Skip to contentSkip to conversation

Yamaha OPL3-Sax support for RH 6.0 kernel=2.2.5-15

DISCUSSION Ac 3 participants 3 messages Oct 13, 1999 – Oct 19, 1999 Sound

lore
Share & tools
lore

Ac

author
#0 lore raw

Hi all,
I'm trying to recompile my kernel (RH 6.0, 2.2.5-15) with Yamaha
OPL3-Sax support built in. I want to know what options to give in make - xconfig/config/menuconfig, since after compilation (make dep,make clean, make bzImage, make modules), when I try to run 'sndconfig' or even 'setup' to configure sound, Tux just freezes and I have to do a cold-boot, which is irritating due to the check-forced mounting I must go through when it boots next time.

I'm sure the problem lies with IRQ's and DMA settings. Also I'm
confused whether to use the MPU-401 during compilation or not.

My computer specs are P-II Celeron 333, 64MB RAM, running Win98 and
loadlin to boot Linux. (and I do copy the updated bzImage to my c:
drive to use with loadlin.). There are no IRQ clashes in Win98, since I'm really using no devices not even a modem.

Hoping some of you out there can help me, so thx. in advance and
peace on this bandwidth.

Bye,
Ac.

#1 lore raw
Quoted 8 lines

Hi all,
I'm trying to recompile my kernel (RH 6.0, 2.2.5-15) with Yamaha
OPL3-Sax support built in. I want to know what options to give in make - xconfig/config/menuconfig, since after compilation (make dep,make clean, make bzImage, make modules), when I try to run 'sndconfig' or even 'setup' to configure sound, Tux just freezes and I have to do a cold-boot, which is irritating due to the check-forced mounting I must go through when it boots next time.

I believe this is from before my fix for OPL3-SA2 non-modular was accepted. Try a more recent kernel - or if you can find a linux-sound archive my patch should be in it about 6 months back.

Quoted 2 lines

I'm sure the problem lies with IRQ's and DMA settings. Also I'm
confused whether to use the MPU-401 during compilation or not.

This is an annoyance. The sound makefile includes the module whether you select it or not, and it is not used (for one of mpu401/uart401 or whatever anyway). This is because the person who wrote the sound makefile only seems to think modular and includes both modules saying "it's only an extra module" when it gets linked into the kernel if you select it non-modular, it happens for a couple of sound drivers.

David.

#2 lore raw
In reply to Ac

[I sent this to the mailing list a few days ago, but I have not seen it appear here... not sure why. Apologies if this is seen twice. Also, David Luyer has just posted a message to the affect that the opl3 driver was broken in redhat's 2.2.5-15 kernel. Nevertheless, I don't believe that this is entirely the problem, and that what I have to say here has some value...]

Quoted 2 lines · Ac

I'm trying to recompile my kernel (RH 6.0, 2.2.5-15) with Yamaha
OPL3-Sax support built in. I want to know what options to give in

It stuns my why anyone would want to build a kernel with any of the sound drivers compiled into it...

Quoted 5 lines

make - xconfig/config/menuconfig, since after compilation (make dep,make clean, make bzImage, make modules), when I try to run 'sndconfig' or even 'setup' to configure sound, Tux just freezes and I have to do a cold-boot, which is irritating due to the check-forced mounting I must go through when it boots next time.

I'm not suprised that it doens't work like you expect it to. These tools won't work at runtime because they expect the sound driver to be available as modules. Running them under other circumstances is just asking for disaster... which is exactly what you are experiencing.

Quoted 2 lines

I'm sure the problem lies with IRQ's and DMA settings. Also I'm
confused whether to use the MPU-401 during compilation or not.

Very likely, but believe me... you are going about it all wrong.

Quoted 7 lines

My computer specs are P-II Celeron 333, 64MB RAM, running Win98 and
loadlin to boot Linux. (and I do copy the updated bzImage to my c:
drive to use with loadlin.). There are no IRQ clashes in Win98, since I'm really using no devices not even a modem.

Hoping some of you out there can help me, so thx. in advance and
peace on this bandwidth.

Unless you really know what you are doing, as much as possible don't build kernels with any hardware drivers compiled into it... as much as possible, make drivers as loadable modules. (Of course there are exceptions, eg: cdrom and ide support, and specialised kernels, etc).

Why?

Because it gives you one hell of a lot more flexability.

I'll give you a good example.

I have in my box at home a Creative/Esonic PCI64 card. The driver it needs is the es1370.o module.

Aside:

It was a step-up from the vibra16 card I've had for a very long time, but being pci it was different and it took me quite a while to start to appreciate these differences. What? No synth device? Argh, what's going on... it's supposed to be the latest and greatest? What's going on here?

As it turned out, no problem - just use something like timidity to convert .midi into a .wav and the result is just spectacular. And there are tools to load in the same wavetable files that are used for windows.

That card can do something very cool... it can do 4-speaker output by turning the input plug into a "rear speaker" output.

BTW, full 4-channel quake with linux is just spectacular... its a real experience :-)

Now, in order to set up the driver to do this, you have to tell it to do that when it is initialised. You also configure the mic and joystick at the same time. For other (sound) drivers, you need to tell it the IRQ, DMA and other such parameter settings.

Now let me repeat that... the point is that it is only possible to configure the driver when you first initialise it.

Stop. Read that again, for when the penny drops you'll realise that what I've just said (err, written:) has some very big implications.

For modules, this is easy... you use insmod or modprobe and set these parameters either on the command line or by some magical incantations cast into /etc/conf.modules. (The redhat sndconfig utility generally does an excellent job of putting exactly the right parameters into that file, after which it works. It has never failed me yet).

If you want to, for example, reconfigure the card into a four-speaker output mode, it's easy to do... just rmmod the driver, then reload it with the necessary parameters.

However, if you have the drivers compiled into the kernel, there are only two ways to configure the driver... either by hacking the default values that you want to use into the kernel sources and compiling these in by default. Or by passing the necessary values to the kernel via lilo, either at the boot prompt or in /etc/lilo.conf

So what happens now when you want to reconfigure the driver? Can you guess? Yes? Ok fine, that's right - you are stuck with having to reboot your system to do it! Yeech, very ugly... smells a lot like a micky$lop, no? :)

And what if you got the compile-time or boot-time parameters wrong? Ugly, very ugly - reboot and try again.

With modules you don't have any of these hassles... if it doesn't work, then simply unload the modules and try again. Simple. Easy. Just like it is supposed to be :)

What I say here holds try for a great many of the other hardware drivers. Ethernet cards for example... it's possible to lock them into 100Mbit full-duplex mode by (re)loading them with the appropriate parameters.

Sometimes drivers are needed at boot time (eg, for scsi-based systems), but that's what initrd images are used for -- pre-loading drivers before system bootup "proper". But I won't elaborate on the magic of using initrd images - there are entire HOWTOs explaining what they are all about and how to use them, and redhat has a very cool utility for easily creating them... /sbin/mkinitrd (it has a man page).

Finally... with modules, you can actually *query* them for what parameters they can take, eg:

$ modinfo -p es1370
joystick int array (min = 1, max = 5), description "if 1 enables joystick interface (still need separate driver)"
lineout int array (min = 1, max = 5), description "if 1 the LINE input is converted to LINE out"
micz int array (min = 1, max = 5), description "changes (??) the microphone impedance"
$ modinfo -p 3c59x 
debug int
options int array (min = 1, max = 8)
full_duplex int array (min = 1, max = 8)
rx_copybreak int
max_interrupt_work int
compaq_ioaddr int
compaq_irq int
compaq_device_id int

Now that is way cool, very handy. (modprobe -h for what else this utility can do).

For more information about what these parameters actually do, have a look in the kernel sources at the actual souce code for the driver you are interested in, or (possibly more productively) in the kernel documentation files, in /usr/src/linux/Documentation/ or in /usr/doc/kernel-doc-<version>/ on redhat boxes.

So, the moral of the story is: as much as possible, compile your hardware drivers as modules rather then directly into the kernel.

Cheers
Tony
-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-
Tony Nugent <[email protected]> Systems Administrator
GrowZone OnLine (a project of) GrowZone Development Network
POBox 475 Toowoomba Oueensland Australia 4350 Ph: 07 4632 8344
-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-=*#*=-