2007-12-26 02:59:22

by Simon Horman

[permalink] [raw]
Subject: [patch] ipv4: Update ip command line processing (take II)

Recently the documentation in Documentation/nfsroot.txt was
update to note that in fact ip=off and ip=::::::off as the
latter is ignored and the default (on) is used.

This was certainly a step in the direction of reducing confusion.
But it seems to me that the code ought to be fixed up so that
ip=::::::off actually turns off ip autoconfiguration.

This patch also notes more specifically that ip=on (aka ip=::::::on)
is the default.

Cc: Amos Waterland <[email protected]>
Signed-off-by: Simon Horman <[email protected]>

---
Fri, 21 Dec 2007 16:48:16 +0900
* Removed check on if (name) in ic_proto_name().
It should have been if (*name == '\0'), but even then its
a bit bogus, providing an early exit for a code-path that
is called approximately once.

Index: net-2.6.25/net/ipv4/ipconfig.c
===================================================================
--- net-2.6.25.orig/net/ipv4/ipconfig.c 2007-12-26 11:56:20.000000000 +0900
+++ net-2.6.25/net/ipv4/ipconfig.c 2007-12-26 11:57:04.000000000 +0900
@@ -1417,6 +1417,10 @@ static int __init ic_proto_name(char *na
if (!strcmp(name, "on") || !strcmp(name, "any")) {
return 1;
}
+ if (!strcmp(name, "off") || !strcmp(name, "none")) {
+ ic_enable = 0;
+ return 1;
+ }
#ifdef CONFIG_IP_PNP_DHCP
else if (!strcmp(name, "dhcp")) {
ic_proto_enabled &= ~IC_RARP;
@@ -1451,12 +1455,6 @@ static int __init ip_auto_config_setup(c

ic_set_manually = 1;

- ic_enable = (*addrs &&
- (strcmp(addrs, "off") != 0) &&
- (strcmp(addrs, "none") != 0));
- if (!ic_enable)
- return 1;
-
if (ic_proto_name(addrs))
return 1;

Index: net-2.6.25/Documentation/nfsroot.txt
===================================================================
--- net-2.6.25.orig/Documentation/nfsroot.txt 2007-12-26 11:56:20.000000000 +0900
+++ net-2.6.25/Documentation/nfsroot.txt 2007-12-26 11:57:04.000000000 +0900
@@ -97,10 +97,6 @@ ip=<client-ip>:<server-ip>:<gw-ip>:<netm
autoconfiguration will take place. The most common way to use this
is "ip=dhcp".

- Note that "ip=off" is not the same thing as "ip=::::::off", because in
- the latter autoconfiguration will take place if any of DHCP, BOOTP or RARP
- are compiled in the kernel.
-
<client-ip> IP address of the client.

Default: Determined using autoconfiguration.
@@ -150,6 +146,7 @@ ip=<client-ip>:<server-ip>:<gw-ip>:<netm

off or none: don't use autoconfiguration
on or any: use any protocol available in the kernel
+ (default)
dhcp: use DHCP
bootp: use BOOTP
rarp: use RARP


2007-12-26 04:56:00

by David Miller

[permalink] [raw]
Subject: Re: [patch] ipv4: Update ip command line processing (take II)

From: Simon Horman <[email protected]>
Date: Wed, 26 Dec 2007 11:59:05 +0900

> Recently the documentation in Documentation/nfsroot.txt was
> update to note that in fact ip=off and ip=::::::off as the
> latter is ignored and the default (on) is used.
>
> This was certainly a step in the direction of reducing confusion.
> But it seems to me that the code ought to be fixed up so that
> ip=::::::off actually turns off ip autoconfiguration.
>
> This patch also notes more specifically that ip=on (aka ip=::::::on)
> is the default.
>
> Cc: Amos Waterland <[email protected]>
> Signed-off-by: Simon Horman <[email protected]>

I applied this to net-2.6 since I think it's a bug fix, and
I also tailored the changelog header to more it clear this
is a fix not just an "Update". :-)

Thanks!

2007-12-26 05:04:52

by Simon Horman

[permalink] [raw]
Subject: Re: [patch] ipv4: Update ip command line processing (take II)

On Tue, Dec 25, 2007 at 08:55:43PM -0800, David Miller wrote:
> From: Simon Horman <[email protected]>
> Date: Wed, 26 Dec 2007 11:59:05 +0900
>
> > Recently the documentation in Documentation/nfsroot.txt was
> > update to note that in fact ip=off and ip=::::::off as the
> > latter is ignored and the default (on) is used.
> >
> > This was certainly a step in the direction of reducing confusion.
> > But it seems to me that the code ought to be fixed up so that
> > ip=::::::off actually turns off ip autoconfiguration.
> >
> > This patch also notes more specifically that ip=on (aka ip=::::::on)
> > is the default.
> >
> > Cc: Amos Waterland <[email protected]>
> > Signed-off-by: Simon Horman <[email protected]>
>
> I applied this to net-2.6 since I think it's a bug fix, and
> I also tailored the changelog header to more it clear this
> is a fix not just an "Update". :-)

Ok, thanks.

--
Horms

2008-01-06 03:59:17

by Amos Waterland

[permalink] [raw]
Subject: [PATCH] Fix regression in ip command line processing

The recent changes for ip command line processing fixed some problems
but unfortunately broke some common usage scenarios. In current
2.6.24-rc6 the following command line results in no IP address
assignment, which is surely a regression:

ip=10.0.2.15::10.0.2.2:255.255.255.0::eth0:off

Please find below a patch that works for all cases I can find.

Signed-off-by: Amos Waterland <[email protected]>

---

For convenience, here are the test cases phrased as qemu invocations:

ADDRESS ASSIGNED

qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=on"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=dhcp"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=both"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=any"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::on"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::dhcp"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=10.0.2.15::10.0.2.2:255.255.255.0::eth0:off"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=10.0.2.15::10.0.2.2:255.255.255.0::eth0:on"

ADDRESS NOT ASSIGNED

qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip="
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=off"
qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::off"

---

Documentation/nfsroot.txt | 1 +
net/ipv4/ipconfig.c | 19 +++++++++++++++----
2 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/Documentation/nfsroot.txt b/Documentation/nfsroot.txt
index c86dd38..1a4d199 100644
--- a/Documentation/nfsroot.txt
+++ b/Documentation/nfsroot.txt
@@ -145,6 +145,7 @@ ip=<client-ip>:<server-ip>:<gw-ip>:<netmask>:<hostname>:<device>:<autoconf>
this option.

off or none: don't use autoconfiguration
+ (do static IP assignment instead)
on or any: use any protocol available in the kernel
(default)
dhcp: use DHCP
diff --git a/net/ipv4/ipconfig.c b/net/ipv4/ipconfig.c
index 56a6757..8563b2e 100644
--- a/net/ipv4/ipconfig.c
+++ b/net/ipv4/ipconfig.c
@@ -1404,8 +1404,7 @@ static int __init ic_proto_name(char *name)
return 1;
}
if (!strcmp(name, "off") || !strcmp(name, "none")) {
- ic_enable = 0;
- return 1;
+ return 0;
}
#ifdef CONFIG_IP_PNP_DHCP
else if (!strcmp(name, "dhcp")) {
@@ -1442,10 +1441,20 @@ static int __init ip_auto_config_setup(char *addrs)
ic_set_manually = 1;
ic_enable = 1;

+ /*
+ * If any dhcp, bootp etc options are set, leave autoconfig on
+ * and skip the below static IP processing.
+ */
if (ic_proto_name(addrs))
return 1;

- /* Parse the whole string */
+ /* If no static IP is given, turn off autoconfig and bail. */
+ if (*addrs == 0 || strcmp(addrs, "off") == 0 || strcmp(addrs, "none") == 0) {
+ ic_enable = 0;
+ return 1;
+ }
+
+ /* Parse string for static IP assignment. */
ip = addrs;
while (ip && *ip) {
if ((cp = strchr(ip, ':')))
@@ -1483,7 +1492,9 @@ static int __init ip_auto_config_setup(char *addrs)
strlcpy(user_dev_name, ip, sizeof(user_dev_name));
break;
case 6:
- ic_proto_name(ip);
+ if (ic_proto_name(ip) == 0 && ic_myaddr == NONE) {
+ ic_enable = 0;
+ }
break;
}
}

2008-01-06 07:20:45

by David Miller

[permalink] [raw]
Subject: Re: [PATCH] Fix regression in ip command line processing

From: Amos Waterland <[email protected]>
Date: Sat, 5 Jan 2008 22:58:16 -0500

> ADDRESS ASSIGNED
>
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=on"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=dhcp"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=both"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=any"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::on"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::dhcp"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=10.0.2.15::10.0.2.2:255.255.255.0::eth0:off"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=10.0.2.15::10.0.2.2:255.255.255.0::eth0:on"
>
> ADDRESS NOT ASSIGNED
>
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip="
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=off"
> qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::off"

I really wish Simon had tested his original patch as extensively
as you have, this is the second major regression it has caused.
:-/

I'll apply your fix, although I'll break up some super long lines
you've added, thanks a lot.

2008-01-06 09:26:05

by Simon Horman

[permalink] [raw]
Subject: Re: [PATCH] Fix regression in ip command line processing

On Sat, Jan 05, 2008 at 11:20:33PM -0800, David Miller wrote:
> From: Amos Waterland <[email protected]>
> Date: Sat, 5 Jan 2008 22:58:16 -0500
>
> > ADDRESS ASSIGNED
> >
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=on"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=dhcp"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=both"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=any"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::on"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::dhcp"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=10.0.2.15::10.0.2.2:255.255.255.0::eth0:off"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=10.0.2.15::10.0.2.2:255.255.255.0::eth0:on"
> >
> > ADDRESS NOT ASSIGNED
> >
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip="
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=off"
> > qemu -kernel x86/arch/i386/boot/bzImage /dev/zero -append "ip=::::::off"
>
> I really wish Simon had tested his original patch as extensively
> as you have, this is the second major regression it has caused.
> :-/

Ok, this time around its an issue with my understainding of how things
are supposed to work. But sorry none the less.

--
Horms