[gentoo-amd64] Re: Alsa Problem With Kernel 2.6.37

[Duncan <1i5t5.duncan@cox.net>]

Frank Peters posted on Sat, 08 Jan 2011 18:11:27 -0500 as excerpted:

>> Curious: why aren't you using udev?
>> 
>> 
> I've been using Linux since 1998 and I like most the opportunity
> for control that Linux allows.  To suit my minimalist needs, I have
> customized my system to a great extent.  For example, I have completely
> eliminated the complex and cumbersome initialization scripts that
> are found in /etc/init.d, /etc/conf.d, and other locations.  In their
> place I wrote my own very simple initialization script that
> automatically boots into console (no login).  From there I can go to X
> (my usual action) or just use the console for configuration,
> troubleshooting, or other tasks.

> I have also eliminated udev as part of this simplification.  I know
> what's on my machine and what has to be done to enable it.  I would
> rather not have actions performed "automagically" for me whenever I plug
> in a USB printer of storage device.  Part of the reason is philosophical
> and part is just a desire to understand the OS better.

Wow. I understand exactly why you do it, because I'm the same way, only 
the Gentoo initscripts are reasonable enough for me and I've enjoyed 
watching openrc/baselayout2 develop -- in fact, since I actually /grok/ 
bash scripting at a reasonable level, its the one area I can quite 
frequently not only submit bug reports, but trace the code and suggest 
patches, which I've done on a number of occasions.  (In one case, once I 
started hitting an error and traced it, it was quite apparent the error 
path/case had never been tested at all, as the logic was apparent but the 
code simply didn't and couldn't work as intended.  By the time the 
original bug was fixed I think I'd had patches for two others taken and 
helped work on a third, in addition to the one for the original triggering 
bug.)

Also, I remember static dev and highly prefer a dynamic dev, since that 
makes it that way, it's a simple binary-case (1) device there, the kernel 
detects the hardware and created a device, any problem must be above that 
level, (0) device not there, kernel/module/detection problem, that you 
don't get if the device node always exists regardless of whether the 
kernel knows what to do with the device or not, since it's a static file.  
But I definitely understand the appeal and simplicity of doing it the 
other way, avoiding all the layers on layers of "automagic".

But it seems I often have at least one and sometimes two or three 
initscripts that I've highly modified, either to cover "advanced" 
functionality the script doesn't offer on its own (back when md-raid first 
got partitioning support, my version of the script handled that way before 
Gentoo's did, and I had a bug open for some time on it before Gentoo 
finally added that feature based on my work and that of another user, as 
well as the md-raid initscript maintainer; I'm doing similar now with 
additional file-specific bind-mounts into the bind/named chroot), or to 
delete complexity I don't need, sometimes both (I simplify the script, 
cutting out what I don't need, to better undestand it before adding my own 
advanced functionality).

> Linux allows many possibilities for customization and, IMO, that is half
> the fun of using it.  Most distributions, including Gentoo, adhere to
> the standard methods.  But if one has the inclination for exploration
> then there is much that can be done to go well beyond the standard
> methods.

And of course, Gentoo makes maintaining customization much easier than 
many distributions. =:^)  I pretty much learned bash, at least the 
scripting side, by rewriting the Mandrake rc.sysinit script, modularizing 
it, etc.  Then I upgraded to the next version, and had to do it again...  
Then I upgraded again, and shortly thereafter, started running cooker 
(rolling "testing" version), at which point I gave up.  Gentoo's rolling 
updates and configuration update management tools make it relatively easy 
to avoid having a custom config broken.

....

Meanwhile, I see you've fixed the problem, but it's worth noting that lspci 
lists devices based on generic pci-bus device queries, not what the kernel 
has drivers for.  Just because lspci lists it doesn't mean the kernel can 
handle that device, only that the device returned certain information when 
the kernel queried for devices on that bus.

Unless you use the -k switch, enabled by default with -v, that is (and 
only on 2.6 kernels, which we're dealing with of course, but...).  -k 
shows kernel drivers handling and/or that can handle a device.  In that 
case, you get that information too if the kernel can handle the device, 
but devices that the kernel knows not how to handle will still be listed, 
just without any assigned kernel driver.

Maybe you meant that lspci -v (or -k) was saying it had assigned a driver 
to it, but that's not what you said, you said lspci wouldn't show the 
sound card as there if the module wasn't loaded, and that's not correct, 
as it'd still be a device on the bus and would respond to queries (with 
the results printed) as such.

This is an important distinction, because lspci can thus be used to see 
devices that the kernel does NOT have drivers for (or doesn't have them 
built and available).  Using lspci to see what hardware a machine has 
(once Linux boots on it at all), so one can google for support status and 
what driver to use, especially on second-hand equipment where a manual 
isn't available, and/or where it's easier to run lspci than to crack open 
the box to look, is a very important usage case for it.  If lspci didn't 
list devices the kernel wasn't configured with drivers for, that wouldn't 
work, so I'm very glad it DOES list them. =:^)

alsa-info.sh, OTOH, I don't know.  You may be correct about it.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman