[arch-dev-public] [signoff] udev 157-1, initscripts 2010.06-1, mkinitcpio 0.6.5-1
You should upgrade and test these together (don't forget to regenerate initramfs when you're upgraded). The new udev should work without the new initscripts and mkinitcpio, but udev-based network device renaming will fail in that case. udev is an upstream update with mostly bugfixes, but some changes in behaviour. Most importantly, NAME= rules are being ignored now, you can not change the kernel's predefined device name anymore. initscripts changes: Dan McGee (1): Include _netdev option in NETFS list when calling fsck Thomas Bächler (2): Only mount /proc, /sys and /dev if they haven't been mounted already, use devtmpfs instead of tmpfs if supported by the kernel Add --action=add to udevadm trigger mkinitcpio changes: Thomas Bächler (5): Do not umount /proc and /sys before switch_root, but mount --move them into the real root Mount tmpfs or (if supported) devtmpfs on /dev, move it into the real root before switch_root udev hook: Add --action=add to udevadm trigger udev hook: Add ata_id, path_id, scsi_id and usb_id to allow complete persistent storage rules Release 0.6.5 For the future, we now fully support devtmpfs. It is recommended (but not required (yet)) by upstream. Using devtmpfs gets rid of a small number of race conditions. I tested devtmpfs on a custom kernel, and with the exception of one short (and harmless) error message (mknod: /dev/rtc0: device exists), it works fine. Getting rid of that error message requires further consideration, as we can fix a bug with some custom kernel configurations at the same time if we do it right. For now, I will ignore this and please ask for testing and signoffs.
On Thu, Jun 3, 2010 at 7:57 AM, Thomas Bächler <thomas@archlinux.org> wrote:
You should upgrade and test these together (don't forget to regenerate initramfs when you're upgraded). The new udev should work without the new initscripts and mkinitcpio, but udev-based network device renaming will fail in that case.
udev is an upstream update with mostly bugfixes, but some changes in behaviour. Most importantly, NAME= rules are being ignored now, you can not change the kernel's predefined device name anymore.
initscripts changes: Dan McGee (1): Include _netdev option in NETFS list when calling fsck
Thomas Bächler (2): Only mount /proc, /sys and /dev if they haven't been mounted already, use devtmpfs instead of tmpfs if supported by the kernel Add --action=add to udevadm trigger
mkinitcpio changes: Thomas Bächler (5): Do not umount /proc and /sys before switch_root, but mount --move them into the real root Mount tmpfs or (if supported) devtmpfs on /dev, move it into the real root before switch_root udev hook: Add --action=add to udevadm trigger udev hook: Add ata_id, path_id, scsi_id and usb_id to allow complete persistent storage rules Release 0.6.5
For the future, we now fully support devtmpfs. It is recommended (but not required (yet)) by upstream. Using devtmpfs gets rid of a small number of race conditions.
I tested devtmpfs on a custom kernel, and with the exception of one short (and harmless) error message (mknod: /dev/rtc0: device exists), it works fine. Getting rid of that error message requires further consideration, as we can fix a bug with some custom kernel configurations at the same time if we do it right. For now, I will ignore this and please ask for testing and signoffs.
Signoff on the whole deal with stock arch kernel on x86_64, so no devtmpfs in use currently. Everything seemed smooth and I didn't see any unexpected errors scroll by. -Dan
On Thu, Jun 3, 2010 at 17:18, Dan McGee <dpmcgee@gmail.com> wrote:
On Thu, Jun 3, 2010 at 7:57 AM, Thomas Bächler <thomas@archlinux.org> wrote:
You should upgrade and test these together (don't forget to regenerate initramfs when you're upgraded). The new udev should work without the new initscripts and mkinitcpio, but udev-based network device renaming will fail in that case.
udev is an upstream update with mostly bugfixes, but some changes in behaviour. Most importantly, NAME= rules are being ignored now, you can not change the kernel's predefined device name anymore.
Signoff on the whole deal with stock arch kernel on x86_64, so no devtmpfs in use currently. Everything seemed smooth and I didn't see any unexpected errors scroll by.
Booted fine under Xen using stock Arch kernel (via pv-grub), so I signoff x86_64. -- Roman Kyrylych (Роман Кирилич)
On Thu, Jun 3, 2010 at 18:36, Roman Kyrylych <roman.kyrylych@gmail.com> wrote:
On Thu, Jun 3, 2010 at 17:18, Dan McGee <dpmcgee@gmail.com> wrote:
On Thu, Jun 3, 2010 at 7:57 AM, Thomas Bächler <thomas@archlinux.org> wrote:
You should upgrade and test these together (don't forget to regenerate initramfs when you're upgraded). The new udev should work without the new initscripts and mkinitcpio, but udev-based network device renaming will fail in that case.
udev is an upstream update with mostly bugfixes, but some changes in behaviour. Most importantly, NAME= rules are being ignored now, you can not change the kernel's predefined device name anymore.
Signoff on the whole deal with stock arch kernel on x86_64, so no devtmpfs in use currently. Everything seemed smooth and I didn't see any unexpected errors scroll by.
Booted fine under Xen using stock Arch kernel (via pv-grub), so I signoff x86_64.
Noticed the following: +if ! /bin/mountpoint -q /dev; then + if grep -q devtmpfs /proc/filesystems 2>/dev/null; then + /bin/mount -n -t devtmpfs udev /dev -o mode=0755,size=10M,nosuid 1) grep should be /bin/grep 2) /dev/null does not exist at that time (if I understand correctly), so if grep prints something to stderr - it will create a /dev/null file (not sure if it is an issue though). -- Roman Kyrylych (Роман Кирилич)
Am 04.06.2010 19:48, schrieb Roman Kyrylych:
Booted fine under Xen using stock Arch kernel (via pv-grub), so I signoff x86_64.
Noticed the following:
+if ! /bin/mountpoint -q /dev; then + if grep -q devtmpfs /proc/filesystems 2>/dev/null; then + /bin/mount -n -t devtmpfs udev /dev -o mode=0755,size=10M,nosuid
1) grep should be /bin/grep
Shouldn't matter actually.
2) /dev/null does not exist at that time (if I understand correctly), so if grep prints something to stderr - it will create a /dev/null file (not sure if it is an issue though).
Well ... this is a corner-case, but it might not exist (in a common setup, it will exist). I'll fix that - we don't really need the 2>/dev/null anyway.
Am 04.06.2010 20:42, schrieb Thomas Bächler:
Am 04.06.2010 19:48, schrieb Roman Kyrylych:
2) /dev/null does not exist at that time (if I understand correctly), so if grep prints something to stderr - it will create a /dev/null file (not sure if it is an issue though).
Well ... this is a corner-case, but it might not exist (in a common setup, it will exist). I'll fix that - we don't really need the 2>/dev/null anyway.
I looked into it, and there is no problem. If /dev is not a separate file system, then the /dev directory on your root device is being used: # mount --bind / /mnt/other # ls -alhF /mnt/other/dev/ total 8.0K drwxr-xr-x 2 root root 4.0K Mar 9 22:27 ./ drwxr-xr-x 21 root root 4.0K May 29 15:50 ../ crw------- 1 root root 5, 1 Mar 9 22:27 console crw-r--r-- 1 root root 1, 3 Mar 9 22:27 null crw-r--r-- 1 root root 1, 5 Mar 9 22:27 zero # umount /mnt/other
On 06/03/2010 03:57 PM, Thomas Bächler wrote:
You should upgrade and test these together (don't forget to regenerate initramfs when you're upgraded). The new udev should work without the new initscripts and mkinitcpio, but udev-based network device renaming will fail in that case.
signoff x86_64 -- Ionut
Am 03.06.2010 14:57, schrieb Thomas Bächler:
You should upgrade and test these together (don't forget to regenerate initramfs when you're upgraded). The new udev should work without the new initscripts and mkinitcpio, but udev-based network device renaming will fail in that case.
udev is an upstream update with mostly bugfixes, but some changes in behaviour. Most importantly, NAME= rules are being ignored now, you can not change the kernel's predefined device name anymore.
initscripts changes: Dan McGee (1): Include _netdev option in NETFS list when calling fsck
Thomas Bächler (2): Only mount /proc, /sys and /dev if they haven't been mounted already, use devtmpfs instead of tmpfs if supported by the kernel Add --action=add to udevadm trigger
mkinitcpio changes: Thomas Bächler (5): Do not umount /proc and /sys before switch_root, but mount --move them into the real root Mount tmpfs or (if supported) devtmpfs on /dev, move it into the real root before switch_root udev hook: Add --action=add to udevadm trigger udev hook: Add ata_id, path_id, scsi_id and usb_id to allow complete persistent storage rules Release 0.6.5
For the future, we now fully support devtmpfs. It is recommended (but not required (yet)) by upstream. Using devtmpfs gets rid of a small number of race conditions.
I tested devtmpfs on a custom kernel, and with the exception of one short (and harmless) error message (mknod: /dev/rtc0: device exists), it works fine. Getting rid of that error message requires further consideration, as we can fix a bug with some custom kernel configurations at the same time if we do it right. For now, I will ignore this and please ask for testing and signoffs.
More small fixes to initscripts. Most importantly, a few errors with devtmpfs are removed. Please sign off again: Gerardo Exequiel Pozzi (1): Add nodevtmpfs at rc.shutdown when umount Thomas Bächler (4): Do not redirect the standard error of fsck to /dev/null by default Rewrite early RTC device creation Initialize /etc/mtab by copying /proc/mounts network, rc.conf: Add NETWORK_PERSIST option
On Thu, Jun 10, 2010 at 10:28, Thomas Bächler <thomas@archlinux.org> wrote:
More small fixes to initscripts. Most importantly, a few errors with devtmpfs are removed. Please sign off again:
Gerardo Exequiel Pozzi (1): Add nodevtmpfs at rc.shutdown when umount
Thomas Bächler (4): Do not redirect the standard error of fsck to /dev/null by default Rewrite early RTC device creation Initialize /etc/mtab by copying /proc/mounts network, rc.conf: Add NETWORK_PERSIST option
Signoff all (x86_64). Tested under Xen with timezone set to UTC, didn't test on a real machine nor with localtime yet. -- Roman Kyrylych (Роман Кирилич)
On Fri, Jun 11, 2010 at 10:14, Roman Kyrylych <roman.kyrylych@gmail.com> wrote:
On Thu, Jun 10, 2010 at 10:28, Thomas Bächler <thomas@archlinux.org> wrote:
More small fixes to initscripts. Most importantly, a few errors with devtmpfs are removed. Please sign off again:
Gerardo Exequiel Pozzi (1): Add nodevtmpfs at rc.shutdown when umount
Thomas Bächler (4): Do not redirect the standard error of fsck to /dev/null by default Rewrite early RTC device creation Initialize /etc/mtab by copying /proc/mounts network, rc.conf: Add NETWORK_PERSIST option
Signoff all (x86_64). Tested under Xen with timezone set to UTC, didn't test on a real machine nor with localtime yet.
Tested on a real hardware (i.e. non-virtual RTC device), still works. No problems with udev & mkinitcpio as well. Noone else uses these packages? ;) -- Roman Kyrylych (Роман Кирилич)
On Sun, Jun 13, 2010 at 12:24 PM, Roman Kyrylych <roman.kyrylych@gmail.com> wrote:
On Fri, Jun 11, 2010 at 10:14, Roman Kyrylych <roman.kyrylych@gmail.com> wrote:
On Thu, Jun 10, 2010 at 10:28, Thomas Bächler <thomas@archlinux.org> wrote:
More small fixes to initscripts. Most importantly, a few errors with devtmpfs are removed. Please sign off again:
Gerardo Exequiel Pozzi (1): Add nodevtmpfs at rc.shutdown when umount
Thomas Bächler (4): Do not redirect the standard error of fsck to /dev/null by default Rewrite early RTC device creation Initialize /etc/mtab by copying /proc/mounts network, rc.conf: Add NETWORK_PERSIST option
Signoff all (x86_64). Tested under Xen with timezone set to UTC, didn't test on a real machine nor with localtime yet.
Tested on a real hardware (i.e. non-virtual RTC device), still works. No problems with udev & mkinitcpio as well.
Noone else uses these packages? ;)
Booted my x86_64 machine without problems. mdadm in use, nothing else spectacular. -Dan
On 17/06/10 15:30, Dan McGee wrote:
On Sun, Jun 13, 2010 at 12:24 PM, Roman Kyrylych <roman.kyrylych@gmail.com> wrote:
On Fri, Jun 11, 2010 at 10:14, Roman Kyrylych<roman.kyrylych@gmail.com> wrote:
On Thu, Jun 10, 2010 at 10:28, Thomas Bächler<thomas@archlinux.org> wrote:
More small fixes to initscripts. Most importantly, a few errors with devtmpfs are removed. Please sign off again:
Gerardo Exequiel Pozzi (1): Add nodevtmpfs at rc.shutdown when umount
Thomas Bächler (4): Do not redirect the standard error of fsck to /dev/null by default Rewrite early RTC device creation Initialize /etc/mtab by copying /proc/mounts network, rc.conf: Add NETWORK_PERSIST option
Signoff all (x86_64). Tested under Xen with timezone set to UTC, didn't test on a real machine nor with localtime yet.
Tested on a real hardware (i.e. non-virtual RTC device), still works. No problems with udev& mkinitcpio as well.
Noone else uses these packages? ;)
Booted my x86_64 machine without problems. mdadm in use, nothing else spectacular.
I have not noticed a difference... signoff i686. Allan
Am 10.06.2010 09:28, schrieb Thomas Bächler:
Am 03.06.2010 14:57, schrieb Thomas Bächler:
You should upgrade and test these together (don't forget to regenerate initramfs when you're upgraded). The new udev should work without the new initscripts and mkinitcpio, but udev-based network device renaming will fail in that case.
udev is an upstream update with mostly bugfixes, but some changes in behaviour. Most importantly, NAME= rules are being ignored now, you can not change the kernel's predefined device name anymore.
initscripts changes: Dan McGee (1): Include _netdev option in NETFS list when calling fsck
Thomas Bächler (2): Only mount /proc, /sys and /dev if they haven't been mounted already, use devtmpfs instead of tmpfs if supported by the kernel Add --action=add to udevadm trigger
mkinitcpio changes: Thomas Bächler (5): Do not umount /proc and /sys before switch_root, but mount --move them into the real root Mount tmpfs or (if supported) devtmpfs on /dev, move it into the real root before switch_root udev hook: Add --action=add to udevadm trigger udev hook: Add ata_id, path_id, scsi_id and usb_id to allow complete persistent storage rules Release 0.6.5
For the future, we now fully support devtmpfs. It is recommended (but not required (yet)) by upstream. Using devtmpfs gets rid of a small number of race conditions.
I tested devtmpfs on a custom kernel, and with the exception of one short (and harmless) error message (mknod: /dev/rtc0: device exists), it works fine. Getting rid of that error message requires further consideration, as we can fix a bug with some custom kernel configurations at the same time if we do it right. For now, I will ignore this and please ask for testing and signoffs.
More small fixes to initscripts. Most importantly, a few errors with devtmpfs are removed. Please sign off again:
Gerardo Exequiel Pozzi (1): Add nodevtmpfs at rc.shutdown when umount
Thomas Bächler (4): Do not redirect the standard error of fsck to /dev/null by default Rewrite early RTC device creation Initialize /etc/mtab by copying /proc/mounts network, rc.conf: Add NETWORK_PERSIST option
One last fix to mkinitcpio: Thomas Bächler (2): Create /dev/{null,zero,mem,console} devices when devtmpfs is missing Release 0.6.6
On 06/19/2010 02:56 AM, Thomas Bächler wrote:
One last fix to mkinitcpio: Thomas Bächler (2): Create /dev/{null,zero,mem,console} devices when devtmpfs is missing Release 0.6.6
this is fixing my dev/null not found problem. signoff x86_64 -- Ionuț
On Sat, Jun 19, 2010 at 02:56, Thomas Bächler <thomas@archlinux.org> wrote:
One last fix to mkinitcpio: Thomas Bächler (2): Create /dev/{null,zero,mem,console} devices when devtmpfs is missing Release 0.6.6
signoff x86_64 -- Roman Kyrylych (Роман Кирилич)
participants (6)
-
Allan McRae
-
Dan McGee
-
Ionut Biru
-
Ionuț Bîru
-
Roman Kyrylych
-
Thomas Bächler