Re: kerneld message / Workaround
On Mon, 17 Jul 2000, brian moore wrote: On Sat, Jul 15, 2000 at 11:51:06AM +0100, Jonathan Heaney wrote: David Wright wrote: The scripts /etc/init.d/{kerneld,modutils} have to be able to handle both 2.0 and 2.2 kernels with kerneld or kmod. You will see they do this by testing for the presence of /proc/sys/kernel/modprobe which only exists under 2.2. But doesn't -always- exist on 2.2: [narvi:/etc/init.d] 11:26:37am 136 % ls -l /proc/sys/kernel/modprobe ls: /proc/sys/kernel/modprobe: No such file or directory [narvi:/etc/init.d] 11:26:39am 137 % uname -a Linux narvi 2.2.16 #6 Fri Jun 23 13:51:08 PDT 2000 i686 unknown You need to have 'CONFIG_KMOD' set in your kernel build to have it, which I don't. Seems to me that the logic on that is broken. /proc/sys/kernel/modprobe is not a good way to determine whether kerneld should be run. What I think is that you are wrong. If you have compiled your kernel with 'CONFIG_KMOD' defined, then you would have kmod built into your kernel. kmod is a REPLACEMENT for kerneld, remember? Perhaps you should file a bug on it? Don't be that fast. Regards, Pavel
Re: kerneld message / Workaround
Quoting brian moore ([EMAIL PROTECTED]): On Sat, Jul 15, 2000 at 11:51:06AM +0100, Jonathan Heaney wrote: David Wright wrote: The scripts /etc/init.d/{kerneld,modutils} have to be able to handle both 2.0 and 2.2 kernels with kerneld or kmod. You will see they do this by testing for the presence of /proc/sys/kernel/modprobe which only exists under 2.2. But doesn't -always- exist on 2.2: [narvi:/etc/init.d] 11:26:37am 136 % ls -l /proc/sys/kernel/modprobe ls: /proc/sys/kernel/modprobe: No such file or directory [narvi:/etc/init.d] 11:26:39am 137 % uname -a Linux narvi 2.2.16 #6 Fri Jun 23 13:51:08 PDT 2000 i686 unknown You need to have 'CONFIG_KMOD' set in your kernel build to have it, which I don't. Thanks for finding that out. Of course, the 2.2 kernel documentation does say that you should configure kmod. Seems to me that the logic on that is broken. /proc/sys/kernel/modprobe is not a good way to determine whether kerneld should be run. Perhaps you should file a bug on it? Maybe, but I tend towards the idea that if you decide to go your own way on this, it's not a lot to expect you to have to modify the startup script too. And what about 2.4test kernels (that I'm running) ? I don't have a /proc/sys/kernel/modprobe - what should I do? How do the magic scripts in /etc/init.d deal with that They run kerneld. Wrongly. Ditto. (Maybe, ... too.) Cheers, -- Email: [EMAIL PROTECTED] Tel: +44 1908 653 739 Fax: +44 1908 655 151 Snail: David Wright, Earth Science Dept., Milton Keynes, England, MK7 6AA Disclaimer: These addresses are only for reaching me, and do not signify official stationery. Views expressed here are either my own or plagiarised.
Re: kerneld message / Workaround
On Tue, Jul 18, 2000 at 01:05:42PM +0300, Pavel M. Penev wrote: On Mon, 17 Jul 2000, brian moore wrote: On Sat, Jul 15, 2000 at 11:51:06AM +0100, Jonathan Heaney wrote: David Wright wrote: The scripts /etc/init.d/{kerneld,modutils} have to be able to handle both 2.0 and 2.2 kernels with kerneld or kmod. You will see they do this by testing for the presence of /proc/sys/kernel/modprobe which only exists under 2.2. But doesn't -always- exist on 2.2: [narvi:/etc/init.d] 11:26:37am 136 % ls -l /proc/sys/kernel/modprobe ls: /proc/sys/kernel/modprobe: No such file or directory [narvi:/etc/init.d] 11:26:39am 137 % uname -a Linux narvi 2.2.16 #6 Fri Jun 23 13:51:08 PDT 2000 i686 unknown You need to have 'CONFIG_KMOD' set in your kernel build to have it, which I don't. Seems to me that the logic on that is broken. /proc/sys/kernel/modprobe is not a good way to determine whether kerneld should be run. What I think is that you are wrong. Well, you're allowed your own opinions, even if they are boneheaded and not backed up by facts. If you have compiled your kernel with 'CONFIG_KMOD' defined, then you would have kmod built into your kernel. kmod is a REPLACEMENT for kerneld, remember? Yes, and 'kerneld' is for 2.0, -NOT- 2.2. | WARNING |kerneld is obsolete as of Linux kernel 2.1.90, it was |replaced by the kmod kernel thread and cron entries. Do |not even think of using kerneld unless you are running a |2.0 kernel. See that? It says DO NOT run kerneld on anything but 2.0. NOT. BAD. Got it? Now, the logic in /etc/init.d/kerneld -WILL- run it if you build a 2.2 (remember that number?) kernel with 'CONFIG_KMOD' undefined. See, in that case, /proc/sys/kernel/modprobe will NOT exist, which is how /etc/init.d/kerneld determines the version of your kernel. Go check the source. The /proc entry for sys/kernel/modprobe is in kernel/sysctl.c: #ifdef CONFIG_KMOD {KERN_MODPROBE, modprobe, modprobe_path, 256, 0644, NULL, proc_dostring, sysctl_string }, #endif See the #ifdef? Yep, /proc/sys/kernel/modprobe only exists if KMOD is defined. Perhaps you should file a bug on it? Don't be that fast. Why not? A bug is a bug. 'man kerneld' is quite clear that kerneld should -not- be run on a 2.2 kernel. Yet /etc/init.d/kerneld is quite happy to run it on a 2.2 kernel because it uses an incorrect method for testing the version of the kernel. Why is that not a bug? -- Brian Moore | Of course vi is God's editor. Sysadmin, C/Perl Hacker | If He used Emacs, He'd still be waiting Usenet Vandal | for it to load on the seventh day. Netscum, Bane of Elves.
Re: kerneld message / Workaround
On Tue, 18 Jul 2000, brian moore wrote: Perhaps you should file a bug on it? Don't be that fast. Why not? A bug is a bug. 'man kerneld' is quite clear that kerneld should -not- be run on a 2.2 kernel. Yet /etc/init.d/kerneld is quite happy to run it on a 2.2 kernel because it uses an incorrect method for testing the version of the kernel. Why is that not a bug? I guess the developers have just left the user decide whether to use kerneld or not. You are right to think that they could have done it nicer by using uname, for example. So, I agree -- send a bug report (with severity of about whishlist) if you would. Pavel
Re: kerneld message / Workaround
Quoting Jonathan Heaney ([EMAIL PROTECTED]): David Wright wrote: The scripts /etc/init.d/{kerneld,modutils} have to be able to handle both 2.0 and 2.2 kernels with kerneld or kmod. You will see they do this by testing for the presence of /proc/sys/kernel/modprobe which only exists under 2.2. Perhaps you have a problem with your /proc filesystem (unless a bug has been introduced into these scripts). There should be no need to move/remove the scripts in /etc/init.d. Well if there is such a bug it's been there since potato day1 and still is. Then I presume no bug has been introduced. Incidentally, does the presence of /proc/sys/kernel/modprobe not depend on what you compile in the kernel??? Well, yes. /proc filesystem support has to be built in. Are you saying that you don't do that? I always roll my own and any time I've installed potato with a 2.2 kernel I've had to shift kerneld. I don't know why your kerneld gets started. Surely you can put the odd echo command into the scripts if you want to find out why. Those of us running systems where it works can't really help you as we are unable to replicate the problem. Maybe if you stick with Debians own kernels it might be there but not everyone does. I gather that there are plenty of people on this list who do not use Debian kernels, but pick up tarballs elsewhere. And what about 2.4test kernels (that I'm running) ? I don't have a /proc/sys/kernel/modprobe - what should I do? How do the magic scripts in /etc/init.d deal with that Well this may well be a completely different problem. What happened with slink (released with 2.0 kernels) and 2.2 kernels is that some kind soul made special versions of the affected packages (like netbase) which had the updated package contents compiled against slink's C library. I have no idea if that sort of thing is necessary here because I don't know how woody will differ fundamentally from potato. You may just eventually get an updated modutils package, but I have no idea if anyone is working on it yet (assuming that is the problem). Cheers, -- Email: [EMAIL PROTECTED] Tel: +44 1908 653 739 Fax: +44 1908 655 151 Snail: David Wright, Earth Science Dept., Milton Keynes, England, MK7 6AA Disclaimer: These addresses are only for reaching me, and do not signify official stationery. Views expressed here are either my own or plagiarised.
Re: kerneld message / Workaround
On Sat, Jul 15, 2000 at 11:51:06AM +0100, Jonathan Heaney wrote: David Wright wrote: The scripts /etc/init.d/{kerneld,modutils} have to be able to handle both 2.0 and 2.2 kernels with kerneld or kmod. You will see they do this by testing for the presence of /proc/sys/kernel/modprobe which only exists under 2.2. But doesn't -always- exist on 2.2: [narvi:/etc/init.d] 11:26:37am 136 % ls -l /proc/sys/kernel/modprobe ls: /proc/sys/kernel/modprobe: No such file or directory [narvi:/etc/init.d] 11:26:39am 137 % uname -a Linux narvi 2.2.16 #6 Fri Jun 23 13:51:08 PDT 2000 i686 unknown You need to have 'CONFIG_KMOD' set in your kernel build to have it, which I don't. Seems to me that the logic on that is broken. /proc/sys/kernel/modprobe is not a good way to determine whether kerneld should be run. Perhaps you should file a bug on it? Perhaps you have a problem with your /proc filesystem (unless a bug has been introduced into these scripts). There should be no need to move/remove the scripts in /etc/init.d. Well if there is such a bug it's been there since potato day1 and still is. Incidentally, does the presence of /proc/sys/kernel/modprobe not depend on what you compile in the kernel??? I always roll my own and any time I've installed potato with a 2.2 kernel I've had to shift kerneld. Maybe if you stick with Debians own kernels it might be there but not everyone does. It does. See above. And what about 2.4test kernels (that I'm running) ? I don't have a /proc/sys/kernel/modprobe - what should I do? How do the magic scripts in /etc/init.d deal with that They run kerneld. Wrongly. -- Brian Moore | Of course vi is God's editor. Sysadmin, C/Perl Hacker | If He used Emacs, He'd still be waiting Usenet Vandal | for it to load on the seventh day. Netscum, Bane of Elves.
Re: kerneld message / Workaround
David Wright wrote: The scripts /etc/init.d/{kerneld,modutils} have to be able to handle both 2.0 and 2.2 kernels with kerneld or kmod. You will see they do this by testing for the presence of /proc/sys/kernel/modprobe which only exists under 2.2. Perhaps you have a problem with your /proc filesystem (unless a bug has been introduced into these scripts). There should be no need to move/remove the scripts in /etc/init.d. Well if there is such a bug it's been there since potato day1 and still is. Incidentally, does the presence of /proc/sys/kernel/modprobe not depend on what you compile in the kernel??? I always roll my own and any time I've installed potato with a 2.2 kernel I've had to shift kerneld. Maybe if you stick with Debians own kernels it might be there but not everyone does. And what about 2.4test kernels (that I'm running) ? I don't have a /proc/sys/kernel/modprobe - what should I do? How do the magic scripts in /etc/init.d deal with that J
Re: kerneld message / Workaround
On Thu, Jul 13, 2000 at 23:41:42 +0200, Erik van der Meulen wrote: After the reboot all seems well, except a message during boot: kerneld: you almost certainly don't want to be running kerneld with = 2.2.x Thanks a lot for the various responses. I did a: mv kerneld kerneld.old in /etc/init.d and now things are fine. I checked the ownership of the file, this was: modutils Now I assume that this had been installed as a dependency consequence of the installation of some kernel packages which I installed prior to compiling my 2.2.15 kernel. Not sure which or if I did something odd or if it could be regarded as a bug? Thanks again. -- Erik van der Meulen [EMAIL PROTECTED]
Re: kerneld message / Workaround
Quoting Erik van der Meulen ([EMAIL PROTECTED]): After the reboot all seems well, except a message during boot: kerneld: you almost certainly don't want to be running kerneld with = 2.2.x Thanks a lot for the various responses. I did a: mv kerneld kerneld.old in /etc/init.d and now things are fine. I checked the ownership of the file, this was: modutils Now I assume that this had been installed as a dependency consequence of the installation of some kernel packages which I installed prior to compiling my 2.2.15 kernel. Not sure which or if I did something odd or if it could be regarded as a bug? The scripts /etc/init.d/{kerneld,modutils} have to be able to handle both 2.0 and 2.2 kernels with kerneld or kmod. You will see they do this by testing for the presence of /proc/sys/kernel/modprobe which only exists under 2.2. Perhaps you have a problem with your /proc filesystem (unless a bug has been introduced into these scripts). There should be no need to move/remove the scripts in /etc/init.d. Cheers, -- Email: [EMAIL PROTECTED] Tel: +44 1908 653 739 Fax: +44 1908 655 151 Snail: David Wright, Earth Science Dept., Milton Keynes, England, MK7 6AA Disclaimer: These addresses are only for reaching me, and do not signify official stationery. Views expressed here are either my own or plagiarised.
Re: kerneld message
Erik van der Meulen [EMAIL PROTECTED] wrote I have just compiled my first kernel (!) on my Debian frozen box. It is 2.2.15 After the reboot all seems well, except a message during boot: kerneld: you almost certainly don't want to be running kerneld with = 2.2.x I assume that to be true. I am not sure if this just appears since the kernel update or that I have missed it earlier on. I have not done a lot of tweaking on this machine, so I have no idea what has caused this. From the kerneld manpage kerneld is obsolete as of Linux kernel 2.1.90, it was replaced by the kmod kernel thread and cron entries. Do not even think of using kerneld unless you are running a 2.0 kernel. HTH -- Ragga
Re: kerneld message
Ragga Muffin wrote: Erik van der Meulen [EMAIL PROTECTED] wrote I have just compiled my first kernel (!) on my Debian frozen box. It is 2.2.15 After the reboot all seems well, except a message during boot: kerneld: you almost certainly don't want to be running kerneld with = 2.2.x I assume that to be true. I am not sure if this just appears since the kernel update or that I have missed it earlier on. I have not done a lot of tweaking on this machine, so I have no idea what has caused this. From the kerneld manpage kerneld is obsolete as of Linux kernel 2.1.90, it was replaced by the kmod kernel thread and cron entries. Do not even think of using kerneld unless you are running a 2.0 kernel. HTH -- Ragga You should check your startup scripts. Somewhere, you have enabled starting kerneld, you need to find that place and disable it. -- Bob McGowan Staff Software Quality Engineer VERITAS Software [EMAIL PROTECTED]
Re: kerneld message
You should check your startup scripts. Somewhere, you have enabled starting kerneld, you need to find that place and disable it. cd /etc/init.d mv kerneld kerneld.old reboot Jonathan