On Tue, Jun 22, 2021 at 09:58:27AM +0200, Takashi Iwai wrote: > On Tue, 22 Jun 2021 09:44:54 +0200, > Geoffrey D. Bennett wrote: > > > > On Tue, Jun 22, 2021 at 09:34:25AM +0200, Takashi Iwai wrote: > > > On Tue, 22 Jun 2021 09:07:20 +0200, > > > Vladimir Sadovnikov wrote: > > > > > > > > Hello Takashi! > > > > > > > > Since Focusrite devices are too advanced in settings, the overall > > > > amount of 256 controls is not enough for these devices (like 18i20). > > > > I would like also to extend this constant up to 1024 or even more > > > > since adding support of software configuration of the device also > > > > can exceed the amount of 512 control elements. > > > > > > This define isn't for the total number of mixer elements. Instead, > > > it's just a size of the bitmap table that contains the head of the > > > linked list for each unit id (in the sense of USB mixer spec). > > > So the number of mixer elements is unlimited. > > > > Sorry I don't understand what's going on then. Am I calling > > snd_usb_mixer_add_control() wrong? Because when I called it more than > > MAX_ID_ELEMS times I got a buffer overflow in mixer->id_elems[] (from > > memory, I can confirm tonight). > > > > snd_usb_create_mixer() has: > > > > mixer->id_elems = kcalloc(MAX_ID_ELEMS, sizeof(*mixer->id_elems), > > GFP_KERNEL); > > > > snd_usb_mixer_add_control() called from mixer_scarlett_gen2.c ends up > > at snd_usb_mixer_add_list() which does: > > > > list->next_id_elem = mixer->id_elems[list->id]; > > mixer->id_elems[list->id] = list; > > > > And list->id was going over MAX_ID_ELEMS. > > Here, list->id is the *USB* mixer unit id, not the ALSA control id or > whatever the internal index. The former should be a byte, hence it > can't be over 256. > > That said, the scarlett2 code calls the function in a wrong way. It > has worked casually, so far, just because the core code doesn't use > the unit id number for significant roles. > > So, as a quick workaround, simply pass 0 or any fixed number under 256 > to list->id (i.e. elem->head.id in scarlett2_add_new_ctl()). That's > all, and the elements are chained in the linked list. Okay, I will fix that tonight. Were patches 1-15 of this set of 31 acceptable? If so, I will send a new set with this fix and the remainder of the patches.
