cortex-m3 on linux-2.6.33

cortex-m3 on linux-2.6.33

[RK Raggit]

Hello Catalin,

I rebased the baseport of my cortex-m3 board on your latest
linux-2.6-stable git (2.6.33 kernel).
My first observation after booting is that the system appears more
stable now. My earlier baseport
based on 2.6.28 kernel was largely unstable with simple commands
crashing the kernel easily.
My first question -

Q1/ could you please summarise significant changes that you made in
the cortex-m3 architecture
port in the 2.6.33 kernel

Now to an issue i am facing with the latest kernel. When i run command
like "ls -l" or
cat /proc/meminfo in an infinite loop on the shell, after running for
about a minute or so they crash
with a dump like below -

-------------------------------8<8<---------------------------
(ls -l dump)
Unhandled exception: IPSR = 00000003 LR = fffffff1
CPU: 0    Not tainted  (2.6.33-arm1 #6)
PC is at copy_namespaces+0xe/0x8c
LR is@copy_process+0x565/0x904
pc : [<800531d6>]    lr : [<80044fd9>]    psr: 2100000b
sp : 83c20238  ip : 83c42848  fp : 83cf0ac0
r10: 83c44a10  r9 : 00000000  r8 : 00000000
r7 : 83c20248  r6 : 00004111  r5 : 00000010  r4 : 460ad0f9
r3 : 83c77000  r2 : 00000000  r1 : 00000010  r0 : 00004111
--------------------------------------------------------------------
(cat /proc/meminfo dump)
Unhandled exception: IPSR = 00000003 LR = fffffff1
CPU: 0    Not tainted  (2.6.33-arm1 #6)
PC is at do_wait+0xf4/0x124
LR is at __call_rcu+0x31/0xa4
pc : [<800473c8>]    lr : [<8005b841>]    psr: 6100000b
sp : 83c202c8  ip : 0000000c  fp : 00000000
r10: 83c1e108  r9 : 0000018c  r8 : 83c20300
r7 : 83c44c00  r6 : 83c1e000  r5 : 83c20000  r4 : 83c202e8
r3 : 1b478423  r2 : 1b478423  r1 : 00000000  r0 : 0000018c
--------------------------------------------------------------------

I am sure dumps are not due to memory constraint since my system has
ample memory (64MB).

Any thoughts ?

--RK

cortex-m3 on linux-2.6.33

[Catalin Marinas]

On Tue, 2010-07-06 at 16:40 +0100, RK Raggit wrote:
> I rebased the baseport of my cortex-m3 board on your latest
> linux-2.6-stable git (2.6.33 kernel).
> My first observation after booting is that the system appears more
> stable now. My earlier baseport
> based on 2.6.28 kernel was largely unstable with simple commands
> crashing the kernel easily.
> My first question -
> 
> Q1/ could you please summarise significant changes that you made in
> the cortex-m3 architecture
> port in the 2.6.33 kernel

The only changes I have made were for solving conflicts with the new
kernel version. A major difference is that the Thumb-2 support (required
for Cortex-M3) is already in mainline 2.6.33 and it may have got various
fixes.

The new feature in the M-profile support is floating point (Cortex-M4).

> Now to an issue i am facing with the latest kernel. When i run command
> like "ls -l" or
> cat /proc/meminfo in an infinite loop on the shell, after running for
> about a minute or so they crash
> with a dump like below -
> 
> -------------------------------8<8<---------------------------
> (ls -l dump)
> Unhandled exception: IPSR = 00000003 LR = fffffff1
> CPU: 0    Not tainted  (2.6.33-arm1 #6)
> PC is at copy_namespaces+0xe/0x8c
> LR is at copy_process+0x565/0x904
> pc : [<800531d6>]    lr : [<80044fd9>]    psr: 2100000b
> sp : 83c20238  ip : 83c42848  fp : 83cf0ac0
> r10: 83c44a10  r9 : 00000000  r8 : 00000000
> r7 : 83c20248  r6 : 00004111  r5 : 00000010  r4 : 460ad0f9
> r3 : 83c77000  r2 : 00000000  r1 : 00000010  r0 : 00004111

IPSR=3 looks like a HardFault. This usually shows unrecoverable system
failures though the docs aren't very clear on what could trigger this
(or I haven't read them in detail).

I can't tell more without further details. What code is around the pc
address? You can also do an "addr2line -e vmlinux -i 800531d6".

-- 
Catalin

cortex-m3 on linux-2.6.33

[RK Raggit]

> IPSR=3 looks like a HardFault. This usually shows unrecoverable system
> failures though the docs aren't very clear on what could trigger this
> (or I haven't read them in detail).
>
> I can't tell more without further details. What code is around the pc
> address? You can also do an "addr2line -e vmlinux -i 800531d6".
>

I ran my testcase (inifinite loop of 'ls -l') 6 times and this is the addr2line
corresponding to the PC i received from the dump each time -

kernel/exit.c:1650
arch/arm/lib/copy_template.S:100
kernel/sched.c:5125
kernel/sched.c:5125
arch/arm/lib/copy_template.S:173
kernel/sched.c:5125

I also checked the hardfault registers during dump - in most of the cases
in turns out to be a bus fault (precise data bus error - 0xE000ED28 reads
0x8200)

--RK

cortex-m3 on linux-2.6.33

[Robert Schwebel]

On Tue, Jul 06, 2010 at 09:10:58PM +0530, RK Raggit wrote:
> I rebased the baseport of my cortex-m3 board on your latest
> linux-2.6-stable git (2.6.33 kernel).

Is there any off-the-shelf hardware out there which could be used to try
uCLinux on Cortex-M3?

rsc
-- 
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 |

cortex-m3 on linux-2.6.33

[Catalin Marinas]

On Tue, 2010-07-06 at 19:09 +0100, Robert Schwebel wrote:
> On Tue, Jul 06, 2010 at 09:10:58PM +0530, RK Raggit wrote:
> > I rebased the baseport of my cortex-m3 board on your latest
> > linux-2.6-stable git (2.6.33 kernel).
> 
> Is there any off-the-shelf hardware out there which could be used to try
> uCLinux on Cortex-M3?

The one we have in ARM is the Microcontroller Prototyping System and it
has 8MB of RAM:

http://www.arm.com/products/DevTools/mps.html

There may be others but I don't actively work on the Cortex-Mx
processors.

-- 
Catalin