PATCH: make h2200-bootloader load & depend on mtdblock

PATCH: make h2200-bootloader load & depend on mtdblock

[Paul Eggleton]

[-- Attachment #1: Type: text/plain, Size: 262 bytes --]

Hi there,

The attached patch fixes bug 4451 by making h2200-bootloader depend on 
kernel-module-mtdblock and load the mtdblock module before attempting to use 
it.

http://bugs.openembedded.net/show_bug.cgi?id=4451

Does this look OK?

Cheers,
Paul

[-- Attachment #2: h2200-suspend-dev.patch --]
[-- Type: text/x-diff, Size: 969 bytes --]

#
# old_revision [87bf384fc6b31f73e5ad8bbd29b113adcc5cd9e0]
#
# patch "packages/h2200-bootloader/h2200-bootloader.bb"
#  from [d015a17dbeb17e4fef0f6d6df88a9ca9620cc88f]
#    to [80838c3d05408a49597b6fad5a441fecf5af0af7]
#
============================================================
--- packages/h2200-bootloader/h2200-bootloader.bb	d015a17dbeb17e4fef0f6d6df88a9ca9620cc88f
+++ packages/h2200-bootloader/h2200-bootloader.bb	80838c3d05408a49597b6fad5a441fecf5af0af7
@@ -1,6 +1,7 @@ DESCRIPTION = "Bootloader firmware extra
 DESCRIPTION = "Bootloader firmware extractor for the h2200 iPAQ"
+RDEPENDS = kernel-module-mtdblock
 ALLOW_EMPTY = "1"
-PR="r2"
+PR="r3"
 
 COMPATIBLE_MACHINE = "h2200"
 #it is a shell script, but lets protect the innocent some more
@@ -9,5 +10,6 @@ mkdir -p /lib/firmware
 pkg_postinst() {
 #!/bin/sh
 mkdir -p /lib/firmware
+modprobe mtdblock
 dd if=/dev/mtdblock0 of=/lib/firmware/h2200_bootloader.bin 2>/dev/null
 }

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Rolf Leggewie]

Hi Paul,

two very minor comments

Paul Eggleton wrote:
> +RDEPENDS = kernel-module-mtdblock

RDEPENDS = "kernel-module-mtdblock"

> +PR="r3"

PR = "r3"

Regards

Rolf

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Koen Kooi]

Paul Eggleton wrote:
> Hi there,
>
> The attached patch fixes bug 4451 by making h2200-bootloader depend on
> kernel-module-mtdblock and load the mtdblock module before attempting to use
> it.
>
> http://bugs.openembedded.net/show_bug.cgi?id=4451
>
> Does this look OK?

No

+RDEPENDS = kernel-module-mtdblock

Never ever rdepend on a kernelmodules, since it might be built in to the 
kernel in the future (like it was in the past), only use

RRECOMMENDS_${PN} = "kernel-module-foo"

regards,

Koen

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Paul Eggleton]

Koen Kooi wrote:
> Never ever rdepend on a kernelmodules, since it might be built in to the
> kernel in the future (like it was in the past), only use
>
> RRECOMMENDS_${PN} = "kernel-module-foo"

OK, so please excuse my ignorance and I don't have OE in front of me to
test it, but does that mean that kernel-module-mtdblock will be included
in images where h2200-bootloader is included? If not, then it needs to be
done some other way.

Cheers,
Paul

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Koen Kooi]

Paul Eggleton wrote:
> Koen Kooi wrote:
>> Never ever rdepend on a kernelmodules, since it might be built in to the
>> kernel in the future (like it was in the past), only use
>>
>> RRECOMMENDS_${PN} = "kernel-module-foo"
>
> OK, so please excuse my ignorance and I don't have OE in front of me to
> test it, but does that mean that kernel-module-mtdblock will be included
> in images where h2200-bootloader is included? If not, then it needs to be
> done some other way.

It means that h2200-bootloader (but not the -dev and -dbg packages) with 
have this in its opkg controlfile:

Recommends: kernel-module-foo

regards,

Koen

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Richard Purdie]

On Thu, 2008-07-24 at 20:08 +1200, Paul Eggleton wrote:
> Koen Kooi wrote:
> > Never ever rdepend on a kernelmodules, since it might be built in to the
> > kernel in the future (like it was in the past), only use
> >
> > RRECOMMENDS_${PN} = "kernel-module-foo"
> 
> OK, so please excuse my ignorance and I don't have OE in front of me to
> test it, but does that mean that kernel-module-mtdblock will be included
> in images where h2200-bootloader is included? If not, then it needs to be
> done some other way.

Yes, its standard practise to RRECOMMEND kernel modules since if they're
built into the kernel you get build failures otherwise. The RRECOMMEND
is a standard approach and well tested.

Regards,

Richard

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Udo van den Heuvel]

Richard Purdie wrote:
> Yes, its standard practise to RRECOMMEND kernel modules since if they're
> built into the kernel you get build failures otherwise. The RRECOMMEND
> is a standard approach and well tested.

Apart from this discussion, sorry to interrupt, but does mtdchar
9related to mtdblock I guess) actually work?
On my H2200('s) it fails to properly flash a kernel.
It doesn't matter much now because we boot from SD, but nevertheless it
is strange.
The mtd folks refuse to help me when I repeatedly queried them.

Udo

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Richard Purdie]

On Thu, 2008-07-24 at 15:04 +0200, Udo van den Heuvel wrote:
> Richard Purdie wrote:
> > Yes, its standard practise to RRECOMMEND kernel modules since if they're
> > built into the kernel you get build failures otherwise. The RRECOMMEND
> > is a standard approach and well tested.
> 
> Apart from this discussion, sorry to interrupt, but does mtdchar
> 9related to mtdblock I guess) actually work?
> On my H2200('s) it fails to properly flash a kernel.
> It doesn't matter much now because we boot from SD, but nevertheless it
> is strange.
> The mtd folks refuse to help me when I repeatedly queried them.

Yes, mtdchar usually works. It may not do exactly what you expect though
as its a bit different from a normal character device.

Cheers,

Richard

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Udo van den Heuvel]

Richard Purdie wrote:
> On Thu, 2008-07-24 at 15:04 +0200, Udo van den Heuvel wrote:
>> Richard Purdie wrote:
>>> Yes, its standard practise to RRECOMMEND kernel modules since if they're
>>> built into the kernel you get build failures otherwise. The RRECOMMEND
>>> is a standard approach and well tested.
>> Apart from this discussion, sorry to interrupt, but does mtdchar
>> 9related to mtdblock I guess) actually work?
>> On my H2200('s) it fails to properly flash a kernel.
>> It doesn't matter much now because we boot from SD, but nevertheless it
>> is strange.
>> The mtd folks refuse to help me when I repeatedly queried them.
> 
> Yes, mtdchar usually works.

'usually' sounds atypical here:

http://bugs.openembedded.net/show_bug.cgi?id=4354
but also:
http://bugs.openembedded.net/show_bug.cgi?id=4324
and:
http://bugs.openembedded.net/show_bug.cgi?id=4362

So a whole slew of flash-related issues, all based on mtd-related code.
My ipaq ran win2003 mobile fine.

> It may not do exactly what you expect though

That is what I observe, others see similar issues.

Udo

Re: PATCH: make h2200-bootloader load & depend on mtdblock

[Paul Eggleton]

On Thursday 24 July 2008, Richard Purdie wrote:
> Yes, its standard practise to RRECOMMEND kernel modules since if they're
> built into the kernel you get build failures otherwise. The RRECOMMEND
> is a standard approach and well tested.

Works like a charm. I have committed the changes including all 
recommendations. Thanks all!

Cheers,
Paul