Hi Jack, Everyone,
The existing x86_64 kdb has two blocks of code which adjust the RSP
value based on the assumption that the RSP and SS are not saved.
I don't think this code is ever need. Once the processor is in
"Long-Mode" it always saves the SS:RSP.
I was having problems with stack trace backs when stopped at a
breakpoint at a function entry. Fixing the RSP values made the
stack trace back work for this case.
The attached patch removes this unnecessary stack adjustment.
Jim Houston - Concurrent Computer Corp.
--
--- arch/x86_64/kdb/kdbasupport.c.0 2005-08-22 17:30:22.000000000 -0400
+++ arch/x86_64/kdb/kdbasupport.c 2005-10-04 10:31:05.000000000 -0400
@@ -699,10 +699,6 @@
if (strcmp(regname, "krsp") == 0) {
*contents = (unsigned long)regs + sizeof(struct pt_regs);
- if ((regs->cs & 0xffff) == __KERNEL_CS) {
- /* rsp and ss are not on stack */
- *contents -= 2*4;
- }
return 0;
}
@@ -734,22 +730,6 @@
if ((i < nkdbreglist)
&& (strlen(kdbreglist[i].reg_name) == strlen(regname))) {
- if ((regs->cs & 0xffff) == __KERNEL_CS) {
- /* No cpl switch, rsp is not on stack */
- if (strcmp(kdbreglist[i].reg_name, "rsp") == 0) {
- *contents = (kdb_machreg_t)regs +
- sizeof(struct pt_regs) - 2*8;
- return(0);
- }
- if (strcmp(kdbreglist[i].reg_name, "ss") == 0) {
- kdb_machreg_t r;
-
- r = (kdb_machreg_t)regs +
- sizeof(struct pt_regs) - 2*8;
- *contents = (kdb_machreg_t)SS(r); /* XXX
*/
- return(0);
- }
- }
*contents = *(unsigned long *)((unsigned long)regs +
kdbreglist[i].reg_offset);
return(0);
---------------------------
Use http://oss.sgi.com/ecartis to modify your settings or to unsubscribe.
