On Tue, Apr 28, 2020 at 09:11:15PM +0300, Cristian Ciocaltea wrote:[...]
On Tue, Apr 28, 2020 at 10:19:21PM +0530, Manivannan Sadhasivam wrote:
On Tue, Apr 28, 2020 at 01:56:12PM +0300, Cristian Ciocaltea wrote:
When the kernel is build with lockdep support and the owl-dma driver is
used, the following message is shown:
The required fix is to use spin_lock_init() on the pchan lock before
attempting to call any spin_lock_irqsave() in owl_dma_get_pchan().
Right, this is a bug. But while looking at the code now, I feel that we don't
need 'pchan->lock'. The idea was to protect 'pchan->vchan', but I think
'od->lock' is the better candidate for that since it already protects it in
'owl_dma_terminate_pchan'.
So I'd be happy if you remove the lock from 'pchan' and just directly use the
one in 'od'.
Out of curiosity, on which platform you're testing this?
Totally agree, I will send a new patch revision as soon as I do some
more testing.
Coo[l], thanks!
I'm currently experimenting on an Actions S500 based board (Roseapple Pi)
trying to extend, if possible, the existing mainline support for those
SoCs.
Awesome! It's great to see that Actions platform is seeing some attention
these days :)
I don't have much progress so far, since I started quite recently
and I also lack experience in the kernel development area, but I do my
best to come back with more patches once I get a consistent functionality.
No worries. Feel free to reach out to me if you have any questions. There is
a lot of work to do and for sure it will be a good learning curve.
We do have an IRC channel (##linux-actions) for quick discussions. Fee[l] free
to join!