Re: [PATCH] amixer: add support for TLV byte control read

[Vinod Koul <vinod.koul@intel.com>]

On Wed, Jan 27, 2016 at 07:49:00PM +0100, Takashi Iwai wrote:
> On Wed, 27 Jan 2016 18:47:52 +0100,
> Vinod Koul wrote:
> > 
> > On Fri, Jan 22, 2016 at 11:50:08AM +0100, Takashi Iwai wrote:
> > > On Fri, 22 Jan 2016 10:56:48 +0100,
> > > Vinod Koul wrote:
> > > > > Hm, it has a check
> > > > > 
> > > > > static int snd_ctl_hw_elem_tlv(snd_ctl_t *handle, int op_flag,
> > > > > 			       unsigned int numid,
> > > > > 			       unsigned int *tlv, unsigned int tlv_size)
> > > > > {
> > > > > 	....
> > > > > 		if (xtlv->tlv[1] + 2 * sizeof(unsigned int) > tlv_size) {
> > > > > 			free(xtlv);
> > > > > 			return -EFAULT;
> > > > > 		}
> > > > > 		memcpy(tlv, xtlv->tlv, xtlv->tlv[1] + 2 * sizeof(unsigned int));
> > > > > 
> > > > > Do you mean somewhere here triggers a crash?
> > > > 
> > > > Yes while it tried to memcpy, in my case 30K sized data to 4K buffer, so I
> > > > allocated right one and got all the data
> > > 
> > > But it has the size check before memcpy()?
> > 
> > Okay I had some issues (yet again) so I rebuilt the ubuntu packages and tools and reran
> > this.
> > 
> > I get:
> > 
> > $ amixer -c0 cget numid=16
> > numid=16,iface=MIXER,name='mdl params'
> >   ; type=BYTES,access=-----RW-,values=30336
> > amixer: Control hw:0 element TLV read error: Bad address
> > 
> > Segmentation fault (core dumped)
> > 
> > And the alsa-lib returns an err and we try to free up tlv on error which
> > should be fine here but fails
> > 
> > Why should free cause a failure here (on 1.0.27)
> > 
> > On latest lib from git. I get same failure but in alsa-lib when it tries to
> > free xtlv on failure of this check
> > 
> > if (xtlv->tlv[1] + 2 * sizeof(unsigned int) > tlv_size) {
> >                    free(xtlv);
> > 
> > Why would that happen...? What am I missing :(
> 
> No idea.  You should check it via gdb.

Yes did, but didn't get a clue. tlv seems to be valid before free and points
to right location and contents

> > Also on second thought, even after working around this by setting right size
> > so that we don't hit failures (this patch), I get my screen spammed by 30K byes so based
> > on TLV access and bytes type I am thinking to not read and display on cget.
> > 
> > We should add new options for these controls ..
> 
> As already mentioned, we may handle differently for ctl elements that
> have no r/w access.  At easiest, just skip tlv read for such an
> element.

Yes will send a patch for this :)

-- 
~Vinod