linux-kernel - Re: [PATCH] memory: renesas-rpc-if: Remove redundant division of dummy

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

Date:   Fri, 20 Jan 2023 18:52:12 +0100
From:   Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To:     Wolfram Sang <wsa+renesas@...g-engineering.com>,
        linux-renesas-soc@...r.kernel.org
Cc:     Mark Brown <broonie@...nel.org>,
        Sergei Shtylyov <sergei.shtylyov@...entembedded.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] memory: renesas-rpc-if: Remove redundant division of
 dummy

On 12/01/2023 10:06, Wolfram Sang wrote:
> From: Cong Dang <cong.dang.xn@...esas.com>
> 
> The dummy cycles value was wrongly calculated if dummy.buswidth > 1,
> which affects QSPI, OSPI, HyperFlash on various SoCs. We're lucky in
> Single SPI case since its dummy.buswidth equals to 1, so the result of
> the division is unchanged
> 
> This issue can be reproduced using something like the following commands
> A. QSPI mode: Mount device with jffs2 format
>     jffs2: CLEANMARKER node found at 0x00000004, not first node in block (0x00000000)
> 
> B. QSPI mode: Write data to mtd10, where mtd10 is a parition on SPI Flash
> storage, defined properly in a device tree
> 
> [Correct fragment, read from SPI Flash]
> 
>   root@v3x:~# echo "hello" > /dev/mtd10
>   root@v3x:~# hexdump -C -n100 /dev/mtd10
>   00000000  68 65 6c 6c 6f 0a ff ff  ff ff ff ff ff ff ff ff  |hello...........|
>   00000010  ff ff ff ff ff ff ff ff  ff ff ff ff ff ff ff ff  |................|
> 
> [Incorrect read of the same fragment: see the difference at offsets 0-3]
> 
>   root@v3x:~# echo "hello" > /dev/mtd10
>   root@v3x:~# hexdump -C -n100 /dev/mtd10
>   00000000  00 00 00 00 68 65 6c 6c  6f 0a ff ff ff ff ff ff  |....hello.......|
>   00000010  ff ff ff ff ff ff ff ff  ff ff ff ff ff ff ff ff  |................|
> 
> As seen from the result, 4 NULL bytes were inserted before the test data.
> Wrong calculation in rpcif_prepare() led to miss of some dummy cycle. A
> division by bus width is redundant because it had been performed already
> in spi-rpc-if.c::rpcif_spi_mem_prepare()
> 
> Fix this by removing the redundant division.
> 
> Fixes: ca7d8b980b67 ("memory: add Renesas RPC-IF driver")
> Signed-off-by: Cong Dang <cong.dang.xn@...esas.com>
> Signed-off-by: Hai Pham <hai.pham.ud@...esas.com>
> Signed-off-by: Wolfram Sang <wsa+renesas@...g-engineering.com>
> ---
> 
> Sadly, I cannot test this patch myself because I don't have access to
> hardware which uses a buswidth > 1 for the dummy read. However, from
> code review, this patch makes sense. The division by buswidth is done
> twice, once in the SPI driver and once in the RPC core. It should stay
> only in the SPI driver.

Any tests or further reviews on this? If not, I'll pick it up in few days.

Best regards,
Krzysztof