Bluetooth: btmtksdio: Fix SKB handling in the TX path
Subsystems (2)
cover Bluetooth: btmtksdio: Fix SKB handling in the TX path
btmtksdio_tx_packet() rounds the SDIO transfer size up to the 256 byte block size, but never grows the SKB accordingly, so the host controller reads up to 255 bytes of uninitialised memory and sends it to the device, and can read past the end of the buffer as well.
Patch 2 fixes that by padding the SKB with zeros. The padding is written behind skb->tail, which is only safe once the driver owns the data buffer, so patch 1 replaces the open-coded headroom check with skb_cow_head() first. Patch 1 on its own changes no observable behaviour, but it is a hard prerequisite, so both patches carry the same
Fixes: tag.
Both patches were previously part of a larger MT7928 series [1]. They are unrelated to MT7928 and to the USB driver, so they are sent separately here. The remaining parts of that series will follow as separate topic branches.
Tested on a Chromebook with MT7921S: Bluetooth power on, then A2DP connect and stream continuously for one hour without failure. The padding added by patch 2 covers every packet whose length is not a multiple of the block size, and the reallocation added by patch 1 covers every HCI command, which hci_send_cmd_sync() always clones into hdev->sent_cmd.
[1] https://lore.kernel.org/linux-bluetooth/[email protected]/
Chris Lu (2):
Bluetooth: btmtksdio: Take exclusive ownership of the SKB before TX
Bluetooth: btmtksdio: Fix out-of-bounds DMA read in the TX path
drivers/bluetooth/btmtksdio.c | 33 ++++++++++++++++++++++++--------- 1 file changed, 24 insertions(+), 9 deletions(-)
Signature · 1 line
2.45.2
PATCH 1/2 Bluetooth: btmtksdio: Take exclusive ownership of the SKB before TX
btmtksdio_tx_packet() prepends the MediaTek SDIO header with skb_push() and writes into that space after only checking the headroom size. On a cloned SKB that headroom belongs to a buffer shared with the other owner, which the driver has no right to write to.
Cloned SKBs do reach this path: hci_send_cmd_sync() keeps a clone of every HCI command in hdev->sent_cmd before handing the SKB to the driver, and l2cap_ertm_send() clones SKBs for retransmission.
Replace the open-coded headroom check with skb_cow_head(), which both guarantees the headroom and reallocates a private buffer when the SKB is cloned. The cost is one reallocation and copy per cloned packet, the usual price of this pattern in network drivers.
This has no observable effect on its own, as the driver only writes in front of skb->data where no other owner looks. It is a prerequisite for "Bluetooth: btmtksdio: Fix out-of-bounds DMA read in the TX path", which writes padding behind skb->tail, and carries the same Fixes: tag so that both are backported together.
--- drivers/bluetooth/btmtksdio.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-)
drivers/bluetooth/btmtksdio.c +6 -7
@@ -274,13 +274,12 @@ static int btmtksdio_tx_packet(struct btmtksdio_dev *bdev,struct mtkbtsdio_hdr *sdio_hdr;int err; -/* Make sure that there are enough rooms for SDIO header */ -if (unlikely (skb_headroom (skb) <sizeof (*sdio_hdr))) { - err =pskb_expand_head (skb,sizeof (*sdio_hdr),0 , -GFP_ATOMIC ); -if (err <0 ) -return err; - } +/* Make sure that the data buffer is not shared with anyone else and +* that there is enough room for the SDIO header +*/ + err =skb_cow_head (skb,sizeof (*sdio_hdr)); +if (err <0 ) +return err;/* Prepend MediaTek SDIO Specific Header */ skb_push (skb,sizeof (*sdio_hdr));
Signature · 1 line
2.45.2
This is automated email and please do not reply to this email!
Dear submitter,
Thank you for submitting the patches to the linux bluetooth mailing list. This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/project/bluetooth/list/?series=1147102
---Test result---
Test Summary:
CheckPatch PASS 1.13 seconds VerifyFixes PASS 0.11 seconds VerifySignedoff PASS 0.11 seconds GitLint FAIL 0.48 seconds SubjectPrefix PASS 0.20 seconds BuildKernel PASS 17.26 seconds CheckAllWarning PASS 18.63 seconds CheckSparse PASS 21.78 seconds BuildKernel32 PASS 16.31 seconds CheckKernelLLVM SKIP 0.00 seconds TestRunnerSetup PASS 298.53 seconds IncrementalBuild PASS 17.27 seconds
Details
##############################
Test: GitLint - FAIL
Desc: Run gitlint
Output:
[2/2] Bluetooth: btmtksdio: Fix out-of-bounds DMA read in the TX path
6: B3 Line contains hard tab characters (\t): " err = sdio_writesb(bdev->func, MTK_REG_CTDR, skb->data,"
7: B3 Line contains hard tab characters (\t): " round_up(skb->len, MTK_SDIO_BLOCK_SIZE));"
##############################
Test: CheckKernelLLVM - SKIP
Desc: Build kernel with LLVM + context analysis
Output:
Clang not found
https://github.com/bluez/bluetooth-next/pull/597
---
Regards,
Linux Bluetooth
PATCH 2/2 Bluetooth: btmtksdio: Fix out-of-bounds DMA read in the TX path
In reply to Chris Lubtmtksdio_tx_packet() rounds the transfer size up to the SDIO block size of 256 bytes, but hands the host controller the SKB buffer as is:
err = sdio_writesb(bdev->func, MTK_REG_CTDR, skb->data, round_up(skb->len, MTK_SDIO_BLOCK_SIZE));
Only skb->len bytes hold packet data, so the controller reads up to 255 bytes of uninitialised memory and sends it to the device over the SDIO bus. Depending on how much tailroom slack the SKB allocation happens to carry, that read can also extend past the end of the buffer.
Compute the padded length up front, ensure the SKB has tailroom for it, and zero-fill the padding with skb_put_zero(). skb->len then covers the padding, so sdio_writesb() no longer needs to round up. byte_tx keeps counting the header and the payload only, and the error path restores the SKB so that the caller can requeue it.
Writing behind skb->tail is only safe because the driver owns the buffer, which "Bluetooth: btmtksdio: Take exclusive ownership of the SKB before TX" ensures.
--- drivers/bluetooth/btmtksdio.c | 26 +++++++++++++++++++++----- 1 file changed, 21 insertions(+), 5 deletions(-)
drivers/bluetooth/btmtksdio.c +21 -5
@@ -272,6 +272,7 @@ static int btmtksdio_tx_packet(struct btmtksdio_dev *bdev, struct sk_buff *skb) {struct mtkbtsdio_hdr *sdio_hdr; +unsigned int len, pad_len;int err;/* Make sure that the data buffer is not shared with anyone else and
@@ -281,6 +282,18 @@ static int btmtksdio_tx_packet(struct btmtksdio_dev *bdev,if (err <0 )return err; +/* The transfer is rounded up to the SDIO block size, so the buffer +* has to provide tailroom for the padding as well +*/ + len = skb->len +sizeof (*sdio_hdr); + pad_len =round_up (len,MTK_SDIO_BLOCK_SIZE ) - len; + +if (unlikely (skb_tailroom (skb) < pad_len)) { + err =pskb_expand_head (skb,0 , pad_len,GFP_ATOMIC ); +if (err <0 ) +return err; + } +/* Prepend MediaTek SDIO Specific Header */ skb_push (skb,sizeof (*sdio_hdr));
@@ -289,19 +302,22 @@ static int btmtksdio_tx_packet(struct btmtksdio_dev *bdev, sdio_hdr->reserved =cpu_to_le16 (0 ); sdio_hdr->bt_type =hci_skb_pkt_type (skb); +/* Zero the padding so that no uninitialised memory is sent out */ +skb_put_zero (skb, pad_len); +clear_bit (BTMTKSDIO_HW_TX_READY , &bdev->tx_state); - err =sdio_writesb (bdev->func,MTK_REG_CTDR , skb->data, -round_up (skb->len,MTK_SDIO_BLOCK_SIZE )); + err =sdio_writesb (bdev->func,MTK_REG_CTDR , skb->data, skb->len);if (err <0 ) - goto err_skb_pull; + goto err_skb_restore; - bdev->hdev->stat.byte_tx += skb->len; + bdev->hdev->stat.byte_tx += len;kfree_skb (skb);return 0 ; -err_skb_pull: +err_skb_restore: +skb_trim (skb, len);skb_pull (skb,sizeof (*sdio_hdr));return err;
Signature · 1 line
2.45.2
This series was applied to bluetooth/bluetooth-next.git (master) by Luiz Augusto von Dentz <[email protected]>:
Hello:
This series was applied to bluetooth/bluetooth-next.git (master) by Luiz Augusto von Dentz <[email protected]>:
Quoted 5 lines · On Mon, 17 Aug 2026 17:53:30 +0800 you wrote:
btmtksdio_tx_packet()rounds the SDIO transfer size up to the 256 byte block size, but never grows the SKB accordingly, so the host controller reads up to 255 bytes of uninitialised memory and sends it to the device, and can read past the end of the buffer as well.
Quoted 8 lines
Patch 2 fixes that by padding the SKB with zeros. The padding is written behind
skb->tail, which is only safe once the driver owns the data buffer, so patch 1 replaces the open-coded headroom check withskb_cow_head()first. Patch 1 on its own changes no observable behaviour, but it is a hard prerequisite, so both patches carry the same
Fixes: tag.[...]
Here is the summary with links:
- [1/2] Bluetooth: btmtksdio: Take exclusive ownership of the SKB before TX
https://git.kernel.org/bluetooth/bluetooth-next/c/951d9f743029 - [2/2] Bluetooth: btmtksdio: Fix out-of-bounds DMA read in the TX path https://git.kernel.org/bluetooth/bluetooth-next/c/262cb784c96c
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html