summaryrefslogtreecommitdiff
path: root/packages/glibc/glibc-2.2.5/linuxthreads-2.2.5-ppc405erratum77.patch
blob: 48127a12814a07f284ee3857520d0c2d7002aac6 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
# see http://bugs.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&pr=4155
# and http://www.kegel.com/xgcc3/ppc405erratum77.html
# See also matching patch for glibc

diff -aur glibc-2.2.5.orig/linuxthreads/sysdeps/powerpc/pt-machine.h glibc-2.2.5/linuxthreads/sysdeps/powerpc/pt-machine.h
--- glibc-2.2.5.orig/linuxthreads/sysdeps/powerpc/pt-machine.h	Thu May 17 12:47:46 2001
+++ glibc-2.2.5/linuxthreads/sysdeps/powerpc/pt-machine.h	Tue Jul 23 05:38:24 2002
@@ -41,6 +41,17 @@
 #define HAS_COMPARE_AND_SWAP_WITH_RELEASE_SEMANTICS
 #define IMPLEMENT_TAS_WITH_CAS
 
+#ifdef __PPC405__
+/* workaround for PPC405 erratum #77 - Mark Hatle, dank.  References:
+   http://www-3.ibm.com/chips/techlib/techlib.nsf/techdocs/89DED00DEBFF54BF87256A8000491BA2/$file/405CR_C_errata_1_2.pdf
+   http://ppc.bkbits.net:8080/linuxppc_2_4_devel/cset@1.489
+   http://www.kegel.com/xgcc3/ppc405erratum77.html
+   FIXME: using dbct instead of sync would be faster  */
+#define __LINUXTHREADS_PPC405_ERR77_SYNC   "sync \n\t"
+#else
+#define __LINUXTHREADS_PPC405_ERR77_SYNC
+#endif
+
 PT_EI int
 __compare_and_swap (long int *p, long int oldval, long int newval)
 {
@@ -50,6 +61,7 @@
 	   "0:    lwarx %0,0,%1 ;"
 	   "      xor. %0,%3,%0;"
 	   "      bne 1f;"
+	   __LINUXTHREADS_PPC405_ERR77_SYNC
 	   "      stwcx. %2,0,%1;"
 	   "      bne- 0b;"
 	   "1:    "