Fwd: the biggest regression at all

Fwd: the biggest regression at all

[werner]

[-- Attachment #1: Type: text/plain, Size: 130 bytes --]

"werner" <w.landgraf@ru.ru>


   --- the forwarded message follows ---
---
Professional hosting for everyone - http://www.host.ru

[-- Attachment #2: Type: message/rfc822, Size: 5611 bytes --]

From: "Rafael J. Wysocki" <rjw@sisk.pl>
To: "werner" <w.landgraf@ru.ru>
Subject: Re: the biggest regression at all
Date: Tue, 29 Dec 2009 23:35:10 +0100
Message-ID: <200912292335.10178.rjw@sisk.pl>

Hi,

First of all, I'm not the right person to report these problems to.

On Tuesday 29 December 2009, werner wrote:
> It's very important your work to collect regressions. 
> Because in the last months accumulate more and more 
> regressions, which gave on more and more computers 
> (especially, laptops) problems.
> 
> There continues two problems which I reported earlier, and 
> which are big problems, making impossible the instalation 
> of new Linux kernels on a good part of laptops, but which 
> were ignored or even negated, anyway not resolved.
> 
> / The earlier hda=scsi command line parameter don't work 
> more correctly or perhaps was substituted (I dont know 
> exactly, because the kernel command line parameter list is 
> obsolete, too).   Anyway, there are laptops, whose IDE 
> devices during booting is reclamed as ocupied, without 
> end.  After some minutes the boot process cancel this and 
> continues, however without assign the device as detected. 
> Then booting fails because the disk and the system isn't 
> found.
> 
> The only manner how I found to get detected the IDE device 
> is with pci=off (I tried plenty other things).  But then 
> other things like sound, grafics etc are not detected (and 
> lspci don't give output) and the system works only in text 
> mode.
> 
> For my opinion, the best solution is to introduce an 
> option pci=noide, so that only the IDE devices will be 
> excluded and then detected or taken over by the BIOS. All 
> other things should be detected normally.

The issue above should better be reported to the people on
linux-ide@vger.kernel.org , as no one else really can take care of it.

Still, I think there's a known way to solve this problem, since I've never
seen it myself and I happen to install Linux quite often on notebooks.

What distribution do you use?

> / When I compile the kernel with a machine with more than 
> 4 GB memory (my big computer has 8 GB) then during 
> makemenuconfig I dont can switch off HIGHMEN64 and PAE . 
> Even when before in config I edit them as not set, then 
> they are setted automatically.
> 
> This is a very serios problem, because this means that 
> kernels compiled on machines with more than 4 GB, always 
> works only on PAE machines, but not on older machines.
> 
> Generally, I suggest to change the kernel makeconfig 
> routine, in the sense that one ALWAYS can switch on or off 
> any configuration but that then all necessary dependencies 
> are switched on/off automatically too. At least this 
> config manner should be alternatively.   Otherwhise often 
> one cannot switch on or of certain parameters which one 
> inconditionally need.

I _think_ the problem is that one option (or more) in your config selects
the HIGHMEM64.  As I told you previously, I've never seen this problem myself,
most likely because I've been building kernels on 64-bit machines for quite
some time now.

Nevertheless, you can report it to Michal Marek <mmarek@suse.cz>, who's our
new kbuild maintainer.

Rafael

Re: Fwd: the biggest regression at all

[Tejun Heo]

Hi,

Can you post the failing dmesg?  You can capture it via serial or
netconsole both of which are documented under the Documentation/
directory.

Thanks.

-- 
tejun