kexec on APQ8074

* kexec on APQ8074
@ 2014-09-11  9:46 Noé RUBINSTEIN
  2014-09-11  9:47 ` Noé RUBINSTEIN
  2014-09-12  2:02 ` Wang, Yalin
  0 siblings, 2 replies; 6+ messages in thread
From: Noé RUBINSTEIN @ 2014-09-11  9:46 UTC (permalink / raw)
  To: linux-arm-msm

Hi,

Has anyone successfully used kexec on MSM8974/APQ8074? Or are there some
outstanding issues that have to be fixed before it is possible?

I’m trying to use Linux with kexec as a second stage bootloader on my
APQ8074-based board.  I fail while loading the second kernel, with no error
message whatsoever (no panic, and I do not have access to JTAG to know what is
going on).

The kernel I use is based upon branch jb_mr1_rb1.47 of
git://codeaurora.org/quic/la/kernel/msm (tag M8974AAAAANLYA31050138)
with these commits cherry-picked from mainline for kexec support:

    2456f44 ARM: 7555/1: kexec: fix segment memory addresses check
    c564df4 ARM: 7540/1: kexec: Check segment memory addresses
    4cabd1d ARM: 7539/1: kexec: scan for dtb magic in segments

The commands I run look like this:

    /bin/kexec \
        --debug \
        --load /system/boot/Image \
        --dtb=/system/boot/dtb \
        --append="console=ttyHSL0,115200,n8 maxcpus=1 msm_rtb.filter=0x37 \
                  earlyprintk debug bootmem_debug"
    /bin/kexec -e

This goes well, but the kernel apparently stops working early-on. With
bootmem_debug and printascii-based tracing, I narrowed the freeze down inside
free_all_bootmem -> free_all_bootmem_core -> __free_pages_bootmem ->
__free_pages -> __free_pages_ok -> free_pages_prepare -> kernel_map_pages
-> poison_pages -> memset) when poisoning a page with physical adress
0x33262000.

If I disable HIGHMEM, I get the freeze a little later, in cma_create_area; the
last line that gets printed to the serial is

    [    0.614040] cma_create_area(base 00032400, count 1000)

To me, it looks like I am somehow stepping on some memory that should be left
alone, but I am not really sure where to go from here. I have a boot log
attached which shows the serial output.

Can anyone tell what might be going on?

Thanks in advance,
Noé.

^ permalink raw reply	[flat|nested] 6+ messages in thread