[PATCH v2 0/8] Raspberry Pi 2 support.

[arnd@arndb.de (Arnd Bergmann)]

On Tuesday 22 December 2015, Alexander Aring wrote:
> On Tue, Dec 22, 2015 at 12:11:32AM +0100, Arnd Bergmann wrote:
> > On Sunday 20 December 2015, Alexander Aring wrote:
> > > On Fri, Dec 18, 2015 at 11:08:26AM +0100, Alexander Aring wrote:
> > > > [    0.000000] Memory policy: Data cache writeback
> > > > [    0.000000] BUG: mapping for 0x3f201000 at 0xf0201000 out of vmalloc space
> > 
> > This is a separate bug, right?
> > 
> 
> Yes, I can remove this BUG when setting:
> 
> VMSPLIT_2G
> 
> or some other option, was "VMSPLIT_3G" before and with this option I
> will get the above message.

Please try to figure out why this happened, it should really work with any
vmsplit, in particular the one that everyone uses.

> > > > [    0.000000] ------------[ cut here ]------------
> > > > [    0.000000] WARNING: CPU: 0 PID: 0 at arch/arm/kernel/devtree.c:149 arm_dt_init_cpu_maps+0x100/0x1a4()
> > > > [    0.000000] DT /cpu 2 nodes greater than max cores 1, capping them
> > > > [    0.000000] Modules linked in:
> > > > [    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.4.0-rc5+ #989
> > > > [    0.000000] Hardware name: BCM2835
> > > > [    0.000000] [<c0016c4c>] (unwind_backtrace) from [<c0013e58>] (show_stack+0x20/0x24)
> > > > [    0.000000] [<c0013e58>] (show_stack) from [<c0247620>] (dump_stack+0x20/0x28)
> > > > [    0.000000] [<c0247620>] (dump_stack) from [<c0023a84>] (warn_slowpath_common+0x9c/0xc4)
> > > > [    0.000000] [<c0023a84>] (warn_slowpath_common) from [<c0023aec>] (warn_slowpath_fmt+0x40/0x48)
> > > > [    0.000000] [<c0023aec>] (warn_slowpath_fmt) from [<c074ef94>] (arm_dt_init_cpu_maps+0x100/0x1a4)
> > > > [    0.000000] [<c074ef94>] (arm_dt_init_cpu_maps) from [<c074e60c>] (setup_arch+0x6f4/0x89c)
> > > > [    0.000000] [<c074e60c>] (setup_arch) from [<c074b9cc>] (start_kernel+0x74/0x3a4)
> > > > [    0.000000] [<c074b9cc>] (start_kernel) from [<00008078>] (0x8078)
> > > > [    0.000000] ---[ end trace cb88537fdc8fa200 ]---
> > > > [    0.000000] DT missing boot CPU MPIDR[23:0], fall back to default cpu_logical_map
> > > > [    0.000000] CPU: All CPU(s) started in SVC mode.
> > > 
> > > I can remove this WARNING when I enable CONFIG_SMP.
> > 
> > I think we should try to change this in the code.
> > 
> 
> mhhh, okay... I am not sure I think there is some missing
> IS_ENABLED(CONFIG_SMP) check.

Makes sense.

> > > > [    0.000000] Virtual kernel memory layout:
> > > > [    0.000000]     vector  : 0xffff0000 - 0xffff1000   (   4 kB)
> > > > [    0.000000]     fixmap  : 0xffc00000 - 0xfff00000   (3072 kB)
> > > > [    0.000000]     vmalloc : 0xf0800000 - 0xff800000   ( 240 MB)
> > > > [    0.000000]     lowmem  : 0xc0000000 - 0xf0000000   ( 768 MB)
> > > > [    0.000000]     modules : 0xbf000000 - 0xc0000000   (  16 MB)
> > > > [    0.000000]       .text : 0xc0008000 - 0xc074a7cc   (7434 kB)
> > > > [    0.000000]       .init : 0xc074b000 - 0xc07b4000   ( 420 kB)
> > > > [    0.000000]       .data : 0xc07b4000 - 0xc081e910   ( 427 kB)
> > > > [    0.000000]        .bss : 0xc081e910 - 0xc08ca80c   ( 688 kB)
> > > > [    0.000000] SLUB: HWalign=64, Order=0-3, MinObjects=0, CPUs=1, Nodes=1
> > > > [    0.000000] NR_IRQS:16 nr_irqs:16 16
> > > > [    0.000000] Kernel panic - not syncing: /soc/local_intc: unable to map local interrupt registers
> > > > [    0.000000] 
> > > > [    0.000000] CPU: 0 PID: 0 Comm: swapper Tainted: G        W       4.4.0-rc5+ #989
> > > > [    0.000000] Hardware name: BCM2835
> > > > [    0.000000] [<c0016c4c>] (unwind_backtrace) from [<c0013e58>] (show_stack+0x20/0x24)
> > > > [    0.000000] [<c0013e58>] (show_stack) from [<c0247620>] (dump_stack+0x20/0x28)
> > > > [    0.000000] [<c0247620>] (dump_stack) from [<c00c17c0>] (panic+0x84/0x210)
> > > > [    0.000000] [<c00c17c0>] (panic) from [<c0771d5c>] (bcm2836_arm_irqchip_l1_intc_of_init+0x94/0x110)
> > > > [    0.000000] [<c0771d5c>] (bcm2836_arm_irqchip_l1_intc_of_init) from [<c0779f04>] (of_irq_init+0x1a0/0x2a8)
> > > > [    0.000000] [<c0779f04>] (of_irq_init) from [<c0771af4>] (irqchip_init+0x14/0x1c)
> > > > [    0.000000] [<c0771af4>] (irqchip_init) from [<c074d7a8>] (init_IRQ+0x28/0x88)
> > > > [    0.000000] [<c074d7a8>] (init_IRQ) from [<c074bb64>] (start_kernel+0x20c/0x3a4)
> > > > [    0.000000] [<c074bb64>] (start_kernel) from [<00008078>] (0x8078)
> > > > [    0.000000] ---[ end Kernel panic - not syncing: /soc/local_intc: unable to map local interrupt registers
> > > > 
> > > 
> > > But still getting this panic, your patches does not contain some
> > > defconfig, so I still try to figure out what I need to enable to get it
> > > working. Maybe it is some missing config entry which should be enabled.
> > > 
> > > But, for me it looks like some devicetree issue, because of_iomem returns
> > > NULL, at [0]. Don't know what I am doing wrong.
> > 
> > That is the most likely cause, yes. You can try replacing it with an ioremap with a
> > hardcoded physical address to see if that works.
> > 
> 
> When I changed to "VMSPLIT_2G" the above issue doesn't occur anymore.

Someone recently had a similar bug on some freescale platform, but unfortunately I don't
remember the details. As above, it should really not depend on that option.

> Later while booting the kernel hangs forever, the solution on my side
> was to enable:
> 
> CONFIG_HAVE_ARM_ARCH_TIMER
> 
> I think this doesn't use the BCM2835 timer anymore and some cortex-a7
> related "generic timer", or?

Right, but I don't see how that relates to the irqchip. Is the BCM2835 timer
connected to a secondary irqchip, while the arch timer connects to the primary GIC?

> btw:
> 
> After the RPi is booted succesful I get several:
> 
> [   35.048513] unexpected IRQ trap at vector 00
> 
> When I type something via uart (after each character).
> 
> 
> Any ideas?

That also indicates that there is something wrong with the irqchip.

	Arnd