Hit-based icache operations may complete before the CM completes
intervention with the local L1. Thus code which invalidates the icache
and then attempts to execute those addresses must include a barrier to
prevent the scenario which:
- icache instruction completes
- icache fetch occurs
- core executes icache data
- CM completes icache invalidate
If the above were allowed to happen then the core would execute stale
instructions from the icache.
A barrier is required to prevent the core i-fetching before the icache
operation has completed. This goes together with the instruction_hazard
to ensure that the pipeline is stalled until the icache operation is
completed and the core will fetch the new instructions.
Suggested-by: Leonid Yegoshin <Leonid.Yegoshin@mips.com>
Signed-off-by: Matt Redfearn <matt.redfearn@mips.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: James Hogan <james.hogan@mips.com>
Cc: stable <stable@vger.kernel.org> # v4.9+
---
arch/mips/mm/c-r4k.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/mips/mm/c-r4k.c b/arch/mips/mm/c-r4k.c
index b7186d47184b..844685e51109 100644
--- a/arch/mips/mm/c-r4k.c
+++ b/arch/mips/mm/c-r4k.c
@@ -763,6 +763,8 @@ static inline void __local_r4k_flush_icache_range(unsigned
long start,
break;
}
}
+ /* Ensure icache operation has completed */
+ mb();
/* Hazard to force new i-fetch */
instruction_hazard();
}
--
2.7.4
|