[gentoo-user] Re: Best way to copy /* ?
Am 27.12.2010 22:23, schrieb Remy Blank: > Marc Blumentritt wrote: >> Have a look at ownership an permissions of virtualbox! > > rsync sets the mtime and permissions of directories *after* recursing > into them. So if you interrupted the process while it was copying the > contents of virtualbox, the permissions of that directory haven't been > set yet. Ah, OK. Didn't know this. Do you know, why it does work like this? And does cp -a work in the same way? Regards, Marc
[gentoo-user] Re: Best way to copy /* ?
Marc Blumentritt wrote: > Ah, OK. Didn't know this. Do you know, why it does work like this? It initially sets the permissions to 600 presumably to limit access to an unfinished copy from other users. The mtime must be set after copying the contents, because adding a file or directory changes the mtime of the parent. > And does cp -a work in the same way? Probably yes, for the same reasons, but I haven't checked. -- Remy signature.asc Description: OpenPGP digital signature
Re: [gentoo-user] Re: Best way to copy /* ?
On Tue, 28 Dec 2010 10:18:46 +0100, Remy Blank wrote: > > Ah, OK. Didn't know this. Do you know, why it does work like this? > > It initially sets the permissions to 600 presumably to limit access to > an unfinished copy from other users. > > The mtime must be set after copying the contents, because adding a file > or directory changes the mtime of the parent. > > > And does cp -a work in the same way? > > Probably yes, for the same reasons, but I haven't checked. No it doesn't. cp leaves the mtime of the directory at the time t copied the files into it. If you want to preserve directory timestamps, use rsync. -- Neil Bothwick The number one cause of computer problems is computer solutions. signature.asc Description: PGP signature
Re: [gentoo-user] Is there a standard sysctl-like way to modify sysfs files at boot time?
On Monday 27 December 2010 11:37:29 Mark David Dumlao wrote: > I want to do this: > http://blog.internetnews.com/skerner/2010/11/forget-200-lines-red-hat-speed > .html > > in userspace, but automate it at boot time. it requires that I create and > mount the cgroup subsystem in sysfs and sounds a lot like something that > I'd do in sysctl for /proc/sys, but for sysfs rather than procfs. > > The only thing that comes to mind is to append to the local init script, > but it's so close to what sysctl does that I feel like someone's probably > written some tool for it. Is there one? If there is, I haven't seen it. Perhaps you can try adding it to /etc/conf.d/local.start and report back if it behaves as expected? -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] kdm keymap
On Friday 24 December 2010 14:12:17 alain.didierj...@free.fr wrote: > I use hal. No more xorg.conf. > Since I updated kdm to v.4.4.5 it use a QWERTY keymap, when it worked OK in > previous 4.4.* version after modifying > /etc/hal/fdi/policy/10-x11-input.fdi. I try adding the following 2 lines > also in > /usr/share/hal/fdi/policy/10osvendor/10-x11-input.fdi, just in case : > key="input.xkb.options"type="string">terminate:ctrl_alt_bksp > fr > > Of course I also checked french as language in the kdl config panel. > Obviously, entering a password using an AZERTY keyboard & a QWERTY keymap > is not that easy. > What can I do ? Couldn't find anything in the doc. > And a happy Xmas to you all. If you haven't yet fixed this, install the latest stable xorg-server and run Xorg -configure to create a new xorg.conf file. Without HAL the .fdi files are no longer the way to configure input devices. Depending on your hardware you will also need to configure your kernel to use KMS. Have a look at this page for more details: http://www.gentoo.org/doc/en/xorg-config.xml -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] Best way to copy /* ?
Mike Edenfield writes: > The tar method you're looking for is: > > tar -C /old cpf - | tar -C /new xvpf - > > You'll probably not want to do the entire / in a single go, > since /proc, /sys, and /dev (at least) should be skipped. > Copy /old/sbin -> /new/sbin, etc. for all of the root > folders that aren't their own partitions. The rest you can > do the entire mount point at once, though I'm not sure you > really need to copy /tmp either. Or bind mount root to somewhere else: mount -o bind / /new /new then contains all of / , but without the filesystems mounted at root. It also has has the original /dev with its necessary entries console and null, but without the stuff udev added. It also takes care of /tmp. Its contents are unimportant, but the permissions of the directory itself matter, the sticky bit needs to be set. Wonko
Re: [gentoo-user] Best way to copy /* ?
On Tue, 28 Dec 2010 00:02:31 -0500, Mike Edenfield wrote: > tar -C /old cpf - | tar -C /new xvpf - > > You'll probably not want to do the entire / in a single go, > since /proc, /sys, and /dev (at least) should be skipped. Which can be done with the -l option. -- Neil Bothwick You can't teach a new mouse old clicks. signature.asc Description: PGP signature
[gentoo-user] Options null for ext4 as root fs
After multiple fs corruptions and nothing untoward shown by smartctl (I think) I decided to move to ext4. Now when I boot I see a message about my / fs which says Opts: (null) dmesg | grep -i EXT4 Command line: root=/dev/sda7 rootfstype=ext4 Kernel command line: root=/dev/sda7 rootfstype=ext4 EXT4-fs (sda7): mounted filesystem with ordered data mode. Opts: (null) VFS: Mounted root (ext4 filesystem) readonly on device 8:7. EXT4-fs (sda7): re-mounted. Opts: (null) EXT4-fs (sda6): mounted filesystem with ordered data mode. Opts: (null) EXT4-fs (sda8): mounted filesystem with ordered data mode. Opts: (null) EXT4-fs (sda9): mounted filesystem with ordered data mode. Opts: (null) EXT4-fs (sda7): re-mounted. Opts: commit=0 EXT4-fs (sda6): re-mounted. Opts: commit=0 EXT4-fs (sda8): re-mounted. Opts: commit=0 EXT4-fs (sda9): re-mounted. Opts: commit=0 EXT4-fs (sda7): re-mounted. Opts: commit=0 EXT4-fs (sda6): re-mounted. Opts: commit=0 EXT4-fs (sda8): re-mounted. Opts: commit=0 EXT4-fs (sda9): re-mounted. Opts: commit=0 What is the meaning of Opts: (null) ? All ext4 partitions were created with a number of options; e.g. Filesystem features: has_journal ext_attr resize_inode dir_index filetype needs_recovery extent flex_bg sparse_super large_file uninit_bg dir_nlink extra_isize but default mount options are shown as: Default mount options:(none) Should I be defining something in fstab? (currently I only have noatime). -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] Re: [EXAMPLE] Configuring xorg without hal
On 12/27/2010 03:52 PM, walt wrote: > On 12/27/2010 06:03 AM, walt wrote: >> ... >> My new (post-hal) mouse config: >> >> Section "InputClass" <- note the new word "Class", not "Device" >> Identifier "trackball" <- can be anything you want >> MatchProduct "ImExPS" <-*new*. Matches the product name! > > Clarification about that MatchProduct keyword: > > Hal and udev differ slightly in the way they identify hardware devices > (is anyone surprised?). > > When I run 'lshal' to display all my hardware, I see this about the mouse: > > info.product = 'ImExPS/2 Logitech Explorer Mouse' > > Now for the confusion. When I run 'udevadm info --export-db" I see: > > E: PRODUCT=11/2/6/6d > E: NAME="ImExPS/2 Logitech Explorer Mouse" > > Note that MatchProduct in the xorg.conf file really wants the udev NAME, > *not* the udev PRODUCT. > > I suppose the MatchProduct keyword was selected back in the day when > hal was still the boss instead of a has-been. "MatchName" would be much > less confusing than "MatchProduct". But don't hold your breath waiting > for it to change. Thanks, Walt, this is great. It's good to see the potholes in the road before the wheel gets in them. These are the kinds of tidbits of information that you really want to find when you are looking for them once you're *in* the pothole, but keeping up with the list and with input from the great people here, it makes the ride much less bumpy.
Re: [gentoo-user] gentoo-sources-2.6.35-r12 causes kernel panic
On Saturday 18 December 2010 14:40:07 Mick wrote: > On 30 November 2010 11:11, Peter Humphrey wrote: > > On Monday 29 November 2010 18:20:56 Mick wrote: > >> Will wait for 2.6.36 series to see if this old PIII will work. > > > > I'm running 2.6.36-r3 at the moment. You only have to add a keyword to > > gentoo-sources. > > Just compiled gentoo-2.6.36-r5. > > Unfortunately, I'm no closer to getting running kernel! :-( > = > ERROR: Unable to locate IOAPIC for GSI 4 > ERROR: Unable to locate IOAPIC for GSI 6 > > ERROR: Unable to locate IOAPIC for GSI 4 > ERROR: Unable to locate IOAPIC for GSI 6 > ERROR: Unable to locate IOAPIC for GSI 3 > ERROR: Unable to locate IOAPIC for GSI 7 > ERROR: Unable to locate IOAPIC for GSI 13 > ERROR: Unable to locate IOAPIC for GSI 8 > ERROR: Unable to locate IOAPIC for GSI 1 > ERROR: Unable to locate IOAPIC for GSI 12 > > kernel oanic -not syncing: VFS: Unable to mount root fs on unknown > block(0,0) Pid: 1, comm: swapper Not taineted 2.6.36-gentoo-r5 > Call trace: > > (some trace messages which contain): > > panic > mount_block_root > kernel_init > prepare_namespace > sys_access > kernel_init > kernel_thread_helper > = > > Any ideas? What a muppet! I had the old root path in GRUB /dev/hda3, instead of the new /dev/sda3 that the new kernel drivers now read. Still getting the errors about "Unable to locate IOAPIC for GSI 4", but they seem to be harmless. Sorry for the noise! -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] kdm keymap
On Friday 24 December 2010 15:12:17 alain.didierj...@free.fr wrote: > I use hal. No more xorg.conf. > Since I updated kdm to v.4.4.5 it use a QWERTY keymap, when it worked OK in > previous 4.4.* version after modifying /etc/hal/fdi/policy/10-x11-input.fdi. > I try adding the following 2 lines also in > /usr/share/hal/fdi/policy/10osvendor/10-x11-input.fdi, just in case : > key="input.xkb.options"type="string">terminate:ctrl_alt_bksp key="input.xkb.layout" type="string">fr > > Of course I also checked french as language in the kdl config panel. > Obviously, entering a password using an AZERTY keyboard & a QWERTY keymap is > not that easy. > What can I do ? Couldn't find anything in the doc. > And a happy Xmas to you all. had a similar problem, finally I did something stupid to solve it: I put setxkbmap de & into /usr/share/config/kdm/Xsetup
Re: [gentoo-user] Best way to copy /* ?
Neil Bothwick wrote: > On Tue, 28 Dec 2010 00:02:31 -0500, Mike Edenfield wrote: > > > tar -C /old cpf - | tar -C /new xvpf - > > > > You'll probably not want to do the entire / in a single go, > > since /proc, /sys, and /dev (at least) should be skipped. > > Which can be done with the -l option. You are mistaken: The -l option causes tar to warn if not all hars links to a file could not be resolved. Jörg -- EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin j...@cs.tu-berlin.de(uni) joerg.schill...@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily
Re: [gentoo-user] Options null for ext4 as root fs
On Tue, Dec 28, 2010 at 7:50 AM, Mick wrote: > What is the meaning of Opts: (null) ? My guess is that it is showing the default mount options as stored in the partition's superblock (set by tune2fs -o x). You can view the current default mount options by using "tune2fs -l /dev/sda1" (or whatever your partition is). Options given on mount commandline (or in fstab) should override the superblock mount options. My dmesg shows (null) as yours does, and tune2fs shows me: Default mount options:(none) > All ext4 partitions were created with a number of options; e.g. > > Filesystem features: has_journal ext_attr resize_inode dir_index filetype > needs_recovery extent flex_bg sparse_super large_file uninit_bg dir_nlink > extra_isize Those are characteristics of your filesystem but I don't think they are considered mount options in this context. > but default mount options are shown as: > > Default mount options: (none) > > Should I be defining something in fstab? (currently I only have noatime). You can view the actual mount options currently in use by doing "cat /proc/mounts" In my fstab I have "defaults,noatime" for my rootfs and in /proc/mounts it shows as: /dev/root / ext4 rw,noatime,barrier=1,data=ordered 0 0 which looks correct to me.
Re: [gentoo-user] Best way to copy /* ?
On Tuesday 28 December 2010 16:26:16 Joerg Schilling wrote: > Neil Bothwick wrote: > > On Tue, 28 Dec 2010 00:02:31 -0500, Mike Edenfield wrote: > > > tar -C /old cpf - | tar -C /new xvpf - > > > > > > You'll probably not want to do the entire / in a single go, > > > since /proc, /sys, and /dev (at least) should be skipped. > > > > Which can be done with the -l option. > > You are mistaken: The -l option causes tar to warn if not all hars links to > a file could not be resolved. I often use this with a LiveCD: cd /home tar --exclude File1 --exclude Dir1 -lcpvSf - . | (cd /mnt/new_partition; tar - xpvf - ) or something like this with star: star -copy -p -xdot -xattr -H=exustar -sparse -M -C /home . /mnt/new_partition (You can use -V -pat=File1 to exclude files or directories with star, use the -M option to avoid following mount points). -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] "
On 12/28/10 20:01:19, Daniel D Jones wrote:
> Getting the following error:
>
> [blocks B ] (" settings-256.52" is blocking x11-drivers/nvidia-drivers-260.19.29)
>
> Total: 150 packages (148 upgrades, 1 new, 1 reinstall), Size of
> downloads:
> 87,737 kB
> Conflict: 2 blocks (1 unsatisfied)
>
> * Error: The above package list contains packages which cannot be
> * installed at the same time on the same system.
>
> (media-video/nvidia-settings-195.36.24, installed) pulled in by
> media-video/nvidia-settings required by (x11-drivers/nvidia-
> drivers-260.19.29, ebuild scheduled for merge)
> media-video/nvidia-settings required by @selected
>
> (x11-drivers/nvidia-drivers-260.19.29, ebuild scheduled for merge)
> pulled in
> by
> x11-drivers/nvidia-drivers required by @selected
> x11-drivers/nvidia-drivers required by (media-video/nvidia-
> settings-195.36.24, installed)
>
>
> Is this just a temporary glitch with the packages that will fix
> itself
> when an
> updated version of nvidia-settings is added to portage or is this
> something
> wrong on my system?
It's a bit strange. nvidia-drivers-260.19.29 is masked unless you
have ACCEPT_KEYWORDS="~amd64" (or ~x86) But in that case you should be
able to emerge media-video/nvidia-settings-260.19.29, as well.
Have you masked/ unmasked only one of them?
Helmut.
Re: [gentoo-user] "
On Tue, Dec 28, 2010 at 1:01 PM, Daniel D Jones wrote: > Getting the following error: > > [blocks B ] settings-256.52" is blocking x11-drivers/nvidia-drivers-260.19.29) > > Total: 150 packages (148 upgrades, 1 new, 1 reinstall), Size of downloads: > 87,737 kB > Conflict: 2 blocks (1 unsatisfied) > > * Error: The above package list contains packages which cannot be > * installed at the same time on the same system. > > (media-video/nvidia-settings-195.36.24, installed) pulled in by > media-video/nvidia-settings required by (x11-drivers/nvidia- > drivers-260.19.29, ebuild scheduled for merge) > media-video/nvidia-settings required by @selected > > (x11-drivers/nvidia-drivers-260.19.29, ebuild scheduled for merge) pulled in > by > x11-drivers/nvidia-drivers required by @selected > x11-drivers/nvidia-drivers required by (media-video/nvidia- > settings-195.36.24, installed) > > > Is this just a temporary glitch with the packages that will fix itself when an > updated version of nvidia-settings is added to portage or is this something > wrong on my system? I have version 260.19.29 of both in my portage tree, maybe try to emerge --sync. I'm using ~amd64 and emerged both of them on 20th of December.
Re: [gentoo-user] Options null for ext4 as root fs
On Tuesday 28 December 2010 17:23:22 Paul Hartman wrote: > On Tue, Dec 28, 2010 at 7:50 AM, Mick wrote: > > What is the meaning of Opts: (null) ? > > My guess is that it is showing the default mount options as stored in > the partition's superblock (set by tune2fs -o x). You can view the > current default mount options by using "tune2fs -l /dev/sda1" (or > whatever your partition is). Options given on mount commandline (or in > fstab) should override the superblock mount options. > > My dmesg shows (null) as yours does, and tune2fs shows me: > Default mount options:(none) OK, same as mine. > > All ext4 partitions were created with a number of options; e.g. > > > > Filesystem features: has_journal ext_attr resize_inode dir_index > > filetype needs_recovery extent flex_bg sparse_super large_file uninit_bg > > dir_nlink extra_isize > > Those are characteristics of your filesystem but I don't think they > are considered mount options in this context. Yes, that's right, but they affect some mount options if the latter are explicitly set at fstab. > You can view the actual mount options currently in use by doing "cat > /proc/mounts" > > In my fstab I have "defaults,noatime" for my rootfs and in > /proc/mounts it shows as: > > /dev/root / ext4 rw,noatime,barrier=1,data=ordered 0 0 > > which looks correct to me. Same here: /dev/root / ext4 rw,noatime,barrier=1,data=ordered 0 0 Is that all that is required for ext4? Do we need to define anything else to take advantage of the various ext4 options (see below)? http://www.mjmwired.net/kernel/Documentation/filesystems/ext4.txt#125 -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] "
On Tue, Dec 28, 2010 at 1:49 PM, Paul Hartman wrote: > On Tue, Dec 28, 2010 at 1:01 PM, Daniel D Jones > wrote: >> Getting the following error: >> >> [blocks B ] > settings-256.52" is blocking x11-drivers/nvidia-drivers-260.19.29) It appears that 260.19.29 nvidia- drivers and nvidia-settings are both stable on amd64, but on x86 the nvidia-drivers is stable but nvidia-settings is ~x86. So if you're on x86 maybe that's the problem. I would hope nvidia-settings is stabilized soon for x86 because there are already a couple bugreports in BGO about it.
Re: [gentoo-user] Best way to copy /* ?
On Tue, 28 Dec 2010 17:26:16 +0100, Joerg Schilling wrote: > > Which can be done with the -l option. > > You are mistaken: The -l option causes tar to warn if not all hars > links to a file could not be resolved. You are right. I was referring to the --one-file-system, which I thought had -l as the short option, but apparently not. Or has that changed in the last couple of years? -- Neil Bothwick I'm Bugs Bunny of Borg. What's up Collective? signature.asc Description: PGP signature
Re: [gentoo-user] Best way to copy /* ?
Joerg Schilling writes: > Neil Bothwick wrote: > > On Tue, 28 Dec 2010 00:02:31 -0500, Mike Edenfield wrote: > > > tar -C /old cpf - | tar -C /new xvpf - > > > > > > You'll probably not want to do the entire / in a single go, > > > since /proc, /sys, and /dev (at least) should be skipped. > > > > Which can be done with the -l option. > > You are mistaken: The -l option causes tar to warn if not all hars links > to a file could not be resolved. We're one the Gentoo Linux list here and so we're talking about GNU tar, where -l is a synonym for --one-file-system, Neil is right. Original Unix tar works like you say, but we don't use it here. Wonko
Re: [gentoo-user] Best way to copy /* ?
On Tuesday 28 December 2010 22:33:53 Neil Bothwick wrote: > On Tue, 28 Dec 2010 17:26:16 +0100, Joerg Schilling wrote: > > > Which can be done with the -l option. > > > > You are mistaken: The -l option causes tar to warn if not all hars > > links to a file could not be resolved. > > You are right. I was referring to the --one-file-system, which I thought > had -l as the short option, but apparently not. Or has that changed in > the last couple of years? Yes, there used to be a warning on the man page saying that --one-file-system should be be used instead of -l. -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] Best way to copy /* ?
Neil Bothwick wrote: > On Tue, 28 Dec 2010 17:26:16 +0100, Joerg Schilling wrote: > > > > Which can be done with the -l option. > > > > You are mistaken: The -l option causes tar to warn if not all hars > > links to a file could not be resolved. > > You are right. I was referring to the --one-file-system, which I thought > had -l as the short option, but apparently not. Or has that changed in > the last couple of years? Tar did not change it's behavior with respect to -l during the past 31 years. On Linux, there is frequently gtar installed as tar and gtar is not respecting standards. Gtar in previous times was e.g. in conflict with the standard regarding to -l. Aprox. 10 years ago, I files a bug report against gtar for this standard deviation and it seems that they fixed this bug. Jörg -- EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin j...@cs.tu-berlin.de(uni) joerg.schill...@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily
Re: [gentoo-user] Best way to copy /* ?
Joerg Schilling writes: > On Linux, there is frequently gtar installed as tar and gtar is not > respecting standards. Gtar in previous times was e.g. in conflict with > the standard regarding to -l. Aprox. 10 years ago, I files a bug report > against gtar for this standard deviation and it seems that they fixed > this bug. Oh, I stand corrected, you're right. I looked at the (german) man page, which has not been updated yet. I sent a mail to bug-...@gnu.org about this. Wonko
Re: [gentoo-user] Is there a standard sysctl-like way to modify sysfs files at boot time?
> On Monday 27 December 2010 11:37:29 Mark David Dumlao wrote: > > I want to do this: > > > http://blog.internetnews.com/skerner/2010/11/forget-200-lines-red-hat-speed > > .html > > > > in userspace, but automate it at boot time. it requires that I create and > > mount the cgroup subsystem in sysfs and sounds a lot like something that > > I'd do in sysctl for /proc/sys, but for sysfs rather than procfs. > > > > The only thing that comes to mind is to append to the local init script, > > but it's so close to what sysctl does that I feel like someone's probably > > written some tool for it. Is there one? > > Have you read; http://en.gentoo-wiki.com/wiki/Improve_responsiveness_with_cgroups
Re: [gentoo-user] Is there a standard sysctl-like way to modify sysfs files at boot time?
On Monday 27 December 2010 19:37:29 Mark David Dumlao wrote: > I want to do this: > http://blog.internetnews.com/skerner/2010/11/forget-200-lines-red-hat-speed. > html > > in userspace, but automate it at boot time. it requires that I create and > mount the cgroup subsystem in sysfs and sounds a lot like something that I'd > do in sysctl for /proc/sys, but for sysfs rather than procfs. > > The only thing that comes to mind is to append to the local init script, but > it's so close to what sysctl does that I feel like someone's probably > written some tool for it. Is there one? why? why not just patch the kernel? or wait for 2.6.37? Why trying the easily broken userspace approach? btw - patch or userspace - what happens with apps not started from a shell?
