Skip to contentSkip to conversation

Bluetooth: btnxpuart: Keep FW dump header in coredump chunks

PATCH In review Ali Ahmet Memis 4 participants 9 messages Aug 18 – Sep 10 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): AS4PR04MB969206113CDC2F5C66994E6EE7A62@AS4PR04MB9692.eurprd04.prod.outlook.com

#1 lore raw

PATCH 1/1 Bluetooth: btnxpuart: Keep FW dump header in coredump chunks

Commit 1fcf216462ec ("Bluetooth: btnxpuart: Validate the FW dump header length") changed nxp_process_fw_dump() to use skb_pull_data() when validating the FW dump header.

This also removes the header from skb->data. Since the skb is cloned for hci_devcd_append() afterwards, the FW dump header is missing from the coredump chunk.

The NXP FW dump analyzer expects nxp_fw_dump_hdr at the beginning of each chunk, so this results in invalid coredumps.

Check the header length without pulling it from the skb instead. This keeps skb->data pointing to the FW dump header when the skb is cloned.

---
Sorry for the late follow-up. I missed your reply on the v2 thread and only noticed it after the change had landed. This patch restores the approach from v2, as you suggested.

 drivers/bluetooth/btnxpuart.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/bluetooth/btnxpuart.c +3 -3
index f2bbe6e462aa..16d0f709d4b7 100644
@@ -1359,14 +1359,14 @@ static int nxp_process_fw_dump(struct hci_dev *hdev, struct sk_buff *skb)
 {
 	struct hci_acl_hdr *acl_hdr = (struct hci_acl_hdr *)skb_pull_data(skb,
 									  sizeof(*acl_hdr));
-	struct nxp_fw_dump_hdr *fw_dump_hdr;
+	struct nxp_fw_dump_hdr *fw_dump_hdr = (struct nxp_fw_dump_hdr *)skb->data;
 	struct btnxpuart_dev *nxpdev = hci_get_drvdata(hdev);
 	__u16 seq_num;
 	__u16 buf_len;
 	int err;
 
-	fw_dump_hdr = skb_pull_data(skb, sizeof(*fw_dump_hdr));
-	if (!fw_dump_hdr) {
+	/* The ACL payload must be long enough to hold the FW dump header */
+	if (skb->len < sizeof(*fw_dump_hdr)) {
 		bt_dev_warn(hdev, "FW dump: invalid or corrupt fw dump chunk");
 		goto free_skb;
 	}
Signature · 1 line
2.55.0
#2 lore raw

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=1147625

---Test result---

Test Summary:

CheckPatch                    PASS      0.80 seconds
VerifyFixes                   PASS      0.14 seconds
VerifySignedoff               PASS      0.59 seconds
GitLint                       PASS      0.35 seconds
SubjectPrefix                 PASS      0.13 seconds
BuildKernel                   PASS      28.64 seconds
CheckAllWarning               PASS      31.09 seconds
CheckSparse                   PASS      29.56 seconds
BuildKernel32                 PASS      27.05 seconds
CheckKernelLLVM               SKIP      0.00 seconds
TestRunnerSetup               PASS      511.83 seconds
IncrementalBuild              PASS      25.96 seconds

Details
##############################
Test: CheckKernelLLVM - SKIP
Desc: Build kernel with LLVM + context analysis
Output:
Clang not found

https://github.com/bluez/bluetooth-next/pull/606

---
Regards,
Linux Bluetooth

maintainer
#3 lore raw

PATCH 1/1 Bluetooth: btnxpuart: Keep FW dump header in coredump chunks

In reply to Ali Ahmet Memis

Hi Ali,

Thank you for the quick fix.

Thanks,
Neeraj

Quoted 62 lines

Commit 1fcf216462ec ("Bluetooth: btnxpuart: Validate the FW dump header length") changed nxp_process_fw_dump() to use skb_pull_data() when validating the FW dump header.

This also removes the header from skb->data. Since the skb is cloned for hci_devcd_append() afterwards, the FW dump header is missing from the coredump chunk.

The NXP FW dump analyzer expects nxp_fw_dump_hdr at the beginning of each chunk, so this results in invalid coredumps.

Check the header length without pulling it from the skb instead. This keeps skb->data pointing to the FW dump header when the skb is cloned.

Fixes 1fcf216462ec ("Bluetooth: btnxpuart: Validate the FW dump header

length")

Link:
https://lore.ke/
rnel.org%2Flinux- bluetooth%2FAS4PR04MB9692EC13E3176B6D7525D097E7A72%40AS4PR04M B9692.eurprd04.prod.outlook.com%2F&data=05%7C02%7Cneeraj.sanjaykale %40nxp.com%7Ce5c09f2752bc463e224608defcfee41a%7C686ea1d3bc2b4c6f a92cd99c5c301635%7C0%7C0%7C639226368879983123%7CUnknown%7CTW FpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW 4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=mEFZvI Xr7Buf8x4p9NDDl3B1mJJR93pBKXZuARuRmX0%3D&reserved=0

---
Sorry for the late follow-up. I missed your reply on the v2 thread and only noticed it after the change had landed. This patch restores the approach from v2, as you suggested.

 drivers/bluetooth/btnxpuart.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/bluetooth/btnxpuart.c +4 -4
index f2bbe6e462aa..16d0f709d4b7 100644
@@ -1359,14 +1359,14 @@ static int nxp_process_fw_dump(struct hci_dev
 *hdev, struct sk_buff *skb)  {
        struct hci_acl_hdr *acl_hdr = (struct hci_acl_hdr *)skb_pull_data(skb,
                                                                          sizeof(*acl_hdr));
-       struct nxp_fw_dump_hdr *fw_dump_hdr;
+       struct nxp_fw_dump_hdr *fw_dump_hdr = (struct nxp_fw_dump_hdr
+ *)skb->data;
        struct btnxpuart_dev *nxpdev = hci_get_drvdata(hdev);
        __u16 seq_num;
        __u16 buf_len;
        int err;
 
-       fw_dump_hdr = skb_pull_data(skb, sizeof(*fw_dump_hdr));
-       if (!fw_dump_hdr) {
+       /* The ACL payload must be long enough to hold the FW dump header
 */
+       if (skb->len < sizeof(*fw_dump_hdr)) {
                bt_dev_warn(hdev, "FW dump: invalid or corrupt fw dump chunk");
                goto free_skb;
        }
--
Quoted 1 line

2.55.0

NXP Confidential

#4 lore raw
In reply to Ali Ahmet Memis

Hi Ali, Neeraj,

Quoted 9 lines · Ali Ahmet Memis

Commit 1fcf216462ec ("Bluetooth: btnxpuart: Validate the FW dump header length") changed nxp_process_fw_dump() to use skb_pull_data() when validating the FW dump header.

This also removes the header from skb->data. Since the skb is cloned for hci_devcd_append() afterwards, the FW dump header is missing from the coredump chunk.

Quoted 5 lines

The NXP FW dump analyzer expects nxp_fw_dump_hdr at the beginning of each chunk, so this results in invalid coredumps.

Check the header length without pulling it from the skb instead. This keeps skb->data pointing to the FW dump header when the skb is cloned.

Ok, but if that is the case why are we parsing the headers in the kernel? I thought the idea was that the kernel would assemble all the segments and then push the dump as a whole. However, the above suggests the NXP FW analyzer expects the headers. In that case I would just have each segment reported on its own rather than appending it to a separate skb including the headers.

Quoted 22 lines

---
Sorry for the late follow-up. I missed your reply on the v2 thread and only noticed it after the change had landed. This patch restores the approach from v2, as you suggested.

 drivers/bluetooth/btnxpuart.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/bluetooth/btnxpuart.c +1 -1
index f2bbe6e462aa..16d0f709d4b7 100644
@@ -1359,14 +1359,14 @@ static int nxp_process_fw_dump(struct hci_dev *hdev, struct sk_buff *skb)
 {
        struct hci_acl_hdr *acl_hdr = (struct hci_acl_hdr *)skb_pull_data(skb,
                                                                          sizeof(*acl_hdr));
-       struct nxp_fw_dump_hdr *fw_dump_hdr;
+       struct nxp_fw_dump_hdr *fw_dump_hdr = (struct nxp_fw_dump_hdr *)skb->data;

drivers/bluetooth/btnxpuart.c static int nxp_process_fw_dump(struct hci_dev *hdev, struct sk_buff *skb) · line 1370

Quoted 10 lines
drivers/bluetooth/btnxpuart.c +2 -2
        struct btnxpuart_dev *nxpdev = hci_get_drvdata(hdev);
        __u16 seq_num;
        __u16 buf_len;
        int err;
 
-       fw_dump_hdr = skb_pull_data(skb, sizeof(*fw_dump_hdr));
-       if (!fw_dump_hdr) {
+       /* The ACL payload must be long enough to hold the FW dump header */
+       if (skb->len < sizeof(*fw_dump_hdr)) {
                bt_dev_warn(hdev, "FW dump: invalid or corrupt fw dump chunk");

Id argue that this should be appended as is then, and the analyzer should be the one checking it, _or_ it needs changing and then it only process the dump _after_ reassemble.

Quoted 5 lines
drivers/bluetooth/btnxpuart.c +0 -1
                goto free_skb;
        }
--

2.55.0

Signature · 1 line
Luiz Augusto von Dentz
#5 lore raw

Hi Luiz,

Quoted 7 lines · Luiz Augusto von Dentz

Ok, but if that is the case why are we parsing the headers in the kernel? I thought the idea was that the kernel would assemble all the segments and then push the dump as a whole. However, the above suggests the NXP FW analyzer expects the headers. In that case I would just have each segment reported on its own rather than appending it to a separate skb including the headers.

The two fields are not just optional parsing for a concatenated dump. They are what the driver uses to handle the dump state.

seq_num == 0x0001 tells us that a new dump has started, so hci_devcd_init() is called. buf_len == 0 marks the end of the dump, so hci_devcd_complete() is called.

The chunks are passed to hci_devcd_append() as they arrive. nxp_process_fw_dump() does not check whether the dump is currently active before calling it. That state is handled by the coredump code itself.

Quoted 3 lines

Id argue that this should be appended as is then, and the analyzer should be the one checking it, _or_ it needs changing and then it only process the dump _after_ reassemble.

The driver also does not otherwise interpret the chunk contents. It does not validate the payload length against buf_len, slice the payload, or remove anything from it. Once the chunk is long enough to contain the header, the whole skb is passed to hci_devcd_append() as-is, including the header.

So the length check is only there to make reading seq_num and buf_len safe. It does not change what gets appended.

The only packets we drop are those that are too short to contain the header. Without the header, there is no seq_num or buf_len, so we cannot determine whether the packet starts or ends a dump.

#6 lore raw

Hi Ali,

Quoted 16 lines · Ali Ahmet Memis

Hi Luiz,

Quoted 7 lines · Luiz Augusto von Dentz

Ok, but if that is the case why are we parsing the headers in the kernel? I thought the idea was that the kernel would assemble all the segments and then push the dump as a whole. However, the above suggests the NXP FW analyzer expects the headers. In that case I would just have each segment reported on its own rather than appending it to a separate skb including the headers.

The two fields are not just optional parsing for a concatenated dump. They are what the driver uses to handle the dump state.

Quoted 8 lines

seq_num == 0x0001 tells us that a new dump has started, so hci_devcd_init() is called. buf_len == 0 marks the end of the dump, so hci_devcd_complete() is called.

The chunks are passed to hci_devcd_append() as they arrive. nxp_process_fw_dump() does not check whether the dump is currently active before calling it. That state is handled by the coredump code itself.

Yeah, and it concatenates the header states into a dump skb, interleaving data with headers. The worst part is that if the analyzer finds an issue with the header it can consider the entire trace malformed. This means both the kernel and userspace try to parse the same thing but may interpret what each chunk means differently.

Quoted 4 lines
Quoted 3 lines

Id argue that this should be appended as is then, and the analyzer should be the one checking it, _or_ it needs changing and then it only process the dump _after_ reassemble.

Quoted 8 lines

The driver also does not otherwise interpret the chunk contents. It does not validate the payload length against buf_len, slice the payload, or remove anything from it. Once the chunk is long enough to contain the header, the whole skb is passed to hci_devcd_append() as-is, including the header.

So the length check is only there to make reading seq_num and buf_len safe. It does not change what gets appended.

Well, it does change, it doesn't append anything if the chunk is less than a header and that won't even show up for the analyzer to analyze. Now, regarding this, I wonder if the driver receives a chunk smaller than the header size, whether we should either append it and leave interpretation to userspace, or 2 drop the entire dump because we can't guarantee its consistency if we suspect data corruption when the firmware is not supposed to generate chunks smaller than the header size.

Quoted 3 lines

The only packets we drop are those that are too short to contain the header. Without the header, there is no seq_num or buf_len, so we cannot determine whether the packet starts or ends a dump.

Yeah, the decision of considering a chunk valid or not seems to be split between kernel and userspace. The kernel evaluates if the chunk is too short but won't check any data consistency after that; userspace will evaluate the headers but won't get a chance to evaluate if any small chunk was dropped in the process.

Signature · 1 line
Luiz Augusto von Dentz
#7 lore raw

Hi Luiz,

Quoted 6 lines · Luiz Augusto von Dentz

Yeah, and it concatenates the header states into a dump skb, interleaving data with headers. The worst part is that if the analyzer finds an issue with the header it can consider the entire trace malformed. This means both the kernel and userspace try to parse the same thing but may interpret what each chunk means differently.

The kernel uses these two fields only for framing. It does not interpret them as part of the dump format or validate their contents. It also does not compare buf_len with the actual payload length.

This cannot be moved to userspace because the driver has no other way to know where a dump starts and ends. Without seq_num and buf_len, it cannot know when to call hci_devcd_init() and hci_devcd_complete().

I agree there is some overlap here, and I think this is also why a short chunk should not be dropped silently.

Quoted 2 lines

Well, it does change, it doesn't append anything if the chunk is less than a header and that won't even show up for the analyzer to analyze.

You're right. That statement was too broad. A chunk shorter than the header is dropped, so the analyzer never sees it.

Quoted 6 lines

Now, regarding this, I wonder if the driver receives a chunk smaller than the header size, whether we should either append it and leave interpretation to userspace, or 2 drop the entire dump because we can't guarantee its consistency if we suspect data corruption when the firmware is not supposed to generate chunks smaller than the header size.

I would go with the second option.

Appending it does not really leave the interpretation to userspace. Each header tells the analyzer how many bytes belong to that chunk. If we add a fragment without a header, all following headers are shifted and the rest of the dump becomes difficult to parse.

Aborting seems like a better fit. The firmware should not generate a chunk smaller than the header, so seeing one means something has already gone wrong. We can no longer rely on the rest of the dump being consistent.

hci_devcd_abort() keeps the data collected so far and marks the dump with HCI_DEVCOREDUMP_ABORT in the State line. Userspace can then see that the dump is truncated. hci_qca and hci_vhci already use it on their error paths.

Quoted 5 lines

Yeah, the decision of considering a chunk valid or not seems to be split between kernel and userspace. The kernel evaluates if the chunk is too short but won't check any data consistency after that; userspace will evaluate the headers but won't get a chance to evaluate if any small chunk was dropped in the process.

I think aborting closes that gap. The kernel decides that a chunk too short to contain a header is invalid, and userspace can see that the dump was aborted instead of silently missing bytes.

Would you prefer this as a separate patch on top? This patch only restores the header in the appended chunk, so its Fixes tag can stay focused on that regression. Handling short chunks with an abort would be a separate behavior change.

Thanks,
Ali

#8 lore raw

Hi Ali,

Quoted 59 lines · Ali Ahmet Memis

Hi Luiz,

Quoted 6 lines · Luiz Augusto von Dentz

Yeah, and it concatenates the header states into a dump skb, interleaving data with headers. The worst part is that if the analyzer finds an issue with the header it can consider the entire trace malformed. This means both the kernel and userspace try to parse the same thing but may interpret what each chunk means differently.

The kernel uses these two fields only for framing. It does not interpret them as part of the dump format or validate their contents. It also does not compare buf_len with the actual payload length.

This cannot be moved to userspace because the driver has no other way to know where a dump starts and ends. Without seq_num and buf_len, it cannot know when to call hci_devcd_init() and hci_devcd_complete().

I agree there is some overlap here, and I think this is also why a short chunk should not be dropped silently.

Quoted 2 lines

Well, it does change, it doesn't append anything if the chunk is less than a header and that won't even show up for the analyzer to analyze.

You're right. That statement was too broad. A chunk shorter than the header is dropped, so the analyzer never sees it.

Quoted 6 lines

Now, regarding this, I wonder if the driver receives a chunk smaller than the header size, whether we should either append it and leave interpretation to userspace, or 2 drop the entire dump because we can't guarantee its consistency if we suspect data corruption when the firmware is not supposed to generate chunks smaller than the header size.

I would go with the second option.

Appending it does not really leave the interpretation to userspace. Each header tells the analyzer how many bytes belong to that chunk. If we add a fragment without a header, all following headers are shifted and the rest of the dump becomes difficult to parse.

Aborting seems like a better fit. The firmware should not generate a chunk smaller than the header, so seeing one means something has already gone wrong. We can no longer rely on the rest of the dump being consistent.

hci_devcd_abort() keeps the data collected so far and marks the dump with HCI_DEVCOREDUMP_ABORT in the State line. Userspace can then see that the dump is truncated. hci_qca and hci_vhci already use it on their error paths.

Quoted 5 lines

Yeah, the decision of considering a chunk valid or not seems to be split between kernel and userspace. The kernel evaluates if the chunk is too short but won't check any data consistency after that; userspace will evaluate the headers but won't get a chance to evaluate if any small chunk was dropped in the process.

Quoted 8 lines

I think aborting closes that gap. The kernel decides that a chunk too short to contain a header is invalid, and userspace can see that the dump was aborted instead of silently missing bytes.

Would you prefer this as a separate patch on top? This patch only restores the header in the appended chunk, so its Fixes tag can stay focused on that regression. Handling short chunks with an abort would be a separate behavior change.

Lets spin a new version aborting if the chunk is too short, I will probably have it fixed up in place with the origina change since that wasn't pulled yet so it makes it easier to not have multiple Fixes tag applying one on top of another.

Quoted 2 lines

Thanks,
Ali

Signature · 1 line
Luiz Augusto von Dentz
Reviewed-by Neeraj Kale on v1 1/1
#9 lore raw
In reply to Ali Ahmet Memis

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/series/1147625/

---Test result---

Test Summary:

CheckPatch                    FAIL      0.85 seconds
VerifyFixes                   PASS      4.71 seconds
VerifySignedoff               PASS      0.13 seconds
GitLint                       PASS      0.29 seconds
SubjectPrefix                 PASS      0.10 seconds
BuildKernel                   PASS      26.84 seconds
CheckAllWarning               PASS      41.92 seconds
CheckSparse                   PASS      32.51 seconds
BuildKernel32                 PASS      27.20 seconds
CheckKernelLLVM               PASS      30.26 seconds
TestRunnerSetup               PASS      674.60 seconds
IncrementalBuild              PASS      26.14 seconds

Details
##############################
Test: CheckPatch - FAIL
Desc: Run checkpatch.pl script
Output:
Bluetooth: btnxpuart: Keep FW dump header in coredump chunks
ERROR: Unrecognized email address: 'Neeraj Kale [email protected]' #100:

total: 1 errors, 0 warnings, 17 lines checked

NOTE: For some of the reported defects, checkpatch may be able to
mechanically convert to the typical style using --fix or --fix-inplace.

/github/workspace/src/patch/14754788.patch has style problems, please review.

NOTE: Ignored message types: UNKNOWN_COMMIT_ID

NOTE: If any of the errors are false positives, please report
them to the maintainer, see CHECKPATCH in MAINTAINERS.

https://github.com/bluez/bluetooth-next/pull/732

---
Regards,
Linux Bluetooth