[RFC PATCH 1/1] mtd: block2mtd: add support for an optional custom MTD label

[Miquel Raynal]

Hi Joachim,

troglobit at gmail.com wrote on Wed, 29 Sep 2021 17:27:20 +0200:

> Hi Miquel!
> 
> On Tue, Sep 28, 2021 at 18:05, Miquel Raynal <miquel.raynal at bootlin.com> wrote:
> > Hi Joachim,
> > troglobit at gmail.com wrote on Tue, 28 Sep 2021 17:49:38 +0200:  
> >> This patch adds support for an optional MTD label for mtd2block emulated
> >> MTD devices.  Useful when, e.g. testing device images using Qemu.  The
> >> following /etc/fstab line in can then be used to mount a file system
> >> regardless of the actual MTD partition number:
> >> 
> >>     mtd:Config  /mnt    jffs2   noatime,nodiratime      0    0
> >> 
> >> Kernel command line syntax:
> >> 
> >>     block2mtd.block2mtd=/dev/sda,,Config
> >> 
> >> The ',,' is the optional erase_size, which like before this patch,
> >> defaults to PAGE_SIZE when omitted.  
> > We already have a label property that can be added in the DT, isn't
> > it enough?  
> 
> Unless I'm missing something, no.  Sorry if if the description was a bit
> unclear.  Let me try expanding on it a bit.
> 
> The block2mtd driver is used a lot when emulating proper MTDs, I'm sure
> we are not alone in building system images that a flashed to an MTD
> device, which we want to be able to test in Qemu before deploying.
> 
> So what we typically do is have a read-only rootfs (squashfs) on one
> partition, and another for config data that uses overlayfs to get a
> writable /etc, for instance.  This is a reduced example, and the order
> of parititions sometimes differ, so we use parition labels. When we test
> in Qemu there's only the automatic DTB from Qemu itself that's injected,
> e.g.
> 
>   qemu-system-aarch64 -M virt -cpu cortex-a53 -nographic       \
> 	-kernel Image -initrd rootfs.img -append               \
>           "root=/dev/ram block2mtd.block2mtd=/dev/vda"         \
>         -drive file=config.img,if=none,format=raw,id=hd0       \
>         -device virtio-blk-device,drive=hd0
> 
> With the default blockt2mtd args we convert the emulated disk to be
> used as an MTD partition, but the name we get in Linux is not the
> name we have in our /etc/fstab on a real target.
> 
>   # cat /proc/mtd 
>   dev:    size   erasesize  name
>   mtd0: 00f00000 00001000 "block2mtd: /dev/vda"
> 
> With the patch we can align the two cases; the embedded system and the
> emulated one, by using the same paritition label.  I.e., with the patch
> and the new argument to the kernel cmdline:
> 
>   qemu-system-aarch64 -M virt -cpu cortex-a53 -nographic       \
> 	-kernel Image -initrd rootfs.img -append               \
>           "root=/dev/ram block2mtd.block2mtd=/dev/vda,,Config" \
>         -drive file=config.img,if=none,format=raw,id=hd0       \
>         -device virtio-blk-device,drive=hd0
> 
> Linux gives the emulated MTD the same partition name we use in the
> real embedded system:
> 
>   # cat /proc/mtd 
>   dev:    size   erasesize  name
>   mtd0: 00f00000 00001000 "Config"

I don't know this portion of the code very well, but indeed if this is
something only declared on the cmdline I understand the request.

Thanks,
Miquèl