[gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[Alan Mackenzie]

Hi, Gentoo!

Just built the new kernel 3.7.9 last night, and it's one of these
"nothing works" situations.

It seems the problems are with the device files, whose ownership is set
to "root root" (rather than, e.g., "root audio") and whose permissions
are set to crw------- (rather than the expected crw-rw----).

I'm still running udev-171-r10.  This might well make a difference.

Needless to say, everything works under kernel 3.6.11.  It would be nice
if there were some mistake in my kernel config.

Could somebody help me get this fixed, please.

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[Alan McKinnon]

On 23/02/2013 19:18, Alan Mackenzie wrote:
> Hi, Gentoo!
> 
> Just built the new kernel 3.7.9 last night, and it's one of these
> "nothing works" situations.
> 
> It seems the problems are with the device files, whose ownership is set
> to "root root" (rather than, e.g., "root audio") and whose permissions
> are set to crw------- (rather than the expected crw-rw----).
> 
> I'm still running udev-171-r10.  This might well make a difference.
> 
> Needless to say, everything works under kernel 3.6.11.  It would be nice
> if there were some mistake in my kernel config.
> 
> Could somebody help me get this fixed, please.


udev sets ownership and permissions on devices nodes, not the kernel.

I'd guess your udev setup is incompatible with 3.7.* so let's start with
the usual:

What is your kernel config for the various TMPFS settings in the kernel?
(grep TMPFS in .config)

What mount options apply to /dev?
Contents of /etc/fstab?
USE flags that apply to udev?

-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[Alan Mackenzie]

On Sun, Feb 24, 2013 at 10:44:19AM +0200, Alan McKinnon wrote:
> On 23/02/2013 19:18, Alan Mackenzie wrote:
> > Hi, Gentoo!

> > Just built the new kernel 3.7.9 last night, and it's one of these
> > "nothing works" situations.

Sorry, I was exaggerating here.  I can work normally as long as I don't
try to use a peripheral, such as printer, audio, or scanner.  My network
connection is working OK.

> > It seems the problems are with the device files, whose ownership is set
> > to "root root" (rather than, e.g., "root audio") and whose permissions
> > are set to crw------- (rather than the expected crw-rw----).

> > I'm still running udev-171-r10.  This might well make a difference.

> > Needless to say, everything works under kernel 3.6.11.  It would be nice
> > if there were some mistake in my kernel config.

> > Could somebody help me get this fixed, please.


> udev sets ownership and permissions on devices nodes, not the kernel.

> I'd guess your udev setup is incompatible with 3.7.* so let's start with
> the usual:

> What is your kernel config for the various TMPFS settings in the kernel?
> (grep TMPFS in .config)

CONFIG_DEVTMPFS=y
CONFIG_DEVTMPFS_MOUNT=y
CONFIG_TMPFS=y
CONFIG_TMPFS_POSIX_ACL=y
CONFIG_TMPFS_XATTR=y

> What mount options apply to /dev?

On my (working) 3.6.11:
devtmpfs on /dev type devtmpfs (rw,relatime,size=1988792k,nr_inodes=497198,mode=755)

On my (non-working) 3.7.9:
devtmpfs on /dev type devtmpfs (rw,relatime,size=1988680k,nr_inodes=497170,mode=755)

> Contents of /etc/fstab?

(Leading comments have been removed.)

#/dev/BOOT              /boot           ext2            noauto,noatime  1 2
/dev/md6                /               ext3            noatime         0 1
/dev/sda2               none            swap            sw              0 0
/dev/sdb2               none            swap            sw              0 0

#/dev/mapper/vg-usr     /usr            ext3            noatime         1 2
/dev/mapper/vg-usrportage       /usr/portage    ext2            noatime         1 2
/dev/mapper/vg-usrportagedistfiles /usr/portage/distfiles ext2  noatime         1 2
/dev/mapper/vg-usrsrc   /usr/src        ext3            noatime         1 2

/dev/mapper/vg-home     /home           ext3            noatime         1 2
/dev/mapper/vg-opt      /opt            ext3            noatime         1 2

/dev/mapper/vg-var      /var            ext3            noatime         1 2
/dev/mapper/vg-varspoolmail     /var/spool/mail ext3            noatime         1 2
/dev/mapper/vg-varspoolnews     /var/spool/news reiserfs        noatime         1 2
/dev/mapper/vg-vardbpkg         /var/db/pkg     ext3            noatime         1 2

/dev/mapper/vg-iso      /iso            ext2            noatime         1 2
/dev/mapper/vg-old      /old            ext2            noatime         1 2
/dev/mapper/vg-vg--backup       /backup         ext2            noatime         1 2

/dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0
/dev/fd0                /mnt/floppy     auto            noauto          0 0

# glibc 2.2 and above expects tmpfs to be mounted at /dev/shm for
# POSIX shared memory (shm_open, shm_unlink).
# (tmpfs is a dynamically expandable/shrinkable ramdisk, and will
#  use almost no memory if not populated with files)
shm                     /dev/shm        tmpfs           nodev,nosuid,noexec     0 0


> USE flags that apply to udev?

[ebuild   R    ] sys-fs/udev-171-r10
    USE="gudev hwdb introspection keymap rule_generator
    -action_modeswitch -build -debug -edd (-extras) -floppy (-selinux) {-test}"

> -- 
> Alan McKinnon
> alan.mckinnon@gmail.com

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[Alan McKinnon]

On 24/02/2013 13:02, Alan Mackenzie wrote:
> On Sun, Feb 24, 2013 at 10:44:19AM +0200, Alan McKinnon wrote:
>> On 23/02/2013 19:18, Alan Mackenzie wrote:
>>> Hi, Gentoo!
> 
>>> Just built the new kernel 3.7.9 last night, and it's one of these
>>> "nothing works" situations.
> 
> Sorry, I was exaggerating here.  I can work normally as long as I don't
> try to use a peripheral, such as printer, audio, or scanner.  My network
> connection is working OK.
> 
>>> It seems the problems are with the device files, whose ownership is set
>>> to "root root" (rather than, e.g., "root audio") and whose permissions
>>> are set to crw------- (rather than the expected crw-rw----).
> 
>>> I'm still running udev-171-r10.  This might well make a difference.
> 
>>> Needless to say, everything works under kernel 3.6.11.  It would be nice
>>> if there were some mistake in my kernel config.
> 
>>> Could somebody help me get this fixed, please.
> 
> 
>> udev sets ownership and permissions on devices nodes, not the kernel.
> 
>> I'd guess your udev setup is incompatible with 3.7.* so let's start with
>> the usual:
> 
>> What is your kernel config for the various TMPFS settings in the kernel?
>> (grep TMPFS in .config)
> 
> CONFIG_DEVTMPFS=y
> CONFIG_DEVTMPFS_MOUNT=y
> CONFIG_TMPFS=y
> CONFIG_TMPFS_POSIX_ACL=y
> CONFIG_TMPFS_XATTR=y
> 
>> What mount options apply to /dev?
> 
> On my (working) 3.6.11:
> devtmpfs on /dev type devtmpfs (rw,relatime,size=1988792k,nr_inodes=497198,mode=755)
> 
> On my (non-working) 3.7.9:
> devtmpfs on /dev type devtmpfs (rw,relatime,size=1988680k,nr_inodes=497170,mode=755)
> 
>> Contents of /etc/fstab?
> 
> (Leading comments have been removed.)
> 
> #/dev/BOOT              /boot           ext2            noauto,noatime  1 2
> /dev/md6                /               ext3            noatime         0 1
> /dev/sda2               none            swap            sw              0 0
> /dev/sdb2               none            swap            sw              0 0
> 
> #/dev/mapper/vg-usr     /usr            ext3            noatime         1 2
> /dev/mapper/vg-usrportage       /usr/portage    ext2            noatime         1 2
> /dev/mapper/vg-usrportagedistfiles /usr/portage/distfiles ext2  noatime         1 2
> /dev/mapper/vg-usrsrc   /usr/src        ext3            noatime         1 2
> 
> /dev/mapper/vg-home     /home           ext3            noatime         1 2
> /dev/mapper/vg-opt      /opt            ext3            noatime         1 2
> 
> /dev/mapper/vg-var      /var            ext3            noatime         1 2
> /dev/mapper/vg-varspoolmail     /var/spool/mail ext3            noatime         1 2
> /dev/mapper/vg-varspoolnews     /var/spool/news reiserfs        noatime         1 2
> /dev/mapper/vg-vardbpkg         /var/db/pkg     ext3            noatime         1 2
> 
> /dev/mapper/vg-iso      /iso            ext2            noatime         1 2
> /dev/mapper/vg-old      /old            ext2            noatime         1 2
> /dev/mapper/vg-vg--backup       /backup         ext2            noatime         1 2
> 
> /dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0
> /dev/fd0                /mnt/floppy     auto            noauto          0 0
> 
> # glibc 2.2 and above expects tmpfs to be mounted at /dev/shm for
> # POSIX shared memory (shm_open, shm_unlink).
> # (tmpfs is a dynamically expandable/shrinkable ramdisk, and will
> #  use almost no memory if not populated with files)
> shm                     /dev/shm        tmpfs           nodev,nosuid,noexec     0 0
> 
> 
>> USE flags that apply to udev?
> 
> [ebuild   R    ] sys-fs/udev-171-r10
>     USE="gudev hwdb introspection keymap rule_generator
>     -action_modeswitch -build -debug -edd (-extras) -floppy (-selinux) {-test}"

That all looks quite normal (or, nothing stands out as being unusual).

I suppose your next step is to examine udev's logs where it creates the
various devices.


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[Alan Mackenzie]

Hi, Alan.

On Sun, Feb 24, 2013 at 06:08:19PM +0200, Alan McKinnon wrote:
> On 24/02/2013 13:02, Alan Mackenzie wrote:
> > On Sun, Feb 24, 2013 at 10:44:19AM +0200, Alan McKinnon wrote:
> >> On 23/02/2013 19:18, Alan Mackenzie wrote:
> >>> Hi, Gentoo!

> >>> Just built the new kernel 3.7.9 last night, and it's one of these
> >>> "nothing works" situations.

> > Sorry, I was exaggerating here.  I can work normally as long as I don't
> > try to use a peripheral, such as printer, audio, or scanner.  My network
> > connection is working OK.

> >>> It seems the problems are with the device files, whose ownership is set
> >>> to "root root" (rather than, e.g., "root audio") and whose permissions
> >>> are set to crw------- (rather than the expected crw-rw----).

> >>> I'm still running udev-171-r10.  This might well make a difference.

> >>> Needless to say, everything works under kernel 3.6.11.  It would be nice
> >>> if there were some mistake in my kernel config.

> >>> Could somebody help me get this fixed, please.

[ .... ]

> I suppose your next step is to examine udev's logs where it creates the
> various devices.

And I promised myself some months ago I would _never_ spend time on udev
innards.  ;-(

So, I set the debugging options inside /etc/conf.d/udev, which causes
/run/udev/udev.log to be created.

On the failing 3.7.9, /run/udev/udev.log terminates abruptly with an
error message, with the entire file, 55 lines, looking like this:

1361736665.547175 [1761] parse_file: reading '/lib/udev/rules.d/10-dm.rules' as rules file
1361736665.547289 [1761] parse_file: reading '/run/udev/rules.d/10-root-link.rules' as rules file
[ ..... ]
1361736665.553901 [1761] parse_file: reading '/lib/udev/rules.d/95-upower-wup.rules' as rules file
1361736665.553924 [1761] parse_file: reading '/lib/udev/rules.d/97-bluetooth-hid2hci.rules' as rules file
1361736665.553971 [1761] udev_rules_new: rules use 108624 bytes tokens (9052 * 12 bytes), 31856 bytes buffer
1361736665.553977 [1761] udev_rules_new: temporary index used 48860 bytes (2443 * 20 bytes)
1361736665.554010 [1761] main: error creating epoll fd: Function not implemented

I hope "epoll fd", whatever that may be, is something critically
important to justify just stopping udev.  But it can't be that important
if kernel 3.6.11 doesn't emit such an event.  It looks like the kernel's
event interface has changed.

It seems that the contents of /dev, such as they were, were created
directly by the kernel when devtmpfs was mounted.  Normally they would be
modified into something usable by udev.

Maybe I should just bite the bullet and go through the pain of updating
my udev to version 197-r8.  :-(

> -- 
> Alan McKinnon
> alan.mckinnon@gmail.com

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[Alan McKinnon]

On 24/02/2013 23:40, Alan Mackenzie wrote:
> Hi, Alan.
> 
> On Sun, Feb 24, 2013 at 06:08:19PM +0200, Alan McKinnon wrote:
>> On 24/02/2013 13:02, Alan Mackenzie wrote:
>>> On Sun, Feb 24, 2013 at 10:44:19AM +0200, Alan McKinnon wrote:
>>>> On 23/02/2013 19:18, Alan Mackenzie wrote:
>>>>> Hi, Gentoo!
> 
>>>>> Just built the new kernel 3.7.9 last night, and it's one of these
>>>>> "nothing works" situations.
> 
>>> Sorry, I was exaggerating here.  I can work normally as long as I don't
>>> try to use a peripheral, such as printer, audio, or scanner.  My network
>>> connection is working OK.
> 
>>>>> It seems the problems are with the device files, whose ownership is set
>>>>> to "root root" (rather than, e.g., "root audio") and whose permissions
>>>>> are set to crw------- (rather than the expected crw-rw----).
> 
>>>>> I'm still running udev-171-r10.  This might well make a difference.
> 
>>>>> Needless to say, everything works under kernel 3.6.11.  It would be nice
>>>>> if there were some mistake in my kernel config.
> 
>>>>> Could somebody help me get this fixed, please.
> 
> [ .... ]
> 
>> I suppose your next step is to examine udev's logs where it creates the
>> various devices.
> 
> And I promised myself some months ago I would _never_ spend time on udev
> innards.  ;-(
> 
> So, I set the debugging options inside /etc/conf.d/udev, which causes
> /run/udev/udev.log to be created.
> 
> On the failing 3.7.9, /run/udev/udev.log terminates abruptly with an
> error message, with the entire file, 55 lines, looking like this:
> 
> 1361736665.547175 [1761] parse_file: reading '/lib/udev/rules.d/10-dm.rules' as rules file
> 1361736665.547289 [1761] parse_file: reading '/run/udev/rules.d/10-root-link.rules' as rules file
> [ ..... ]
> 1361736665.553901 [1761] parse_file: reading '/lib/udev/rules.d/95-upower-wup.rules' as rules file
> 1361736665.553924 [1761] parse_file: reading '/lib/udev/rules.d/97-bluetooth-hid2hci.rules' as rules file
> 1361736665.553971 [1761] udev_rules_new: rules use 108624 bytes tokens (9052 * 12 bytes), 31856 bytes buffer
> 1361736665.553977 [1761] udev_rules_new: temporary index used 48860 bytes (2443 * 20 bytes)
> 1361736665.554010 [1761] main: error creating epoll fd: Function not implemented
> 
> I hope "epoll fd", whatever that may be, is something critically
> important to justify just stopping udev.  But it can't be that important
> if kernel 3.6.11 doesn't emit such an event.  It looks like the kernel's
> event interface has changed.
> 
> It seems that the contents of /dev, such as they were, were created
> directly by the kernel when devtmpfs was mounted.  Normally they would be
> modified into something usable by udev.
> 
> Maybe I should just bite the bullet and go through the pain of updating
> my udev to version 197-r8.  :-(


Sounds logical to me, I'd conclude the same.

Perhaps you should report your findings to b.g.o just in case the
DEPENDS and config options between kernel and udev need to be tweaked in
the ebuilds (I'd hate to see many users have to go through the same as
yourself)

Got it on the pain you feel, I avoid looking into udev like it was the
plague. Some days I wonder if going back to MAKEDEV wouldn't be far
simpler in Gentoo land.....


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Kernel 3.7.9: Lots of devices are root root rw-------.

[Alan Mackenzie]

Hi, Alan.

On Mon, Feb 25, 2013 at 01:47:55AM +0200, Alan McKinnon wrote:
> On 24/02/2013 23:40, Alan Mackenzie wrote:
> > On Sun, Feb 24, 2013 at 06:08:19PM +0200, Alan McKinnon wrote:
> >> On 24/02/2013 13:02, Alan Mackenzie wrote:
> >>> On Sun, Feb 24, 2013 at 10:44:19AM +0200, Alan McKinnon wrote:
> >>>> On 23/02/2013 19:18, Alan Mackenzie wrote:
> >>>>> Hi, Gentoo!

> >>>>> Just built the new kernel 3.7.9 last night, and it's one of these
> >>>>> "nothing works" situations.

> >>> Sorry, I was exaggerating here.  I can work normally as long as I don't
> >>> try to use a peripheral, such as printer, audio, or scanner.  My network
> >>> connection is working OK.

> >>>>> It seems the problems are with the device files, whose ownership is set
> >>>>> to "root root" (rather than, e.g., "root audio") and whose permissions
> >>>>> are set to crw------- (rather than the expected crw-rw----).

> >>>>> I'm still running udev-171-r10.  This might well make a difference.

> >>>>> Needless to say, everything works under kernel 3.6.11.  It would be nice
> >>>>> if there were some mistake in my kernel config.

> >>>>> Could somebody help me get this fixed, please.

> > [ .... ]

> >> I suppose your next step is to examine udev's logs where it creates the
> >> various devices.

> > And I promised myself some months ago I would _never_ spend time on udev
> > innards.  ;-(

> > So, I set the debugging options inside /etc/conf.d/udev, which causes
> > /run/udev/udev.log to be created.

> > On the failing 3.7.9, /run/udev/udev.log terminates abruptly with an
> > error message, with the entire file, 55 lines, looking like this:

> > 1361736665.547175 [1761] parse_file: reading '/lib/udev/rules.d/10-dm.rules' as rules file
> > 1361736665.547289 [1761] parse_file: reading '/run/udev/rules.d/10-root-link.rules' as rules file
> > [ ..... ]
> > 1361736665.553901 [1761] parse_file: reading '/lib/udev/rules.d/95-upower-wup.rules' as rules file
> > 1361736665.553924 [1761] parse_file: reading '/lib/udev/rules.d/97-bluetooth-hid2hci.rules' as rules file
> > 1361736665.553971 [1761] udev_rules_new: rules use 108624 bytes tokens (9052 * 12 bytes), 31856 bytes buffer
> > 1361736665.553977 [1761] udev_rules_new: temporary index used 48860 bytes (2443 * 20 bytes)
> > 1361736665.554010 [1761] main: error creating epoll fd: Function not implemented

> > I hope "epoll fd", whatever that may be, is something critically
> > important to justify just stopping udev.  But it can't be that important
> > if kernel 3.6.11 doesn't emit such an event.  It looks like the kernel's
> > event interface has changed.

> > It seems that the contents of /dev, such as they were, were created
> > directly by the kernel when devtmpfs was mounted.  Normally they would be
> > modified into something usable by udev.

> > Maybe I should just bite the bullet and go through the pain of updating
> > my udev to version 197-r8.  :-(


> Sounds logical to me, I'd conclude the same.

> Perhaps you should report your findings to b.g.o just in case the
> DEPENDS and config options between kernel and udev need to be tweaked in
> the ebuilds (I'd hate to see many users have to go through the same as
> yourself)

I did that this morning (European time).  I suspect I'll get an answer
along the lines of "what do you expect if you use old versions?".

> Got it on the pain you feel, I avoid looking into udev like it was the
> plague. Some days I wonder if going back to MAKEDEV wouldn't be far
> simpler in Gentoo land.....

MAKEDEV??

I spent some time this afternoon making sure I knew how to chroot
effectively, using my old "mdev system" as a rescue system.  Probably
updating udev will just go OK for me.  I'll wait until I here back from
the Gentoo guys on my bug report.

Thanks for the help in sorting this out.

> -- 
> Alan McKinnon
> alan.mckinnon@gmail.com

-- 
Alan Mackenzie (Nuremberg, Germany).