kernel BUG while reading from SPI into static buffer

kernel BUG while reading from SPI into static buffer

[Felix Varghese]

Hi,

I am trying to read some data via SPI on an a modified (custom SPI device
attached) AT91SAM9G20-EK board. If I pass a local array buffer, declared as
"char buffer[100];" as the rx_buf pointer for the spi transfer, the code
works fine. But if I just change the declaration to "static char
buffer[100]" instead, I get the following crash:

kernel BUG at arch/arm/mm/dma-mapping.c:425!
Unable to handle kernel NULL pointer dereference at virtual address 00000000
pgd = c0004000
[00000000] *pgd=00000000
Internal error: Oops: 817 [#1]
last sysfs file:
Modules linked in: testmod [last unloaded: testmod]
CPU: 0    Not tainted  (2.6.37.2 #8)
PC is at __bug+0x1c/0x28
LR is at __bug+0x18/0x28
pc : [<c0031d88>]    lr : [<c0031d84>]    psr: 20000093
sp : c3a6fea8  ip : 00001e7b  fp : 00000000
r10: ffffffff  r9 : 00000000  r8 : c38ca6a0
r7 : c395e4e8  r6 : c3a6ff28  r5 : 00000032  r4 : c3a6ff54
r3 : 00000000  r2 : 00000001  r1 : 60000093  r0 : 00000033
Flags: nzCv  IRQs off  FIQs on  Mode SVC_32  ISA ARM  Segment kernel
Control: 0005317f  Table: 23a3c000  DAC: 00000017
Process tx_thread (pid: 460, stack limit = 0xc3a6e270)
Stack: (0xc3a6fea8 to 0xc3a70000)
fea0:                   c3a6ff24 c0034344 c0274848 c3a6ff54 bf00c7f8
c0189dd4
fec0: c395e4e8 60000013 c3a6ff28 c395e400 c395e400 00000000 00000000
c0188bf0
fee0: c3a6ff2c 60000013 c3a6ff28 c395e400 c38ca6a0 c0188c54 c3a6ff04
c0188e38
ff00: 00000000 00000000 c3a6ff08 c3a6ff08 c3a6ff94 c3a6ff70 c3a6ff28
bf00c7f8
ff20: 00000032 bf00a0d4 c3a6ff94 c3a6ff70 c38ca6a0 00000000 c0188fdc
c3a6ff04
ff40: 00000000 ffffff8d 00000000 00000000 00000000 00000000 bf00c7f8
00000032
ff60: ffffffff ffffffff 00000000 00000000 c3a6ff28 c3a6ff94 c3a6ff9f
c3a6ff9f
ff80: 00000001 23a6ff9f 23a6ff9f 00000000 00000000 c3a6ff70 c3a6ff28
20a6ff50
ffa0: c3a6ffd4 c39bfdc4 c38ca6a0 bf00a0e0 00000000 bf00a168 720a0000
01a6cafe
ffc0: c3a6ffd4 c00556e0 c002f884 00000000 c38ca6a0 00000000 c3a6ffd8
c3a6ffd8
ffe0: 00000000 c39bfdc4 c0055660 c002f884 00000013 c002f884 0001151c
000000d0
[<c0031d88>] (__bug+0x1c/0x28) from [<c0034344>]
(___dma_single_cpu_to_dev+0x3c/0x68)
[<c0034344>] (___dma_single_cpu_to_dev+0x3c/0x68) from [<c0189dd4>]
(atmel_spi_transfer+0xf8/0x1cc)
[<c0189dd4>] (atmel_spi_transfer+0xf8/0x1cc) from [<c0188bf0>]
(__spi_async+0xa0/0xb0)
[<c0188bf0>] (__spi_async+0xa0/0xb0) from [<c0188c54>]
(spi_async_locked+0x14/0x2c)
[<c0188c54>] (spi_async_locked+0x14/0x2c) from [<c0188e38>]
(__spi_sync+0x60/0xa0)
[<c0188e38>] (__spi_sync+0x60/0xa0) from [<bf00a0d4>] (read_bytes+0xac/0xb8
[testmod])
[<bf00a0d4>] (read_bytes+0xac/0xb8 [testmod]) from [<bf00a168>]
(tx_thread+0x88/0x110 [testmod])
[<bf00a168>] (tx_thread+0x88/0x110 [testmod]) from [<c00556e0>]
(kthread+0x80/0x88)
[<c00556e0>] (kthread+0x80/0x88) from [<c002f884>]
(kernel_thread_exit+0x0/0x8)
Code: e1a01000 e59f000c eb090854 e3a03000 (e5833000)
---[ end trace 0dac538caa941b38 ]---

Does this behaviour make sense to anyone? Is this a bug or is it just me
doing the wrong thing?

Regards,
Felix.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20110513/7015febe/attachment.html 

kernel BUG while reading from SPI into static buffer

[Mulyadi Santosa]

Hi....

On Fri, May 13, 2011 at 14:03, Felix Varghese <felixv1986@gmail.com> wrote:
> Hi,
>
> I am trying to read some data via SPI on an a modified (custom SPI device
> attached) AT91SAM9G20-EK board. If I pass a local array buffer, declared as
> "char buffer[100];" as the rx_buf pointer for the spi transfer, the code
> works fine. But if I just change the declaration to "static char
> buffer[100]" instead, I get the following crash:
>
> kernel BUG at arch/arm/mm/dma-mapping.c:425!
> Unable to handle kernel NULL pointer dereference at virtual address 00000000


Hmm, I think this is the corresponding line:
BUG_ON(!virt_addr_valid(kaddr) || !virt_addr_valid(kaddr + size - 1)    );

alright...so it roughly says that the address (buffer) falls into
invalid virt address. And, virt_addr_valid is:
static inline int virt_addr_valid(const volatile void *kaddr)
{
        extern void *high_memory;  /* copied from <linux/mm.h> */
       return ((unsigned long)kaddr >= PAGE_OFFSET && kaddr < high_memory);
}

OK, a bit clear now.... so, I can say base on the above code snippet,
buffer...when declared as static, somehow falls into non
ZONE_NORMAL......i.e non directly mapped memory area...

Feel free to try to observer whether this statement is correct or
not... and you should check the address of "buffer" when declared as
static I guess....

-- 
regards,

Mulyadi Santosa
Freelance Linux trainer and consultant

blog: the-hydra.blogspot.com
training: mulyaditraining.blogspot.com

kernel BUG while reading from SPI into static buffer

[Mulyadi Santosa]

Need to add...

On Fri, May 13, 2011 at 21:32, Mulyadi Santosa
<mulyadi.santosa@gmail.com> wrote:
> OK, a bit clear now.... so, I can say base on the above code snippet,
> buffer...when declared as static, somehow falls into non
> ZONE_NORMAL......i.e non directly mapped memory area...

what I precisely mean, it falls into non DMA able zone...

-- 
regards,

Mulyadi Santosa
Freelance Linux trainer and consultant

blog: the-hydra.blogspot.com
training: mulyaditraining.blogspot.com

kernel BUG while reading from SPI into static buffer

[Greg KH]

On Fri, May 13, 2011 at 12:33:15PM +0530, Felix Varghese wrote:
> Hi,
> 
> I am trying to read some data via SPI on an a modified (custom SPI device
> attached) AT91SAM9G20-EK board. If I pass a local array buffer, declared as
> "char buffer[100];" as the rx_buf pointer for the spi transfer, the code works
> fine. But if I just change the declaration to "static char buffer[100]"
> instead, I get the following crash:

Don't statically allocate memory for spi, you need to dynamically
allocate it with 'kmalloc'.

The fact that the first time didn't crash for you was just lucky.

hope this helps,

greg k-h

kernel BUG while reading from SPI into static buffer

[Felix Varghese]

On 13 May 2011 20:02, Mulyadi Santosa <mulyadi.santosa@gmail.com> wrote:
> Feel free to try to observer whether this statement is correct or
> not... and you should check the address of "buffer" when declared as
> static I guess....

I did, and here's the result:

Local Buffer - 0xc3a61edd
Static Buffer - 0xbf03eac1

So they indeed seem to be way apart in memory :).

On 13 May 2011 20:38, Greg KH <greg@kroah.com> wrote:
> Don't statically allocate memory for spi, you need to dynamically
> allocate it with 'kmalloc'.
>
> The fact that the first time didn't crash for you was just lucky.

I'm guessing you say this because static variables are not in DMA
capable memory. But how exactly do we figure out which part of memory
is DMA-capable and which isn't? Also, is this a restriction imposed by
the kernel or by the hardware?

Thanks & Regards,
Felix.

kernel BUG while reading from SPI into static buffer

[Greg KH]

On Mon, May 16, 2011 at 11:25:32AM +0530, Felix Varghese wrote:
> On 13 May 2011 20:38, Greg KH <greg@kroah.com> wrote:
> > Don't statically allocate memory for spi, you need to dynamically
> > allocate it with 'kmalloc'.
> >
> > The fact that the first time didn't crash for you was just lucky.
> 
> I'm guessing you say this because static variables are not in DMA
> capable memory. But how exactly do we figure out which part of memory
> is DMA-capable and which isn't?

If you create the memory with a call to kmalloc() it is dma-capable.  So
always do that.

> Also, is this a restriction imposed by the kernel or by the hardware?

The hardware.

USB also has this rule.

thanks,

greg k-h