mm/damon/core: fix damon_call()/damos_walk() vs kdmond exit race

damon_call() and damos_walk() can leak memory and/or deadlock when they race with kdamond terminations. Fix those.

Changes from RFC v3
(https://lore.kernel.org/[email protected])
- Drop RFC tag.
- Rebase to latest mm-new.
Changes from RFC v2
(https://lore.kernel.org/[email protected])
- Update and wordsmith commit message.
- Add damos_walk() race fix.
Changes from RFC v1
(https://lore.kernel.org/[email protected])
- Clarify damon_call() call condition.
- Init call_controls_obsolete before kdamond_started completion.
- Wordsmith commit message.
- Split out repeat_call_control leak fix from the series.

SeongJae Park (2):
mm/damon/core: fix damon_call() vs kdamond_fn() exit race mm/damon/core: fix damos_walk() vs kdamond_fn() exit race

 include/linux/damon.h |  2 ++
 mm/damon/core.c       | 66 ++++++++++++++++++-------------------------
 2 files changed, 30 insertions(+), 38 deletions(-)

base-commit: 305aff97ab8306284a0aa85f9128403b50c89019

2.47.3

When kdamond_fn() main loop is finished, the function cancels all remaining damon_call() requests and unset the damon_ctx->kdamond so that API callers and API functions themselves can know the context is terminated. damon_call() adds the caller's request to the queue first. After that, it shows if the kdamond of the damon_ctx is still running (damon_ctx->kdamond is set). Only if the kdamond is running, damon_call() starts waiting for the kdamond's handling of the newly added request.

The damon_call() requests registration and damon_ctx->kdamond unset are protected by different mutexes, though. Hence, damon_call() could race with damon_ctx->kdamond unset, and result in deadlocks.

For example, let's suppose kdamond successfully finished the damon_call() requests cancelling. Right after that, damon_call() is called for the context. It registers the new request, and shows the context is still running, because damon_ctx->kdamond unset is not yet done. Hence the damon_call() caller starts waiting for the handling of the request. However, the kdamond is already on the termination steps, so it never handles the new request. As a result, the damon_call() caller threads infinitely waits.

Fix this by introducing another damon_ctx field, namely call_controls_obsolete. It is protected by the damon_ctx->call_controls_lock, which protects damon_call() requests registration. Initialize (unset) it in kdamond_fn() before letting damon_start() returns and set it just before the cancelling of remaining damon_call() requests is executed. damon_call() reads the obsolete field under the lock and avoids adding a new request.

After this change, only requests that are guaranteed to be handled or cancelled are registered. Hence the after-registration DAMON context termination check is no longer needed. Remove it together.

Note that the deadlock will not happen when damon_call() is called for repeat mode request. In tis case, damon_call() returns instead of waiting for the handling when the request registration succeeds and it shows the kdamond is running. However, if the request also has dealloc_on_cancel, the request memory would be leaked.

The issue is found by sashiko [1].

[1] https://lore.kernel.org/[email protected]

---
 include/linux/damon.h |  1 +
 mm/damon/core.c       | 45 ++++++++++++++-----------------------------
 2 files changed, 15 insertions(+), 31 deletions(-)

index d9a3babbafc1..5129de70e7b7 100644
@@ -818,6 +818,7 @@ struct damon_ctx {
 
 	/* lists of &struct damon_call_control */
 	struct list_head call_controls;
+	bool call_controls_obsolete;
 	struct mutex call_controls_lock;
 
 	struct damos_walk_control *walk_control;

2.47.3

Forwarding Sashiko review for doing discussions via mails.

# review url: https://sashiko.dev/#/patchset/[email protected]
# start of sashiko.dev inline review
commit 9cf54c95f0c4bda07c5dbaf4019d0d731f6c58fd
Author: SeongJae Park <[email protected]>

mm/damon/core: fix damon_call() vs kdamond_fn() exit race

This commit introduces a new call_controls_obsolete flag to prevent damon_call() from adding requests to a terminating kdamond context. It removes the after-registration DAMON context termination check.

[ ... ]

    list_del(&control->list);
    kfree(control);

    mutex_unlock(&ctx->call_controls_lock);
    if (control->repeat)

# end of sashiko.dev inline review
# review url: https://sashiko.dev/#/patchset/[email protected]
#
# hkml [1] generated a draft of this mail.  It can be regenerated
# using below command:
#
#     hkml patch sashiko_dev --for_forwarding \
#             [email protected]
#
# [1] https://github.com/sjp38/hackermail

No. kdamond_call(ctx, true) is called only after call_controls_obsolete is set, under the call_control_lock. And damon_call() reads the call_controls_obsolete just after acquiring the call_control_lock, and return if it is set.

Hence CPU2 in this scenario cannot execute the mutex_unlock() part after CPU1 entered kdamond_call().

So this is also wrong.

Thanks,
SJ

[...]

When kdamond_fn() main loop is finished, the function cancels remaining damos_walk() request and unset the damon_ctx->kdamond so that API callers and API functions themselves can show the context is terminated. damos_walk() adds the caller's request to the queue first. After that, it shows if the kdamond of the damon_ctx is still running (damon_ctx->kdamond is set). Only if the kdamond is running, damos_walk() starts waiting for the kdamond's handling of the newly added request.

The damos_walk() requests registration and damon_ctx->kdamond unset are protected by different mutexes, though. Hence, damos_walk() could race with damon_ctx->kdamond unset, and result in deadlocks.

For example, let's suppose kdamond successfully finished the damow_walk() request cancelling. Right after that, damos_walk() is called for the context. It registers the new request, and shows the context is still running, because damon_ctx->kdamond unset is not yet done. Hence the damos_walk() caller starts waiting for the handling of the request. However, the kdamond is already on the termination steps, so it never handles the new request. As a result, the damos_walk() caller thread infinitely waits.

Fix this by introducing another damon_ctx field, namely walk_control_obsolete. It is protected by the damon_ctx->walk_control_lock, which protects damos_walk() request registration. Initialize (unset) it in kdamond_fn() before letting damon_start() returns and set it just before the cancelling of the remaining damos_walk() request is executed. damos_walk() reads the obsolete field under the lock and avoids adding a new request.

After this change, only requests that are guaranteed to be handled or cancelled are registered. Hence the after-registration DAMON context termination check is no longer needed. Remove it together.

The issue is found by sashiko [1].

[1] https://lore.kernel.org/[email protected]

---
 include/linux/damon.h |  1 +
 mm/damon/core.c       | 21 ++++++++++++++-------
 2 files changed, 15 insertions(+), 7 deletions(-)

index 5129de70e7b7..f2cdb7c3f5e6 100644
@@ -822,6 +822,7 @@ struct damon_ctx {
 	struct mutex call_controls_lock;
 
 	struct damos_walk_control *walk_control;
+	bool walk_control_obsolete;
 	struct mutex walk_control_lock;
 
 	/*

index 9bcda2765ac9..ddabb93f2377 100644
@@ -1637,6 +1637,10 @@ int damon_call(struct damon_ctx *ctx, struct damon_call_control *control)
  * passed at least one &damos->apply_interval_us, kdamond marks the request as
  * completed so that damos_walk() can wakeup and return.
  *
+ * Note that this function should be called only after damon_start() with the
+ * @ctx has succeeded.  Otherwise, this function could fall into an indefinite
+ * wait.
+ *
  * Return: 0 on success, negative error code otherwise.
  */
 int damos_walk(struct damon_ctx *ctx, struct damos_walk_control *control)

@@ -1644,19 +1648,16 @@ int damos_walk(struct damon_ctx *ctx, struct damos_walk_control *control)
 	init_completion(&control->completion);
 	control->canceled = false;
 	mutex_lock(&ctx->walk_control_lock);
+	if (ctx->walk_control_obsolete) {
+		mutex_unlock(&ctx->walk_control_lock);
+		return -ECANCELED;
+	}
 	if (ctx->walk_control) {
 		mutex_unlock(&ctx->walk_control_lock);
 		return -EBUSY;
 	}
 	ctx->walk_control = control;
 	mutex_unlock(&ctx->walk_control_lock);
-	if (!damon_is_running(ctx)) {
-		mutex_lock(&ctx->walk_control_lock);
-		if (ctx->walk_control == control)
-			ctx->walk_control = NULL;
-		mutex_unlock(&ctx->walk_control_lock);
-		return -EINVAL;
-	}
 	wait_for_completion(&control->completion);
 	if (control->canceled)
 		return -ECANCELED;

@@ -2932,6 +2933,9 @@ static int kdamond_fn(void *data)
 	mutex_lock(&ctx->call_controls_lock);
 	ctx->call_controls_obsolete = false;
 	mutex_unlock(&ctx->call_controls_lock);
+	mutex_lock(&ctx->walk_control_lock);
+	ctx->walk_control_obsolete = false;
+	mutex_unlock(&ctx->walk_control_lock);
 	complete(&ctx->kdamond_started);
 	kdamond_init_ctx(ctx);
 

@@ -3046,6 +3050,9 @@ static int kdamond_fn(void *data)
 	ctx->call_controls_obsolete = true;
 	mutex_unlock(&ctx->call_controls_lock);
 	kdamond_call(ctx, true);
+	mutex_lock(&ctx->walk_control_lock);
+	ctx->walk_control_obsolete = true;
+	mutex_unlock(&ctx->walk_control_lock);
 	damos_walk_cancel(ctx);
 
 	pr_debug("kdamond (%d) finishes\n", current->pid);

2.47.3

Forwarding sashiko.dev review status for this thread.

# review url: https://sashiko.dev/#/patchset/[email protected]

- [PATCH 1/2] mm/damon/core: fix damon_call() vs kdamond_fn() exit race
- status: Reviewed
- review: ISSUES MAY FOUND
- [PATCH 2/2] mm/damon/core: fix damos_walk() vs kdamond_fn() exit race
- status: Reviewed
- review: No issues found.

# hkml [1] generated a draft of this mail.  It can be regenerated
# using below command:
#
#     hkml patch sashiko_dev --thread_status --for_forwarding \
#             [email protected]
#
# [1] https://github.com/sjp38/hackermail

Sent using hkml (https://github.com/sjp38/hackermail)