PATCH] Revert "bridge: include in6.h in if_bridge.h for struct in6_addr"

PATCH In review Stephen Hemminger 2 participants 6 messages Dec 10, 2014 – Dec 11, 2014

Subsystems (3)

Networking Network drivers Ethernet bridge

lore

Share & tools

lore

Stephen Hemminger

author

2014-12-10 04:30 UTC #0 lore raw

This reverts commit 66f1c44887ba4f47d617f8ae21cf8e04e1892bd7.

iproute2 uses sanitized kernel headers, and this change broke the build. The problem is that the iproute2 ip/iplink_bridge_slave.c include if_bridge.h but this causes a redefintion error when Glibc <netlink/in.h> is included.

Since iproute2 is arguably the most important user of network headers, the uapi header change should be reverted.

Signed-off-by Stephen Hemminger <[email protected]>

---
Please update stable as well.

 include/uapi/linux/if_bridge.h | 1 -
 1 file changed, 1 deletion(-)

include/uapi/linux/if_bridge.h +0 -1

index b03ee8f..439fd0d 100644
@@ -15,7 +15,6 @@
 
 #include <linux/types.h>
 #include <linux/if_ether.h>
-#include <linux/in6.h>
 
 #define SYSFS_BRIDGE_ATTR	"bridge"
 #define SYSFS_BRIDGE_FDB	"brforward"

Signature · 1 line

2.1.3

David Miller

maintainer

2014-12-10 04:49 UTC #1 lore raw

From: Stephen Hemminger <[email protected]>
Date: Tue, 9 Dec 2014 20:30:07 -0800

Quoted 10 lines

This reverts commit 66f1c44887ba4f47d617f8ae21cf8e04e1892bd7.
iproute2 uses sanitized kernel headers, and this change broke the build. The problem is that the iproute2 ip/iplink_bridge_slave.c include if_bridge.h but this causes a redefintion error when Glibc <netlink/in.h> is included.
Since iproute2 is arguably the most important user of network headers, the uapi header change should be reverted.
Signed-off-by Stephen Hemminger <[email protected]>

Stephen, please understand first the reason for this change.

GLIBC and the kernel fight over various definitions of core networking types.

The only way to resolve this in the long term is this change in question.

If you can make iproute2 work with both the new and the old mechanims, please do so instead of reverting this important change.

Stephen Hemminger

author

2014-12-10 05:41 UTC #2 lore raw

PATCH v2 1/1 if_bridge: fix conflict with glibc

Even though kernel and glibc headers have same effective values Gcc complains about redefinitions. Since this is a header expected to be used by userspace, use glibc header.

This supersedes change in commit 66f1c44887ba4f47d617f8ae21cf8e04e1892bd7 and fixes build of iproute2 with Glibc-2.19. This follows similar usage in include/uapi/linux for l2tp.h

Signed-off-by Stephen Hemminger <[email protected]>

include/uapi/linux/if_bridge.h +3 -1

@@ -15,7 +15,9 @@
 
 #include <linux/types.h>
 #include <linux/if_ether.h>
-#include <linux/in6.h>
+#ifndef __KERNEL__
+#include <netinet/in.h>
+#endif
 
 #define SYSFS_BRIDGE_ATTR	"bridge"
 #define SYSFS_BRIDGE_FDB	"brforward"

David Miller

maintainer

2014-12-10 18:34 UTC #3 lore raw

Re: [PATCH v2] if_bridge: fix conflict with glibc

From: Stephen Hemminger <[email protected]>
Date: Tue, 9 Dec 2014 21:41:22 -0800

Quoted 13 lines

Even though kernel and glibc headers have same effective values Gcc complains about redefinitions. Since this is a header expected to be used by userspace, use glibc header.
This supersedes change in commit 66f1c44887ba4f47d617f8ae21cf8e04e1892bd7 and fixes build of iproute2 with Glibc-2.19. This follows similar usage in include/uapi/linux for l2tp.h
Signed-off-by Stephen Hemminger <[email protected]>

include/uapi/linux/if_bridge.h line 20

Quoted 7 lines

include/uapi/linux/if_bridge.h +3 -1

@@ -15,7 +15,9 @@
 
 #include <linux/types.h>
 #include <linux/if_ether.h>
-#include <linux/in6.h>
+#ifndef __KERNEL__
+#include <netinet/in.h>
+#endif

No, we really want to incluse the linux/in6.h header, as that's where all the special GLIBC CPP checks are, such as:

#if __UAPI_DEF_IN6_ADDR_ALT

Please research how we have resolved the conflict between GLIBC and the kernel's exported headers. We really need to use linux/in6.h for all of this to work.

I understand that it is upsetting that iproute2 stopped building for you, but I'd like to kindly ask that you look more deeply into this and think more longer term, rather than having a knee jerk reaction and looking for quick fixes.

Thanks.

Stephen Hemminger

author

2014-12-10 23:57 UTC #4 lore raw

Re: [PATCH v2] if_bridge: fix conflict with glibc

Quoted 29 lines · On Wed, 10 Dec 2014 13:34:17 -0500 (EST) David Miller <[email protected]>…

From: Stephen Hemminger <[email protected]>
Date: Tue, 9 Dec 2014 21:41:22 -0800
Quoted 13 lines
Even though kernel and glibc headers have same effective values Gcc complains about redefinitions. Since this is a header expected to be used by userspace, use glibc header.
This supersedes change in commit 66f1c44887ba4f47d617f8ae21cf8e04e1892bd7 and fixes build of iproute2 with Glibc-2.19. This follows similar usage in include/uapi/linux for l2tp.h
Signed-off-by Stephen Hemminger <[email protected]>
include/uapi/linux/if_bridge.h line 20
Quoted 7 lines
include/uapi/linux/if_bridge.h +3 -1
@@ -15,7 +15,9 @@
 
 #include <linux/types.h>
 #include <linux/if_ether.h>
-#include <linux/in6.h>
+#ifndef __KERNEL__
+#include <netinet/in.h>
+#endif
No, we really want to incluse the linux/in6.h header, as that's where all the special GLIBC CPP checks are, such as:
#if __UAPI_DEF_IN6_ADDR_ALT

Quoted 8 lines

Please research how we have resolved the conflict between GLIBC and the kernel's exported headers. We really need to use linux/in6.h for all of this to work.
I understand that it is upsetting that iproute2 stopped building for you, but I'd like to kindly ask that you look more deeply into this and think more longer term, rather than having a knee jerk reaction and looking for quick fixes.

I don't have the time to understand the intricacies of glibc headers. The problem is that Gcc warns about duplicate definitions in headers; this is a useful warning and not something that I want to disable. Hacks with #undef seem to be heading the wrong way.

David Miller

maintainer

2014-12-11 01:44 UTC #5 lore raw

Re: [PATCH v2] if_bridge: fix conflict with glibc

From: Stephen Hemminger <[email protected]>
Date: Wed, 10 Dec 2014 15:57:45 -0800

Quoted 17 lines

Quoted 5 lines · On Wed, 10 Dec 2014 13:34:17 -0500 (EST) David Miller <[email protected]>…
No, we really want to incluse the linux/in6.h header, as that's where all the special GLIBC CPP checks are, such as:
#if __UAPI_DEF_IN6_ADDR_ALT
Quoted 8 lines
Please research how we have resolved the conflict between GLIBC and the kernel's exported headers. We really need to use linux/in6.h for all of this to work.
I understand that it is upsetting that iproute2 stopped building for you, but I'd like to kindly ask that you look more deeply into this and think more longer term, rather than having a knee jerk reaction and looking for quick fixes.

Quoted 4 lines

I don't have the time to understand the intricacies of glibc headers. The problem is that Gcc warns about duplicate definitions in headers; this is a useful warning and not something that I want to disable. Hacks with #undef seem to be heading the wrong way.

GLIBC and the kernel have unconditionally defined various types in their copies of in.h, and the only way to resolve this moving forward is to have some communication between the two so we can know who is in fact in charge of instantianting the type.

The new defines, such as __UAPI_DEF_IN6_ADDR_ALT, are that new mechanism.