&fscache_n_cop_write_page),
+		   atomic_read(&fscache_n_cop_uncache_page),
+		   atomic_read(&fscache_n_cop_dissociate_pages));
 	return 0;
 }
 
author maintainer
#12 lore raw

PATCH 11/28 FS-Cache: Clear netfs pointers in cookie after detaching object, not before

In reply to David Howells

Clear the pointers from the fscache_cookie struct to netfs private data after clearing the pointer to the cookie from the fscache_object struct and releasing the object lock, rather than before.

This allows the netfs private data pointers to be relied on simply by holding the object lock, rather than having to hold the cookie lock. This is makes things simpler as the cookie lock has to be taken before the object lock, but sometimes the object pointer is all that the code has.

---

 fs/fscache/cookie.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)
fs/fscache/cookie.c +4 -4
index 432482e..b1870a6 100644
@@ -437,12 +437,8 @@ void __fscache_relinquish_cookie(struct fscache_cookie *cookie, int retire)
 
 	event = retire ? FSCACHE_OBJECT_EV_RETIRE : FSCACHE_OBJECT_EV_RELEASE;
 
-	/* detach pointers back to the netfs */
 	spin_lock(&cookie->lock);
 
-	cookie->netfs_data	= NULL;
-	cookie->def		= NULL;
-
 	/* break links with all the active objects */
 	while (!hlist_empty(&cookie->backing_objects)) {
 		object = hlist_entry(cookie->backing_objects.first,
@@ -465,6 +461,10 @@ void __fscache_relinquish_cookie(struct fscache_cookie *cookie, int retire)
 			BUG();
 	}
 
+	/* detach pointers back to the netfs */
+	cookie->netfs_data	= NULL;
+	cookie->def		= NULL;
+
 	spin_unlock(&cookie->lock);
 
 	if (cookie->parent) {
author maintainer
#13 lore raw

PATCH 12/28 FS-Cache: Use radix tree preload correctly in tracking of pages to be stored

In reply to David Howells

__fscache_write_page() attempts to load the radix tree preallocation pool for the CPU it is on before calling radix_tree_insert(), as the insertion must be done inside a pair of spinlocks.

Use of the preallocation pool, however, is contingent on the radix tree being initialised without __GFP_WAIT specified. __fscache_acquire_cookie() was passing GFP_NOFS to INIT_RADIX_TREE() - but that includes __GFP_WAIT.

The solution is to AND out __GFP_WAIT.

Additionally, the banner comment to radix_tree_preload() is altered to make note of this prerequisite. Possibly there should be a WARN_ON() too.

Without this fix, I have seen the following recursive deadlock caused by radix_tree_insert() attempting to allocate memory inside the spinlocked region, which resulted in FS-Cache being called back into to release memory - which required the spinlock already held.

=============================================
[ INFO: possible recursive locking detected ]
2.6.32-rc6-cachefs #24
---------------------------------------------
nfsiod/7916 is trying to acquire lock:
(&cookie->lock){+.+.-.}, at: [<ffffffffa0076872>] __fscache_uncache_page+0xdb/0x160 [fscache]

but task is already holding lock:
(&cookie->lock){+.+.-.}, at: [<ffffffffa0076acc>] __fscache_write_page+0x15c/0x3f3 [fscache]

other info that might help us debug this:
5 locks held by nfsiod/7916:

 #0:  (nfsiod){+.+.+.}, at: [<ffffffff81048290>] worker_thread+0x19a/0x2e2
 #1:  (&task->u.tk_work#2){+.+.+.}, at: [<ffffffff81048290>] worker_thread+0x19a/0x2e2
 #2:  (&cookie->lock){+.+.-.}, at: [<ffffffffa0076acc>] __fscache_write_page+0x15c/0x3f3 [fscache]
 #3:  (&object->lock#2){+.+.-.}, at: [<ffffffffa0076b07>] __fscache_write_page+0x197/0x3f3 [fscache]
 #4:  (&cookie->stores_lock){+.+...}, at: [<ffffffffa0076b0f>] __fscache_write_page+0x19f/0x3f3 [fscache]
stack backtrace:
Pid: 7916, comm: nfsiod Not tainted 2.6.32-rc6-cachefs #24
Call Trace:
 [<ffffffff8105ac7f>] __lock_acquire+0x1649/0x16e3
 [<ffffffff81059ded>] ? __lock_acquire+0x7b7/0x16e3
 [<ffffffff8100e27d>] ? dump_trace+0x248/0x257
 [<ffffffff8105ad70>] lock_acquire+0x57/0x6d
 [<ffffffffa0076872>] ? __fscache_uncache_page+0xdb/0x160 [fscache]
 [<ffffffff8135467c>] _spin_lock+0x2c/0x3b
 [<ffffffffa0076872>] ? __fscache_uncache_page+0xdb/0x160 [fscache]
 [<ffffffffa0076872>] __fscache_uncache_page+0xdb/0x160 [fscache]
 [<ffffffffa0077eb7>] ? __fscache_check_page_write+0x0/0x71 [fscache]
 [<ffffffffa00b4755>] nfs_fscache_release_page+0x86/0xc4 [nfs]
 [<ffffffffa00907f0>] nfs_release_page+0x3c/0x41 [nfs]
 [<ffffffff81087ffb>] try_to_release_page+0x32/0x3b
 [<ffffffff81092c2b>] shrink_page_list+0x316/0x4ac
 [<ffffffff81058a9b>] ? mark_held_locks+0x52/0x70
 [<ffffffff8135451b>] ? _spin_unlock_irq+0x2b/0x31
 [<ffffffff81093153>] shrink_inactive_list+0x392/0x67c
 [<ffffffff81058a9b>] ? mark_held_locks+0x52/0x70
 [<ffffffff810934ca>] shrink_list+0x8d/0x8f
 [<ffffffff81093744>] shrink_zone+0x278/0x33c
 [<ffffffff81052c70>] ? ktime_get_ts+0xad/0xba
 [<ffffffff8109453b>] try_to_free_pages+0x22e/0x392
 [<ffffffff8109184c>] ? isolate_pages_global+0x0/0x212
 [<ffffffff8108e16b>] __alloc_pages_nodemask+0x3dc/0x5cf
 [<ffffffff810ae24a>] cache_alloc_refill+0x34d/0x6c1
 [<ffffffff811bcf74>] ? radix_tree_node_alloc+0x52/0x5c
 [<ffffffff810ae929>] kmem_cache_alloc+0xb2/0x118
 [<ffffffff811bcf74>] radix_tree_node_alloc+0x52/0x5c
 [<ffffffff811bcfd5>] radix_tree_insert+0x57/0x19c
 [<ffffffffa0076b53>] __fscache_write_page+0x1e3/0x3f3 [fscache]
 [<ffffffffa00b4248>] __nfs_readpage_to_fscache+0x58/0x11e [nfs]
 [<ffffffffa009bb77>] nfs_readpage_release+0x34/0x9b [nfs]
 [<ffffffffa009c0d9>] nfs_readpage_release_full+0x32/0x4b [nfs]
 [<ffffffffa0006cff>] rpc_release_calldata+0x12/0x14 [sunrpc]
 [<ffffffffa0006e2d>] rpc_free_task+0x59/0x61 [sunrpc]
 [<ffffffffa0006f03>] rpc_async_release+0x10/0x12 [sunrpc]
 [<ffffffff810482e5>] worker_thread+0x1ef/0x2e2
 [<ffffffff81048290>] ? worker_thread+0x19a/0x2e2
 [<ffffffff81352433>] ? thread_return+0x3e/0x101
 [<ffffffffa0006ef3>] ? rpc_async_release+0x0/0x12 [sunrpc]
 [<ffffffff8104bff5>] ? autoremove_wake_function+0x0/0x34
 [<ffffffff81058d25>] ? trace_hardirqs_on+0xd/0xf
 [<ffffffff810480f6>] ? worker_thread+0x0/0x2e2
 [<ffffffff8104bd21>] kthread+0x7a/0x82
 [<ffffffff8100beda>] child_rip+0xa/0x20
 [<ffffffff8100b87c>] ? restore_args+0x0/0x30
 [<ffffffff8104c2b9>] ? add_wait_queue+0x15/0x44
 [<ffffffff8104bca7>] ? kthread+0x0/0x82
 [<ffffffff8100bed0>] ? child_rip+0x0/0x20

---

 fs/fscache/cookie.c |    4 +++-
 lib/radix-tree.c    |    3 +++
 2 files changed, 6 insertions(+), 1 deletions(-)
fs/fscache/cookie.c +3 -1
index b1870a6..e6854f5 100644
@@ -102,7 +102,9 @@ struct fscache_cookie *__fscache_acquire_cookie(
 	cookie->netfs_data	= netfs_data;
 	cookie->flags		= 0;
 
-	INIT_RADIX_TREE(&cookie->stores, GFP_NOFS);
+	/* radix tree insertion won't use the preallocation pool unless it's
+	 * told it may not wait */
+	INIT_RADIX_TREE(&cookie->stores, GFP_NOFS & ~__GFP_WAIT);
 
 	switch (cookie->def->type) {
 	case FSCACHE_COOKIE_TYPE_INDEX:
lib/radix-tree.c +3 -0
index 23abbd9..ae61068 100644
@@ -200,6 +200,9 @@ radix_tree_node_free(struct radix_tree_node *node)
  * ensure that the addition of a single element in the tree cannot fail.  On
  * success, return zero, with preemption disabled.  On error, return -ENOMEM
  * with preemption not disabled.
+ *
+ * To make use of this facility, the radix tree must be initialised without
+ * __GFP_WAIT being passed to INIT_RADIX_TREE().
  */
 int radix_tree_preload(gfp_t gfp_mask)
 {
author maintainer
#14 lore raw

PATCH 13/28 FS-Cache: Permit cache retrieval ops to be interrupted in the initial wait phase

In reply to David Howells

Permit the operations to retrieve data from the cache or to allocate space in the cache for future writes to be interrupted whilst they're waiting for permission for the operation to proceed. Typically this wait occurs whilst the cache object is being looked up on disk in the background.

If an interruption occurs, and the operation has not yet been given the go-ahead to run, the operation is dequeued and cancelled, and control returns to the read operation of the netfs routine with none of the requested pages having been read or in any way marked as known by the cache.

This means that the initial wait is done interruptibly rather than uninterruptibly.

In addition, extra stats values are made available to show the number of ops cancelled and the number of cache space allocations interrupted.

---

 Documentation/filesystems/caching/fscache.txt |    2 +
 fs/fscache/internal.h                         |    3 +
 fs/fscache/operation.c                        |   82 ++++++++++++++++---------
 fs/fscache/page.c                             |   55 +++++++++++++++--
 fs/fscache/stats.c                            |   12 ++--
 5 files changed, 115 insertions(+), 39 deletions(-)
Documentation/filesystems/caching/fscache.txt +2 -0
index 472398f..21d20bf 100644
@@ -250,6 +250,7 @@ proc files.
 		ok=N	Number of successful alloc reqs
 		wt=N	Number of alloc reqs that waited on lookup completion
 		nbf=N	Number of alloc reqs rejected -ENOBUFS
+		int=N	Number of alloc reqs aborted -ERESTARTSYS
 		ops=N	Number of alloc reqs submitted
 		owt=N	Number of alloc reqs waited for CPU time
 	Retrvls	n=N	Number of retrieval (read) requests seen
@@ -271,6 +272,7 @@ proc files.
 	Ops	pend=N	Number of times async ops added to pending queues
 		run=N	Number of times async ops given CPU time
 		enq=N	Number of times async ops queued for processing
+		can=N	Number of async ops cancelled
 		dfr=N	Number of async ops queued for deferred release
 		rel=N	Number of async ops released
 		gc=N	Number of deferred-release async ops garbage collected
fs/fscache/internal.h +3 -0
index b85cc89..50324ad 100644
@@ -112,6 +112,7 @@ extern int fscache_submit_exclusive_op(struct fscache_object *,
 				       struct fscache_operation *);
 extern int fscache_submit_op(struct fscache_object *,
 			     struct fscache_operation *);
+extern int fscache_cancel_op(struct fscache_operation *);
 extern void fscache_abort_object(struct fscache_object *);
 extern void fscache_start_operations(struct fscache_object *);
 extern void fscache_operation_gc(struct work_struct *);
@@ -140,6 +141,7 @@ extern atomic_t fscache_n_op_enqueue;
 extern atomic_t fscache_n_op_deferred_release;
 extern atomic_t fscache_n_op_release;
 extern atomic_t fscache_n_op_gc;
+extern atomic_t fscache_n_op_cancelled;
 
 extern atomic_t fscache_n_attr_changed;
 extern atomic_t fscache_n_attr_changed_ok;
@@ -151,6 +153,7 @@ extern atomic_t fscache_n_allocs;
 extern atomic_t fscache_n_allocs_ok;
 extern atomic_t fscache_n_allocs_wait;
 extern atomic_t fscache_n_allocs_nobufs;
+extern atomic_t fscache_n_allocs_intr;
 extern atomic_t fscache_n_alloc_ops;
 extern atomic_t fscache_n_alloc_op_waits;
 
fs/fscache/operation.c +54 -28

116 lines not on this page · raw message

fs/fscache/page.c +48 -7

91 lines not on this page · raw message

fs/fscache/stats.c +8 -4
index 20233fb..4c07439 100644
@@ -25,6 +25,7 @@ atomic_t fscache_n_op_requeue;
 atomic_t fscache_n_op_deferred_release;
 atomic_t fscache_n_op_release;
 atomic_t fscache_n_op_gc;
+atomic_t fscache_n_op_cancelled;
 
 atomic_t fscache_n_attr_changed;
 atomic_t fscache_n_attr_changed_ok;
@@ -36,6 +37,7 @@ atomic_t fscache_n_allocs;
 atomic_t fscache_n_allocs_ok;
 atomic_t fscache_n_allocs_wait;
 atomic_t fscache_n_allocs_nobufs;
+atomic_t fscache_n_allocs_intr;
 atomic_t fscache_n_alloc_ops;
 atomic_t fscache_n_alloc_op_waits;
 
@@ -169,11 +171,12 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_attr_changed_nomem),
 		   atomic_read(&fscache_n_attr_changed_calls));
 
-	seq_printf(m, "Allocs : n=%u ok=%u wt=%u nbf=%u\n",
+	seq_printf(m, "Allocs : n=%u ok=%u wt=%u nbf=%u int=%u\n",
 		   atomic_read(&fscache_n_allocs),
 		   atomic_read(&fscache_n_allocs_ok),
 		   atomic_read(&fscache_n_allocs_wait),
-		   atomic_read(&fscache_n_allocs_nobufs));
+		   atomic_read(&fscache_n_allocs_nobufs),
+		   atomic_read(&fscache_n_allocs_intr));
 	seq_printf(m, "Allocs : ops=%u owt=%u\n",
 		   atomic_read(&fscache_n_alloc_ops),
 		   atomic_read(&fscache_n_alloc_op_waits));
@@ -201,10 +204,11 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_store_ops),
 		   atomic_read(&fscache_n_store_calls));
 
-	seq_printf(m, "Ops    : pend=%u run=%u enq=%u\n",
+	seq_printf(m, "Ops    : pend=%u run=%u enq=%u can=%u\n",
 		   atomic_read(&fscache_n_op_pend),
 		   atomic_read(&fscache_n_op_run),
-		   atomic_read(&fscache_n_op_enqueue));
+		   atomic_read(&fscache_n_op_enqueue),
+		   atomic_read(&fscache_n_op_cancelled));
 	seq_printf(m, "Ops    : dfr=%u rel=%u gc=%u\n",
 		   atomic_read(&fscache_n_op_deferred_release),
 		   atomic_read(&fscache_n_op_release),
author maintainer
#15 lore raw

PATCH 14/28 FS-Cache: The object-available state can't rely on the cookie to be available

In reply to David Howells

The object-available state in the object processing state machine (as processed by fscache_object_available()) can't rely on the cookie to be available because the FSCACHE_COOKIE_CREATING bit may have been cleared by fscache_obtained_object() prior to the object being put into the FSCACHE_OBJECT_AVAILABLE state.

Clearing the FSCACHE_COOKIE_CREATING bit on a cookie permits __fscache_relinquish_cookie() to proceed and detach the cookie from the object.

To deal with this, we don't dereference object->cookie in fscache_object_available() if the object has already been detached.

In addition, a couple of assertions are added into fscache_drop_object() to make sure the object is unbound from the cookie before it gets there.

---

 fs/fscache/object.c |    9 +++++++--
 1 files changed, 7 insertions(+), 2 deletions(-)
fs/fscache/object.c +7 -2
index 0d65c0c..1a1afa8 100644
@@ -158,7 +158,8 @@ static void fscache_object_state_machine(struct fscache_object *object)
 
 		spin_lock(&object->lock);
 		object->state = FSCACHE_OBJECT_DYING;
-		if (test_and_clear_bit(FSCACHE_COOKIE_CREATING,
+		if (object->cookie &&
+		    test_and_clear_bit(FSCACHE_COOKIE_CREATING,
 				       &object->cookie->flags))
 			wake_up_bit(&object->cookie->flags,
 				    FSCACHE_COOKIE_CREATING);
@@ -594,7 +595,8 @@ static void fscache_object_available(struct fscache_object *object)
 
 	spin_lock(&object->lock);
 
-	if (test_and_clear_bit(FSCACHE_COOKIE_CREATING, &object->cookie->flags))
+	if (object->cookie &&
+	    test_and_clear_bit(FSCACHE_COOKIE_CREATING, &object->cookie->flags))
 		wake_up_bit(&object->cookie->flags, FSCACHE_COOKIE_CREATING);
 
 	fscache_done_parent_op(object);
@@ -631,6 +633,9 @@ static void fscache_drop_object(struct fscache_object *object)
 
 	_enter("{OBJ%x,%d}", object->debug_id, object->n_children);
 
+	ASSERTCMP(object->cookie, ==, NULL);
+	ASSERT(hlist_unhashed(&object->cookie_link));
+
 	spin_lock(&cache->object_list_lock);
 	list_del_init(&object->cache_link);
 	spin_unlock(&cache->object_list_lock);
author maintainer
#16 lore raw

PATCH 15/28 FS-Cache: Fix lock misorder in fscache_write_op()

In reply to David Howells

FS-Cache has two structs internally for keeping track of the internal state of a cached file: the fscache_cookie struct, which represents the netfs's state, and fscache_object struct, which represents the cache's state. Each has a pointer that points to the other (when both are in existence), and each has a spinlock for pointer maintenance.

Since netfs operations approach these structures from the cookie side, they get the cookie lock first, then the object lock. Cache operations, on the other hand, approach from the object side, and get the object lock first. It is not then permitted for a cache operation to get the cookie lock whilst it is holding the object lock lest deadlock occur; instead, it must do one of two things:

(1) increment the cookie usage counter, drop the object lock and then get both locks in order, or

(2) simply hold the object lock as certain parts of the cookie may not be altered whilst the object lock is held.

It is also not permitted to follow either pointer without holding the lock at the end you start with. To break the pointers between the cookie and the object, both locks must be held.

fscache_write_op(), however, violates the locking rules: It attempts to get the cookie lock without (a) checking that the cookie pointer is a valid pointer, and (b) holding the object lock to protect the cookie pointer whilst it follows it. This is so that it can access the pending page store tree without interference from __fscache_write_page().

This is fixed by splitting the cookie lock, such that the page store tracking tree is protected by its own lock, and checking that the cookie pointer is non-NULL before we attempt to follow it whilst holding the object lock.

The new lock is subordinate to both the cookie lock and the object lock, and so should be taken after those.

---

 Documentation/filesystems/caching/fscache.txt |    3 +
 fs/fscache/cookie.c                           |    1 
 fs/fscache/internal.h                         |    4 ++
 fs/fscache/page.c                             |   52 +++++++++++++++++--------
 fs/fscache/stats.c                            |   10 ++++-
 include/linux/fscache-cache.h                 |    1 
 6 files changed, 52 insertions(+), 19 deletions(-)
Documentation/filesystems/caching/fscache.txt +3 -0
index 21d20bf..cd2b4c9 100644
@@ -269,6 +269,9 @@ proc files.
 		oom=N	Number of store reqs failed -ENOMEM
 		ops=N	Number of store reqs submitted
 		run=N	Number of store reqs granted CPU time
+		pgs=N	Number of pages given store req processing time
+		rxd=N	Number of store reqs deleted from tracking tree
+		olm=N	Number of store reqs over store limit
 	Ops	pend=N	Number of times async ops added to pending queues
 		run=N	Number of times async ops given CPU time
 		enq=N	Number of times async ops queued for processing
fs/fscache/cookie.c +1 -0
index e6854f5..f979659 100644
@@ -36,6 +36,7 @@ void fscache_cookie_init_once(void *_cookie)
 
 	memset(cookie, 0, sizeof(*cookie));
 	spin_lock_init(&cookie->lock);
+	spin_lock_init(&cookie->stores_lock);
 	INIT_HLIST_HEAD(&cookie->backing_objects);
 }
 
fs/fscache/internal.h +4 -0
index 50324ad..ba1853f 100644
@@ -17,6 +17,7 @@
  * - cache->object_list_lock
  * - object->lock
  * - object->parent->lock
+ * - cookie->stores_lock
  * - fscache_thread_lock
  *
  */
@@ -174,6 +175,9 @@ extern atomic_t fscache_n_stores_nobufs;
 extern atomic_t fscache_n_stores_oom;
 extern atomic_t fscache_n_store_ops;
 extern atomic_t fscache_n_store_calls;
+extern atomic_t fscache_n_store_pages;
+extern atomic_t fscache_n_store_radix_deletes;
+extern atomic_t fscache_n_store_pages_over_limit;
 
 extern atomic_t fscache_n_marks;
 extern atomic_t fscache_n_uncaches;
fs/fscache/page.c +35 -17

131 lines not on this page · raw message

fs/fscache/stats.c +8 -2
index 4c07439..1d53ea6 100644
@@ -58,6 +58,9 @@ atomic_t fscache_n_stores_nobufs;
 atomic_t fscache_n_stores_oom;
 atomic_t fscache_n_store_ops;
 atomic_t fscache_n_store_calls;
+atomic_t fscache_n_store_pages;
+atomic_t fscache_n_store_radix_deletes;
+atomic_t fscache_n_store_pages_over_limit;
 
 atomic_t fscache_n_marks;
 atomic_t fscache_n_uncaches;
@@ -200,9 +203,12 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_stores_again),
 		   atomic_read(&fscache_n_stores_nobufs),
 		   atomic_read(&fscache_n_stores_oom));
-	seq_printf(m, "Stores : ops=%u run=%u\n",
+	seq_printf(m, "Stores : ops=%u run=%u pgs=%u rxd=%u olm=%u\n",
 		   atomic_read(&fscache_n_store_ops),
-		   atomic_read(&fscache_n_store_calls));
+		   atomic_read(&fscache_n_store_calls),
+		   atomic_read(&fscache_n_store_pages),
+		   atomic_read(&fscache_n_store_radix_deletes),
+		   atomic_read(&fscache_n_store_pages_over_limit));
 
 	seq_printf(m, "Ops    : pend=%u run=%u enq=%u can=%u\n",
 		   atomic_read(&fscache_n_op_pend),
include/linux/fscache-cache.h +1 -0
index 184cbdf..f3aa4bd 100644
@@ -310,6 +310,7 @@ struct fscache_cookie {
 	atomic_t			usage;		/* number of users of this cookie */
 	atomic_t			n_children;	/* number of children of this cookie */
 	spinlock_t			lock;
+	spinlock_t			stores_lock;	/* lock on page store tree */
 	struct hlist_head		backing_objects; /* object(s) backing this file/index */
 	const struct fscache_cookie_def	*def;		/* definition */
 	struct fscache_cookie		*parent;	/* parent of this entry */
author maintainer
#17 lore raw

PATCH 16/28 FS-Cache: Don't delete pending pages from the page-store tracking tree

In reply to David Howells

Don't delete pending pages from the page-store tracking tree, but rather send them for another write as they've presumably been updated.

---

 fs/fscache/page.c |    7 +++++--
 lib/radix-tree.c  |    2 --
 2 files changed, 5 insertions(+), 4 deletions(-)
fs/fscache/page.c +5 -2
index 3ea8897..022a5da 100644
@@ -57,8 +57,11 @@ static void fscache_end_page_write(struct fscache_object *object,
 		/* delete the page from the tree if it is now no longer
 		 * pending */
 		spin_lock(&cookie->stores_lock);
-		fscache_stat(&fscache_n_store_radix_deletes);
-		xpage = radix_tree_delete(&cookie->stores, page->index);
+		if (!radix_tree_tag_get(&cookie->stores, page->index,
+					FSCACHE_COOKIE_PENDING_TAG)) {
+			fscache_stat(&fscache_n_store_radix_deletes);
+			xpage = radix_tree_delete(&cookie->stores, page->index);
+		}
 		spin_unlock(&cookie->stores_lock);
 		wake_up_bit(&cookie->flags, 0);
 	}
lib/radix-tree.c +0 -2
index ae61068..92cdd99 100644
@@ -546,7 +546,6 @@ out:
 }
 EXPORT_SYMBOL(radix_tree_tag_clear);
 
-#ifndef __KERNEL__	/* Only the test harness uses this at present */
 /**
  * radix_tree_tag_get - get a tag on a radix tree node
  * @root:		radix tree root
@@ -609,7 +608,6 @@ int radix_tree_tag_get(struct radix_tree_root *root,
 	}
 }
 EXPORT_SYMBOL(radix_tree_tag_get);
-#endif
 
 /**
  *	radix_tree_next_hole    -    find the next hole (not-present entry)
author maintainer
#18 lore raw

PATCH 17/28 FS-Cache: Handle read request vs lookup, creation or other cache failure

In reply to David Howells

FS-Cache doesn't correctly handle the netfs requesting a read from the cache on an object that failed or was withdrawn by the cache. A trace similar to the following might be seen:

	CacheFiles: Lookup failed error -105
	[exe   ] unexpected submission OP165afe [OBJ6cac OBJECT_LC_DYING]
	[exe   ] objstate=OBJECT_LC_DYING [OBJECT_LC_DYING]
	[exe   ] objflags=0
	[exe   ] objevent=9 [fffffffffffffffb]
	[exe   ] ops=0 inp=0 exc=0
	Pid: 6970, comm: exe Not tainted 2.6.32-rc6-cachefs #50
	Call Trace:
	 [<ffffffffa0076477>] fscache_submit_op+0x3ff/0x45a [fscache]
	 [<ffffffffa0077997>] __fscache_read_or_alloc_pages+0x187/0x3c4 [fscache]
	 [<ffffffffa00b6480>] ? nfs_readpage_from_fscache_complete+0x0/0x66 [nfs]
	 [<ffffffffa00b6388>] __nfs_readpages_from_fscache+0x7e/0x176 [nfs]
	 [<ffffffff8108e483>] ? __alloc_pages_nodemask+0x11c/0x5cf
	 [<ffffffffa009d796>] nfs_readpages+0x114/0x1d7 [nfs]
	 [<ffffffff81090314>] __do_page_cache_readahead+0x15f/0x1ec
	 [<ffffffff81090228>] ? __do_page_cache_readahead+0x73/0x1ec
	 [<ffffffff810903bd>] ra_submit+0x1c/0x20
	 [<ffffffff810906bb>] ondemand_readahead+0x227/0x23a
	 [<ffffffff81090762>] page_cache_sync_readahead+0x17/0x19
	 [<ffffffff8108a99e>] generic_file_aio_read+0x236/0x5a0
	 [<ffffffffa00937bd>] nfs_file_read+0xe4/0xf3 [nfs]
	 [<ffffffff810b2fa2>] do_sync_read+0xe3/0x120
	 [<ffffffff81354cc3>] ? _spin_unlock_irq+0x2b/0x31
	 [<ffffffff8104c0f1>] ? autoremove_wake_function+0x0/0x34
	 [<ffffffff811848e5>] ? selinux_file_permission+0x5d/0x10f
	 [<ffffffff81352bdb>] ? thread_return+0x3e/0x101
	 [<ffffffff8117d7b0>] ? security_file_permission+0x11/0x13
	 [<ffffffff810b3b06>] vfs_read+0xaa/0x16f
	 [<ffffffff81058df0>] ? trace_hardirqs_on_caller+0x10c/0x130
	 [<ffffffff810b3c84>] sys_read+0x45/0x6c
	 [<ffffffff8100ae2b>] system_call_fastpath+0x16/0x1b

The object state might also be OBJECT_DYING or OBJECT_WITHDRAWING.

This should be handled by simply rejecting the new operation with ENOBUFS. There's no need to log an error for it. Events of this type now appear in the stats file under Ops:rej.

---

 Documentation/filesystems/caching/fscache.txt |    1 +
 fs/fscache/internal.h                         |    1 +
 fs/fscache/operation.c                        |    5 +++++
 fs/fscache/stats.c                            |    6 ++++--
 4 files changed, 11 insertions(+), 2 deletions(-)
Documentation/filesystems/caching/fscache.txt +1 -0
index cd2b4c9..8889d6a 100644
@@ -276,6 +276,7 @@ proc files.
 		run=N	Number of times async ops given CPU time
 		enq=N	Number of times async ops queued for processing
 		can=N	Number of async ops cancelled
+		rej=N	Number of async ops rejected due to object lookup/create failure
 		dfr=N	Number of async ops queued for deferred release
 		rel=N	Number of async ops released
 		gc=N	Number of deferred-release async ops garbage collected
fs/fscache/internal.h +1 -0
index ba1853f..a076987 100644
@@ -143,6 +143,7 @@ extern atomic_t fscache_n_op_deferred_release;
 extern atomic_t fscache_n_op_release;
 extern atomic_t fscache_n_op_gc;
 extern atomic_t fscache_n_op_cancelled;
+extern atomic_t fscache_n_op_rejected;
 
 extern atomic_t fscache_n_attr_changed;
 extern atomic_t fscache_n_attr_changed_ok;
fs/fscache/operation.c +5 -0
index 296492e..313e79a 100644
@@ -232,6 +232,11 @@ int fscache_submit_op(struct fscache_object *object,
 		list_add_tail(&op->pend_link, &object->pending_ops);
 		fscache_stat(&fscache_n_op_pend);
 		ret = 0;
+	} else if (object->state == FSCACHE_OBJECT_DYING ||
+		   object->state == FSCACHE_OBJECT_LC_DYING ||
+		   object->state == FSCACHE_OBJECT_WITHDRAWING) {
+		fscache_stat(&fscache_n_op_rejected);
+		ret = -ENOBUFS;
 	} else if (!test_bit(FSCACHE_IOERROR, &object->cache->flags)) {
 		fscache_report_unexpected_submission(object, op, ostate);
 		ASSERT(!fscache_object_is_active(object));
fs/fscache/stats.c +4 -2
index 1d53ea6..045ba39 100644
@@ -26,6 +26,7 @@ atomic_t fscache_n_op_deferred_release;
 atomic_t fscache_n_op_release;
 atomic_t fscache_n_op_gc;
 atomic_t fscache_n_op_cancelled;
+atomic_t fscache_n_op_rejected;
 
 atomic_t fscache_n_attr_changed;
 atomic_t fscache_n_attr_changed_ok;
@@ -210,11 +211,12 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_store_radix_deletes),
 		   atomic_read(&fscache_n_store_pages_over_limit));
 
-	seq_printf(m, "Ops    : pend=%u run=%u enq=%u can=%u\n",
+	seq_printf(m, "Ops    : pend=%u run=%u enq=%u can=%u rej=%u\n",
 		   atomic_read(&fscache_n_op_pend),
 		   atomic_read(&fscache_n_op_run),
 		   atomic_read(&fscache_n_op_enqueue),
-		   atomic_read(&fscache_n_op_cancelled));
+		   atomic_read(&fscache_n_op_cancelled),
+		   atomic_read(&fscache_n_op_rejected));
 	seq_printf(m, "Ops    : dfr=%u rel=%u gc=%u\n",
 		   atomic_read(&fscache_n_op_deferred_release),
 		   atomic_read(&fscache_n_op_release),
author maintainer
#19 lore raw

PATCH 18/28 FS-Cache: Handle pages pending storage that get evicted under OOM conditions

In reply to David Howells

Handle netfs pages that the vmscan algorithm wants to evict from the pagecache under OOM conditions, but that are waiting for write to the cache. Under these conditions, vmscan calls the releasepage() function of the netfs, asking if a page can be discarded.

The problem is typified by the following trace of a stuck process:

	kslowd005     D 0000000000000000     0  4253      2 0x00000080
	 ffff88001b14f370 0000000000000046 ffff880020d0d000 0000000000000007
	 0000000000000006 0000000000000001 ffff88001b14ffd8 ffff880020d0d2a8
	 000000000000ddf0 00000000000118c0 00000000000118c0 ffff880020d0d2a8
	Call Trace:
	 [<ffffffffa00782d8>] __fscache_wait_on_page_write+0x8b/0xa7 [fscache]
	 [<ffffffff8104c0f1>] ? autoremove_wake_function+0x0/0x34
	 [<ffffffffa0078240>] ? __fscache_check_page_write+0x63/0x70 [fscache]
	 [<ffffffffa00b671d>] nfs_fscache_release_page+0x4e/0xc4 [nfs]
	 [<ffffffffa00927f0>] nfs_release_page+0x3c/0x41 [nfs]
	 [<ffffffff810885d3>] try_to_release_page+0x32/0x3b
	 [<ffffffff81093203>] shrink_page_list+0x316/0x4ac
	 [<ffffffff8109372b>] shrink_inactive_list+0x392/0x67c
	 [<ffffffff813532fa>] ? __mutex_unlock_slowpath+0x100/0x10b
	 [<ffffffff81058df0>] ? trace_hardirqs_on_caller+0x10c/0x130
	 [<ffffffff8135330e>] ? mutex_unlock+0x9/0xb
	 [<ffffffff81093aa2>] shrink_list+0x8d/0x8f
	 [<ffffffff81093d1c>] shrink_zone+0x278/0x33c
	 [<ffffffff81052d6c>] ? ktime_get_ts+0xad/0xba
	 [<ffffffff81094b13>] try_to_free_pages+0x22e/0x392
	 [<ffffffff81091e24>] ? isolate_pages_global+0x0/0x212
	 [<ffffffff8108e743>] __alloc_pages_nodemask+0x3dc/0x5cf
	 [<ffffffff81089529>] grab_cache_page_write_begin+0x65/0xaa
	 [<ffffffff8110f8c0>] ext3_write_begin+0x78/0x1eb
	 [<ffffffff81089ec5>] generic_file_buffered_write+0x109/0x28c
	 [<ffffffff8103cb69>] ? current_fs_time+0x22/0x29
	 [<ffffffff8108a509>] __generic_file_aio_write+0x350/0x385
	 [<ffffffff8108a588>] ? generic_file_aio_write+0x4a/0xae
	 [<ffffffff8108a59e>] generic_file_aio_write+0x60/0xae
	 [<ffffffff810b2e82>] do_sync_write+0xe3/0x120
	 [<ffffffff8104c0f1>] ? autoremove_wake_function+0x0/0x34
	 [<ffffffff810b18e1>] ? __dentry_open+0x1a5/0x2b8
	 [<ffffffff810b1a76>] ? dentry_open+0x82/0x89
	 [<ffffffffa00e693c>] cachefiles_write_page+0x298/0x335 [cachefiles]
	 [<ffffffffa0077147>] fscache_write_op+0x178/0x2c2 [fscache]
	 [<ffffffffa0075656>] fscache_op_execute+0x7a/0xd1 [fscache]
	 [<ffffffff81082093>] slow_work_execute+0x18f/0x2d1
	 [<ffffffff8108239a>] slow_work_thread+0x1c5/0x308
	 [<ffffffff8104c0f1>] ? autoremove_wake_function+0x0/0x34
	 [<ffffffff810821d5>] ? slow_work_thread+0x0/0x308
	 [<ffffffff8104be91>] kthread+0x7a/0x82
	 [<ffffffff8100beda>] child_rip+0xa/0x20
	 [<ffffffff8100b87c>] ? restore_args+0x0/0x30
	 [<ffffffff8102ef83>] ? tg_shares_up+0x171/0x227
	 [<ffffffff8104be17>] ? kthread+0x0/0x82
	 [<ffffffff8100bed0>] ? child_rip+0x0/0x20

In the above backtrace, the following is happening:

(1) A page storage operation is being executed by a slow-work thread

     (fscache_write_op()).

(2) FS-Cache farms the operation out to the cache to perform

     (cachefiles_write_page()).

(3) CacheFiles is then calling Ext3 to perform the actual write, using Ext3's standard write (do_sync_write()) under KERNEL_DS directly from the netfs page.

(4) However, for Ext3 to perform the write, it must allocate some memory, in particular, it must allocate at least one page cache page into which it can copy the data from the netfs page.

(5) Under OOM conditions, the memory allocator can't immediately come up with a page, so it uses vmscan to find something to discard (try_to_free_pages()).

(6) vmscan finds a clean netfs page it might be able to discard (possibly the one it's trying to write out).

(7) The netfs is called to throw the page away (nfs_release_page()) - but it's called with __GFP_WAIT, so the netfs decides to wait for the store to complete (__fscache_wait_on_page_write()).

(8) This blocks a slow-work processing thread - possibly against itself.

The system ends up stuck because it can't write out any netfs pages to the cache without allocating more memory.

To avoid this, we make FS-Cache cancel some writes that aren't in the middle of actually being performed. This means that some data won't make it into the cache this time. To support this, a new FS-Cache function is added fscache_maybe_release_page() that replaces what the netfs releasepage() functions used to do with respect to the cache.

The decisions fscache_maybe_release_page() makes are counted and displayed through /proc/fs/fscache/stats on a line labelled "VmScan". There are four counters provided: "nos=N" - pages that weren't pending storage; "gon=N" - pages that were pending storage when we first looked, but weren't by the time we got the object lock; "bsy=N" - pages that we ignored as they were actively being written when we looked; and "can=N" - pages that we cancelled the storage of.

What I'd really like to do is alter the behaviour of the cancellation heuristics, depending on how necessary it is to expel pages. If there are plenty of other pages that aren't waiting to be written to the cache that could be ejected first, then it would be nice to hold up on immediate cancellation of cache writes - but I don't see a way of doing that.

---

 Documentation/filesystems/caching/fscache.txt   |    4 +
 Documentation/filesystems/caching/netfs-api.txt |   21 ++++++
 fs/9p/cache.c                                   |   14 ----
 fs/afs/file.c                                   |   15 +---
 fs/fscache/internal.h                           |    5 +
 fs/fscache/page.c                               |   79 ++++++++++++++++++++++-
 fs/fscache/stats.c                              |   11 +++
 fs/nfs/fscache.c                                |   10 +--
 include/linux/fscache-cache.h                   |    1 
 include/linux/fscache.h                         |   27 ++++++++
 10 files changed, 152 insertions(+), 35 deletions(-)
Documentation/filesystems/caching/fscache.txt +4 -0
index 8889d6a..655569e 100644
@@ -272,6 +272,10 @@ proc files.
 		pgs=N	Number of pages given store req processing time
 		rxd=N	Number of store reqs deleted from tracking tree
 		olm=N	Number of store reqs over store limit
+	VmScan	nos=N	Number of release reqs against pages with no pending store
+		gon=N	Number of release reqs against pages stored by time lock granted
+		bsy=N	Number of release reqs ignored due to in-progress store
+		can=N	Number of page stores cancelled due to release req
 	Ops	pend=N	Number of times async ops added to pending queues
 		run=N	Number of times async ops given CPU time
 		enq=N	Number of times async ops queued for processing
Documentation/filesystems/caching/netfs-api.txt +20 -1
index 2666b1e..1902c57 100644
@@ -641,7 +641,7 @@ data file must be retired (see the relinquish cookie function below).
 
 Furthermore, note that this does not cancel the asynchronous read or write
 operation started by the read/alloc and write functions, so the page
-invalidation and release functions must use:
+invalidation functions must use:
 
 	bool fscache_check_page_write(struct fscache_cookie *cookie,
 				      struct page *page);
@@ -654,6 +654,25 @@ to see if a page is being written to the cache, and:
 to wait for it to finish if it is.
 
 
+When releasepage() is being implemented, a special FS-Cache function exists to
+manage the heuristics of coping with vmscan trying to eject pages, which may
+conflict with the cache trying to write pages to the cache (which may itself
+need to allocate memory):
+
+	bool fscache_maybe_release_page(struct fscache_cookie *cookie,
+					struct page *page,
+					gfp_t gfp);
+
+This takes the netfs cookie, and the page and gfp arguments as supplied to
+releasepage().  It will return false if the page cannot be released yet for
+some reason and if it returns true, the page has been uncached and can now be
+released.
+
+To make a page available for release, this function may wait for an outstanding
+storage request to complete, or it may attempt to cancel the storage request -
+in which case the page will not be stored in the cache this time.
+
+
 ==========================
 INDEX AND DATA FILE UPDATE
 ==========================
fs/9p/cache.c +1 -13
index 51c94e2..bcc5357 100644
@@ -343,18 +343,7 @@ int __v9fs_fscache_release_page(struct page *page, gfp_t gfp)
 
 	BUG_ON(!vcookie->fscache);
 
-	if (PageFsCache(page)) {
-		if (fscache_check_page_write(vcookie->fscache, page)) {
-			if (!(gfp & __GFP_WAIT))
-				return 0;
-			fscache_wait_on_page_write(vcookie->fscache, page);
-		}
-
-		fscache_uncache_page(vcookie->fscache, page);
-		ClearPageFsCache(page);
-	}
-
-	return 1;
+	return fscache_maybe_release_page(vnode->cache, page, gfp);
 }
 
 void __v9fs_fscache_invalidate_page(struct page *page)
@@ -368,7 +357,6 @@ void __v9fs_fscache_invalidate_page(struct page *page)
 		fscache_wait_on_page_write(vcookie->fscache, page);
 		BUG_ON(!PageLocked(page));
 		fscache_uncache_page(vcookie->fscache, page);
-		ClearPageFsCache(page);
 	}
 }
 
fs/afs/file.c +3 -12
index 681c2a7..39b3016 100644
@@ -315,7 +315,6 @@ static void afs_invalidatepage(struct page *page, unsigned long offset)
 			struct afs_vnode *vnode = AFS_FS_I(page->mapping->host);
 			fscache_wait_on_page_write(vnode->cache, page);
 			fscache_uncache_page(vnode->cache, page);
-			ClearPageFsCache(page);
 		}
 #endif
 
@@ -349,17 +348,9 @@ static int afs_releasepage(struct page *page, gfp_t gfp_flags)
 	/* deny if page is being written to the cache and the caller hasn't
 	 * elected to wait */
 #ifdef CONFIG_AFS_FSCACHE
-	if (PageFsCache(page)) {
-		if (fscache_check_page_write(vnode->cache, page)) {
-			if (!(gfp_flags & __GFP_WAIT)) {
-				_leave(" = F [cache busy]");
-				return 0;
-			}
-			fscache_wait_on_page_write(vnode->cache, page);
-		}
-
-		fscache_uncache_page(vnode->cache, page);
-		ClearPageFsCache(page);
+	if (!fscache_maybe_release_page(vnode->cache, page, gfp_flags)) {
+		_leave(" = F [cache busy]");
+		return 0;
 	}
 #endif
 
fs/fscache/internal.h +5 -0
index a076987..e504651 100644
@@ -180,6 +180,11 @@ extern atomic_t fscache_n_store_pages;
 extern atomic_t fscache_n_store_radix_deletes;
 extern atomic_t fscache_n_store_pages_over_limit;
 
+extern atomic_t fscache_n_store_vmscan_not_storing;
+extern atomic_t fscache_n_store_vmscan_gone;
+extern atomic_t fscache_n_store_vmscan_busy;
+extern atomic_t fscache_n_store_vmscan_cancelled;
+
 extern atomic_t fscache_n_marks;
 extern atomic_t fscache_n_uncaches;
 
fs/fscache/page.c +77 -2

97 lines not on this page · raw message

fs/fscache/stats.c +11 -0
index 045ba39..cda6999 100644
@@ -63,6 +63,11 @@ atomic_t fscache_n_store_pages;
 atomic_t fscache_n_store_radix_deletes;
 atomic_t fscache_n_store_pages_over_limit;
 
+atomic_t fscache_n_store_vmscan_not_storing;
+atomic_t fscache_n_store_vmscan_gone;
+atomic_t fscache_n_store_vmscan_busy;
+atomic_t fscache_n_store_vmscan_cancelled;
+
 atomic_t fscache_n_marks;
 atomic_t fscache_n_uncaches;
 
@@ -211,6 +216,12 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_store_radix_deletes),
 		   atomic_read(&fscache_n_store_pages_over_limit));
 
+	seq_printf(m, "VmScan : nos=%u gon=%u bsy=%u can=%u\n",
+		   atomic_read(&fscache_n_store_vmscan_not_storing),
+		   atomic_read(&fscache_n_store_vmscan_gone),
+		   atomic_read(&fscache_n_store_vmscan_busy),
+		   atomic_read(&fscache_n_store_vmscan_cancelled));
+
 	seq_printf(m, "Ops    : pend=%u run=%u enq=%u can=%u rej=%u\n",
 		   atomic_read(&fscache_n_op_pend),
 		   atomic_read(&fscache_n_op_run),
fs/nfs/fscache.c +3 -7
index 70fad69..fa58800 100644
@@ -359,17 +359,13 @@ int nfs_fscache_release_page(struct page *page, gfp_t gfp)
 
 	BUG_ON(!cookie);
 
-	if (fscache_check_page_write(cookie, page)) {
-		if (!(gfp & __GFP_WAIT))
-			return 0;
-		fscache_wait_on_page_write(cookie, page);
-	}
-
 	if (PageFsCache(page)) {
 		dfprintk(FSCACHE, "NFS: fscache releasepage (0x%p/0x%p/0x%p)\n",
 			 cookie, page, nfsi);
 
-		fscache_uncache_page(cookie, page);
+		if (!fscache_maybe_release_page(cookie, page, gfp))
+			return 0;
+
 		nfs_add_fscache_stats(page->mapping->host,
 				      NFSIOS_FSCACHE_PAGES_UNCACHED, 1);
 	}
include/linux/fscache-cache.h +1 -0
index f3aa4bd..4750d5f 100644
@@ -317,6 +317,7 @@ struct fscache_cookie {
 	void				*netfs_data;	/* back pointer to netfs */
 	struct radix_tree_root		stores;		/* pages to be stored on this cookie */
 #define FSCACHE_COOKIE_PENDING_TAG	0		/* pages tag: pending write to cache */
+#define FSCACHE_COOKIE_STORING_TAG	1		/* pages tag: writing to cache */
 
 	unsigned long			flags;
 #define FSCACHE_COOKIE_LOOKING_UP	0	/* T if non-index cookie being looked up still */
include/linux/fscache.h +27 -0
index 6d8ee46..595ce49 100644
@@ -202,6 +202,8 @@ extern int __fscache_write_page(struct fscache_cookie *, struct page *, gfp_t);
 extern void __fscache_uncache_page(struct fscache_cookie *, struct page *);
 extern bool __fscache_check_page_write(struct fscache_cookie *, struct page *);
 extern void __fscache_wait_on_page_write(struct fscache_cookie *, struct page *);
+extern bool __fscache_maybe_release_page(struct fscache_cookie *, struct page *,
+					 gfp_t);
 
 /**
  * fscache_register_netfs - Register a filesystem as desiring caching services
@@ -615,4 +617,29 @@ void fscache_wait_on_page_write(struct fscache_cookie *cookie,
 		__fscache_wait_on_page_write(cookie, page);
 }
 
+/**
+ * fscache_maybe_release_page - Consider releasing a page, cancelling a store
+ * @cookie: The cookie representing the cache object
+ * @page: The netfs page that is being cached.
+ * @gfp: The gfp flags passed to releasepage()
+ *
+ * Consider releasing a page for the vmscan algorithm, on behalf of the netfs's
+ * releasepage() call.  A storage request on the page may cancelled if it is
+ * not currently being processed.
+ *
+ * The function returns true if the page no longer has a storage request on it,
+ * and false if a storage request is left in place.  If true is returned, the
+ * page will have been passed to fscache_uncache_page().  If false is returned
+ * the page cannot be freed yet.
+ */
+static inline
+bool fscache_maybe_release_page(struct fscache_cookie *cookie,
+				struct page *page,
+				gfp_t gfp)
+{
+	if (fscache_cookie_valid(cookie) && PageFsCache(page))
+		return __fscache_maybe_release_page(cookie, page, gfp);
+	return false;
+}
+
 #endif /* _LINUX_FSCACHE_H */
author maintainer
#20 lore raw

PATCH 19/28 FS-Cache: Add a retirement stat counter

In reply to David Howells

Add a stat counter to count retirement events rather than ordinary release events (the retire argument to fscache_relinquish_cookie()).

---

 fs/fscache/cookie.c   |    2 ++
 fs/fscache/internal.h |    1 +
 fs/fscache/stats.c    |    6 ++++--
 3 files changed, 7 insertions(+), 2 deletions(-)
fs/fscache/cookie.c +2 -0
index f979659..9905350 100644
@@ -415,6 +415,8 @@ void __fscache_relinquish_cookie(struct fscache_cookie *cookie, int retire)
 	unsigned long event;
 
 	fscache_stat(&fscache_n_relinquishes);
+	if (retire)
+		fscache_stat(&fscache_n_relinquishes_retire);
 
 	if (!cookie) {
 		fscache_stat(&fscache_n_relinquishes_null);
fs/fscache/internal.h +1 -0
index e504651..2bf463d 100644
@@ -202,6 +202,7 @@ extern atomic_t fscache_n_updates_run;
 extern atomic_t fscache_n_relinquishes;
 extern atomic_t fscache_n_relinquishes_null;
 extern atomic_t fscache_n_relinquishes_waitcrt;
+extern atomic_t fscache_n_relinquishes_retire;
 
 extern atomic_t fscache_n_cookie_index;
 extern atomic_t fscache_n_cookie_data;
fs/fscache/stats.c +4 -2
index cda6999..9e15289 100644
@@ -85,6 +85,7 @@ atomic_t fscache_n_updates_run;
 atomic_t fscache_n_relinquishes;
 atomic_t fscache_n_relinquishes_null;
 atomic_t fscache_n_relinquishes_waitcrt;
+atomic_t fscache_n_relinquishes_retire;
 
 atomic_t fscache_n_cookie_index;
 atomic_t fscache_n_cookie_data;
@@ -168,10 +169,11 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_updates_null),
 		   atomic_read(&fscache_n_updates_run));
 
-	seq_printf(m, "Relinqs: n=%u nul=%u wcr=%u\n",
+	seq_printf(m, "Relinqs: n=%u nul=%u wcr=%u rtr=%u\n",
 		   atomic_read(&fscache_n_relinquishes),
 		   atomic_read(&fscache_n_relinquishes_null),
-		   atomic_read(&fscache_n_relinquishes_waitcrt));
+		   atomic_read(&fscache_n_relinquishes_waitcrt),
+		   atomic_read(&fscache_n_relinquishes_retire));
 
 	seq_printf(m, "AttrChg: n=%u ok=%u nbf=%u oom=%u run=%u\n",
 		   atomic_read(&fscache_n_attr_changed),
author maintainer
#21 lore raw

PATCH 20/28 FS-Cache: Make sure FSCACHE_COOKIE_LOOKING_UP cleared on lookup failure

In reply to David Howells

We must make sure that FSCACHE_COOKIE_LOOKING_UP is cleared on lookup failure (if an object reaches the LC_DYING state), and we should clear it before clearing FSCACHE_COOKIE_CREATING.

If this doesn't happen then fscache_wait_for_deferred_lookup() may hold allocation and retrieval operations indefinitely until they're interrupted by signals - which in turn pins the dying object until they go away.

---

 fs/fscache/object.c |   17 ++++++++++++-----
 1 files changed, 12 insertions(+), 5 deletions(-)
fs/fscache/object.c +12 -5
index 1a1afa8..74bc562 100644
@@ -105,6 +105,7 @@ static inline void fscache_done_parent_op(struct fscache_object *object)
 static void fscache_object_state_machine(struct fscache_object *object)
 {
 	enum fscache_object_state new_state;
+	struct fscache_cookie *cookie;
 
 	ASSERT(object != NULL);
 
@@ -158,11 +159,17 @@ static void fscache_object_state_machine(struct fscache_object *object)
 
 		spin_lock(&object->lock);
 		object->state = FSCACHE_OBJECT_DYING;
-		if (object->cookie &&
-		    test_and_clear_bit(FSCACHE_COOKIE_CREATING,
-				       &object->cookie->flags))
-			wake_up_bit(&object->cookie->flags,
-				    FSCACHE_COOKIE_CREATING);
+		cookie = object->cookie;
+		if (cookie) {
+			if (test_and_clear_bit(FSCACHE_COOKIE_LOOKING_UP,
+					       &cookie->flags))
+				wake_up_bit(&cookie->flags,
+					    FSCACHE_COOKIE_LOOKING_UP);
+			if (test_and_clear_bit(FSCACHE_COOKIE_CREATING,
+					       &cookie->flags))
+				wake_up_bit(&cookie->flags,
+					    FSCACHE_COOKIE_CREATING);
+		}
 		spin_unlock(&object->lock);
 
 		fscache_done_parent_op(object);
author maintainer
#22 lore raw

PATCH 21/28 FS-Cache: Start processing an object's operations on that object's death

In reply to David Howells

Start processing an object's operations when that object moves into the DYING state as the object cannot be destroyed until all its outstanding operations have completed.

Furthermore, make sure that read and allocation operations handle being woken up on a dead object. Such events are recorded in the Allocs.abt and Retrvls.abt statistics as viewable through /proc/fs/fscache/stats.

The code for waiting for object activation for the read and allocation operations is also extracted into its own function as it is much the same in all cases, differing only in the stats incremented.

---

 Documentation/filesystems/caching/fscache.txt |    2 
 fs/fscache/internal.h                         |    5 +
 fs/fscache/object.c                           |    1 
 fs/fscache/page.c                             |  112 ++++++++++++-------------
 fs/fscache/stats.c                            |   12 ++-
 include/linux/fscache-cache.h                 |    4 +
 6 files changed, 75 insertions(+), 61 deletions(-)
Documentation/filesystems/caching/fscache.txt +2 -0
index 655569e..692ffef 100644
@@ -253,6 +253,7 @@ proc files.
 		int=N	Number of alloc reqs aborted -ERESTARTSYS
 		ops=N	Number of alloc reqs submitted
 		owt=N	Number of alloc reqs waited for CPU time
+		abt=N	Number of alloc reqs aborted due to object death
 	Retrvls	n=N	Number of retrieval (read) requests seen
 		ok=N	Number of successful retr reqs
 		wt=N	Number of retr reqs that waited on lookup completion
@@ -262,6 +263,7 @@ proc files.
 		oom=N	Number of retr reqs failed -ENOMEM
 		ops=N	Number of retr reqs submitted
 		owt=N	Number of retr reqs waited for CPU time
+		abt=N	Number of retr reqs aborted due to object death
 	Stores	n=N	Number of storage (write) requests seen
 		ok=N	Number of successful store reqs
 		agn=N	Number of store reqs on a page already pending storage
fs/fscache/internal.h +5 -0
index 2bf463d..5b49a37 100644
@@ -156,6 +156,7 @@ extern atomic_t fscache_n_allocs_ok;
 extern atomic_t fscache_n_allocs_wait;
 extern atomic_t fscache_n_allocs_nobufs;
 extern atomic_t fscache_n_allocs_intr;
+extern atomic_t fscache_n_allocs_object_dead;
 extern atomic_t fscache_n_alloc_ops;
 extern atomic_t fscache_n_alloc_op_waits;
 
@@ -166,6 +167,7 @@ extern atomic_t fscache_n_retrievals_nodata;
 extern atomic_t fscache_n_retrievals_nobufs;
 extern atomic_t fscache_n_retrievals_intr;
 extern atomic_t fscache_n_retrievals_nomem;
+extern atomic_t fscache_n_retrievals_object_dead;
 extern atomic_t fscache_n_retrieval_ops;
 extern atomic_t fscache_n_retrieval_op_waits;
 
@@ -249,9 +251,12 @@ static inline void fscache_stat_d(atomic_t *stat)
 	atomic_dec(stat);
 }
 
+#define __fscache_stat(stat) (stat)
+
 extern const struct file_operations fscache_stats_fops;
 #else
 
+#define __fscache_stat(stat) (NULL)
 #define fscache_stat(stat) do {} while (0)
 #endif
 
fs/fscache/object.c +1 -0
index 74bc562..c85c9f5 100644
@@ -201,6 +201,7 @@ static void fscache_object_state_machine(struct fscache_object *object)
 		}
 		spin_unlock(&object->lock);
 		fscache_enqueue_dependents(object);
+		fscache_start_operations(object);
 		goto terminal_transit;
 
 		/* handle an abort during initialisation */
fs/fscache/page.c +55 -57

136 lines not on this page · raw message

fs/fscache/stats.c +8 -4
index 9e15289..05f77ca 100644
@@ -39,6 +39,7 @@ atomic_t fscache_n_allocs_ok;
 atomic_t fscache_n_allocs_wait;
 atomic_t fscache_n_allocs_nobufs;
 atomic_t fscache_n_allocs_intr;
+atomic_t fscache_n_allocs_object_dead;
 atomic_t fscache_n_alloc_ops;
 atomic_t fscache_n_alloc_op_waits;
 
@@ -49,6 +50,7 @@ atomic_t fscache_n_retrievals_nodata;
 atomic_t fscache_n_retrievals_nobufs;
 atomic_t fscache_n_retrievals_intr;
 atomic_t fscache_n_retrievals_nomem;
+atomic_t fscache_n_retrievals_object_dead;
 atomic_t fscache_n_retrieval_ops;
 atomic_t fscache_n_retrieval_op_waits;
 
@@ -188,9 +190,10 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_allocs_wait),
 		   atomic_read(&fscache_n_allocs_nobufs),
 		   atomic_read(&fscache_n_allocs_intr));
-	seq_printf(m, "Allocs : ops=%u owt=%u\n",
+	seq_printf(m, "Allocs : ops=%u owt=%u abt=%u\n",
 		   atomic_read(&fscache_n_alloc_ops),
-		   atomic_read(&fscache_n_alloc_op_waits));
+		   atomic_read(&fscache_n_alloc_op_waits),
+		   atomic_read(&fscache_n_allocs_object_dead));
 
 	seq_printf(m, "Retrvls: n=%u ok=%u wt=%u nod=%u nbf=%u"
 		   " int=%u oom=%u\n",
@@ -201,9 +204,10 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_retrievals_nobufs),
 		   atomic_read(&fscache_n_retrievals_intr),
 		   atomic_read(&fscache_n_retrievals_nomem));
-	seq_printf(m, "Retrvls: ops=%u owt=%u\n",
+	seq_printf(m, "Retrvls: ops=%u owt=%u abt=%u\n",
 		   atomic_read(&fscache_n_retrieval_ops),
-		   atomic_read(&fscache_n_retrieval_op_waits));
+		   atomic_read(&fscache_n_retrieval_op_waits),
+		   atomic_read(&fscache_n_retrievals_object_dead));
 
 	seq_printf(m, "Stores : n=%u ok=%u agn=%u nbf=%u oom=%u\n",
 		   atomic_read(&fscache_n_stores),
include/linux/fscache-cache.h +4 -0
index 4750d5f..907bb56 100644
@@ -404,6 +404,10 @@ extern const char *fscache_object_states[];
 	 (obj)->state >= FSCACHE_OBJECT_AVAILABLE &&	      \
 	 (obj)->state < FSCACHE_OBJECT_DYING)
 
+#define fscache_object_is_dead(obj)				\
+	(test_bit(FSCACHE_IOERROR, &(obj)->cache->flags) &&	\
+	 (obj)->state >= FSCACHE_OBJECT_DYING)
+
 extern const struct slow_work_ops fscache_object_slow_work_ops;
 
 /**
author maintainer
#23 lore raw

PATCH 22/28 FS-Cache: Actually requeue an object when requested

In reply to David Howells

FS-Cache objects have an FSCACHE_OBJECT_EV_REQUEUE event that can theoretically be raised to ask the state machine to requeue the object for further processing before the work function returns to the slow-work facility.

However, fscache_object_work_execute() was clearing that bit before checking the event mask to see whether the object has any pending events that require it to be requeued immediately.

Instead, the bit should be cleared after the check and enqueue.

---

 fs/fscache/object.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)
fs/fscache/object.c +1 -2
index c85c9f5..f3f952c 100644
@@ -353,13 +353,12 @@ static void fscache_object_slow_work_execute(struct slow_work *work)
 
 	_enter("{OBJ%x}", object->debug_id);
 
-	clear_bit(FSCACHE_OBJECT_EV_REQUEUE, &object->events);
-
 	start = jiffies;
 	fscache_object_state_machine(object);
 	fscache_hist(fscache_objs_histogram, start);
 	if (object->events & object->event_mask)
 		fscache_enqueue_object(object);
+	clear_bit(FSCACHE_OBJECT_EV_REQUEUE, &object->events);
 }
 
 /*
author maintainer
#24 lore raw

PATCH 23/28 CacheFiles: Don't write a full page if there's only a partial page to cache

In reply to David Howells

cachefiles_write_page() writes a full page to the backing file for the last page of the netfs file, even if the netfs file's last page is only a partial page.

This causes the EOF on the backing file to be extended beyond the EOF of the netfs, and thus the backing file will be truncated by cachefiles_attr_changed() called from cachefiles_lookup_object().

So we need to limit the write we make to the backing file on that last page such that it doesn't push the EOF too far.

Also, if a backing file that has a partial page at the end is expanded, we discard the partial page and refetch it on the basis that we then have a hole in the file with invalid data, and should the power go out... A better way to deal with this could be to record a note that the partial page contains invalid data until the correct data is written into it.

This isn't a problem for netfs's that discard the whole backing file if the file size changes (such as NFS).

---

 fs/cachefiles/interface.c     |   20 +++++++++++++++++---
 fs/cachefiles/rdwr.c          |   23 +++++++++++++++++++----
 include/linux/fscache-cache.h |    3 +++
 3 files changed, 39 insertions(+), 7 deletions(-)
fs/cachefiles/interface.c +17 -3
index dd7f852..8e67abf 100644
@@ -404,12 +404,26 @@ static int cachefiles_attr_changed(struct fscache_object *_object)
 	if (oi_size == ni_size)
 		return 0;
 
-	newattrs.ia_size = ni_size;
-	newattrs.ia_valid = ATTR_SIZE;
-
 	cachefiles_begin_secure(cache, &saved_cred);
 	mutex_lock(&object->backer->d_inode->i_mutex);
+
+	/* if there's an extension to a partial page at the end of the backing
+	 * file, we need to discard the partial page so that we pick up new
+	 * data after it */
+	if (oi_size & ~PAGE_MASK && ni_size > oi_size) {
+		_debug("discard tail %llx", oi_size);
+		newattrs.ia_valid = ATTR_SIZE;
+		newattrs.ia_size = oi_size & PAGE_MASK;
+		ret = notify_change(object->backer, &newattrs);
+		if (ret < 0)
+			goto truncate_failed;
+	}
+
+	newattrs.ia_valid = ATTR_SIZE;
+	newattrs.ia_size = ni_size;
 	ret = notify_change(object->backer, &newattrs);
+
+truncate_failed:
 	mutex_unlock(&object->backer->d_inode->i_mutex);
 	cachefiles_end_secure(cache, saved_cred);
 
fs/cachefiles/rdwr.c +19 -4
index 3304646..5a84fd7 100644
@@ -803,7 +803,8 @@ int cachefiles_write_page(struct fscache_storage *op, struct page *page)
 	struct cachefiles_cache *cache;
 	mm_segment_t old_fs;
 	struct file *file;
-	loff_t pos;
+	loff_t pos, eof;
+	size_t len;
 	void *data;
 	int ret;
 
@@ -837,15 +838,29 @@ int cachefiles_write_page(struct fscache_storage *op, struct page *page)
 		ret = -EIO;
 		if (file->f_op->write) {
 			pos = (loff_t) page->index << PAGE_SHIFT;
+
+			/* we mustn't write more data than we have, so we have
+			 * to beware of a partial page at EOF */
+			eof = object->fscache.store_limit_l;
+			len = PAGE_SIZE;
+			if (eof & ~PAGE_MASK) {
+				ASSERTCMP(pos, <, eof);
+				if (eof - pos < PAGE_SIZE) {
+					_debug("cut short %llx to %llx",
+					       pos, eof);
+					len = eof - pos;
+					ASSERTCMP(pos + len, ==, eof);
+				}
+			}
+
 			data = kmap(page);
 			old_fs = get_fs();
 			set_fs(KERNEL_DS);
 			ret = file->f_op->write(
-				file, (const void __user *) data, PAGE_SIZE,
-				&pos);
+				file, (const void __user *) data, len, &pos);
 			set_fs(old_fs);
 			kunmap(page);
-			if (ret != PAGE_SIZE)
+			if (ret != len)
 				ret = -EIO;
 		}
 		fput(file);
include/linux/fscache-cache.h +3 -0
index 907bb56..5db5000 100644
@@ -395,6 +395,7 @@ struct fscache_object {
 	struct rb_node		objlist_link;	/* link in global object list */
 #endif
 	pgoff_t			store_limit;	/* current storage limit */
+	loff_t			store_limit_l;	/* current storage limit */
 };
 
 extern const char *fscache_object_states[];
@@ -439,6 +440,7 @@ void fscache_object_init(struct fscache_object *object,
 	object->events = object->event_mask = 0;
 	object->flags = 0;
 	object->store_limit = 0;
+	object->store_limit_l = 0;
 	object->cache = cache;
 	object->cookie = cookie;
 	object->parent = NULL;
@@ -491,6 +493,7 @@ static inline void fscache_object_lookup_error(struct fscache_object *object)
 static inline
 void fscache_set_store_limit(struct fscache_object *object, loff_t i_size)
 {
+	object->store_limit_l = i_size;
 	object->store_limit = i_size >> PAGE_SHIFT;
 	if (i_size & ~PAGE_MASK)
 		object->store_limit++;
author maintainer
#25 lore raw

PATCH 24/28 CacheFiles: Handle truncate unlocking the page we're reading

In reply to David Howells

Handle truncate unlocking the page we're attempting to read from the backing device before the read has completed.

This was causing reports like the following to occur:

	Pid: 4765, comm: kslowd Not tainted 2.6.30.1 #1
	Call Trace:
	 [<ffffffffa0331d7a>] ? cachefiles_read_waiter+0xd9/0x147 [cachefiles]
	 [<ffffffff804b74bd>] ? __wait_on_bit+0x60/0x6f
	 [<ffffffff8022bbbb>] ? __wake_up_common+0x3f/0x71
	 [<ffffffff8022cc32>] ? __wake_up+0x30/0x44
	 [<ffffffff8024a41f>] ? __wake_up_bit+0x28/0x2d
	 [<ffffffffa003a793>] ? ext3_truncate+0x4d7/0x8ed [ext3]
	 [<ffffffff80281f90>] ? pagevec_lookup+0x17/0x1f
	 [<ffffffff8028c2ff>] ? unmap_mapping_range+0x59/0x1ff
	 [<ffffffff8022cc32>] ? __wake_up+0x30/0x44
	 [<ffffffff8028e286>] ? vmtruncate+0xc2/0xe2
	 [<ffffffff802b82cf>] ? inode_setattr+0x22/0x10a
	 [<ffffffffa003baa5>] ? ext3_setattr+0x17b/0x1e6 [ext3]
	 [<ffffffff802b853d>] ? notify_change+0x186/0x2c9
	 [<ffffffffa032d9de>] ? cachefiles_attr_changed+0x133/0x1cd [cachefiles]
	 [<ffffffffa032df7f>] ? cachefiles_lookup_object+0xcf/0x12a [cachefiles]
	 [<ffffffffa0318165>] ? fscache_lookup_object+0x110/0x122 [fscache]
	 [<ffffffffa03188c3>] ? fscache_object_slow_work_execute+0x590/0x6bc
	[fscache]
	 [<ffffffff80278f82>] ? slow_work_thread+0x285/0x43a
	 [<ffffffff8024a446>] ? autoremove_wake_function+0x0/0x2e
	 [<ffffffff80278cfd>] ? slow_work_thread+0x0/0x43a
	 [<ffffffff8024a317>] ? kthread+0x54/0x81
	 [<ffffffff8020c93a>] ? child_rip+0xa/0x20
	 [<ffffffff8024a2c3>] ? kthread+0x0/0x81
	 [<ffffffff8020c930>] ? child_rip+0x0/0x20
	CacheFiles: I/O Error: Readpage failed on backing file 200000000000810
	FS-Cache: Cache cachefiles stopped due to I/O error

---

 fs/cachefiles/rdwr.c |   99 +++++++++++++++++++++++++++++++++++++++++++++++---
 1 files changed, 93 insertions(+), 6 deletions(-)
fs/cachefiles/rdwr.c +93 -6
index 5a84fd7..1d83325 100644
@@ -40,8 +40,10 @@ static int cachefiles_read_waiter(wait_queue_t *wait, unsigned mode,
 
 	_debug("--- monitor %p %lx ---", page, page->flags);
 
-	if (!PageUptodate(page) && !PageError(page))
-		dump_stack();
+	if (!PageUptodate(page) && !PageError(page)) {
+		/* unlocked, not uptodate and not erronous? */
+		_debug("page probably truncated");
+	}
 
 	/* remove from the waitqueue */
 	list_del(&wait->task_list);
@@ -61,6 +63,84 @@ static int cachefiles_read_waiter(wait_queue_t *wait, unsigned mode,
 }
 
 /*
+ * handle a probably truncated page
+ * - check to see if the page is still relevant and reissue the read if
+ *   possible
+ * - return -EIO on error, -ENODATA if the page is gone, -EINPROGRESS if we
+ *   must wait again and 0 if successful
+ */
+static int cachefiles_read_reissue(struct cachefiles_object *object,
+				   struct cachefiles_one_read *monitor)
+{
+	struct address_space *bmapping = object->backer->d_inode->i_mapping;
+	struct page *backpage = monitor->back_page, *backpage2;
+	int ret;
+
+	kenter("{ino=%lx},{%lx,%lx}",
+	       object->backer->d_inode->i_ino,
+	       backpage->index, backpage->flags);
+
+	/* skip if the page was truncated away completely */
+	if (backpage->mapping != bmapping) {
+		kleave(" = -ENODATA [mapping]");
+		return -ENODATA;
+	}
+
+	backpage2 = find_get_page(bmapping, backpage->index);
+	if (!backpage2) {
+		kleave(" = -ENODATA [gone]");
+		return -ENODATA;
+	}
+
+	if (backpage != backpage2) {
+		put_page(backpage2);
+		kleave(" = -ENODATA [different]");
+		return -ENODATA;
+	}
+
+	/* the page is still there and we already have a ref on it, so we don't
+	 * need a second */
+	put_page(backpage2);
+
+	INIT_LIST_HEAD(&monitor->op_link);
+	add_page_wait_queue(backpage, &monitor->monitor);
+
+	if (trylock_page(backpage)) {
+		ret = -EIO;
+		if (PageError(backpage))
+			goto unlock_discard;
+		ret = 0;
+		if (PageUptodate(backpage))
+			goto unlock_discard;
+
+		kdebug("reissue read");
+		ret = bmapping->a_ops->readpage(NULL, backpage);
+		if (ret < 0)
+			goto unlock_discard;
+	}
+
+	/* but the page may have been read before the monitor was installed, so
+	 * the monitor may miss the event - so we have to ensure that we do get
+	 * one in such a case */
+	if (trylock_page(backpage)) {
+		_debug("jumpstart %p {%lx}", backpage, backpage->flags);
+		unlock_page(backpage);
+	}
+
+	/* it'll reappear on the todo list */
+	kleave(" = -EINPROGRESS");
+	return -EINPROGRESS;
+
+unlock_discard:
+	unlock_page(backpage);
+	spin_lock_irq(&object->work_lock);
+	list_del(&monitor->op_link);
+	spin_unlock_irq(&object->work_lock);
+	kleave(" = %d", ret);
+	return ret;
+}
+
+/*
  * copy data from backing pages to netfs pages to complete a read operation
  * - driven by FS-Cache's thread pool
  */
@@ -92,20 +172,26 @@ static void cachefiles_read_copier(struct fscache_operation *_op)
 
 		_debug("- copy {%lu}", monitor->back_page->index);
 
-		error = -EIO;
+	recheck:
 		if (PageUptodate(monitor->back_page)) {
 			copy_highpage(monitor->netfs_page, monitor->back_page);
 
 			pagevec_add(&pagevec, monitor->netfs_page);
 			fscache_mark_pages_cached(monitor->op, &pagevec);
 			error = 0;
-		}
-
-		if (error)
+		} else if (!PageError(monitor->back_page)) {
+			/* the page has probably been truncated */
+			error = cachefiles_read_reissue(object, monitor);
+			if (error == -EINPROGRESS)
+				goto next;
+			goto recheck;
+		} else {
 			cachefiles_io_error_obj(
 				object,
 				"Readpage failed on backing file %lx",
 				(unsigned long) monitor->back_page->flags);
+			error = -EIO;
+		}
 
 		page_cache_release(monitor->back_page);
 
@@ -114,6 +200,7 @@ static void cachefiles_read_copier(struct fscache_operation *_op)
 		fscache_put_retrieval(op);
 		kfree(monitor);
 
+	next:
 		/* let the thread pool have some air occasionally */
 		max--;
 		if (max < 0 || need_resched()) {
author maintainer
#26 lore raw

PATCH 25/28 CacheFiles: Mark parent directory locks as I_MUTEX_PARENT to keep lockdep happy

In reply to David Howells

Mark parent directory locks as I_MUTEX_PARENT in the callers of cachefiles_bury_object() so that lockdep doesn't complain when that invokes vfs_unlink():

=============================================
[ INFO: possible recursive locking detected ]
2.6.32-rc6-cachefs #47
---------------------------------------------
kslowd002/3089 is trying to acquire lock:
(&sb->s_type->i_mutex_key#7){+.+.+.}, at: [<ffffffff810bbf72>] vfs_unlink+0x8b/0x128

but task is already holding lock:
(&sb->s_type->i_mutex_key#7){+.+.+.}, at: [<ffffffffa00e4e61>] cachefiles_walk_to_object+0x1b0/0x831 [cachefiles]

other info that might help us debug this:
1 lock held by kslowd002/3089:
#0: (&sb->s_type->i_mutex_key#7){+.+.+.}, at: [<ffffffffa00e4e61>] cachefiles_walk_to_object+0x1b0/0x831 [cachefiles]

stack backtrace:
Pid: 3089, comm: kslowd002 Not tainted 2.6.32-rc6-cachefs #47
Call Trace:
 [<ffffffff8105ad7b>] __lock_acquire+0x1649/0x16e3
 [<ffffffff8118170e>] ? inode_has_perm+0x5f/0x61
 [<ffffffff8105ae6c>] lock_acquire+0x57/0x6d
 [<ffffffff810bbf72>] ? vfs_unlink+0x8b/0x128
 [<ffffffff81353ac3>] mutex_lock_nested+0x54/0x292
 [<ffffffff810bbf72>] ? vfs_unlink+0x8b/0x128
 [<ffffffff8118179e>] ? selinux_inode_permission+0x8e/0x90
 [<ffffffff8117e271>] ? security_inode_permission+0x1c/0x1e
 [<ffffffff810bb4fb>] ? inode_permission+0x99/0xa5
 [<ffffffff810bbf72>] vfs_unlink+0x8b/0x128
 [<ffffffff810adb19>] ? kfree+0xed/0xf9
 [<ffffffffa00e3f00>] cachefiles_bury_object+0xb6/0x420 [cachefiles]
 [<ffffffff81058e21>] ? trace_hardirqs_on+0xd/0xf
 [<ffffffffa00e7e24>] ? cachefiles_check_object_xattr+0x233/0x293 [cachefiles]
 [<ffffffffa00e51b0>] cachefiles_walk_to_object+0x4ff/0x831 [cachefiles]
 [<ffffffff81032238>] ? finish_task_switch+0x0/0xb2
 [<ffffffffa00e3429>] cachefiles_lookup_object+0xac/0x12a [cachefiles]
 [<ffffffffa00741e9>] fscache_lookup_object+0x1c7/0x214 [fscache]
 [<ffffffffa0074fc5>] fscache_object_state_machine+0xa5/0x52d [fscache]
 [<ffffffffa00754ac>] fscache_object_slow_work_execute+0x5f/0xa0 [fscache]
 [<ffffffff81082093>] slow_work_execute+0x18f/0x2d1
 [<ffffffff8108239a>] slow_work_thread+0x1c5/0x308
 [<ffffffff8104c0f1>] ? autoremove_wake_function+0x0/0x34
 [<ffffffff810821d5>] ? slow_work_thread+0x0/0x308
 [<ffffffff8104be91>] kthread+0x7a/0x82
 [<ffffffff8100beda>] child_rip+0xa/0x20
 [<ffffffff8100b87c>] ? restore_args+0x0/0x30
 [<ffffffff8104be17>] ? kthread+0x0/0x82
 [<ffffffff8100bed0>] ? child_rip+0x0/0x20

---

 fs/cachefiles/namei.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
fs/cachefiles/namei.c +2 -2
index 4ce818a..3df8695 100644
@@ -254,7 +254,7 @@ int cachefiles_delete_object(struct cachefiles_cache *cache,
 
 	dir = dget_parent(object->dentry);
 
-	mutex_lock(&dir->d_inode->i_mutex);
+	mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT);
 	ret = cachefiles_bury_object(cache, dir, object->dentry);
 
 	dput(dir);
@@ -307,7 +307,7 @@ lookup_again:
 	/* search the current directory for the element name */
 	_debug("lookup '%s'", name);
 
-	mutex_lock(&dir->d_inode->i_mutex);
+	mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT);
 
 	start = jiffies;
 	next = lookup_one_len(name, dir, nlen);
author maintainer
#27 lore raw

PATCH 26/28 CacheFiles: Better showing of debugging information in active object problems

In reply to David Howells

Show more debugging information if cachefiles_mark_object_active() is asked to activate an active object.

This may happen, for instance, if the netfs tries to register an object with the same key multiple times.

The code is changed to (a) get the appropriate object lock to protect the cookie pointer whilst we dereference it, and (b) get and display the cookie key if available.

---

 fs/cachefiles/namei.c |  102 ++++++++++++++++++++++++++++++++++++-------------
 1 files changed, 75 insertions(+), 27 deletions(-)
fs/cachefiles/namei.c +75 -27
index 3df8695..00a0cda 100644
@@ -27,6 +27,76 @@ static int cachefiles_wait_bit(void *flags)
 	return 0;
 }
 
+#define CACHEFILES_KEYBUF_SIZE 512
+
+/*
+ * dump debugging info about an object
+ */
+static noinline
+void __cachefiles_printk_object(struct cachefiles_object *object,
+				const char *prefix,
+				u8 *keybuf)
+{
+	struct fscache_cookie *cookie;
+	unsigned keylen, loop;
+
+	printk(KERN_ERR "%sobject: OBJ%x\n",
+	       prefix, object->fscache.debug_id);
+	printk(KERN_ERR "%sobjstate=%s fl=%lx swfl=%lx ev=%lx[%lx]\n",
+	       prefix, fscache_object_states[object->fscache.state],
+	       object->fscache.flags, object->fscache.work.flags,
+	       object->fscache.events,
+	       object->fscache.event_mask & FSCACHE_OBJECT_EVENTS_MASK);
+	printk(KERN_ERR "%sops=%u inp=%u exc=%u\n",
+	       prefix, object->fscache.n_ops, object->fscache.n_in_progress,
+	       object->fscache.n_exclusive);
+	printk(KERN_ERR "%sparent=%p\n",
+	       prefix, object->fscache.parent);
+
+	spin_lock(&object->fscache.lock);
+	cookie = object->fscache.cookie;
+	if (cookie) {
+		printk(KERN_ERR "%scookie=%p [pr=%p nd=%p fl=%lx]\n",
+		       prefix,
+		       object->fscache.cookie,
+		       object->fscache.cookie->parent,
+		       object->fscache.cookie->netfs_data,
+		       object->fscache.cookie->flags);
+		if (keybuf)
+			keylen = cookie->def->get_key(cookie->netfs_data, keybuf,
+						      CACHEFILES_KEYBUF_SIZE);
+		else
+			keylen = 0;
+	} else {
+		printk(KERN_ERR "%scookie=NULL\n", prefix);
+		keylen = 0;
+	}
+	spin_unlock(&object->fscache.lock);
+
+	if (keylen) {
+		printk(KERN_ERR "%skey=[%u] '", prefix, keylen);
+		for (loop = 0; loop < keylen; loop++)
+			printk("%02x", keybuf[loop]);
+		printk("'\n");
+	}
+}
+
+/*
+ * dump debugging info about a pair of objects
+ */
+static noinline void cachefiles_printk_object(struct cachefiles_object *object,
+					      struct cachefiles_object *xobject)
+{
+	u8 *keybuf;
+
+	keybuf = kmalloc(CACHEFILES_KEYBUF_SIZE, GFP_NOIO);
+	if (object)
+		__cachefiles_printk_object(object, "", keybuf);
+	if (xobject)
+		__cachefiles_printk_object(xobject, "x", keybuf);
+	kfree(keybuf);
+}
+
 /*
  * record the fact that an object is now active
  */
@@ -42,8 +112,11 @@ static void cachefiles_mark_object_active(struct cachefiles_cache *cache,
 try_again:
 	write_lock(&cache->active_lock);
 
-	if (test_and_set_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags))
+	if (test_and_set_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags)) {
+		printk(KERN_ERR "CacheFiles: Error: Object already active\n");
+		cachefiles_printk_object(object, NULL);
 		BUG();
+	}
 
 	dentry = object->dentry;
 	_p = &cache->active_nodes.rb_node;
@@ -76,32 +149,7 @@ wait_for_old_object:
 		printk(KERN_ERR "\n");
 		printk(KERN_ERR "CacheFiles: Error:"
 		       " Unexpected object collision\n");
-		printk(KERN_ERR "xobject: OBJ%x\n",
-		       xobject->fscache.debug_id);
-		printk(KERN_ERR "xobjstate=%s\n",
-		       fscache_object_states[xobject->fscache.state]);
-		printk(KERN_ERR "xobjflags=%lx\n", xobject->fscache.flags);
-		printk(KERN_ERR "xobjevent=%lx [%lx]\n",
-		       xobject->fscache.events, xobject->fscache.event_mask);
-		printk(KERN_ERR "xops=%u inp=%u exc=%u\n",
-		       xobject->fscache.n_ops, xobject->fscache.n_in_progress,
-		       xobject->fscache.n_exclusive);
-		printk(KERN_ERR "xcookie=%p [pr=%p nd=%p fl=%lx]\n",
-		       xobject->fscache.cookie,
-		       xobject->fscache.cookie->parent,
-		       xobject->fscache.cookie->netfs_data,
-		       xobject->fscache.cookie->flags);
-		printk(KERN_ERR "xparent=%p\n",
-		       xobject->fscache.parent);
-		printk(KERN_ERR "object: OBJ%x\n",
-		       object->fscache.debug_id);
-		printk(KERN_ERR "cookie=%p [pr=%p nd=%p fl=%lx]\n",
-		       object->fscache.cookie,
-		       object->fscache.cookie->parent,
-		       object->fscache.cookie->netfs_data,
-		       object->fscache.cookie->flags);
-		printk(KERN_ERR "parent=%p\n",
-		       object->fscache.parent);
+		cachefiles_printk_object(object, xobject);
 		BUG();
 	}
 	atomic_inc(&xobject->usage);
author maintainer
#28 lore raw

PATCH 27/28 CacheFiles: Catch an overly long wait for an old active object

In reply to David Howells

Catch an overly long wait for an old, dying active object when we want to replace it with a new one. The probability is that all the slow-work threads are hogged, and the delete can't get a look in.

What we do instead is:

(1) if there's nothing in the slow work queue, we sleep until either the dying object has finished dying or there is something in the slow work queue behind which we can queue our object.

(2) if there is something in the slow work queue, we return ETIMEDOUT to fscache_lookup_object(), which then puts us back on the slow work queue, presumably behind the deletion that we're blocked by. We are then deferred for a while until we work our way back through the queue - without blocking a slow-work thread unnecessarily.

A backtrace similar to the following may appear in the log without this patch:

	INFO: task kslowd004:5711 blocked for more than 120 seconds.
	"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
	kslowd004     D 0000000000000000     0  5711      2 0x00000080
	 ffff88000340bb80 0000000000000046 ffff88002550d000 0000000000000000
	 ffff88002550d000 0000000000000007 ffff88000340bfd8 ffff88002550d2a8
	 000000000000ddf0 00000000000118c0 00000000000118c0 ffff88002550d2a8
	Call Trace:
	 [<ffffffff81058e21>] ? trace_hardirqs_on+0xd/0xf
	 [<ffffffffa011c4d8>] ? cachefiles_wait_bit+0x0/0xd [cachefiles]
	 [<ffffffffa011c4e1>] cachefiles_wait_bit+0x9/0xd [cachefiles]
	 [<ffffffff81353153>] __wait_on_bit+0x43/0x76
	 [<ffffffff8111ae39>] ? ext3_xattr_get+0x1ec/0x270
	 [<ffffffff813531ef>] out_of_line_wait_on_bit+0x69/0x74
	 [<ffffffffa011c4d8>] ? cachefiles_wait_bit+0x0/0xd [cachefiles]
	 [<ffffffff8104c125>] ? wake_bit_function+0x0/0x2e
	 [<ffffffffa011bc79>] cachefiles_mark_object_active+0x203/0x23b [cachefiles]
	 [<ffffffffa011c209>] cachefiles_walk_to_object+0x558/0x827 [cachefiles]
	 [<ffffffffa011a429>] cachefiles_lookup_object+0xac/0x12a [cachefiles]
	 [<ffffffffa00aa1e9>] fscache_lookup_object+0x1c7/0x214 [fscache]
	 [<ffffffffa00aafc5>] fscache_object_state_machine+0xa5/0x52d [fscache]
	 [<ffffffffa00ab4ac>] fscache_object_slow_work_execute+0x5f/0xa0 [fscache]
	 [<ffffffff81082093>] slow_work_execute+0x18f/0x2d1
	 [<ffffffff8108239a>] slow_work_thread+0x1c5/0x308
	 [<ffffffff8104c0f1>] ? autoremove_wake_function+0x0/0x34
	 [<ffffffff810821d5>] ? slow_work_thread+0x0/0x308
	 [<ffffffff8104be91>] kthread+0x7a/0x82
	 [<ffffffff8100beda>] child_rip+0xa/0x20
	 [<ffffffff8100b87c>] ? restore_args+0x0/0x30
	 [<ffffffff8104be17>] ? kthread+0x0/0x82
	 [<ffffffff8100bed0>] ? child_rip+0x0/0x20
	1 lock held by kslowd004/5711:
	 #0:  (&sb->s_type->i_mutex_key#7/1){+.+.+.}, at: [<ffffffffa011be64>] cachefiles_walk_to_object+0x1b3/0x827 [cachefiles]

---

 Documentation/filesystems/caching/fscache.txt |    1 
 fs/cachefiles/interface.c                     |    6 +-
 fs/cachefiles/namei.c                         |   87 ++++++++++++++++++++-----
 fs/fscache/internal.h                         |    1 
 fs/fscache/object.c                           |   10 +++
 fs/fscache/stats.c                            |    4 +
 include/linux/fscache-cache.h                 |    6 +-
 7 files changed, 90 insertions(+), 25 deletions(-)
Documentation/filesystems/caching/fscache.txt +1 -0
index 692ffef..571e989 100644
@@ -235,6 +235,7 @@ proc files.
 		neg=N	Number of negative lookups made
 		pos=N	Number of positive lookups made
 		crt=N	Number of objects created by lookup
+		tmo=N	Number of lookups timed out and requeued
 	Updates	n=N	Number of update cookie requests seen
 		nul=N	Number of upd reqs given a NULL parent
 		run=N	Number of upd reqs granted CPU time
fs/cachefiles/interface.c +4 -2
index 8e67abf..9d3c426 100644
@@ -114,8 +114,9 @@ nomem_lookup_data:
 
 /*
  * attempt to look up the nominated node in this cache
+ * - return -ETIMEDOUT to be scheduled again
  */
-static void cachefiles_lookup_object(struct fscache_object *_object)
+static int cachefiles_lookup_object(struct fscache_object *_object)
 {
 	struct cachefiles_lookup_data *lookup_data;
 	struct cachefiles_object *parent, *object;
@@ -145,13 +146,14 @@ static void cachefiles_lookup_object(struct fscache_object *_object)
 	    object->fscache.cookie->def->type != FSCACHE_COOKIE_TYPE_INDEX)
 		cachefiles_attr_changed(&object->fscache);
 
-	if (ret < 0) {
+	if (ret < 0 && ret != -ETIMEDOUT) {
 		printk(KERN_WARNING "CacheFiles: Lookup failed error %d\n",
 		       ret);
 		fscache_object_lookup_error(&object->fscache);
 	}
 
 	_leave(" [%d]", ret);
+	return ret;
 }
 
 /*
fs/cachefiles/namei.c +68 -19

143 lines not on this page · raw message

fs/fscache/internal.h +1 -0
index 5b49a37..0ca2566 100644
@@ -215,6 +215,7 @@ extern atomic_t fscache_n_object_no_alloc;
 extern atomic_t fscache_n_object_lookups;
 extern atomic_t fscache_n_object_lookups_negative;
 extern atomic_t fscache_n_object_lookups_positive;
+extern atomic_t fscache_n_object_lookups_timed_out;
 extern atomic_t fscache_n_object_created;
 extern atomic_t fscache_n_object_avail;
 extern atomic_t fscache_n_object_dead;
fs/fscache/object.c +9 -1
index f3f952c..e513ac5 100644
@@ -468,6 +468,7 @@ static void fscache_lookup_object(struct fscache_object *object)
 {
 	struct fscache_cookie *cookie = object->cookie;
 	struct fscache_object *parent;
+	int ret;
 
 	_enter("");
 
@@ -493,12 +494,19 @@ static void fscache_lookup_object(struct fscache_object *object)
 
 	fscache_stat(&fscache_n_object_lookups);
 	fscache_stat(&fscache_n_cop_lookup_object);
-	object->cache->ops->lookup_object(object);
+	ret = object->cache->ops->lookup_object(object);
 	fscache_stat_d(&fscache_n_cop_lookup_object);
 
 	if (test_bit(FSCACHE_OBJECT_EV_ERROR, &object->events))
 		set_bit(FSCACHE_COOKIE_UNAVAILABLE, &cookie->flags);
 
+	if (ret == -ETIMEDOUT) {
+		/* probably stuck behind another object, so move this one to
+		 * the back of the queue */
+		fscache_stat(&fscache_n_object_lookups_timed_out);
+		set_bit(FSCACHE_OBJECT_EV_REQUEUE, &object->events);
+	}
+
 	_leave("");
 }
 
fs/fscache/stats.c +3 -1
index 05f77ca..46435f3 100644
@@ -98,6 +98,7 @@ atomic_t fscache_n_object_no_alloc;
 atomic_t fscache_n_object_lookups;
 atomic_t fscache_n_object_lookups_negative;
 atomic_t fscache_n_object_lookups_positive;
+atomic_t fscache_n_object_lookups_timed_out;
 atomic_t fscache_n_object_created;
 atomic_t fscache_n_object_avail;
 atomic_t fscache_n_object_dead;
@@ -160,10 +161,11 @@ static int fscache_stats_show(struct seq_file *m, void *v)
 		   atomic_read(&fscache_n_acquires_nobufs),
 		   atomic_read(&fscache_n_acquires_oom));
 
-	seq_printf(m, "Lookups: n=%u neg=%u pos=%u crt=%u\n",
+	seq_printf(m, "Lookups: n=%u neg=%u pos=%u crt=%u tmo=%u\n",
 		   atomic_read(&fscache_n_object_lookups),
 		   atomic_read(&fscache_n_object_lookups_negative),
 		   atomic_read(&fscache_n_object_lookups_positive),
+		   atomic_read(&fscache_n_object_lookups_timed_out),
 		   atomic_read(&fscache_n_object_created));
 
 	seq_printf(m, "Updates: n=%u nul=%u run=%u\n",
include/linux/fscache-cache.h +4 -2
index 5db5000..7be0c6f 100644
@@ -234,8 +234,10 @@ struct fscache_cache_ops {
 	struct fscache_object *(*alloc_object)(struct fscache_cache *cache,
 					       struct fscache_cookie *cookie);
 
-	/* look up the object for a cookie */
-	void (*lookup_object)(struct fscache_object *object);
+	/* look up the object for a cookie
+	 * - return -ETIMEDOUT to be requeued
+	 */
+	int (*lookup_object)(struct fscache_object *object);
 
 	/* finished looking up */
 	void (*lookup_complete)(struct fscache_object *object);
author maintainer
#29 lore raw

PATCH 28/28 CacheFiles: Don't log lookup/create failing with ENOBUFS

In reply to David Howells

Don't log the CacheFiles lookup/create object routined failing with ENOBUFS as under high memory load or high cache load they can do this quite a lot. This error simply means that the requested object cannot be created on disk due to lack of space, or due to failure of the backing filesystem to find sufficient resources.

---

 fs/cachefiles/interface.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)
fs/cachefiles/interface.c +3 -2
index 9d3c426..2708931 100644
@@ -147,8 +147,9 @@ static int cachefiles_lookup_object(struct fscache_object *_object)
 		cachefiles_attr_changed(&object->fscache);
 
 	if (ret < 0 && ret != -ETIMEDOUT) {
-		printk(KERN_WARNING "CacheFiles: Lookup failed error %d\n",
-		       ret);
+		if (ret != -ENOBUFS)
+			printk(KERN_WARNING
+			       "CacheFiles: Lookup failed error %d\n", ret);
 		fscache_object_lookup_error(&object->fscache);
 	}
 
author maintainer
#32 lore raw

cover Fixes for FS-Cache and CacheFiles

In reply to David Howells

The associated patches provide three sets of changes:

(1) More ways to get debugging information from the slow-work, FS-Cache and CacheFiles facilities.

(2) Fixes for all facilities.

(3) Some extensions for slow-work for users other than FS-Cache/CacheFiles that aren't included in this series, but upon which other patches have been built.

The patches can also be found in:

http://people.redhat.com/~dhowells/fscache/patches/fscache-fixes.tar.bz2

The bugs fixed were found by running a simple stress tester:

http://people.redhat.com/~dhowells/slurp.c

Build it and point it at a bunch of NFS (or AFS or whatever) directory trees, and it forks and execs a copy of itself to handle each directory that it is given. Whilst scanning a directory, for each subdir of that directory a new copy forked and exec'd, and then each file in that directory is read from end to end.

This results in a large number of processes reading in parallel a large number of files. If the data set is large enough, it applies pressure to both RAM space and cache space, and can cause OOMs to occur, and can cause cachefilesd to run in parallel trying to cull the cache as well.

With these patches applied, you can see more counters in /proc/fs/fscache/stats and new sources of information exist too:

(*) /proc/slow_work_rq

The items currently being processed by or currently queued for processing by the slow-work facility.

(*) /proc/fs/fscache/objects

A list of current fscache objects and their states. This can be configured by loading a key into your session keyring before viewing the file, for instance:

keyctl add user fscache:objlist KB @s

This is described in Documentation/filesystems/caching/fscache.txt

David
---

David Howells (3):
FS-Cache: Provide nop fscache_stat_d() if CONFIG_FSCACHE_STATS=n
SLOW_WORK: Fix GFS2 to #include <linux/module.h> before using THIS_MODULE
SLOW_WORK: Fix CIFS to pass THIS_MODULE to slow_work_register_user()

 fs/cifs/cifsfs.c      |    2 +-
 fs/fscache/internal.h |    1 +
 fs/gfs2/recovery.c    |    1 +
 3 files changed, 3 insertions(+), 1 deletions(-)
author maintainer
#33 lore raw

PATCH 1/3 SLOW_WORK: Fix CIFS to pass THIS_MODULE to slow_work_register_user()

As of the patch:

SLOW_WORK: Wait for outstanding work items belonging to a module to clear

Wait for outstanding slow work items belonging to a module to clear when unregistering that module as a user of the facility. This prevents the put_ref code of a work item from being taken away before it returns.

slow_work_register_user() takes a module pointer as an argument. CIFS must now pass THIS_MODULE as that argument, lest the following error be observed:

fs/cifs/cifsfs.c: In function 'init_cifs':
fs/cifs/cifsfs.c:1040: error: too few arguments to function 'slow_work_register_user'

---

 fs/cifs/cifsfs.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
fs/cifs/cifsfs.c +1 -1
index 9a5e4f5..29f1da7 100644
@@ -1037,7 +1037,7 @@ init_cifs(void)
 	if (rc)
 		goto out_unregister_key_type;
 #endif
-	rc = slow_work_register_user();
+	rc = slow_work_register_user(THIS_MODULE);
 	if (rc)
 		goto out_unregister_resolver_key;
 
author maintainer
#34 lore raw

PATCH 2/3 SLOW_WORK: Fix GFS2 to #include <linux/module.h> before using THIS_MODULE

In reply to David Howells

GFS2 has been altered to pass THIS_MODULE to slow_work_register_user(), but hasn't been altered to #include <linux/module.h> to provide it, resulting in the following error:

fs/gfs2/recovery.c:596: error: 'THIS_MODULE' undeclared here (not in a function)

Add the missing #include.

---

 fs/gfs2/recovery.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
fs/gfs2/recovery.c +1 -0
index b2bb779..09fa319 100644
@@ -7,6 +7,7 @@
  * of the GNU General Public License version 2.
  */
 
+#include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/spinlock.h>
 #include <linux/completion.h>
author maintainer
#35 lore raw

PATCH 3/3 FS-Cache: Provide nop fscache_stat_d() if CONFIG_FSCACHE_STATS=n

In reply to David Howells

Provide nop fscache_stat_d() macro if CONFIG_FSCACHE_STATS=n lest errors like the following occur:

fs/fscache/cache.c: In function 'fscache_withdraw_cache':
fs/fscache/cache.c:386: error: implicit declaration of function 'fscache_stat_d' fs/fscache/cache.c:386: error: 'fscache_n_cop_sync_cache' undeclared (first use in this function) fs/fscache/cache.c:386: error: (Each undeclared identifier is reported only once fs/fscache/cache.c:386: error: for each function it appears in.) fs/fscache/cache.c:392: error: 'fscache_n_cop_dissociate_pages' undeclared (first use in this function)

---

 fs/fscache/internal.h |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
fs/fscache/internal.h +1 -0
index 0ca2566..edd7434 100644
@@ -259,6 +259,7 @@ extern const struct file_operations fscache_stats_fops;
 
 #define __fscache_stat(stat) (NULL)
 #define fscache_stat(stat) do {} while (0)
+#define fscache_stat_d(stat) do {} while (0)
 #endif
 
 /*