net interface renaming issue (+fix?)

DISCUSSION Thomas de Grenier de Latour 5 participants 9 messages Apr 9, 2006 – Apr 13, 2006

Subsystems (2)

Networking Network drivers

lore

Share & tools

lore

Thomas de Grenier de Latour

author

2006-04-09 17:21 UTC #0 lore raw

Hi,

I'm running Gentoo Linux, a 2.6.16-ck kernel, and udev-0.89-r2, and have had hard time with network interfaces renaming through udev rules. The first thing i've tried were rules like this one:

SUBSYSTEM="net", KERNEL="eth*", SYSFS{address}="00:0d:60:12:75:0a", NAME="lan"

Plus this one from the standard early rules:

ACTION="add", SUBSYSTEM="net", WAIT_FOR_SYSFS="address"

It doesn't work when i "modprobe e1000" (my ethernet driver):
...
udevd[21612]: udev_event_run: seq 956 forked, pid [21816], 'add' 'net', 0 seconds old udevd-event[21816]: wait_for_sysfs: file '/sys/class/net/eth0/address' appeared after 0 loops udevd-event[21816]: udev_rules_get_name: no node name set, will use kernel name 'eth0'
...
But if i later do a "echo add > /sys/class/net/eth0/uevent", then
interface is properly renamed. It also works fine if i start with
the module initialy not loaded, and then trigger the uevent on the corresponding pci device (will load the module, etc.)

Then i've tried replacing the address match by DRIVER="e1000", with
no more success. But i've noticed something interesting in debug
output, when the rule matcher walks up in parent devices to check
the driver:
% grep -i driver /var/tmp/udev-net-debug.log
...
udevd-event[18411]: match_key: key DRIVER value='e1000'
udevd-event[18411]: match_key: match DRIVER 'e1000' <-> ''
udevd-event[18411]: match_key: DRIVER is false
udevd-event[18411]: sysfs_device_get: add to cache 'devpath=/devices/pci0000:00/0000:00:1e.0/0000:02:01.0', subsystem='pci', driver='' udevd-event[18411]: match_key: key DRIVER value='e1000'
udevd-event[18411]: match_key: match DRIVER 'e1000' <-> ''
udevd-event[18411]: match_key: DRIVER is false
...
This is weird, because if i latter look in /sys, the "driver" links
are here. Which made me think it was a race, so i've added this rule:

ACTION="add", SUBSYSTEM="net", WAIT_FOR_SYSFS="device/driver"

And it fixed the problem:
...
udevd[21612]: udev_event_run: seq 950 forked, pid [21790], 'add' 'net', 0 seconds old udevd-event[21790]: wait_for_sysfs: file '/sys/class/net/eth0/address' appeared after 0 loops udevd-event[21790]: wait_for_sysfs: wait for '/sys/class/net/eth0/device/driver' for 20 mseconds udevd-event[21790]: wait_for_sysfs: file '/sys/class/net/eth0/device/driver' appeared after 1 loops udevd-event[21790]: udev_rules_get_name: rule applied, 'eth0' becomes 'lan' udevd-event[21790]: rename_net_if: changing net interface name from 'eth0' to 'lan' udevd-event[21790]: udev_add_device: renamed netif to 'lan'
...

It also fixes the problem with using a SYSFS{address} match btw.
With no such wait, i can see in debug that "address" is found in sysfs,
but with no value:
...
udevd-event[21977]: sysfs_attr_get_value: open '/class/net/eth0'/'address' udevd-event[21977]: sysfs_attr_get_value: new uncached attribute '/sys/class/net/eth0/address' udevd-event[21977]: sysfs_attr_get_value: add to cache '/sys/class/net/eth0/address' udevd-event[21977]: sysfs_attr_get_value: open '/class/net/eth0'/'address'
...
Whereas with the wait-for-driver trick, i can see it read with a
useful value:
...
udevd-event[21954]: sysfs_attr_get_value: open '/class/net/eth0'/'address' udevd-event[21954]: sysfs_attr_get_value: new uncached attribute '/sys/class/net/eth0/address' udevd-event[21954]: sysfs_attr_get_value: add to cache '/sys/class/net/eth0/address' udevd-event[21954]: sysfs_attr_get_value: cache '/sys/class/net/eth0/address' with value '00:0d:60:12:75:0a'
...

So i wonder, maybe such a rule should be added to the standard early ones? It should maybe use more checks though, to be sure there is actually a driver to wait. Something like ENV{PHYSDEVPATH}="?*" and/or ENV{PHYSDEVDRIVER}="?*".

Btw, using ENV{PHYSDEVDRIVER}="e1000" in my renaming rule was working fine, with no trick (this variables are correctly set, like 'udevmonitor --env' shows).

So, what do you think, does such a rule makes sense?
Or is "address" being added to sysfs with no useful value yet the real issue, and my rule only an ugly workround?

Thanks,

Signature · 12 lines

TGL.


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid0944&bid$1720&dat1642
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Sergey Vlasov

2006-04-09 17:56 UTC #1 lore raw

Quoted 77 lines · On Sun, 9 Apr 2006 19:21:40 +0200 Thomas de Grenier de Latour wrote:

Hi,
I'm running Gentoo Linux, a 2.6.16-ck kernel, and udev-0.89-r2, and have had hard time with network interfaces renaming through udev rules. The first thing i've tried were rules like this one:
SUBSYSTEM=="net", KERNEL=="eth*", SYSFS{address}=="00:0d:60:12:75:0a", NAME="lan"
Plus this one from the standard early rules:
ACTION=="add", SUBSYSTEM=="net", WAIT_FOR_SYSFS="address"
It doesn't work when i "modprobe e1000" (my ethernet driver):
...
udevd[21612]: udev_event_run: seq 956 forked, pid [21816], 'add' 'net', 0 seconds old udevd-event[21816]: wait_for_sysfs: file '/sys/class/net/eth0/address' appeared after 0 loops udevd-event[21816]: udev_rules_get_name: no node name set, will use kernel name 'eth0'
...
But if i later do a "echo add > /sys/class/net/eth0/uevent", then
interface is properly renamed. It also works fine if i start with
the module initialy not loaded, and then trigger the uevent on the corresponding pci device (will load the module, etc.)
Then i've tried replacing the address match by DRIVER=="e1000", with
no more success. But i've noticed something interesting in debug
output, when the rule matcher walks up in parent devices to check
the driver:
% grep -i driver /var/tmp/udev-net-debug.log
...
udevd-event[18411]: match_key: key DRIVER value='e1000'
udevd-event[18411]: match_key: match DRIVER 'e1000' <-> ''
udevd-event[18411]: match_key: DRIVER is false
udevd-event[18411]: sysfs_device_get: add to cache 'devpath=/devices/pci0000:00/0000:00:1e.0/0000:02:01.0', subsystem='pci', driver='' udevd-event[18411]: match_key: key DRIVER value='e1000'
udevd-event[18411]: match_key: match DRIVER 'e1000' <-> ''
udevd-event[18411]: match_key: DRIVER is false
...
This is weird, because if i latter look in /sys, the "driver" links
are here. Which made me think it was a race, so i've added this rule:
ACTION=="add", SUBSYSTEM=="net", WAIT_FOR_SYSFS="device/driver"
And it fixed the problem:
...
udevd[21612]: udev_event_run: seq 950 forked, pid [21790], 'add' 'net', 0 seconds old udevd-event[21790]: wait_for_sysfs: file '/sys/class/net/eth0/address' appeared after 0 loops udevd-event[21790]: wait_for_sysfs: wait for '/sys/class/net/eth0/device/driver' for 20 mseconds udevd-event[21790]: wait_for_sysfs: file '/sys/class/net/eth0/device/driver' appeared after 1 loops udevd-event[21790]: udev_rules_get_name: rule applied, 'eth0' becomes 'lan' udevd-event[21790]: rename_net_if: changing net interface name from 'eth0' to 'lan' udevd-event[21790]: udev_add_device: renamed netif to 'lan'
...
It also fixes the problem with using a SYSFS{address} match btw.
With no such wait, i can see in debug that "address" is found in sysfs,
but with no value:
...
udevd-event[21977]: sysfs_attr_get_value: open '/class/net/eth0'/'address' udevd-event[21977]: sysfs_attr_get_value: new uncached attribute '/sys/class/net/eth0/address' udevd-event[21977]: sysfs_attr_get_value: add to cache '/sys/class/net/eth0/address' udevd-event[21977]: sysfs_attr_get_value: open '/class/net/eth0'/'address'
...
Whereas with the wait-for-driver trick, i can see it read with a
useful value:
...
udevd-event[21954]: sysfs_attr_get_value: open '/class/net/eth0'/'address' udevd-event[21954]: sysfs_attr_get_value: new uncached attribute '/sys/class/net/eth0/address' udevd-event[21954]: sysfs_attr_get_value: add to cache '/sys/class/net/eth0/address' udevd-event[21954]: sysfs_attr_get_value: cache '/sys/class/net/eth0/address' with value '00:0d:60:12:75:0a'
...
So i wonder, maybe such a rule should be added to the standard early ones? It should maybe use more checks though, to be sure there is actually a driver to wait. Something like ENV{PHYSDEVPATH}=="?*" and/or ENV{PHYSDEVDRIVER}=="?*".

Quoted 8 lines

Btw, using ENV{PHYSDEVDRIVER}=="e1000" in my renaming rule was working fine, with no trick (this variables are correctly set, like 'udevmonitor --env' shows).
So, what do you think, does such a rule makes sense?
Or is "address" being added to sysfs with no useful value yet the real issue, and my rule only an ugly workround?

(quoting the whole message for netdev)

Apparently there is a race there. The "add" uevent for /class/net/eth0 is emitted by class_device_register(), which is called by netdev_register_sysfs(). Class device attributes (like "address") are also added by class_device_register() (really class_device_add(), which is invoked from there), but this is done before generating the uevent, so at the first glance there is no race here (and waiting for "address" is unnecessary). However, show_address() does not output anything unless dev->reg_state == NETREG_REGISTERED - and this state is set by netdev_run_todo() only after netdev_register_sysfs() returns, so in the meantime (while netdev_register_sysfs() is busy adding the "statistics" attribute group) some process may see an empty "address" attribute.

Waiting for "device/driver" ensures that udevd continues to process the uevent only after the probe function completes, which ensures completion of the registration (netdev_run_todo() is invoked by rtnl_unlock() at the end of register_netdev() and processes all pending requests).

Andrey Borzenkov

2006-04-09 18:00 UTC #2 lore raw

In reply to Thomas de Grenier de Latour

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 13 lines · Thomas de Grenier de Latour

Hi,
I'm running Gentoo Linux, a 2.6.16-ck kernel, and udev-0.89-r2, and have had hard time with network interfaces renaming through udev rules. The first thing i've tried were rules like this one:
SUBSYSTEM="net", KERNEL="eth*", SYSFS{address}="00:0d:60:12:75:0a", NAME="lan"
Plus this one from the standard early rules:
ACTION="add", SUBSYSTEM="net", WAIT_FOR_SYSFS="address"

Quoted 11 lines

It doesn't work when i "modprobe e1000" (my ethernet driver):
....
udevd[21612]: udev_event_run: seq 956 forked, pid [21816], 'add' 'net', 0 seconds old udevd-event[21816]: wait_for_sysfs: file '/sys/class/net/eth0/address' appeared after 0 loops udevd-event[21816]:
udev_rules_get_name: no node name set, will use kernel name 'eth0' .... But if i later do a "echo add > /sys/class/net/eth0/uevent", then interface is properly renamed. It also works fine if i start with the module initialy not loaded, and then trigger the uevent on the corresponding pci device (will load the module, etc.)

There appears to be race between creating 'address' and getting (usable) value. See below.

[...]

Quoted 6 lines

So i wonder, maybe such a rule should be added to the standard early ones? It should maybe use more checks though, to be sure there is actually a driver to wait. Something like ENV{PHYSDEVPATH}="?*" and/or ENV{PHYSDEVDRIVER}="?*".

You probably get reply that they are obsolete and should not be used in new rules :)
[...]

Quoted 3 lines

Or is "address" being added to sysfs with no useful value yet the real issue, and my rule only an ugly workround?

Seems so.

net/core/net-syfs.c:dev_isalive():

        return dev->reg_state = NETREG_REGISTERED;

net/core/net-syfs.c:show_address():

        if (dev_isalive(net))
            ret = format_addr(buf, net->dev_addr, net->addr_len);

net/core/dev.c:netdev_run_todo():
err = netdev_register_sysfs(dev);
if (err)
printk(KERN_ERR "%s: failed sysfs registration
(%d)\n",

                                       dev->name, err);
                        dev->reg_state = NETREG_REGISTERED;
                        break;

So basically there is a window between net device appearing in sysfs and 'address' returning usable value. To test you could move dev->reg_state assignment before netdev_register_sysfs call; if this helps it makes sense ask on net list if this is safe change.

- -andrey
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)

iD8DBQFEOUvTR6LMutpd94wRAg+nAKCXnI9mrChpQ9TAz4VVQhSaP/MQjgCfbX6n 5lcJ30K1qpTtbxISei6pN9I=Rwsc
-----END PGP SIGNATURE-----

-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language that extends applications into web and mobile media. Attend the live webcast and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid0944&bid$1720&dat1642
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Thomas de Grenier de Latour

author

2006-04-09 19:28 UTC #3 lore raw

In reply to Thomas de Grenier de Latour

Quoted 5 lines · Andrey Borzenkov

net/core/dev.c:netdev_run_todo():
err = netdev_register_sysfs(dev); if (err)
printk(KERN_ERR "%s: failed sysfs
registration (%d)\n",

Quoted 8 lines

                                       dev->name, err);
                        dev->reg_state = NETREG_REGISTERED;
                        break;
So basically there is a window between net device appearing in sysfs and 'address' returning usable value. To test you could move dev->reg_state assignment before netdev_register_sysfs call; if this helps it makes sense ask on net list if this is safe change.

As far as i can test, yes, it works fine. I will forward your suggestion to -netdev@.

Btw, many thanks to both of you, Andrey and Sergey, for the fast answers and crystal-clear explanations of the actual issue.

Signature · 12 lines

TGL.


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid0944&bid$1720&dat1642
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Marco d'Itri

2006-04-13 08:39 UTC #4 lore raw

In reply to Thomas de Grenier de Latour

Quoted 8 lines · Andrey Borzenkov

Quoted 5 lines
So i wonder, maybe such a rule should be added to the standard early ones? It should maybe use more checks though, to be sure there is actually a driver to wait. Something like ENV{PHYSDEVPATH}="?*" and/or ENV{PHYSDEVDRIVER}="?*".
You probably get reply that they are obsolete and should not be used in new rules :)

So, is this an acceptable workaround until the kernel is fixed or not? I implemented interfaces persistence for the Debian package, but users will get mad if the rules are randomly ignored (which apparently is already happening).

Also, is there a kernel patch yet?

Signature · 13 lines

ciao,
Marco


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid0944&bid$1720&dat1642
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Thomas de Grenier de Latour

author

2006-04-13 09:06 UTC #5 lore raw

In reply to Thomas de Grenier de Latour

Quoted 3 lines · On Thu, 13 Apr 2006 10:39:53 +0200, [email protected] (Marco d'Itri) wrote:

Also, is there a kernel patch yet?

Yes, this one has been applied:
http://kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blobdiff;h'31570eba5b35a21c311dd587057c39805082f1;hp�b62998866ae2e298139164a85ec0757b7f3fc7;hb�69d458b90bfb9117cbb488cfa645d94c3921b1;f=net/core/dev.c

Signature · 12 lines

TGL.


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid0944&bid$1720&dat1642
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Kay Sievers

2006-04-13 10:29 UTC #6 lore raw

In reply to Thomas de Grenier de Latour

Quoted 8 lines · Marco d'Itri

Quoted 6 lines · Andrey Borzenkov
Quoted 5 lines
So i wonder, maybe such a rule should be added to the standard early ones? It should maybe use more checks though, to be sure there is actually a driver to wait. Something like ENV{PHYSDEVPATH}="?*" and/or ENV{PHYSDEVDRIVER}="?*".

Quoted 4 lines

Quoted 2 lines
You probably get reply that they are obsolete and should not be used in new rules :)
So, is this an acceptable workaround until the kernel is fixed or not?

Sure, it will not go away soon, and for temporary fixes, if that helps working around something that is on the way to be fixed, it's fine.

But right, the longer term goal is that the notion of "physical devices" and "virtual devices" should not be exported that way by the kernel and therefore the use of these values, or the "device" link, or the <subsystem:kernel_name> symlinks, or a hardcoded sequence of sysfs parent devices must be avoided for other things than temporary workarounds.

Thanks,
Kay

Marco d'Itri

2006-04-13 10:36 UTC #7 lore raw

In reply to Thomas de Grenier de Latour

Quoted 5 lines · Kay Sievers

But right, the longer term goal is that the notion of "physical devices" and "virtual devices" should not be exported that way by the kernel and therefore the use of these values, or the "device" link, or the <subsystem:kernel_name> symlinks, or a hardcoded sequence of sysfs parent devices must be avoided for other things than temporary workarounds.

How should I rewrite this code then?

device_description() {
local bus=$(sysreadlink device/bus)
bus=${bus##*/}

  if   [ "$bus" = pci ]; then
    local vendor_id=$(sysread device/vendor)
    local device_id=$(sysread device/device)
    echo -n "PCI device ${vendor_id#0x}:${device_id#0x}"
  elif [ "$bus" = usb ]; then
    local device=$(sysreadlink device)
    if [ "$device" -a -e $device/../idVendor ]; then
      local idVendor idProduct
      read idVendor  < $device/../idVendor  || true
      read idProduct < $device/../idProduct || true
      printf 'USB device %x/%x' 0x$idVendor 0x$idProduct
    else
      echo -n "UNKNOWN USB device ($DEVPATH)"
    fi
  elif [ "$bus" ]; then
    echo -n "UNKNOWN $bus device ($DEVPATH)"
  else
    echo -n "UNKNOWN device ($DEVPATH)"
  fi

local driver=$(sysreadlink device/driver)
if [ "$driver" ]; then echo -n " (${driver##*/})"; fi
}

Signature · 2 lines

ciao,
Marco

Kay Sievers

2006-04-13 11:10 UTC #8 lore raw

In reply to Thomas de Grenier de Latour

Quoted 34 lines · Marco d'Itri

Quoted 5 lines · Kay Sievers
But right, the longer term goal is that the notion of "physical devices" and "virtual devices" should not be exported that way by the kernel and therefore the use of these values, or the "device" link, or the <subsystem:kernel_name> symlinks, or a hardcoded sequence of sysfs parent devices must be avoided for other things than temporary workarounds.
How should I rewrite this code then?
device_description() {
local bus=$(sysreadlink device/bus)
bus=${bus##*/}
  if   [ "$bus" = pci ]; then
    local vendor_id=$(sysread device/vendor)
    local device_id=$(sysread device/device)
    echo -n "PCI device ${vendor_id#0x}:${device_id#0x}"
  elif [ "$bus" = usb ]; then
    local device=$(sysreadlink device)
    if [ "$device" -a -e $device/../idVendor ]; then
      local idVendor idProduct
      read idVendor  < $device/../idVendor  || true
      read idProduct < $device/../idProduct || true
      printf 'USB device %x/%x' 0x$idVendor 0x$idProduct
    else
      echo -n "UNKNOWN USB device ($DEVPATH)"
    fi
  elif [ "$bus" ]; then
    echo -n "UNKNOWN $bus device ($DEVPATH)"
  else
    echo -n "UNKNOWN device ($DEVPATH)"
  fi

Quoted 3 lines

local driver=$(sysreadlink device/driver)
if [ "$driver" ]; then echo -n " (${driver##*/})"; fi
}

In the end, we will have only one tree in sysfs and the devpath will contain the whole dependency of devices for every device. eth0 will have:
/devices/pci0000:00/0000:00:1c.0/0000:02:00.0/eth0
as devpath instead of the current /class/net/eth0, which will only be a silent symlink to that location. When you get the event, you have to walk up the chain of parents in the devpath until you find the device you are looking for, you can't rely on a specific sequence, there may be a device inserted if the kernel device model needs that. That's why the "device" link does not make sense at that point.

So you would walk up until you find a device of the subsystem you are looking for like "pci" or "usb" and then read these values. See the usb_id logic:
http://www.kernel.org/git/?p=linux/hotplug/udev.git;a=blob;h…c6beda40e38d307d355aed4d8b789bb193cca4;hb4a805bccbc52e48f972a87f83007fdb0ffe19e;f=extras/usb_id/usb_id.c#l250
It does not use any of these values and requests parent devices only by subsystem and can skip newly inserted parents in the devpath without breakage.

Udev already does all that today, so it may make sense to get udev to read these values and pass them to the script.

Or we export them with rules to the environment (needs a change to udev to apply the format to the ENV key at that time).

Or we provide some logic from udev_sysfs.c as a similar shell function to include by script users:
http://www.kernel.org/git/?p=linux/hotplug/udev.git;a=blob;hÝc0b46790066e30bede2c9bb0026a271633795c;hb4a805bccbc52e48f972a87f83007fdb0ffe19e;f=udev_sysfs.c#l196

Let me know if we should add something to udev itself to solve that.

Thanks,
Kay