fix/update BLE log SPI LBM (IDFGH-17969)

[lifeiyu-u]

Description

Related

Testing

---

Checklist

Before submitting a Pull Request, please ensure the following:

• 🚨 This PR does not introduce breaking changes.
• All CI checks (GH Actions) pass.
• Documentation is updated as needed.
• Tests are updated or added as necessary.
• Code is well-commented, especially in complex areas.
• Git history is clean — commits are squashed to the minimum necessary.

---

Note

High Risk
Large concurrency and lifecycle changes in ISR/task logging, flush/deinit, and sdkconfig tuning; mis-tuning or races could cause dropped logs, stalls, or use-after-free if shutdown ordering regresses.

Overview
Replaces the old multi-LBM design (separate atomic/spin/Link Layer managers and per-LBM transport sizing) with a single global pool of transport buffers shared by task, ISR, and Link Layer writers.

Kconfig drops per-LBM options (auto-flush, atomic LBM counts, common/LL total memory) and adds pool tuning: BLE_LOG_POOL_TRANS_CNT, ISR reserve count, per-buffer size, optional ISR OPEN-scan cap, and task backpressure timeout.

Allocation path uses per-buffer atomic_lock, FREE/OPEN/SENDING lifecycle, and free/open hint bitmaps with round-robin scan. Task writers can block on a counting semaphore (with timeout) when the pool is full; ISR writers use shared buffers first, then a reserved ISR region, and never block. Recycle is centralized in ble_log_lbm_recycle_trans() from runtime drain and UART/SPI/dummy tx-done or send-failure paths.

Flush/deinit wake blocked tasks, wait on both ref count and waiting_task_count, seal all OPEN buffers, then wait for pool inflight to drain; per-LBM utilization reporting collapses to one pool record. UART port-0 redirection stays on a separate ble_log_redir_t (not pool IDs). ble_log_write_final_stat / ble_log_write_internal and the old LBM acquire/release write path are removed in favor of direct pool acquire + ble_log_pool_write_frame.

^{Reviewed by Cursor Bugbot for commit 77586fb. Bugbot is set up for automated code reviews on this repo. Configure here.}

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 3 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 77586fb. Configure here.}

[cursor]

LL task writers can block

High Severity

ble_log_write_hex_ll in task context now goes through ble_log_pool_acquire with backpressure enabled, so the Link Layer or HCI task can block on xSemaphoreTake when the pool is empty. Previously those paths used dedicated non-blocking LBMs and dropped on exhaustion. Blocking the controller task under log pressure can stall BLE scheduling.

Additional Locations (1)

• components/bt/common/ble_log/src/ble_log_lbm.c#L324-L363

 

^{Reviewed by Cursor Bugbot for commit 77586fb. Configure here.}

[cursor]

Flush lock spin can hang

High Severity

ble_log_flush busy-waits on ble_log_cas_acquire with no yield or delay. If a lower-priority writer holds atomic_lock and flush runs at higher priority, the writer never resumes, the spin never exits, and the system hangs until the watchdog.

 

^{Reviewed by Cursor Bugbot for commit 77586fb. Configure here.}

[cursor]

HCI upstream limit skipped

Medium Severity

When BLE_LOG_LL_HCI_LOG_PAYLOAD_LEN_LIMIT_ENABLED is set, truncation only runs for BLE_LOG_SRC_LL_HCI. Upstream HCI logs (BLE_LOG_SRC_HCI / BLE_LOG_LL_FLAG_HCI_UPSTREAM) no longer get the limit, so large len_append values can exceed the intended cap or fail to fit a transport.

 

^{Reviewed by Cursor Bugbot for commit 77586fb. Configure here.}