Skip to contentSkip to conversation

if_bridge.h: include in6.h for struct in6_addr use

PATCH Thomas Backlund 12 participants 45 messages Jan 13, 2013 – Mar 13, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

patch both inline and attached as thunderbird tends to mess up ...
-----

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

--

include/uapi/linux/if_bridge.h 2013-01-13 20:15:04.153676151 +0200 renamed from include/uapi/linux/if_bridge.h 2013-01-13 20:09:54.257271755 +0200 +1 -0
diff -Nurp linux-3.8-rc3/include/uapi/linux/if_bridge.h linux-3.8-rc3.fix/include/uapi/linux/if_bridge.h
@@ -14,6 +14,7 @@
 #define _UAPI_LINUX_IF_BRIDGE_H
 
 #include <linux/types.h>
+#include <linux/in6.h>
 
 #define SYSFS_BRIDGE_ATTR      "bridge"
 #define SYSFS_BRIDGE_FDB       "brforward"

-----
Thomas

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

--

include/uapi/linux/if_bridge.h +1 -0
diff -Nurp linux-3.8-rc3/include/uapi/linux/if_bridge.h linux-3.8-rc3.fix/include/uapi/linux/if_bridge.h
@@ -14,6 +14,7 @@
 #define _UAPI_LINUX_IF_BRIDGE_H
 
 #include <linux/types.h>
+#include <linux/in6.h>
 
 #define SYSFS_BRIDGE_ATTR	"bridge"
 #define SYSFS_BRIDGE_FDB	"brforward"
#1 lore raw

Thomas Backlund skrev 13.1.2013 20:38:

Quoted 9 lines

patch both inline and attached as thunderbird tends to mess up ...

-----

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

Ok,
ignore this patch, it's not the correct fix as it introduces redefinitions...

Btw, the error that I hit that made me suggest this fix was libvirt config check bailing out:

config.log:/usr/include/linux/if_bridge.h:173:20: error: field 'ip6' has incomplete type

Quoted 23 lines

--

diff -Nurp linux-3.8-rc3/include/uapi/linux/if_bridge.h linux-3.8-rc3.fix/include/uapi/linux/if_bridge.h
--- linux-3.8-rc3/include/uapi/linux/if_bridge.h 2013-01-13
20:09:54.257271755 +0200
+++ linux-3.8-rc3.fix/include/uapi/linux/if_bridge.h 2013-01-13
20:15:04.153676151 +0200

include/uapi/linux/if_bridge.h 2013-01-13 20:15:04.153676151 +0200 renamed from include/uapi/linux/if_bridge.h 2013-01-13 20:09:54.257271755 +0200 +1 -1
@@ -14,6 +14,7 @@
  #define _UAPI_LINUX_IF_BRIDGE_H
 
  #include <linux/types.h>
+#include <linux/in6.h>
 
  #define SYSFS_BRIDGE_ATTR      "bridge"
  #define SYSFS_BRIDGE_FDB       "brforward"
 
 
-----
Quoted 2 lines

Thomas

--
Thomas

#2 lore raw
Quoted 15 lines · Thomas Backlund

Thomas Backlund skrev 13.1.2013 20:38:

Quoted 9 lines

patch both inline and attached as thunderbird tends to mess up ...

-----

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

Ok,
ignore this patch, it's not the correct fix as it introduces redefinitions...

Quoted 5 lines

Btw, the error that I hit that made me suggest this fix was libvirt config check bailing out:

config.log:/usr/include/linux/if_bridge.h:173:20: error: field 'ip6' has incomplete type

Hmm, just now noticing this thread, after independently hitting and and having to work around the same problem in libvirt:
https://www.redhat.com/archives/libvir-list/2013-January/msg00930.html
https://bugzilla.redhat.com/show_bug.cgi?id=895141

Quoted 10 lines
Quoted 10 lines

--- linux-3.8-rc3/include/uapi/linux/if_bridge.h 2013-01-13
20:09:54.257271755 +0200
+++ linux-3.8-rc3.fix/include/uapi/linux/if_bridge.h 2013-01-13
20:15:04.153676151 +0200

include/uapi/linux/if_bridge.h 2013-01-13 20:15:04.153676151 +0200 renamed from include/uapi/linux/if_bridge.h 2013-01-13 20:09:54.257271755 +0200 +1 -0
@@ -14,6 +14,7 @@
  #define _UAPI_LINUX_IF_BRIDGE_H
 
  #include <linux/types.h>
+#include <linux/in6.h>
 
Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
author Acked-by Thomas Graf
#3 lore raw

the patch "bridge: export multicast database via netlink" broke kernel 3.8 uapi (was: Re: [libvirt] if_bridge.h: include in6.h for struct in6_addr use)

Eric Blake skrev 15.1.2013 01:57:

Quoted 19 lines · Thomas Backlund
Quoted 18 lines

Thomas Backlund skrev 13.1.2013 20:38:

Quoted 9 lines

patch both inline and attached as thunderbird tends to mess up ...

-----

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

Ok,
ignore this patch, it's not the correct fix as it introduces redefinitions...

Btw, the error that I hit that made me suggest this fix was libvirt config check bailing out:

Quoted 7 lines
Quoted 2 lines

config.log:/usr/include/linux/if_bridge.h:173:20: error: field 'ip6' has incomplete type

Hmm, just now noticing this thread, after independently hitting and and having to work around the same problem in libvirt:
https://www.redhat.com/archives/libvir-list/2013-January/msg00930.html
https://bugzilla.redhat.com/show_bug.cgi?id=895141

Yep,

and the commit breaking uapi headers is by using "struct in6_addr ip6" is:

From ee07c6e7a6f8a25c18f0a6b18152fbd7499245f6 Mon Sep 17 00:00:00 2001
From: Cong Wang <[email protected]>
Date: Fri, 7 Dec 2012 00:04:48 +0000
Subject: [PATCH] bridge: export multicast database via netlink

V5: fix two bugs pointed out by Thomas
remove seq check for now, mark it as TODO

V4: remove some useless #include
some coding style fix

V3: drop debugging printk's
update selinux perm table as well

V2: drop patch 1/2, export ifindex directly
Redesign netlink attributes
Improve netlink seq check
Handle IPv6 addr as well

This patch exports bridge multicast database via netlink message type RTM_GETMDB. Similar to fdb, but currently bridge-specific. We may need to support modify multicast database too (RTM_{ADD,DEL}MDB).

(Thanks to Thomas for patient reviews)

Cc Herbert Xu <[email protected]>
Cc Stephen Hemminger <[email protected]>
Cc "David S. Miller" <[email protected]>
Cc Thomas Graf <[email protected]>
Cc Jesper Dangaard Brouer <[email protected]>
---
 include/uapi/linux/if_bridge.h |   55 ++++++++++++++
 include/uapi/linux/rtnetlink.h |    3 +
 net/bridge/Makefile            |    2 +-
 net/bridge/br_mdb.c            |  163 
++++++++++++++++++++++++++++++++++++++++
 net/bridge/br_multicast.c      |    1 +
 net/bridge/br_private.h        |    1 +
 security/selinux/nlmsgtab.c    |    1 +
 7 files changed, 225 insertions(+), 1 deletions(-)
 create mode 100644 net/bridge/br_mdb.c

--
Thomas

#4 lore raw

Re: the patch "bridge: export multicast database via netlink" broke kernel 3.8 uapi (was: Re: [libvirt] if_bridge.h: include in6.h for struct in6_addr use)

Quoted 31 lines · Thomas Backlund

Eric Blake skrev 15.1.2013 01:57:

Quoted 19 lines · Thomas Backlund
Quoted 18 lines

Thomas Backlund skrev 13.1.2013 20:38:

Quoted 9 lines

patch both inline and attached as thunderbird tends to mess up ...

-----

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

Ok,
ignore this patch, it's not the correct fix as it introduces redefinitions...

Btw, the error that I hit that made me suggest this fix was libvirt config check bailing out:

Quoted 7 lines
Quoted 2 lines

config.log:/usr/include/linux/if_bridge.h:173:20: error: field 'ip6' has incomplete type

Hmm, just now noticing this thread, after independently hitting and and having to work around the same problem in libvirt:
https://www.redhat.com/archives/libvir-list/2013-January/msg00930.html
https://bugzilla.redhat.com/show_bug.cgi?id=895141

Yep,

Quoted 7 lines

and the commit breaking uapi headers is by using "struct in6_addr ip6" is:

From ee07c6e7a6f8a25c18f0a6b18152fbd7499245f6 Mon Sep 17 00:00:00 2001
From: Cong Wang <[email protected]>
Date: Fri, 7 Dec 2012 00:04:48 +0000
Subject: [PATCH] bridge: export multicast database via netlink

Does the following patch help?

$ git diff include/uapi/linux/if_bridge.h
(quoted diff) +0 -0

b/include/uapi/linux/if_bridge.h
index 5db2975..653db23 100644

include/uapi/linux/if_bridge.h +1 -0
@@ -14,6 +14,7 @@
 #define _UAPI_LINUX_IF_BRIDGE_H
 
 #include <linux/types.h>
+#include <linux/in6.h>
 
 #define SYSFS_BRIDGE_ATTR      "bridge"
 #define SYSFS_BRIDGE_FDB       "brforward"
#5 lore raw

Re: the patch "bridge: export multicast database via netlink" broke kernel 3.8 uapi

Cong Wang skrev 15.1.2013 12:11:

Quoted 50 lines · Thomas Backlund
Quoted 31 lines

Eric Blake skrev 15.1.2013 01:57:

Quoted 19 lines · Thomas Backlund
Quoted 18 lines

Thomas Backlund skrev 13.1.2013 20:38:

Quoted 9 lines

patch both inline and attached as thunderbird tends to mess up ...

-----

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

Ok,
ignore this patch, it's not the correct fix as it introduces redefinitions...

Btw, the error that I hit that made me suggest this fix was libvirt config check bailing out:

Quoted 7 lines
Quoted 2 lines

config.log:/usr/include/linux/if_bridge.h:173:20: error: field 'ip6' has incomplete type

Hmm, just now noticing this thread, after independently hitting and and having to work around the same problem in libvirt:
https://www.redhat.com/archives/libvir-list/2013-January/msg00930.html
https://bugzilla.redhat.com/show_bug.cgi?id=895141

Yep,

Quoted 7 lines

and the commit breaking uapi headers is by using "struct in6_addr ip6" is:

From ee07c6e7a6f8a25c18f0a6b18152fbd7499245f6 Mon Sep 17 00:00:00 2001
From: Cong Wang <[email protected]>
Date: Fri, 7 Dec 2012 00:04:48 +0000
Subject: [PATCH] bridge: export multicast database via netlink

Does the following patch help?

$ git diff include/uapi/linux/if_bridge.h
include/uapi/linux/if_bridge.h +0 -0
index 5db2975..653db23 100644

include/uapi/linux/if_bridge.h line 20

Quoted 7 lines
include/uapi/linux/if_bridge.h +1 -0
@@ -14,6 +14,7 @@
  #define _UAPI_LINUX_IF_BRIDGE_H
 
  #include <linux/types.h>
+#include <linux/in6.h>
 
  #define SYSFS_BRIDGE_ATTR      "bridge"
  #define SYSFS_BRIDGE_FDB       "brforward"

Well, I suggested the same fix in the beginning of the thread on netdev and lkml: "if_bridge.h: include in6.h for struct in6_addr use"

as it seemed to fix the libvirt case

but then asked it to be ignored after I tried to build connman, and hit this conflict with glibc-2.17:

In file included from /usr/include/arpa/inet.h:22:0,

                 from ./include/connman/inet.h:25,
                 from src/connman.h:128,
                 from src/tethering.c:40:

/usr/include/netinet/in.h:35:5: error: expected identifier before numeric constant /usr/include/netinet/in.h:197:8: error: redefinition of 'struct in6_addr' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:30:8: note: originally defined here
In file included from /usr/include/arpa/inet.h:22:0,

                 from ./include/connman/inet.h:25,
                 from src/connman.h:128,
                 from src/tethering.c:40:

/usr/include/netinet/in.h:238:8: error: redefinition of 'struct sockaddr_in6' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:46:8: note: originally defined here
In file included from /usr/include/arpa/inet.h:22:0,

                 from ./include/connman/inet.h:25,
                 from src/connman.h:128,
                 from src/tethering.c:40:

/usr/include/netinet/in.h:274:8: error: redefinition of 'struct ipv6_mreq' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:54:8: note: originally defined here make[1]: *** [src/src_connmand-tethering.o] Error 1

So I'm not sure it's the right one...

--
Thomas

#6 lore raw

Re: the patch "bridge: export multicast database via netlink" broke kernel 3.8 uapi

Quoted 29 lines · Thomas Backlund

as it seemed to fix the libvirt case

but then asked it to be ignored after I tried to build connman, and hit this conflict with glibc-2.17:

In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:35:5: error: expected identifier before numeric constant
/usr/include/netinet/in.h:197:8: error: redefinition of 'struct in6_addr' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:30:8: note: originally defined here In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:238:8: error: redefinition of 'struct sockaddr_in6'
In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:46:8: note: originally defined here In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:
Quoted 8 lines

/usr/include/netinet/in.h:274:8: error: redefinition of 'struct ipv6_mreq' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:54:8: note: originally defined here make[1]: *** [src/src_connmand-tethering.o] Error 1

So I'm not sure it's the right one...

% rpm -q --whatprovides /usr/include/netinet/in.h
glibc-headers-2.14.90-24.fc16.9.x86_64

Seems we have conflicts with glibc headers...

#7 lore raw

Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Thomas Backlund

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

So, how can we handle this? /me is wondering why we didn't see this before.

Thanks.

Quoted 60 lines · Thomas Backlund

Cong Wang skrev 15.1.2013 12:11:

Quoted 11 lines

Does the following patch help?

$ git diff include/uapi/linux/if_bridge.h
include/uapi/linux/if_bridge.h +0 -0
index 5db2975..653db23 100644

include/uapi/linux/if_bridge.h line 20

Quoted 7 lines
include/uapi/linux/if_bridge.h +1 -0
@@ -14,6 +14,7 @@
  #define _UAPI_LINUX_IF_BRIDGE_H
 
  #include <linux/types.h>
+#include <linux/in6.h>
 
  #define SYSFS_BRIDGE_ATTR      "bridge"
  #define SYSFS_BRIDGE_FDB       "brforward"

Well, I suggested the same fix in the beginning of the thread on netdev and lkml: "if_bridge.h: include in6.h for struct in6_addr use"

as it seemed to fix the libvirt case

but then asked it to be ignored after I tried to build connman, and hit this conflict with glibc-2.17:

In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:35:5: error: expected identifier before numeric constant
/usr/include/netinet/in.h:197:8: error: redefinition of 'struct in6_addr' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:30:8: note: originally defined here In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:238:8: error: redefinition of 'struct sockaddr_in6'
In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:46:8: note: originally defined here In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:274:8: error: redefinition of 'struct ipv6_mreq' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:54:8: note: originally defined here make[1]: *** [src/src_connmand-tethering.o] Error 1

So I'm not sure it's the right one...

#8 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 8 lines · Cong Wang

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

Quoted 65 lines

So, how can we handle this? /me is wondering why we didn't see this before.

Thanks.

Quoted 59 lines · Thomas Backlund

Cong Wang skrev 15.1.2013 12:11:

Quoted 11 lines

Does the following patch help?

$ git diff include/uapi/linux/if_bridge.h
include/uapi/linux/if_bridge.h +0 -0
index 5db2975..653db23 100644

include/uapi/linux/if_bridge.h line 20

Quoted 7 lines
include/uapi/linux/if_bridge.h +1 -0
@@ -14,6 +14,7 @@
  #define _UAPI_LINUX_IF_BRIDGE_H
 
  #include <linux/types.h>
+#include <linux/in6.h>
 
  #define SYSFS_BRIDGE_ATTR      "bridge"
  #define SYSFS_BRIDGE_FDB       "brforward"

Well, I suggested the same fix in the beginning of the thread on netdev and lkml: "if_bridge.h: include in6.h for struct in6_addr use"

as it seemed to fix the libvirt case

but then asked it to be ignored after I tried to build connman, and hit this conflict with glibc-2.17:

In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:35:5: error: expected identifier before numeric constant
/usr/include/netinet/in.h:197:8: error: redefinition of 'struct in6_addr' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:30:8: note: originally defined here In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:238:8: error: redefinition of 'struct sockaddr_in6'
In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:46:8: note: originally defined here In file included from /usr/include/arpa/inet.h:22:0,

                  from ./include/connman/inet.h:25,
                  from src/connman.h:128,
                  from src/tethering.c:40:

/usr/include/netinet/in.h:274:8: error: redefinition of 'struct ipv6_mreq' In file included from /usr/include/linux/if_bridge.h:17:0,
from src/tethering.c:38:
/usr/include/linux/in6.h:54:8: note: originally defined here make[1]: *** [src/src_connmand-tethering.o] Error 1

So I'm not sure it's the right one...

This is not a new issue. In addition to this,
netinet/in.h also conflits with linux/in.h.

We might have
#if !defined(__GLIBC__) || !defined(_NETINET_IN_H) :
#endif
around those conflicting definitions in uapi/linux/in{,6}.h.

--yoshfuji

#9 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 11 lines · YOSHIFUJI Hideaki
Quoted 10 lines · Cong Wang

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

So, how can we handle this? /me is wondering why we didn't see this before.

[...]

Quoted 8 lines

This is not a new issue. In addition to this,
netinet/in.h also conflits with linux/in.h.

We might have
#if !defined(__GLIBC__) || !defined(_NETINET_IN_H) :
#endif
around those conflicting definitions in uapi/linux/in{,6}.h.

This only solves half the problem, as <netinet/in.h> might be included after <linux/in.h>. Also, not all Linux userland uses glibc.

Ben.

Signature · 3 lines
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
#10 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 20 lines · On Wednesday 16 January 2013 10:47:12 Ben Hutchings wrote:
Quoted 11 lines · YOSHIFUJI Hideaki
Quoted 10 lines · Cong Wang

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

So, how can we handle this? /me is wondering why we didn't see this before.

[...]

Quoted 5 lines

This is not a new issue. In addition to this,
netinet/in.h also conflits with linux/in.h.

We might have

Quoted 8 lines
Quoted 5 lines

#if !defined(__GLIBC__) || !defined(_NETINET_IN_H)

#endif

around those conflicting definitions in uapi/linux/in{,6}.h.

This only solves half the problem, as <netinet/in.h> might be included after <linux/in.h>. Also, not all Linux userland uses glibc.

certainly true, but the current expectation is that you don't mix your ABIs. if you're programming with the C library API, then use the C library headers. if you're banging directly on the kernel, then use the kernel headers. not saying it's a perfect solution, but it works for the vast majority of use cases.
-mike

#11 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 23 lines · Mike Frysinger
Quoted 22 lines · On Wednesday 16 January 2013 10:47:12 Ben Hutchings wrote:
Quoted 11 lines · YOSHIFUJI Hideaki
Quoted 10 lines · Cong Wang

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

So, how can we handle this? /me is wondering why we didn't see this before.

[...]

Quoted 7 lines

This is not a new issue. In addition to this,
netinet/in.h also conflits with linux/in.h.

We might have

#if !defined(__GLIBC__) || !defined(_NETINET_IN_H)

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

#endif

around those conflicting definitions in uapi/linux/in{,6}.h.

This only solves half the problem, as <netinet/in.h> might be included after <linux/in.h>. Also, not all Linux userland uses glibc.

certainly true, but the current expectation is that you don't mix your ABIs.

Whose expectation? Which ABIs are being mixed?

Quoted 4 lines

if you're programming with the C library API, then use the C library headers. if you're banging directly on the kernel, then use the kernel headers. not saying it's a perfect solution, but it works for the vast majority of use cases.

In practice most C programs for Linux will use a mixture of thinly wrapped system calls and higher-level APIs from the C library, and never really call the kernel directly (as that requires inline assembler). Userland programmers will work around this historical mess by tweaking the #include order or splitting source files. But they shouldn't have to.

Ben.

Signature · 3 lines
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
#12 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 28 lines · On Wednesday 16 January 2013 12:10:11 Ben Hutchings wrote:
Quoted 27 lines · Mike Frysinger
Quoted 26 lines · On Wednesday 16 January 2013 10:47:12 Ben Hutchings wrote:
Quoted 11 lines · YOSHIFUJI Hideaki
Quoted 10 lines · Cong Wang

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

So, how can we handle this? /me is wondering why we didn't see this before.

[...]

Quoted 10 lines

This is not a new issue. In addition to this,
netinet/in.h also conflits with linux/in.h.

We might have

#if !defined(__GLIBC__) || !defined(_NETINET_IN_H)

#endif

around those conflicting definitions in uapi/linux/in{,6}.h.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

This only solves half the problem, as <netinet/in.h> might be included after <linux/in.h>. Also, not all Linux userland uses glibc.

certainly true, but the current expectation is that you don't mix your ABIs.

Whose expectation? Which ABIs are being mixed?

the kernel's view of the world and the C library's view of the world. there is a lot of overlap, but it's the C library's job to provide a POSIX compliant interface (and then some).

obvious examples:
- the stat structures are not the same and require translation
- ptrace structures are not the same and you really need to use the C lib one
- the readdir function is completely different

but to this specific question, if you're calling the kernel's network functions directly, then you need to include the kernel's headers for its definition of how things are passed. if you're calling the C library's network functions, then use the C library's headers.

Quoted 11 lines
Quoted 4 lines

if you're programming with the C library API, then use the C library headers. if you're banging directly on the kernel, then use the kernel headers. not saying it's a perfect solution, but it works for the vast majority of use cases.

In practice most C programs for Linux will use a mixture of thinly wrapped system calls and higher-level APIs from the C library, and never really call the kernel directly (as that requires inline assembler). Userland programmers will work around this historical mess by tweaking the #include order or splitting source files. But they shouldn't have to.

if you're not calling the kernel directly, why are you including the kernel headers ? what is the problem people are actually trying to address here (and no, "i want to include both headers" is not the answer) ?
-mike

maintainer
#20 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Mike Frysinger

From: Mike Frysinger <[email protected]>
Date: Wed, 16 Jan 2013 12:04:56 -0500

Quoted 5 lines

certainly true, but the current expectation is that you don't mix your ABIs. if you're programming with the C library API, then use the C library headers. if you're banging directly on the kernel, then use the kernel headers. not saying it's a perfect solution, but it works for the vast majority of use cases.

This isn't how real life works.

GLIBC itself brings in some of the kernel headers, as do various library headers for libraries other than glibc.

So you can get these conflicting headers included indirectly, and it is of no fault of any of the various parties involved.

We have to make them work when included at the same time somehow, and this is totally unavoidable.

maintainer
#13 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Mike Frysinger

From: Mike Frysinger <[email protected]>
Date: Wed, 16 Jan 2013 12:28:39 -0500

Quoted 3 lines

if you're not calling the kernel directly, why are you including the kernel headers ? what is the problem people are actually trying to address here (and no, "i want to include both headers" is not the answer) ?

When GLIBC doesn't provide it's own definition of some networking macros or interfaces that the kernel provides, people include the kernel header.

This has been done for decades, wake up.

#14 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 9 lines · On Wednesday 16 January 2013 13:59:59 David Miller wrote:

From: Mike Frysinger <[email protected]>

Quoted 4 lines

if you're not calling the kernel directly, why are you including the kernel headers ? what is the problem people are actually trying to address here (and no, "i want to include both headers" is not the answer) ?

When GLIBC doesn't provide it's own definition of some networking macros or interfaces that the kernel provides, people include the kernel header.

sounds like glibc's headers are out of date and we should update them to include the missing definitions

but this is still too vague. what headers/definitions do people want to see simultaneously included ? changes would be needed on both sides (kernel & C library).

Quoted 1 line

This has been done for decades, wake up.

and it's been broken for just as long. no need to be a dick. -mike

maintainer
#15 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

From: Mike Frysinger <[email protected]>
Date: Wed, 16 Jan 2013 14:22:16 -0500

Quoted 4 lines · On Wednesday 16 January 2013 13:59:59 David Miller wrote:
Quoted 1 line

This has been done for decades, wake up.

and it's been broken for just as long. no need to be a dick.

By being ignorant and having such a simplistic view of the situation, you're the one who's actually the dick.

#21 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to David Miller
Quoted 7 lines · On Wednesday 16 January 2013 13:57:44 David Miller wrote:

From: Mike Frysinger <[email protected]>

Quoted 5 lines

certainly true, but the current expectation is that you don't mix your ABIs. if you're programming with the C library API, then use the C library headers. if you're banging directly on the kernel, then use the kernel headers. not saying it's a perfect solution, but it works for the vast majority of use cases.

Quoted 7 lines

This isn't how real life works.

GLIBC itself brings in some of the kernel headers, as do various library headers for libraries other than glibc.

So you can get these conflicting headers included indirectly, and it is of no fault of any of the various parties involved.

the headers glibc includes tend to be pretty stand alone specifically so that it doesn't matter

Quoted 2 lines

We have to make them work when included at the same time somehow, and this is totally unavoidable.

"them" is vague. saying that every kernel header has to be usable in the same compilation unit as every C library header regardless of order is unrealistic (at least it is today). there are cases where they define the same structure different because the structure as the C library expects is different from what the kernel syscall expects. you could avoid that on the kernel side by giving them all prefixes (like __kernel_), but that didn't seem entirely palpable to the kernel folks. i couldn't even get them to remove crap that breaks non- glibc C libraries (e.g. uapi/linux/stat.h -- looks like someone inadvertently fixed uapi/linux/socket.h finally).

for many networking headers, the C library will provide enums & defines while the kernel only provides enums. including the kernel after the C library one leads to parsing errors as the defines expand in the enum and kill it. like linux/in.h and netinet/in.h and IPPROTO_*.
-mike

maintainer
#36 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Ben Hutchings

From: Ben Hutchings <[email protected]>
Date: Wed, 16 Jan 2013 15:47:12 +0000

Quoted 16 lines · YOSHIFUJI Hideaki
Quoted 11 lines
Quoted 10 lines · Cong Wang

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

So, how can we handle this? /me is wondering why we didn't see this before.

[...]

Quoted 3 lines

This is not a new issue. In addition to this,
netinet/in.h also conflits with linux/in.h.

Quoted 8 lines
Quoted 5 lines

We might have
#if !defined(__GLIBC__) || !defined(_NETINET_IN_H) :
#endif
around those conflicting definitions in uapi/linux/in{,6}.h.

This only solves half the problem, as <netinet/in.h> might be included after <linux/in.h>. Also, not all Linux userland uses glibc.

So I've been looking at reasonable ways to fix this.

What would be really nice is if GCC treated multiple identical definitions of structures the same way it handles multiple identical definitions of CPP defines. Which is to silently accept them.

But that's not the case, so we need a different solution.

Another thing to do is to use a new structure for in6_addr in kernel headers exported to userland.

If we were to make the structure members and accessor macros identical we could avoid breaking most if not all apps.

However the type name is different so:

	struct in6_addr *p = &kern_struct->member;

wouldn't work so well. And you couldn't fix up the sources to these kinds of accesses in a way that work cleanly both before and after changing the kernel headers.

I'm out of ideas for today.

#37 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 44 lines · David Miller

From: Ben Hutchings <[email protected]>
Date: Wed, 16 Jan 2013 15:47:12 +0000

Quoted 16 lines · YOSHIFUJI Hideaki
Quoted 11 lines
Quoted 10 lines · Cong Wang

(Cc'ing some glibc developers...)

Hello,

In glibc source file inet/netinet/in.h and kernel source file include/uapi/linux/in6.h, both define struct in6_addr, and both are visible to user applications. Thomas reported a conflict below.

So, how can we handle this? /me is wondering why we didn't see this before.

[...]

Quoted 3 lines

This is not a new issue. In addition to this,
netinet/in.h also conflits with linux/in.h.

Quoted 8 lines
Quoted 5 lines

We might have
#if !defined(__GLIBC__) || !defined(_NETINET_IN_H) :
#endif
around those conflicting definitions in uapi/linux/in{,6}.h.

This only solves half the problem, as <netinet/in.h> might be included after <linux/in.h>. Also, not all Linux userland uses glibc.

So I've been looking at reasonable ways to fix this.

What would be really nice is if GCC treated multiple identical definitions of structures the same way it handles multiple identical definitions of CPP defines. Which is to silently accept them.

But that's not the case, so we need a different solution.

Another thing to do is to use a new structure for in6_addr in kernel headers exported to userland.

If we were to make the structure members and accessor macros identical we could avoid breaking most if not all apps.

However the type name is different so:

Quoted 7 lines
	struct in6_addr *p = &kern_struct->member;

wouldn't work so well. And you couldn't fix up the sources to these kinds of accesses in a way that work cleanly both before and after changing the kernel headers.

I'm out of ideas for today.

So I just went down the rabbit hole, and the further I get the closer I get to having two exact copies of the same definitions in both glibc and the kernel and using whichever one was included first.

Is anyone opposed to that kind of solution?

There is some ugliness there.

Cheers,
Carlos.

maintainer
#38 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 20:58:47 -0500

Quoted 6 lines

So I just went down the rabbit hole, and the further I get the closer I get to having two exact copies of the same definitions in both glibc and the kernel and using whichever one was included first.

Is anyone opposed to that kind of solution?

Sounds interesting, please share :-)

#22 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to David Miller
Quoted 11 lines · David Miller

From: Mike Frysinger <[email protected]>
Date: Wed, 16 Jan 2013 12:04:56 -0500

Quoted 5 lines

certainly true, but the current expectation is that you don't mix your ABIs. if you're programming with the C library API, then use the C library headers. if you're banging directly on the kernel, then use the kernel headers. not saying it's a perfect solution, but it works for the vast majority of use cases.

This isn't how real life works.

Quoted 9 lines

GLIBC itself brings in some of the kernel headers, as do various library headers for libraries other than glibc.

So you can get these conflicting headers included indirectly, and it is of no fault of any of the various parties involved.

We have to make them work when included at the same time somehow, and this is totally unavoidable.

Just to put some code behind the comments I made earlier.

Solution:
=========

- Synchronize linux's include/uapi/linux/in6.h' with glibc's inet/netinet/in.h'.
- Synchronize glibc's inet/netinet/in.h with linux's include/uapi/linux/in6.h'.
- Allow including the headers in either other.
- First header included defines the structures and macros.

Details:
========

The kernel promises not to break the UAPI ABI so I don't see why we can't just have the two userspace headers coordinate?

If you include the kernel headers first you get those, and if you include the glibc headers first you get those, and the following patch arranges a coordination and synchronization between the two.

Let's handle include/uapi/linux/in6.h' from linux, and inet/netinet/in.h' from glibc and ensure they compile in any order and preserve the required ABI.

These two patches pass the following compile tests:

cat >> test1.c <<EOF
#include <netinet/in.h>
#include <linux/in6.h>
int main (void) {
return 0;
}
EOF
gcc -c test1.c

cat >> test2.c <<EOF
#include <linux/in6.h>
#include <netinet/in.h>
int main (void) {
return 0;
}
EOF
gcc -c test2.c

One wrinkle is that the kernel has a different name for one of the members in ipv6_mreq. In the kernel patch we create a macro to cover the uses of the old name, and while that's not entirely clean it's one of the best solutions (aside from an anonymous union which has other issues).

I've reviewed the code and it looks to me like the ABI is assured and everything matches on both sides.

Comments?

Notes:
- You want netinet/in.h to include bits/in.h as early as possible, but it needs in_addr so define in_addr early.
- You want bits/in.h included as early as possible so you can use the linux specific code to define __USE_KERNEL_DEFS based on the _UAPI_* macro definition and use those to cull in.h.
- glibc was missing IPPROTO_MH, added here.

glibc/

2012-01-16 Carlos O'Donell <[email protected]>

* sysdeps/unix/sysv/linux/bits/in.h
[_UAPI_LINUX_IN6_H]: Define __USE_KERNEL_IPV6_DEFS.
* inet/netinet/in.h: Move in_addr definition and bits/in.h inclusion before __USE_KERNEL_IPV6_DEFS uses.
* inet/netinet/in.h [!__USE_KERNEL_IPV6_DEFS]: Define IPPROTO_MH, and IPPROTO_BEETPH.
[__USE_KERNEL_IPV6_DEFS]: Don't define any of IPPROTO_*, in6_addr, sockaddr_in6, or ipv6_mreq.

inet/netinet/in.h +25 -19
index 89e3813..882739d 100644
@@ -26,6 +26,20 @@
 
 __BEGIN_DECLS
 
+/* Internet address.  */
+typedef uint32_t in_addr_t;
+struct in_addr
+  {
+    in_addr_t s_addr;
+  };
+
+/* Get system-specific definitions.  */
+#include <bits/in.h>
+
+/* If __USER_KERNEL_IPV6_DEFS is defined then the user has included the kernel
+   network headers first and we should use those ABI-identical definitions
+   instead of our own.  */
+#ifndef __USE_KERNEL_IPV6_DEFS
 /* Standard well-defined IP protocols.  */
 enum
   {
@@ -75,6 +89,8 @@ enum
 #define IPPROTO_DSTOPTS		IPPROTO_DSTOPTS
     IPPROTO_MTP = 92,	   /* Multicast Transport Protocol.  */
 #define IPPROTO_MTP		IPPROTO_MTP
+    IPPROTO_BEETPH = 94,   /* IP option pseudo header for BEET.  */
+#define IPPROTO_BEETPH		IPPROTO_BEETPH
     IPPROTO_ENCAP = 98,	   /* Encapsulation Header.  */
 #define IPPROTO_ENCAP		IPPROTO_ENCAP
     IPPROTO_PIM = 103,	   /* Protocol Independent Multicast.  */
@@ -83,13 +99,15 @@ enum
 #define IPPROTO_COMP		IPPROTO_COMP
     IPPROTO_SCTP = 132,	   /* Stream Control Transmission Protocol.  */
 #define IPPROTO_SCTP		IPPROTO_SCTP
+    IPPROTO_MH = 135,      /* IPv6 mobility header.  */
+#define IPPROTO_MH		IPPROTO_MH
     IPPROTO_UDPLITE = 136, /* UDP-Lite protocol.  */
 #define IPPROTO_UDPLITE		IPPROTO_UDPLITE
     IPPROTO_RAW = 255,	   /* Raw IP packets.  */
 #define IPPROTO_RAW		IPPROTO_RAW
     IPPROTO_MAX
   };
-
+#endif /* !__USE_KERNEL_IPV6_DEFS */
 
 /* Type to represent a port.  */
 typedef uint16_t in_port_t;
@@ -134,15 +152,6 @@ enum
     IPPORT_USERRESERVED = 5000
   };
 
-
-/* Internet address.  */
-typedef uint32_t in_addr_t;
-struct in_addr
-  {
-    in_addr_t s_addr;
-  };
-
-
 /* Definitions of the bits in an Internet address integer.
 
    On subnets, host and network parts are found according to
@@ -191,7 +200,7 @@ struct in_addr
 #define INADDR_ALLRTRS_GROUP    ((in_addr_t) 0xe0000002) /* 224.0.0.2 */
 #define INADDR_MAX_LOCAL_GROUP  ((in_addr_t) 0xe00000ff) /* 224.0.0.255 */
 
-
+#ifndef __USE_KERNEL_IPV6_DEFS
 /* IPv6 address */
 struct in6_addr
   {
@@ -209,6 +218,7 @@ struct in6_addr
 # define s6_addr32		__in6_u.__u6_addr32
 #endif
   };
+#endif /* !__USE_KERNEL_IPV6_DEFS */
 
 extern const struct in6_addr in6addr_any;        /* :: */
 extern const struct in6_addr in6addr_loopback;   /* ::1 */
@@ -218,7 +228,6 @@ extern const struct in6_addr in6addr_loopback;   /* ::1 */
 #define INET_ADDRSTRLEN 16
 #define INET6_ADDRSTRLEN 46
 
-
 /* Structure describing an Internet socket address.  */
 struct sockaddr_in
   {
@@ -233,6 +242,7 @@ struct sockaddr_in
 			   sizeof (struct in_addr)];
   };
 
+#ifndef __USE_KERNEL_IPV6_DEFS
 /* Ditto, for IPv6.  */
 struct sockaddr_in6
   {
@@ -242,7 +252,7 @@ struct sockaddr_in6
     struct in6_addr sin6_addr;	/* IPv6 address */
     uint32_t sin6_scope_id;	/* IPv6 scope-id */
   };
-
+#endif /* !__USE_KERNEL_IPV6_DEFS */
 
 #if defined __USE_MISC || defined __USE_GNU
 /* IPv4 multicast request.  */
@@ -268,7 +278,7 @@ struct ip_mreq_source
   };
 #endif
 
-
+#ifndef __USE_KERNEL_IPV6_DEFS
 /* Likewise, for IPv6.  */
 struct ipv6_mreq
   {
@@ -278,7 +288,7 @@ struct ipv6_mreq
     /* local interface */
     unsigned int ipv6mr_interface;
   };
-
+#endif /* !__USE_KERNEL_IPV6_DEFS */
 
 #if defined __USE_MISC || defined __USE_GNU
 /* Multicast group request.  */
@@ -349,10 +359,6 @@ struct group_filter
 				      * sizeof (struct sockaddr_storage)))
 #endif
 
-
-/* Get system-specific definitions.  */
-#include <bits/in.h>
-
 /* Functions to convert between host and network byte order.
 
    Please note that these functions normally take `unsigned long int' or
sysdeps/unix/sysv/linux/bits/in.h +9 -0
index e959b33..36bb72b 100644
@@ -21,6 +21,15 @@
 # error "Never use <bits/in.h> directly; include <netinet/in.h> instead."
 #endif
 
+/* If the application has already included linux/in6.h from a linux-based
+   kernel then we will not define IPPROTO_* defines, in6_addr (nor the 
+   defines), sockaddr_in6, or ip_mreq.  The ABI used by the linux-kernel
+   and glibc match exactly.  Neither the linux kernel nor glibc should 
+   break this ABI without coordination.  */
+#ifdef _UAPI_LINUX_IN6_H
+# define __USE_KERNEL_IPV6_DEFS 
+#endif
+
 /* Options for use with `getsockopt' and `setsockopt' at the IP level.
    The first word in the comment at the right is the data type used;
    "bool" means a boolean value stored in an `int'.  */

---

linux/

Compile tested and inspected.

 include/uapi/linux/in.h      |   32 +-----------------------
 include/uapi/linux/in6.h     |   50 +++++++++++++++++++++++++++++++------
 include/uapi/linux/ipproto.h |    0 
 include/uapi/linux/ipproto.h |   57 +++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 101 insertions(+), 38 deletions(-)
include/uapi/linux/in.h +2 -30
index 9edb441..998ecb2 100644
@@ -21,36 +21,8 @@
 #include <linux/types.h>
 #include <linux/socket.h>
 
-/* Standard well-defined IP protocols.  */
-enum {
-  IPPROTO_IP = 0,		/* Dummy protocol for TCP		*/
-  IPPROTO_ICMP = 1,		/* Internet Control Message Protocol	*/
-  IPPROTO_IGMP = 2,		/* Internet Group Management Protocol	*/
-  IPPROTO_IPIP = 4,		/* IPIP tunnels (older KA9Q tunnels use 94) */
-  IPPROTO_TCP = 6,		/* Transmission Control Protocol	*/
-  IPPROTO_EGP = 8,		/* Exterior Gateway Protocol		*/
-  IPPROTO_PUP = 12,		/* PUP protocol				*/
-  IPPROTO_UDP = 17,		/* User Datagram Protocol		*/
-  IPPROTO_IDP = 22,		/* XNS IDP protocol			*/
-  IPPROTO_DCCP = 33,		/* Datagram Congestion Control Protocol */
-  IPPROTO_RSVP = 46,		/* RSVP protocol			*/
-  IPPROTO_GRE = 47,		/* Cisco GRE tunnels (rfc 1701,1702)	*/
-
-  IPPROTO_IPV6	 = 41,		/* IPv6-in-IPv4 tunnelling		*/
-
-  IPPROTO_ESP = 50,            /* Encapsulation Security Payload protocol */
-  IPPROTO_AH = 51,             /* Authentication Header protocol       */
-  IPPROTO_BEETPH = 94,	       /* IP option pseudo header for BEET */
-  IPPROTO_PIM    = 103,		/* Protocol Independent Multicast	*/
-
-  IPPROTO_COMP   = 108,                /* Compression Header protocol */
-  IPPROTO_SCTP   = 132,		/* Stream Control Transport Protocol	*/
-  IPPROTO_UDPLITE = 136,	/* UDP-Lite (RFC 3828)			*/
-
-  IPPROTO_RAW	 = 255,		/* Raw IP packets			*/
-  IPPROTO_MAX
-};
-
+/* Include IPPROTO_* defines. */
+#include <linux/ipproto.h>
 
 /* Internet address. */
 struct in_addr {
include/uapi/linux/in6.h +42 -8

85 lines not on this page · raw message

include/uapi/linux/ipproto.h new file +57 -0

57 lines not on this page · raw message

---

Cheers,
Carlos.

#23 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

include/uapi/linux/in6.h

Quoted 4 lines · Carlos O'Donell
include/uapi/linux/in6.h +0 -0
index f79c372..a2b16a5 100644

:

include/uapi/linux/in6.h struct in6_flowlabel_req { · line 144

Quoted 11 lines

#define IPV6_PRIORITY_14 0x0e00
#define IPV6_PRIORITY_15 0x0f00

include/uapi/linux/in6.h +6 -0
+
+#ifndef _NETINET_IN_H
+#if defined (__GLIBC__)
+/* Include all of the other IPPROTO_* defines for userspace. */
+#include <linux/ipproto.h>
+#endif
 /*
  *	IPV6 extension headers

Overall, it looks good, but why including linux/ipproto.h?

--yoshfuji

maintainer
#24 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Carlos O'Donell

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 21:15:03 -0500

include/uapi/linux/in6.h line 32

Quoted 7 lines
include/uapi/linux/in6.h +7 -0
+/* If a glibc-based userspace has already included in.h, then we will not 
+ * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The
+ * ABI used by the kernel and by glibc match exactly. Neither the kernel
+ * nor glibc should break this ABI without coordination.
+ */
+#ifndef _NETINET_IN_H
+

I think we should shoot for a non-glibc-centric solution.

I can't imagine that other libc's won't have the same exact problem with their netinet/in.h conflicting with the kernel's, redefining structures like in6_addr, that we'd want to provide a protection scheme for here as well.

Let's pick some more generic names and themes for the CPP macros and comments we use to protect the header file blocks and describe that protection scheme.

Thanks.

#34 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Carlos O'Donell
Quoted 8 lines · Carlos O'Donell
include/uapi/linux/in6.h +0 -0
index f79c372..a2b16a5 100644
@@ -23,6 +23,13 @@
 
 #include <linux/types.h>
 

include/uapi/linux/in6.h line 35

Quoted 10 lines
include/uapi/linux/in6.h +7 -0
+/* If a glibc-based userspace has already included in.h, then we will not 
+ * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The
+ * ABI used by the kernel and by glibc match exactly. Neither the kernel
+ * nor glibc should break this ABI without coordination.
+ */
+#ifndef _NETINET_IN_H
+
 /*
  *	IPv6 address structure
  */

This should be
#if !defined(__GLIBC__) || !defined(_NETINET_IN_H)

Quoted 9 lines
include/uapi/linux/in6.h +3 -0
@@ -30,12 +37,20 @@
 struct in6_addr {
 	union {
 		__u8		u6_addr8[16];
+#if !defined(__GLIBC__) \
+    || (defined(__GLIBC__) && (defined(__USE_MISC) || defined(__USE_GNU))) \
+    || defined(__KERNEL__)
 		__be16		u6_addr16[8];
 		__be32		u6_addr32[4];

include/uapi/linux/in6.h line 54

Quoted 10 lines
include/uapi/linux/in6.h +5 -0
+#endif
 	} in6_u;
+#if !defined(__GLIBC__) \
+    || (defined(__GLIBC__) && (defined(__USE_MISC) || defined(__USE_GNU))) \
+    || defined(__KERNEL__)
 #define s6_addr			in6_u.u6_addr8
 #define s6_addr16		in6_u.u6_addr16
 #define s6_addr32		in6_u.u6_addr32
+#endif
 };

2nd "if" be after s6_addr?

--yoshfuji

#16 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Mike Frysinger
Quoted 5 lines · Mike Frysinger

but this is still too vague. what headers/definitions do people want to see simultaneously included ? changes would be needed on both sides (kernel & C library).

Hi, Mike,

Please take a look at my first email in this thread. The user application includes <linux/if_bridge.h> and <netinet/in.h>.

<linux/if_bridge.h> uses struct_in6 but doesn't include <linux/in6.h> (this is my bad, sorry), an obvious fix is just including <linux/in6.h>. But this immediately breaks applications which include <linux/if_bridge.h> and <netinet/in.h>, just as what Thomas reported.

And if_bridge.h is kernel-specific, there is no corresponding glibc one, so you can't blame applications which include both of them.

Thanks.

#17 lore raw

Re: [libvirt] Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to David Miller
Quoted 5 lines · David Miller

When GLIBC doesn't provide it's own definition of some networking macros or interfaces that the kernel provides, people include the kernel header.

Recently I got a problem when copying a structure from kernel to userspace, after debugging I found:

kernel: include/linux/inet.h

#define INET6_ADDRSTRLEN (48)

glibc: /usr/include/netinet/in.h

#define INET6_ADDRSTRLEN 46

Any reason to differentiate them from each other?

Signature · 2 lines
Thanks,
Jike
#18 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 13 lines · Jike Song
Quoted 5 lines · David Miller

When GLIBC doesn't provide it's own definition of some networking macros or interfaces that the kernel provides, people include the kernel header.

Recently I got a problem when copying a structure from kernel to userspace, after debugging I found:

kernel: include/linux/inet.h

Quoted 9 lines

#define INET6_ADDRSTRLEN (48)

glibc: /usr/include/netinet/in.h

#define INET6_ADDRSTRLEN 46

Any reason to differentiate them from each other?

I see no reason, even although I don't know why it is 46 instead of 40.

But include/linux/inet.h is not exported to user-space, AFAIK.

#19 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

----- Original Message -----

Quoted 3 lines

I see no reason, even although I don't know why it is 46 instead of 40.

Ok, for "0000:0000:0000:0000:0000:0000:255.255.255.255".

#39 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to David Miller
Quoted 11 lines · David Miller

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 20:58:47 -0500

Quoted 6 lines

So I just went down the rabbit hole, and the further I get the closer I get to having two exact copies of the same definitions in both glibc and the kernel and using whichever one was included first.

Is anyone opposed to that kind of solution?

Sounds interesting, please share :-)

iptables has the same issue, and solved it its way. (uapi/)linux/netfilter.h is used to get at things like union nf_inet_addr. This union contains struct in6_addr. There is no include for in6_addr in netfilter.h itself. This may break the "standalone compilation" test, but at least allows for specifying the environment-specific header for in6_addr in the C file:

a. userspace: #include <netinet/in.h> before <linux/netfilter.h> b. kernel parts: #include <linux/in6.h> before <linux/netfilter.h>

#40 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to David Miller
Quoted 8 lines · On Wednesday 16 January 2013 16:45:11 David Miller wrote:

What would be really nice is if GCC treated multiple identical definitions of structures the same way it handles multiple identical definitions of CPP defines. Which is to silently accept them.

But that's not the case, so we need a different solution.

Another thing to do is to use a new structure for in6_addr in kernel headers exported to userland.

the kernel already exports many types with a __kernel_ prefix. i changed the kernel headers in Gentoo for a few releases (2.6.28 - 2.6.34) to do the same thing to pretty much all the networking headers. a few packages broke, but the number was low, and trivial to fix (a sed would do it most of the time).

it's also trivial for userland packages to detect that they need to do this sort of thing in a local header by using linux/version.h and a set of defines to redirect the new structure name back to the old one.

would be a lot cleaner to just break it and be done. -mike

#35 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to YOSHIFUJI Hideaki
Quoted 9 lines · YOSHIFUJI Hideaki
Quoted 8 lines · Carlos O'Donell
include/uapi/linux/in6.h +0 -0
index f79c372..a2b16a5 100644
@@ -23,6 +23,13 @@
 
 #include <linux/types.h>
 
Quoted 13 lines

include/uapi/linux/in6.h line 35

Quoted 10 lines
include/uapi/linux/in6.h +7 -0
+/* If a glibc-based userspace has already included in.h, then we will not
+ * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The
+ * ABI used by the kernel and by glibc match exactly. Neither the kernel
+ * nor glibc should break this ABI without coordination.
+ */
+#ifndef _NETINET_IN_H
+
 /*
  *	IPv6 address structure
  */

This should be
#if !defined(__GLIBC__) || !defined(_NETINET_IN_H)

Correct. If it's non-glibc we want these defines to be used even if a non-glibc defined _NETINET_IN_H.

Quoted 13 lines
Quoted 13 lines
include/uapi/linux/in6.h +6 -0
@@ -30,12 +37,20 @@
 struct in6_addr {
 	union {
 		__u8		u6_addr8[16];
+#if !defined(__GLIBC__) \
+    || (defined(__GLIBC__) && (defined(__USE_MISC) || defined(__USE_GNU))) \
+    || defined(__KERNEL__)
 		__be16		u6_addr16[8];
 		__be32		u6_addr32[4];
+#endif
 	} in6_u;
+#if !defined(__GLIBC__) \
+    || (defined(__GLIBC__) && (defined(__USE_MISC) || defined(__USE_GNU))) \
Quoted 8 lines

include/uapi/linux/in6.h line 54

Quoted 6 lines
include/uapi/linux/in6.h +2 -0
+    || defined(__KERNEL__)
 #define s6_addr			in6_u.u6_addr8
 #define s6_addr16		in6_u.u6_addr16
 #define s6_addr32		in6_u.u6_addr32
+#endif
 };

2nd "if" be after s6_addr?

Correct. We want to unconditinally define s6_addr in this case.

I've fixed both of those cases in the next version of the patch. Thanks.

Cheers,
Carlos.

#25 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to David Miller
Quoted 13 lines · On Wednesday 16 January 2013 22:15:38 David Miller wrote:

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 21:15:03 -0500

include/uapi/linux/in6.h line 32

Quoted 7 lines

+/* If a glibc-based userspace has already included in.h, then we will not + * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The + * ABI used by the kernel and by glibc match exactly. Neither the kernel + * nor glibc should break this ABI without coordination.
+ */

include/uapi/linux/in6.h +2 -0
+#ifndef _NETINET_IN_H
+

I think we should shoot for a non-glibc-centric solution.

Quoted 4 lines

I can't imagine that other libc's won't have the same exact problem with their netinet/in.h conflicting with the kernel's, redefining structures like in6_addr, that we'd want to provide a protection scheme for here as well.

yes, the kernel's use of __GLIBC__ in exported headers has already caused problems in the past. fortunately, it's been reduced down to just one case now (stat.h). let's not balloon it back up.
-mike

#26 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 19 lines · Mike Frysinger
Quoted 13 lines · On Wednesday 16 January 2013 22:15:38 David Miller wrote:

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 21:15:03 -0500

include/uapi/linux/in6.h line 32

Quoted 7 lines

+/* If a glibc-based userspace has already included in.h, then we will not + * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The + * ABI used by the kernel and by glibc match exactly. Neither the kernel + * nor glibc should break this ABI without coordination.
+ */

include/uapi/linux/in6.h +2 -0
+#ifndef _NETINET_IN_H
+

I think we should shoot for a non-glibc-centric solution.

Quoted 4 lines

I can't imagine that other libc's won't have the same exact problem with their netinet/in.h conflicting with the kernel's, redefining structures like in6_addr, that we'd want to provide a protection scheme for here as well.

Quoted 4 lines

yes, the kernel's use of __GLIBC__ in exported headers has already caused problems in the past. fortunately, it's been reduced down to just one case now (stat.h). let's not balloon it back up.
-mike

I also see coda.h has grown a __GLIBC__ usage.

In the next revision of the patch I created a single libc-compat.h header which encompasses the logic for any libc that wants to coordinate with the kernel headers.

It's simple enough to move all of the __GLIBC__ uses into libc-compat.h, then you control userspace libc coordination from one file.

Cheers,
Carlos.

#27 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 20 lines · On Thursday 17 January 2013 23:22:26 Carlos O'Donell wrote:
Quoted 19 lines · Mike Frysinger
Quoted 17 lines · On Wednesday 16 January 2013 22:15:38 David Miller wrote:

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 21:15:03 -0500

include/uapi/linux/in6.h line 32

Quoted 7 lines

+/* If a glibc-based userspace has already included in.h, then we will not + * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The + * ABI used by the kernel and by glibc match exactly. Neither the kernel + * nor glibc should break this ABI without coordination.
+ */

include/uapi/linux/in6.h +2 -0
+#ifndef _NETINET_IN_H
+

I think we should shoot for a non-glibc-centric solution.

I can't imagine that other libc's won't have the same exact problem with their netinet/in.h conflicting with the kernel's, redefining structures like in6_addr, that we'd want to provide a protection scheme for here as well.

Quoted 5 lines
Quoted 3 lines

yes, the kernel's use of __GLIBC__ in exported headers has already caused problems in the past. fortunately, it's been reduced down to just one case now (stat.h). let's not balloon it back up.

I also see coda.h has grown a __GLIBC__ usage.

that file is just a pile of cruft :). it's something that'd be rejected by today's kernel standard as it's full of OS shim code. "#ifdef DOS" ? comeon!

fortunately, coda is pretty uncommon, so this issue doesn't bite most people, and i've never bothered with it. the same cannot be said of stat.h. -mike

maintainer
#41 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Mike Frysinger

From: Mike Frysinger <[email protected]>
Date: Thu, 17 Jan 2013 23:14:31 -0500

Quoted 10 lines

the kernel already exports many types with a __kernel_ prefix. i changed the kernel headers in Gentoo for a few releases (2.6.28 - 2.6.34) to do the same thing to pretty much all the networking headers. a few packages broke, but the number was low, and trivial to fix (a sed would do it most of the time).

it's also trivial for userland packages to detect that they need to do this sort of thing in a local header by using linux/version.h and a set of defines to redirect the new structure name back to the old one.

would be a lot cleaner to just break it and be done.

We are not at liberty to break something that has legitimately compiled successfully for two decades.

Your __kernel_ prefix idea breaks compilation just as equally.

One thing you certainly don't have to be is happy about this header file situation, but you cannot use that dissatisfaction as justification to make the situation worse by breaking the build for people outside of the world you directly control.

#42 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 7 lines · On Thursday 17 January 2013 23:55:24 David Miller wrote:

From: Mike Frysinger <[email protected]>

Quoted 6 lines

the kernel already exports many types with a __kernel_ prefix. i changed the kernel headers in Gentoo for a few releases (2.6.28 - 2.6.34) to do the same thing to pretty much all the networking headers. a few packages broke, but the number was low, and trivial to fix (a sed would do it most of the time).

Quoted 8 lines
Quoted 5 lines

it's also trivial for userland packages to detect that they need to do this sort of thing in a local header by using linux/version.h and a set of defines to redirect the new structure name back to the old one.

would be a lot cleaner to just break it and be done.

We are not at liberty to break something that has legitimately compiled successfully for two decades.

i doubt there are code bases that have compiled successfully for that long w/out being changed. in fact, the kernel user headers (until recently) have been a volatile mess requiring constant care & feeding of user programs to keep compiling. if anything, this would simply be maintaining the status quo for a bit longer.

Quoted 4 lines

One thing you certainly don't have to be is happy about this header file situation, but you cannot use that dissatisfaction as justification to make the situation worse by breaking the build for people outside of the world you directly control.

i think you're making wrong assumptions about my position on the topic. i never said i was against cleaning up the kernel user headers (and in fact, you can find many commits from me doing exactly that over the years in the kernel). -mike

#28 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Carlos O'Donell
Quoted 27 lines · Carlos O'Donell
Quoted 19 lines · Mike Frysinger
Quoted 13 lines · On Wednesday 16 January 2013 22:15:38 David Miller wrote:

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 21:15:03 -0500

include/uapi/linux/in6.h line 32

Quoted 7 lines

+/* If a glibc-based userspace has already included in.h, then we will not + * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The + * ABI used by the kernel and by glibc match exactly. Neither the kernel + * nor glibc should break this ABI without coordination.
+ */

include/uapi/linux/in6.h +2 -0
+#ifndef _NETINET_IN_H
+

I think we should shoot for a non-glibc-centric solution.

Quoted 4 lines

I can't imagine that other libc's won't have the same exact problem with their netinet/in.h conflicting with the kernel's, redefining structures like in6_addr, that we'd want to provide a protection scheme for here as well.

Quoted 4 lines

yes, the kernel's use of __GLIBC__ in exported headers has already caused problems in the past. fortunately, it's been reduced down to just one case now (stat.h). let's not balloon it back up.
-mike

I also see coda.h has grown a __GLIBC__ usage.

Quoted 7 lines

In the next revision of the patch I created a single libc-compat.h header which encompasses the logic for any libc that wants to coordinate with the kernel headers.

It's simple enough to move all of the __GLIBC__ uses into libc-compat.h, then you control userspace libc coordination from one file.

How about just deciding on a single macro/symbol both the kernel and libc (any libc that needs this) define? Something like both the kernel and userland doing:

#ifndef __IPV6_BITS_DEFINED
#define __IPV6_BITS_DEFINED
...
define in6_addr, sockaddr_in6, ipv6_mreq, whatnot #endif

So whichever the application includes first, wins. Too naive? I didn't see this option being discarded, so not sure it was considered.

Signature · 1 line
Pedro Alves
#29 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 46 lines · Pedro Alves
Quoted 27 lines · Carlos O'Donell
Quoted 19 lines · Mike Frysinger
Quoted 13 lines · On Wednesday 16 January 2013 22:15:38 David Miller wrote:

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 21:15:03 -0500

include/uapi/linux/in6.h line 32

Quoted 7 lines

+/* If a glibc-based userspace has already included in.h, then we will not + * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The + * ABI used by the kernel and by glibc match exactly. Neither the kernel + * nor glibc should break this ABI without coordination.
+ */

include/uapi/linux/in6.h +2 -0
+#ifndef _NETINET_IN_H
+

I think we should shoot for a non-glibc-centric solution.

Quoted 4 lines

I can't imagine that other libc's won't have the same exact problem with their netinet/in.h conflicting with the kernel's, redefining structures like in6_addr, that we'd want to provide a protection scheme for here as well.

Quoted 4 lines

yes, the kernel's use of __GLIBC__ in exported headers has already caused problems in the past. fortunately, it's been reduced down to just one case now (stat.h). let's not balloon it back up.
-mike

I also see coda.h has grown a __GLIBC__ usage.

Quoted 7 lines

In the next revision of the patch I created a single libc-compat.h header which encompasses the logic for any libc that wants to coordinate with the kernel headers.

It's simple enough to move all of the __GLIBC__ uses into libc-compat.h, then you control userspace libc coordination from one file.

How about just deciding on a single macro/symbol both the kernel and libc (any libc that needs this) define? Something like both the kernel and userland doing:

#ifndef __IPV6_BITS_DEFINED
#define __IPV6_BITS_DEFINED
...
define in6_addr, sockaddr_in6, ipv6_mreq, whatnot #endif

Quoted 3 lines

So whichever the application includes first, wins. Too naive? I didn't see this option being discarded, so not sure it was considered.

Too naive, but *close* to what my patch does :-)

The kernel definitions when included first, and in a glibc userspace, must try to mimic the glibc userspace headers and we need more than one guard macro to do that effectively.

The reason I jumped into the code is because this kind of problem is easy to talk about, but the devil is in the details.

There are certainly some compromises on both sides, but the solution promises to solve this problem.

Honestly without UAPI this would have been an impossible task.

Cheers,
Carlos.

#30 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 41 lines · Carlos O'Donell
Quoted 40 lines · Pedro Alves
Quoted 27 lines · Carlos O'Donell
Quoted 19 lines · Mike Frysinger
Quoted 13 lines · On Wednesday 16 January 2013 22:15:38 David Miller wrote:

From: Carlos O'Donell <[email protected]>
Date: Wed, 16 Jan 2013 21:15:03 -0500

include/uapi/linux/in6.h line 32

Quoted 7 lines

+/* If a glibc-based userspace has already included in.h, then we will not + * define in6_addr (nor the defines), sockaddr_in6, or ipv6_mreq. The + * ABI used by the kernel and by glibc match exactly. Neither the kernel + * nor glibc should break this ABI without coordination.
+ */

include/uapi/linux/in6.h +2 -0
+#ifndef _NETINET_IN_H
+

I think we should shoot for a non-glibc-centric solution.

Quoted 4 lines

I can't imagine that other libc's won't have the same exact problem with their netinet/in.h conflicting with the kernel's, redefining structures like in6_addr, that we'd want to provide a protection scheme for here as well.

Quoted 4 lines

yes, the kernel's use of __GLIBC__ in exported headers has already caused problems in the past. fortunately, it's been reduced down to just one case now (stat.h). let's not balloon it back up.
-mike

I also see coda.h has grown a __GLIBC__ usage.

Quoted 7 lines

In the next revision of the patch I created a single libc-compat.h header which encompasses the logic for any libc that wants to coordinate with the kernel headers.

It's simple enough to move all of the __GLIBC__ uses into libc-compat.h, then you control userspace libc coordination from one file.

How about just deciding on a single macro/symbol both the kernel and libc (any libc that needs this) define? Something like both the kernel and userland doing:

Quoted 5 lines
Quoted 5 lines

#ifndef __IPV6_BITS_DEFINED
#define __IPV6_BITS_DEFINED
...
define in6_addr, sockaddr_in6, ipv6_mreq, whatnot
#endif

Hmm, how should we handle future structs/enums then? For example, if I want to have in6_flowlabel_req{} defined in glibc, what should we do?

We probably want to have __LIBC_HAS_STRUCT_IN6_FLOWLABEL_REQ defined.

--yoshfuji

#31 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 13 lines · YOSHIFUJI Hideaki
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines

It's simple enough to move all of the __GLIBC__ uses into libc-compat.h, then you control userspace libc coordination from one file.

How about just deciding on a single macro/symbol both the kernel and libc (any libc that needs this) define? Something like both the kernel and userland doing:

Quoted 5 lines
Quoted 5 lines

#ifndef __IPV6_BITS_DEFINED
#define __IPV6_BITS_DEFINED
...
define in6_addr, sockaddr_in6, ipv6_mreq, whatnot
#endif

Quoted 3 lines

Hmm, how should we handle future structs/enums then? For example, if I want to have in6_flowlabel_req{} defined in glibc, what should we do?

Include the glibc header first? Or is this some other use case?

The point wasn't that you'd have only one macro for all structs/enums (you could split into __IPV6_IN6_ADDR_DEFINED, __IPV6_SOCKADDR_IN6_DEFINED, etc.) but to have the kernel and libc agree on guard macros, instead of having the kernel do #ifdef __GLIBC__ and glibc doing #ifdef _NETINET_IN_H.

But as Carlos says, the devil is in the details, and I sure am not qualified on the details here.

Signature · 1 line
Pedro Alves
#32 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

Quoted 22 lines · Pedro Alves
Quoted 13 lines · YOSHIFUJI Hideaki
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines

It's simple enough to move all of the __GLIBC__ uses into libc-compat.h, then you control userspace libc coordination from one file.

How about just deciding on a single macro/symbol both the kernel and libc (any libc that needs this) define? Something like both the kernel and userland doing:

Quoted 5 lines
Quoted 5 lines

#ifndef __IPV6_BITS_DEFINED
#define __IPV6_BITS_DEFINED
...
define in6_addr, sockaddr_in6, ipv6_mreq, whatnot
#endif

Quoted 3 lines

Hmm, how should we handle future structs/enums then? For example, if I want to have in6_flowlabel_req{} defined in glibc, what should we do?

Include the glibc header first? Or is this some other use case?

Quoted 8 lines

The point wasn't that you'd have only one macro for all structs/enums (you could split into __IPV6_IN6_ADDR_DEFINED, __IPV6_SOCKADDR_IN6_DEFINED, etc.) but to have the kernel and libc agree on guard macros, instead of having the kernel do #ifdef __GLIBC__ and glibc doing #ifdef _NETINET_IN_H.

But as Carlos says, the devil is in the details, and I sure am not qualified on the details here.

My plan is to have one _UAPI_DEF_xxx macro to guard each problematic definition in the kernel UAPI headers.

Then userspace can check for those macros and act appropriately.

The kernel likewise when detecting glibc headers included first can use the _UAPI_DEF_xxx macro guards to disable problematic definitions knowing that glibc has identical ABI and API-compatible versions that the program can use without breaking.

Cheers,
Carlos.

#33 lore raw

Re: Redefinition of struct in6_addr in <netinet/in.h> and <linux/in6.h>

In reply to Pedro Alves
Quoted 22 lines · On Friday 18 January 2013 09:36:35 Pedro Alves wrote:
Quoted 14 lines · YOSHIFUJI Hideaki
Quoted 8 lines
Quoted 8 lines
Quoted 3 lines

It's simple enough to move all of the __GLIBC__ uses into libc-compat.h, then you control userspace libc coordination from one file.

How about just deciding on a single macro/symbol both the kernel and libc (any libc that needs this) define? Something like both the kernel and userland doing:

Quoted 5 lines
Quoted 5 lines

#ifndef __IPV6_BITS_DEFINED
#define __IPV6_BITS_DEFINED
...
define in6_addr, sockaddr_in6, ipv6_mreq, whatnot
#endif

Quoted 3 lines

Hmm, how should we handle future structs/enums then? For example, if I want to have in6_flowlabel_req{} defined in glibc, what should we do?

Include the glibc header first? Or is this some other use case?

Quoted 8 lines

The point wasn't that you'd have only one macro for all structs/enums (you could split into __IPV6_IN6_ADDR_DEFINED, __IPV6_SOCKADDR_IN6_DEFINED, etc.) but to have the kernel and libc agree on guard macros, instead of having the kernel do #ifdef __GLIBC__ and glibc doing #ifdef _NETINET_IN_H.

But as Carlos says, the devil is in the details, and I sure am not qualified on the details here.

i shipped some kernel header versions in Gentoo where the linux network headers would include the C library's header (as defined by POSIX) so the structs that were the same would only come from glibc. the only reported breakage was due to defines that the kernel provided but glibc did not. -mike

#43 lore raw
In reply to Eric Blake
Quoted 34 lines · Eric Blake
Quoted 15 lines · Thomas Backlund

Thomas Backlund skrev 13.1.2013 20:38:

Quoted 9 lines

patch both inline and attached as thunderbird tends to mess up ...

-----

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

Ok,
ignore this patch, it's not the correct fix as it introduces redefinitions...

Quoted 5 lines

Btw, the error that I hit that made me suggest this fix was libvirt config check bailing out:

config.log:/usr/include/linux/if_bridge.h:173:20: error: field 'ip6' has incomplete type

Hmm, just now noticing this thread, after independently hitting and and having to work around the same problem in libvirt:
https://www.redhat.com/archives/libvir-list/2013-January/msg00930.html
https://bugzilla.redhat.com/show_bug.cgi?id=895141

Quoted 7 lines
Quoted 7 lines

--- linux-3.8-rc3/include/uapi/linux/if_bridge.h 2013-01-13
20:09:54.257271755 +0200
+++ linux-3.8-rc3.fix/include/uapi/linux/if_bridge.h 2013-01-13
20:15:04.153676151 +0200

(quoted diff) +0 -0
@@ -14,6 +14,7 @@
 #define _UAPI_LINUX_IF_BRIDGE_H
 
Quoted 3 lines
Quoted 3 lines

include/uapi/linux/if_bridge.h 2013-01-13 20:15:04.153676151 +0200 line 18

Quoted 3 lines
include/uapi/linux/if_bridge.h 2013-01-13 20:15:04.153676151 +0200 renamed from include/uapi/linux/if_bridge.h 2013-01-13 20:09:54.257271755 +0200 +1 -0
 #include <linux/types.h>
+#include <linux/in6.h>
 

Did this ever get resolved ?

- k

#44 lore raw
Quoted 9 lines · Kumar Gala
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines

if_bridge.h uses struct in6_addr ip6; but does not include the in6.h header.

Found by trying to build libvirt and connman against 3.8-rc3 headers.

Did this ever get resolved ?

Libvirt has managed to work around the kernel issue in the meantime. But just today, I hit the same issue with the latest kernel-headers-3.8.2-206.fc18.x86_64 on Fedora 18 when backporting to an older libvirt branch that did not have the workaround. A kernel person would have to speak up for certain, but I think it is still a problem.

Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org