[PATCH v2 6/6] drm/bridge: cdns-mhdp8546: Fix the interrupt enable/disable

[Nikhil Devshatwar]

On 14:32-20201110, Swapnil Jakhade wrote:
> 
> 
> > -----Original Message-----
> > From: Nikhil Devshatwar <nikhil.nd at ti.com>
> > Sent: Tuesday, November 10, 2020 7:23 PM
> > To: Tomi Valkeinen <tomi.valkeinen at ti.com>; Swapnil Kashinath Jakhade
> > <sjakhade at cadence.com>; Yuti Suresh Amonkar <yamonkar at cadence.com>
> > Cc: dri-devel at lists.freedesktop.org; Swapnil Kashinath Jakhade
> > <sjakhade at cadence.com>; Sekhar Nori <nsekhar at ti.com>; Laurent Pinchart
> > <laurent.pinchart at ideasonboard.com>; Yuti Suresh Amonkar
> > <yamonkar at cadence.com>
> > Subject: Re: [PATCH v2 6/6] drm/bridge: cdns-mhdp8546: Fix the interrupt
> > enable/disable
> > 
> > EXTERNAL MAIL
> > 
> > 
> > On 14:27-20201110, Tomi Valkeinen wrote:
> > > On 10/11/2020 12:27, Nikhil Devshatwar wrote:
> > > > On 11:21-20201110, Tomi Valkeinen wrote:
> > > >> On 09/11/2020 19:06, Nikhil Devshatwar wrote:
> > > >>> When removing the tidss driver, there is a warning reported by
> > > >>> kernel about an unhandled interrupt for mhdp driver.
> > > >>>
> > > >>> [   43.238895] irq 31: nobody cared (try booting with the "irqpoll"
> > option)
> > > >>> ... [snipped backtrace]
> > > >>> [   43.330735] handlers:
> > > >>> [   43.333020] [<000000005367c4f9>] irq_default_primary_handler
> > threaded [<000000007e02b601>]
> > > >>> cdns_mhdp_irq_handler [cdns_mhdp8546]
> > > >>> [   43.344607] Disabling IRQ #31
> > > >>>
> > > >>> This happens because as part of cdns_mhdp_bridge_hpd_disable,
> > > >>> driver tries to disable the interrupts. While disabling the
> > > >>> SW_EVENT interrupts, it accidentally enables the MBOX interrupts,
> > > >>> which are not handled by the driver.
> > > >>>
> > > >>> Fix this with a read-modify-write to update only required bits.
> > > >>> Do the same for enabling interrupts as well.
> > > >>>
> > > >>> Signed-off-by: Nikhil Devshatwar <nikhil.nd at ti.com>
> > > >>> ---
> > > >>>  drivers/gpu/drm/bridge/cadence/cdns-mhdp8546-core.c | 7 +++++--
> > > >>>  1 file changed, 5 insertions(+), 2 deletions(-)
> > > >>>
> > > >>> diff --git a/drivers/gpu/drm/bridge/cadence/cdns-mhdp8546-core.c
> > > >>> b/drivers/gpu/drm/bridge/cadence/cdns-mhdp8546-core.c
> > > >>> index 2cd809eed827..6beccd2a408e 100644
> > > >>> --- a/drivers/gpu/drm/bridge/cadence/cdns-mhdp8546-core.c
> > > >>> +++ b/drivers/gpu/drm/bridge/cadence/cdns-mhdp8546-core.c
> > > >>> @@ -2146,7 +2146,8 @@ static void
> > > >>> cdns_mhdp_bridge_hpd_enable(struct drm_bridge *bridge)
> > > >>>
> > > >>>  	/* Enable SW event interrupts */
> > > >>>  	if (mhdp->bridge_attached)
> > > >>> -		writel(~(u32)CDNS_APB_INT_MASK_SW_EVENT_INT,
> > > >>> +		writel(readl(mhdp->regs + CDNS_APB_INT_MASK) &
> > > >>> +		       ~CDNS_APB_INT_MASK_SW_EVENT_INT,
> > > >>>  		       mhdp->regs + CDNS_APB_INT_MASK);  }
> > > >>>
> > > >>> @@ -2154,7 +2155,9 @@ static void
> > > >>> cdns_mhdp_bridge_hpd_disable(struct drm_bridge *bridge)  {
> > > >>>  	struct cdns_mhdp_device *mhdp = bridge_to_mhdp(bridge);
> > > >>>
> > > >>> -	writel(CDNS_APB_INT_MASK_SW_EVENT_INT, mhdp->regs +
> > CDNS_APB_INT_MASK);
> > > >>> +	writel(readl(mhdp->regs + CDNS_APB_INT_MASK) |
> > > >>> +	       CDNS_APB_INT_MASK_SW_EVENT_INT,
> > > >>> +	       mhdp->regs + CDNS_APB_INT_MASK);
> > > >>>  }
> > > >>>
> > > >>>  static const struct drm_bridge_funcs cdns_mhdp_bridge_funcs = {
> > > >>
> > > >> Good catch. I wonder why we need the above functions... We already
> > > >> enable and disable the interrupts when attaching/detaching the
> > > >> driver. And I think we want to get the interrupt even if we won't report
> > HPD (but I think we always do report it), as we need the interrupts to track
> > the link status.
> > > >>
> > > >
> > > > I read from the code that there is TODO for handling the mailbox
> > > > interrupts in the driver. Once that is supported, you will be able
> > > > to explictily enable/disable interrupts for SW_EVENTS (like hotplug)
> > > > as well as mailbox events. This enabling specific bits in the
> > > > interrupt status.
> > >
> > > But SW_EVENTS is not the same as HPD, at least in theory. If we
> > > disable SW_EVENT_INT in hpd_disable(), we lose all SW_EVENT interrupts.
> > 
> > I am not sure, what exactly is covered in the SW events apart from the
> > hotplug.
> > 
> > Swapnil, Yuti, Please fill in..
> 
> hpd_enable/hpd_disable callbacks were implemented as a part of supporting
> DRM_BRIDGE_OP_HPD bridge operation. The existing implementation could
> work with current features set supported by MHDP driver. But Tomi's point is
> valid, as there are some HDCP interrupts which are part of SW_EVENT interrupts
> and this might not be the control to just enable/disable HPD.

So do you think something should change now?
I assume that you will modify this part when supporting HDCP.
Also, please provide your ack if this patch is the right fix.


Nikhil D
> 
> Swapnil
> 
> > 
> > Nikhil D
> > >
> > >  Tomi
> > >
> > > --
> > > Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
> > > Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki