Add huge page support to ARM64, different huge page sizes are
supported depending on the size of normal pages:

PAGE_SIZE is 4KB:
   2MB - (pmds) these can be allocated at any time.
1024MB - (puds) usually allocated on bootup with the command line
         with something like: hugepagesz=1G hugepages=6

PAGE_SIZE is 64KB:
 512MB - (pmds) usually allocated on bootup via command line.

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>

Under ARM64, PTEs can be broadly categorised as follows:
   - Present and valid: Bit #0 is set. The PTE is valid and memory
     access to the region may fault.

   - Present and invalid: Bit #0 is clear and bit #1 is set.
     Represents present memory with PROT_NONE protection. The PTE
     is an invalid entry, and the user fault handler will raise a
     SIGSEGV.

   - Not present (file or swap): Bits #0 and #1 are clear.
     Memory represented has been paged out. The PTE is an invalid
     entry, and the fault handler will try and re-populate the
     memory where necessary.

Huge PTEs are block descriptors that have bit #1 clear. If we wish
to represent PROT_NONE huge PTEs we then run into a problem as
there is no way to distinguish between regular and huge PTEs if we
set bit #1.

To resolve this ambiguity this patch moves PTE_PROT_NONE from
bit #1 to bit #2 and moves PTE_FILE from bit #2 to bit #3. The
number of swap/file bits is reduced by 1 as a consequence, leaving
60 bits for file and swap entries.

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>

If we consider the following code sequence:

	my_pte = pte_modify(entry, myprot);
	x = pte_write(my_pte);
	y = pte_exec(my_pte);

If myprot comes from a PROT_NONE page, then x and y will both be
true which is undesireable behaviour.

This patch sets the no-execute and read-only bits for PAGE_NONE
such that the code above will return false for both x and y.

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>

In paging_init the memblock limit is set to restrict any addresses
returned by early_alloc to fit within the initial direct kernel
mapping in swapper_pg_dir. This allows map_mem to allocate puds,
pmds and ptes from the initial direct kernel mapping.

The limit stays low after paging_init() though, meaning any
bootmem allocations will be from a restricted subset of memory.
Gigabyte huge pages, for instance, are normally allocated from
bootmem as their order (18) is too large for the default buddy
allocator (MAX_ORDER = 11).

This patch restores the memblock limit when map_mem has finished,
allowing gigabyte huge pages (and other objects) to be allocated
from all of bootmem.

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>

All Transparent Huge Pages are allocated by the buddy allocator.

A compile time check is in place that fails when the order of a
transparent huge page is too large to be allocated by the buddy
allocator. Unfortunately that compile time check passes when:
HPAGE_PMD_ORDER == MAX_ORDER
( which is incorrect as the buddy allocator can only allocate
memory of order strictly less than MAX_ORDER. )

This patch updates the compile time check to fail in the above
case.

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>

huge_pte_alloc, huge_pte_offset and follow_huge_p[mu]d have
already been copied over to mm.

This patch removes the x86 copies of these functions and activates
the general ones by enabling:
CONFIG_ARCH_WANT_GENERAL_HUGETLB

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>

The huge_pte_alloc, huge_pte_offset and follow_huge_p[mu]d
functions in x86/mm/hugetlbpage.c do not rely on any architecture
specific knowledge other than the fact that pmds and puds can be
treated as huge ptes.

To allow other architectures to use this code (and reduce the need
for code duplication), this patch copies these functions into mm,
replaces the use of pud_large with pud_huge and provides a config
flag to activate them:
CONFIG_ARCH_WANT_GENERAL_HUGETLB

If CONFIG_ARCH_WANT_HUGE_PMD_SHARE is also active then the
huge_pmd_share code will be called by huge_pte_alloc (othewise we
call pmd_alloc and skip the sharing code).

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>

The huge_pmd_share code has been copied over to mm/hugetlb.c to
make it accessible to other architectures.

Remove the x86 copy of the huge_pmd_share code and enable the
ARCH_WANT_HUGE_PMD_SHARE config flag. That way we reference the
general one.

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>

Under x86, multiple puds can be made to reference the same bank of
huge pmds provided that they represent a full PUD_SIZE of shared
huge memory that is aligned to a PUD_SIZE boundary.

The code to share pmds does not require any architecture specific
knowledge other than the fact that pmds can be indexed, thus can
be beneficial to some other architectures.

This patch copies the huge pmd sharing (and unsharing) logic from
x86/ to mm/ and introduces a new config option to activate it:
CONFIG_ARCH_WANTS_HUGE_PMD_SHARE

Signed-off-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>

Pull parisc fixes from Helge Deller:
 "This patcheset includes fixes for:

   - the PCI/LBA which brings back the stifb graphics framebuffer
     console
   - possible memory overflows in parisc kernel init code
   - parport support on older GSC machines
   - avoids that users by mistake enable PARPORT_PC_SUPERIO on parisc
   - MAINTAINERS file list updates for parisc."

* 'for-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: parport0: fix this legacy no-device port driver!
  parport_pc: disable PARPORT_PC_SUPERIO on parisc architecture
  parisc/PCI: lba: fix: convert to pci_create_root_bus() for correct root bus resources (v2)
  parisc/PCI: Set type for LBA bus_num resource
  MAINTAINERS: update parisc architecture file list
  parisc: kernel: using strlcpy() instead of strcpy()
  parisc: rename "CONFIG_PA7100" to "CONFIG_PA7000"
  parisc: fix kernel BUG at arch/parisc/include/asm/mmzone.h:50
  parisc: memory overflow, 'name' length is too short for using

Fix the above kernel error from parport_announce_port() on 32bit GSC
machines (e.g. B160L). The parport driver requires now a pointer to the
device struct.

Signed-off-by: Helge Deller <deller@gmx.de>

If enabled, CONFIG_PARPORT_PC_SUPERIO scans on PC-like hardware for
various super-io chips by accessing i/o ports in a range which will
crash any parisc hardware at once.

In addition, parisc has it's own incompatible superio chip
(CONFIG_SUPERIO), so if we disable PARPORT_PC_SUPERIO completely for
parisc we can avoid that people by accident enable the parport_pc
superio option too.

Signed-off-by: Helge Deller <deller@gmx.de>

commit dc7dce28


Author: Bjorn Helgaas <bhelgaas@google.com>
Date:   Fri Oct 28 16:27:27 2011 -0600
   parisc/PCI: lba: convert to pci_create_root_bus() for correct root bus
                    resources

  Supply root bus resources to pci_create_root_bus() so they're correct
  immediately.  This fixes the problem of "early" and "header" quirks seeing
  incorrect root bus resources.

added tests for elmmio_space.start while it should use
elmmio_space.flags.  This for example led to incorrect resource
assignments and a non-working stifb framebuffer on most parisc machines.

LBA 10:1: PCI host bridge to bus 0000:01
pci_bus 0000:01: root bus resource [io  0x12000-0x13fff] (bus address [0x2000-0x3fff])
pci_bus 0000:01: root bus resource [mem 0xfffffffffa000000-0xfffffffffbffffff] (bus address [0xfa000000-0xfbffffff])
pci_bus 0000:01: root bus resource [mem 0xfffffffff4800000-0xfffffffff4ffffff] (bus address [0xf4800000-0xf4ffffff])
pci_bus 0000:01: root bus resource [??? 0x00000001 flags 0x0]

Signed-off-by: Helge Deller <deller@gmx.de>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>

The non-PAT resource probing code failed to set the type of the LBA bus_num
resource (30aa80da

 "parisc/PCI: register busn_res for root buses" did
the corresponding thing for the PAT case).

This causes incorrect resource assignments and a non-working stifb
framebuffer on most parisc machines.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Helge Deller <deller@gmx.de>

Signed-off-by: Helge Deller <deller@gmx.de>

'boot_args' is an input args, and 'boot_command_line' has a fix length.
So use strlcpy() instead of strcpy() to avoid memory overflow.

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Acked-by: Kyle McMartin <kyle@mcmartin.ca>
Signed-off-by: Helge Deller <deller@gmx.de>

There's a Makefile line setting cflags for CONFIG_PA7100. But that
Kconfig macro doesn't exist. There is a Kconfig symbol PA7000, which
covers both PA7000 and PA7100 processors. So let's use the corresponding
Kconfig macro.

Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Helge Deller <deller@gmx.de>

With CONFIG_DISCONTIGMEM=y and multiple physical memory areas,
cat /proc/kpageflags triggers this kernel bug:

kernel BUG at arch/parisc/include/asm/mmzone.h:50!
CPU: 2 PID: 7848 Comm: cat Tainted: G      D W 3.10.0-rc3-64bit #44
 IAOQ[0]: kpageflags_read0x128/0x238
 IAOQ[1]: kpageflags_read0x12c/0x238
 RP(r2): proc_reg_read0xbc/0x130
Backtrace:
 [<00000000402ca2d4>] proc_reg_read0xbc/0x130
 [<0000000040235bcc>] vfs_read0xc4/0x1d0
 [<0000000040235f0c>] SyS_read0x94/0xf0
 [<0000000040105fc0>] syscall_exit0x0/0x14

kpageflags_read() walks through the whole memory, even if some memory
areas are physically not available. So, we should better not BUG on an
unavailable pfn in pfn_to_nid() but just return the expected value -1 or
0.

Signed-off-by: Helge Deller <deller@gmx.de>

'path.bc[i]' can be asigned by PCI_SLOT() which can '> 10', so sizeof(6
* "%u:" + "%u" + '\0') may be 21.

Since 'name' length is 20, it may be memory overflow.

And 'path.bc[i]' is 'unsigned char' for printing, we can be sure the
max length of 'name' must be less than 28.

So simplify thinking, we can use 28 instead of 20 directly, and do not
think of whether 'patchc.bc[i]' can '> 100'.

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Helge Deller <deller@gmx.de>

Pull powerpc fixes from Ben Herrenschmidt:
 "Here are a few more fixes for powerpc 3.10.  It's a bit more than I
  would have liked this late in the game but I suppose that's what
  happens with a brand new chip generation coming out.

  A few regression fixes, some last minute fixes for new P8 features
  such as transactional memory,...

  There's also one powerpc KVM patch that I requested that adds two
  missing functions to our in-kernel interrupt controller support which
  is itself a new 3.10 feature.  These are defined by the base
  hypervisor specification.  We didn't implement them originally because
  Linux doesn't use them but they are simple and I'm not comfortable
  having a half-implemented interface in 3.10 and having to deal with
  versionning etc...  later when something starts needing those calls.
  They cannot be emulated in qemu when using in-kernel interrupt
  controller (not enough shared state).

  Just added a last minute patch to fix a typo introducing a breakage in
  our cputable for Power7+ processors, sorry about that, but the
  regression it fixes just hurt me :-)"

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc/cputable: Fix typo on P7+ cputable entry
  powerpc/perf: Add missing SIER support
  powerpc/perf: Revert to original NO_SIPR logic
  powerpc/pci: Remove the unused variables in pci_process_bridge_OF_ranges
  powerpc/pci: Remove the stale comments of pci_process_bridge_OF_ranges
  powerpc/pseries: Always enable CONFIG_HOTPLUG_CPU on PSERIES SMP
  powerpc/kvm/book3s: Add support for H_IPOLL and H_XIRR_X in XICS emulation
  powerpc/32bit:Store temporary result in r0 instead of r8
  powerpc/mm: Always invalidate tlb on hpte invalidate and update
  powerpc/pseries: Improve stream generation comments in copypage/user
  powerpc/pseries: Kill all prefetch streams on context switch
  powerpc/cputable: Fix oprofile_cpu_type on power8
  powerpc/mpic: Fix irq distribution problem when MPIC_SINGLE_DEST_CPU
  powerpc/tm: Fix userspace stack corruption on signal delivery for active transactions
  powerpc/tm: Move TM abort cause codes to uapi
  powerpc/tm: Abort on emulation and alignment faults
  powerpc/tm: Update cause codes documentation
  powerpc/tm: Make room for hypervisor in abort cause codes

Pull scsi target fixes from Nicholas Bellinger:
 "The highlights include:

   - Re-instate sess->wait_list in target_wait_for_sess_cmds() for
     active I/O shutdown handling in fabrics using se_cmd->cmd_kref
   - Make ib_srpt call target_sess_cmd_list_set_waiting() during session
     shutdown
   - Fix FILEIO off-by-one READ_CAPACITY bug for !S_ISBLK export
   - Fix iscsi-target login error heap buffer overflow (Kees)
   - Fix iscsi-target active I/O shutdown handling regression in
     v3.10-rc1

  A big thanks to Kees Cook for fixing a long standing login error
  buffer overflow bug.

  All patches are CC'ed to stable with the exception of the v3.10-rc1
  specific regression + other minor target cleanup."

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  iscsi-target: Fix iscsit_free_cmd() se_cmd->cmd_kref shutdown handling
  target: Propigate up ->cmd_kref put return via transport_generic_free_cmd
  iscsi-target: fix heap buffer overflow on error
  target/file: Fix off-by-one READ_CAPACITY bug for !S_ISBLK export
  ib_srpt: Call target_sess_cmd_list_set_waiting during shutdown_session
  target: Re-instate sess_wait_list for target_wait_for_sess_cmds
  target: Remove unused wait_for_tasks bit in target_wait_for_sess_cmds

Pull clock subsystem fixes from Mike Turquette:
 "A mix of small fixes affecting mostly ARM platforms as well as a
  discrete clock expander chip.  Most fixes are corrections to lousy
  clock data of one form or another."

* tag 'clk-fixes-for-linus' of git://git.linaro.org/people/mturquette/linux:
  clk: mxs: Include clk mxs header file
  clk: vt8500: Fix unbalanced spinlock in vt8500_dclk_set_rate()
  clk: si5351: Set initial clkout rate when defined in platform data.
  clk: si5351: Fix clkout rate computation.
  clk: samsung: Add CLK_IGNORE_UNUSED flag for the sysreg clocks
  clk: ux500: clk-sysctrl: handle clocks with no parents
  clk: ux500: Provide device enumeration number suffix for SMSC911x

Merge tag 'fbdev-for-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/plagnioj/linux-fbdev

Pull fbdev fixes from Jean-Christophe PLAGNIOL-VILLARD:
 "This contains some small fixes

   - Atmel LCDC: fix blank the backlight on remove
   - ps3fb: fix compile warning
   - OMAPDSS: Fix crash with DT boot"

* tag 'fbdev-for-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/plagnioj/linux-fbdev:
  atmel_lcdfb: blank the backlight on remove
  trivial: atmel_lcdfb: add missing error message
  OMAPDSS: Fix crash with DT boot
  fbdev/ps3fb: fix compile warning

Pull assorted fixes from Al Viro:
 "There'll be more - I'm trying to dig out from under the pile of mail
  (a couple of weeks of something flu-like ;-/) and there's several more
  things waiting for review; this is just the obvious stuff."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  zoran: racy refcount handling in vm_ops ->open()/->close()
  befs_readdir(): do not increment ->f_pos if filldir tells us to stop
  hpfs: deadlock and race in directory lseek()
  qnx6: qnx6_readdir() has a braino in pos calculation
  fix buffer leak after "scsi: saner replacements for ->proc_info()"
  vfs: Fix invalid ida_remove() call

Pull two NFS client fixes from Trond Myklebust:
 - Fix a regression that broke NFS mounting using klibc and busybox
 - Stable fix to check access modes correctly on NFSv4 delegated open()

* tag 'nfs-for-3.10-4' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFS: Fix security flavor negotiation with legacy binary mounts
  NFSv4: Fix a thinko in nfs4_try_open_cached

Fix a typo in setting COMMON_USER2_POWER7 bits to .cpu_user_features2
cpu specs table.

Signed-off-by: Will Schmidt <will_schmidt@vnet.ibm.com>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Commit 8f61aa32

 "Add support for SIER" missed updates to siar_valid()
and perf_get_data_addr().

In both cases we need to check the SIER instead of mmcra.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

This is a revert and then some of commit 860aad71

 "Add regs_no_sipr()".
This workaround was only needed on early chip versions.

As before NO_SIPR becomes a static flag of the PMU struct.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

The codes which ever used these two variables have gone. Throw away
them too.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

These comments already don't apply to the current code. So just remove
them.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Adam Lackorzynski reported the following build failure on
!CONFIG_HOTPLUG_CPU configuration:

  CC      arch/powerpc/kernel/rtas.o
arch/powerpc/kernel/rtas.c: In function ‘rtas_cpu_state_change_mask’:
arch/powerpc/kernel/rtas.c:843:4: error: implicit declaration of function ‘cpu_down’ [-Werror=implicit-function-declaration]
cc1: all warnings being treated as errors
make[1]: *** [arch/powerpc/kernel/rtas.o] Error 1
make: *** [arch/powerpc/kernel] Error 2

The build fails because cpu_down() is defined only under CONFIG_HOTPLUG_CPU.

Looking further, the mobility code in pseries is one of the call-sites which
uses rtas_ibm_suspend_me(), which in turn calls rtas_cpu_state_change_mask().
And the mobility code is unconditionally compiled-in (it does not fall under
any Kconfig option). And commit 120496ac

 (powerpc: Bring all threads online
prior to migration/hibernation) which introduced this build regression is
critical for the proper functioning of the migration code. So it appears
that the only solution to this problem is to enable CONFIG_HOTPLUG_CPU if
SMP is enabled on PPC_PSERIES platforms. So make that change in the Kconfig.

Reported-by: Adam Lackorzynski <adam@os.inf.tu-dresden.de>
Cc: stable@vger.kernel.org
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

This adds the remaining two hypercalls defined by PAPR for manipulating
the XICS interrupt controller, H_IPOLL and H_XIRR_X.  H_IPOLL returns
information about the priority and pending interrupts for a virtual
cpu, without changing any state.  H_XIRR_X is like H_XIRR in that it
reads and acknowledges the highest-priority pending interrupt, but it
also returns the timestamp (timebase register value) from when the
interrupt was first received by the hypervisor.  Currently we just
return the current time, since we don't do any software queueing of
virtual interrupts inside the XICS emulation code.

These hcalls are not currently used by Linux guests, but may be in
future.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Commit a9c4e541


"powerpc/kprobe: Complete kprobe and migrate exception frame"
introduced a regression:

While returning from exception handling in case of PREEMPT enabled,
_TIF_NEED_RESCHED bit is checked in TI_FLAGS (thread_info flag) of current
task. Only if this bit is set, it should continue with the process of
calling preempt_schedule_irq() to schedule highest priority task if
available.

Current code assumes that r8 contains TI_FLAGS and check this for
_TIF_NEED_RESCHED, but as r8 is modified in the code which executes before
this check, r8 no longer contains the expected TI_FLAGS information.

As a result check for comparison with _TIF_NEED_RESCHED was failing even if
NEED_RESCHED bit is set in the current thread_info flag. Due to this,
preempt_schedule_irq() and in turn scheduler was not getting called even if
highest priority task is ready for execution.

So, store temporary results in r0 instead of r8 to prevent r8 from getting
modified as subsequent code is dependent on its value.

Signed-off-by: Priyanka Jain <Priyanka.Jain@freescale.com>
CC: <stable@vger.kernel.org> [v3.7+]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

If a hash bucket gets full, we "evict" a more/less random entry from it.
When we do that we don't invalidate the TLB (hpte_remove) because we assume
the old translation is still technically "valid". This implies that when
we are invalidating or updating pte, even if HPTE entry is not valid
we should do a tlb invalidate.

This was a regression introduced by b1022fbd



Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

No code changes, just documenting what's happening a little better.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

On context switch, we should have no prefetch streams leak from one
userspace process to another.  This frees up prefetch resources for the
next process.

Based on patch from Milton Miller.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Maynard informed me that neither the oprofile kernel module nor oprofile
userspace has been updated to support that "legacy" oprofile module
interface for power8, which is indicated by "ppc64/power8." This results
in no samples. The solution is to default to the "timer" type, instead.
The raw entry also should be updated, as "ppc64/ibm-compat-v1" indicates
to oprofile userspace to use "compatibility events" which are obsolete
in ISA 2.07.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

For the mpic with a flag MPIC_SINGLE_DEST_CPU, only one bit should be
set in interrupt destination registers.

The code is applicable to 64-bit platforms as well as 32-bit.

Signed-off-by: Zhao Chenhui <chenhui.zhao@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

When in an active transaction that takes a signal, we need to be careful with
the stack.  It's possible that the stack has moved back up after the tbegin.
The obvious case here is when the tbegin is called inside a function that
returns before a tend.  In this case, the stack is part of the checkpointed
transactional memory state.  If we write over this non transactionally or in
suspend, we are in trouble because if we get a tm abort, the program counter
and stack pointer will be back at the tbegin but our in memory stack won't be
valid anymore.

To avoid this, when taking a signal in an active transaction, we need to use
the stack pointer from the checkpointed state, rather than the speculated
state.  This ensures that the signal context (written tm suspended) will be
written below the stack required for the rollback.  The transaction is aborted
becuase of the treclaim, so any memory written between the tbegin and the
signal will be rolled back anyway.

For signals taken in non-TM or suspended mode, we use the
normal/non-checkpointed stack pointer.

Tested with 64 and 32 bit signals

Signed-off-by: Michael Neuling <mikey@neuling.org>
Cc: <stable@vger.kernel.org> # v3.9
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>