From: Dave Taht dave.taht@gmail.com
commit 65cab850f0eeaa9180bd2e10a231964f33743edf upstream.
While most distributions long ago switched to the iproute2 suite of utilities, which allow class-e (240.0.0.0/4) address assignment, distributions relying on busybox, toybox and other forms of ifconfig cannot assign class-e addresses without this kernel patch.
While CIDR has been obsolete for 2 decades, and a survey of all the open source code in the world shows the IN_whatever macros are also obsolete... rather than obsolete CIDR from this ioctl entirely, this patch merely enables class-e assignment, sanely.
Signed-off-by: Dave Taht dave.taht@gmail.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Linus Walleij linus.walleij@linaro.org --- ChangeLog v1->v2: - Add SoB-line
- This commit is upstream in v4.20 - This should ve applied for stable v4.19.y, v4.14.y and v4.9.y --- include/uapi/linux/in.h | 10 +++++++--- net/ipv4/devinet.c | 5 +++-- net/ipv4/ipconfig.c | 2 ++ 3 files changed, 12 insertions(+), 5 deletions(-)
diff --git a/include/uapi/linux/in.h b/include/uapi/linux/in.h index 48e8a225b985..f6052e70bf40 100644 --- a/include/uapi/linux/in.h +++ b/include/uapi/linux/in.h @@ -266,10 +266,14 @@ struct sockaddr_in {
#define IN_CLASSD(a) ((((long int) (a)) & 0xf0000000) == 0xe0000000) #define IN_MULTICAST(a) IN_CLASSD(a) -#define IN_MULTICAST_NET 0xF0000000 +#define IN_MULTICAST_NET 0xe0000000
-#define IN_EXPERIMENTAL(a) ((((long int) (a)) & 0xf0000000) == 0xf0000000) -#define IN_BADCLASS(a) IN_EXPERIMENTAL((a)) +#define IN_BADCLASS(a) ((((long int) (a) ) == 0xffffffff) +#define IN_EXPERIMENTAL(a) IN_BADCLASS((a)) + +#define IN_CLASSE(a) ((((long int) (a)) & 0xf0000000) == 0xf0000000) +#define IN_CLASSE_NET 0xffffffff +#define IN_CLASSE_NSHIFT 0
/* Address to accept any incoming messages. */ #define INADDR_ANY ((unsigned long int) 0x00000000) diff --git a/net/ipv4/devinet.c b/net/ipv4/devinet.c index ea4bd8a52422..e38042933a27 100644 --- a/net/ipv4/devinet.c +++ b/net/ipv4/devinet.c @@ -941,17 +941,18 @@ static int inet_abc_len(__be32 addr) { int rc = -1; /* Something else, probably a multicast. */
- if (ipv4_is_zeronet(addr)) + if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr)) rc = 0; else { __u32 haddr = ntohl(addr); - if (IN_CLASSA(haddr)) rc = 8; else if (IN_CLASSB(haddr)) rc = 16; else if (IN_CLASSC(haddr)) rc = 24; + else if (IN_CLASSE(haddr)) + rc = 32; }
return rc; diff --git a/net/ipv4/ipconfig.c b/net/ipv4/ipconfig.c index 88212615bf4c..2393e5c106bf 100644 --- a/net/ipv4/ipconfig.c +++ b/net/ipv4/ipconfig.c @@ -429,6 +429,8 @@ static int __init ic_defaults(void) ic_netmask = htonl(IN_CLASSB_NET); else if (IN_CLASSC(ntohl(ic_myaddr))) ic_netmask = htonl(IN_CLASSC_NET); + else if (IN_CLASSE(ntohl(ic_myaddr))) + ic_netmask = htonl(IN_CLASSE_NET); else { pr_err("IP-Config: Unable to guess netmask for address %pI4\n", &ic_myaddr);
On Thu, Feb 14, 2019 at 02:16:04PM +0100, Linus Walleij wrote:
From: Dave Taht dave.taht@gmail.com
commit 65cab850f0eeaa9180bd2e10a231964f33743edf upstream.
While most distributions long ago switched to the iproute2 suite of utilities, which allow class-e (240.0.0.0/4) address assignment, distributions relying on busybox, toybox and other forms of ifconfig cannot assign class-e addresses without this kernel patch.
While CIDR has been obsolete for 2 decades, and a survey of all the open source code in the world shows the IN_whatever macros are also obsolete... rather than obsolete CIDR from this ioctl entirely, this patch merely enables class-e assignment, sanely.
Signed-off-by: Dave Taht dave.taht@gmail.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Linus Walleij linus.walleij@linaro.org
Is this really a fix? This isn't something that ever worked.
Either way, David Miller will need to sign off on this since he manages net/ -stable patches.
-- Thanks, Sasha
On Sun, Feb 17, 2019 at 7:22 PM Sasha Levin sashal@kernel.org wrote:
On Thu, Feb 14, 2019 at 02:16:04PM +0100, Linus Walleij wrote:
From: Dave Taht dave.taht@gmail.com
commit 65cab850f0eeaa9180bd2e10a231964f33743edf upstream.
While most distributions long ago switched to the iproute2 suite of utilities, which allow class-e (240.0.0.0/4) address assignment, distributions relying on busybox, toybox and other forms of ifconfig cannot assign class-e addresses without this kernel patch.
While CIDR has been obsolete for 2 decades, and a survey of all the open source code in the world shows the IN_whatever macros are also obsolete... rather than obsolete CIDR from this ioctl entirely, this patch merely enables class-e assignment, sanely.
Signed-off-by: Dave Taht dave.taht@gmail.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Linus Walleij linus.walleij@linaro.org
Is this really a fix? This isn't something that ever worked.
I guess it fits the bill of "a real bug that bothers people", what OpenWrt wants is to use their simple userspace to set up class-e addresses and without this patch, this doesn't work for them, and IIUC it could be expected to work.
I guess it is an ontological question whether OpenWrt are "fixning" or "stabilizing" or "adding features" by making class-e networks work with good old ifconfig. The maintainer(s) will decide.
Either way, David Miller will need to sign off on this since he manages net/ -stable patches.
I was unaware of different route points for stable patches, but it makes sense.
Maybe we should add some kind of tagging entries to MAINTAINERS so it is clear where to route stable material? Right now I guess it is another one of these undocumented rules that one is supposed to pick up by first annoying everyone :D
Are there other subsystems that have a second-level maintainer for stable, so I know before annoying someone else?
Yours, Linus Walleij
On Sun, Feb 17, 2019 at 10:18:29PM +0100, Linus Walleij wrote:
On Sun, Feb 17, 2019 at 7:22 PM Sasha Levin sashal@kernel.org wrote:
Either way, David Miller will need to sign off on this since he manages net/ -stable patches.
I was unaware of different route points for stable patches, but it makes sense.
Maybe we should add some kind of tagging entries to MAINTAINERS so it is clear where to route stable material? Right now I guess it is another one of these undocumented rules that one is supposed to pick up by first annoying everyone :D
Are there other subsystems that have a second-level maintainer for stable, so I know before annoying someone else?
I really think that this is the only exception to the procees, and given that David is only looking at the last 2 -stable releases (that would be 4.20 and 4.19 right now) it gets a bit tricky for folks who are not intimate with the process. So no, you're not annoying anyone :)
FWIW, the process is documented here: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Docu...
-- Thanks, Sasha
From: Sasha Levin sashal@kernel.org Date: Sun, 17 Feb 2019 13:22:41 -0500
On Thu, Feb 14, 2019 at 02:16:04PM +0100, Linus Walleij wrote:
From: Dave Taht dave.taht@gmail.com
commit 65cab850f0eeaa9180bd2e10a231964f33743edf upstream.
While most distributions long ago switched to the iproute2 suite of utilities, which allow class-e (240.0.0.0/4) address assignment, distributions relying on busybox, toybox and other forms of ifconfig cannot assign class-e addresses without this kernel patch.
While CIDR has been obsolete for 2 decades, and a survey of all the open source code in the world shows the IN_whatever macros are also obsolete... rather than obsolete CIDR from this ioctl entirely, this patch merely enables class-e assignment, sanely.
Signed-off-by: Dave Taht dave.taht@gmail.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Linus Walleij linus.walleij@linaro.org
Is this really a fix? This isn't something that ever worked.
Either way, David Miller will need to sign off on this since he manages net/ -stable patches.
I don't think this is -stable material at all.
linux-stable-mirror@lists.linaro.org
-
David Miller -
Linus Walleij -
Sasha Levin