[PATCH v4 0/6] Add support for IPQ5018 Bluetooth
From: George Moussalem via B4 Relay
Date: Mon Jul 06 2026 - 13:34:48 EST
Hello,
This patch series introduces Bluetooth support for IPQ5018.
Bluetooth firmware is loaded by the host into a dedicated reserved
memory carveout and authenticated by TrustZone. A Secure Channel Manager
(SCM) call safely brings the peripheral core out of reset.
A shared memory ring buffer topology handles runtime data frame
transport between the host APSS and the Bluetooth controller.
An outgoing APCS IPC bit and an incoming GIC interrupt handle
host/guest signaling.
This series has been tested and verified on various IPQ5018 router
boards utilizing firmware extracted from GPL distributions, using both
mdt and mbn file formats.
[ 14.781511] Bluetooth: hci0: QCA Product ID :0x00000016
[ 14.781583] Bluetooth: hci0: QCA SOC Version :0x20180100
[ 14.785926] Bluetooth: hci0: QCA ROM Version :0x00000100
[ 14.791546] Bluetooth: hci0: QCA Patch Version:0x00003ded
[ 14.796698] Bluetooth: hci0: QCA controller version 0x01000100
[ 14.802217] Bluetooth: hci0: QCA Downloading qca/mpnv10.bin
[ 16.393850] Bluetooth: hci0: QCA FW Build: BTFW.MAPLE.1.0.0-00102-MPL_ROM_PATCHZ-1
Best regards,
George Moussalem
Signed-off-by: George Moussalem <george.moussalem@xxxxxxxxxxx>
---
Changes in v4:
- Ordered switch statements in btqca.c. NOTE: left enum qca_bt_soc_type
unordered in the header file as the btqca driver has conditionally
checks whether bt_soc_type is above a certain value:
'if (soc_type >= QCA_WCN3991)' is found in 5 places
- Fixed TX path serialization by protecting btqcomipc_send and
btss_send ring access with desc->lock, preventing concurrent updates
of ring indices and free counters from TX and worker contexts.
- Fixed and simplified firmware ring traversal to decode little-endian
next and translate firmware offsets via TO_APPS_ADDR in all ring
walks, avoiding invalid pointer dereferences.
- Added short-message RX length validation (msg_len must not exceed
IPC_MSG_PLD_SZ when not long message), preventing out-of-bounds read
from inline shared-memory payloads.
- Fixed RX allocation failure behavior to advance ring index and
continue, preventing permanent RX ring stall on transient kmalloc
failures.
- Replace macro for calculting large buffer size that performed unsafe
arithmetic by an inline function that safely calculates the size by
reading values using endian functions le**_to_cpu().
- Declared aux_ptr as a zero-initialized struct as opposed to using
memset and changed struct definition by updating its member to use
u32 as opposed to __le32 as these values aren't set by firmware and
need not be validated on endianness.
- Fixed long-message slot leak when short TX ring is full by
snapshotting and rolling back long-buffer accounting on
allocation-success/short-ring-fail path.
- Updated firmware load logic to remove call to qcom_mdt_pas_init for
which the export was recently removed and replaced it by calls to
read the metadata and the validation of it using SCM.
NOTE: Sashiko correctly pointed out that the firmware ELF/mdt headers
weren't validated in the driver. However, this is done by the call to
qcom_mdt_read_metadata which returns -EINVAL if the mdt header isn't
valid.
- Hardened firmware segment bounds checks against integer overflow by
using subtraction form (filesz <= mem_size - vaddr).
- Replaced unsafe firmware split name rewrite with validated stem
handling and kasprintf formatting, removing underflow/overflow risk.
- Converted workqueue allocation to devm_alloc_ordered_workqueue to fix
workqueue leak in IRQ registration failure path.
- Added cleanup path in btqcomipc_open to shut down controller on
startup timeout, avoiding leaving peripheral powered after failed open
- Reordered probe sequence to allocate/attach hdev before IRQ
registration, eliminating early-spurious-IRQ dereference window for
desc->hdev.
- Reordered remove path to unregister/free hdev before driver deinit,
preventing send path from racing against queue/work teardown.
- Corrected HCI command packet length parsing to use 8-bit plen directly
instead of using le16_to_cpu.
- Fixed endianness and address translation in TX ACK processing path.
Converted firmware-owned ring metadata before use in btss_process_ack:
rbuf base offset, msg header, long-message pointer, and message length
are now read via le32_to_cpu or le16_to_cpu prior to reclaim logic.
- Added explicit producer ordering barrier in TX ring publish path.
Inserted dma_wmb between payload or header writes and widx update so
firmware cannot observe producer index advancement before data becomes
visible.
- Removed info message printing FW build info as that was recently
introduced in the kernel in commit bdea21b3be43
- Added dependency on OF and ARCH_QCOM symbols to Kconfig. In addition,
also select QCOM_MDT_LOADER and QCOM_SCM.
- Removed syscon compatible from mailbox node, avoiding dt_binding_check
failures and update the syscon acquisition logic in driver probe
accordingly.
- Link to v3: https://lore.kernel.org/r/20260703-ipq5018-bluetooth-v3-0-62da72818ab3@xxxxxxxxxxx
Changes in v3:
- Consolidated the SCM and helper call into one function which now
checks availability of the SCM call inline.
- Changed argument to boolean type to set ECO power mode on or off as
opposed to passing a raw power mode value.
- Removed unneeded parentheses in if statement
- Declared aux_ptr as a zero-initialized struct as opposed to using
memset
- Changed for loop to a while loop for better readability
- Reordered function definition of btqcomipc_update_stats to avoid the
need for a forward declaration.
- Removed unneeded header files from btqcomipc.c (bits.h, device.h,
init.h, kernel.h, of_irq.h, types.h)
- Link to v2: https://lore.kernel.org/r/20260629-ipq5018-bluetooth-v2-0-02770f03b6bb@xxxxxxxxxxx
Changes in v2:
- removed unused struct member btss_reset from m0_btss struct.
- used devm_reset_control_get_exclusive_deasserted to acquire and
deassert btss_reset in one call. No need to explicitly assert upon
unbind since devres will handle it.
- As per Bjorn's comments, collapsed the rproc and BT driver into one
which now handles firmware loading and transport.
- As per further review comments and above change, squashed the two
devicetree bindings into one to represent the Bluetooth controller
- Added new SCM call to drive Bluetooth power modes. Based on testing on
more router boards, this is required during start/stop sequences to
avoid the controller hanging. This is only supported on relatively
newer QSEE versions, so the SCM call is checked for availability
before use.
- Refactored resource lifecycle management in the driver so that IRQ and
work queue are managed at the driver level (probe/remove) while FW
loading and powering on/off the controller is handled upon hdev
open/close.
- Consolidated TX send functions for custom IPC and HCI frames into one
and solved a potential NULL pointer dereference issue under TX
pressure.
- Replace code to load and initialize firmware metadate by existing
qcom_mdt_pas_init function.
- Solved an off byone calculation error in the RX parser.
- Added entry to MAINTAINERS file
- Link to v1: https://lore.kernel.org/r/20260625-ipq5018-bluetooth-v1-0-d999be0e04f7@xxxxxxxxxxx
---
George Moussalem (6):
dt-bindings: net: bluetooth: Document Qualcomm IPQ5018 Bluetooth controller
Bluetooth: btqca: Add IPQ5018 support
firmware: qcom: scm: Add support for setting Bluetooth power modes
Bluetooth: Introduce Qualcomm IPQ5018 IPC based HCI driver
arm64: dts: qcom: ipq5018: add nodes required for Bluetooth support
MAINTAINERS: Add entry for Qualcomm IPQ5018 Bluetooth driver
.../bindings/net/bluetooth/qcom,ipq5018-bt.yaml | 86 ++
MAINTAINERS | 7 +
arch/arm64/boot/dts/qcom/ipq5018.dtsi | 22 +
drivers/bluetooth/Kconfig | 14 +
drivers/bluetooth/Makefile | 1 +
drivers/bluetooth/btqca.c | 16 +
drivers/bluetooth/btqca.h | 3 +
drivers/bluetooth/btqcomipc.c | 1148 ++++++++++++++++++++
drivers/firmware/qcom/qcom_scm.c | 44 +
drivers/firmware/qcom/qcom_scm.h | 1 +
include/linux/firmware/qcom/qcom_scm.h | 1 +
11 files changed, 1343 insertions(+)
---
base-commit: 4d1ab324fcb7d20df5a071edb0304461846fdc12
change-id: 20260625-ipq5018-bluetooth-06ff66c9d753
prerequisite-message-id: <20260703-block-as-nvmem-v8-0-98ae32bfc49a@xxxxxxxxxxxxxxxx>
prerequisite-patch-id: 467ba5e0fde0e9a1e9d0a6faf3d22ba2126c62a7
prerequisite-patch-id: 9168930e40551e842c8171d5433a6f39ad4b78a4
prerequisite-patch-id: 64fecfbd1e085d7d2ab0ae23295ca34ec8e14c5e
prerequisite-patch-id: 7bf282ed9312b2ffdbe420429bbe355622620272
prerequisite-patch-id: 29a7d652948d13f91bdd4fc5fb8e6dbcd6e5fcc5
prerequisite-patch-id: c03136820a7749fd9630080e6d4459a298b3b630
prerequisite-patch-id: 43975fc425bf2f5dd5c1de66aaa78c4b4bde568c
prerequisite-patch-id: 75caa99e3bbcdf41b6462b9f5f703bea1d4a65fa
prerequisite-patch-id: b9b2a2f80a4827b457166b20f54644e57d90b824
Best regards,
--
George Moussalem <george.moussalem@xxxxxxxxxxx>