[patch] mips-work-around-R10k-ll-sc-errata.patch

0001-mips-work-around-R10k-ll-sc-errata.patch (text/plain), 4.10 KB, created by Matt Turner on 2011-06-28 15:40:13 UTC

From c804165aae6e4055a7a8c42cf66ea0b1a29ef3ed Mon Sep 17 00:00:00 2001
From: Matt Turner <[e-mail]>
Date: Tue, 28 Jun 2011 11:30:53 -0400
Subject: [PATCH] mips: work-around R10k ll/sc errata

The R10000 needs an LL/SC Workaround. If not applied all R10k before
rev 3.0 misbehave on atomic ops and rev 2.6 and lower (e.g. SGI IP28)
die after very few seconds with a deadlock due to even more erratas.
---
 ChangeLog.mips             |    4 ++++
 sysdeps/mips/bits/atomic.h |   38 ++++++++++++++++++++++++++++++++------
 2 files changed, 36 insertions(+), 6 deletions(-)

diff --git a/sysdeps/mips/bits/atomic.h b/sysdeps/mips/bits/atomic.h
index 11b3467..2bd723d 100644
--- a/sysdeps/mips/bits/atomic.h
+++ b/sysdeps/mips/bits/atomic.h
@@ -49,6 +49,32 @@ typedef uintmax_t uatomic_max_t;
 # define MIPS_SYNC	sync
 #endif
 
+/* Certain revisions of the R10000 Processor need an LL/SC Workaround
+   enabled.  Revisions before 3.0 misbehave on atomic operations, and
+   Revs 2.6 and lower deadlock after several seconds due to other errata.
+
+   To quote the R10K Errata:
+      Workaround: The basic idea is to inhibit the four instructions
+      from simultaneously becoming active in R10000. Padding all
+      ll/sc sequences with nops or changing the looping branch in the
+      routines to a branch likely (which is always predicted taken
+      by R10000) will work. The nops should go after the loop, and the
+      number of them should be 28. This number could be decremented for
+      each additional instruction in the ll/sc loop such as the lock
+      modifier(s) between the ll and sc, the looping branch and its
+      delay slot. For typical short routines with one ll/sc loop, any
+      instructions after the loop could also count as a decrement. The
+      nop workaround pollutes the cache more but would be a few cycles
+      faster if all the code is in the cache and the looping branch
+      is predicted not taken.  */
+
+
+#ifdef _MIPS_ARCH_R10000
+#define R10K_BEQZ_INSN "beqzl"
+#else
+#define R10K_BEQZ_INSN "beqz"
+#endif
+
 #define MIPS_SYNC_STR_2(X) #X
 #define MIPS_SYNC_STR_1(X) MIPS_SYNC_STR_2(X)
 #define MIPS_SYNC_STR MIPS_SYNC_STR_1(MIPS_SYNC)
@@ -74,7 +100,7 @@ typedef uintmax_t uatomic_max_t;
      "bne	%0,%3,2f\n\t"						      \
      "move	%1,%4\n\t"						      \
      "sc	%1,%2\n\t"						      \
-     "beqz	%1,1b\n"						      \
+     R10K_BEQZ_INSN"	%1,1b\n"					      \
      acq	"\n\t"							      \
      ".set	pop\n"							      \
      "2:\n\t"								      \
@@ -98,7 +124,7 @@ typedef uintmax_t uatomic_max_t;
      "bne	%0,%3,2f\n\t"						      \
      "move	%1,%4\n\t"						      \
      "scd	%1,%2\n\t"						      \
-     "beqz	%1,1b\n"						      \
+     R10K_BEQZ_INSN"	%1,1b\n"					      \
      acq	"\n\t"							      \
      ".set	pop\n"							      \
      "2:\n\t"								      \
@@ -192,7 +218,7 @@ typedef uintmax_t uatomic_max_t;
      "ll	%0,%4\n\t"						      \
      "move	%1,%3\n\t"						      \
      "sc	%1,%2\n\t"						      \
-     "beqz	%1,1b\n"						      \
+     R10K_BEQZ_INSN"	%1,1b\n"					      \
      acq	"\n\t"							      \
      ".set	pop\n"							      \
      "2:\n\t"								      \
@@ -216,7 +242,7 @@ typedef uintmax_t uatomic_max_t;
      "lld	%0,%4\n\t"						      \
      "move	%1,%3\n\t"						      \
      "scd	%1,%2\n\t"						      \
-     "beqz	%1,1b\n"						      \
+     R10K_BEQZ_INSN"	%1,1b\n"					      \
      acq	"\n\t"							      \
      ".set	pop\n"							      \
      "2:\n\t"								      \
@@ -251,7 +277,7 @@ typedef uintmax_t uatomic_max_t;
      "ll	%0,%4\n\t"						      \
      "addu	%1,%0,%3\n\t"						      \
      "sc	%1,%2\n\t"						      \
-     "beqz	%1,1b\n"						      \
+     R10K_BEQZ_INSN"	%1,1b\n"					      \
      acq	"\n\t"							      \
      ".set	pop\n"							      \
      "2:\n\t"								      \
@@ -275,7 +301,7 @@ typedef uintmax_t uatomic_max_t;
      "lld	%0,%4\n\t"						      \
      "daddu	%1,%0,%3\n\t"						      \
      "scd	%1,%2\n\t"						      \
-     "beqz	%1,1b\n"						      \
+     R10K_BEQZ_INSN"	%1,1b\n"					      \
      acq	"\n\t"							      \
      ".set	pop\n"							      \
      "2:\n\t"								      \
-- 
1.7.3.4