Re: 2.5.73 -- Uninitialised timer! (i386)

[Mikael Pettersson]

Pete Clements writes:
 > Quoting Andrew Morton
 >   > 
 >   > Well it beats me.  That timer is clearly initialised OK.
 >   > 
 >   > What compiler version are you using?  Can you try
 >   > a different one>?
 > 
 > Reading specs from /usr/lib/gcc-lib/i386-linux/2.95.4/specs
 > gcc version 2.95.4 20011002 (Debian prerelease)
 > 
 > Only compiler currently installed.  Have four systems (3 single
 > processor, 1 dual) all running Debian--woody with same versions.
 > On the UP systems, will see several of these traces during boot.
 > After that, it is very seldom (0 to 3 in 12 hours). Have seen none
 > on the SMP system.  Recompiled one of the UP systems with SMP
 > enabled and no longer saw the trace during boot and post. Have
 > not seen this prior to 2.5.73.

Apply the patch below (which I posted to LKML yesterday btw).
2.5.73 incorrectly removed the workaround needed to prevent
gcc-2.95.x from miscompiling spinlocks on UP (they become
empty structs, and gcc-2.95.x has problems with those).

/Mikael

--- linux-2.5.73/include/linux/spinlock.h.~1~	2003-06-23 13:07:39.000000000 +0200
+++ linux-2.5.73/include/linux/spinlock.h	2003-06-23 22:58:29.000000000 +0200
@@ -146,8 +146,13 @@
 /*
  * gcc versions before ~2.95 have a nasty bug with empty initializers.
  */
-typedef struct { } spinlock_t;
-#define SPIN_LOCK_UNLOCKED (spinlock_t) { }
+#if (__GNUC__ > 2)
+  typedef struct { } spinlock_t;
+  #define SPIN_LOCK_UNLOCKED (spinlock_t) { }
+#else
+  typedef struct { int gcc_is_buggy; } spinlock_t;
+  #define SPIN_LOCK_UNLOCKED (spinlock_t) { 0 }
+#endif
 
 /*
  * If CONFIG_SMP is unset, declare the _raw_* definitions as nops