Re: [gentoo-user] udev broken...
On Sunday 29 November 2009 03:07:14 BRM wrote: If not, fixing it is quite trivially easy: Get a copy of any recent liveCD or rescue image that you can boot, and boot into it. It will find your drives using whatever conventions it uses, and let you mount your gentoo partitions just like you would do with installs. chroot lets you test stuff and you can also use the compiler on the rescue disk to build a new kernel and store it in /boot Then boot into that new kernel, everything ought to start properly, and immediately rebuild that kernel using your gentoo system compiler. Along the way you might have to edit your fstab to use sda devices instead of hda ones. Thanks! That seems to be a good plan. I built it earlier, but for some reason grub won't boot it - perhaps b/c I gzip compress the kernel (kernel option)? Not sure. Going to figure it out though. More likely you got the chipset drivers wrong. There's been a lot of changes in that area over the past 18 months or so. gzip is the default compression for the kernel, I can't think of any reason why a kernel cannot decompress itself. As a side note: I always keep a rescue USB disk handy in my box of tricks. I use RIPLinux (there are many alternatives) as it supports all imaginable disk hardware, plus software raid, lvm and who knows what else. I keep it up to date with latest current version, this little gadget has saved many a machine from a reinstall and data loss. -- alan dot mckinnon at gmail dot com
Re: [gentoo-user] udev broken...
On Fri, 2009-11-27 at 20:06 -0800, BRM wrote: [... way too much background info removed] BREVITY! We don't want to have to read about what you had on your sandwich for lunch or the fight you had with your girl to get to the meat of what you're trying to say ;-). Since you didn't paste the link to where you found info on the web, I'll go ahead and paste what I found for those who may want to follow up: http://bugs.gentoo.org/281312 To be brief, it basically mentions that udev-145 and kernel-2.6.25 are not compatible. Looks like there is a possible patch (to glibc) but the BZ doesn't really say if/when the patch will be applied in Gentoo. Rather than waiting on a fix for nearly-obsolete software, here is my advice: 1. Boot into a Live CD/stick/whatever. 2. Chroot into your environment as laid out in the Handbook 3. Do an emerge --sync 4. Upgrade *at least* the following: * kernel sources (might as well configure/compile it too, but don't reboot yet) * linux-headers * glibc (do this *after* linux headers) * udev (do this *after* glibc) Reboot and pray. HTH, -a
Re: [gentoo-user] udev broken...
On Saturday 28 November 2009 06:06:59 BRM wrote: So, I have been running my laptop for quite a while with the current software - it's been well over a week since I last synced and installed software - when I upgraded to KDE4; and I do believe I've rebooted several times since. Today, I rebooted back into my old Win2k partition - to do some checking around to clean it up and prep for removal/conversion to a VM image since I've been using gentoo on the laptop for well over a year, and haven't touched the Win2k side for a long time. Having cleaned it up, I rebooted back to gentoo, only to be faced with cascading errors during reboot due to udevd not starting up and mapping the drives, etc. During boot, udevd (version 146, btw) complains about error getting signalfd. I did some basic hunting and this seems to have been a big problem over the last year. I'm running kernel 2.6.25, built on 9/27, from the gentoo source tree. The system then breaks while trying to do some drive mounts, and I end up in maintenance mode - with read-only partitions. Most seem to have resolved the issue by moving back to udev 141. I noticed that newer kernels are suppose to work with it, starting with the 2.6.25; so I _should_ have been okay. Needless to say, right now I'm stuck writing this e-mail from Win2k. The system seems to have the correct partition mounted for the root partition, but it doesn't report it as /dev/hdaX yet. However, I need access to the other partitions to get to portage. (Due to size of portage, and other complications, I've taken to putting it on another partition and mapping it. Usually this hasn't been a problem.) So I think I have a couple options: 1) Figure out how to mount the other partitions, and then revert to an older udev 2) Upgrade to a newer kernel - I do have sources for the 2.6.30 kernels. Either way, I need to figure out how to get read-access to the root partition again. Any advice on either of the above (or other options), and more importantly (since any options depend on it) how to get read-write access to the root partition again? Actual error messages please :-) Let's start with what is in dmesg. You also mention /dev/hda and the context implies it is a physical disk. Unless you have ancient disk hardware and unusual module setup, your disks will be /dev/sda. Do you have references to /dev/dh** in /etc/fstab? That won;t work as udev will not name them that way -- alan dot mckinnon at gmail dot com
Re: [gentoo-user] udev broken...
- Original Message From: Alan McKinnon alan.mckin...@gmail.com On Saturday 28 November 2009 06:06:59 BRM wrote: During boot, udevd (version 146, btw) complains about error getting signalfd. I did some basic hunting and this seems to have been a big problem over the last year. I'm running kernel 2.6.25, built on 9/27, from the gentoo source tree. The system then breaks while trying to do some drive mounts, and I end up in maintenance mode - with read-only partitions. snip The system seems to have the correct partition mounted for the root partition, but it doesn't report it as /dev/hdaX yet. However, I need access to the other partitions to get to portage. (Due to size of portage, and other complications, I've taken to putting it on another partition and mapping it. Usually this hasn't been a problem.) Actual error messages please :-) Let's start with what is in dmesg. dmesg doesn't provide any useful information for this error. Nor could I find anything in the /var/log/messages. The only error I got was the error getting signalfd from udev's startup at boot. You also mention /dev/hda and the context implies it is a physical disk. Unless you have ancient disk hardware and unusual module setup, your disks will be /dev/sda. Do you have references to /dev/dh** in /etc/fstab? That won;t work as udev will not name them that way Actually, yes - it is a 2003 Dell D600 with a standard ATA/IDE hard drive. So yes - it would be /dev/hda; and yes, udev has been working fine until this issue. Ben
Re: [gentoo-user] udev broken...
On Saturday 28 November 2009 17:04:10 BRM wrote: You also mention /dev/hda and the context implies it is a physical disk. Unless you have ancient disk hardware and unusual module setup, your disks will be /dev/sda. Do you have references to /dev/dh** in /etc/fstab? That won;t work as udev will not name them that way Actually, yes - it is a 2003 Dell D600 with a standard ATA/IDE hard drive. So yes - it would be /dev/hda; and yes, udev has been working fine until this issue. For quite some time now IDE drives have been handled below the SCSI subsytem so you do in fact get a /dev/sda, except when using the old deprectaed IDE driver that has been around for ages. That one uses /dev/hda, and it's very unusual these days to find it. You should check what the kernek you are running is using and what udev calls those things as it very likely is not the same as what it was before your kernel udev upgrade. I want to eliminate obvious things before we go looking for exotic things -- alan dot mckinnon at gmail dot com
Re: [gentoo-user] udev broken...
- Original Message From: Alan McKinnon alan.mckin...@gmail.com On Saturday 28 November 2009 17:04:10 BRM wrote: You also mention /dev/hda and the context implies it is a physical disk. Unless you have ancient disk hardware and unusual module setup, your disks will be /dev/sda. Do you have references to /dev/dh** in /etc/fstab? That won;t work as udev will not name them that way Actually, yes - it is a 2003 Dell D600 with a standard ATA/IDE hard drive. So yes - it would be /dev/hda; and yes, udev has been working fine until this issue. For quite some time now IDE drives have been handled below the SCSI subsytem so you do in fact get a /dev/sda, except when using the old deprectaed IDE driver that has been around for ages. That one uses /dev/hda, and it's very unusual these days to find it. You should check what the kernek you are running is using and what udev calls those things as it very likely is not the same as what it was before your kernel udev upgrade. Okay - booted back over to it to do some checking: - trying to use /dev/sda1 as the root device (kernel command-line) won't work. - exact kernel version: 2.6.25-gentoo-r7 - there are no drives (hda, sda, etc.) listed under /dev - kind of expected since udevd isn't running. I do have sources for linux kernel 2.6.30-gentoo-r8 available, but then I need to be able to write to the read-only fs. Guess I could probably do that using the kernel command-line, no? (Haven't done that before, so I'm not sure what the correct option would be.) I want to eliminate obvious things before we go looking for exotic things Sounds like a good plan. TIA, Ben
Re: [gentoo-user] udev broken...
On Saturday 28 November 2009 18:31:04 BRM wrote: - Original Message From: Alan McKinnon alan.mckin...@gmail.com On Saturday 28 November 2009 17:04:10 BRM wrote: You also mention /dev/hda and the context implies it is a physical disk. Unless you have ancient disk hardware and unusual module setup, your disks will be /dev/sda. Do you have references to /dev/dh** in /etc/fstab? That won;t work as udev will not name them that way Actually, yes - it is a 2003 Dell D600 with a standard ATA/IDE hard drive. So yes - it would be /dev/hda; and yes, udev has been working fine until this issue. For quite some time now IDE drives have been handled below the SCSI subsytem so you do in fact get a /dev/sda, except when using the old deprectaed IDE driver that has been around for ages. That one uses /dev/hda, and it's very unusual these days to find it. You should check what the kernek you are running is using and what udev calls those things as it very likely is not the same as what it was before your kernel udev upgrade. Okay - booted back over to it to do some checking: - trying to use /dev/sda1 as the root device (kernel command-line) won't work. - exact kernel version: 2.6.25-gentoo-r7 - there are no drives (hda, sda, etc.) listed under /dev - kind of expected since udevd isn't running. I do have sources for linux kernel 2.6.30-gentoo-r8 available, but then I need to be able to write to the read-only fs. Guess I could probably do that using the kernel command-line, no? (Haven't done that before, so I'm not sure what the correct option would be.) Before these troubles started, did you build a 2.6.30 kernel? If so, you can just boot it, editing the grub command line at boot time as necessary. If not, fixing it is quite trivially easy: Get a copy of any recent liveCD or rescue image that you can boot, and boot into it. It will find your drives using whatever conventions it uses, and let you mount your gentoo partitions just like you would do with installs. chroot lets you test stuff and you can also use the compiler on the rescue disk to build a new kernel and store it in /boot Then boot into that new kernel, everything ought to start properly, and immediately rebuild that kernel using your gentoo system compiler. Along the way you might have to edit your fstab to use sda devices instead of hda ones. btw, this is exactly the reason why user-oriented distros like Ubuntu mount system partitions using the fs GUID, not the kernel device name. It gets around this kind of trouble quite elegantly -- alan dot mckinnon at gmail dot com
Re: [gentoo-user] udev broken...
On Fri, 27 Nov 2009 20:06:59 -0800 (PST) BRM wrote: ...[snip]... Either way, I need to figure out how to get read-access to the root partition again. Any advice on either of the above (or other options), and more importantly (since any options depend on it) how to get read-write access to the root partition again? I've encountered the root is read-only and I need read-write problem. My solution is the script below. #!/bin/sh sync /bin/mount -o remount,rw / /bin/mount -o remount,rw /boot Of course you can type the commands by hand. Since you've only got one partition mounted and it's ro, you don't need the sync. HTH, David
Re: [gentoo-user] udev broken...
- Original Message From: Alan McKinnon alan.mckin...@gmail.com On Saturday 28 November 2009 18:31:04 BRM wrote: snip I do have sources for linux kernel 2.6.30-gentoo-r8 available, but then I need to be able to write to the read-only fs. Guess I could probably do that using the kernel command-line, no? (Haven't done that before, so I'm not sure what the correct option would be.) Before these troubles started, did you build a 2.6.30 kernel? If so, you can just boot it, editing the grub command line at boot time as necessary. Unfortunately not. I have been thinking lately that I should upgrade to a new kernel - but I don't get around to it very often. So this is probably a good opportunity to do so. If not, fixing it is quite trivially easy: Get a copy of any recent liveCD or rescue image that you can boot, and boot into it. It will find your drives using whatever conventions it uses, and let you mount your gentoo partitions just like you would do with installs. chroot lets you test stuff and you can also use the compiler on the rescue disk to build a new kernel and store it in /boot Then boot into that new kernel, everything ought to start properly, and immediately rebuild that kernel using your gentoo system compiler. Along the way you might have to edit your fstab to use sda devices instead of hda ones. Thanks! That seems to be a good plan. I built it earlier, but for some reason grub won't boot it - perhaps b/c I gzip compress the kernel (kernel option)? Not sure. Going to figure it out though. Right now, I'm using a vintage 2007 live CD; but chrooting over to the partitions on the hard drive. I don't have network since it won't recognize the firmware needed for my wireless (b43legacy driver). Perhaps I'll try a newer image... btw, this is exactly the reason why user-oriented distros like Ubuntu mount system partitions using the fs GUID, not the kernel device name. It gets around this kind of trouble quite elegantly May solve some headaches, but it also creates an equal number of others - like identifying the partition that matches the GUID. Thanks! Ben
[gentoo-user] udev broken...
So, I have been running my laptop for quite a while with the current software - it's been well over a week since I last synced and installed software - when I upgraded to KDE4; and I do believe I've rebooted several times since. Today, I rebooted back into my old Win2k partition - to do some checking around to clean it up and prep for removal/conversion to a VM image since I've been using gentoo on the laptop for well over a year, and haven't touched the Win2k side for a long time. Having cleaned it up, I rebooted back to gentoo, only to be faced with cascading errors during reboot due to udevd not starting up and mapping the drives, etc. During boot, udevd (version 146, btw) complains about error getting signalfd. I did some basic hunting and this seems to have been a big problem over the last year. I'm running kernel 2.6.25, built on 9/27, from the gentoo source tree. The system then breaks while trying to do some drive mounts, and I end up in maintenance mode - with read-only partitions. Most seem to have resolved the issue by moving back to udev 141. I noticed that newer kernels are suppose to work with it, starting with the 2.6.25; so I _should_ have been okay. Needless to say, right now I'm stuck writing this e-mail from Win2k. The system seems to have the correct partition mounted for the root partition, but it doesn't report it as /dev/hdaX yet. However, I need access to the other partitions to get to portage. (Due to size of portage, and other complications, I've taken to putting it on another partition and mapping it. Usually this hasn't been a problem.) So I think I have a couple options: 1) Figure out how to mount the other partitions, and then revert to an older udev 2) Upgrade to a newer kernel - I do have sources for the 2.6.30 kernels. Either way, I need to figure out how to get read-access to the root partition again. Any advice on either of the above (or other options), and more importantly (since any options depend on it) how to get read-write access to the root partition again? TIA, Ben
