This patch should fix kgdb to work without highmem_start_page. 
However, I don't have a kgdb setup, and I'd appreciate someone
testing this to make sure my selection of a new variable is OK.

Also, should this function possibly be using system_state, instead?

Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
---

 memhotplug-dave/arch/i386/lib/kgdb_serial.c |    2 +-
 1 files changed, 1 insertion(+), 1 deletion(-)

diff -puN arch/i386/lib/kgdb_serial.c~A1-no-highmem-mess-kgdb arch/i386/lib/kgdb_serial.c
--- memhotplug/arch/i386/lib/kgdb_serial.c~A1-no-highmem-mess-kgdb	2004-11-12 16:51:37.000000000 -0800
+++ memhotplug-dave/arch/i386/lib/kgdb_serial.c	2004-11-12 16:51:38.000000000 -0800
@@ -407,7 +407,7 @@ void shutdown_for_kgdb(struct async_stru
 #ifdef CONFIG_DISCONTIGMEM
 static inline int kgdb_mem_init_done(void)
 {
-	return highmem_start_page != NULL;
+	return initsize != NULL;
 }
 #else
 static inline int kgdb_mem_init_done(void)
_