[PATCH v06 30/36] uapi rdma/rdma_user_rxe.h: include in.h and in6.h

[PATCH v06 30/36] uapi rdma/rdma_user_rxe.h: include in.h and in6.h

[Mikko Rapeli]

Fixes these userspace compilation errors when rdma_user_rxe.h is compiled
alone:

rdma/rdma_user_rxe.h:59:20: error: field ‘_sockaddr’ has incomplete type
   struct sockaddr  _sockaddr;
                    ^~~~~~~~~
rdma/rdma_user_rxe.h:60:22: error: field ‘_sockaddr_in’ has incomplete type
   struct sockaddr_in _sockaddr_in;
                      ^~~~~~~~~~~~
rdma/rdma_user_rxe.h:61:23: error: field ‘_sockaddr_in6’ has incomplete type
   struct sockaddr_in6 _sockaddr_in6;
                       ^~~~~~~~~~~~~

Signed-off-by: Mikko Rapeli <mikko.rapeli-X3B1VOXEql0@public.gmane.org>
Acked-by: Moni Shoua <monis-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Acked-by: Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
Cc: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Cc: Hal Rosenstock <hal.rosenstock-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
---
 include/uapi/rdma/rdma_user_rxe.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/uapi/rdma/rdma_user_rxe.h b/include/uapi/rdma/rdma_user_rxe.h
index 1de99cfdaf7d..ce6c929ca331 100644
--- a/include/uapi/rdma/rdma_user_rxe.h
+++ b/include/uapi/rdma/rdma_user_rxe.h
@@ -34,6 +34,8 @@
 #define RDMA_USER_RXE_H
 
 #include <linux/types.h>
+#include <linux/in.h>
+#include <linux/in6.h>
 
 union rxe_gid {
 	__u8	raw[16];
-- 
2.13.3

Re: [PATCH v06 30/36] uapi rdma/rdma_user_rxe.h: include in.h and in6.h

[Jason Gunthorpe]

On Sun, Aug 06, 2017 at 06:44:21PM +0200, Mikko Rapeli wrote:
> Fixes these userspace compilation errors when rdma_user_rxe.h is compiled
> alone:
> 
> rdma/rdma_user_rxe.h:59:20: error: field ???_sockaddr??? has incomplete type
>    struct sockaddr  _sockaddr;
>                     ^~~~~~~~~
> rdma/rdma_user_rxe.h:60:22: error: field ???_sockaddr_in??? has incomplete type
>    struct sockaddr_in _sockaddr_in;
>                       ^~~~~~~~~~~~
> rdma/rdma_user_rxe.h:61:23: error: field ???_sockaddr_in6??? has incomplete type
>    struct sockaddr_in6 _sockaddr_in6;
>                        ^~~~~~~~~~~~~

I'm not sure this is a good idea, linux/in.h should not be included in
userspace users of this file, 'sockaddr_in' needs to come from glibc's
'netinet/in.h' instead..

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

Re: [PATCH v06 30/36] uapi rdma/rdma_user_rxe.h: include in.h and in6.h

[Moni Shoua]

>
> I'm not sure this is a good idea, linux/in.h should not be included in
> userspace users of this file, 'sockaddr_in' needs to come from glibc's
> 'netinet/in.h' instead..
>
> Jason
Is it wrong to include include/uapi/linux/in.h from userspace?

Re: [PATCH v06 30/36] uapi rdma/rdma_user_rxe.h: include in.h and in6.h

[Jason Gunthorpe]

On Wed, Aug 09, 2017 at 04:48:08PM +0300, Moni Shoua wrote:
> >
> > I'm not sure this is a good idea, linux/in.h should not be included in
> > userspace users of this file, 'sockaddr_in' needs to come from glibc's
> > 'netinet/in.h' instead..

> Is it wrong to include include/uapi/linux/in.h from userspace?

Generally speaking userspace should always use the glibc version of
the headers and not headers from under linux/

Even if it is OK to include both today, that may not be true as glibc
or the kernel changes, and we end up with broken builds.

For instance look at this commit for a real problem caused by mixing
net/if.h and linux/if.h:

commit dc386fcf4a6d5dc0cc0052437c1cb161b7781592
Author: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Date:   Fri Oct 14 12:04:42 2016 -0600

    verbs: Use cmake to detect if net/if.h vs netling/route/link.h is broken
    
    If not then just use the header directly, otherwise use the work around.
    
    The issue is that old libnl libraries include linux/if.h and the
    declarations there conflict with net/if.h. New libraries do not
    do this.
    
    Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>

Re: [PATCH v06 30/36] uapi rdma/rdma_user_rxe.h: include in.h and in6.h

[Dmitry V. Levin]

[-- Attachment #1: Type: text/plain, Size: 718 bytes --]

On Wed, Aug 09, 2017 at 09:52:41AM -0600, Jason Gunthorpe wrote:
> On Wed, Aug 09, 2017 at 04:48:08PM +0300, Moni Shoua wrote:
> > >
> > > I'm not sure this is a good idea, linux/in.h should not be included in
> > > userspace users of this file, 'sockaddr_in' needs to come from glibc's
> > > 'netinet/in.h' instead..
> 
> > Is it wrong to include include/uapi/linux/in.h from userspace?
> 
> Generally speaking userspace should always use the glibc version of
> the headers and not headers from under linux/

Sorry?  Isn't the main idea of UAPI in providing the Linux interface for
userspace, unlike glibc headers that provide other more or less standard
interfaces like GNU, POSIX, etc?


-- 
ldv

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH v06 30/36] uapi rdma/rdma_user_rxe.h: include in.h and in6.h

[Jason Gunthorpe]

On Thu, Aug 10, 2017 at 05:47:48AM +0300, Dmitry V. Levin wrote:
> On Wed, Aug 09, 2017 at 09:52:41AM -0600, Jason Gunthorpe wrote:
> > On Wed, Aug 09, 2017 at 04:48:08PM +0300, Moni Shoua wrote:
> > > >
> > > > I'm not sure this is a good idea, linux/in.h should not be included in
> > > > userspace users of this file, 'sockaddr_in' needs to come from glibc's
> > > > 'netinet/in.h' instead..
> > 
> > > Is it wrong to include include/uapi/linux/in.h from userspace?
> > 
> > Generally speaking userspace should always use the glibc version of
> > the headers and not headers from under linux/
> 
> Sorry?  Isn't the main idea of UAPI in providing the Linux interface for
> userspace, unlike glibc headers that provide other more or less standard
> interfaces like GNU, POSIX, etc?

The issue is when glibc and linux provide the same standards based API
(eg in.h) - then the uapi version of in.h is really only for
consumption by glibc and other very special cases.

Forcing apps that need to access the rdma uapi that is not implemented
in glibc to deal with that mess is what is problematic. The uapi
system has a number of issues like this :|

If you make it so I cannot include rdma_user_rxe.h and netinet/in.h,
(for instance) in the same file then you broke user space use of the
header and things are worse off..

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