Skip to contentSkip to conversation

Problems with 2.6.23-rc6 on AMD Geode LX800

DISCUSSION Joerg Pommnitz 3 participants 3 messages Sep 20, 2007

lore
Share & tools
lore
#0 lore raw

Hello all,
yesterday I tried to boot a kernel built from the current wireless-dev git tree (ath5k branch) on a MSEP800/A board (see http://www.milesie.co.uk/pdf/MSEP800.pdf). The board contains an AMD Geode LX800 CPU.
The wireless-dev tree is up to date with Linus kernel 2.6.23-rc6.

Attached is a photographic screen shot. The EIP value of c0378dd6 seems to correspond with the reserve_bootmem_core from System.map:

c0378d51 t free_bootmem_core
c0378da7 T free_bootmem
c0378db2 T free_bootmem_node
c0378dba t reserve_bootmem_core
c0378e14 T reserve_bootmem
c0378e1f T reserve_bootmem_node

Additionally I'm sending the .config file used to compile the kernel.

Additional information:
* The last known good kernel on the MSEP800/A is 2.6.20.14
* The exact same kernel binary boots fine (well, until it tries to mount /) on a standard desktop PC with a 2.5GHz Celeron CPU.

Am I doing something stupid or is Linux currently broken on this CPU?

Thanks in advance and kind regards
Joerg

      __________________________________  

Alles was der Gesundheit und Entspannung dient. BE A BETTER MEDIZINMANN! www.yahoo.de/clever

#1 lore raw
Quoted 9 lines · Joerg Pommnitz

Hello all,
yesterday I tried to boot a kernel built from the current wireless-dev git tree (ath5k branch) on a MSEP800/A board (see http://www.milesie.co.uk/pdf/MSEP800.pdf). The board contains an AMD Geode LX800 CPU.
The wireless-dev tree is up to date with Linus kernel 2.6.23-rc6.

Attached is a photographic screen shot. The EIP value of c0378dd6 seems to correspond with the reserve_bootmem_core from System.map:

Quoted 7 lines

c0378d51 t free_bootmem_core
c0378da7 T free_bootmem
c0378db2 T free_bootmem_node
c0378dba t reserve_bootmem_core
c0378e14 T reserve_bootmem
c0378e1f T reserve_bootmem_node

Can you post disassembled code for that function?

#2 lore raw
Quoted 21 lines · Chuck Ebbert
Quoted 12 lines · Joerg Pommnitz

Hello all,
yesterday I tried to boot a kernel built from the current wireless-dev git tree (ath5k branch)
on a MSEP800/A board (see http://www.milesie.co.uk/pdf/MSEP800.pdf). The board
contains an AMD Geode LX800 CPU.
The wireless-dev tree is up to date with Linus kernel 2.6.23-rc6.

Attached is a photographic screen shot. The EIP value of c0378dd6 seems to correspond with the
reserve_bootmem_core from System.map:

Quoted 7 lines

c0378d51 t free_bootmem_core
c0378da7 T free_bootmem
c0378db2 T free_bootmem_node
c0378dba t reserve_bootmem_core
c0378e14 T reserve_bootmem
c0378e1f T reserve_bootmem_node

Quoted 1 line

Can you post disassembled code for that function?

Its hitting a bug - specifically (from bootmem.c:125):
BUG_ON(PFN_DOWN(addr) >= bdata->node_low_pfn);

I hit this problem on a db800 last week. It went away with a newer version of the BIOS, which doesn't help Joerg any, since its a different board (though I think it is the same BIOS vendor). Other BIOSes work just fine with the same kernel image (including known troublemakers like LinuxBIOS). I believe that 2.6.22 was good, so some change must have come along in 2.6.23-pre to cause the pain. Or, it may have exposed old breakage in the BIOS that was later repaired.

I'll do the math to figure out whats happening - and I'll check the release notes to see what changed in the BIOS between the failing and working version. If anybody familiar with arch/i386 can think of something new in the kernel that may have precipitated this, do let me know. :)

Jordan

Signature · 3 lines
Jordan Crouse
Systems Software Development Engineer 
Advanced Micro Devices, Inc.