Hi,

Dave Anderson wrote:
 > Xavier also mentioned in a private email that "foreach bt" shows
 > nothing for all non-active tasks, so it appears that something has
 > changed affecting the ia64 unwind facility.  Has anybody else
 > running ia64 on 2.6.14 kernels seen this?

While 'bt' just works on my 2.6.15 vanilla kernel (16KB-page),
I noticed Xavier used 64KB-page kernel but crash assumed
stacksize was 32KB.

 > xb wrote:
 >>          pagesize: 65536
...
 >>         stacksize: 32768

Can this be a cause of the problem?
In kernel, the stack size is 64KB under 64KB-page.
 

Anyway, it sure looks like it could be the issue here.

Xavier, please try something like this and let us know:

--- ia64.c.orig 2006-01-10 14:07:07.000000000 -0500
+++ ia64.c      2006-01-10 14:07:31.000000000 -0500
@@ -92,6 +92,9 @@ ia64_init(int when)
                case 16384:
                        machdep->stacksize = (power(2, 1) * PAGESIZE());
                        break;
+               case 65536:
+                       machdep->stacksize = (power(2, 0) * PAGESIZE());
+                       break;
                default:
                        machdep->stacksize = 32*1024;
                        break;

Thanks again, Jun-ichi,
  Dave
 

 
In crash/ia64.c:ia64_init(),
..
       case 4096:
               machdep->stacksize = (power(2, 3) * PAGESIZE());
               break;
       case 8192:
               machdep->stacksize = (power(2, 2) * PAGESIZE());
               break;
       case 16384:
               machdep->stacksize = (power(2, 1) * PAGESIZE());
               break;
       default:
               machdep->stacksize = 32*1024;
               break;
..

In include/asm-ia64/ptrace.h of kernel,
   #if defined(CONFIG_IA64_PAGE_SIZE_4KB)
   # define KERNEL_STACK_SIZE_ORDER                3
   #elif defined(CONFIG_IA64_PAGE_SIZE_8KB)
   # define KERNEL_STACK_SIZE_ORDER                2
   #elif defined(CONFIG_IA64_PAGE_SIZE_16KB)
   # define KERNEL_STACK_SIZE_ORDER                1
   #else
   # define KERNEL_STACK_SIZE_ORDER                0
   #endif

Thanks,
Jun'ichi "Nick" Nomura

--
Crash-utility mailing list
Crash-utility@redhat.com
https://www.redhat.com/mailman/listinfo/crash-utility