What the semantics of DSP_API?
Víctor M. Jáquez L.
I was testing the kernel 3.3.0-rc2 and gst-dsp 0.10.1 and I got stuck in
MMUFAULT errors.
After digging I found that I should set DSP_API=1 to fix that behavior, that
means do not use the DMA.
What's the meaning of DSP_API?
Is there a relation between the socket node release and the DSP_API? If so, I
think it should be documented somewhere.
vmjl
Felipe Contreras
2012/2/2 Víctor M. Jáquez L. <vja...@igalia.com>:
> I was testing the kernel 3.3.0-rc2 and gst-dsp 0.10.1 and I got stuck in
> MMUFAULT errors.
>
> After digging I found that I should set DSP_API=1 to fix that behavior, that
> means do not use the DMA.
>
> What's the meaning of DSP_API?
It's the interface to the tidspbridge driver.
> Is there a relation between the socket node release and the DSP_API? If so, I
> think it should be documented somewhere.
No, there shouldn't be, that's what SN_API is for.
Generally you should care about this, just use the default (latest and
greatest).
However, the issues might have been triggered by:
https://github.com/felipec/gst-dsp/commit/b60bd3e573b1a2199bfc30a1ee4acff58a2639e1
I haven't tested that mode myself all that much, but should work, as
flush/invalidate are supposed to be only wrappers around the begin/end
ioctls.
Cheers.
--
Felipe Contreras
Víctor M. Jáquez L.
> Hi,
>
> 2012/2/2 V�ctor M. J�quez L. <vja...@igalia.com>:
> > I was testing the kernel 3.3.0-rc2 and gst-dsp 0.10.1 and I got stuck in
> > MMUFAULT errors.
> >
> > After digging I found that I should set DSP_API=1 to fix that behavior, that
> > means do not use the DMA.
> >
> > What's the meaning of DSP_API?
>
> It's the interface to the tidspbridge driver.
>
> > Is there a relation between the socket node release and the DSP_API? If so, I
> > think it should be documented somewhere.
>
> No, there shouldn't be, that's what SN_API is for.
>
> Generally you should care about this, just use the default (latest and
> greatest).
>
> However, the issues might have been triggered by:
> https://github.com/felipec/gst-dsp/commit/b60bd3e573b1a2199bfc30a1ee4acff58a2639e1
Yep, precisely that commit is triggering the MMUFAULT in every pipeline I
ran.
Perhaps there's a lost commit in the kernel 3.3.0-rc2
vmjl