Bug 132301

Summary: 2.6.14 does not boot
Product: [openSUSE] SUSE Linux 10.1 Reporter: Hannes Reinecke <hare>
Component: KernelAssignee: Torsten Duwe <duwe>
Status: RESOLVED FIXED QA Contact: E-mail List <qa-bugs>
Severity: Blocker    
Priority: P5 - None CC: stefan.fent, trenn
Version: unspecified   
Target Milestone: ---   
Hardware: x86-64   
OS: Other   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---

Description Hannes Reinecke 2005-11-04 12:04:31 UTC 
The current kernel (2.6.14-3-smp) does not boot on my machine. If fails just after 

Booting the kernel.
kernel direct mapping tables upto ffff810100000000 @ 8000-c000
PANIC: early exception rip ffffffff8023da09 error 0 cr2 0
PANIC: early exception rip ffffffff8011ba7a error 0 cr2 ffffffffff5fd023

-> dead after that.

Machine:
2 x Dual Core AMD Opteron(tm) Processor 860 HE, 2G RAM.
Iwill board. More details to follow.
Comment 1 Hannes Reinecke 2005-11-04 12:06:23 UTC 
So, board is Iwill DK8-HTX.
Comment 2 Hannes Reinecke 2005-11-04 13:47:15 UTC 
Bad news. Intel evaluation mainboard (Dual-core w/ HT) also fails.
This time I get:

kernel direct mapping tables upto ffff81003ff00000 @ 8000 - a0000
PANIC: early exception rip ffffffff8023da09 error 0 cr2 0
PANIC: early exception rip ffffffff8011ba7a error 0 cr2 ffffffffff5fd023

So the panic occurs at exactly the same location.

Probably not a BIOS bug then ...
Comment 3 Hannes Reinecke 2005-11-04 14:34:34 UTC 
Tja. If I were suspicious I'd say it's bloody ACPI again:

arch/x86_64/kernel/setup.c:

	init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT));

#ifdef CONFIG_ACPI
	/*
	 * Initialize the ACPI boot-time table parser (gets the RSDP and SDT).
	 * Call this early for SRAT node setup.
	 */
	acpi_boot_table_init();
#endif

'kernel direct mapping ...' is printed last from init_memory_mapping();
acpi_boot_table_init() should print something like

RSDP @ XXXX

but it doesn't. Curious though, 'acpi=off' doesn't help.
Comment 4 Hannes Reinecke 2005-11-04 14:48:27 UTC 
Compiling vanilla 2.6.14 now to see whether it's a generic problem.
Comment 5 Hannes Reinecke 2005-11-04 15:00:31 UTC 
Cf. e.g.

http://www.uwsg.iu.edu/hypermail/linux/kernel/0510.3/1417.html

Maybe we should change 'smp' configuration to discontigmem?
Comment 6 Andreas Kleen 2005-11-04 15:47:37 UTC 
Oops yes. Where the hell enabled sparsemem?
That option is totally broken and even if it worked it would be a waste of memory right now.
Comment 7 Torsten Duwe 2005-11-04 17:20:30 UTC 
The changelog shows nothing.
Fixed in CVS.