[PATCH v2] allow ifup bring up network manually even without netroot

[PATCH v2] allow ifup bring up network manually even without netroot

[Dave Young]

For kdump we need scp vmcore to remote machine, the nic to be used is
not limited to netroot one. we need a feature for manually bringing up
 network interface. Also it is useful for emergency shell with
ssh-client for recovery or test purpose

I implement this by adding one argument to ifup script, user can use
`/sbin/ifup eth0 -m` to bring up eth0, note ifup will regard it a
manual operation for the nic specified in 1st argument if there's
the 2nd argument.

If same nic is used for netroot the 2nd argument will be ignored,
in this case we will leave netroot bring up it automatically to
avoid side effect. And in this case hooks such as kdump will need to
execute after netroot mounted.

`ifup eth0 -m` will create /tmp/net.eth0.manualup stamp file,
later dhclient-script can check this and pass $2 to netroot,
then netroot script will bring eth0 up

Thanks for comments and suggestions from David Dillow.

Signed-off-by: Dave Young <dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
---
 modules.d/40network/dhclient-script |    8 ++
 modules.d/40network/ifup            |   97 ++++++++++++++++++++----------------
 modules.d/40network/netroot         |   86 ++++++++++++++++++-------------
 3 files changed, 111 insertions(+), 80 deletions(-)

Index: dracut/modules.d/40network/dhclient-script
===================================================================
--- dracut.orig/modules.d/40network/dhclient-script
+++ dracut/modules.d/40network/dhclient-script
@@ -74,7 +74,13 @@ case $reason in
             echo "$line"
         done >/tmp/dhclient.$netif.dhcpopts
         echo online > /sys/class/net/$netif/uevent
-        initqueue --onetime --name netroot-$netif netroot $netif
+
+        if [ -e /tmp/net.$netif.manualup ]; then
+            /sbin/netroot $netif -m
+            rm -f /tmp/net.$netif.manualup
+        else
+            initqueue --onetime --name netroot-$netif netroot $netif
+        fi
         ;;
     *) echo "dhcp: $reason";;
 esac
Index: dracut/modules.d/40network/ifup
===================================================================
--- dracut.orig/modules.d/40network/ifup
+++ dracut/modules.d/40network/ifup
@@ -5,9 +5,54 @@
 # We don't need to check for ip= errors here, that is handled by the
 # cmdline parser script
 #
+# without $2 means this is for real netroot case
+# or it is for manually bring up network ie. for kdump scp vmcore
 PATH=/usr/sbin:/usr/bin:/sbin:/bin
 
 type getarg >/dev/null 2>&1 || . /lib/dracut-lib.sh
+export PS4="ifup.$1.$$ + "
+exec >>/run/initramfs/loginit.pipe 2>>/run/initramfs/loginit.pipe
+type getarg >/dev/null 2>&1 || . /lib/dracut-lib.sh
+
+# Huh? No $1?
+[ -z "$1" ] && exit 1
+
+# $netif reads easier than $1
+netif=$1
+
+# enslave this interface to bond?
+if [ -e /tmp/bond.info ]; then
+    . /tmp/bond.info
+    for slave in $bondslaves ; do
+        if [ "$netif" = "$slave" ] ; then
+            netif=$bondname
+        fi
+    done
+fi
+
+# bridge this interface?
+if [ -e /tmp/bridge.info ]; then
+    . /tmp/bridge.info
+    if [ "$netif" = "$ethname" ]; then
+        if [ "$netif" = "$bondname" ] && [ -n "$DO_BOND_SETUP" ] ; then
+            : # We need to really setup bond (recursive call)
+        else
+            netif="$bridgename"
+        fi
+    fi
+fi
+
+# bail immediately if the interface is already up
+# or we don't need the network
+[ -f "/tmp/net.$netif.up" ] && exit 0
+[ -f "/tmp/root.info" ] || exit 0
+. /tmp/root.info
+
+# disable manual ifup while netroot is set for simplifying our logic
+# in netroot case we prefer netroot to bringup $netif automaticlly
+[ -n "$2" ] && [ -z "$netroot" ] && manualup="$2"
+[ -z "$netroot" ] && [ -z "$manualup" ] && exit 0
+[ -n "$manualup" ] && >/tmp/net.$netif.manualup
 
 # Run dhclient
 do_dhcp() {
@@ -50,7 +95,11 @@ do_ipv6auto() {
 
 
     echo online > /sys/class/net/$netif/uevent
-    initqueue --onetime --name netroot-$netif netroot $netif
+    if [ -n "$manualup" ]; then
+        /sbin/netroot $netif -m
+    else
+        initqueue --onetime --name netroot-$netif netroot $netif
+    fi
 }
 
 # Handle static ip configuration
@@ -77,47 +126,12 @@ do_static() {
     fi >> /tmp/net.$netif.resolv.conf
 
     echo online > /sys/class/net/$netif/uevent
-    initqueue --onetime --name netroot-$netif netroot $netif
-}
-
-export PS4="ifup.$1.$$ + "
-exec >>/run/initramfs/loginit.pipe 2>>/run/initramfs/loginit.pipe
-type getarg >/dev/null 2>&1 || . /lib/dracut-lib.sh
-
-# Huh? No $1?
-[ -z "$1" ] && exit 1
-
-# $netif reads easier than $1
-netif=$1
-
-# enslave this interface to bond?
-if [ -e /tmp/bond.info ]; then
-    . /tmp/bond.info
-    for slave in $bondslaves ; do
-        if [ "$netif" = "$slave" ] ; then
-            netif=$bondname
-        fi
-    done
-fi
-
-# bridge this interface?
-if [ -e /tmp/bridge.info ]; then
-    . /tmp/bridge.info
-    if [ "$netif" = "$ethname" ]; then
-        if [ "$netif" = "$bondname" ] && [ -n "$DO_BOND_SETUP" ] ; then
-            : # We need to really setup bond (recursive call)
-        else
-            netif="$bridgename"
-        fi
+    if [ -n "$manualup" ]; then
+        /sbin/netroot $netif -m
+    else
+        initqueue --onetime --name netroot-$netif netroot $netif
     fi
-fi
-
-# bail immediately if the interface is already up
-# or we don't need the network
-[ -f "/tmp/net.$netif.up" ] && exit 0
-[ -f "/tmp/root.info" ] || exit 0
-. /tmp/root.info
-[ -z "$netroot" ] && exit 0
+}
 
 # loopback is always handled the same way
 if [ "$netif" = "lo" ] ; then
@@ -228,5 +242,4 @@ for p in $(getargs ip=); do
     esac
     break
 done
-
 exit 0
Index: dracut/modules.d/40network/netroot
===================================================================
--- dracut.orig/modules.d/40network/netroot
+++ dracut/modules.d/40network/netroot
@@ -3,7 +3,6 @@
 # ex: ts=8 sw=4 sts=4 et filetype=sh
 
 PATH=/usr/sbin:/usr/bin:/sbin:/bin
-
 type getarg >/dev/null 2>&1 || . /lib/dracut-lib.sh
 
 # Huh? Empty $1?
@@ -12,12 +11,16 @@ type getarg >/dev/null 2>&1 || . /lib/dr
 # Huh? No interface config?
 [ ! -e /tmp/net.$1.up ] && exit 1
 
-# There's no sense in doing something if no (net)root info is available
+# [ ! -z $2 ] means this is for manually bringing up network
+# instead of real netroot; If It's called without $2, then there's
+# no sense in doing something if no (net)root info is available
 # or root is already there
 [ -e /tmp/root.info ] || exit 1
 . /tmp/root.info
-[ -d $NEWROOT/proc ] && exit 0
-[ -z "$netroot" ] && exit 1
+if [ -z "$2" ]; then
+    [ -d $NEWROOT/proc ] && exit 0
+    [ -z "$netroot" ] && exit 1
+fi
 
 # Let's see if we have to wait for other interfaces
 # Note: exit works just fine, since the last interface to be
@@ -31,43 +34,49 @@ done
 netif=$1
 [ -e "/tmp/net.bootdev" ] && read netif < /tmp/net.bootdev
 
+if [ -e /tmp/net.$netif.manualup ]; then
+    rm -f /tmp/net.$netif.manualup
+fi
+
 # Figure out the handler for root=dhcp by recalling all netroot cmdline
-# handlers
-if [ "$netroot" = "dhcp" ] || [ "$netroot" = "dhcp6" ] ; then
-    # Unset root so we can check later
-    unset root
-
-    # Load dhcp options
-    [ -e /tmp/dhclient.$netif.dhcpopts ] && . /tmp/dhclient.$netif.dhcpopts
-
-    # If we have a specific bootdev with no dhcpoptions or empty root-path,
-    # we die. Otherwise we just warn
-    if [ -z "$new_root_path" ] ; then
-        [ -n "$BOOTDEV" ] && die "No dhcp root-path received for '$BOOTDEV'"
-        warn "No dhcp root-path received for '$BOOTDEV' trying other interfaces if available"
-        exit 1
-    fi
+# handlers when this is not called from manually network bringing up.
+if [ -z "$2" ]; then
+    if [ "$netroot" = "dhcp" ] || [ "$netroot" = "dhcp6" ] ; then
+        # Unset root so we can check later
+        unset root
+
+        # Load dhcp options
+        [ -e /tmp/dhclient.$netif.dhcpopts ] && . /tmp/dhclient.$netif.dhcpopts
+
+        # If we have a specific bootdev with no dhcpoptions or empty root-path,
+        # we die. Otherwise we just warn
+        if [ -z "$new_root_path" ] ; then
+            [ -n "$BOOTDEV" ] && die "No dhcp root-path received for '$BOOTDEV'"
+            warn "No dhcp root-path received for '$BOOTDEV' trying other interfaces if available"
+            exit 1
+        fi
 
-    # Set netroot to new_root_path, so cmdline parsers don't call
-    netroot=$new_root_path
+        # Set netroot to new_root_path, so cmdline parsers don't call
+        netroot=$new_root_path
 
-    # FIXME!
-    for f in $hookdir/cmdline/90*.sh; do
-        [ -f "$f" ] && . "$f";
-    done
-else
-    rootok="1"
-fi
+        # FIXME!
+        for f in $hookdir/cmdline/90*.sh; do
+            [ -f "$f" ] && . "$f";
+        done
+    else
+        rootok="1"
+    fi
 
-# Check: do we really know how to handle (net)root?
-[ -z "$root" ] && die "No or empty root= argument"
-[ -z "$rootok" ] && die "Don't know how to handle 'root=$root'"
-
-handler=${netroot%%:*}
-handler=${handler%%4}
-handler=$(command -v ${handler}root)
-if [ -z "$netroot" ] || [ ! -e "$handler" ] ; then
-    die "No handler for netroot type '$netroot'"
+    # Check: do we really know how to handle (net)root?
+    [ -z "$root" ] && die "No or empty root= argument"
+    [ -z "$rootok" ] && die "Don't know how to handle 'root=$root'"
+
+    handler=${netroot%%:*}
+    handler=${handler%%4}
+    handler=$(command -v ${handler}root)
+    if [ -z "$netroot" ] || [ ! -e "$handler" ] ; then
+        die "No handler for netroot type '$netroot'"
+    fi
 fi
 
 # We're here, so we can assume that upping interfaces is now ok
@@ -122,6 +131,9 @@ if [ -n "$dest" ] && ! arping -q -f -w 6
     dinfo "Resolving $dest via ARP on $netif failed"
 fi
 
+# exit in case manually bring up network
+[ -n "$2" ] && exit 0
+
 # Source netroot hooks before we start the handler
 source_all netroot
 

Re: [PATCH v2] allow ifup bring up network manually even without netroot

[Cong Wang]

On Fri, 30 Dec 2011 at 08:38 GMT, Dave Young <dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
> For kdump we need scp vmcore to remote machine, the nic to be used is
> not limited to netroot one. we need a feature for manually bringing up
>  network interface. Also it is useful for emergency shell with
> ssh-client for recovery or test purpose
>
> I implement this by adding one argument to ifup script, user can use
> `/sbin/ifup eth0 -m` to bring up eth0, note ifup will regard it a
> manual operation for the nic specified in 1st argument if there's
> the 2nd argument.
>
> If same nic is used for netroot the 2nd argument will be ignored,
> in this case we will leave netroot bring up it automatically to
> avoid side effect. And in this case hooks such as kdump will need to
> execute after netroot mounted.
>
> `ifup eth0 -m` will create /tmp/net.eth0.manualup stamp file,
> later dhclient-script can check this and pass $2 to netroot,
> then netroot script will bring eth0 up
>

As we talked on IRC, we still need to discuss if this is really
needed, from kdump perspective. If we need to invoke `ifup` by
ourselves, then we also need to determine which NIC we should bring
up, this will make kdump code larger.

So, I hope, we can have some kernel cmdline (like ip=...) to tell
dracut which NIC we want to bring up in the second kernel, provided by
either ssh-client module or nfs module.

Let's see what Harald thinks...

Re: [PATCH v2] allow ifup bring up network manually even without netroot

[Dave Young]

Hi, I have some additional comments

On 01/10/2012 11:41 PM, Cong Wang wrote:

> On Fri, 30 Dec 2011 at 08:38 GMT, Dave Young <dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
>> For kdump we need scp vmcore to remote machine, the nic to be used is
>> not limited to netroot one. we need a feature for manually bringing up
>>  network interface. Also it is useful for emergency shell with
>> ssh-client for recovery or test purpose
>>
>> I implement this by adding one argument to ifup script, user can use
>> `/sbin/ifup eth0 -m` to bring up eth0, note ifup will regard it a
>> manual operation for the nic specified in 1st argument if there's
>> the 2nd argument.
>>
>> If same nic is used for netroot the 2nd argument will be ignored,
>> in this case we will leave netroot bring up it automatically to
>> avoid side effect. And in this case hooks such as kdump will need to
>> execute after netroot mounted.
>>
>> `ifup eth0 -m` will create /tmp/net.eth0.manualup stamp file,
>> later dhclient-script can check this and pass $2 to netroot,
>> then netroot script will bring eth0 up
>>
> 
> As we talked on IRC, we still need to discuss if this is really
> needed, from kdump perspective.


manually ifup nic also benefit to emergency shell use case. I want to
use dracut initramfs as a minimal system without real root for kernel
testing

> If we need to invoke `ifup` by
> ourselves, then we also need to determine which NIC we should bring
> up, this will make kdump code larger.
> 
> So, I hope, we can have some kernel cmdline (like ip=...) to tell
> dracut which NIC we want to bring up in the second kernel, provided by
> either ssh-client module or nfs module.


If we need to pass the nic name to cmdline then we still need to
determine which one should be use.

If we just bring up all nics, then use the manual ifup is also easy.

current dracut udev rules is: if there's no nic specified it will ifup
all possible nics, but all will be put in initqueue which finally be
brought up by netroot script.

> 
> Let's see what Harald thinks...

> 
> --
> To unsubscribe from this list: send the line "unsubscribe initramfs" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html



-- 
Thanks
Dave

Re: [PATCH v2] allow ifup bring up network manually even without netroot

[Dave Young]

On 01/11/2012 10:05 AM, Dave Young wrote:

> Hi, I have some additional comments
> 
> On 01/10/2012 11:41 PM, Cong Wang wrote:
> 
>> On Fri, 30 Dec 2011 at 08:38 GMT, Dave Young <dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
>>> For kdump we need scp vmcore to remote machine, the nic to be used is
>>> not limited to netroot one. we need a feature for manually bringing up
>>>  network interface. Also it is useful for emergency shell with
>>> ssh-client for recovery or test purpose
>>>
>>> I implement this by adding one argument to ifup script, user can use
>>> `/sbin/ifup eth0 -m` to bring up eth0, note ifup will regard it a
>>> manual operation for the nic specified in 1st argument if there's
>>> the 2nd argument.
>>>
>>> If same nic is used for netroot the 2nd argument will be ignored,
>>> in this case we will leave netroot bring up it automatically to
>>> avoid side effect. And in this case hooks such as kdump will need to
>>> execute after netroot mounted.
>>>
>>> `ifup eth0 -m` will create /tmp/net.eth0.manualup stamp file,
>>> later dhclient-script can check this and pass $2 to netroot,
>>> then netroot script will bring eth0 up
>>>
>>
>> As we talked on IRC, we still need to discuss if this is really
>> needed, from kdump perspective.
> 
> 
> manually ifup nic also benefit to emergency shell use case. I want to
> use dracut initramfs as a minimal system without real root for kernel
> testing
> 
>> If we need to invoke `ifup` by
>> ourselves, then we also need to determine which NIC we should bring
>> up, this will make kdump code larger.
>>
>> So, I hope, we can have some kernel cmdline (like ip=...) to tell
>> dracut which NIC we want to bring up in the second kernel, provided by
>> either ssh-client module or nfs module.
> 
> 
> If we need to pass the nic name to cmdline then we still need to
> determine which one should be use.
> 
> If we just bring up all nics, then use the manual ifup is also easy.
> 
> current dracut udev rules is: if there's no nic specified it will ifup
> all possible nics, but all will be put in initqueue which finally be
> brought up by netroot script.


And if without netroot cmdline set netroot script will just fail and
return, so in this case ifup will act as nop

> 
>>
>> Let's see what Harald thinks...
> 
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe initramfs" in
>> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> 
> 



-- 
Thanks
Dave

Re: [PATCH v2] allow ifup bring up network manually even without netroot

[Cong Wang]

On 01/11/2012 10:05 AM, Dave Young wrote:
> Hi, I have some additional comments
>
> On 01/10/2012 11:41 PM, Cong Wang wrote:
>
>> On Fri, 30 Dec 2011 at 08:38 GMT, Dave Young<dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>  wrote:
>>> For kdump we need scp vmcore to remote machine, the nic to be used is
>>> not limited to netroot one. we need a feature for manually bringing up
>>>   network interface. Also it is useful for emergency shell with
>>> ssh-client for recovery or test purpose
>>>
>>> I implement this by adding one argument to ifup script, user can use
>>> `/sbin/ifup eth0 -m` to bring up eth0, note ifup will regard it a
>>> manual operation for the nic specified in 1st argument if there's
>>> the 2nd argument.
>>>
>>> If same nic is used for netroot the 2nd argument will be ignored,
>>> in this case we will leave netroot bring up it automatically to
>>> avoid side effect. And in this case hooks such as kdump will need to
>>> execute after netroot mounted.
>>>
>>> `ifup eth0 -m` will create /tmp/net.eth0.manualup stamp file,
>>> later dhclient-script can check this and pass $2 to netroot,
>>> then netroot script will bring eth0 up
>>>
>>
>> As we talked on IRC, we still need to discuss if this is really
>> needed, from kdump perspective.
>
>
> manually ifup nic also benefit to emergency shell use case. I want to
> use dracut initramfs as a minimal system without real root for kernel
> testing


For emergency shell, yeah, it make senses.

>
>> If we need to invoke `ifup` by
>> ourselves, then we also need to determine which NIC we should bring
>> up, this will make kdump code larger.
>>
>> So, I hope, we can have some kernel cmdline (like ip=...) to tell
>> dracut which NIC we want to bring up in the second kernel, provided by
>> either ssh-client module or nfs module.
>
>
> If we need to pass the nic name to cmdline then we still need to
> determine which one should be use.


Sure, but this will go to ssh-client or nfs module, not kdump.

>
> If we just bring up all nics, then use the manual ifup is also easy.
>


If there are lots of NIC, and we only use one (for remote kdump),
bringing all NIC up is a waste. Nor to say when we only use eth0, you 
also bring up bond0, br0 etc...

> current dracut udev rules is: if there's no nic specified it will ifup
> all possible nics, but all will be put in initqueue which finally be
> brought up by netroot script.


We need to specify at least one, the one that reach the remote dump server.

Thanks.

Re: [PATCH v2] allow ifup bring up network manually even without netroot

[Dave Young]

On 01/11/2012 05:48 PM, Cong Wang wrote:

> On 01/11/2012 10:05 AM, Dave Young wrote:
>> Hi, I have some additional comments
>>
>> On 01/10/2012 11:41 PM, Cong Wang wrote:
>>
>>> On Fri, 30 Dec 2011 at 08:38 GMT, Dave Young<dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>  wrote:
>>>> For kdump we need scp vmcore to remote machine, the nic to be used is
>>>> not limited to netroot one. we need a feature for manually bringing up
>>>>   network interface. Also it is useful for emergency shell with
>>>> ssh-client for recovery or test purpose
>>>>
>>>> I implement this by adding one argument to ifup script, user can use
>>>> `/sbin/ifup eth0 -m` to bring up eth0, note ifup will regard it a
>>>> manual operation for the nic specified in 1st argument if there's
>>>> the 2nd argument.
>>>>
>>>> If same nic is used for netroot the 2nd argument will be ignored,
>>>> in this case we will leave netroot bring up it automatically to
>>>> avoid side effect. And in this case hooks such as kdump will need to
>>>> execute after netroot mounted.
>>>>
>>>> `ifup eth0 -m` will create /tmp/net.eth0.manualup stamp file,
>>>> later dhclient-script can check this and pass $2 to netroot,
>>>> then netroot script will bring eth0 up
>>>>
>>>
>>> As we talked on IRC, we still need to discuss if this is really
>>> needed, from kdump perspective.
>>
>>
>> manually ifup nic also benefit to emergency shell use case. I want to
>> use dracut initramfs as a minimal system without real root for kernel
>> testing
> 
> 
> For emergency shell, yeah, it make senses.
> 
>>
>>> If we need to invoke `ifup` by
>>> ourselves, then we also need to determine which NIC we should bring
>>> up, this will make kdump code larger.
>>>
>>> So, I hope, we can have some kernel cmdline (like ip=...) to tell
>>> dracut which NIC we want to bring up in the second kernel, provided by
>>> either ssh-client module or nfs module.
>>
>>
>> If we need to pass the nic name to cmdline then we still need to
>> determine which one should be use.
> 
> 
> Sure, but this will go to ssh-client or nfs module, not kdump.


ssh-client does not pass any cmdline, instead the scripts of call dracut
to generate initrd do this. In case of kdump it's the kdump init scripts
or simplified mkdumprd, so we could not skip this step.

> 
>>
>> If we just bring up all nics, then use the manual ifup is also easy.
>>
> 
> 
> If there are lots of NIC, and we only use one (for remote kdump),
> bringing all NIC up is a waste. Nor to say when we only use eth0, you
> also bring up bond0, br0 etc...
> 
>> current dracut udev rules is: if there's no nic specified it will ifup
>> all possible nics, but all will be put in initqueue which finally be
>> brought up by netroot script.
> 
> 
> We need to specify at least one, the one that reach the remote dump server.
> 
> Thanks.



-- 
Thanks
Dave

Re: [PATCH v2] allow ifup bring up network manually even without netroot

[Cong Wang]

On 01/12/2012 10:07 AM, Dave Young wrote:
>>>> If we need to invoke `ifup` by
>>>> ourselves, then we also need to determine which NIC we should bring
>>>> up, this will make kdump code larger.
>>>>
>>>> So, I hope, we can have some kernel cmdline (like ip=...) to tell
>>>> dracut which NIC we want to bring up in the second kernel, provided by
>>>> either ssh-client module or nfs module.
>>>
>>>
>>> If we need to pass the nic name to cmdline then we still need to
>>> determine which one should be use.
>>
>>
>> Sure, but this will go to ssh-client or nfs module, not kdump.
>
>
> ssh-client does not pass any cmdline, instead the scripts of call dracut
> to generate initrd do this. In case of kdump it's the kdump init scripts
> or simplified mkdumprd, so we could not skip this step.

Well, I still think ssh-client or nfs module should handle the network, 
not kdump's mkdumprd, like lvm module for lvm.* cmdline.

But anyway, I will take this patch to my kdump tree.

Thanks.