[RFC PATCH] ARM: Fix the bug in dcache flush all API.

Will Deacon will.deacon at arm.com
Fri Dec 16 06:26:40 EST 2011

On Fri, Dec 16, 2011 at 10:57:07AM +0000, Will Deacon wrote:
> I'll take another look at your patch.

So on a second look, I'm failing to see the problem you're describing. The
thing that I reckon needs changing is this:

diff --git a/arch/arm/mm/cache-v7.S b/arch/arm/mm/cache-v7.S
index 07c4bc8..5ca3503 100644
--- a/arch/arm/mm/cache-v7.S
+++ b/arch/arm/mm/cache-v7.S
@@ -58,7 +58,7 @@ loop1:
        isb                                     @ isb to sych the new cssr&csidr
        mrc     p15, 1, r1, c0, c0, 0           @ read the new csidr
        and     r2, r1, #7                      @ extract the length of the cache lines
-       add     r2, r2, #4                      @ add 4 (line length offset)
+       add     r2, r2, #2                      @ add 2 (line length offset)
        ldr     r4, =0x3ff
        ands    r4, r4, r1, lsr #3              @ find maximum number on the way size
        clz     r5, r4                          @ find bit position of way size increment

but this function is crazily hard to read and is taken verbatim from the
ARM ARM, so I'll try and find the original author and ask them about it.

If you're convinced it's broken, please could you describe the problem in
more depth? (your original commit message says we use bits [25:23] for loc,
which isn't true - we use [26:24] but only shift it down to bit 1, which
we've masked to 0).


More information about the linux-arm-kernel mailing list