Re: can't boot kernel 4.x on SX164

On Sat, 20 Feb 2016, Helge Deller wrote:

> On 20.02.2016 08:41, Michael Cree wrote:
>> On Sat, Feb 20, 2016 at 12:09:49AM -0500, Alex Winbow wrote:
>>> 	I'm unable to boot kernel 4.2 or 4.4 on SX164. aboot loads the 
>>> kernel and the initrd successfully, then returns to the SRM prompt -- 
>>> no error messages. (I'm using kernel 3.16 at present.)
>>> 	Are there any known issues with Linux 4.x over 3.x on alpha, or on
>>> SX164/EB164 in particular?
>> Yes, I have seen the same thing with the Debian 4.3 generic kernel.
>> But a self-built 4.3 kernel boots fine.
> It might be related, but on parisc I'm seeing the same problem that
> kernels above 4.3 crash and won't boot. There have been a few reports
> about it, e.g.  https://www.spinics.net/lists/linux-parisc/msg06624.html
> I'm trying to debug it, but like as for Michael the debian kernel crashes,
> while my self-built 4.3 kernel boots fine. It seems to be related if drivers
> are built-in or loaded as modules during bootup.

Hi, Michael, Helge,

 	Thanks for responding!

 	I built kernel 4.4.2 with the Debian 'generic' config options 
(took 33 hours!), and it exhibits exactly the same symptom as the 4.2 and 
4.4 packages do: aboot loads the kernel, then the initrd, jumps to the 
kernel, and then comes back to the SRM prompt with "BOOT FAILURE" and no 
error messages from the kernel at all.

 	Unlike the parisc case and the dmesg output referred to in that 
link, the kernel seemingly doesn't boot at all, never gets to the point of 
initializing drivers.

 	Helge, I'm not sure that this is related to whether drivers are 
built-in or loaded as modules, because I don't think it's getting anywhere 
near that far in the kernel boot. I'm guessing the problem is in the 
bootstrap sequence -- not a single printk() gets out.

 	I'm glad to hear that self-built kernels do boot for both of 
y'all, so there must be something in the kernel config that relates to the 
very early kernel startup that is different for the generic Debian config 
vs. your own configs. Any ideas what that might be? I guess I'll start by 
setting machinetype from generic to SX164 and rebuilding.


Possibly related:
 	Are you both using aboot 0.9b? I found the following post on LKML:
https://lkml.org/lkml/2007/8/14/259
 	... which refers to a patch to aboot-1.0_pre20040408 (but that is 
back in the linux 2.6 days).


> One assumption I have is that it mainly affects parisc and alpha, 
> because those two arches don't have ARCH_HAS_SG_CHAIN defined. I 
> stumbled over commit 10c95ed9aa2970e05fedb4ac8b3ce1b934dab17b (which 
> seems fine in general) 
> [https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=10c95ed9aa2970e05fedb4ac8b3ce1b934dab17b] 
> but maybe the crypto modules and the scsi libs are involved here and the 
> scatterlist changes lead to memory corruption somewhere.
 	Not sure what to make of that. Again, I think we're well before 
crypto and scsi libs become relevant, but I've zero knowloedge of how 
early these "scatterlists" become relevant in the bootstrap sequence.

 			Thanks,
 			-Alex