[PATCH] Standard rules for AoE devices.

[PATCH] Standard rules for AoE devices.

[Ryan Thomas]

In the interest of standardizing udev rules, please consider the 
following patch that adds udev rules for the ATA over Ethernet character 
and block devices.  The aoe module has been a long-time member of the 
kernel and needs inclusion in the standard udev rules.

Thanks,
Ryan Thomas

---
  rules/packages/40-aoe.rules |    9 +++++++++
  1 files changed, 9 insertions(+), 0 deletions(-)
  create mode 100644 rules/packages/40-aoe.rules

diff --git a/rules/packages/40-aoe.rules b/rules/packages/40-aoe.rules
new file mode 100644
index 0000000..38c0e00
--- /dev/null
+++ b/rules/packages/40-aoe.rules
@@ -0,0 +1,9 @@
+# aoe char devices
+SUBSYSTEM="aoe", KERNEL="discover",   NAME="etherd/%k", GROUP="disk", 
MODE="0220"
+SUBSYSTEM="aoe", KERNEL="err",        NAME="etherd/%k", GROUP="disk", 
MODE="0440"
+SUBSYSTEM="aoe", KERNEL="interfaces", NAME="etherd/%k", GROUP="disk", 
MODE="0220"
+SUBSYSTEM="aoe", KERNEL="revalidate", NAME="etherd/%k", GROUP="disk", 
MODE="0220"
+SUBSYSTEM="aoe", KERNEL="flush",      NAME="etherd/%k", GROUP="disk", 
MODE="0220"
+
+# aoe block devices
+SUBSYSTEM="aoe", KERNEL="etherd*",    NAME="%k", GROUP="disk"
-- 
1.5.6.5

Re: [PATCH] Standard rules for AoE devices.

[Kay Sievers]

On Tue, Dec 9, 2008 at 01:10, Ryan Thomas <ryan@coraid.com> wrote:
> In the interest of standardizing udev rules, please consider the following
> patch that adds udev rules for the ATA over Ethernet character and block
> devices.  The aoe module has been a long-time member of the kernel and needs
> inclusion in the standard udev rules.

Applied.

Thanks,
Kay

Re: [PATCH] Standard rules for AoE devices.

[Marco d'Itri]

On Dec 09, Ryan Thomas <ryan@coraid.com> wrote:

> In the interest of standardizing udev rules, please consider the  
> following patch that adds udev rules for the ATA over Ethernet character  
> and block devices.  The aoe module has been a long-time member of the  
> kernel and needs inclusion in the standard udev rules.

Debian so far has used these rules which are more generic and much
simpler. Are they incorrect in some way?

# AOE character devices
SUBSYSTEM="aoe",		NAME="etherd/%k"

SUBSYSTEM="aoe",               MODE="0220",    GROUP="disk"
SUBSYSTEM="aoe", KERNEL="err", MODE="0440",   GROUP="disk"

-- 
ciao,
Marco

Re: [PATCH] Standard rules for AoE devices.

[Kay Sievers]

On Tue, Dec 9, 2008 at 01:49, Marco d'Itri <md@linux.it> wrote:
> On Dec 09, Ryan Thomas <ryan@coraid.com> wrote:
>
>> In the interest of standardizing udev rules, please consider the
>> following patch that adds udev rules for the ATA over Ethernet character
>> and block devices.  The aoe module has been a long-time member of the
>> kernel and needs inclusion in the standard udev rules.
>
> Debian so far has used these rules which are more generic and much
> simpler. Are they incorrect in some way?
>
> # AOE character devices
> SUBSYSTEM="aoe",               NAME="etherd/%k"
>
> SUBSYSTEM="aoe",               MODE="0220",    GROUP="disk"
> SUBSYSTEM="aoe", KERNEL="err", MODE="0440",   GROUP="disk"

That looks fine.

I wonder why we have this comment:
  +# aoe block devices
  +SUBSYSTEM="aoe", KERNEL="etherd*",    NAME="%k", GROUP="disk"

They can not be block devices if they have SUBSYSTEM="aoe". Do the
etherd* devices belong into the subdir or not?

We have:
  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk"
in the current default rules.

I guess doing:
  # ATA over Ethernet
  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk", MODE="0220"
  SUBSYSTEM="aoe", KERNEL="err", MODE="0440"
would be all we need.

Kay

Re: [PATCH] Standard rules for AoE devices.

[Ryan Thomas]

Kay Sievers wrote:
> On Tue, Dec 9, 2008 at 01:49, Marco d'Itri <md@linux.it> wrote:
>> On Dec 09, Ryan Thomas <ryan@coraid.com> wrote:
>>
>>> In the interest of standardizing udev rules, please consider the
>>> following patch that adds udev rules for the ATA over Ethernet character
>>> and block devices.  The aoe module has been a long-time member of the
>>> kernel and needs inclusion in the standard udev rules.
>> Debian so far has used these rules which are more generic and much
>> simpler. Are they incorrect in some way?
>>
>> # AOE character devices
>> SUBSYSTEM="aoe",               NAME="etherd/%k"
>>
>> SUBSYSTEM="aoe",               MODE="0220",    GROUP="disk"
>> SUBSYSTEM="aoe", KERNEL="err", MODE="0440",   GROUP="disk"
> 
> That looks fine.
> 
> I wonder why we have this comment:
>   +# aoe block devices
>   +SUBSYSTEM="aoe", KERNEL="etherd*",    NAME="%k", GROUP="disk"
> 
> They can not be block devices if they have SUBSYSTEM="aoe". Do the
> etherd* devices belong into the subdir or not?
> 
> We have:
>   SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk"
> in the current default rules.
> 
> I guess doing:
>   # ATA over Ethernet
>   SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk", MODE="0220"
>   SUBSYSTEM="aoe", KERNEL="err", MODE="0440"
> would be all we need.
> 
> Kay
> 

My previous patch should be backed out as you are right that it's not 
needed.

A patch to the default rules to correct the mode on the etherd/err 
character device is all that is needed.

Thanks for the help straightening this (and me) out.
Ryan

---
  rules/rules.d/50-udev-default.rules |    2 +-
  1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/rules/rules.d/50-udev-default.rules 
b/rules/rules.d/50-udev-default.rules
index 2743dcb..0f27982 100644
--- a/rules/rules.d/50-udev-default.rules
+++ b/rules/rules.d/50-udev-default.rules
@@ -87,7 +87,7 @@ KERNEL="pktcdvd[0-9]*",      NAME="pktcdvd/%k"
  KERNEL="pktcdvd",             NAME="pktcdvd/control"
  KERNEL="qft0",                        SYMLINK+="ftape"
  SUBSYSTEM="bsg",              NAME="bsg/%k"
-SUBSYSTEM="aoe",              NAME="etherd/%k", GROUP="disk"
+SUBSYSTEM="aoe",              NAME="etherd/%k", GROUP="disk", MODE="0220"
  SUBSYSTEM="aoe", KERNEL="err", MODE="0440"

  # network
-- 
1.5.6.5

Re: [PATCH] Standard rules for AoE devices.

[Kay Sievers]

On Tue, Dec 9, 2008 at 04:07, Ryan Thomas <ryan@coraid.com> wrote:
> Kay Sievers wrote:
>> On Tue, Dec 9, 2008 at 01:49, Marco d'Itri <md@linux.it> wrote:
>>> On Dec 09, Ryan Thomas <ryan@coraid.com> wrote:
>>>
>>>> In the interest of standardizing udev rules, please consider the
>>>> following patch that adds udev rules for the ATA over Ethernet character
>>>> and block devices.  The aoe module has been a long-time member of the
>>>> kernel and needs inclusion in the standard udev rules.
>>>
>>> Debian so far has used these rules which are more generic and much
>>> simpler. Are they incorrect in some way?
>>>
>>> # AOE character devices
>>> SUBSYSTEM="aoe",               NAME="etherd/%k"
>>>
>>> SUBSYSTEM="aoe",               MODE="0220",    GROUP="disk"
>>> SUBSYSTEM="aoe", KERNEL="err", MODE="0440",   GROUP="disk"
>>
>> That looks fine.
>>
>> I wonder why we have this comment:
>>  +# aoe block devices
>>  +SUBSYSTEM="aoe", KERNEL="etherd*",    NAME="%k", GROUP="disk"
>>
>> They can not be block devices if they have SUBSYSTEM="aoe". Do the
>> etherd* devices belong into the subdir or not?
>>
>> We have:
>>  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk"
>> in the current default rules.
>>
>> I guess doing:
>>  # ATA over Ethernet
>>  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk", MODE="0220"
>>  SUBSYSTEM="aoe", KERNEL="err", MODE="0440"
>> would be all we need.

> My previous patch should be backed out as you are right that it's not
> needed.
>
> A patch to the default rules to correct the mode on the etherd/err character
> device is all that is needed.

Applied. Please check:
  http://git.kernel.org/?p=linux/hotplug/udev.git;a=commitdiff;h/c635fb67dbd509f4ee532c8bf0d3885cc71989

The block device are never moved to the subdir, they all just stay in
/dev, right?

Thanks,
Kay

Re: [PATCH] Standard rules for AoE devices.

[Ryan Thomas]

Kay Sievers wrote:
> On Tue, Dec 9, 2008 at 04:07, Ryan Thomas <ryan@coraid.com> wrote:
>> Kay Sievers wrote:
>>> On Tue, Dec 9, 2008 at 01:49, Marco d'Itri <md@linux.it> wrote:
>>>> On Dec 09, Ryan Thomas <ryan@coraid.com> wrote:
>>>>
>>>>> In the interest of standardizing udev rules, please consider the
>>>>> following patch that adds udev rules for the ATA over Ethernet character
>>>>> and block devices.  The aoe module has been a long-time member of the
>>>>> kernel and needs inclusion in the standard udev rules.
>>>> Debian so far has used these rules which are more generic and much
>>>> simpler. Are they incorrect in some way?
>>>>
>>>> # AOE character devices
>>>> SUBSYSTEM="aoe",               NAME="etherd/%k"
>>>>
>>>> SUBSYSTEM="aoe",               MODE="0220",    GROUP="disk"
>>>> SUBSYSTEM="aoe", KERNEL="err", MODE="0440",   GROUP="disk"
>>> That looks fine.
>>>
>>> I wonder why we have this comment:
>>>  +# aoe block devices
>>>  +SUBSYSTEM="aoe", KERNEL="etherd*",    NAME="%k", GROUP="disk"
>>>
>>> They can not be block devices if they have SUBSYSTEM="aoe". Do the
>>> etherd* devices belong into the subdir or not?
>>>
>>> We have:
>>>  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk"
>>> in the current default rules.
>>>
>>> I guess doing:
>>>  # ATA over Ethernet
>>>  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk", MODE="0220"
>>>  SUBSYSTEM="aoe", KERNEL="err", MODE="0440"
>>> would be all we need.
> 
>> My previous patch should be backed out as you are right that it's not
>> needed.
>>
>> A patch to the default rules to correct the mode on the etherd/err character
>> device is all that is needed.
> 
> Applied. Please check:
>   http://git.kernel.org/?p=linux/hotplug/udev.git;a=commitdiff;h/c635fb67dbd509f4ee532c8bf0d3885cc71989
> 
> The block device are never moved to the subdir, they all just stay in
> /dev, right?
> 

The block device does go in the subdir.  The kernel name for the block 
device is "etherd!eX.Y" which means that the subdir is automatically 
taken care of.

Ryan

Re: [PATCH] Standard rules for AoE devices.

[Kay Sievers]

On Tue, Dec 9, 2008 at 12:23, Ryan Thomas <ryan@coraid.com> wrote:
> Kay Sievers wrote:
>> On Tue, Dec 9, 2008 at 04:07, Ryan Thomas <ryan@coraid.com> wrote:
>>> Kay Sievers wrote:
>>>> On Tue, Dec 9, 2008 at 01:49, Marco d'Itri <md@linux.it> wrote:
>>>>> On Dec 09, Ryan Thomas <ryan@coraid.com> wrote:
>>>>>
>>>>>> In the interest of standardizing udev rules, please consider the
>>>>>> following patch that adds udev rules for the ATA over Ethernet
>>>>>> character
>>>>>> and block devices.  The aoe module has been a long-time member of the
>>>>>> kernel and needs inclusion in the standard udev rules.
>>>>>
>>>>> Debian so far has used these rules which are more generic and much
>>>>> simpler. Are they incorrect in some way?
>>>>>
>>>>> # AOE character devices
>>>>> SUBSYSTEM="aoe",               NAME="etherd/%k"
>>>>>
>>>>> SUBSYSTEM="aoe",               MODE="0220",    GROUP="disk"
>>>>> SUBSYSTEM="aoe", KERNEL="err", MODE="0440",   GROUP="disk"
>>>>
>>>> That looks fine.
>>>>
>>>> I wonder why we have this comment:
>>>>  +# aoe block devices
>>>>  +SUBSYSTEM="aoe", KERNEL="etherd*",    NAME="%k", GROUP="disk"
>>>>
>>>> They can not be block devices if they have SUBSYSTEM="aoe". Do the
>>>> etherd* devices belong into the subdir or not?
>>>>
>>>> We have:
>>>>  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk"
>>>> in the current default rules.
>>>>
>>>> I guess doing:
>>>>  # ATA over Ethernet
>>>>  SUBSYSTEM="aoe", NAME="etherd/%k", GROUP="disk", MODE="0220"
>>>>  SUBSYSTEM="aoe", KERNEL="err", MODE="0440"
>>>> would be all we need.
>>
>>> My previous patch should be backed out as you are right that it's not
>>> needed.
>>>
>>> A patch to the default rules to correct the mode on the etherd/err
>>> character
>>> device is all that is needed.
>>
>> Applied. Please check:
>>
>>  http://git.kernel.org/?p=linux/hotplug/udev.git;a=commitdiff;h/c635fb67dbd509f4ee532c8bf0d3885cc71989
>>
>> The block device are never moved to the subdir, they all just stay in
>> /dev, right?
>
> The block device does go in the subdir.  The kernel name for the block
> device is "etherd!eX.Y" which means that the subdir is automatically taken
> care of.

Ah, nice, so it should be fine.

The char devices should have done just the same '!' thing, then no
rule, besides the permissions would be needed. :)

Thanks,
Kay