Kernel crash on mem= argument

Kernel crash on mem= argument

[Clint Thomas]

So, I am able to get my system to autoboot with a ramdisk out of flash
if I leave the mem=3D argument out of my u-boot bootargs, however since =
I
need to define the range of real memory for my kernel to address,
leaving that argument out is not an option for me.=20

When I add it in, this is the error I get.

NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
RAMDISK: Couldn't find valid RAM disk image starting at 0.
Freeing initrd memory: 3938k freed
Oops: kernel access of bad area, sig: 11
NIP: C002EB90 XER: 20000000 LR: C0030120 SP: C0419DB0 REGS: c0419d00
TRAP: 0800    Not tainted
MSR: 00009000 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 00
DEAR: 8048486C, ESR: 00800000
TASK =3D c0418000[1] 'swapper' Last syscall: 6
last math 00000000 last altivec 00000000 last spe 00000000
GPR00: 00000000 C0419DB0 C0418000 C0584820 FFFFFFFF 00000001 C058483C
00000080
GPR08: C0584838 80484868 C0291000 C0584808 00000000 00000000 10001000
00000000
GPR16: 00000001 00000001 FFFFFFFF 007FFF00 00009002 C0419EA0 00000000
C0003DEC
GPR24: C0003B40 007FFEB0 00000000 C0240000 C0240000 CFFAEB6B C0584820
CFBD7000
Call backtrace:
00000000 C0030120 C000D94C C00F86E0 C003A388 C0038D24 C0003B9C
C020019C C0001E98 C0001ACC C00064B8
Kernel panic: Attempted to kill init!
 <0>Rebooting in 180 seconds..
=20
Suddenly, the kernel goes from finding, mounting and loading it's
ramdisk RFS, to not being able to find it at all. I have not moved the
ramdisk in flash, so i'm unsure why adding a mem=3D argument on the =
kernel
bootargs would do this. Any ideas?


Clinton Thomas

Re: Kernel crash on mem= argument

[Frank D Lombardo]

Clint Thomas wrote:
> So, I am able to get my system to autoboot with a ramdisk out of flash
> if I leave the mem= argument out of my u-boot bootargs, however since I
> need to define the range of real memory for my kernel to address,
> leaving that argument out is not an option for me. 
>
> When I add it in, this is the error I get.
>
> NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
> RAMDISK: Couldn't find valid RAM disk image starting at 0.
> Freeing initrd memory: 3938k freed
> Oops: kernel access of bad area, sig: 11
> NIP: C002EB90 XER: 20000000 LR: C0030120 SP: C0419DB0 REGS: c0419d00
> TRAP: 0800    Not tainted
> MSR: 00009000 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 00
> DEAR: 8048486C, ESR: 00800000
> TASK = c0418000[1] 'swapper' Last syscall: 6
> last math 00000000 last altivec 00000000 last spe 00000000
> GPR00: 00000000 C0419DB0 C0418000 C0584820 FFFFFFFF 00000001 C058483C
> 00000080
> GPR08: C0584838 80484868 C0291000 C0584808 00000000 00000000 10001000
> 00000000
> GPR16: 00000001 00000001 FFFFFFFF 007FFF00 00009002 C0419EA0 00000000
> C0003DEC
> GPR24: C0003B40 007FFEB0 00000000 C0240000 C0240000 CFFAEB6B C0584820
> CFBD7000
> Call backtrace:
> 00000000 C0030120 C000D94C C00F86E0 C003A388 C0038D24 C0003B9C
> C020019C C0001E98 C0001ACC C00064B8
> Kernel panic: Attempted to kill init!
>  <0>Rebooting in 180 seconds..
>  
> Suddenly, the kernel goes from finding, mounting and loading it's
> ramdisk RFS, to not being able to find it at all. I have not moved the
> ramdisk in flash, so i'm unsure why adding a mem= argument on the kernel
> bootargs would do this. Any ideas?
>
>
> Clinton Thomas
> _______________________________________________
> Linuxppc-embedded mailing list
> Linuxppc-embedded@ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-embedded
>   
Make sure that u-boot does not load the initrd in the space above the 
limit set with the mem variable.  See the u-boot README description of 
CONFIG_PRAM.  That should help.

Frank

RE: Kernel crash on mem= argument

[Clint Thomas]

My u-boot config does not currently have a CONFIG_PRAM variable defined.
Is doing this as easy as dropping in a #define in my board config? Or do
I need to address this in another part of the code? Up until now, I just
assumed that if I passed mem=3D200M to the kernel, it would allocate as
necessary, then I could use the remaining RAM as I saw fit, but since
this is not working I need to deal with it appropriately.

Clint

-----Original Message-----
From: Frank D Lombardo [mailto:lombardo@mdivac.com]=20
Sent: Thursday, April 19, 2007 2:39 PM
To: Clint Thomas
Cc: linuxppc-embedded@ozlabs.org
Subject: Re: Kernel crash on mem=3D argument

Clint Thomas wrote:
> So, I am able to get my system to autoboot with a ramdisk out of flash

> if I leave the mem=3D argument out of my u-boot bootargs, however =
since=20
> I need to define the range of real memory for my kernel to address,=20
> leaving that argument out is not an option for me.
>
> When I add it in, this is the error I get.
>
> NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
> RAMDISK: Couldn't find valid RAM disk image starting at 0.
> Freeing initrd memory: 3938k freed
> Oops: kernel access of bad area, sig: 11
> NIP: C002EB90 XER: 20000000 LR: C0030120 SP: C0419DB0 REGS: c0419d00
> TRAP: 0800    Not tainted
> MSR: 00009000 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 00
> DEAR: 8048486C, ESR: 00800000
> TASK =3D c0418000[1] 'swapper' Last syscall: 6 last math 00000000 last =

> altivec 00000000 last spe 00000000
> GPR00: 00000000 C0419DB0 C0418000 C0584820 FFFFFFFF 00000001 C058483C=20
> 00000080
> GPR08: C0584838 80484868 C0291000 C0584808 00000000 00000000 10001000=20
> 00000000
> GPR16: 00000001 00000001 FFFFFFFF 007FFF00 00009002 C0419EA0 00000000=20
> C0003DEC
> GPR24: C0003B40 007FFEB0 00000000 C0240000 C0240000 CFFAEB6B C0584820=20
> CFBD7000 Call backtrace:
> 00000000 C0030120 C000D94C C00F86E0 C003A388 C0038D24 C0003B9C=20
> C020019C C0001E98 C0001ACC C00064B8 Kernel panic: Attempted to kill=20
> init!
>  <0>Rebooting in 180 seconds..
> =20
> Suddenly, the kernel goes from finding, mounting and loading it's=20
> ramdisk RFS, to not being able to find it at all. I have not moved the

> ramdisk in flash, so i'm unsure why adding a mem=3D argument on the=20
> kernel bootargs would do this. Any ideas?
>
>
> Clinton Thomas
> _______________________________________________
> Linuxppc-embedded mailing list
> Linuxppc-embedded@ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-embedded
>  =20
Make sure that u-boot does not load the initrd in the space above the
limit set with the mem variable.  See the u-boot README description of
CONFIG_PRAM.  That should help.

Frank

Re: Kernel crash on mem= argument

[Frank D Lombardo]

Clint Thomas wrote:
> My u-boot config does not currently have a CONFIG_PRAM variable defined.
> Is doing this as easy as dropping in a #define in my board config? Or do
> I need to address this in another part of the code? Up until now, I just
> assumed that if I passed mem=200M to the kernel, it would allocate as
> necessary, then I could use the remaining RAM as I saw fit, but since
> this is not working I need to deal with it appropriately.
>
> Clint
>
> -----Original Message-----
> From: Frank D Lombardo [mailto:lombardo@mdivac.com] 
> Sent: Thursday, April 19, 2007 2:39 PM
> To: Clint Thomas
> Cc: linuxppc-embedded@ozlabs.org
> Subject: Re: Kernel crash on mem= argument
>
> Clint Thomas wrote:
>   
Clint,

I believe whats happening is that u-boot is loading the initrd in the 
region of RAM above 200M.  Then you pass mem=200M to the kernel, and it 
can't access the initrd.  By adding the appropriate CONFIG_PRAM define 
to your board config file, you let u-boot know not to use the memory 
above 200M either.  That way u-boot will load the initrd in memory that 
is usable by the kernel.

Frank