Skip to contentSkip to conversation

Clean up sys_shutdown

PATCH In review Andi Kleen 2 participants 3 messages Apr 7, 2009 – Apr 10, 2009

lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

PATCH 1/1 Clean up sys_shutdown

Clean up sys_shutdown exit path

Impact: cleanup, fix

Clean up sys_shutdown exit path. Factor out common code. Return correct error code instead of always 0 on failure.

---
 kernel/sys.c |   24 +++++++++---------------
 1 file changed, 9 insertions(+), 15 deletions(-)
kernel/sys.c +9 -15
Index: linux/kernel/sys.c
===================================================================
@@ -360,6 +360,7 @@
 		void __user *, arg)
 {
 	char buffer[256];
+	int ret = 0;
 
 	/* We only trust the superuser with rebooting the system. */
 	if (!capable(CAP_SYS_BOOT))
@@ -397,7 +398,7 @@
 		kernel_halt();
 		unlock_kernel();
 		do_exit(0);
-		break;
+		panic("cannot halt");
 
 	case LINUX_REBOOT_CMD_POWER_OFF:
 		kernel_power_off();
@@ -417,29 +418,22 @@
 
 #ifdef CONFIG_KEXEC
 	case LINUX_REBOOT_CMD_KEXEC:
-		{
-			int ret;
-			ret = kernel_kexec();
-			unlock_kernel();
-			return ret;
-		}
+		ret = kernel_kexec();
+		break;
 #endif
 
 #ifdef CONFIG_HIBERNATION
 	case LINUX_REBOOT_CMD_SW_SUSPEND:
-		{
-			int ret = hibernate();
-			unlock_kernel();
-			return ret;
-		}
+		ret = hibernate();
+		break;
 #endif
 
 	default:
-		unlock_kernel();
-		return -EINVAL;
+		ret = -EINVAL;
+		break;
 	}
 	unlock_kernel();
-	return 0;
+	return ret;
 }
 
 static void deferred_cad(struct work_struct *dummy)
#1 lore raw
Quoted 5 lines · On Wed, 8 Apr 2009 00:36:26 +0200 Andi Kleen <[email protected]> wrote:

Clean up sys_shutdown exit path

Impact: cleanup, fix

Clean up sys_shutdown exit path. Factor out common code.

OK.

Quoted 2 lines

Return
correct error code instead of always 0 on failure.

oh goody, that makes the patch a bugfix.

Quoted 44 lines
kernel/sys.c +6 -12
@@ -360,6 +360,7 @@
 		void __user *, arg)
 {
 	char buffer[256];
+	int ret = 0;
 
 	/* We only trust the superuser with rebooting the system. */
 	if (!capable(CAP_SYS_BOOT))
@@ -397,7 +398,7 @@
 		kernel_halt();
 		unlock_kernel();
 		do_exit(0);
-		break;
+		panic("cannot halt");
 
 	case LINUX_REBOOT_CMD_POWER_OFF:
 		kernel_power_off();
@@ -417,29 +418,22 @@
 
 #ifdef CONFIG_KEXEC
 	case LINUX_REBOOT_CMD_KEXEC:
-		{
-			int ret;
-			ret = kernel_kexec();
-			unlock_kernel();
-			return ret;
-		}
+		ret = kernel_kexec();
+		break;
 #endif
 
 #ifdef CONFIG_HIBERNATION
 	case LINUX_REBOOT_CMD_SW_SUSPEND:
-		{
-			int ret = hibernate();
-			unlock_kernel();
-			return ret;
-		}
+		ret = hibernate();
+		break;
 #endif
 

kernel/sys.c line 437

Quoted 10 lines
kernel/sys.c +3 -3
 	default:
-		unlock_kernel();
-		return -EINVAL;
+		ret = -EINVAL;
+		break;
 	}
 	unlock_kernel();
-	return 0;
+	return ret;
 }

I wonder why this code uses lock_kernel().

author
#2 lore raw
Quoted 1 line · Andrew Morton

I wonder why this code uses lock_kernel().

I suspect because there's no other lock and parallel shutdowns would be bad, but I wouldn't be surprised if there aren't some caller without.

We could just convert it to a new global mutex or something. I think it would be safe on x86, but I'm not sure about all the other architectures.

Actually a simple flag would be enough that returns when one is already in progress (that wouldn't work if parallel ones fail, but that's presumably unlikely)

-Andi

Signature · 1 line
[email protected] -- Speaking for myself only.