| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CABBYNZJgpjNQ4X7=9g58nLH1GfsaybLegKsUmqwAt70YMiCUmw@mail.gmail.com>
Date: Fri, 6 Jan 2023 12:27:16 -0800
From: Luiz Augusto von Dentz <luiz.dentz@...il.com>
To: hildawu@...ltek.com
Cc: marcel@...tmann.org, johan.hedberg@...il.com,
linux-bluetooth@...r.kernel.org, linux-kernel@...r.kernel.org,
apusaka@...omium.org, yinghsu@...omium.org, max.chou@...ltek.com,
alex_lu@...lsil.com.cn, kidman@...ltek.com
Subject: Re: [PATCH v5] Bluetooth: btusb: Work around spotty SCO quality
Hi Hilda,
On Sun, Dec 25, 2022 at 11:48 PM <hildawu@...ltek.com> wrote:
>
> From: Hilda Wu <hildawu@...ltek.com>
>
> When streaming HFP, every few minutes a brief pause in audio can be
> heard on some platforms with Realtek Bluetooth. Linux based products
> may be encountered, but because of the different implementation of
> upper-level SCO services, the situation would not necessarily impact
> customer experience. But when the issue occurs, the system will see
> the SCO packet for unknown connection handle messages.
Im not really sure what you are talking about here, the ISO packet
fragmentation over HCI is part of the BT standard so if you are not
following it then your controller is not compliant.
> Note: This issue affects (e)SCO only, does not affect ACLs.
> The duplicate data affected the invalid connection handle only
> occurs in Realtek BT.
> This is to filter out the duplicate packet for avoiding influence.
>
> The btmon trace gives a better idea of what we're filtering.
> The following excerpts are part of SCO packets in the HCI log:
>
> > SCO Data RX: Handle 11 flags 0x00 dlen 72 #23327 [hci0] 132.343418
> 8c a3 55 4f 8a d5 56 e9 35 56 37 8d 55 87 53 55 ..UO..V.5V7.U.SU
> 59 66 d5 57 1d b5 54 00 01 08 ad 00 00 e0 10 00 Yf.W..T.........
> 00 00 85 c6 d5 60 e9 b5 52 94 6d 54 e4 9b 55 b1 .....`..R.mT..U.
> b6 d5 62 91 b5 57 84 6d 56 e4 5b 55 75 c6 d5 51 ..b..W.mV.[Uu..Q
> 2d b5 53 9a 6d 54 a5 1b -.S.mT..
> < SCO Data TX: Handle 11 flags 0x00 dlen 72 #23328 [hci0] 132.343600
> 01 c8 ad 00 00 aa db ba aa a9 72 b4 d9 5d af 14 ..........r..]..
> 53 0c 75 b0 a6 f3 8a 51 b3 54 17 b1 a6 d5 62 c5 S.u....Q.T....b.
> d5 6b 35 29 8d c5 1c 56 4c 24 96 9b 8d b5 d7 1a .k5)...VL$......
> b2 8d bc da 3b 8c 46 ae 1d 4d a4 04 01 f8 ad 00 ....;.F..M......
> 00 3d ec bb a9 98 8b 28 .=.....(
> > SCO Data RX: Handle 11 flags 0x00 dlen 72 #23329 [hci0] 132.353419
> 55 55 c6 d5 62 29 b5 57 b2 6d 54 00 01 38 ad 00 UU..b).W.mT..8..
> 00 e0 10 00 00 00 0b 00 d5 62 55 c6 57 b2 29 b5 .........bU.W.).
> 00 01 6d 54 00 00 38 ad 00 00 e0 10 00 00 00 92 ..mT..8.........
> 36 d5 5a ed b5 58 6c 6d 55 b3 1b 55 6b 26 d5 52 6.Z..XlmU..Uk&.R
> d1 b5 54 23 6d 56 82 db ..T#mV..
> < SCO Data TX: Handle 11 flags 0x00 dlen 72 #23330 [hci0] 132.353581
> 6d 5b be db 89 34 66 e9 fa 99 a6 6e e5 6d 9f 1a m[...4f....n.m..
> 1c 57 d2 66 92 63 98 99 a9 3b 8a 6c 3e 5b 5a 34 .W.f.c...;.l>[Z4
> a4 96 e2 21 21 8c f8 88 0f 3d e0 52 48 85 18 00 ...!!....=.RH...
> 01 08 ad 00 00 0c eb ba a9 a8 28 ca 9a d0 3c 33 ..........(...<3
> 45 4a f9 90 fb ca 4b 39 EJ....K9
> > SCO Data RX: Handle 2901 flags 0x0a dlen 54 #23331 [hci0] 132.373416
> d5 48 a9 b5 56 aa 6d 56 d2 db 55 75 36 d5 56 2d .H..V.mV..Uu6.V-
> b5 57 5b 6d 54 00 0b 00 48 01 c8 ad 00 00 e0 10 .W[mT...H.......
> 00 00 00 5e c6 d5 56 e1 b5 56 43 6d 55 ca db 55 ...^..V..VCmU..U
> 7d c6 d5 5b 31 b5
>
> We handle is HCI SCO Data RX packets.
> The packet 23327 was a normal HCI SCO Data RX packet.
> The packet 23329 was the abnormal HCI SCO Data RX packet.
> The packet 23331 was the invalid connection handle affected by the
> packet 23329 abnormal HCI SCO Data RX packet.
>
> So we expect to filter the packet 23329 SCO data RX packet case.
> As you can see the packet 23329, the packet's connection handle
> (0x0B 00/11) and length (0x48/72) is normal.
> This btmon trace is SCO packets in USB alternate setting 3, payload
> length is 72 bytes that consist of three SCO data packets.
> After our investigation, we found that the anomaly is due to the
> intermediate composition data.
> There is duplicate data in the intermediate composition data, but it
> affects packet combination. The system parses the next packet of the
> connection handle mistake, so the system shows unknown connection
> handle messages.
>
> This commit can estimate and find out its abnormal rule to filter the
> duplicate packet out for avoiding influence.
> Check fragments and filter out the abnormal packet, and then it will
> not affect the system parsing of the connection handle subsequent.
> This commit can filter out the invalid connection handle and avoid the
> spotty SCO quality.
>
> Signed-off-by: Alex Lu <alex_lu@...lsil.com.cn>
> Signed-off-by: Hilda Wu <hildawu@...ltek.com>
> ---
> Changes in v5:
> - Fix typos in the commit log.
>
> Changes in v4:
> - Fix typos and add more in commit description.
> - Modify comparison method.
>
> Changes in v3:
> - Use the vendor function to replace btus_recv_isoc
> - Additional info: The comparison of btrtl_usb_recv_isoc here is
> for invalid handle, the invalid handle shouldn't appear.
> So we try to find out the rule and filter out this.
>
> Changes in v2:
> - Separate commits for functions
> ---
> ---
> drivers/bluetooth/btrtl.c | 33 ++++++++++++++++
> drivers/bluetooth/btrtl.h | 7 ++++
> drivers/bluetooth/btusb.c | 80 ++++++++++++++++++++++++++++++++++++++-
> 3 files changed, 119 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/bluetooth/btrtl.c b/drivers/bluetooth/btrtl.c
> index 69c3fe649ca7..a0a33d2467ef 100644
> --- a/drivers/bluetooth/btrtl.c
> +++ b/drivers/bluetooth/btrtl.c
> @@ -944,6 +944,39 @@ int btrtl_get_uart_settings(struct hci_dev *hdev,
> }
> EXPORT_SYMBOL_GPL(btrtl_get_uart_settings);
>
> +int btrtl_usb_recv_isoc(u16 pos, u8 *data, u8 *p, u16 wMaxPacketSize)
> +{
> + u8 *prev;
> + u8 tmp[8];
> + u8 zero[1] = {0};
> + u32 *prev_frag_a;
> + u32 *prev_frag_b;
> +
> + /* Issue was found when alt was set to 3 or bigger */
> + if (wMaxPacketSize < 25)
> + return 0;
> +
> + /* The first fragment is always correct. */
> + if (pos < wMaxPacketSize)
> + return 0;
> +
> + if (memcmp(p, zero, sizeof(zero))) {
> + prev = data + (pos - wMaxPacketSize);
> + memcpy(tmp, prev + 4, 8);
> + prev_frag_a = (u32 *)(tmp);
> + prev_frag_b = (u32 *)(tmp + 4);
> + *prev_frag_a = swahw32(*prev_frag_a);
> + *prev_frag_b = swahw32(*prev_frag_b);
What is this information? You seem to be looking into the part of data
that is supposed to be codec specific in which case you will need to
know what codec has been negotiated, in that this would be a layer
violation and we cannot really do this here, instead it would probably
need to be handled in userspace component handling the SCO socket
data.
> +
> + /* Check the current fragment with the previous one. */
> + if (!memcmp(p, prev, 2) && !memcmp(p + 2, tmp, 8))
> + return -1;
> + }
Besides the problems above I'm still not convinced this won't cause
false positives, meaning you would detect a frame as duplicated when
in fact it isn't, and btw if you know about this problem why aren't
you fixing it in the firmware?
> +
> + return 0;
> +}
> +EXPORT_SYMBOL_GPL(btrtl_usb_recv_isoc);
> +
> MODULE_AUTHOR("Daniel Drake <drake@...lessm.com>");
> MODULE_DESCRIPTION("Bluetooth support for Realtek devices ver " VERSION);
> MODULE_VERSION(VERSION);
> diff --git a/drivers/bluetooth/btrtl.h b/drivers/bluetooth/btrtl.h
> index ebf0101c959b..893f1eae57e2 100644
> --- a/drivers/bluetooth/btrtl.h
> +++ b/drivers/bluetooth/btrtl.h
> @@ -83,6 +83,7 @@ int btrtl_get_uart_settings(struct hci_dev *hdev,
> struct btrtl_device_info *btrtl_dev,
> unsigned int *controller_baudrate,
> u32 *device_baudrate, bool *flow_control);
> +int btrtl_usb_recv_isoc(u16 pos, u8 *data, u8 *buffer, u16 wMaxPacketSize);
>
> #else
>
> @@ -126,4 +127,10 @@ static inline int btrtl_get_uart_settings(struct hci_dev *hdev,
> return -ENOENT;
> }
>
> +static inline int btrtl_usb_recv_isoc(u16 pos, u8 *data, u8 *buffer,
> + u16 wMaxPacketSize)
> +{
> + return -EOPNOTSUPP;
> +}
> +
> #endif
> diff --git a/drivers/bluetooth/btusb.c b/drivers/bluetooth/btusb.c
> index afd2f08ffe30..47e4dcaec948 100644
> --- a/drivers/bluetooth/btusb.c
> +++ b/drivers/bluetooth/btusb.c
> @@ -792,6 +792,7 @@ struct btusb_data {
> int (*recv_event)(struct hci_dev *hdev, struct sk_buff *skb);
> int (*recv_acl)(struct hci_dev *hdev, struct sk_buff *skb);
> int (*recv_bulk)(struct btusb_data *data, void *buffer, int count);
> + int (*recv_isoc)(struct btusb_data *data, void *buffer, int count);
>
> int (*setup_on_usb)(struct hci_dev *hdev);
>
> @@ -1394,7 +1395,7 @@ static void btusb_isoc_complete(struct urb *urb)
>
> hdev->stat.byte_rx += length;
>
> - if (btusb_recv_isoc(data, urb->transfer_buffer + offset,
> + if (data->recv_isoc(data, urb->transfer_buffer + offset,
> length) < 0) {
> bt_dev_err(hdev, "corrupted SCO packet");
> hdev->stat.err_rx++;
> @@ -2492,6 +2493,80 @@ static int btusb_setup_realtek(struct hci_dev *hdev)
> return ret;
> }
>
> +static int btusb_recv_isoc_realtek(struct btusb_data *data, void *buffer,
> + int count)
> +{
> + struct sk_buff *skb;
> + unsigned long flags;
> + int err = 0;
> + u16 wMaxPacketSize = le16_to_cpu(data->isoc_rx_ep->wMaxPacketSize);
> +
> + spin_lock_irqsave(&data->rxlock, flags);
> + skb = data->sco_skb;
> +
> + while (count) {
> + int len;
> +
> + if (!skb) {
> + skb = bt_skb_alloc(HCI_MAX_SCO_SIZE, GFP_ATOMIC);
> + if (!skb) {
> + err = -ENOMEM;
> + break;
> + }
> +
> + hci_skb_pkt_type(skb) = HCI_SCODATA_PKT;
> + hci_skb_expect(skb) = HCI_SCO_HDR_SIZE;
> + }
> +
> + len = min_t(uint, hci_skb_expect(skb), count);
> +
> + /* Gaps in audio could be heard while streaming WBS using USB
> + * alt settings 3 on some platforms, since this is only used
> + * with RTK chips so let vendor function detect it.
> + */
> + if (!btusb_find_altsetting(data, 6) &&
> + test_bit(BTUSB_USE_ALT3_FOR_WBS, &data->flags)) {
> + err = btrtl_usb_recv_isoc(skb->len, skb->data, buffer,
> + wMaxPacketSize);
> + if (err)
> + break;
> + }
> +
> + skb_put_data(skb, buffer, len);
> +
> + count -= len;
> + buffer += len;
> + hci_skb_expect(skb) -= len;
> +
> + if (skb->len == HCI_SCO_HDR_SIZE) {
> + /* Complete SCO header */
> + struct hci_sco_hdr *hdr = hci_sco_hdr(skb);
> +
> + hci_skb_expect(skb) = hdr->dlen;
> +
> + if (skb_tailroom(skb) < hci_skb_expect(skb) ||
> + !btusb_validate_sco_handle(data->hdev, hdr)) {
> + kfree_skb(skb);
> + skb = NULL;
> +
> + err = -EILSEQ;
> + break;
> + }
> + }
> +
> + if (!hci_skb_expect(skb)) {
> + /* Complete frame */
> + hci_recv_frame(data->hdev, skb);
> + skb = NULL;
> + }
> + }
> +
> + data->sco_skb = skb;
> + spin_unlock_irqrestore(&data->rxlock, flags);
> +
> + return err;
> +}
> +
> /* UHW CR mapping */
> #define MTK_BT_MISC 0x70002510
> #define MTK_BT_SUBSYS_RST 0x70002610
> @@ -3924,6 +3999,7 @@ static int btusb_probe(struct usb_interface *intf,
>
> data->recv_event = hci_recv_frame;
> data->recv_bulk = btusb_recv_bulk;
> + data->recv_isoc = btusb_recv_isoc;
>
> if (id->driver_info & BTUSB_INTEL_COMBINED) {
> /* Allocate extra space for Intel device */
> @@ -4097,6 +4173,8 @@ static int btusb_probe(struct usb_interface *intf,
> hdev->shutdown = btrtl_shutdown_realtek;
> hdev->cmd_timeout = btusb_rtl_cmd_timeout;
>
> + data->recv_isoc = btusb_recv_isoc_realtek;
> +
> /* Realtek devices need to set remote wakeup on auto-suspend */
> set_bit(BTUSB_WAKEUP_AUTOSUSPEND, &data->flags);
> set_bit(BTUSB_USE_ALT3_FOR_WBS, &data->flags);
> --
> 2.17.1
>
--
Luiz Augusto von Dentz
Powered by blists - more mailing lists