Re: [coreboot] initrd in 4.4 versus head

[Trammell Hudson]

On Thu, Jul 28, 2016 at 10:04:56PM +0200, Stefan Reinauer wrote:
> * Trammell Hudson  [160727 13:58]:
> > It looks like 4.4 is adding the initrd as a separate section
> > named "(empty)" with type "null" and the kernel can't find it:
> 
> (empty) is indeed what it claims to be, empty space in the image.
> There is no initrd in there.

Ah, yes.  I see that the (empty) space is the amount remaining
from the end of the payload to the bootblock at the top of the ROM
image.

So that then leads to the question, where is the initrd image
when I build with coreboot-4.4 with the same .config file?  

Building with head runs:

build/util/cbfstool/cbfstool build/coreboot.pre.tmp add-payload -f bzImage -n 
fallback/payload -t payload -c none  -r COREBOOT-I "initrd.img"

While 4.4 executes:

build/util/cbfstool/cbfstool build/coreboot.pre.tmp add-payload -f ./bzImage -n 
fallback/payload -t payload -c none  -r COREBOOT

Looks like the fix happened when the definition of
$(CONFIG_CBFS_PREFIX)/payload-options was moved from ./Makefile.inc
to ./payloads/external/Makefile.inc.  In 4.4 the definition happens before
ADDITIONAL_PAYLOAD_CONFIG is defined, so it will always be empty
and initrd is ignored.

-- 
Trammell

-- 
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot

Re: [coreboot] initrd in 4.4 versus head

[Stefan Reinauer]

* Trammell Hudson  [160727 13:58]:
> I see a difference in the way 4.4 handles initrd images for linux
> payloads versus the way it is done in head.  With 4.4 my Linux
> kernel can not find the external initrd, so it is necessary to
> build it as part of the kernel.  With head it works fine.
> 
> It looks like 4.4 is adding the initrd as a separate section
> named "(empty)" with type "null" and the kernel can't find it:

(empty) is indeed what it claims to be, empty space in the image.
There is no initrd in there.

> performing operation on 'COREBOOT' region...
> Name   Offset Type Size
> cbfs master header 0x0cbfs header  32
> cpu_microcode_blob.bin 0x80   microcode22528
> cmos.default   0x5900 cmos_default 256
> cmos_layout.bin0x5a40 cmos_layout  1948
> fallback/dsdt.aml  0x6240 raw  13847
> (empty)0x98c0 null 26264
> fallback/romstage  0xff80 stage74020
> (empty)0x22140null 56664
> mrc.cache  0x2fec0mrc_cache65536
> fallback/ramstage  0x3ff00stage84790
> fallback/payload   0x54a80payload  1618769
> (empty)0x1dfe40   null 2226328
> bootblock  0x3ff700   bootblock1952
> 
> While in head it is bundling them together into the payload
> region (3.9 MB == bzImage + initrd.img) -- the kernel can
> find the image and use it:
> 
> Performing operation on 'COREBOOT' region...
> Name   Offset Type Size
> cbfs master header 0x0cbfs header  32
> fallback/romstage  0x80   stage14620
> cpu_microcode_blob.bin 0x3a00 microcode22528
> fallback/ramstage  0x9280 stage43781
> cmos_layout.bin0x13dc0cmos_layout  1948
> fallback/dsdt.aml  0x145c0raw  4021
> fallback/payload   0x155c0payload  3906169
> (empty)0x3cf080   null 199256
> bootblock  0x3ffb00   bootblock960
> 
> I don't see any changes in the util/cbfstool/cbfs-payload-linux.c
> file between these two versions.  Is there something else
> that changed?

It seems these two images are significantly different from each other,
apart from the payload. Almost all the stages are about twice as big in
the first image.

Did you compare the .config files for both images? Did you use the same
compiler to produce them? Are these for the same mainboard? Which one?

Stefan




-- 
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot

[coreboot] initrd in 4.4 versus head

[Trammell Hudson]

I see a difference in the way 4.4 handles initrd images for linux
payloads versus the way it is done in head.  With 4.4 my Linux
kernel can not find the external initrd, so it is necessary to
build it as part of the kernel.  With head it works fine.

It looks like 4.4 is adding the initrd as a separate section
named "(empty)" with type "null" and the kernel can't find it:

performing operation on 'COREBOOT' region...
Name   Offset Type Size
cbfs master header 0x0cbfs header  32
cpu_microcode_blob.bin 0x80   microcode22528
cmos.default   0x5900 cmos_default 256
cmos_layout.bin0x5a40 cmos_layout  1948
fallback/dsdt.aml  0x6240 raw  13847
(empty)0x98c0 null 26264
fallback/romstage  0xff80 stage74020
(empty)0x22140null 56664
mrc.cache  0x2fec0mrc_cache65536
fallback/ramstage  0x3ff00stage84790
fallback/payload   0x54a80payload  1618769
(empty)0x1dfe40   null 2226328
bootblock  0x3ff700   bootblock1952

While in head it is bundling them together into the payload
region (3.9 MB == bzImage + initrd.img) -- the kernel can
find the image and use it:

Performing operation on 'COREBOOT' region...
Name   Offset Type Size
cbfs master header 0x0cbfs header  32
fallback/romstage  0x80   stage14620
cpu_microcode_blob.bin 0x3a00 microcode22528
fallback/ramstage  0x9280 stage43781
cmos_layout.bin0x13dc0cmos_layout  1948
fallback/dsdt.aml  0x145c0raw  4021
fallback/payload   0x155c0payload  3906169
(empty)0x3cf080   null 199256
bootblock  0x3ffb00   bootblock960

I don't see any changes in the util/cbfstool/cbfs-payload-linux.c
file between these two versions.  Is there something else
that changed?

-- 
Trammell

-- 
coreboot mailing list: coreboot@coreboot.org
https://www.coreboot.org/mailman/listinfo/coreboot