[PATCH v2] wifi: ath9k: fix out-of-bounds endpoint index in htc_issue_send()

[PATCH v2] wifi: ath9k: fix out-of-bounds endpoint index in htc_issue_send()

[Dmitry Antipov]

Syzbot has reported the following UBSAN splat:

UBSAN: array-index-out-of-bounds in drivers/net/wireless/ath/ath9k/htc_hst.c:26:35
index 255 is out of range for type 'struct htc_endpoint[22]'
...
Call Trace:
 ...
 __ubsan_handle_out_of_bounds+0x115/0x140
 htc_send_epid+0x1e0/0x2d0
 ath9k_wmi_cmd+0x31f/0x590
 ath9k_regread+0xdc/0x140
 ? __pfx_ath9k_regread+0x10/0x10
 ? _raw_spin_unlock_irqrestore+0xae/0x110
 ? __pfx__raw_spin_unlock_irqrestore+0x10/0x10
 ? __pfx_ath9k_regread+0x10/0x10
 ath9k_hw_init+0x4b0/0x2230
 ? init_timer_key+0x171/0x2d0
 ath9k_htc_probe_device+0xa66/0x1fe0
 ? __pfx_ath9k_htc_probe_device+0x10/0x10
 ? usb_hcd_submit_urb+0x345/0x1ae0
 ? usb_free_urb+0x9d/0x120
 ? ath9k_hif_usb_alloc_urbs+0x1077/0x1150
 ath9k_htc_hw_init+0x32/0x70
 ath9k_hif_usb_firmware_cb+0x240/0x4c0
 ? __pfx_ath9k_hif_usb_firmware_cb+0x10/0x10
 request_firmware_work_func+0x106/0x1c0
 ...

IIUC https://syzkaller.appspot.com/text?tag=ReproC&x=1734b9f1180000 makes
an attempt to emulate ath9k USB device in somewhat weird and obscure way
so I'm not sure whether this may happen on a real hardware. OTOH whether
ending with ENDPOINT_UNUSED in 'htc_connect_service()' may indicate the
malfunctioned real device, it may be better to reject such a device with
-ENODEV.

Reported-by: syzbot+93cbd5fbb85814306ba1@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=93cbd5fbb85814306ba1
Fixes: fb9987d0f748 ("ath9k_htc: Support for AR9271 chipset.")
Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>
---
v2: avoid conditional expression (Kalle) and return
    -ENODEV without touching *conn_rsp_epid (Toke)
---
 drivers/net/wireless/ath/ath9k/htc_hst.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/net/wireless/ath/ath9k/htc_hst.c b/drivers/net/wireless/ath/ath9k/htc_hst.c
index eb631fd3336d..a5bc0a5dfe7c 100644
--- a/drivers/net/wireless/ath/ath9k/htc_hst.c
+++ b/drivers/net/wireless/ath/ath9k/htc_hst.c
@@ -294,6 +294,9 @@ int htc_connect_service(struct htc_target *target,
 		return -ETIMEDOUT;
 	}
 
+	if (unlikely(target->conn_rsp_epid == ENDPOINT_UNUSED))
+		return -ENODEV;
+
 	*conn_rsp_epid = target->conn_rsp_epid;
 	return 0;
 err:
-- 
2.47.0

Re: [PATCH v2] wifi: ath9k: fix out-of-bounds endpoint index in htc_issue_send()

[Toke Høiland-Jørgensen]

Dmitry Antipov <dmantipov@yandex.ru> writes:

> Syzbot has reported the following UBSAN splat:
>
> UBSAN: array-index-out-of-bounds in drivers/net/wireless/ath/ath9k/htc_hst.c:26:35
> index 255 is out of range for type 'struct htc_endpoint[22]'
> ...
> Call Trace:
>  ...
>  __ubsan_handle_out_of_bounds+0x115/0x140
>  htc_send_epid+0x1e0/0x2d0
>  ath9k_wmi_cmd+0x31f/0x590
>  ath9k_regread+0xdc/0x140
>  ? __pfx_ath9k_regread+0x10/0x10
>  ? _raw_spin_unlock_irqrestore+0xae/0x110
>  ? __pfx__raw_spin_unlock_irqrestore+0x10/0x10
>  ? __pfx_ath9k_regread+0x10/0x10
>  ath9k_hw_init+0x4b0/0x2230
>  ? init_timer_key+0x171/0x2d0
>  ath9k_htc_probe_device+0xa66/0x1fe0
>  ? __pfx_ath9k_htc_probe_device+0x10/0x10
>  ? usb_hcd_submit_urb+0x345/0x1ae0
>  ? usb_free_urb+0x9d/0x120
>  ? ath9k_hif_usb_alloc_urbs+0x1077/0x1150
>  ath9k_htc_hw_init+0x32/0x70
>  ath9k_hif_usb_firmware_cb+0x240/0x4c0
>  ? __pfx_ath9k_hif_usb_firmware_cb+0x10/0x10
>  request_firmware_work_func+0x106/0x1c0
>  ...
>
> IIUC https://syzkaller.appspot.com/text?tag=ReproC&x=1734b9f1180000 makes
> an attempt to emulate ath9k USB device in somewhat weird and obscure way
> so I'm not sure whether this may happen on a real hardware. OTOH whether
> ending with ENDPOINT_UNUSED in 'htc_connect_service()' may indicate the
> malfunctioned real device, it may be better to reject such a device with
> -ENODEV.
>
> Reported-by: syzbot+93cbd5fbb85814306ba1@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=93cbd5fbb85814306ba1
> Fixes: fb9987d0f748 ("ath9k_htc: Support for AR9271 chipset.")
> Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>

Heh, so pwcli was complaining that the patch doesn't apply, which turns
out to be because of this fix for the same thing:

8619593634cb ("wifi: ath9k: add range check for conn_rsp_epid in htc_connect_service()")

So dropping this...

-Toke