bt/ble_log: release fallback mutex on host write errors (IDFGH-17872)

[Old-Ding]

Summary

This fixes the fallback path in ble_log_spi_out_host_write() so that LOG_MODULE_MUTEX(ul) is released on host write errors.

When no host task mapping is available, the function falls back to the ul log buffer and takes LOG_MODULE_MUTEX(ul). Two error paths after that point returned -1 directly, leaving the fallback mutex locked.

Changes

• Use a single local exit path for ble_log_spi_out_host_write() after fallback lock ownership can exist.
• Preserve the existing successful write behavior.
• Keep the change limited to the deprecated BLE log SPI output implementation.

Testing

• git diff --check origin/master..HEAD
• rg -n "ble_log_spi_out_host_write|fallback|LOG_MODULE_MUTEX\\(ul\\)|return -1|goto exit|ret =" components/bt/common/ble_log/deprecated/ble_log_spi_out.c
• git diff --stat origin/master..HEAD

Build not run locally: idf.py is not available in this Windows environment.

---

Note

Low Risk
Small, localized fix in deprecated BLE log SPI code; behavior on success is unchanged and only error cleanup is corrected.

Overview
Fixes a mutex leak in deprecated ble_log_spi_out_host_write() when the NimBLE fallback path is used (no host task mapping → takes LOG_MODULE_MUTEX(ul) and uses the ul log buffer).

Previously, oversized prefix and failed string formatting (total_len == 0) returned -1 immediately and never called xSemaphoreGive, which could deadlock other BLE log SPI output on ul.

The change routes those failures through a shared exit label, sets ret to 0 only on success, and always releases the fallback mutex before returning.

^{Reviewed by Cursor Bugbot for commit e349ca3. Bugbot is set up for automated code reviews on this repo. Configure here.}

[CLAassistant]

All committers have signed the CLA.

[Old-Ding]

Maintainer ping: CLA and the current checks are passing, the branch is mergeable, and this is still at Status: Opened. Could this be routed to the Bluetooth owner/reviewer for triage when possible?