On Fri, 19 Jul 2019 17:08:48 +0200, Geoffrey D. Bennett wrote:
Hi Takashi,
On Thu, Jul 18, 2019 at 05:53:13PM +0200, Takashi Iwai wrote: [...]
Thanks, the patch looks almost good, but it's already too late for 5.3, so I'm going to queue this for 5.4 after 5.3 merge window is closed in this week.
Thank you. I really appreciate your assistance and feedback.
But, before that, maybe one more refresh would be appreciated. Namely,
- We need a verification of the fixed pipe before actually submitting urb. scarlett2_usb() calls with usb_sndctrlpipe(), and this pipe has to be verified beforehand. See the commit 801ebf1043ae for details.
I had a look at that commit, and I think I don't need to change scarlett2_usb() because it calls snd_usb_ctl_msg() which already calls the snd_usb_pipe_sanity_check() function to verify the pipe.
Ah right, then it's superfluous.
I'm thinking though that scarlett2_mixer_status_create() which does something like this:
struct usb_device *dev = mixer->chip->dev; unsigned int pipe = usb_rcvintpipe(dev, SCARLETT2_USB_INTERRUPT_ENDPOINT); ... usb_fill_int_urb(mixer->urb, dev, pipe, ...); usb_submit_urb(mixer->urb, GFP_KERNEL);
probably needs this added:
if (snd_usb_pipe_sanity_check(dev, pipe)) return -EINVAL;
Do you agree?
Yes, if you call directly, you need the verification.
thanks,
Takashi