Tpowa's "Simple Hardware Detection"

tpowa · 2005-11-29 07:09:15

andy tell me where does it hang?
during modules loading, during starting other daemons?
are you running a usb keyboard with usb legacy enabled?
what are your entries in blacklist or modules?
please send me your rc.conf
please send me a log file where a hang occurs in .gz format

shadowhand:
are you running a custom kernel?
do you still get hangs during bootup?

thanks for your answers
greetings
tpowa

shadowhand · 2005-11-29 07:46:59

tpowa wrote:

shadowhand:
are you running a custom kernel?
do you still get hangs during bootup?

I'm running the latest Arch -mm kernel. It doesn't hang at all.. it just doesn't do anything.

tpowa · 2005-11-29 08:18:32

the syntax is hwdetect --show-modules.

andyprior · 2005-11-29 10:36:10

OK. Here goes.

>>andy tell me where does it hang?
>>during modules loading, during starting other daemons?
It varies, see below.

>>are you running a usb keyboard with usb legacy enabled? "
No

>>what are your entries in blacklist or modules?
MOD_BLACKLIST=()
MODULES=(!usbserial nvnet 3c59x)

>>please send me your rc.conf
>>please send me a log file where a hang occurs in .gz format
will send in e-mail.

Meanwhile these are the results of a set of 20 reboots:

1. hangs on nfsd
2. hangs on nfsd
3. OK
4. OK
5. OK
6. hangs on nfsd
7. OK
8. hangs on nfsd
9. OK
10. hangs on nfsd
11. hangs on nfsd
12. OK
DISABLE nfslock, nfsd,netfs in rc.conf
13. hangs on cups
14. hangs on cups
15. hangs on cups
16. OK
DISABLE cups
17. hangs on kdm
18. OK
DISABLE all except syslog,network,crond,kdm
19. hangs on rc.local (ntpd)
20 OK

As you can see there is no single daemon associated with the hangs from which I conclude that they have nothing to do with the problem. I think it is within the script and the point at which the hang occurs is a function of the time it takes to execute the script to the point of failure. But why the intermittent nature of the failure?

I will continue to investigate later today and let you know of any revelations.

Andy.

tpowa · 2005-11-29 15:23:44

your problems sounds really network related.
why do you load your network modules in modules?
is it because of eth1 and eth0 mixup if hwdetect does the job,
then you could try to add aliases to modprobe.conf
eg.:
alias eth0 nvnet
alias eth1 <yourothermodule>

andyprior · 2005-11-29 19:25:45

The network modules are loaded in MODULES= because this forces the order of installation and appears to work. If hwdetect does its own thing, eth1 fails even though the module was loaded by hwdetect.

So, as you suggested tried:
in modprobe.conf
alias eth0 3c59x
alias eth1 nvnet
in rc.conf:
MODULES=()

First reboot, it hung at rc.multi.
Second reboot was successful but eth1 was unavailable despite a loaded module. Manually deleting the modules and re-installing them brought up the network.

Further tried:
MOD_BLACKLIST=(nvnet 3c59x)

Both modules loaded but eth1 unavailable.
Hangs just after hwdetect 3 out of 6 reboots.

More and more curious.

tpowa · 2005-11-29 19:39:15

try the other thing i mailed you, with blacklist and so on

tpowa · 2005-11-29 21:09:13

ok could it help if you unplug all of your usb devices and then try booting?

andyprior · 2005-11-30 11:26:07

Tried everything that has been suggested, and a few more, to no avail. At the moment I am inclined to think that the problem lies with the onboard LAN interfaces which only seem to be happy when the modules (nvnet and 3c59x) are loaded in order. Exactly why I cannot force this with combinations of MOD_BLACKLIST and MODULES= I do not know.

My workaround is to put the following in rc.local:
modprobe -r 3c59x nvnet
modprobe nvnet
modprobe 3c59x
This, combined with default settings in rc.conf for hwdetect, seems to do the trick. At least I can have a working system if hwdetect becomes the preferred module detection script in place of hotplug with which I have not had any problems. An alternative solution is to disable the nforce LAN in the bios and install another LAN card but that seems defeatist.

Snowman · 2005-11-30 15:49:09

Maybe you have already tried that:
Put both nvnet and 3c59x in the MOD_BLACKLIST. *Also*, put them in the MODULES array in the right order. That way, they will be loaded in the rigth order (unless I'm wrong).

andyprior · 2005-11-30 17:04:09

Yes, that was any early attempt and no it doesn't work reliably - hangs intermittently during boot - but I agree it really should do what I want. Elsewhere on this bbs there are some references to nforce LANs with reports of difficulty/failure. What puzzles me is that I have not had any trouble with Hotplug. Meanwhile I have just borrowed a pci LAN card (Belkin, no idea what module is required yet) which I will try to see if the problem is solely with the nforce chips.

tpowa · 2005-11-30 17:34:01

why do you use nvnet? and not forcedeth is there a reaon for it?
perhaps nvidia is bugged, they do this some time you know
greetings
tpowa

andyprior · 2005-12-01 06:20:51

No particular reason. When I first installed arch, about a year ago, I don't think forcedeth was available so I compiled the nvnet module from the Nvidia source. Now forcedeth is included in the kernel and nforce has appeared as an arch package (I wonder if there is a clue there). Anyway, I have tried both modules and they seem to behave identically in respect of the problems I have been describing although I admit to not having been entirely methodical in using one module then the other for all variations I have tried.

tpowa · 2005-12-01 07:14:11

you could also try this:
blacklist your network modules, remove them from modules=
add this to /etc/rc.local:
modprobe nvnet
sleep 2
modprobe yourothermodule

perhaps it's just a timing issue.
greetings
tpowa

andyprior · 2005-12-01 11:09:21

Yes, we are thinking along the same lines and I am about to try something similar to your suggestion. Will let you know the outcome.

lilsirecho · 2005-12-01 16:28:57

tpowa;

Have both systems up and running with initscripts-15.

Both systems have scroll mouse running with new KDE upgrade.

I appreciate all the effort you put into this project which is all OK at this location.

Many thanks

andyprior · 2005-12-02 19:33:25

I have a delusion of comprehension.
With my 2 onboard LANs, hotplug loads the nvnet module before 3c59x, with hwdetect 3c59x loads before nvnet. Consequently eth0 and eth1 become reversed which means that all references to ethx need to be swapped, including any alias.
I think I am correct in saying that hwdetect sorts and loads related modules alphabetically whereas hotplug loads in the order in which they are found on the pci bus. Hence the reversal of eth0/1 allocation.

Does this make sense to anyone?

tpowa · 2005-12-03 08:35:51

that i said in some posts before.
hwdetect does alphabetical loading, i don't know how hotlpug does it but pci bus makes sense.

Blind · 2005-12-03 09:22:48

One question: if I change to this detection, then do I still need hotplug? What are the implications on udev (which depends on hotplug)? I mean, could I just force-deinstall hotplug? Or does the detection of a hotplugged usb/firewire device not work anymore, then?
I guess this was more than one question Thx,

Blind

stavrosg · 2005-12-03 09:26:41

You don't need to have hotplug detect your hardware during startup (aka coldplugging), but you need it when you connect devices during operation (aka hotplugging )

tpowa · 2005-12-03 09:28:31

blind:
hwdetect replaces hotplug (the dameon script) so it doesn't make sense to have both detections on in rc.conf.
don't deinstall hotplug!
the detection of usb/firewire will work if hwdetect will load the hostcontrollers of fw and usb.
hwdetect needs a recent kernel >= 2.6.12

Blind · 2005-12-03 10:23:38

ok - just so I get this right:
* use hwdetect for bootup module loading
* keep hotplug for 'hotplugging' devices

So in rc.conf, I just change 'hotplug' to 'hwdetect', fiddle with some modules I do/dont want loaded and (hopefully) everything works (phew, I am getting old).

I just got a kernel panic without changing anything! :shock:
I.e. that the normal 'hotplug' daemon was started - not using 'hwdetect'.
I never ever had that before!

Greetings,
Blind

tpowa · 2005-12-03 10:36:46

do you have you updated to a new kernel if yes to which one?
don't change hotplug to hwdetect!
just put a ! in front of hotplug --> !hotplug
add this to rc.conf:
# Scan hardware and load required modules at bootup
MOD_AUTOLOAD="yes"
# Module Blacklist - modules in this list will never be loaded by hwdetect
MOD_BLACKLIST=()

that's all to get hwdetect running
greetings
tpowa

Blind · 2005-12-03 10:45:03

Wouldn't know what to do without you
Alright!
I have homebrewn kernel 2.6.14.2, with all my drivers compiled as modules (except evdev (compiled in), which caused some problems lately) - you obviously had some problems there, too, as I can see from your beautiful script.
I shall do as you say now, give it a shot and post my results here. The kernel panic could have also come from my laptop, which behaves strange at times - never had a kernel panic, though.
Cheers and thank you!!!
Blind

Blind · 2005-12-03 11:34:04

Worked very nicely (first module to get blacklistet was pcspkr ).
eth0 and eth1 changed names.
Thats it, everything else is beautiful.
Great stuff!!!
Blind

Arch Linux

#126 2005-11-29 07:09:15

Re: Tpowa's "Simple Hardware Detection"

#127 2005-11-29 07:46:59

Re: Tpowa's "Simple Hardware Detection"

#128 2005-11-29 08:18:32

Re: Tpowa's "Simple Hardware Detection"

#129 2005-11-29 10:36:10

Re: Tpowa's "Simple Hardware Detection"

#130 2005-11-29 15:23:44

Re: Tpowa's "Simple Hardware Detection"

#131 2005-11-29 19:25:45

Re: Tpowa's "Simple Hardware Detection"

#132 2005-11-29 19:39:15

Re: Tpowa's "Simple Hardware Detection"

#133 2005-11-29 21:09:13

Re: Tpowa's "Simple Hardware Detection"

#134 2005-11-30 11:26:07

Re: Tpowa's "Simple Hardware Detection"

#135 2005-11-30 15:49:09

Re: Tpowa's "Simple Hardware Detection"

#136 2005-11-30 17:04:09

Re: Tpowa's "Simple Hardware Detection"

#137 2005-11-30 17:34:01

Re: Tpowa's "Simple Hardware Detection"

#138 2005-12-01 06:20:51

Re: Tpowa's "Simple Hardware Detection"

#139 2005-12-01 07:14:11

Re: Tpowa's "Simple Hardware Detection"

#140 2005-12-01 11:09:21

Re: Tpowa's "Simple Hardware Detection"

#141 2005-12-01 16:28:57

Re: Tpowa's "Simple Hardware Detection"

#142 2005-12-02 19:33:25

Re: Tpowa's "Simple Hardware Detection"

#143 2005-12-03 08:35:51

Re: Tpowa's "Simple Hardware Detection"

#144 2005-12-03 09:22:48

Re: Tpowa's "Simple Hardware Detection"

#145 2005-12-03 09:26:41

Re: Tpowa's "Simple Hardware Detection"

#146 2005-12-03 09:28:31

Re: Tpowa's "Simple Hardware Detection"

#147 2005-12-03 10:23:38

Re: Tpowa's "Simple Hardware Detection"

#148 2005-12-03 10:36:46

Re: Tpowa's "Simple Hardware Detection"

#149 2005-12-03 10:45:03

Re: Tpowa's "Simple Hardware Detection"

#150 2005-12-03 11:34:04

Re: Tpowa's "Simple Hardware Detection"

Board footer