svn commit: r323254 - head/sys/compat/freebsd32

[Ian Lepore]

On Fri, 2017-09-08 at 08:14 -0400, John Baldwin wrote:
> On 9/7/17 12:29 AM, Maxim Sobolev wrote:
> > 
> > Author: sobomax
> > Date: Thu Sep  7 04:29:57 2017
> > New Revision: 323254
> > URL: https://svnweb.freebsd.org/changeset/base/323254
> > 
> > Log:
> >   In the recvmsg32() system call iterate over returned structure(s)
> >   and convert any messages of types SCM_BINTIME, SCM_TIMESTAMP,
> >   SCM_REALTIME and SCM_MONOTONIC from 64-bit to its 32-bit
> >   representation. Otherwise we either run out of user-supplied
> >   buffer to copy those out resulting in the MSG_CTRUNC or simply
> >   return values that the userland 32-bit code is not going
> >   to parse correctly. This fixes at least two regression tests
> >   failing to function properly in 32-bit compat mode:
> >   
> >       tools/regression/sockets/udp_pingpong
> >       tools/regression/sockets/unix_cmsg
> >   
> >   PR:             kern/222039
> >   MFC after:	30 days
> Is this correct on !amd64?  Other 32-bit platforms use a 64-bit time_t
> (note the time32_t type defined earlier in freebsd32.h).  struct bintime32
> should use time32_t for the seconds field, not uint32_t.  I think that
> will be sufficient to make this correct on !amd64 (it also means that
> bintime32 == bintime on !amd64 so you could perhaps use a simpler BT_CP
> for !amd64, but the existing one is probably ok).
> 

The existing one now does *(uint64_t *) on a value that's only aligned
to a 32-bit boundary.  That will work in practice because only i386 has
a 32-bit time_t that will use this code, and it's not a strict-
alignment platform.  It may still cause compiler warnings about
alignment.

-- Ian