large initramfs causes h/w reset after decompressing

large initramfs causes h/w reset after decompressing

[John Z. Bohach]

I have a large initramfs cpio file, about 890 MB, which expands to 2.7 
GB, and am running on linux-2.6.36.1, in 64-bit mode with 8 GB of 
physical RAM.

Upon loading the initramfs, I see the message "Decompressing..." for a 
few seconds, and then comes a reset.

This question has come up before...is there some sort of max size. that 
I am exceeding?  With 8 GB of physical memory and 64-bit arch. 
(AMD-II), certainly seems that there should be enough memory, no?

Thanks...

large initramfs causes h/w reset after decompressing

[John Z. Bohach]

I have a large initramfs cpio file, about 890 MB, which expands to 2.7 
GB, and am running on linux-2.6.36.1, in 64-bit mode with 8 GB of 
physical RAM.

Upon loading the initramfs, I see the message "Decompressing..." for a 
few seconds, and then comes a reset.

This question has come up before...is there some sort of max size. that 
I am exceeding?  With 8 GB of physical memory and 64-bit arch. 
(AMD-II), certainly seems that there should be enough memory, no?

Thanks...

Re: large initramfs causes h/w reset after decompressing

[H. Peter Anvin]

On 07/14/2011 09:06 AM, John Z. Bohach wrote:
> I have a large initramfs cpio file, about 890 MB, which expands to 2.7 
> GB, and am running on linux-2.6.36.1, in 64-bit mode with 8 GB of 
> physical RAM.
> 
> Upon loading the initramfs, I see the message "Decompressing..." for a 
> few seconds, and then comes a reset.
> 
> This question has come up before...is there some sort of max size. that 
> I am exceeding?  With 8 GB of physical memory and 64-bit arch. 
> (AMD-II), certainly seems that there should be enough memory, no?
> 
> Thanks...

What bootloader are you using?  There are some bootloaders which don't
allocate the initramfs in a very good spot.

	-hpa

Re: large initramfs causes h/w reset after decompressing

[John Z. Bohach]

On Thursday 14 July 2011 02:14:08 pm H. Peter Anvin wrote:
> On 07/14/2011 09:06 AM, John Z. Bohach wrote:
> > I have a large initramfs cpio file, about 890 MB, which expands to
> > 2.7 GB, and am running on linux-2.6.36.1, in 64-bit mode with 8 GB
> > of physical RAM.
> >
> > Upon loading the initramfs, I see the message "Decompressing..."
> > for a few seconds, and then comes a reset.
> >
> > This question has come up before...is there some sort of max size.
> > that I am exceeding?  With 8 GB of physical memory and 64-bit arch.
> > (AMD-II), certainly seems that there should be enough memory, no?
> >
> > Thanks...
>
> What bootloader are you using?  There are some bootloaders which
> don't allocate the initramfs in a very good spot.
>
> 	-hpa

I've actually gotten the improbable to work, and am PXE booting via 
syslinux and dhcp.

It all works fine with an initramfs of ~20 MB, but that's quite a diff. 
from 2.7 GB of expanded /root (890 MB cpio.tar.gz file).

But excuse my ignorance, I don't see the relevance of your 
question...the initramfs is part of the kernel itself...I simply load 
the kernel bzImage (all 890 MB) and that's where its all at...this is 
_not_ an initrd.

--john

Re: large initramfs causes h/w reset after decompressing

[H. Peter Anvin]

On 07/14/2011 03:32 PM, John Z. Bohach wrote:
> On Thursday 14 July 2011 02:14:08 pm H. Peter Anvin wrote:
>> On 07/14/2011 09:06 AM, John Z. Bohach wrote:
>>> I have a large initramfs cpio file, about 890 MB, which expands to
>>> 2.7 GB, and am running on linux-2.6.36.1, in 64-bit mode with 8 GB
>>> of physical RAM.
>>>
>>> Upon loading the initramfs, I see the message "Decompressing..."
>>> for a few seconds, and then comes a reset.
>>>
>>> This question has come up before...is there some sort of max size.
>>> that I am exceeding?  With 8 GB of physical memory and 64-bit arch.
>>> (AMD-II), certainly seems that there should be enough memory, no?
>>>
>>> Thanks...
>>
>> What bootloader are you using?  There are some bootloaders which
>> don't allocate the initramfs in a very good spot.
>>
>> 	-hpa
> 
> I've actually gotten the improbable to work, and am PXE booting via 
> syslinux and dhcp.
> 
> It all works fine with an initramfs of ~20 MB, but that's quite a diff. 
> from 2.7 GB of expanded /root (890 MB cpio.tar.gz file).
> 
> But excuse my ignorance, I don't see the relevance of your 
> question...the initramfs is part of the kernel itself...I simply load 
> the kernel bzImage (all 890 MB) and that's where its all at...this is 
> _not_ an initrd.
> 

Ah, interesting.  You didn't say that.  That does change some things...
I would not at all be surprised if the kernel decompressor doesn't
handle that very well...

	-hpa

Re: large initramfs causes h/w reset after decompressing

[John Z. Bohach]

> > But excuse my ignorance, I don't see the relevance of your
> > question...the initramfs is part of the kernel itself...I simply
> > load the kernel bzImage (all 890 MB) and that's where its all
> > at...this is _not_ an initrd.
>
> Ah, interesting.  You didn't say that.  That does change some
> things... I would not at all be surprised if the kernel decompressor
> doesn't handle that very well...
>

Sorry, I thought initramfs would be clear on its own that its not 
initrd, but I guess some people do use it interchangeably.  Anyway, I 
have some experience in programming, so is there anything you could 
suggest that I might tweak to test your theory on the decompressor?  Or 
would this be more of a kernel VM allocator issue?  Also, would 
ramdisk_size have any bearing on this?

--john

Re: large initramfs causes h/w reset after decompressing

[H. Peter Anvin]

On 07/14/2011 04:35 PM, John Z. Bohach wrote:
>>> But excuse my ignorance, I don't see the relevance of your
>>> question...the initramfs is part of the kernel itself...I simply
>>> load the kernel bzImage (all 890 MB) and that's where its all
>>> at...this is _not_ an initrd.
>>
>> Ah, interesting.  You didn't say that.  That does change some
>> things... I would not at all be surprised if the kernel decompressor
>> doesn't handle that very well...
>>
> 
> Sorry, I thought initramfs would be clear on its own that its not 
> initrd, but I guess some people do use it interchangeably.  Anyway, I 
> have some experience in programming, so is there anything you could 
> suggest that I might tweak to test your theory on the decompressor?  Or 
> would this be more of a kernel VM allocator issue?  Also, would 
> ramdisk_size have any bearing on this?
> 

initramfs can be built in or loaded via the initrd protocol.

Anyway, it's hard to say for sure what to look for, but the first thing
is to try to "fence" the problem... that is, establish the latest point
which is successfully crossed and the earliest point which is not.

	-hpa

Re: large initramfs causes h/w reset after decompressing

[John Z. Bohach]

> > Sorry, I thought initramfs would be clear on its own that its not
> > initrd, but I guess some people do use it interchangeably.  Anyway,
> > I have some experience in programming, so is there anything you
> > could suggest that I might tweak to test your theory on the
> > decompressor?  Or would this be more of a kernel VM allocator
> > issue?  Also, would ramdisk_size have any bearing on this?
>
> initramfs can be built in or loaded via the initrd protocol.
>
> Anyway, it's hard to say for sure what to look for, but the first
> thing is to try to "fence" the problem... that is, establish the
> latest point which is successfully crossed and the earliest point
> which is not.

Fair enough...I'll start with a 2 GB boundary and see where it takes 
me...thanks for you feedback.

--john