kexec and kdump in imx6ull with kernel 4.9.11

[arthur_rythm]

Hi,

I tried to work with kexec and kdump on kernel rel_imx_4.1.15_2.1.0_ga without success:

https://community.nxp.com/message/982634?commentID=982634#comment-982634 

As sugested by FabioEstevam‌ on this comment. I start a new topic and run the test with 4.9.11 release instead.

Problem is that with this kernel I am not able to init crashkernel anymore :

# dmesg | grep -i crash
crashkernel reservation failed - memory is in use.

I was using the same kernel parameter than before : crashkernel=50M@0x81000000

A message on mailing list kernel is talking about CMA and crashkernel : Proposal for run time memory allocation for crash kernel . Its seems that we have to use memory block in CMA block ? I have a CMA in my kernel command (cma=96M). Message from kernel which show the cma :

# dmesg | grep -i cma
Reserved memory: created CMA memory pool at 0x8a000000, size 96 MiB

So I tried to init crashkernel with : crashkernel=50M@0x8a000000 but the result is the same.

Unfortunately I dont have anymore an imx6ull or imx6ul evk to run the test, my evk are broken. I will order a new one today. I will try to test kdump on evk with official release as soon as possible.

Thanks,

Arthur.

[fabio_estevam]

Could you try with kernel 4.16 and see if it works there?

[arthur_rythm]

I am able to load the kexec kernel correctly after the kernel panic now.

I need to check now if I am able to read the dump.

I am using the mainline kernel v4.16-rc7 +  imx_v6_v7 defconfig. it reminds me this nice topic : 

https://community.nxp.com/thread/385542 

I will try to align defconfig and perhaps kernel/kexec_core to find the issue ?

[Carlos_Musich]

Hi Arthur,

I'm not quite sure what the problem is with crashkernel on your system. Seems to work fine on my 4.1.15 kernel on an i.MX 6QP SDB.

root@imx6qpsabresd:~# dmesg | egrep -i crash
[    0.000000] Reserving 50MB of memory at 960MB for crashkernel (System RAM: 1024MB)

However, it doesn't seem to matter what I put in the cma= kernel parameter. It always sets it to 320 MiB at 0x3c000000. Below the size is ignored, but it seems to ignore addresses too.

root@imx6qpsabresd:~# dmesg | egrep -i cma
[    0.000000] Reserved memory: created CMA memory pool at 0x3c000000, size 320 MiB
[    0.000000] Reserved memory: initialized node linux,cma, compatible id shared-dma-pool
[    0.000000] Kernel command line: console=ttymxc0,115200 root=/dev/mmcblk2p2 rootwait rw video=mxcfb0:dev=hdmi,1920x1080M@60,if=RGB24 psplash=false vt.global_cursor_default=0 cma=96M@0x3c000000 crashkernel=50M@0x3c000000
[    0.000000] Memory: 697660K/1048576K available (8616K kernel code, 526K rwdata, 3144K rodata, 436K init, 458K bss, 23236K reserved, 327680K cma-reserved, 0K highmem)

Looking forward to your EVK results.

Regards,

Carlos

[billchen]

thanks for your reply , I confirm it has good result. I couldn't execute "kexec -p " successfully before that. 

[billchen]

but after I run "echo c >/proc/sysrq-trigger" 

It stopped at 

"Loading crashdump kernel...

Bye! " 

                                  "

[arthur_rythm]

Sorry for the delay I was out of office for few weeks. So I received my imx6ull evk. 

Test result are exactly the same on evk.

For both test :

device tree : imx6ull-14x14-evk

kernel defconfig : imx_v7

uboot git : git://git.freescale.com/imx/uboot-imx.git

uboot release : rel_imx_4.1.15_2.1.0_ga

git kernel : git://git.freescale.com/imx/linux-imx.git

tag release : rel_imx_4.1.15_2.0.0_ga

Result : load kernel with kexec. Simulate kernel panic then freeze after this output :

[<803476f8>] (sysrq_handle_crash) from [<00000000>] ( (null))
Code: e5c32000 e8bd8010 e3a03000 e3a02001 (e5c32000)
Loading crashdump kernel...
Bye

git kernel : git://git.freescale.com/imx/linux-imx.git

tag release : rel_imx_4.9.x_1.0.0_ga

Result : crashkernel init failed :

# dmesg | grep -i crash
crashkernel reservation failed - memory is in use.

[arthur_rythm]

Anyone to test on his own imx6ull evk to confirm my result ? 

[fabio_estevam]

Hi Arthur,

Tested on a 4.9 NXP kernel and it worked fine:

# dmesg | grep crash
Reserving 50MB of memory at 2176MB for crashkernel (System RAM: 512MB)
Kernel command line: console=ttymxc0,115200 root=/dev/mmcblk1p2 rootwait rw crashkernel=50M

Please note that I use 'crashkernel=50M' with no extra parameters.

[arthur_rythm]

Always KO.

Topic on linux arm mailing list :

issue with kexec/kdump on imx6ull - Patchwork 

I was able to fix an issue related to kexec thanks to Russel. ARM support is broken in official kexec repository. But my test always failed even after the fix. I have probably more issue on my evk.

I tried to enable earlyprintk to try to debug kernel hang but no output :

Loading crashdump kernel...
Bye!
Uncompressing Linux... done, booting the kernel.

Kexec is probably not able to find the kernel.