It seems there is a thinko in the TLB invalidation code that makes the
tlbie in the loop executed just once. The intended check was probably
'gt', not 'lt'.
Signed-off-by: Anton Vorontsov <avoront...@ru.mvista.com>
---
Btw, 'lis r4,0x1000' seems needlessly too high. e300 core reference
manual says:
The [...]tlbie[...] instruction invalidates the TLB entry indexed by
the EA, and operates on both the instruction and data TLBs
simultaneously invalidating four TLB entries (both sets in each TLB).
The index corresponds to bits 15-19 of the EA. To invalidate all
entries within both TLBs, 32 tlbie instructions should be issued,
incrementing this field by one each time.
So, shouldn't 0x1000 be 0x0002? Are there any classic PowerPCs
for which this wouldn't work?
arch/powerpc/kernel/swsusp_32.S | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/arch/powerpc/kernel/swsusp_32.S b/arch/powerpc/kernel/swsusp_32.S
index b47d8ce..b0754e2 100644
--- a/arch/powerpc/kernel/swsusp_32.S
+++ b/arch/powerpc/kernel/swsusp_32.S
@@ -303,7 +303,7 @@ END_MMU_FTR_SECTION_IFSET(MMU_FTR_USE_HIGH_BATS)
lis r4,0x1000
1: addic. r4,r4,-0x1000
tlbie r4
- blt 1b
+ bgt 1b
sync
/* restore the MSR and turn on the MMU */
--
1.6.5.7
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
