Skip to contentSkip to conversation

Suppress switch warnings while debugging

PATCH Superseded Jan Kiszka 2 participants 6 messages Mar 10, 2009

lore
Share & tools
lore

Jan Kiszka

author
#0 lore raw

PATCH 1/1 Suppress switch warnings while debugging

Don't raise SIGXCPU while the process is being debugged. These mode changes are expected, and reporting them doesn't provide any helpful information to the application. Rather, it may raise error storms on the application side.

---

 ksrc/nucleus/shadow.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
ksrc/nucleus/shadow.c +2 -1
index bcf3b8b..91cf499 100644
@@ -1082,7 +1082,8 @@ void xnshadow_relax(int notify)
 
 	xnstat_counter_inc(&thread->stat.ssw);	/* Account for secondary mode switch. */
 
-	if (notify && xnthread_test_state(thread, XNTRAPSW))
+	if (notify && xnthread_test_state(thread, XNTRAPSW) &&
+	    !xnthread_test_state(thread, XNDEBUG))
 		/* Help debugging spurious relaxes. */
 		send_sig(SIGXCPU, current, 1);
 
#1 lore raw
Quoted 16 lines · Jan Kiszka

Don't raise SIGXCPU while the process is being debugged. These mode changes are expected, and reporting them doesn't provide any helpful information to the application. Rather, it may raise error storms on the application side.

---

 ksrc/nucleus/shadow.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
ksrc/nucleus/shadow.c +0 -0
index bcf3b8b..91cf499 100644

ksrc/nucleus/shadow.c void xnshadow_relax(int notify) · line 1089

Quoted 9 lines
ksrc/nucleus/shadow.c +2 -1
@@ -1082,7 +1082,8 @@ void xnshadow_relax(int notify)
 
 	xnstat_counter_inc(&thread->stat.ssw);	/* Account for secondary mode switch. */
 
-	if (notify && xnthread_test_state(thread, XNTRAPSW))
+	if (notify && xnthread_test_state(thread, XNTRAPSW) &&
+	    !xnthread_test_state(thread, XNDEBUG))
 		/* Help debugging spurious relaxes. */
 		send_sig(SIGXCPU, current, 1);
 

I would rather identify the source of the switch and clear the notify flag appropriately from the relax call site.

Quoted 6 lines

_______________________________________________
Xenomai-core mailing list
[email protected]
https://mail.gna.org/listinfo/xenomai-core

Signature · 1 line
Philippe.
author
#2 lore raw
Quoted 21 lines · Philippe Gerum
Quoted 20 lines · Jan Kiszka

Don't raise SIGXCPU while the process is being debugged. These mode changes are expected, and reporting them doesn't provide any helpful information to the application. Rather, it may raise error storms on the application side.

---

 ksrc/nucleus/shadow.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
ksrc/nucleus/shadow.c +0 -0
index bcf3b8b..91cf499 100644
@@ -1082,7 +1082,8 @@ void xnshadow_relax(int notify)
 
     xnstat_counter_inc(&thread->stat.ssw);    /* Account for

secondary mode switch. */

Quoted 9 lines

ksrc/nucleus/shadow.c void xnshadow_relax(int notify) · line 1088

Quoted 6 lines
ksrc/nucleus/shadow.c +2 -1
-    if (notify && xnthread_test_state(thread, XNTRAPSW))
+    if (notify && xnthread_test_state(thread, XNTRAPSW) &&
+        !xnthread_test_state(thread, XNDEBUG))
         /* Help debugging spurious relaxes. */
         send_sig(SIGXCPU, current, 1);

I would rather identify the source of the switch and clear the notify flag appropriately from the relax call site.

Sorry, don't get this: What flag do you want to clear? XNTRAPSW?

Jan

Signature · 2 lines
Siemens AG, Corporate Technology, CT SE 2
Corporate Competence Center Embedded Linux
#3 lore raw
Quoted 31 lines · Jan Kiszka
Quoted 21 lines · Philippe Gerum
Quoted 20 lines · Jan Kiszka

Don't raise SIGXCPU while the process is being debugged. These mode changes are expected, and reporting them doesn't provide any helpful information to the application. Rather, it may raise error storms on the application side.

---

 ksrc/nucleus/shadow.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
ksrc/nucleus/shadow.c +0 -0
index bcf3b8b..91cf499 100644
@@ -1082,7 +1082,8 @@ void xnshadow_relax(int notify)
 
     xnstat_counter_inc(&thread->stat.ssw);    /* Account for

secondary mode switch. */

Quoted 8 lines

ksrc/nucleus/shadow.c void xnshadow_relax(int notify) · line 1088

Quoted 6 lines
ksrc/nucleus/shadow.c +2 -1
-    if (notify && xnthread_test_state(thread, XNTRAPSW))
+    if (notify && xnthread_test_state(thread, XNTRAPSW) &&
+        !xnthread_test_state(thread, XNDEBUG))
         /* Help debugging spurious relaxes. */
         send_sig(SIGXCPU, current, 1);

I would rather identify the source of the switch and clear the notify flag appropriately from the relax call site.

Quoted 2 lines

Sorry, don't get this: What flag do you want to clear? XNTRAPSW?

The "notify" parameter should set by the call site to tell xnshadow_relax() whether we want the SIGXCPU signal to be sent or not. I would rather identify which call site introduces the issue, and fix the notify parameter accordingly. Did you track the issue down to request_syscall_restart() for instance? In which case, we should rather test XNDEBUG there, and set the notify flag passed to xnshadow_relax() appropriately.

Quoted 2 lines

Jan

Signature · 1 line
Philippe.
author
#4 lore raw
Quoted 36 lines · Philippe Gerum
Quoted 32 lines · Jan Kiszka
Quoted 22 lines · Philippe Gerum
Quoted 21 lines · Jan Kiszka

Don't raise SIGXCPU while the process is being debugged. These mode changes are expected, and reporting them doesn't provide any helpful information to the application. Rather, it may raise error storms on the
application side.

---

 ksrc/nucleus/shadow.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
ksrc/nucleus/shadow.c +0 -0
index bcf3b8b..91cf499 100644
@@ -1082,7 +1082,8 @@ void xnshadow_relax(int notify)
 
     xnstat_counter_inc(&thread->stat.ssw);    /* Account for

secondary mode switch. */

Quoted 8 lines

ksrc/nucleus/shadow.c void xnshadow_relax(int notify) · line 1088

Quoted 6 lines
ksrc/nucleus/shadow.c +2 -1
-    if (notify && xnthread_test_state(thread, XNTRAPSW))
+    if (notify && xnthread_test_state(thread, XNTRAPSW) &&
+        !xnthread_test_state(thread, XNDEBUG))
         /* Help debugging spurious relaxes. */
         send_sig(SIGXCPU, current, 1);

I would rather identify the source of the switch and clear the notify flag appropriately from the relax call site.

Quoted 2 lines

Sorry, don't get this: What flag do you want to clear? XNTRAPSW?

Quoted 7 lines

The "notify" parameter should set by the call site to tell xnshadow_relax() whether we want the SIGXCPU signal to be sent or not. I would rather identify which call site introduces the issue, and fix the notify parameter accordingly. Did you track the issue down to request_syscall_restart() for instance? In which case, we should rather test XNDEBUG there, and set the notify flag passed to xnshadow_relax() appropriately.

OK, now I understand. I didn't track this down yet, but it must be request_syscall_restart, because that should be the only site that relaxes due to pending signals. Will update the patch.

Jan

Signature · 2 lines
Siemens AG, Corporate Technology, CT SE 2
Corporate Competence Center Embedded Linux
author
#5 lore raw

PATCH v2 1/1 Suppress switch warnings while debugging

In reply to Jan Kiszka

Don't raise SIGXCPU while the process is being debugged. These mode changes are expected, and reporting them doesn't provide any helpful information to the application. Rather, it may raise error storms on the application side.

---

 ksrc/nucleus/shadow.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
ksrc/nucleus/shadow.c +1 -1
index bcf3b8b..12b4073 100644
@@ -667,7 +667,7 @@ static inline void request_syscall_restart(xnthread_t *thread,
 			__xn_error_return(regs,
 					  (sysflags & __xn_exec_norestart) ?
 					  -ERESTARTNOHAND : -ERESTARTSYS);
-			notify = 1;
+			notify = !xnthread_test_state(thread, XNDEBUG);
 		}
 
 		xnthread_clear_info(thread, XNKICKED);