kernel/watch_queue: Clean up some code
cover kernel/watch_queue: Clean up some code
There is a dangling reference to pipe in a watch_queue after clearing it. Thus, NULL that pointer while clearing. This can be thought of as a v4 of the patches I had sent earlier.
This change renders wqueue->defunct superfluous, as the latter is only used to check if watch_queue is cleared. With this change, the pipe is NULL'd while clearing, so we can just check if the pipe is NULL.
Extending comment for watch_queue->pipe in the definition of watch_queue made the comment conventionally too long (it was already past 80 chars), so I have changed the struct annotations to be doxygen-styled, so that I can extend the comment mentioning that the pipe is NULL when watch_queue is cleared.
Siddh Raman Pant (3):
kernel/watch_queue: Remove dangling pipe reference while clearing
watch_queue
kernel/watch_queue: Improve struct annotation formatting kernel/watch_queue: Remove wqueue->defunct and use pipe for clear
check
include/linux/watch_queue.h | 95 +++++++++++++++++++++++++++---------- kernel/watch_queue.c | 11 ++--- 2 files changed, 75 insertions(+), 31 deletions(-)
Signature · 7 lines
2.35.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH v4 1/3 kernel/watch_queue: Remove dangling pipe reference while clearing watch_queue
If not done, a reference to a freed pipe remains in the watch_queue, as this function is called before freeing a pipe in free_pipe_info() (see line 834 of fs/pipe.c).
We also need to use READ_ONCE() in post_one_notification() to prevent the compiler from optimising and loading a non-NULL value from wqueue->pipe.
---
Changes in v4:
- Brought the lines towards the start rather than the end.
- Removed incorrect NULLing of wqueue->pipe->watch_queue. The latter was pointed out by Eric Biggers <[email protected]> in reply to v3.
Changes in v3:
- Restored the original unlock order, and clear before unlock.
- Used READ_ONCE() in post path.
This was explained by David Howells <[email protected]> in reply to v1.
Changes in v2:
- Removed the superfluous ifdef guard.
kernel/watch_queue.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/watch_queue.c +4 -1
@@ -99,7 +99,7 @@ static bool post_one_notification(struct watch_queue *wqueue, struct watch_notification *n) {void *p; -struct pipe_inode_info *pipe = wqueue->pipe; +struct pipe_inode_info *pipe =READ_ONCE (wqueue->pipe);struct pipe_buffer *buf;struct page *page;unsigned int head, tail, mask, note, offset, len;
@@ -606,6 +606,9 @@ void watch_queue_clear(struct watch_queue *wqueue)/* Prevent new notifications from being stored. */ wqueue->defunct = true; +/* This pipe will get freed by caller, and we are anyways clearing. */ + wqueue->pipe =NULL ; +while (!hlist_empty (&wqueue->watches)) { watch = hlist_entry(wqueue->watches.first, struct watch, queue_node);hlist_del_init_rcu (&watch->queue_node);
Signature · 7 lines
2.35.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH 3/3 kernel/watch_queue: Remove wqueue->defunct and use pipe for clear check
In reply to Siddh Raman PantThe sole use of wqueue->defunct is for checking if the watch queue has been cleared, but wqueue->pipe is also NULL'd while clearing.
Thus, wqueue->defunct is superfluous, as wqueue->pipe can be checked for NULL.
--- include/linux/watch_queue.h | 3 +-- kernel/watch_queue.c | 14 +++++--------- 2 files changed, 6 insertions(+), 11 deletions(-)
include/linux/watch_queue.h +1 -2
@@ -55,7 +55,7 @@ struct watch_filter {* * @rcu: RCU head * @filter: Filter on watch_notification::info -* @pipe: The pipe we're using as a buffer. +* @pipe: The pipe we're using as a buffer, NULL when queue is cleared/closed * @watches: Contributory watches * @notes: Preallocated notifications * @notes_bitmap: Allocation bitmap for notes
@@ -63,7 +63,6 @@ struct watch_filter {* @lock: Spinlock * @nr_notes: Number of notes * @nr_pages: Number of pages in notes[] -* @defunct: True when queues closed */ struct watch_queue {struct rcu_head rcu;
kernel/watch_queue.c +5 -9
@@ -43,7 +43,7 @@ MODULE_LICENSE("GPL");static inline bool lock_wqueue (struct watch_queue *wqueue) {spin_lock_bh (&wqueue->lock); -if (unlikely (wqueue->defunct)) { +if (unlikely (READ_ONCE (wqueue->pipe) ==NULL )) {spin_unlock_bh (&wqueue->lock);return false; }
@@ -99,15 +99,12 @@ static bool post_one_notification(struct watch_queue *wqueue, struct watch_notification *n) {void *p; -struct pipe_inode_info *pipe =READ_ONCE (wqueue->pipe); +struct pipe_inode_info *pipe = wqueue->pipe;struct pipe_buffer *buf;struct page *page;unsigned int head, tail, mask, note, offset, len;bool done = false; -if (!pipe) -return false; -spin_lock_irq (&pipe->rd_wait.lock); mask = pipe->ring_size -1 ;
@@ -603,10 +600,9 @@ void watch_queue_clear(struct watch_queue *wqueue)rcu_read_lock ();spin_lock_bh (&wqueue->lock); -/* Prevent new notifications from being stored. */ - wqueue->defunct = true; - -/* This pipe will get freed by caller, and we are anyways clearing. */ +/* This pipe will get freed by the caller free_pipe_info(). +* Removing this reference also prevents new notifications. +*/ wqueue->pipe =NULL ;while (!hlist_empty (&wqueue->watches)) {
Signature · 7 lines
2.35.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH 2/3 kernel/watch_queue: Improve struct annotation formatting
In reply to Siddh Raman PantImprove formatting struct annotations in watch_queue.h, so that they fall in the preferred 80 character limit.
--- include/linux/watch_queue.h | 96 +++++++++++++++++++++++++++---------- 1 file changed, 71 insertions(+), 25 deletions(-)
include/linux/watch_queue.h +71 -25
@@ -18,57 +18,103 @@struct cred ; +/** +* watch_type_filter - Filter on watch type +* +* @type: Type of watch_notification +* @subtype_filter: Bitmask of subtypes to filter on +* @info_filter: Filter on watch_notification::info +* @info_mask: Mask of relevant bits in info_filter +*/ struct watch_type_filter {enum watch_notification_type type; -__u32 subtype_filter[1 ];/* Bitmask of subtypes to filter on */ -__u32 info_filter;/* Filter on watch_notification::info */ -__u32 info_mask;/* Mask of relevant bits in info_filter */ +__u32 subtype_filter[1 ]; +__u32 info_filter; +__u32 info_mask; }; +/** +* watch_filter - Filter on watch +* +* @rcu: (union) RCU head +* @type_filter: (union) Bitmask of accepted types +* @nr_filters: Number of filters +* @filters: Array of watch_type_filter +*/ struct watch_filter {union {struct rcu_head rcu; -/* Bitmask of accepted types */ DECLARE_BITMAP(type_filter, WATCH_TYPE__NR); }; -u32 nr_filters;/* Number of filters */ +u32 nr_filters;struct watch_type_filter filters[]; }; +/** +* watch_queue - General notification queue +* +* @rcu: RCU head +* @filter: Filter on watch_notification::info +* @pipe: The pipe we're using as a buffer. +* @watches: Contributory watches +* @notes: Preallocated notifications +* @notes_bitmap: Allocation bitmap for notes +* @usage: Object usage count +* @lock: Spinlock +* @nr_notes: Number of notes +* @nr_pages: Number of pages in notes[] +* @defunct: True when queues closed +*/ struct watch_queue {struct rcu_head rcu;struct watch_filter __rcu *filter; -struct pipe_inode_info *pipe;/* The pipe we're using as a buffer */ -struct hlist_head watches;/* Contributory watches */ -struct page **notes;/* Preallocated notifications */ -unsigned long *notes_bitmap;/* Allocation bitmap for notes */ -struct kref usage;/* Object usage count */ +struct pipe_inode_info *pipe; +struct hlist_head watches; +struct page **notes; +unsigned long *notes_bitmap; +struct kref usage;spinlock_t lock; -unsigned int nr_notes;/* Number of notes */ -unsigned int nr_pages;/* Number of pages in notes[] */ -bool defunct;/* T when queues closed */ +unsigned int nr_notes; +unsigned int nr_pages; +bool defunct; }; -/* -* Representation of a watch on an object. +/** +* watch - Representation of a watch on an object. +* +* @rcu: (union) RCU head +* @info_id: (union) ID to be OR'd in to info field +* @queue: Queue to post events to +* @queue_node: Link in queue->watches +* @watch_list: Link in watch_list->watchers +* @list_node: The list node +* @cred: Creds of the owner of the watch +* @private: Private data for the watched object +* @id: Internal identifier +* @usage: Object usage count */ struct watch {union {struct rcu_head rcu; -u32 info_id;/* ID to be OR'd in to info field */ +u32 info_id; }; -struct watch_queue __rcu *queue;/* Queue to post events to */ -struct hlist_node queue_node;/* Link in queue->watches */ +struct watch_queue __rcu *queue; +struct hlist_node queue_node;struct watch_list __rcu *watch_list; -struct hlist_node list_node;/* Link in watch_list->watchers */ -const struct cred *cred;/* Creds of the owner of the watch */ -void *private;/* Private data for the watched object */ -u64 id;/* Internal identifier */ -struct kref usage;/* Object usage count */ +struct hlist_node list_node; +const struct cred *cred; +void *private; +u64 id; +struct kref usage; }; -/* -* List of watches on an object. +/** +* watch_list - List of watches on an object. +* +* @rcu: RCU head +* @watchers: List head +* @release_watch: Function to release watch +* @lock: Spinlock */ struct watch_list {struct rcu_head rcu;
Signature · 7 lines
2.35.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH v2 3/3 kernel/watch_queue: Remove wqueue->defunct and use pipe for clear check
In reply to Siddh Raman PantThe sole use of wqueue->defunct is for checking if the watch queue has been cleared, but wqueue->pipe is also NULL'd while clearing.
Thus, wqueue->defunct is superfluous, as wqueue->pipe can be checked for NULL.
---
Changes in v2:
- The sent patch had accidentally missed removing the member from struct, fix that.
- Use !READ_ONCE() instead of == NULL, as said by checkpatch.pl.
include/linux/watch_queue.h | 4 +--- kernel/watch_queue.c | 14 +++++--------- 2 files changed, 6 insertions(+), 12 deletions(-)
include/linux/watch_queue.h +1 -3
@@ -55,7 +55,7 @@ struct watch_filter {* * @rcu: RCU head * @filter: Filter on watch_notification::info -* @pipe: The pipe we're using as a buffer. +* @pipe: The pipe we're using as a buffer, NULL when queue is cleared/closed * @watches: Contributory watches * @notes: Preallocated notifications * @notes_bitmap: Allocation bitmap for notes
@@ -63,7 +63,6 @@ struct watch_filter {* @lock: Spinlock * @nr_notes: Number of notes * @nr_pages: Number of pages in notes[] -* @defunct: True when queues closed */ struct watch_queue {struct rcu_head rcu;
@@ -76,7 +75,6 @@ struct watch_queue {spinlock_t lock;unsigned int nr_notes;unsigned int nr_pages; -bool defunct; };/**
kernel/watch_queue.c +5 -9
@@ -43,7 +43,7 @@ MODULE_LICENSE("GPL");static inline bool lock_wqueue (struct watch_queue *wqueue) {spin_lock_bh (&wqueue->lock); -if (unlikely (wqueue->defunct)) { +if (unlikely (!READ_ONCE (wqueue->pipe))) {spin_unlock_bh (&wqueue->lock);return false; }
@@ -99,15 +99,12 @@ static bool post_one_notification(struct watch_queue *wqueue, struct watch_notification *n) {void *p; -struct pipe_inode_info *pipe =READ_ONCE (wqueue->pipe); +struct pipe_inode_info *pipe = wqueue->pipe;struct pipe_buffer *buf;struct page *page;unsigned int head, tail, mask, note, offset, len;bool done = false; -if (!pipe) -return false; -spin_lock_irq (&pipe->rd_wait.lock); mask = pipe->ring_size -1 ;
@@ -603,10 +600,9 @@ void watch_queue_clear(struct watch_queue *wqueue)rcu_read_lock ();spin_lock_bh (&wqueue->lock); -/* Prevent new notifications from being stored. */ - wqueue->defunct = true; - -/* This pipe will get freed by caller, and we are anyways clearing. */ +/* This pipe will get freed by the caller free_pipe_info(). +* Removing this reference also prevents new notifications. +*/ wqueue->pipe =NULL ;while (!hlist_empty (&wqueue->watches)) {
Signature · 7 lines
2.35.1 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 21 lines · Siddh Raman Pant
There is a dangling reference to pipe in a
watch_queueafter clearing it. Thus, NULL that pointer while clearing. This can be thought of as a v4 of the patches I had sent earlier.This change renders
wqueue->defunctsuperfluous, as the latter is only used to check ifwatch_queueis cleared. With this change, the pipe is NULL'd while clearing, so we can just check if the pipe is NULL.Extending comment for
watch_queue->pipein the definition ofwatch_queuemade the comment conventionally too long (it was already past 80 chars), so I have changed the struct annotations to be doxygen-styled, so that I can extend the comment mentioning that the pipe is NULL whenwatch_queueis cleared.Siddh Raman Pant (3):
kernel/watch_queue: Remove dangling pipe reference while clearing
watch_queue
kernel/watch_queue: Improve struct annotation formattingkernel/watch_queue: Removewqueue->defunctand use pipe for clear
check
Quoted 3 lines
include/linux/watch_queue.h | 95 +++++++++++++++++++++++++++---------- kernel/watch_queue.c | 11 ++--- 2 files changed, 75 insertions(+), 31 deletions(-)
I think patches 1 and 3 should be merged together.
Also, please use a consistent version number for all patches in the series. You have a version 1, version 2, and version 4 patch all in the same series, which is very confusing.
- Eric
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [PATCH 2/3] kernel/watch_queue: Improve struct annotation formatting
In reply to Siddh Raman PantQuoted 4 lines · Siddh Raman Pant
Improve formatting struct annotations in
watch_queue.h, so that they fall in the preferred 80 character limit.Signed-off-by Siddh Raman Pant <[email protected]>
This patch isn't just fixing overly long lines, but rather is introducing kerneldoc comments and documenting things that weren't documented before. That's fine, but please make the commit message accurately describe the patch.
include/linux/watch_queue.h line 22
Quoted 10 lines
include/linux/watch_queue.h +2
-0@@ -18,57 +18,103 @@struct cred ; +/** +* watch_type_filter - Filter on watch type
If you're going to use kerneldoc comments, they should be correctly formatted. This is not, since it's missing the word struct. You can run this command to see the kerneldoc warnings:
./scripts/kernel-doc -v -none include/linux/watch_queue.h
include/linux/watch_queue.h line 23
Quoted 1 line
include/linux/watch_queue.h +1
-0+* @lock: Spinlock
Please make sure that comments provide useful information and don't just repeat what the code says.
- Eric
Re: [PATCH v2 3/3] kernel/watch_queue: Remove wqueue->defunct and use pipe for clear check
In reply to Siddh Raman Pantkernel/watch_queue.c MODULE_LICENSE("GPL"); · line 49
Quoted 8 lines · Siddh Raman Pant
static inline bool
lock_wqueue(struct watch_queue *wqueue){kernel/watch_queue.c +1
-1spin_lock_bh (&wqueue->lock); -if (unlikely (wqueue->defunct)) { +if (unlikely (!READ_ONCE (wqueue->pipe))) {spin_unlock_bh (&wqueue->lock);return false; }
Why is the READ_ONCE() needed? Doesn't wqueue->lock protect wqueue->pipe?
kernel/watch_queue.c void watch_queue_clear(struct watch_queue *wqueue) · line 605
Quoted 3 lines
kernel/watch_queue.c +3
-0+/* This pipe will get freed by the caller free_pipe_info(). +* Removing this reference also prevents new notifications. +*/
This isn't the correct block comment format; it should look like:
/* * This pipe will get freed by the caller free_pipe_info(). * Removing this reference also prevents new notifications. */
- Eric
Quoted 8 lines · On Fri, 05 Aug 2022 12:46:17 +0530 Eric Biggers wrote:
I think patches 1 and 3 should be merged together.
Also, please use a consistent version number for all patches in the series. You have a version 1, version 2, and version 4 patch all in the same series, which is very confusing.
- Eric
Will do.
Sorry for the confusion.
Thanks,
Siddh
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [PATCH 2/3] kernel/watch_queue: Improve struct annotation formatting
In reply to Eric BiggersQuoted 25 lines · On Fri, 05 Aug 2022 12:52:11 +0530 Eric Biggers wrote:
Quoted 4 lines · Siddh Raman Pant
Improve formatting struct annotations in
watch_queue.h, so that they fall in the preferred 80 character limit.Signed-off-by Siddh Raman Pant [email protected]>This patch isn't just fixing overly long lines, but rather is introducing kerneldoc comments and documenting things that weren't documented before. That's fine, but please make the commit message accurately describe the patch.
include/linux/watch_queue.h line 22
Quoted 10 lines
include/linux/watch_queue.h +2
-0@@ -18,57 +18,103 @@struct cred ; +/** +* watch_type_filter - Filter on watch type If you're going to use kerneldoc comments, they should be correctly formatted. This is not, since it's missing the word struct. You can run this command to see the kerneldoc warnings:
Quoted 9 lines
./scripts/kernel-doc-v -noneinclude/linux/watch_queue.hinclude/linux/watch_queue.h line 23
Quoted 1 line
include/linux/watch_queue.h +1
-0+* @lock: Spinlock Please make sure that comments provide useful information and don't just repeat what the code says.
- Eric
Okay, will do.
Thanks,
Siddh
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [PATCH v2 3/3] kernel/watch_queue: Remove wqueue->defunct and use pipe for clear check
In reply to Eric BiggersQuoted 1 line · On Fri, 05 Aug 2022 12:54:31 +0530 Eric Biggers wrote:
Why is the
READ_ONCE()needed? Doesn'twqueue->lockprotectwqueue->pipe?
We are changing the pointer while a notification can be potentially waiting to be posted to the pipe. So a barrier is needed to prevent compiler magic from reloading the value.
This was remarked by David Howells here:
https://lore.kernel.org/lkml/[email protected]/
Quoted 9 lines
This isn't the correct block comment format; it should look like:
/*
* This pipe will get freed by the callerfree_pipe_info().
* Removing this reference also prevents new notifications. */- Eric
Okay, will make the change.
Thanks,
Siddh
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [PATCH v2 3/3] kernel/watch_queue: Remove wqueue->defunct and use pipe for clear check
Quoted 7 lines · Siddh Raman Pant
Quoted 1 line · On Fri, 05 Aug 2022 12:54:31 +0530 Eric Biggers wrote:
Why is the
READ_ONCE()needed? Doesn'twqueue->lockprotectwqueue->pipe?We are changing the pointer while a notification can be potentially waiting to be posted to the pipe. So a barrier is needed to prevent compiler magic from reloading the value.
wqueue->pipe is only read or written while wqueue->lock is held, so that is not an issue at all.
- Eric
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [PATCH v2 3/3] kernel/watch_queue: Remove wqueue->defunct and use pipe for clear check
Quoted 4 lines · On Sat, 06 Aug 2022 00:03:20 +0530 Eric Biggers wrote:
wqueue->pipeis only read or written whilewqueue->lockis held, so that is not an issue at all.- Eric
Thanks for explaining. I will send the v2 now.
Thanks,
Siddh
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees