bcm2835_pinctrl_probe: WARNING: CPU: 0 PID: 1 at kernel/irq/chip.c:850

[Stefan Wahren]

Hi Linus,

> Linus Walleij <linus.walleij at linaro.org> hat am 20. Juni 2017 um 13:32 geschrieben:
> 
> 
> On Mon, Jun 19, 2017 at 10:59 PM, Stefan Wahren <stefan.wahren at i2se.com> wrote:
> 
> > i'm currently working Raspberry Pi Zero W support and had problems to boot with Mini UART
> > (no sign of life after startup). After switching to pl011 with EARLY_PRINTK, the Zero W boots
> > and i get the following output:
> 
> (...)
> > [    0.255621] ------------[ cut here ]------------
> > [    0.255742] WARNING: CPU: 0 PID: 1 at kernel/irq/chip.c:850 __irq_do_set_handler+0x4c/0x14c
> > [    0.255811] Modules linked in:
> > [    0.255862] CPU: 0 PID: 1 Comm: swapper Not tainted 4.12.0-rc5-next-20170616+ #3
> > [    0.255924] Hardware name: Generic DT based system
> > [    0.256009] [<c010f64c>] (unwind_backtrace) from [<c010c604>] (show_stack+0x20/0x24)
> > [    0.256109] [<c010c604>] (show_stack) from [<c037bb88>] (dump_stack+0x20/0x28)
> > [    0.256204] [<c037bb88>] (dump_stack) from [<c011e76c>] (__warn+0xe4/0x110)
> > [    0.256270] [<c011e76c>] (__warn) from [<c011e868>] (warn_slowpath_null+0x30/0x38)
> > [    0.256355] [<c011e868>] (warn_slowpath_null) from [<c01590f0>] (__irq_do_set_handler+0x4c/0x14c)
> > [    0.256444] [<c01590f0>] (__irq_do_set_handler) from [<c01592dc>] (irq_set_chained_handler_and_data+0x4c/0x64)
> > [    0.256548] [<c01592dc>] (irq_set_chained_handler_and_data) from [<c03ae534>] (gpiochip_set_chained_irqchip+0x80/0xa0)
> > [    0.256641] [<c03ae534>] (gpiochip_set_chained_irqchip) from [<c03aa9e8>] (bcm2835_pinctrl_probe+0x31c/0x3c8)
> > [    0.256746] [<c03aa9e8>] (bcm2835_pinctrl_probe) from [<c044f03c>] (platform_drv_probe+0x60/0xac)
> > [    0.256840] [<c044f03c>] (platform_drv_probe) from [<c044d434>] (driver_probe_device+0x1d8/0x438)
> > [    0.256929] [<c044d434>] (driver_probe_device) from [<c044d730>] (__driver_attach+0x9c/0x100)
> > [    0.257014] [<c044d730>] (__driver_attach) from [<c044b6e4>] (bus_for_each_dev+0x84/0x98)
> > [    0.257098] [<c044b6e4>] (bus_for_each_dev) from [<c044cd7c>] (driver_attach+0x28/0x30)
> > [    0.257181] [<c044cd7c>] (driver_attach) from [<c044c7dc>] (bus_add_drcm2835_pinctrl_driver_init+0x20/0x28)
> > [    0.257535] [<c0b2c9b0>] (bcm2835_pinctrl_driver_init) from [<c0101ac0>] (do_one_initcall+0xc0/0x188)
> > [    0.257635] [<c0101ac0>] (do_one_initcall) from [<c0b00f64>] (kernel_init_freeable+0x170/0x234)
> > [    0.257728] [<c0b00f64>] (kernel_init_freeable) from [<c07366b4>] (kernel_init+0x18/0x11c)
> > [    0.257822] [<c07366b4>] (kernel_init) from [<c0107f38>] (ret_from_fork+0x14/0x3c)
> > [    0.257904] random: init_oops_id+0x30/0x4c get_random_bytes called with crng_init=0
> > [    0.257980] ---[ end trace 9c0064da26f1596a ]---
> > [    0.261962] Serial: 8250/16550 driver, 1 ports, IRQ sharing enabled
> > [    0.265460] libphy: Fixed MDIO Bus: probed
> > ...
> >
> > According to the source code, this is the relevant line:
> >
> > if (WARN_ON(!irq_data || irq_data->chip == &no_irq_chip))
> >         return;
> 
> Isn't that because something is using the irqchip before the irqchip
> is registered, so an initcall
> problem?
> 
> I don't really know how some stuff like irqchips deal with deferred
> probe or if they even can.
> 
> I had some similar patch the other day fixing something like this in
> another driver,
> I guess you can try reordering the initcalls. (Sorry.)
> 

thanks, but it seems more trivial. The return value of irq_of_parse_and_map() wasn't check before calling gpiochip_set_chained_irqchip().

So this change (similar to a downstream patch) avoid this warning:

diff --git a/drivers/pinctrl/bcm/pinctrl-bcm2835.c b/drivers/pinctrl/bcm/pinctrl-bcm2835.c
index 1eb7a1a..2a29128 100644
--- a/drivers/pinctrl/bcm/pinctrl-bcm2835.c
+++ b/drivers/pinctrl/bcm/pinctrl-bcm2835.c
@@ -1048,6 +1048,10 @@ static int bcm2835_pinctrl_probe(struct platform_device *pdev)
        for (i = 0; i < BCM2835_NUM_IRQS; i++) {
                pc->irq[i] = irq_of_parse_and_map(np, i);
                pc->irq_group[i] = i;
+
+               if (pc->irq[i] == 0)
+                       continue;
+
                /*
                 * Use the same handler for all groups: this is necessary
                 * since we use one gpiochip to cover all lines - the

> Yours,
> Linus Walleij