On Thu, Jul 17, 2025 at 12:11:16AM GMT, Ivan Pravdin wrote: > On Wed, Jul 16, 2025 at 10:11:01AM GMT, Luiz Augusto von Dentz wrote: > > Hi Ivan, > > > > On Tue, Jul 15, 2025 at 8:38 PM Ivan Pravdin > > <ipravdin.official@xxxxxxxxx> wrote: > > > > > > Create and use new vmalloc'ed buffer with dev_coredumpv. From > > > dev_coredumpv documentation: > > > > > > `This function takes ownership of the vmalloc'ed data and will free > > > it when it is no longer used.` > > > > > > As hdev->dump is used after dev_coredumpv, create temporary buffer to > > > hold hdev->dump data. > > > > > > Reported-by: syzbot+ac3c79181f6aecc5120c@xxxxxxxxxxxxxxxxxxxxxxxxx > > > Closes: https://lore.kernel.org/all/67eaa688.050a0220.1547ec.014a.GAE@xxxxxxxxxx > > > > It should probably close the syzkaller issue, which I assume is this one: > > > > https://syzkaller.appspot.com/bug?extid=ac3c79181f6aecc5120c > > I will change it in v3. > > > > > > Fixes: b257e02ecc46 ("HCI: coredump: Log devcd dumps into the monitor") > > > Signed-off-by: Ivan Pravdin <ipravdin.official@xxxxxxxxx> > > > --- > > > v1 -> v2: Changed subject prefix to Bluetooth: > > > > > > net/bluetooth/coredump.c | 7 ++++++- > > > 1 file changed, 6 insertions(+), 1 deletion(-) > > > > > > diff --git a/net/bluetooth/coredump.c b/net/bluetooth/coredump.c > > > index 819eacb38762..1232c9a94f95 100644 > > > --- a/net/bluetooth/coredump.c > > > +++ b/net/bluetooth/coredump.c > > > @@ -243,6 +243,7 @@ static void hci_devcd_handle_pkt_pattern(struct hci_dev *hdev, > > > static void hci_devcd_dump(struct hci_dev *hdev) > > > { > > > struct sk_buff *skb; > > > + char *coredump; > > > u32 size; > > > > > > bt_dev_dbg(hdev, "state %d", hdev->dump.state); > > > @@ -250,7 +251,11 @@ static void hci_devcd_dump(struct hci_dev *hdev) > > > size = hdev->dump.tail - hdev->dump.head; > > > > > > /* Emit a devcoredump with the available data */ > > > - dev_coredumpv(&hdev->dev, hdev->dump.head, size, GFP_KERNEL); > > > + coredump = vmalloc(size); > > > + if (coredump) { > > > + memcpy(coredump, hdev->dump.head, size); > > > + dev_coredumpv(&hdev->dev, coredump, size, GFP_KERNEL); > > > + } > > > > > > > > > /* Send a copy to monitor as a diagnostic packet */ > > > skb = bt_skb_alloc(size, GFP_ATOMIC); > > > -- > > > 2.45.2 > > > > > > > What is wrong here the is code that attempt to send a copy to the > > monitor uses dump.head _after_ dev_coredumpv has freed it, so just > > changing the order shall make it work properly: > > > > diff --git a/net/bluetooth/coredump.c b/net/bluetooth/coredump.c > > index 819eacb38762..720cb79adf96 100644 > > --- a/net/bluetooth/coredump.c > > +++ b/net/bluetooth/coredump.c > > @@ -249,15 +249,15 @@ static void hci_devcd_dump(struct hci_dev *hdev) > > > > size = hdev->dump.tail - hdev->dump.head; > > > > - /* Emit a devcoredump with the available data */ > > - dev_coredumpv(&hdev->dev, hdev->dump.head, size, GFP_KERNEL); > > - > > /* Send a copy to monitor as a diagnostic packet */ > > skb = bt_skb_alloc(size, GFP_ATOMIC); > > if (skb) { > > skb_put_data(skb, hdev->dump.head, size); > > hci_recv_diag(hdev, skb); > > } > > + > > + /* Emit a devcoredump with the available data */ > > + dev_coredumpv(&hdev->dev, hdev->dump.head, size, GFP_KERNEL); > > } > > > > static void hci_devcd_handle_pkt_complete(struct hci_dev *hdev > > Unfortunately this does not work. This was my initial attempt but it > still reproduces the error [2]. The reason for that is that > dev_coredumpv takes the ownership of the buffer. From the comment > above dev_coredumpm_timeout (which is used by dev_coredumpv): > > > Creates a new device coredump for the given device. If a previous one hasn't > > been read yet, the new coredump is discarded. The data lifetime is determined > > by the device coredump framework and when it is no longer needed the @free > > function will be called to free the data. > > If hci_devcd_dump is called multiple times, first time hdev->dump.head > will be kept. However, calling hci_devcd_dump second time before data is > read by userspace will free it, leading to uaf error like the one > observed in the report. > > As hci_devcd_dump need not be called only one time before hdev is freed, > calling it multiple times will lead to uaf, no matter in what order > skb_put_data and dev_coredumpv occur in the function. > > To solve this problem, a temporary buffer should be used. Such buffer > should copy data from hdev->dump.head and be surrendered to > dev_coredumpv without any possibility to be altered or freed. > > For reference, here are some examples of using temporary buffer with > dev_coredumpv in Intel AVS driver: > > [bluetoot-next 6f40d15c70bc] sound/soc/intel/avs/apl.c:183 > [bluetoot-next 6f40d15c70bc] sound/soc/intel/avs/skl.c:147 > > [2] https://lore.kernel.org/all/6845dc6b.050a0220.daf97.0af4.GAE@xxxxxxxxxx/ > Nevermind... I was under impression that hdev->dump is allocated only once and is reused until hdev is freed. The link I have copied actually shows another error that is solved by another patch [1]. I have tested your suggestion on top of unintegrated patch and it did not reproduce any issues [2]. I will include your suggestion in v3. Thanks. [1] https://lore.kernel.org/all/20250716004252.125466-2-ipravdin.official@xxxxxxxxx/ [2] https://lore.kernel.org/all/gkct6i53onzjuzugsor6pnrhjfef7trwhgbjrgfie25xirebhp@dpqxvkguqdnf/ > > > > > > -- > > Luiz Augusto von Dentz > > Thank you for your feedback. > > Ivan Pravdin Ivan Pravdin
