Re: clang compiled kernel panic when mounting zfs root on i386

On Tue, Nov 27, 2012 at 08:21:05AM +1100, Bruce Evans wrote:
> On Mon, 26 Nov 2012, Konstantin Belousov wrote:
> 
> > On Mon, Nov 26, 2012 at 06:31:34AM -0800, sig6247 wrote:
> >>
> >> Just checked out r243529, this only happens when the kernel is compiled
> >> by clang, and only on i386, either recompiling the kernel with gcc or
> >> booting from a UFS root works fine. Is it a known problem?
> > It looks like that clang uses more stack than gcc, and zfs makes quite
> > deep call chains.
...
> It would be useful if the stack trace printed the the stack pointer
> on every function call, so that you could see how much stack each
> function used.

Please apply the patch below and obtain the backtrace of the double fault
panic again. I will commit the patch later.

diff --git a/sys/amd64/amd64/db_trace.c b/sys/amd64/amd64/db_trace.c
index cba90f2..2c81f87 100644
--- a/sys/amd64/amd64/db_trace.c
+++ b/sys/amd64/amd64/db_trace.c
_at__at_ -186,7 +186,8 _at__at_ db_ss(struct db_variable *vp, db_expr_t *valuep, int op)
 
 static void db_nextframe(struct amd64_frame **, db_addr_t *, struct thread *);
 static int db_numargs(struct amd64_frame *);
-static void db_print_stack_entry(const char *, int, char **, long *, db_addr_t);
+static void db_print_stack_entry(const char *, int, char **, long *, db_addr_t,
+    void *);
 static void decode_syscall(int, struct thread *);
 
 static const char * watchtype_str(int type);
_at__at_ -230,12 +231,13 _at__at_ db_numargs(fp)
 }
 
 static void
-db_print_stack_entry(name, narg, argnp, argp, callpc)
+db_print_stack_entry(name, narg, argnp, argp, callpc, frame)
 	const char *name;
 	int narg;
 	char **argnp;
 	long *argp;
 	db_addr_t callpc;
+	void *frame;
 {
 	db_printf("%s(", name);
 #if 0
_at__at_ -250,6 +252,8 _at__at_ db_print_stack_entry(name, narg, argnp, argp, callpc)
 #endif
 	db_printf(") at ");
 	db_printsym(callpc, DB_STGY_PROC);
+	if (frame != NULL)
+		db_printf("/frame 0x%lx", (register_t)frame);
 	db_printf("\n");
 }
 
_at__at_ -341,7 +345,7 _at__at_ db_nextframe(struct amd64_frame **fp, db_addr_t *ip, struct thread *td)
 		return;
 	}
 
-	db_print_stack_entry(name, 0, 0, 0, rip);
+	db_print_stack_entry(name, 0, 0, 0, rip, &(*fp)->f_frame);
 
 	/*
 	 * Point to base of trapframe which is just above the
_at__at_ -437,7 +441,8 _at__at_ db_backtrace(struct thread *td, struct trapframe *tf,
 				 * Don't try to walk back on a stack for a
 				 * process that hasn't actually been run yet.
 				 */
-				db_print_stack_entry(name, 0, 0, 0, pc);
+				db_print_stack_entry(name, 0, 0, 0, pc,
+				    actframe);
 				break;
 			}
 			first = FALSE;
_at__at_ -451,7 +456,7 _at__at_ db_backtrace(struct thread *td, struct trapframe *tf,
 			narg = db_numargs(frame);
 		}
 
-		db_print_stack_entry(name, narg, argnp, argp, pc);
+		db_print_stack_entry(name, narg, argnp, argp, pc, actframe);
 
 		if (actframe != frame) {
 			/* `frame' belongs to caller. */
_at__at_ -465,7 +470,7 _at__at_ db_backtrace(struct thread *td, struct trapframe *tf,
 		if (INKERNEL((long)pc) && !INKERNEL((long)frame)) {
 			sym = db_search_symbol(pc, DB_STGY_ANY, &offset);
 			db_symbol_values(sym, &name, NULL);
-			db_print_stack_entry(name, 0, 0, 0, pc);
+			db_print_stack_entry(name, 0, 0, 0, pc, frame);
 			break;
 		}
 		if (!INKERNEL((long) frame)) {
diff --git a/sys/i386/i386/db_trace.c b/sys/i386/i386/db_trace.c
index 445d9c5..822cc56 100644
--- a/sys/i386/i386/db_trace.c
+++ b/sys/i386/i386/db_trace.c
_at__at_ -176,7 +176,8 _at__at_ db_ss(struct db_variable *vp, db_expr_t *valuep, int op)
 
 static void db_nextframe(struct i386_frame **, db_addr_t *, struct thread *);
 static int db_numargs(struct i386_frame *);
-static void db_print_stack_entry(const char *, int, char **, int *, db_addr_t);
+static void db_print_stack_entry(const char *, int, char **, int *, db_addr_t,
+    void *);
 static void decode_syscall(int, struct thread *);
 
 static const char * watchtype_str(int type);
_at__at_ -220,12 +221,13 _at__at_ retry:
 }
 
 static void
-db_print_stack_entry(name, narg, argnp, argp, callpc)
+db_print_stack_entry(name, narg, argnp, argp, callpc, frame)
 	const char *name;
 	int narg;
 	char **argnp;
 	int *argp;
 	db_addr_t callpc;
+	void *frame;
 {
 	int n = narg >= 0 ? narg : 5;
 
_at__at_ -242,6 +244,8 _at__at_ db_print_stack_entry(name, narg, argnp, argp, callpc)
 		db_printf(",...");
 	db_printf(") at ");
 	db_printsym(callpc, DB_STGY_PROC);
+	if (frame != NULL)
+		db_printf("/frame 0x%r", (register_t)frame);
 	db_printf("\n");
 }
 
_at__at_ -326,7 +330,7 _at__at_ db_nextframe(struct i386_frame **fp, db_addr_t *ip, struct thread *td)
 		return;
 	}
 
-	db_print_stack_entry(name, 0, 0, 0, eip);
+	db_print_stack_entry(name, 0, 0, 0, eip, &(*fp)->f_frame);
 
 	/*
 	 * For a double fault, we have to snag the values from the
_at__at_ -467,7 +471,8 _at__at_ db_backtrace(struct thread *td, struct trapframe *tf, struct i386_frame *frame,
 				 * Don't try to walk back on a stack for a
 				 * process that hasn't actually been run yet.
 				 */
-				db_print_stack_entry(name, 0, 0, 0, pc);
+				db_print_stack_entry(name, 0, 0, 0, pc,
+				    actframe);
 				break;
 			}
 			first = FALSE;
_at__at_ -481,7 +486,7 _at__at_ db_backtrace(struct thread *td, struct trapframe *tf, struct i386_frame *frame,
 			narg = db_numargs(frame);
 		}
 
-		db_print_stack_entry(name, narg, argnp, argp, pc);
+		db_print_stack_entry(name, narg, argnp, argp, pc, actframe);
 
 		if (actframe != frame) {
 			/* `frame' belongs to caller. */
_at__at_ -495,7 +500,7 _at__at_ db_backtrace(struct thread *td, struct trapframe *tf, struct i386_frame *frame,
 		if (INKERNEL((int)pc) && !INKERNEL((int) frame)) {
 			sym = db_search_symbol(pc, DB_STGY_ANY, &offset);
 			db_symbol_values(sym, &name, NULL);
-			db_print_stack_entry(name, 0, 0, 0, pc);
+			db_print_stack_entry(name, 0, 0, 0, pc, frame);
 			break;
 		}
 		if (!INKERNEL((int) frame)) {