unexport profile_pc

I haven't found any modular usage of profile_pc in the kernel.

Is the patch below correct?

---

 arch/arm/kernel/time.c     |    1 -
 arch/i386/kernel/time.c    |    1 -
 arch/ppc/kernel/time.c     |    1 -
 arch/ppc64/kernel/time.c   |    1 -
 arch/sparc64/kernel/time.c |    1 -
 arch/x86_64/kernel/time.c  |    1 -
 6 files changed, 6 deletions(-)

@@ -209,7 +209,6 @@
 
 	return pc;
 }
-EXPORT_SYMBOL(profile_pc);
 #endif
 
 /*

@@ -86,7 +86,6 @@
 		return regs->u_regs[UREG_RETPC];
 	return pc;
 }
-EXPORT_SYMBOL(profile_pc);
 #endif
 
 static void tick_disable_protection(void)

@@ -69,7 +69,6 @@
 
 	return pc;
 }
-EXPORT_SYMBOL(profile_pc);
 #endif
 
 /*

@@ -119,7 +119,6 @@
 
 	return pc;
 }
-EXPORT_SYMBOL(profile_pc);
 #endif
 
 /*

@@ -206,7 +206,6 @@
 
 	return pc;
 }
-EXPORT_SYMBOL(profile_pc);
 #endif
 
 #ifdef CONFIG_PPC_ISERIES

@@ -199,7 +199,6 @@
 	}
 	return pc;
 }
-EXPORT_SYMBOL(profile_pc);
 
 /*
  * In order to set the CMOS clock precisely, set_rtc_mmss has to be called 500

In theory, /proc/ can be modular. In practice...

-- wli

Oprofile?

Not even in theory.

I have vague recollections of modular /proc/ "fixes" being posted. I don't have any vested (or other) interest in the thing being modular and generally avoid modules altogether myself. fs/Kconfig of course reveals that this "option" has since been removed, to my complete indifference.

-- wli

We don't actually use it, but it looks like maybe we should? It seems unfortunate that readprofile and OProfile should disagree here.

john

We really should be using it otherwise you get profile hits in the spin_lock functions, which isn't really helpful. I recall sending patches for this, i may have to resend.