AT91RM9200 hang in atmel_serial_putc

AT91RM9200 hang in atmel_serial_putc

[Peter Kardos]

Greetings,

I'm looking for a AT91 specialist that could help me debug barebox v2016.08.

I have a custom board running AT91RM9200 that came with a really old 
custom u-boot (v1.1.4, no ubifs support)...

The first bootloader I've ported (barebox v2015.07, using AT91RM9200ek 
as template) almost works; I'm having trouble getting networking (ping, 
dhcp) working with Micrel KSZ8721 phy. But it boots, works with the 
Flash, etc...
With hopes, that this (networking issue) is resolved with the latest 
barebox release (using the same patchset and config) I've found (??) a 
another issue.

At first it seemed the board just does not boot, after having a look 
with a j-link it seems the board "hangs" in atmel_serial_putc().
I've tried to look at the AT91RM9200 init (lowlevel, peripherals, 
driver) and could not find any significant change (rewrite, define 
changes, etc)...

Any help, ideas, even flame is welcome.

Thanks in advance.
Cheers, Peter

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Sascha Hauer]

Hi Peter,

On Thu, Aug 18, 2016 at 03:08:36AM +0200, Peter Kardos wrote:
> Greetings,
> 
> I'm looking for a AT91 specialist that could help me debug barebox v2016.08.
> 
> I have a custom board running AT91RM9200 that came with a really old custom
> u-boot (v1.1.4, no ubifs support)...
> 
> The first bootloader I've ported (barebox v2015.07, using AT91RM9200ek as
> template) almost works; I'm having trouble getting networking (ping, dhcp)
> working with Micrel KSZ8721 phy. But it boots, works with the Flash, etc...
> With hopes, that this (networking issue) is resolved with the latest barebox
> release (using the same patchset and config) I've found (??) a another
> issue.
> 
> At first it seemed the board just does not boot, after having a look with a
> j-link it seems the board "hangs" in atmel_serial_putc().
> I've tried to look at the AT91RM9200 init (lowlevel, peripherals, driver)
> and could not find any significant change (rewrite, define changes, etc)...

So you have a jtag debugger attached, does it support showing registers?
What is the value of uart->base? Is it correct?
Also you could compare the UART registers with the ones from a working
U-Boot/barebox. Does atmel_serial_init_port() do everything correctly?
Finally it may be that the clock is disabled for some reason, then the
values written to the UART would never be shifted out.
Which instance of the UART are you using? It seems we have:

static struct clk_lookup usart_clocks_lookups[] = {
	CLKDEV_CON_DEV_ID("usart", "atmel_usart0", &mck),
	CLKDEV_CON_DEV_ID("usart", "atmel_usart1", &usart0_clk),
	CLKDEV_CON_DEV_ID("usart", "atmel_usart2", &usart1_clk),
	CLKDEV_CON_DEV_ID("usart", "atmel_usart3", &usart2_clk),
	CLKDEV_CON_DEV_ID("usart", "atmel_usart4", &usart3_clk),
};

The first clock seems to be always-on, but the others could be turned off.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Peter Kardos]

Hi Sascha,

> So you have a jtag debugger attached, does it support showing registers?
> What is the value of uart->base? Is it correct?
> Also you could compare the UART registers with the ones from a working
> U-Boot/barebox.
Yes, I have, it does. I've checked, and it seems to be correct, pointing 
to 0xFFFFF200.

However I didn't think to check the register offset and mask (it was late).

> Which instance of the UART are you using? It seems we have:
I'm using the debug UART (USART0)...

> Does atmel_serial_init_port() do everything correctly?
> Finally it may be that the clock is disabled for some reason,
Was thinking the same thing.

I'll check it (compare reg dumps, console init) tonight and report my 
findings.


Cheers, Peter

On 2016-08-18 09:38, Sascha Hauer wrote:
> Hi Peter,
>
> On Thu, Aug 18, 2016 at 03:08:36AM +0200, Peter Kardos wrote:
>> Greetings,
>>
>> I'm looking for a AT91 specialist that could help me debug barebox v2016.08.
>>
>> I have a custom board running AT91RM9200 that came with a really old custom
>> u-boot (v1.1.4, no ubifs support)...
>>
>> The first bootloader I've ported (barebox v2015.07, using AT91RM9200ek as
>> template) almost works; I'm having trouble getting networking (ping, dhcp)
>> working with Micrel KSZ8721 phy. But it boots, works with the Flash, etc...
>> With hopes, that this (networking issue) is resolved with the latest barebox
>> release (using the same patchset and config) I've found (??) a another
>> issue.
>>
>> At first it seemed the board just does not boot, after having a look with a
>> j-link it seems the board "hangs" in atmel_serial_putc().
>> I've tried to look at the AT91RM9200 init (lowlevel, peripherals, driver)
>> and could not find any significant change (rewrite, define changes, etc)...
> So you have a jtag debugger attached, does it support showing registers?
> What is the value of uart->base? Is it correct?
> Also you could compare the UART registers with the ones from a working
> U-Boot/barebox. Does atmel_serial_init_port() do everything correctly?
> Finally it may be that the clock is disabled for some reason, then the
> values written to the UART would never be shifted out.
> Which instance of the UART are you using? It seems we have:
>
> static struct clk_lookup usart_clocks_lookups[] = {
> 	CLKDEV_CON_DEV_ID("usart", "atmel_usart0", &mck),
> 	CLKDEV_CON_DEV_ID("usart", "atmel_usart1", &usart0_clk),
> 	CLKDEV_CON_DEV_ID("usart", "atmel_usart2", &usart1_clk),
> 	CLKDEV_CON_DEV_ID("usart", "atmel_usart3", &usart2_clk),
> 	CLKDEV_CON_DEV_ID("usart", "atmel_usart4", &usart3_clk),
> };
>
> The first clock seems to be always-on, but the others could be turned off.
>
> Sascha
>


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Peter Kardos]

Hi Sascha,

I may have something. It seems the memory (MMU?) gets "messed" up;

Reading the debug uart registers (v2015.07) gives reasonable results, like
     (gdb) x/32w 0xfffff200
     0xfffff200:    0x00000000    0x00000800    0x00000000 0x00000000
     0xfffff210:    0x00000000    0x40001a1a    0x00000000 0x00000000
     0xfffff220:    0x00000021    0x00000000    0x00000000 0x00000000
     0xfffff230:    0x00000000    0x00000000    0x00000000 0x00000000
     0xfffff240:    0x09290781    0x00000000    0x00000000 0x00000000

However the content from v2016.08 gives

0xFFFFF200    D78D7E1F    F1139ADE    413E0FE5    BBFB6DF2
0xFFFFF210    7D78666E    79CBDEA6    8FB2CB03    BEF6C2B7
0xFFFFF220    C9071D17    FA1EFA2D    C4BCD95E    27D73C7C
0xFFFFF230    727C3437    DFBDEBED    69C45C2A    7F5958F6
0xFFFFF240    834B237E    F8B8A211    1AC74D66    FAE06274

Peter

On 8/18/2016 10:00 AM, Peter Kardos wrote:
> Hi Sascha,
>
>> So you have a jtag debugger attached, does it support showing registers?
>> What is the value of uart->base? Is it correct?
>> Also you could compare the UART registers with the ones from a working
>> U-Boot/barebox.
> Yes, I have, it does. I've checked, and it seems to be correct, 
> pointing to 0xFFFFF200.
>
> However I didn't think to check the register offset and mask (it was 
> late).
>
>> Which instance of the UART are you using? It seems we have:
> I'm using the debug UART (USART0)...
>
>> Does atmel_serial_init_port() do everything correctly?
>> Finally it may be that the clock is disabled for some reason,
> Was thinking the same thing.
>
> I'll check it (compare reg dumps, console init) tonight and report my 
> findings.
>
>
> Cheers, Peter
>
> On 2016-08-18 09:38, Sascha Hauer wrote:
>> Hi Peter,
>>
>> On Thu, Aug 18, 2016 at 03:08:36AM +0200, Peter Kardos wrote:
>>> Greetings,
>>>
>>> I'm looking for a AT91 specialist that could help me debug barebox 
>>> v2016.08.
>>>
>>> I have a custom board running AT91RM9200 that came with a really old 
>>> custom
>>> u-boot (v1.1.4, no ubifs support)...
>>>
>>> The first bootloader I've ported (barebox v2015.07, using 
>>> AT91RM9200ek as
>>> template) almost works; I'm having trouble getting networking (ping, 
>>> dhcp)
>>> working with Micrel KSZ8721 phy. But it boots, works with the Flash, 
>>> etc...
>>> With hopes, that this (networking issue) is resolved with the latest 
>>> barebox
>>> release (using the same patchset and config) I've found (??) a another
>>> issue.
>>>
>>> At first it seemed the board just does not boot, after having a look 
>>> with a
>>> j-link it seems the board "hangs" in atmel_serial_putc().
>>> I've tried to look at the AT91RM9200 init (lowlevel, peripherals, 
>>> driver)
>>> and could not find any significant change (rewrite, define changes, 
>>> etc)...
>> So you have a jtag debugger attached, does it support showing registers?
>> What is the value of uart->base? Is it correct?
>> Also you could compare the UART registers with the ones from a working
>> U-Boot/barebox. Does atmel_serial_init_port() do everything correctly?
>> Finally it may be that the clock is disabled for some reason, then the
>> values written to the UART would never be shifted out.
>> Which instance of the UART are you using? It seems we have:
>>
>> static struct clk_lookup usart_clocks_lookups[] = {
>>     CLKDEV_CON_DEV_ID("usart", "atmel_usart0", &mck),
>>     CLKDEV_CON_DEV_ID("usart", "atmel_usart1", &usart0_clk),
>>     CLKDEV_CON_DEV_ID("usart", "atmel_usart2", &usart1_clk),
>>     CLKDEV_CON_DEV_ID("usart", "atmel_usart3", &usart2_clk),
>>     CLKDEV_CON_DEV_ID("usart", "atmel_usart4", &usart3_clk),
>> };
>>
>> The first clock seems to be always-on, but the others could be turned 
>> off.
>>
>> Sascha
>>
>


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Sascha Hauer]

On Thu, Aug 18, 2016 at 11:15:33PM +0200, Peter Kardos wrote:
> Hi Sascha,
> 
> I may have something. It seems the memory (MMU?) gets "messed" up;
> 
> Reading the debug uart registers (v2015.07) gives reasonable results, like
>     (gdb) x/32w 0xfffff200
>     0xfffff200:    0x00000000    0x00000800    0x00000000 0x00000000
>     0xfffff210:    0x00000000    0x40001a1a    0x00000000 0x00000000
>     0xfffff220:    0x00000021    0x00000000    0x00000000 0x00000000
>     0xfffff230:    0x00000000    0x00000000    0x00000000 0x00000000
>     0xfffff240:    0x09290781    0x00000000    0x00000000 0x00000000
> 
> However the content from v2016.08 gives
> 
> 0xFFFFF200    D78D7E1F    F1139ADE    413E0FE5    BBFB6DF2
> 0xFFFFF210    7D78666E    79CBDEA6    8FB2CB03    BEF6C2B7
> 0xFFFFF220    C9071D17    FA1EFA2D    C4BCD95E    27D73C7C
> 0xFFFFF230    727C3437    DFBDEBED    69C45C2A    7F5958F6
> 0xFFFFF240    834B237E    F8B8A211    1AC74D66    FAE06274

Uh, this indeed seems to be messed up by the MMU, more specifically
during setup of the vector table. Could you try the attached patch?
It's not a solution, but is a clear indication that the bug is in this
area.

Sascha

---------------------------8<--------------------------------

From eb66f09db694a0bc1fc88cde8d86e47faf6debf9 Mon Sep 17 00:00:00 2001
From: Sascha Hauer <s.hauer@pengutronix.de>
Date: Mon, 22 Aug 2016 08:05:38 +0200
Subject: [PATCH] ARM: Disable vector table (tmp)

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/mmu.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/arm/cpu/mmu.c b/arch/arm/cpu/mmu.c
index a31bce4..fc26077 100644
--- a/arch/arm/cpu/mmu.c
+++ b/arch/arm/cpu/mmu.c
@@ -289,6 +289,8 @@ static void create_vector_table(unsigned long adr)
 	u32 *exc;
 	int idx;
 
+	return;
+
 	vectors_sdram = request_sdram_region("vector table", adr, SZ_4K);
 	if (vectors_sdram) {
 		/*
-- 
2.8.1

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Peter Kardos]

During the weekend I've played with this some more...
It seems that the "breakage" was introduced with v2016.05. This version
introduced the "exception vector remapping to 0xFFFF*".
Unfortunately AT91RM9200 has peripherals in this region.

I'll test the patch tonight when i get home.

Peter

On 2016-08-22 08:10, Sascha Hauer wrote:
> On Thu, Aug 18, 2016 at 11:15:33PM +0200, Peter Kardos wrote:
>> Hi Sascha,
>>
>> I may have something. It seems the memory (MMU?) gets "messed" up;
>>
>> Reading the debug uart registers (v2015.07) gives reasonable results, like
>>      (gdb) x/32w 0xfffff200
>>      0xfffff200:    0x00000000    0x00000800    0x00000000 0x00000000
>>      0xfffff210:    0x00000000    0x40001a1a    0x00000000 0x00000000
>>      0xfffff220:    0x00000021    0x00000000    0x00000000 0x00000000
>>      0xfffff230:    0x00000000    0x00000000    0x00000000 0x00000000
>>      0xfffff240:    0x09290781    0x00000000    0x00000000 0x00000000
>>
>> However the content from v2016.08 gives
>>
>> 0xFFFFF200    D78D7E1F    F1139ADE    413E0FE5    BBFB6DF2
>> 0xFFFFF210    7D78666E    79CBDEA6    8FB2CB03    BEF6C2B7
>> 0xFFFFF220    C9071D17    FA1EFA2D    C4BCD95E    27D73C7C
>> 0xFFFFF230    727C3437    DFBDEBED    69C45C2A    7F5958F6
>> 0xFFFFF240    834B237E    F8B8A211    1AC74D66    FAE06274
> Uh, this indeed seems to be messed up by the MMU, more specifically
> during setup of the vector table. Could you try the attached patch?
> It's not a solution, but is a clear indication that the bug is in this
> area.
>
> Sascha
>
> ---------------------------8<--------------------------------
>
>  From eb66f09db694a0bc1fc88cde8d86e47faf6debf9 Mon Sep 17 00:00:00 2001
> From: Sascha Hauer <s.hauer@pengutronix.de>
> Date: Mon, 22 Aug 2016 08:05:38 +0200
> Subject: [PATCH] ARM: Disable vector table (tmp)
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
>   arch/arm/cpu/mmu.c | 2 ++
>   1 file changed, 2 insertions(+)
>
> diff --git a/arch/arm/cpu/mmu.c b/arch/arm/cpu/mmu.c
> index a31bce4..fc26077 100644
> --- a/arch/arm/cpu/mmu.c
> +++ b/arch/arm/cpu/mmu.c
> @@ -289,6 +289,8 @@ static void create_vector_table(unsigned long adr)
>   	u32 *exc;
>   	int idx;
>   
> +	return;
> +
>   	vectors_sdram = request_sdram_region("vector table", adr, SZ_4K);
>   	if (vectors_sdram) {
>   		/*


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Peter Kardos]

Hi Sascha,

I've tested your workaround and it seems to work; the board starts and 
attempts
to boot.

Peter

On 8/22/2016 11:00 AM, Peter Kardos wrote:
> During the weekend I've played with this some more...
> It seems that the "breakage" was introduced with v2016.05. This version
> introduced the "exception vector remapping to 0xFFFF*".
> Unfortunately AT91RM9200 has peripherals in this region.
>
> I'll test the patch tonight when i get home.
>
> Peter
>
> On 2016-08-22 08:10, Sascha Hauer wrote:
>> On Thu, Aug 18, 2016 at 11:15:33PM +0200, Peter Kardos wrote:
>>> Hi Sascha,
>>>
>>> I may have something. It seems the memory (MMU?) gets "messed" up;
>>>
>>> Reading the debug uart registers (v2015.07) gives reasonable 
>>> results, like
>>>      (gdb) x/32w 0xfffff200
>>>      0xfffff200:    0x00000000    0x00000800    0x00000000 0x00000000
>>>      0xfffff210:    0x00000000    0x40001a1a    0x00000000 0x00000000
>>>      0xfffff220:    0x00000021    0x00000000    0x00000000 0x00000000
>>>      0xfffff230:    0x00000000    0x00000000    0x00000000 0x00000000
>>>      0xfffff240:    0x09290781    0x00000000    0x00000000 0x00000000
>>>
>>> However the content from v2016.08 gives
>>>
>>> 0xFFFFF200    D78D7E1F    F1139ADE    413E0FE5    BBFB6DF2
>>> 0xFFFFF210    7D78666E    79CBDEA6    8FB2CB03    BEF6C2B7
>>> 0xFFFFF220    C9071D17    FA1EFA2D    C4BCD95E    27D73C7C
>>> 0xFFFFF230    727C3437    DFBDEBED    69C45C2A    7F5958F6
>>> 0xFFFFF240    834B237E    F8B8A211    1AC74D66    FAE06274
>> Uh, this indeed seems to be messed up by the MMU, more specifically
>> during setup of the vector table. Could you try the attached patch?
>> It's not a solution, but is a clear indication that the bug is in this
>> area.
>>
>> Sascha
>>
>> ---------------------------8<--------------------------------
>>
>>  From eb66f09db694a0bc1fc88cde8d86e47faf6debf9 Mon Sep 17 00:00:00 2001
>> From: Sascha Hauer <s.hauer@pengutronix.de>
>> Date: Mon, 22 Aug 2016 08:05:38 +0200
>> Subject: [PATCH] ARM: Disable vector table (tmp)
>>
>> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
>> ---
>>   arch/arm/cpu/mmu.c | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/arch/arm/cpu/mmu.c b/arch/arm/cpu/mmu.c
>> index a31bce4..fc26077 100644
>> --- a/arch/arm/cpu/mmu.c
>> +++ b/arch/arm/cpu/mmu.c
>> @@ -289,6 +289,8 @@ static void create_vector_table(unsigned long adr)
>>       u32 *exc;
>>       int idx;
>>   +    return;
>> +
>>       vectors_sdram = request_sdram_region("vector table", adr, SZ_4K);
>>       if (vectors_sdram) {
>>           /*
>


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Sascha Hauer]

Hi Peter,

On Mon, Aug 22, 2016 at 10:39:24PM +0200, Peter Kardos wrote:
> Hi Sascha,
> 
> I've tested your workaround and it seems to work; the board starts and
> attempts
> to boot.

I just sent a proper solution for that problem, you're on Cc. I'm
confident that this solves your problem. Could you give this a test
please?

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: AT91RM9200 hang in atmel_serial_putc

[Peter Kardos]

Hi Sascha,

Can confirm, with v2016.08 + patch the system boots and seems to work as 
expected.

Thanks for your help with this issue.

Cheers, Peter


On 8/24/2016 2:24 PM, Sascha Hauer wrote:
> Hi Peter,
>
> On Mon, Aug 22, 2016 at 10:39:24PM +0200, Peter Kardos wrote:
>> Hi Sascha,
>>
>> I've tested your workaround and it seems to work; the board starts and
>> attempts
>> to boot.
> I just sent a proper solution for that problem, you're on Cc. I'm
> confident that this solves your problem. Could you give this a test
> please?
>
> Sascha
>


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox