[IA64] enable dumps to capture second page of kernel stack
authorCliff Wickman <cpw@sgi.com>
Tue, 25 Apr 2006 15:47:48 +0000 (10:47 -0500)
committerTony Luck <tony.luck@intel.com>
Thu, 27 Apr 2006 21:31:16 +0000 (14:31 -0700)
In SLES10 (2.6.16) crash dumping (in my experience, LKCD) is unable to
capture the second page of the 2-page task/stack allocation.
This is particularly troublesome for dump analysis, as the stack traceback
cannot be done.
  (A similar convention is probably needed throughout the kernel to make
   kernel multi-page allocations detectable for dumping)

Multi-page kernel allocations are represented by the single page structure
associated with the first page of the allocation.  The page structures
associated with the other pages are unintialized.

If the dumper is selecting only kernel pages it has no way to identify
any but the first page of the allocation.

The fix is to make the task/stack allocation a compound page.

Signed-off-by: Cliff Wickman <cpw@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
include/asm-ia64/thread_info.h

index 56394a2c7055daed97e57b1fb3966e2acfd19622..e5392c4d30c69fd73bfefb47da912a36212ad53b 100644 (file)
@@ -67,7 +67,7 @@ struct thread_info {
 #define end_of_stack(p) (unsigned long *)((void *)(p) + IA64_RBS_OFFSET)
 
 #define __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
-#define alloc_task_struct()    ((task_t *)__get_free_pages(GFP_KERNEL, KERNEL_STACK_SIZE_ORDER))
+#define alloc_task_struct()    ((task_t *)__get_free_pages(GFP_KERNEL | __GFP_COMP, KERNEL_STACK_SIZE_ORDER))
 #define free_task_struct(tsk)  free_pages((unsigned long) (tsk), KERNEL_STACK_SIZE_ORDER)
 
 #endif /* !__ASSEMBLY */