issue:trying to load the second kernel to capture the crash dump when panic ic occurred
performed following pre conditions in menuconfig-
CONFIG_KEXEC=y
CONFIG_SYSFS=y
CONFIG_DEBUG_INFO=Y
CONFIG_CRASH_DUMP=y
CONFIG_PROC_VMCORE=y
CONFIG_PROC_DEVICETREE=y
also allocated the memory for crashkernel :crashkernel=16M@900M
Using kexec-tool 2.0.6 cross compiled for arm platform.
Executed below command to load/boot the second kernel on panic/crash but we are getting "unrecoverable error:" not enough data for mem property

kexec -p uImage_m14a0 --command-line="$(cat /proc/cmdline)"
error:
Unable to find /proc/device-tree//chosen/linux,stdout-path, printing from purgatory is disabled
"unrecoverable error:" not enough data for mem property

please suggest any idea to resolve this issue.

Thanks
Manju

Hi Manju,

I am trying to load the second kernel to capture the crash dump when panic occurs.
I am trying this on ARMHF platform with ubuntu.

kexec -p is giving me below error :

sudo kexec -p -d uImage
kernel: 0x36b96008 kernel_size: 0x2cbf00
phys_offset: 0
get_crash_notes_per_cpu: crash_notes addr = 2e9db088, size = 180
Elf header: p_type = 4, p_offset = 0x2e9db088 p_paddr = 0x2e9db088 p_vaddr = 0x0 p_filesz = 0xb4 p_memsz = 0xb4
get_crash_notes_per_cpu: crash_notes addr = 2e9e3088, size = 180
Elf header: p_type = 4, p_offset = 0x2e9e3088 p_paddr = 0x2e9e3088 p_vaddr = 0x0 p_filesz = 0xb4 p_memsz = 0xb4
vmcoreinfo header: p_type = 4, p_offset = 0x5ac8d8 p_paddr = 0x5ac8d8 p_vaddr = 0x0 p_filesz = 0x1024 p_memsz = 0x1024
Elf header: p_type = 1, p_offset = 0x0 p_paddr = 0x0 p_vaddr = 0xc0000000 p_filesz = 0x0 p_memsz = 0x0
Elf header: p_type = 1, p_offset = 0x8000000 p_paddr = 0x8000000 p_vaddr = 0xc8000000 p_filesz = 0x28000000 p_memsz = 0x28000000
elfcorehdr: 0x7f00000
crashkernel: [0 - 0x7ffffff] (128M)
memory range: [0 - 0xffffffffffffffff] (0M)
memory range: [0x8000000 - 0x2fffffff] (640M)
kernel command line: " elfcorehdr=0x7f00000 mem=130048K"
Modified cmdline: elfcorehdr=0x7f00000 mem=130048K root=/dev/mmcblk0p2
unrecoverable error: not enough data for mem property


How did you resolve your issue?
I think it is similar to mine.

Please let me know if you got the resolution.

Thanks,
Samiksha

Hi
I tried to operate kdump on ubunto with success.
kdump-config show return ready for kdump.

but in arm imx6 i get the
unrecoverable error: not enough data for mem property

Note that i did managed to run kexec -e for a loaded kernel with kexec -l vmlinux --cmdline...
kdump-config load failed with that error.

Did someone find why?
Thanks!
Eli

Hi Eli,
I checked the kexec-tools code and it was
failing to create flattened devicetree. I gave my own device tree as argument in /etc/default/kdump-tools.config and passed
--dtb argument. It worked for me after these changes. I am successfully able to get the kdump now on ARM.

Regards,
Samiksha

1 members found this post helpful.

Hi Samiksha

from the source of kexec , the problem is here:
if (len < sizeof(buf))
{
printf("len = %d\n" , len);
printf("size of buf = %d\n" , sizeof(buf));
die("unrecoverable error: not enough data for mem property\n");
}

And why it happened?
because the device tree file /proc/device-tree/memory/reg is actually 8 bytes and not 16 bytes
the size of buf is 2 x 64 bit = 16 bytes
so why the device-tree/memory/reg in the imx arm linux is just 8 bytes ( 4 for offset and 4 for size) and not 8 for offset
and 8 for size.

any case , i compiled the sources , modified the code to load the correct address
i got kdump ready for dump
but it is not recording the memory when i trigger sysrq panic.

I guess a kexec for arm , that read the /proc/device-tree/memory/reg device tree should exists somewhere.

Any help will be great!
Thanks,
Eli