Bluetooth: Controller: Replace prepare pipeline with ordered list by cvinayak · Pull Request #79444 · zephyrproject-rtos/zephyr

cvinayak · 2024-10-05T03:52:44Z

Replace prepare pipeline with ordered linked list.

sonarqubecloud · 2025-08-09T06:32:50Z

Quality Gate passed

Issues
20 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
5.0% Duplication on New Code

See analysis details on SonarQube Cloud

cvinayak · 2025-10-22T16:29:36Z

I can’t post comments to GitHub from here. Please paste the following into a PR comment:

data:
- url: "https://github.com/zephyrproject-rtos/zephyr/pull/79444"
  state: "open"
  draft: true
  title: "Bluetooth: Controller: Replace prepare pipeline with ordered list"
  number: 79444
  created_at: "05 October 2024"
  closed_at: ""
  merged_at: ""
  labels:
  author: "cvinayak"
  comments: 2
  assignees_avatar_urls:
- url: "https://github.com/zephyrproject-rtos/zephyr/pull/79444"
  state: "open"
  draft: true
  title: "Bluetooth: Controller: Replace prepare pipeline with ordered list"
  number: 79444
  created_at: "05 October 2024"
  closed_at: ""
  merged_at: ""
  labels:
  author: "cvinayak"
  comments: 2
  assignees_avatar_urls:

Summary of what changed

Replaced the MFIFO-based prepare pipeline with an explicitly ordered singly linked list in subsys/bluetooth/controller/ll_sw/ull.c.
- New pipeline structure: pool-backed nodes storing a next pointer plus struct lll_event.
- Enqueue is now order-aware by ticks_at_expire and treats resume events specially.
- Dequeue/iteration now uses void** as an iterator cursor (NULL means start/end), replacing uint8_t indices.
API updates:
- ull_prepare_dequeue_iter signature changed from uint8_t* to void** (see subsys/bluetooth/controller/ll_sw/lll.h). All call sites in Nordic and OpenISA LLL updated accordingly.
Nordic LLL changes (subsys/bluetooth/controller/ll_sw/nordic/lll/lll.c):
- prepare_dequeue_iter_ready_get now filters out aborted, resume, and defer entries.
- lll_disable, lll_prepare_resolve, and preempt reworked for the new iterator. Notably, the prior workaround that aborted earlier prepares to make room for “short” prepares has been removed; ordering is handled by the list.
OpenISA LLL changes (subsys/bluetooth/controller/ll_sw/openisa/lll/lll.c):
- Equivalent iterator and control-flow updates to match the new pipeline semantics.
Scan updates:
- lll_scan.h adds a persistent lll_prepare_param field when advanced scheduler is enabled.
- lll_scan.c and lll_scan_aux.c copy timing fields into lll->prepare_param and pass that to mayfly, avoiding lifetime issues from passing a stack/local param.
Initialization:
- The pipeline memory is initialized in init_reset via mem_init; MFIFO prep init removed.

Why this is a good change

Correctness and determinism: The prepare pipeline is now ordered by scheduling time instead of FIFO plus ad-hoc aborts. This removes the previous “abort to reorder” workaround and should reduce spurious aborts and reschedules, especially for short preemptors.
Simpler preemption logic: Preempt handling is cleaner; instead of scanning/aborting earlier entries to accommodate a short event, the ordered structure ensures correct sequencing.
Safer parameter passing in scan paths: Copying timing/param state into lll->prepare_param before scheduling mayfly avoids dangling references to transient stack data.
Clearer iteration semantics: Using a pointer cursor for ull_prepare_dequeue_iter is a better fit for list traversal than synthetic indices, reducing off-by-one/overflow risks.

Things to double-check

Concurrency/atomicity: MFIFO provided a well-tested queue; the new list relies on mem_acquire/mem_release and pointer manipulation. Ensure all accesses occur under the same context/locking assumptions as before and that no ISR/LLL context can mutate the list concurrently with ULL.
Ordering conditions: The insertion logic relies on ticker_ticks_diff_get and MSBIT checks. These are notoriously subtle with wrap-around; unit/soak tests around boundary conditions (near wrap and equal times) would be valuable.
Resume event handling: The enqueue code treats resume specially (often appended). The dequeue path enqueues at most two non-resume events before switching behavior; verify this preserves prior fairness/progress guarantees and doesn’t starve resumes or normal events under load.
Memory footprint: Each pipeline node adds a pointer over the previous structure. The increase is small but present; confirm this fits tight controller builds.
API ripple: All call sites appear updated for the new iterator type in Nordic and OpenISA; ensure there are no out-of-tree or less commonly built targets relying on the old signature.

Overall assessment

This is a solid improvement. It addresses a known limitation of the FIFO pipeline, removes complex workarounds, and makes scheduling behavior more predictable. With careful validation around concurrency and time-diff edge cases, it looks like a positive and maintainable change.

sonarqubecloud · 2025-11-19T05:17:57Z

Quality Gate passed

Issues
17 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
4.0% Duplication on New Code

See analysis details on SonarQube Cloud

Copilot

Pull Request Overview

This PR replaces the MFIFO (multi-producer, multi-consumer FIFO) prepare pipeline with an ordered linked list implementation in the Bluetooth controller. The new implementation maintains events sorted by their ticks_at_expire value, eliminating the need for O(n) searches through the pipeline.

Key Changes:

Replaced MFIFO-based prepare pipeline with a custom ordered linked list structure
Updated iterator functions to use void ** instead of uint8_t indices for traversal
Simplified preemption logic in nordic/lll/lll.c by removing complex short-prepare search code

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 12 comments.

File	Description
subsys/bluetooth/controller/ll_sw/ull.c	Implements new ordered linked list structure, enqueue/dequeue logic with time-based ordering, and memory management for pipeline events
subsys/bluetooth/controller/ll_sw/openisa/lll/lll.c	Updates iterator usage from uint8_t index to void* pointer and adds list termination checks
subsys/bluetooth/controller/ll_sw/nordic/lll/lll.c	Updates iterator usage, simplifies preemption logic by removing short-prepare search loops, and converts do-while to while loop in prepare_dequeue_iter_ready_get
subsys/bluetooth/controller/ll_sw/lll.h	Updates function signature for ull_prepare_dequeue_iter to accept void** instead of uint8_t*

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-11-19T05:40:45Z

subsys/bluetooth/controller/ll_sw/ull.c

+	/* Initialize prepare pipeline list */
+	mem_init(pipeline.mem.pool, EVENT_PIPELINE_EVENT_SIZE,
+		 sizeof(pipeline.mem.pool) / EVENT_PIPELINE_EVENT_SIZE,
+		 &pipeline.mem.free);


Missing initialization of pipeline.head and pipeline.tail to NULL. These pointers should be explicitly initialized to NULL during reset to ensure proper operation of the linked list, especially since ull_prepare_enqueue checks if (pipeline.tail) at line 2153.

Suggested change

&pipeline.mem.free);

&pipeline.mem.free);

pipeline.head = NULL;

pipeline.tail = NULL;

Copilot · 2025-11-19T05:40:46Z

subsys/bluetooth/controller/ll_sw/ull.c

+					if (!*next) {
+						pipeline.tail = next;
+					}


Potential issue with the insertion logic when placing a new event at the head of the list. At line 2204, the check if (!*next) is intended to handle the case where the list becomes empty after moving to head, but this condition can never be true here. The *next was just set to pipeline.head at line 2174, and if pipeline.head was NULL, the outer condition at line 2153 (if (pipeline.tail)) would have been false. Consider removing this unreachable condition or documenting why it exists.

Suggested change

if (!*next) {

pipeline.tail = next;

}

Copilot · 2025-11-19T05:40:46Z

subsys/bluetooth/controller/ll_sw/ull.c

-	struct lll_event *next;
+	static memq_link_t link;
+	static struct mayfly mfy = {0, 0, &link, NULL, lll_resume};
+	uint8_t enqueued_loop = 2U;


The variable enqueued_loop is initialized to 2U but its purpose and why it's set to 2 specifically is unclear. This magic number should be documented with a comment explaining why 2 iterations are allowed before setting is_enqueued flag, or defined as a named constant.

Copilot · 2025-11-19T05:40:46Z

subsys/bluetooth/controller/ll_sw/ull.c

+				/* Trigger the LLL resume */
+				mfy.param = e;
+				ret = mayfly_enqueue(caller_id, TICKER_USER_ID_LLL, 0, &mfy);
+				LL_ASSERT(!ret);


Inconsistent assertion macro usage: LL_ASSERT is used here (line 2344) while LL_ASSERT_ERR is used elsewhere in the same function (line 2306). Consider using the same assertion macro consistently throughout the function, or document why different macros are needed in different locations.

Suggested change

LL_ASSERT(!ret);

LL_ASSERT_ERR(!ret);

Copilot · 2025-11-19T05:40:46Z

subsys/bluetooth/controller/ll_sw/nordic/lll/lll.c

-		}
+		/* Start the preempt timeout for ready event */
+		ret = preempt_ticker_start(ready, NULL, ready);
+		LL_ASSERT((ret == TICKER_STATUS_SUCCESS) ||


Inconsistent assertion macro usage: LL_ASSERT is used here (line 1232) while LL_ASSERT_ERR was used in the previous code (visible in the diff context). Consider using the same assertion macro consistently, or document why different macros are needed.

Suggested change

LL_ASSERT((ret == TICKER_STATUS_SUCCESS) ||

LL_ASSERT_ERR((ret == TICKER_STATUS_SUCCESS) ||

Copilot · 2025-11-19T05:40:47Z

subsys/bluetooth/controller/ll_sw/ull.c

-	return MFIFO_DEQUEUE_GET(prep);
+	void *e;
+
+	/* peak lll_event */


Typo in comment: "peak" should be "peek". The function is inspecting/peeking at the head element, not finding a peak/maximum.

Copilot · 2025-11-19T05:40:47Z

subsys/bluetooth/controller/ll_sw/ull.c

+		return NULL;
+	}
+
+	/* Peak lll_event */


Typo in comment: "Peak" should be "Peek". The function is inspecting/peeking at the event, not finding a peak/maximum.

Copilot · 2025-11-19T05:40:48Z

subsys/bluetooth/controller/ll_sw/ull.c

+		/* Trigger the LLL resume */
+		mfy.param = e;
+		ret = mayfly_enqueue(caller_id, TICKER_USER_ID_LLL, 0, &mfy);
+		LL_ASSERT(!ret);


Inconsistent assertion macro usage: LL_ASSERT is used here (line 2375) while LL_ASSERT_ERR is used at line 2306. Consider using the same assertion macro consistently throughout the function, or document why different macros are needed in different locations.

Suggested change

LL_ASSERT(!ret);

LL_ASSERT_ERR(!ret);

Copilot · 2025-11-19T05:40:48Z

subsys/bluetooth/controller/ll_sw/nordic/lll/lll.c

 preempt_abort_resume:
-		/* Abort any duplicate non-resume, that they get dequeued */
-		iter_idx = UINT8_MAX;
+		/* Abort any duplicates so that they get dequeued */


[nitpick] Comment text update has a minor issue: "Abort any duplicate non-resume" (old) was changed to "Abort any duplicates" (new), but the old comment was grammatically incomplete. The new comment is better but could be more precise: "Abort any duplicate events so that they get dequeued" would be clearer.

Suggested change

/* Abort any duplicates so that they get dequeued */

/* Abort any duplicate events so that they get dequeued */

Copilot · 2025-11-19T05:40:48Z

subsys/bluetooth/controller/ll_sw/ull.c

+					if (!e_curr->is_aborted && !e_curr->is_resume) {
+						prev = curr;
+					}


Critical bug in list insertion ordering: The prev pointer is only updated for non-aborted, non-resume events (lines 2184-2186), but the list should maintain all nodes including aborted ones until they are explicitly dequeued. If the list contains aborted events followed by a valid event that should come after the new event (e.g., [aborted1] -> [aborted2] -> [valid_event]), the loop will skip updating prev while traversing aborted events, leaving prev as NULL. When the loop exits upon finding valid_event, prev is still NULL, causing the new event to be incorrectly inserted at the head (lines 2202-2203) instead of before valid_event. The fix is to unconditionally update prev = curr before advancing curr at line 2188, regardless of the event's state.

Suggested change

if (!e_curr->is_aborted && !e_curr->is_resume) {

prev = curr;

}

prev = curr;

github-actions · 2026-01-19T00:47:25Z

This pull request has been marked as stale because it has been open (more than) 60 days with no activity. Remove the stale label or add a comment saying that you would like to have the label removed otherwise this pull request will automatically be closed in 14 days. Note, that you can always re-open a closed pull request at any time.

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 7 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-25T21:11:35Z

subsys/bluetooth/controller/ll_sw/openisa/lll/lll.c

 				 *       through the prepare pipeline.
 				 */
-				idx = UINT8_MAX;
+				idx = NULL;


Inside the iter_idx traversal loop, the code resets idx = NULL after abort_cb() to restart iteration, but the iterator variable being used is iter_idx. As written, the loop won't actually restart from the head after the pipeline is modified. This looks like it should reset iter_idx instead of idx.

Suggested change

idx = NULL;

iter_idx = NULL;

Copilot · 2026-01-25T21:11:35Z

subsys/bluetooth/controller/ll_sw/ull.c

+#define EVENT_PIPELINE_EVENT_SIZE (sizeof(struct lll_event) + sizeof(void *))
+
+static struct {
+	struct {
+		void *free;
+		uint8_t pool[EVENT_PIPELINE_MAX * EVENT_PIPELINE_EVENT_SIZE];
+	} mem;


EVENT_PIPELINE_EVENT_SIZE and the pipeline node layout don't account for the mem_* allocator's reserved header (next pointer + free-count stored at MROUND(sizeof(void *)), see util/mem.c). With the current layout (lll_event starts at + sizeof(void *)), mem_release() will overwrite the first bytes of struct lll_event, corrupting event contents if the block is released while still referenced (e.g., passed to mayfly). Consider placing struct lll_event after MEM_FREE_RESERVED_SIZE (and aligning as needed) and updating EVENT_PIPELINE_EVENT_SIZE/offset calculations accordingly.

Copilot · 2026-01-25T21:11:36Z

subsys/bluetooth/controller/ll_sw/ull.c

 		return NULL;
 	}

+	e = (void *)((uint8_t *)next + sizeof(void *));


e is derived as next + sizeof(void *), but the mem_* pool uses the bytes immediately after the next pointer to store the free-count (mem_release() writes it at MROUND(sizeof(void *))). If this block is released back to the pool before the consumer is done, e->prepare_param will be corrupted. Adjust the start offset of lll_event to be after the allocator's reserved header (e.g., MEM_FREE_RESERVED_SIZE, possibly rounded up for alignment).

Suggested change

e = (void *)((uint8_t *)next + sizeof(void *));

e = (void *)((uint8_t *)next + MEM_FREE_RESERVED_SIZE);

Copilot · 2026-01-25T21:11:36Z

subsys/bluetooth/controller/ll_sw/ull.c

+				/* Trigger the LLL resume */
+				mfy.param = e;
+				ret = mayfly_enqueue(caller_id, TICKER_USER_ID_LLL, 0, &mfy);
+				LL_ASSERT_ERR(!ret);
+
+				/* Pipeline could be modified, restart at head to remove all aborted
+				 * prepare.
+				 */
+				e_prev = NULL;
+				e_curr = pipeline.head;
 			}
+
+			/* Free event memory */
+			mem_release(head, &pipeline.mem.free);
+


ull_prepare_dequeue() passes e to mayfly_enqueue() (which may defer execution when caller/callee priorities are not equal), but then immediately mem_release(head, ...) frees the backing storage. Since mem_release() overwrites the freed block's header (and free-count), lll_resume can observe corrupted struct lll_event data if the mayfly call is not executed inline. Keep the event storage valid until lll_resume has run (e.g., release in/after lll_resume, or only release when mayfly_enqueue executed inline / assert on inline-only usage).

Copilot · 2026-01-25T21:11:36Z

subsys/bluetooth/controller/ll_sw/ull.c

+		/* Release the memory to pool */
+		mem_release(head, &pipeline.mem.free);


Same lifetime issue as above for resume prepares: the resume lll_event is passed to mayfly_enqueue() and then immediately mem_release(head, ...) frees/overwrites the memory. If the mayfly call is deferred, lll_resume may dereference invalid/corrupted data. Defer releasing the node until after lll_resume executes.

Suggested change

/* Release the memory to pool */

mem_release(head, &pipeline.mem.free);

/* Note: Do not release the resume event memory here.

* The node must remain valid until after lll_resume has executed.

* It will be released as part of the resume/done handling.

*/

Copilot · 2026-01-25T21:11:36Z

subsys/bluetooth/controller/ll_sw/ull.c

+	/* peak lll_event */
+	if (pipeline.head) {


Typo in comment: peak lll_event should be peek lll_event.

Copilot · 2026-01-25T21:11:37Z

subsys/bluetooth/controller/ll_sw/ull.c

+	/* Peak lll_event */
+	e = (uint8_t *)*idx + sizeof(void *);


Typo in comment: Peak lll_event should be Peek lll_event.

…os#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

…ct-rtos#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

…os#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

…ct-rtos#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

Replace prepare pipeline with ordered linked list. Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>

sonarqubecloud · 2026-02-14T06:05:46Z

Quality Gate passed

Issues
2 New issues
9 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
3.7% Duplication on New Code

See analysis details on SonarQube Cloud

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch 5 times, most recently from d3fd40c to d377960 Compare October 7, 2024 13:52

cvinayak closed this Oct 13, 2024

cvinayak reopened this Aug 9, 2025

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch 2 times, most recently from 01d6254 to 5224961 Compare August 9, 2025 06:11

cvinayak closed this Aug 21, 2025

cvinayak reopened this Sep 19, 2025

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch from 5224961 to d5aea11 Compare September 19, 2025 19:51

cvinayak closed this Sep 20, 2025

cvinayak mentioned this pull request Oct 11, 2025

Bluetooth: Controller: In-system profiling and CPU overhead assertions #95391

Merged

cvinayak reopened this Oct 17, 2025

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch 2 times, most recently from 37f9ac4 to 7ba61e4 Compare October 18, 2025 05:58

cvinayak mentioned this pull request Oct 20, 2025

Bluetooth: Controller: Add code comments related to short prepare #97863

Merged

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch 3 times, most recently from 1f03b00 to 3308317 Compare October 22, 2025 16:17

cvinayak marked this pull request as ready for review October 22, 2025 16:58

zephyrbot added area: Bluetooth area: Bluetooth Controller labels Oct 22, 2025

zephyrbot requested review from carlescufi, erbr-ot, ppryga and thoh-ot October 22, 2025 17:00

zephyrbot requested review from Tronil and wopu-ot October 22, 2025 17:00

zephyrbot assigned cvinayak Oct 22, 2025

cvinayak added the Enhancement Changes/Updates/Additions to existing features label Oct 27, 2025

cvinayak added this to the v4.4.0 milestone Oct 27, 2025

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch from 3308317 to ed1da0f Compare October 28, 2025 12:18

cvinayak mentioned this pull request Oct 30, 2025

Bluetooth: Controller: Place holder for a list of known assertions #82399

Open

15 tasks

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch from ed1da0f to 8c40860 Compare November 19, 2025 05:13

cvinayak requested a review from Copilot November 19, 2025 05:33

Copilot started reviewing on behalf of cvinayak November 19, 2025 05:34 View session

Copilot finished reviewing on behalf of cvinayak November 19, 2025 05:39

Copilot AI reviewed Nov 19, 2025

View reviewed changes

github-actions bot added the Stale label Jan 19, 2026

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch from 8c40860 to 92fb0f9 Compare January 25, 2026 20:54

cvinayak requested a review from Copilot January 25, 2026 21:01

Copilot started reviewing on behalf of cvinayak January 25, 2026 21:01 View session

Copilot AI reviewed Jan 25, 2026

View reviewed changes

github-actions bot removed the Stale label Jan 26, 2026

Copilot AI mentioned this pull request Jan 26, 2026

Add unit tests for prepare pipeline ordered list implementation (PR #79444) cvinayak/zephyr#12

Draft

52 tasks

Copilot AI added a commit to cvinayak/zephyr that referenced this pull request Jan 26, 2026

Add comprehensive unit tests for prepare pipeline PR zephyrproject-rt…

62ee1ec

…os#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

Copilot AI added a commit to cvinayak/zephyr that referenced this pull request Jan 26, 2026

Add integration guide and clarify tests are for future PR zephyrproje…

dabcbdc

…ct-rtos#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

cvinayak added a commit to cvinayak/zephyr that referenced this pull request Jan 27, 2026

Add comprehensive unit tests for prepare pipeline PR zephyrproject-rt…

9c0287d

…os#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

cvinayak added a commit to cvinayak/zephyr that referenced this pull request Jan 27, 2026

Add integration guide and clarify tests are for future PR zephyrproje…

d48b2f8

…ct-rtos#79444 Co-authored-by: cvinayak <6350656+cvinayak@users.noreply.github.com>

Bluetooth: Controller: Replace prepare pipeline with ordered list

997a1e0

Replace prepare pipeline with ordered linked list. Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>

cvinayak force-pushed the github_bt_ctlr_pipeline_fix branch from 92fb0f9 to 997a1e0 Compare February 14, 2026 06:00

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Bluetooth: Controller: Replace prepare pipeline with ordered list#79444

Bluetooth: Controller: Replace prepare pipeline with ordered list#79444
cvinayak wants to merge 1 commit intozephyrproject-rtos:mainfrom
cvinayak:github_bt_ctlr_pipeline_fix

cvinayak commented Oct 5, 2024

sonarqubecloud bot commented Aug 9, 2025

cvinayak commented Oct 22, 2025

sonarqubecloud bot commented Nov 19, 2025

Copilot AI left a comment

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

Copilot AI Nov 19, 2025

github-actions bot commented Jan 19, 2026

Copilot AI left a comment

Copilot AI Jan 25, 2026

Copilot AI Jan 25, 2026

Copilot AI Jan 25, 2026

Copilot AI Jan 25, 2026

Copilot AI Jan 25, 2026

Copilot AI Jan 25, 2026

Copilot AI Jan 25, 2026

sonarqubecloud bot commented Feb 14, 2026

Labels

3 participants

	LL_ASSERT((ret == TICKER_STATUS_SUCCESS) \|\|
	LL_ASSERT_ERR((ret == TICKER_STATUS_SUCCESS) \|\|

	/* Abort any duplicates so that they get dequeued */
	/* Abort any duplicate events so that they get dequeued */

	e = (void )((uint8_t )next + sizeof(void *));
	e = (void )((uint8_t )next + MEM_FREE_RESERVED_SIZE);

		/* Release the memory to pool */
		mem_release(head, &pipeline.mem.free);

-		/* Release the memory to pool */
-		mem_release(head, &pipeline.mem.free);
+		/* Note: Do not release the resume event memory here.
+		 * The node must remain valid until after lll_resume has executed.
+		 * It will be released as part of the resume/done handling.
+		 */

Conversation

cvinayak commented Oct 5, 2024

sonarqubecloud bot commented Aug 9, 2025

Quality Gate passed

cvinayak commented Oct 22, 2025

sonarqubecloud bot commented Nov 19, 2025

Quality Gate passed

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Key Changes:

Reviewed Changes

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

github-actions bot commented Jan 19, 2026

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Copilot AI Jan 25, 2026

Choose a reason for hiding this comment

Copilot AI Jan 25, 2026

Choose a reason for hiding this comment

Copilot AI Jan 25, 2026

Choose a reason for hiding this comment

Copilot AI Jan 25, 2026

Choose a reason for hiding this comment

Copilot AI Jan 25, 2026

Choose a reason for hiding this comment

Copilot AI Jan 25, 2026

Choose a reason for hiding this comment

Copilot AI Jan 25, 2026

Choose a reason for hiding this comment

sonarqubecloud bot commented Feb 14, 2026

Quality Gate passed

Labels

3 participants