linux-kernel - [PATCH] media: venus: helpers: Fix ALIGN() of non power of two

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Date:   Sat, 12 Sep 2020 21:03:01 +0200
From:   Rikard Falkeborn <rikard.falkeborn@...il.com>
To:     Aniket Masule <amasule@...eaurora.org>,
        Stanimir Varbanov <stanimir.varbanov@...aro.org>
Cc:     Bjorn Andersson <bjorn.andersson@...aro.org>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        linux-media@...r.kernel.org, linux-arm-msm@...r.kernel.org,
        linux-kernel@...r.kernel.org, Andy Gross <agross@...nel.org>,
        Rikard Falkeborn <rikard.falkeborn@...il.com>
Subject: [PATCH] media: venus: helpers: Fix ALIGN() of non power of two

ALIGN() expects its second argument to be a power of 2, otherwise
incorrect results are produced for some inputs. The output can be
both larger or smaller than what is expected.

For example, ALIGN(304, 192) equals 320 instead of 384, and
ALIGN(65, 192) equals 256 instead of 192.

However, nestling two ALIGN() as is done in this case seem to only
produce results equal to or bigger than the expected result if ALIGN()
had handled non powers of two, and that in turn results in framesizes
that are either the correct size or too large.

Fortunately, since 192 * 4 / 3 equals 256, it turns out that one ALIGN()
is sufficient.

Fixes: ab1eda449c6e ("media: venus: vdec: handle 10bit bitstreams")
Signed-off-by: Rikard Falkeborn <rikard.falkeborn@...il.com>
---
I'm fairly certain this patch does the right thing, but I have only
compile-tested it (I don't have the hardware to test on). The only
reason I spotted it is that I tried implementing compile-time checking
of arguments to ALIGN (and some other functions) to check that arguments
that are supposed to be powers of two really are powers of two, and it
found this.

 drivers/media/platform/qcom/venus/helpers.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/media/platform/qcom/venus/helpers.c b/drivers/media/platform/qcom/venus/helpers.c
index 7147871d9dc1..194c5dd08803 100644
--- a/drivers/media/platform/qcom/venus/helpers.c
+++ b/drivers/media/platform/qcom/venus/helpers.c
@@ -678,8 +678,8 @@ static u32 get_framesize_raw_yuv420_tp10_ubwc(u32 width, u32 height)
 	u32 extradata = SZ_16K;
 	u32 size;
 
-	y_stride = ALIGN(ALIGN(width, 192) * 4 / 3, 256);
-	uv_stride = ALIGN(ALIGN(width, 192) * 4 / 3, 256);
+	y_stride = ALIGN(width * 4 / 3, 256);
+	uv_stride = ALIGN(width * 4 / 3, 256);
 	y_sclines = ALIGN(height, 16);
 	uv_sclines = ALIGN((height + 1) >> 1, 16);
 
-- 
2.28.0