ux500 breaking multi-platform kernels

ux500 breaking multi-platform kernels

[Rob Herring]

Lee, Linus,

There are some initcalls on ux500 that are not properly
conditionalized. The first is in ux500_l2x0_init and here is a fix:

diff --git a/arch/arm/mach-ux500/cache-l2x0.c b/arch/arm/mach-ux500/cache-l2x0.c
index f58615b..e066d32 100644
--- a/arch/arm/mach-ux500/cache-l2x0.c
+++ b/arch/arm/mach-ux500/cache-l2x0.c
@@ -39,10 +39,10 @@ static int __init ux500_l2x0_init(void)
 {
        u32 aux_val = 0x3e000000;

-       if (cpu_is_u8500_family() || cpu_is_ux540_family())
-               l2x0_base = __io_address(U8500_L2CC_BASE);
-       else
-               ux500_unknown_soc();
+       if (!(cpu_is_u8500_family() || cpu_is_ux540_family()))
+               return -ENODEV;
+
+       l2x0_base = __io_address(U8500_L2CC_BASE);

        /* Unlock before init */
        ux500_l2x0_unlock();


Here is the second problem with cpuidle driver:

[    4.514645] Unable to handle kernel NULL pointer dereference at
virtual address 000000fc
[    4.522729] pgd = c0004000
[    4.525429] [000000fc] *pgd=00000000
[    4.529009] Internal error: Oops: 5 [#1] SMP ARM
[    4.533620] Modules linked in:
[    4.536679] CPU: 0 PID: 1 Comm: swapper/0 Tainted: G        W
3.10.0-rc1-00013-g5d170cd-dirty #527
[    4.545893] task: ed888000 ti: ed884000 task.ti: ed884000
[    4.551297] PC is at config_wakeups+0x6c/0x14c
[    4.555740] LR is at db8500_prcmu_enable_wakeups+0x64/0x7c
[    4.561223] pc : [<c06f4948>]    lr : [<c06f64f4>]    psr: 20000193
[    4.561223] sp : ed885ec8  ip : ed885ef8  fp : ed885ef4
[    4.572693] r10: c2d69494  r9 : 00000000  r8 : c31d92bc
[    4.577911] r7 : 00028881  r6 : 00000000  r5 : c31d92bc  r4 : 00000000
[    4.584431] r3 : 00000000  r2 : 00000000  r1 : 00000000  r0 : 60000113
[    4.590954] Flags: nzCv  IRQs off  FIQs on  Mode SVC_32  ISA ARM
Segment kernel
[    4.598342] Control: 10c53c7d  Table: 0000404a  DAC: 00000017
[    4.604082] Process swapper/0 (pid: 1, stack limit = 0xed884240)
[    4.610083] Stack: (0xed885ec8 to 0xed886000)
[    4.614436] 5ec0:                   00000000 04019494 c31d92bc
00020081 60000113 c3137440
[    4.622611] 5ee0: 00000000 c2d69494 ed885f14 ed885ef8 c06f64f4
c06f48e8 ed884000 00000006
[    4.630786] 5f00: c2d69488 c2c9abdc ed885f24 ed885f18 c2c9abf0
c06f649c ed885f5c ed885f28
[    4.638960] 5f20: c0008a78 c2c9abe8 00000be1 c2a527f4 ed885f5c
00000006 c2d69484 00000006
[    4.647134] 5f40: c2d69488 c2df9efc c3137440 00000be1 ed885f94
ed885f60 c2c71a98 c00089bc
[    4.655307] 5f60: 00000006 00000006 c2c711ac c010ed98 00000000
c1c95000 00000000 00000000
[    4.663479] 5f80: 00000000 00000000 ed885fac ed885f98 c1c95018
c2c718f8 00000000 00000000
[    4.671651] 5fa0: 00000000 ed885fb0 c000ed98 c1c9500c 00000000
00000000 00000000 00000000
[    4.679824] 5fc0: 00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000
[    4.687997] 5fe0: 00000000 00000000 00000000 00000000 00000013
00000000 00000000 00000000
[    4.696183] [<c06f4948>] (config_wakeups+0x6c/0x14c) from
[<c06f64f4>] (db8500_prcmu_enable_wakeups+0x64/0x7c)
[    4.706189] [<c06f64f4>] (db8500_prcmu_enable_wakeups+0x64/0x7c)
from [<c2c9abf0>] (ux500_idle_init+0x14/0x2c)
[    4.716195] [<c2c9abf0>] (ux500_idle_init+0x14/0x2c) from
[<c0008a78>] (do_one_initcall+0xc8/0x184)
[    4.725247] [<c0008a78>] (do_one_initcall+0xc8/0x184) from
[<c2c71a98>] (kernel_init_freeable+0x1ac/0x284)
[    4.734906] [<c2c71a98>] (kernel_init_freeable+0x1ac/0x284) from
[<c1c95018>] (kernel_init+0x18/0xf4)
[    4.744128] [<c1c95018>] (kernel_init+0x18/0xf4) from [<c000ed98>]
(ret_from_fork+0x14/0x20)
[    4.752565] Code: ea000001 e59f50dc e1a08005 e5953054 (e59320fc)
[    4.758656] ---[ end trace 47a577e4e96d6f0b ]---
[    4.763269] Kernel panic - not syncing: Fatal exception

ux500 breaking multi-platform kernels

[Lee Jones]

On Wed, 15 May 2013, Rob Herring wrote:

> Lee, Linus,
> 
> There are some initcalls on ux500 that are not properly
> conditionalized. The first is in ux500_l2x0_init and here is a fix:
> 
> diff --git a/arch/arm/mach-ux500/cache-l2x0.c b/arch/arm/mach-ux500/cache-l2x0.c
> index f58615b..e066d32 100644
> --- a/arch/arm/mach-ux500/cache-l2x0.c
> +++ b/arch/arm/mach-ux500/cache-l2x0.c
> @@ -39,10 +39,10 @@ static int __init ux500_l2x0_init(void)
>  {
>         u32 aux_val = 0x3e000000;
> 
> -       if (cpu_is_u8500_family() || cpu_is_ux540_family())
> -               l2x0_base = __io_address(U8500_L2CC_BASE);
> -       else
> -               ux500_unknown_soc();
> +       if (!(cpu_is_u8500_family() || cpu_is_ux540_family()))
> +               return -ENODEV;
> +
> +       l2x0_base = __io_address(U8500_L2CC_BASE);
> 
>         /* Unlock before init */
>         ux500_l2x0_unlock();

Ah, I see. Nice solution.

Are you going to forward this as a patch? Do you require an Ack?

> Here is the second problem with cpuidle driver:

Can you re-run this and send me your vmlinux file and this output for
'that run' please? I'll endeavour to fix.

> [    4.514645] Unable to handle kernel NULL pointer dereference at
> virtual address 000000fc
> [    4.522729] pgd = c0004000
> [    4.525429] [000000fc] *pgd=00000000
> [    4.529009] Internal error: Oops: 5 [#1] SMP ARM
> [    4.533620] Modules linked in:
> [    4.536679] CPU: 0 PID: 1 Comm: swapper/0 Tainted: G        W
> 3.10.0-rc1-00013-g5d170cd-dirty #527
> [    4.545893] task: ed888000 ti: ed884000 task.ti: ed884000
> [    4.551297] PC is at config_wakeups+0x6c/0x14c
> [    4.555740] LR is at db8500_prcmu_enable_wakeups+0x64/0x7c
> [    4.561223] pc : [<c06f4948>]    lr : [<c06f64f4>]    psr: 20000193
> [    4.561223] sp : ed885ec8  ip : ed885ef8  fp : ed885ef4
> [    4.572693] r10: c2d69494  r9 : 00000000  r8 : c31d92bc
> [    4.577911] r7 : 00028881  r6 : 00000000  r5 : c31d92bc  r4 : 00000000
> [    4.584431] r3 : 00000000  r2 : 00000000  r1 : 00000000  r0 : 60000113
> [    4.590954] Flags: nzCv  IRQs off  FIQs on  Mode SVC_32  ISA ARM
> Segment kernel
> [    4.598342] Control: 10c53c7d  Table: 0000404a  DAC: 00000017
> [    4.604082] Process swapper/0 (pid: 1, stack limit = 0xed884240)
> [    4.610083] Stack: (0xed885ec8 to 0xed886000)
> [    4.614436] 5ec0:                   00000000 04019494 c31d92bc
> 00020081 60000113 c3137440
> [    4.622611] 5ee0: 00000000 c2d69494 ed885f14 ed885ef8 c06f64f4
> c06f48e8 ed884000 00000006
> [    4.630786] 5f00: c2d69488 c2c9abdc ed885f24 ed885f18 c2c9abf0
> c06f649c ed885f5c ed885f28
> [    4.638960] 5f20: c0008a78 c2c9abe8 00000be1 c2a527f4 ed885f5c
> 00000006 c2d69484 00000006
> [    4.647134] 5f40: c2d69488 c2df9efc c3137440 00000be1 ed885f94
> ed885f60 c2c71a98 c00089bc
> [    4.655307] 5f60: 00000006 00000006 c2c711ac c010ed98 00000000
> c1c95000 00000000 00000000
> [    4.663479] 5f80: 00000000 00000000 ed885fac ed885f98 c1c95018
> c2c718f8 00000000 00000000
> [    4.671651] 5fa0: 00000000 ed885fb0 c000ed98 c1c9500c 00000000
> 00000000 00000000 00000000
> [    4.679824] 5fc0: 00000000 00000000 00000000 00000000 00000000
> 00000000 00000000 00000000
> [    4.687997] 5fe0: 00000000 00000000 00000000 00000000 00000013
> 00000000 00000000 00000000
> [    4.696183] [<c06f4948>] (config_wakeups+0x6c/0x14c) from
> [<c06f64f4>] (db8500_prcmu_enable_wakeups+0x64/0x7c)
> [    4.706189] [<c06f64f4>] (db8500_prcmu_enable_wakeups+0x64/0x7c)
> from [<c2c9abf0>] (ux500_idle_init+0x14/0x2c)
> [    4.716195] [<c2c9abf0>] (ux500_idle_init+0x14/0x2c) from
> [<c0008a78>] (do_one_initcall+0xc8/0x184)
> [    4.725247] [<c0008a78>] (do_one_initcall+0xc8/0x184) from
> [<c2c71a98>] (kernel_init_freeable+0x1ac/0x284)
> [    4.734906] [<c2c71a98>] (kernel_init_freeable+0x1ac/0x284) from
> [<c1c95018>] (kernel_init+0x18/0xf4)
> [    4.744128] [<c1c95018>] (kernel_init+0x18/0xf4) from [<c000ed98>]
> (ret_from_fork+0x14/0x20)
> [    4.752565] Code: ea000001 e59f50dc e1a08005 e5953054 (e59320fc)
> [    4.758656] ---[ end trace 47a577e4e96d6f0b ]---
> [    4.763269] Kernel panic - not syncing: Fatal exception

-- 
Lee Jones
Linaro ST-Ericsson Landing Team Lead
Linaro.org ? Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog

ux500 breaking multi-platform kernels

[Lee Jones]

> > Here is the second problem with cpuidle driver:
> 
> Can you re-run this and send me your vmlinux file and this output for
> 'that run' please? I'll endeavour to fix.

Scrap that. I see what's happening. I'll fix.

-- 
Lee Jones
Linaro ST-Ericsson Landing Team Lead
Linaro.org ? Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog

ux500 breaking multi-platform kernels

[Linus Walleij]

On Wed, May 15, 2013 at 10:30 PM, Rob Herring <robherring2@gmail.com> wrote:

> There are some initcalls on ux500 that are not properly
> conditionalized.

Oh wow you actually compiled it in and tried to boot it on something
else :-D

Highbank I suppose?

> The first is in ux500_l2x0_init and here is a fix:

I guess this is all -rc material. Can you send a proper patch (set)?

Unfortunately I don't have any other ARMv7 system than ux500 so
I cannot test multiplatform on any other machine myself.
(I have several ARMv5:s...)

Yours,
Linus Walleij

ux500 breaking multi-platform kernels

[Lee Jones]

On Fri, 17 May 2013, Linus Walleij wrote:

> On Wed, May 15, 2013 at 10:30 PM, Rob Herring <robherring2@gmail.com> wrote:
> 
> > There are some initcalls on ux500 that are not properly
> > conditionalized.
> 
> Oh wow you actually compiled it in and tried to boot it on something
> else :-D
> 
> Highbank I suppose?
> 
> > The first is in ux500_l2x0_init and here is a fix:
> 
> I guess this is all -rc material. Can you send a proper patch (set)?

Feel free to add my fix for cpuidle to this set if it works for you.

If it doesn't I'll re-work.

-- 
Lee Jones
Linaro ST-Ericsson Landing Team Lead
Linaro.org ? Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog

ux500 breaking multi-platform kernels

[Srinivas KANDAGATLA]

On 15/05/13 21:30, Rob Herring wrote:
> @@ -39,10 +39,10 @@ static int __init ux500_l2x0_init(void)
>  {
>         u32 aux_val = 0x3e000000;
> 
> -       if (cpu_is_u8500_family() || cpu_is_ux540_family())
> -               l2x0_base = __io_address(U8500_L2CC_BASE);
> -       else
> -               ux500_unknown_soc();
> +       if (!(cpu_is_u8500_family() || cpu_is_ux540_family()))
> +               return -ENODEV;
> +
> +       l2x0_base = __io_address(U8500_L2CC_BASE);

I think moving
early_initcall(ux500_l2x0_init);

to
MACHINE
.init_early = ux500_l2x0_init

should solve this issue neatly.


--srini

ux500 breaking multi-platform kernels

[Rob Herring]

On Fri, May 17, 2013 at 1:29 AM, Linus Walleij <linus.walleij@linaro.org> wrote:
> On Wed, May 15, 2013 at 10:30 PM, Rob Herring <robherring2@gmail.com> wrote:
>
>> There are some initcalls on ux500 that are not properly
>> conditionalized.
>
> Oh wow you actually compiled it in and tried to boot it on something
> else :-D
>
> Highbank I suppose?
>
>> The first is in ux500_l2x0_init and here is a fix:
>
> I guess this is all -rc material. Can you send a proper patch (set)?

It is. I don't really think this is the ideal fix. It would be better
if you could make this not be an unconditional initcall.

Rob