Re: [PATCH 3/3] amba: Properly handle device probe without IRQ domain

[Saravana Kannan <saravanak@google.com>]

On Tue, Aug 24, 2021 at 9:05 PM Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
>
>
> On 2021/8/25 4:08, Saravana Kannan wrote:
> > On Tue, Aug 24, 2021 at 1:05 PM Rob Herring <robh+dt@kernel.org> wrote:
> >> +Saravana
> >>
> >> Saravana mentioned to me there may be some issues with this one...
> >>
> >>
> >> On Mon, Aug 16, 2021 at 2:43 AM Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
> >>> of_amba_device_create() uses irq_of_parse_and_map() to translate
> >>> a DT interrupt specification into a Linux virtual interrupt number.
> >>>
> >>> But it doesn't properly handle the case where the interrupt controller
> >>> is not yet available, eg, when pl011 interrupt is connected to MBIGEN
> >>> interrupt controller, because the mbigen initialization is too late,
> >>> which will lead to no IRQ due to no IRQ domain found, log is shown below,
> >>>    "irq: no irq domain found for uart0 !"
> >>>
> >>> use of_irq_get() to return -EPROBE_DEFER as above, and in the function
> >>> amba_device_try_add()/amba_device_add(), it will properly handle in such
> >>> case, also return 0 in other fail cases to be consistent as before.
> >>>
> >>> Cc: Russell King <linux@armlinux.org.uk>
> >>> Cc: Rob Herring <robh+dt@kernel.org>
> >>> Cc: Frank Rowand <frowand.list@gmail.com>
> >>> Reported-by: Ruizhe Lin <linruizhe@huawei.com>
> >>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> >>> ---
> >>>   drivers/amba/bus.c    | 27 +++++++++++++++++++++++++++
> >>>   drivers/of/platform.c |  6 +-----
> >>>   2 files changed, 28 insertions(+), 5 deletions(-)
> >>>
> >>> diff --git a/drivers/amba/bus.c b/drivers/amba/bus.c
> >>> index 36f2f42c8014..720aa6cdd402 100644
> >>> --- a/drivers/amba/bus.c
> >>> +++ b/drivers/amba/bus.c
> >>> @@ -19,6 +19,7 @@
> >>>   #include <linux/clk/clk-conf.h>
> >>>   #include <linux/platform_device.h>
> >>>   #include <linux/reset.h>
> >>> +#include <linux/of_irq.h>
> >>>
> >>>   #include <asm/irq.h>
> >>>
> >>> @@ -371,12 +372,38 @@ static void amba_device_release(struct device *dev)
> >>>          kfree(d);
> >>>   }
> >>>
> >>> +static int of_amba_device_decode_irq(struct amba_device *dev)
> >>> +{
> >>> +       struct device_node *node = dev->dev.of_node;
> >>> +       int i, irq = 0;
> >>> +
> >>> +       if (IS_ENABLED(CONFIG_OF_IRQ) && node) {
> >>> +               /* Decode the IRQs and address ranges */
> >>> +               for (i = 0; i < AMBA_NR_IRQS; i++) {
> >>> +                       irq = of_irq_get(node, i);
> >>> +                       if (irq < 0) {
> >>> +                               if (irq == -EPROBE_DEFER)
> >>> +                                       return irq;
> >>> +                               irq = 0;
> >>> +                       }
> >>> +
> >>> +                       dev->irq[i] = irq;
> >>> +               }
> >>> +       }
> >>> +
> >>> +       return 0;
> >>> +}
> >>> +
> >>>   static int amba_device_try_add(struct amba_device *dev, struct resource *parent)
> >>>   {
> >>>          u32 size;
> >>>          void __iomem *tmp;
> >>>          int i, ret;
> >>>
> >>> +       ret = of_amba_device_decode_irq(dev);
> >>> +       if (ret)
> >>> +               goto err_out;
> >>> +
> > Similar to other resources the AMBA bus "gets" for the device, I think
> > this should be moved into amba_probe() and not here. There's no reason
> > to delay the addition of the device (and loading its module) because
> > the IRQ isn't ready yet.
>
> The following code in the amba_device_try_add() will be called, it uses irq[0]
> and irq[1], so I put of_amba_device_decode_irq() into amba_device_try_add().
>
> 470         if (dev->irq[0])
> 471                 ret = device_create_file(&dev->dev, &dev_attr_irq0);
> 472         if (ret == 0 && dev->irq[1])
> 473                 ret = device_create_file(&dev->dev, &dev_attr_irq1);
> 474         if (ret == 0)
> 475                 return ret;
>
> of_amba_device_decode_irq() in amba_device_try_add() won't lead to issue,
> only delay the device add, right?

But delaying the device add is the issue. For example, adding a device
could trigger the loading of the corresponding module using uevents.
But now this change would delay that step. That can have other
unintended consequences -- slowing down boot, what if the driver was
working fine without the IRQ, etc.

> If make it into amba_probe(), the above code should be moved too, could we
> make a new patch to move both of them, or don't move them?

I'd say move them both. If Russell hasn't already picked this up, then
I'd say redo your Patch 3/3.

Btw, I've been working on [1] cleaning up the one-off deferred probe
solution that we have for amba devices. That causes a bunch of other
headaches. Your patch 3/3 takes us further in the wrong direction by
adding more reasons for delaying the addition of the device.

-Saravana

[1] - https://lore.kernel.org/lkml/CAGETcx8b228nDUho3cX9AAQ-pXOfZTMv8cj2vhdx9yc_pk8q+A@mail.gmail.com/

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel