On 2016-08-29 08:10, Clemens Gruber wrote:
When using gzwrite to eMMC on an i.MX6Q board, the following warning occurs repeatedly: CACHE: Misaligned operation at range [4fd63318, 4fe63318]
This patch cache-aligns the memory allocation for the gzwrite writebuf, therefore avoiding the misaligned dcache flush and the warning from check_cache_range.
Signed-off-by: Clemens Gruber clemens.gruber@pqgruber.com
lib/gunzip.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/lib/gunzip.c b/lib/gunzip.c index bc746d6..832b306 100644 --- a/lib/gunzip.c +++ b/lib/gunzip.c @@ -11,6 +11,7 @@ #include <console.h> #include <image.h> #include <malloc.h> +#include <memalign.h> #include <u-boot/zlib.h> #include <div64.h>
@@ -193,7 +194,7 @@ int gzwrite(unsigned char *src, int len,
s.next_in = src + i; s.avail_in = payload_size+8;
- writebuf = (unsigned char *)malloc(szwritebuf);
writebuf = (unsigned char *)malloc_cache_aligned(szwritebuf);
/* decompress until deflate stream ends or end of file */ do {
I wondered about the length of the buffer, but szwritebuf is forced to be a multiple of the device block size, which is typically quite a bit bigger than a cache line, so I think we are fine there.
Reviewed-by: Stefan Agner stefan.agner@toradex.com
-- Stefan