Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
Date:   Fri, 04 Nov 2022 11:04:57 +0100
From:   =?UTF-8?Q?Rafa=C5=82_Mi=C5=82ecki?= <rafal@milecki.pl>
To:     Christian Lamparter <chunkeey@gmail.com>
Cc:     linux-kernel@vger.kernel.org, srinivas.kandagatla@linaro.org,
        gregkh@linuxfoundation.org, a.fatoum@pengutronix.de
Subject: Re: [PATCH v1] nvmem: address crc32 check on redundant-layout powerpc
 machines
In-Reply-To: <41e7ab94800b4a9aef73e212464bd94117429bec.1667555240.git.chunkeey@gmail.com>
References: <41e7ab94800b4a9aef73e212464bd94117429bec.1667555240.git.chunkeey@gmail.com>
User-Agent: Roundcube Webmail/1.4.13
Message-ID: <e22067325dffb02739a7bd7c20419b43@milecki.pl>
Content-Type: text/plain; charset=US-ASCII;
 format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2022-11-04 10:47, Christian Lamparter wrote:
> The Western Digital MyBook Live (PowerPC 464/APM82181)
> has a set of redundant u-boot-env. Loading up the driver
> causes it to error out with:
> 
> | u_boot_env: Invalid calculated CRC32: 0x4f8f2c86 (expected: 
> 0x98b14514)
> | u_boot_env: probe of partition@1e000 failed with error -22
> 
> Looking up the userspace libubootenv utilities source [0],
> it looks like the "mark" or "flag" is not part of the
> crc32 sum... which is unfortunate :(
> 
> |static int libuboot_load(struct uboot_ctx *ctx)
> |{
> |[...]
> |       if (ctx->redundant) {
> |		[...]
> |               offsetdata = offsetof(struct uboot_env_redund, data);
> |		[...]
> |       }
> |       usable_envsize = ctx->size - offsetdata;
> |       buf[0] = malloc(bufsize);
> |[...]
> |	for (i = 0; i < copies; i++) {
> |		data = (uint8_t *)(buf[i] + offsetdata);
> |               uint32_t crc;
> |
> |		ret = devread(ctx, i, buf[i]);
> |		[...]
> |		crc = *(uint32_t *)(buf[i] + offsetcrc);
> |               dev->crc = crc32(0, (uint8_t *)data, usable_envsize);
> |

Thanks for the fix, it may be I didn't actually test that code.

That change has nothing to do with byte swapping so I' suggest TWO
separated patches.


> Now, this alone didn't fully fix the kernel's uboot-env nvmem
> driver. The driver then ran into an endian error on the
> big-endian powerpc device:
> 
> | u_boot_env: Invalid calculated CRC32: 0x1445b198 (expected: 
> 0x98b14514)
> 
> however the __le32 type for the crc32 value is justified because the
> the crc32() is just a macro for crc32_le(). So, to side-step that
> problem, the crc32 check gets extended to also accept a byteswapped
> crc32.

Can you try this pending patch, please?
[PATCH v2] nvmem: u-boot-env: align endianness of crc32 values
https://lore.kernel.org/lkml/20221012155133.287-1-musashino.open@gmail.com/