Add support for the x32 VDSO.  The x32 VDSO takes advantage of the
similarity between the x86-64 and the x32 ABIs to contain the same
content, only the container is different, as the x32 VDSO obviously is
an x32 shared object.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

At this point, one should be able to build an x32 kernel.

Note that for now we depend on CONFIG_IA32_EMULATION.  Long term, x32
and IA32 should be detangled.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

If CONFIG_X86_X32_ABI is defined, add the x32 system calls to the
system call tables.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Allow an x32 process to be started.
Originally-by: H. J. Lu <hjl.tools@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>

x32 uses the 64-bit signal frame format, obviously, but there are some
structures which mixes that with pointers or sizeof(long) types, as
such we have to create a handful of system calls specific to x32.  By
and large these are a mixture of the 64-bit and the compat system
calls.
Originally-by: H. J. Lu <hjl.tools@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Unfortunately a lot of the compat types are guarded with CONFIG_COMPAT
or the equivalent, so add a similar guard to <asm/sys_ia32.h> to avoid
compilation failures when CONFIG_COMPAT=n.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

x32 shares most system calls with x86-64, but unfortunately some
subsystem (the input subsystem is the chief offender) which require
is_compat() when operating with a 32-bit userspace.  The input system
actually has text files in sysfs whose meaning is dependent on
sizeof(long) in userspace!

We could solve this by having two completely disjoint system call
tables; requiring that each system call be duplicated.  This patch
takes a different approach: we add a flag to the system call number;
this flag doesn't affect the system call dispatch but requests compat
treatment from affected subsystems for the duration of the system call.

The change of cmpq to cmpl is safe since it immediately follows the
and.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Add rt_sigframe_x32 to <asm/sigframe.h>.  Unfortunately we can't just
define all the data structures unconditionally, due to the #ifdef
CONFIG_COMPAT in <linux/compat.h> and its trickle-down effects, hence
the #ifdef mess.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Add a definition for struct ucontext_x32; this is inherently a mix of
the 32- and 64-bit versions.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Export setup_sigcontext() and restore_sigcontext() from signal.c, so
we can use the 64-bit versions verbatim for x32.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

There are some definitions which are duplicated between
kernel/signal.c and ia32/ia32_signal.c; move them to a common header
file.

Rather than adding stuff to existing header files which contain data
structures, create a new header file; hence the slightly odd name
("all the good ones were taken.")

Note: nothing relied on signal_fault() being defined in
<asm/ptrace.h>.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Generate macros for the *kernel* code to use to refer to x32 system
calls.  These have an __NR_x32_ prefix and do not include
__X32_SYSCALL_BIT.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Generate <asm/unistd_x32.h>; this exports x32 system call numbers to
user space.

[ v2: Enclose all arguments to syshdr in '' so empty arguments aren't
      dropped on the floor. ]
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Split the 64-bit system calls into "64" (64-bit only) and "common"
(64-bit or x32) and add the x32 system call numbers.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

On x86, the only difference between sys_rt_sigprocmask and
sys32_rt_sigprocmask is the alignment of the data structures.
However, x86 allows data accesses with arbitrary alignment, and
therefore there is no reason for this code to be different.
Reported-by: Gregory M. Lueck <gregory.m.lueck@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

An x32 process is *almost* the same thing as a 64-bit process with a
32-bit address limit, but there are a few minor differences -- in
particular core dumps are 32 bits and signal handling is different.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

So far this has only been used in process_64.c, but the x32 code will
need it in additional code.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Allow some core dump-related fields to be overridden.  This allows
core dumps to work correctly for x32.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Roland McGrath <roland@redhat.com>
Cc: Oleg Nesterov <oleg@redhat.com>

For 32-bit ABIs which have real 64-bit registers, we don't want to
break the position argument into two.  However, we still need compat
support to deal with 32-bit pointers, so we can't just use
sys_p{read,write} directly.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Rather than using "unsigned long" which is ABI-dependent, use
__kernel_ulong_t to define the externally visible type aio_context_t.

Note: the change in this form will cause unsigned long/unsigned int
differences on existing ABIs.  If that is unacceptable we may have to
define a new type.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>

Use helper functions aware of COMPAT_USE_64BIT_TIME to write struct
timeval and struct timespec to userspace in net/socket.c.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Handle 64-bit time structures in the networking core compat code.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: David S. Miller <davem@davemloft.net>

Enable the Bluetooth subsystem to be used with a compat ABI with
64-bit time.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Marcel Holtmann <marcel@holtmann.org>
Cc: Gustavo F. Padovan <padovan@profusion.mobi>
Cc: David S. Miller <davem@davemloft.net>

Enable the input system to be used with a compat ABI with 64-bit time.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Enable the lp driver to be used with a compat ABI with 64-bit time.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <greg@kroah.com>

Add helper functions to read and write struct timeval and struct
timespec from userspace.  We already had helper functions for reading
and writing struct compat_timespec; add a set of functions to do the
same with struct timeval, and add a second suite of functions which
can be sensitive to COMPAT_USE_64BIT_TIME and access either 32- or
64-bit time structures.

This also exports these helper functions to modules.

Rename the existing inlines for converting between struct
compat_timeval and native struct timespec so we can have a saner
naming convention for the exported functions.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Allow a compatibility ABI to use a 64-bit time_t and 64-bit members in
struct timeval and struct timespec to avoid the Y2038 problem.

This will be used for the x32 ABI.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Change <linux/sysinfo.h> to use explicitly sized types.  Replace
long/unsigned long with __kernel_[u]long_t so that a non-legacy 32-bit
ABI running on a 64-bit kernel can export those as 64-bit types.
Originally-by: H. J. Lu <hjl.tools@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

This is the same as the 64-bit posix_types.h, except that
__kernel_[u]long_t is defined to be [unsigned] long long and therefore
64 bits.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Introduce __kernel_[u]long_t, which allows an ABI to override all
defaults of type [unsigned] long.

This enables x32 and potentially other 32-bit userspace on 64-bit
kernel ABIs.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

struct sysinfo is just about the only thing exported to userspace from
<linux/kernel.h>, so move it into a separate header file with a
residual #include in <linux/kernel.h>.
Originally-by: H. J. Lu <hjl.tools@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: http://lkml.kernel.org/n/tip-4pr1xnnksprt7t0h3w5fw4rv@git.kernel.org

Use explicit sizes (__u64) instead of implicit sizes (unsigned long)
in the definition for sigcontext.h; this will allow this structure to
be shared between the x86-64 native ABI and the x32 ABI.
Originally-by: H. J. Lu <hjl.tools@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: http://lkml.kernel.org/n/tip-4pr1xnnksprt7t0h3w5fw4rv@git.kernel.org

Factor out IA32 (compatibility instruction set) from 32-bit address
space in the thread_info flags; this is a precondition patch for x32
support.
Originally-by: H. J. Lu <hjl.tools@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: http://lkml.kernel.org/n/tip-4pr1xnnksprt7t0h3w5fw4rv@git.kernel.org

Delete the __FD_*() functions for operating on fd_set structs from
linux/time.h as they're no longer used within the kernel with the preceding
patch and are not exported to userspace.

Whilst linux/time.h *does* export the FD_*() equivalents as wrappers around
__FD_*(), userspace provides its own definition of __FD_*().

Note that the definition of FD_ZERO() in linux/time.h may not be used with the
fd_sets associated with struct fdtable as the fd_set may have been allocated in
a truncated fashion.
Signed-off-by: David Howells <dhowells@redhat.com>
Link: http://lkml.kernel.org/r/20120216175006.23314.18984.stgit@warthog.procyon.org.ukSigned-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>

Replace the fd_sets in struct fdtable with an array of unsigned longs and then
use the standard non-atomic bit operations rather than the FD_* macros.

This:

 (1) Removes the abuses of struct fd_set:

     (a) Since we don't want to allocate a full fd_set the vast majority of the
     	 time, we actually, in effect, just allocate a just-big-enough array of
     	 unsigned longs and cast it to an fd_set type - so why bother with the
     	 fd_set at all?

     (b) Some places outside of the core fdtable handling code (such as
     	 SELinux) want to look inside the array of unsigned longs hidden inside
     	 the fd_set struct for more efficient iteration over the entire set.

 (2) Eliminates the use of FD_*() macros in the kernel completely.

 (3) Permits the __FD_*() macros to be deleted entirely where not exposed to
     userspace.
Signed-off-by: David Howells <dhowells@redhat.com>
Link: http://lkml.kernel.org/r/20120216174954.23314.48147.stgit@warthog.procyon.org.ukSigned-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>

Wrap accesses to the fd_sets in struct fdtable (for recording open files and
close-on-exec flags) so that we can move away from using fd_sets since we
abuse the fd_set structs by not allocating the full-sized structure under
normal circumstances and by non-core code looking at the internals of the
fd_sets.

The first abuse means that use of FD_ZERO() on these fd_sets is not permitted,
since that cannot be told about their abnormal lengths.

This introduces six wrapper functions for setting, clearing and testing
close-on-exec flags and fd-is-open flags:

	void __set_close_on_exec(int fd, struct fdtable *fdt);
	void __clear_close_on_exec(int fd, struct fdtable *fdt);
	bool close_on_exec(int fd, const struct fdtable *fdt);
	void __set_open_fd(int fd, struct fdtable *fdt);
	void __clear_open_fd(int fd, struct fdtable *fdt);
	bool fd_is_open(int fd, const struct fdtable *fdt);

Note that I've prepended '__' to the names of the set/clear functions because
they require the caller to hold a lock to use them.

Note also that I haven't added wrappers for looking behind the scenes at the
the array.  Possibly that should exist too.
Signed-off-by: David Howells <dhowells@redhat.com>
Link: http://lkml.kernel.org/r/20120216174942.23314.1364.stgit@warthog.procyon.org.ukSigned-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>

<asm/posix_types.h> includes a set of macros that operate on file
descriptors.  Way long ago those were exported to user space, but
nowadays they are #ifdef __KERNEL__.

However, they are nothing but standard (nonatomic) bit operations, and
we already have optimized versions of bit operations in the kernel.
We can't include <linux/bitops.h> in <asm/posix_types.h> but we can
move the definitions to <linux/time.h> and define them there in terms
of standard kernel bitops.

[ v2: folds the following fixes in:

  a) Stray space in __FD_SET(), reported by Andrew Morton
  b) #include <linux/string.h> needed for memset(), reported by Tony Luck ]
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: http://lkml.kernel.org/r/1328677745-20121-22-git-send-email-hpa@zytor.com
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>

Change the xtensa architecture to use <asm-generic/posix_types.h>.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: http://lkml.kernel.org/r/1328677745-20121-21-git-send-email-hpa@zytor.com
Cc: Chris Zankel <chris@zankel.net>

Change the x86 architecture to use <asm-generic/posix_types.h>.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Link: http://lkml.kernel.org/r/1328677745-20121-20-git-send-email-hpa@zytor.com
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>

Change the sparc architecture to use <asm-generic/posix_types.h>.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Acked-by: "David S. Miller" <davem@davemloft.net>
Link: http://lkml.kernel.org/r/1328677745-20121-19-git-send-email-hpa@zytor.com