[PATCH v2] mtd: cmdlinepart: fix the wrong partitions number when truncating occurs

[Huang Shijie]

On Sun, Aug 26, 2012 at 2:06 AM, Shmulik Ladkani
<shmulik.ladkani at gmail.com> wrote:
> Hi,
>
> On Sat, 25 Aug 2012 05:26:51 -0400 Huang Shijie <shijie8 at gmail.com> wrote:
>> On Sat, Aug 25, 2012 at 5:02 AM, Shmulik Ladkani
>> <shmulik.ladkani at gmail.com> wrote:
>> > Your analysis seems right, but let me offer an alternative approach.
>> >
>> > I would simply:
>> >
>> > -                                       part->num_parts = i;
>> your code does not wors in such kernel command line(also with the 1GB
>> nand chip):
>> #gpmi-nand:100m(root),100m(kernel),1g(rootfs),1g(user),-(rest)
>>
>
> Can you please detail what do you mean by "not work"?
sorry. :)

My meaning was the result is unreadable. It looks too strange. That's
why i use the 'break'
to shortcut the loop.

>
> To my understanding, in this example, according to my suggestion, the
> resulting partitions would be:
>
> root    100m at 0
> kernel  100m at 100m
> rootfs  800m at 200m (truncated)
> user    0 at 1g (truncated)
> rest    0 at 1g
>

yes, the result is like this.

> Reasonable IMO, given the fact that the mtd device size is smaller than

> the specified parts.
>
> I saw you submitted a patch which sorts the cmdline parts; I don't
> understand why this is necessary.
> Also, sorting might not be desirable, as the user specified the unsorted
> partitions might have _wanted_ them to appear in that order.
>
> Now lets focus on your original suggestion and its consequences:
>
> - Orignal code STOPPED parsing at the 1st truncated partition,
>   this partition WAS NOT returned to the caller
> - Your patch STOPS AFTER parsing the 1st truncated partition,
>   this partiton IS returned to the caller (but partitions specified
>   later are no longer parsed)
> - My suggestion CONTINUES parsing all partitions.
>   So later partitions (specified with the 'size' but *without* 'offset')
>   will be truncated AND presented to the caller.
>   AND, if later partitions are specified using the 'size at offset'
>   explicit format, they are parsed normally.
>

Could Artem or David point us a direction about this?
[1]  Should the unsorted partitions be supported?
[2] And what should we do when we meet a 1GB nand, and the cmdline is
      gpmi-nand:100m(root),100m(kernel),1g(rootfs),1g(user),-(rest)

thanks a lot
Huang Shijie