We only need to call the pre/post transtion methods when we know the
power state is changing.  First, split up the pre/post transition
calls to be per-powerdomain, and then make them conditional on whether
the power domain is actually changing states.
Reviewed-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Tested-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Tested-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Kevin Hilman <khilman@ti.com>

Iteration over all power domains in the idle path is unnecessary since
only power domains that are transitioning need to be accounted for.
Also PRCM register accesses are known to be expensive, so the
additional latency added to the idle path is signficiant.

In order allow the pre/post transitions to be isolated and called
per-pwrdm, change the API so passing in a specific power domain will
trigger the pre/post transtion accounting for only that specific power
domain.  Passing NULL means iterating over all power domains as is
current behavior.
Reviewed-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Tested-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Tested-by: Grazvydas Ignotas <notasas@gmail.com>
Acked-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Kevin Hilman <khilman@ti.com>

Merge tag 'omap-devel-c-for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/pjw/omap-pending into devel-pm

Reimplement the OMAP PRCM I/O chain code.  Needed for I/O wakeups to
work correctly.

Conflicts:
	arch/arm/mach-omap2/prm2xxx_3xxx.c

Coult -> Could
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull KVM fixes from Avi Kivity:
 "Fixing a scheduling-while-atomic bug in the ppc code, and a bug which
  allowed pci bridges to be assigned to guests."

* git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: PPC: Book3S HV: Drop locks around call to kvmppc_pin_guest_page
  KVM: Fix PCI header check on device assignment

Pull InfiniBand/RDMA fixes from Roland Dreier:
 - Fixes to new ocrdma driver
 - Typo in test in CMA

* tag 'rdma-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  RDMA/cma: QP type check on received REQs should be AND not OR
  RDMA/ocrdma: Fix off by one in ocrdma_query_gid()
  RDMA/ocrdma: Fixed RQ error CQE polling
  RDMA/ocrdma: Correct queue SGE calculation
  RDMA/ocrdma: Correct reported max queue sizes
  RDMA/ocrdma: Fixed GID table for vlan and events

Pull ARM SoC fixes from Olof Johansson:
 "Nothing very controversial in here.  Most of the fixes are for OMAP
  this time around, with some orion/kirkwood and a tegra patch mixed in."

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: Orion: Fix Virtual/Physical mixup with watchdog
  ARM: Kirkwood: clk_register_gate_fn: add fn assignment
  ARM: Orion5x - Restore parts of io.h, with rework
  ARM: OMAP4: hwmod data: Force HDMI in no-idle while enabled
  ARM: OMAP2+: mux: fix sparse warning
  ARM: OMAP2+: CM: increase the module disable timeout
  ARM: OMAP4: clock data: add clockdomains for clocks used as main clocks
  ARM: OMAP4: hwmod data: fix 32k sync timer idle modes
  ARM: OMAP4+: hwmod: fix issue causing IPs not going back to Smart-Standby
  ARM: OMAP: Fix Beagleboard DVI reset gpio
  arm/dts: OMAP2: Fix interrupt controller binding
  ARM: OMAP2: Fix tusb6010 GPIO interrupt for n8x0
  ARM: OMAP2+: Fix MUSB ifdefs for platform init code
  ARM: tegra: make tegra_cpu_reset_handler_enable() __init
  ARM: OMAP: PM: Lock clocks list while generating summary
  ARM: iconnect: Remove include of removed linux/spi/orion_spi.h

Pull drm fixes from Dave Airlie:
 "Nothing major in here, one radeon SI fix for tiling, and one uninit
  var fix, two minor header file fixes."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm: drop comment about this header being autogenerated.
  drm/edid: don't return stack garbage from supports_rb
  vga_switcheroo: Add include guard
  drm/radeon: SI tiling fixes for display

The orion watchdog is expecting to be passed the physcial address of
the hardware, and will ioremap() it to give a virtual address it will
use as the base address for the hardware. However, when creating the
platform resource record, a virtual address was being used.

Add the necassary #define's so we can pass the physical address as
expected.

Tested on Kirkwood and Orion5x.

Cc: stable <stable@vger.kernel.org>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Olof Johansson <olof@lixom.net>

In commit:
    98d9986c ARM: Kirkwood: Replace clock gating
the kirkwood clock gating has been reworked. A custom variant of
clock gating, that calls a custom function before gating the clock
off, has been introduced. However in clk_register_gate_fn() this
custom function "fn" is never assigned.

This patch adds the missing fn assignment.

Cc: stable <stable@vger.kernel.org>
Signed-off-by: Marc Kleine-Budde <mkl@blackshift.org>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Olof Johansson <olof@lixom.net>

Commit 4d5fc58d (ARM: remove bunch of
now unused mach/io.h files) removed the orion5x io.h. Unfortunately,
this is still needed for the definition of IO_SPACE_LIMIT which
overrides the default 64K. All Orion based systems have 1Mbyte of IO
space per PCI[e] bus, and try to request_resource() this size. Orion5x
has two such PCI buses.

It is likely that the original, removed version, was broken. This
version might be less broken. However, it has not been tested on
hardware with a PCI card, let alone hardware with a PCI card with IO
capabilities.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Rob Herring <rob.herring@calxeda.com>
Signed-off-by: Olof Johansson <olof@lixom.net>

Merge tag 'omap-fixes-a-for-3.5rc' of git://git.kernel.org/pub/scm/linux/kernel/git/pjw/omap-pending into fixes

From Paul Walmsley (as per Tony Lindgren's request):
 "Some uncontroversial OMAP clock, hwmod, and compiler warning fixes for 3.5-rc"

* tag 'omap-fixes-a-for-3.5rc' of git://git.kernel.org/pub/scm/linux/kernel/git/pjw/omap-pending:
  ARM: OMAP4: hwmod data: Force HDMI in no-idle while enabled
  ARM: OMAP2+: mux: fix sparse warning
  ARM: OMAP2+: CM: increase the module disable timeout
  ARM: OMAP4: clock data: add clockdomains for clocks used as main clocks
  ARM: OMAP4: hwmod data: fix 32k sync timer idle modes
  ARM: OMAP4+: hwmod: fix issue causing IPs not going back to Smart-Standby
  ARM: OMAP: PM: Lock clocks list while generating summary

Merge tag 'omap-fixes-for-v3.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap into fixes

From Tony Lindgren:
"Here are a few fixes with the biggest one being fix for Beagle DVI
 reset. All of them are regression fixes, except for the missing omap2
 interrupt controller binding that somehow got missed earlier."

* tag 'omap-fixes-for-v3.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap:
  ARM: OMAP: Fix Beagleboard DVI reset gpio
  arm/dts: OMAP2: Fix interrupt controller binding
  ARM: OMAP2: Fix tusb6010 GPIO interrupt for n8x0
  ARM: OMAP2+: Fix MUSB ifdefs for platform init code

Pull Ceph fixes from Sage Weil:
 "There are a couple of fixes from Yan for bad pointer dereferences in
  the messenger code and when fiddling with page->private after page
  migration, a fix from Alex for a use-after-free in the osd client
  code, and a couple fixes for the message refcounting and shutdown
  ordering."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  libceph: flush msgr queue during mon_client shutdown
  rbd: Clear ceph_msg->bio_iter for retransmitted message
  libceph: use con get/put ops from osd_client
  libceph: osd_client: don't drop reply reference too early
  ceph: check PG_Private flag before accessing page->private

Pull XFS fixes from Ben Myers:
 - Fix stale data exposure with unwritten extents
 - Fix a warning in xfs_alloc_vextent with ODEBUG
 - Fix overallocation and alignment of pages for xfs_bufs
 - Fix a cursor leak
 - Fix a log hang
 - Fix a crash related to xfs_sync_worker
 - Rename xfs log structure from struct log to struct xlog so we can use
   crash dumps effectively

* tag 'for-linus-Jun-21-2012' of git://oss.sgi.com/xfs/xfs:
  xfs: rename log structure to xlog
  xfs: shutdown xfs_sync_worker before the log
  xfs: Fix overallocation in xfs_buf_allocate_memory()
  xfs: fix allocbt cursor leak in xfs_alloc_ag_vextent_near
  xfs: check for stale inode before acquiring iflock on push
  xfs: fix debug_object WARN at xfs_alloc_vextent()
  xfs: xfs_vm_writepage clear iomap_valid when !buffer_uptodate (REV2)

Pull perf updates from Ingo Molnar.

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  ftrace: Make all inline tags also include notrace
  perf: Use css_tryget() to avoid propping up css refcount
  perf tools: Fix synthesizing tracepoint names from the perf.data headers
  perf stat: Fix default output file
  perf tools: Fix endianity swapping for adds_features bitmask

As IO Daisy chain sequence is triggered via hwmod mux, there is no need to
control it from cpuidle path for OMAP3.

Also as omap3_disable_io_chain is no longer being used, just remove the
function.
Signed-off-by: Vishwanath BS <vishwanath.bs@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Reviewed-by: Rajendra Nayak <rnayak@ti.com>
Signed-off-by: Paul Walmsley <paul@pwsan.com>

IO Daisychain feature has to be triggered whenever there is a change in
device's mux configuration (See section 3.9.4 in OMAP4 Public TRM vP).

Now devices can idle independent of the powerdomain, there can be a
window where device is idled and corresponding powerdomain can be
ON/INACTIVE state. In such situations, since both module wake up is
enabled at padlevel as well as io daisychain sequence is triggered,
there will be 2 PRCM interrupts (Module async wake up via swakeup and
IO Pad interrupt). But as PRCM Interrupt handler clears the Module
Padlevel WKST bit in the first interrupt, module specific interrupt
handler will not triggered for the second time

Also look at detailed explanation given by Rajendra at
http://www.spinics.net/lists/linux-serial/msg04480.htmlSigned-off-by: Vishwanath BS <vishwanath.bs@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Reviewed-by: Rajendra Nayak <rnayak@ti.com>
[paul@pwsan.com: remove dependency on pm.c & pm.h; add kerneldoc]
Signed-off-by: Paul Walmsley <paul@pwsan.com>

Enable IO Wake up for OMAP3/4 as part of PRM Init. Currently this has been
managed in cpuidle path which is not the right place. Subsequent patch
will remove IO Daisy chain handling in cpuidle path once daisy chain is
handled as part of hwmod mux. This patch also moves the OMAP4 IO wakeup
enable code from the trigger function to init time setup.
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Reviewed-by: Rajendra Nayak <rnayak@ti.com>
[paul@pwsan.com: harmonize function names with other PRM functions; add
 kerneldoc; resolve checkpatch warnings]
Signed-off-by: Paul Walmsley <paul@pwsan.com>

IO daisychain is a mechanism that allows individual IO pads to generate
wakeup events on their own based on a switch of an input signal level.
This allows the hardware module behind the pad to be powered down, but
still have device level capability to detect IO events, and once this
happens the module can be powered back up to resume IO. See section
3.9.4 in OMAP4430 Public TRM for details.
Signed-off-by: Rajendra Nayak <rnayak@ti.com>
Signed-off-by: Vishwanath BS <vishwanath.bs@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
[paul@pwsan.com: use the shared MAX_IOPAD_LATCH_TIME declaration; renamed
 omap4_trigger_io_chain() to conform to other PRM function names;
 added kerneldoc; resolved checkpatch warnings]
Signed-off-by: Paul Walmsley <paul@pwsan.com>

Since IO Daisychain modifies only PRM registers, it makes sense to move
it to PRM File. Also changed the timeout value for IO chain enable to
100us and added a wait for status disable at the end.

Thanks to Nishanth Menon <nm@ti.com> for contributing a fix to the
timeout code waiting for WUCLKOUT to go high.
Signed-off-by: Vishwanath BS <vishwanath.bs@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Cc: Nishanth Menon <nm@ti.com>
Reviewed-by: Rajendra Nayak <rnayak@ti.com>
[paul@pwsan.com: renamed omap3_trigger_io_chain() to better describe the
 end result and to match other PRM functions; removed
 omap3_disable_io_chain(); moved MAX_IOPAD_LATCH_TIME to prcm-common as it
 will also be used by the OMAP4 code; removed unnecessary barrier;
 added kerneldoc; added credit for fix from Nishanth]
Signed-off-by: Paul Walmsley <paul@pwsan.com>

Currently the enabling and disabling of IO Daisy chain is not
according to the TRM. The below steps are followed to enable/
disable the IO chain, based loosely on the "Sec 3.5.7.2.2
I/O Wake-Up Mechanism" section in OMAP3630 Public TRM[1].

Steps to enable IO chain:
[a] Set PM_WKEN_WKUP.EN_IO bit
[b] Set the PM_WKEN_WKUP.EN_IO_CHAIN bit
[c] Poll for PM_WKST_WKUP.ST_IO_CHAIN.
[d] When ST_IO_CHAIN bit set to 1, clear PM_WKEN_WKUP.EN_IO_CHAIN
[e] Clear ST_IO_CHAIN bit.

Steps to disable IO chain:
[a] Clear PM_WKEN_WKUP.EN_IO_CHAIN bit
[b] Clear PM_WKEN_WKUP.EN_IO bit
[c] Clear PM_WKST_WKUP.ST_IO bit by writing 1 to it.

Step [e] & [c] in each case can be skipped, as these are handled
by the PRCM interrupt handler later.

[1] http://focus.ti.com/pdfs/wtbu/OMAP36xx_ES1.x_PUBLIC_TRM_vV.zipSigned-off-by: Mohan V <mohanv@ti.com>
Signed-off-by: Vishwanath BS <vishwanath.bs@ti.com>
[paul@pwsan.com: modified commit message to clarify that these steps are
 based loosely on the TRM section, rather than documented exactly]
Reviewed-by: Rajendra Nayak <rnayak@ti.com>
[paul@pwsan.com: resolved new warnings from checkpatch]
Signed-off-by: Paul Walmsley <paul@pwsan.com>

For OMAP4 only builds, the omap2_prm_* functions have dummy wrappers
to detect incorrect usage.  However, several unrelated omap3 PRM
functions have made it inside the #else clause of the #ifdef wrapping
the omap2_prm stubs, causing them to disappear on OMAP4-only builds.

This was unnoticed until the IO chain support was added and introduced
a new function in this section which is referenced by omap_hwmod.c:

arch/arm/mach-omap2/omap_hwmod.c: In function '_reconfigure_io_chain':
arch/arm/mach-omap2/omap_hwmod.c:1665:3: error: implicit declaration of function 'omap3xxx_prm_reconfigure_io_chain' [-Werror=implicit-function-declaration]

Fix by using the #ifdef to only wrap the omap2_prm functions that
need stubs on OMAP4-only builds.

Cc: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Kevin Hilman <khilman@ti.com>
[paul@pwsan.com: fixed checkpatch warnings for patch description]
Signed-off-by: Paul Walmsley <paul@pwsan.com>

This comment is well out of date.
Signed-off-by: Dave Airlie <airlied@redhat.com>

As per the OMAP4 documentation, audio over HDMI must be transmitted in
no-idle mode. This patch adds the HWMOD_SWSUP_SIDLE so that omap_hwmod uses
no-idle/force-idle settings instead of smart-idle mode.

This is required as the DSS interface clock is used as functional clock
for the HDMI wrapper audio FIFO. If no-idle mode is not used, audio could
be choppy, have bad quality or not be audible at all.
Signed-off-by: Ricardo Neri <ricardo.neri@ti.com>
[b-cousson@ti.com: Update the subject and align the .flags
location with the script template]
Signed-off-by: Benoit Cousson <b-cousson@ti.com>
Signed-off-by: Paul Walmsley <paul@pwsan.com>

Commit bbd707ac ("ARM: omap2: use
machine specific hook for late init") resulted in the addition of this
sparse warning:

arch/arm/mach-omap2/mux.c:791:12: warning: symbol 'omap_mux_late_init' was not declared. Should it be static?

Fix by including the header file containing the prototype.
Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Shawn Guo <shawn.guo@linaro.org>
Cc: Tony Lindgren <tony@atomide.com>

Increase the timeout for disabling an IP block to five milliseconds.
This is to handle the usb_host_fs idle latency, which takes almost
four milliseconds after a host controller reset.

This is the second of two patches needed to resolve the following
boot warning:

omap_hwmod: usb_host_fs: _wait_target_disable failed

Thanks to Sergei Shtylyov <sshtylyov@mvista.com> for finding
an unrelated hunk in a previous version of this patch.
Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Sergei Shtylyov <sshtylyov@mvista.com>
Cc: Tero Kristo <t-kristo@ti.com>

Until the OMAP4 code is converted to disable the use of the clock
framework-based clockdomain enable/disable sequence, any clock used as
a hwmod main_clk must have a clockdomain associated with it.  This
patch populates some clock structure clockdomain names to resolve the
following warnings during kernel init:

omap_hwmod: dpll_mpu_m2_ck: missing clockdomain for dpll_mpu_m2_ck.
omap_hwmod: trace_clk_div_ck: missing clockdomain for trace_clk_div_ck.
omap_hwmod: l3_div_ck: missing clockdomain for l3_div_ck.
omap_hwmod: ddrphy_ck: missing clockdomain for ddrphy_ck.
Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Rajendra Nayak <rnayak@ti.com>
Cc: Benoît Cousson <b-cousson@ti.com>

The 32k sync timer IP block target idle modes in the hwmod data are
incorrect.  The IP block does not support any smart-idle modes.
Update the data to reflect the correct modes.

This problem was initially identified and a diff fragment posted to
the lists by Benoît Cousson <b-cousson@ti.com>.  A patch description
bug in the first version was also identified by Benoît.
Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>
Cc: Tero Kristo <t-kristo@ti.com>

If an IP is configured in Smart-Standby-Wakeup, when disabling wakeup feature the
IP will not go back to Smart-Standby, but will remain in Smart-Standby-Wakeup.
Signed-off-by: Djamil Elaidi <d-elaidi@ti.com>
Signed-off-by: Paul Walmsley <paul@pwsan.com>

Pull NFS client bugfixes from Trond Myklebust:
 - Fix a write hang due to an uninitalised variable when
   !defined(CONFIG_NFS_V4)
 - Address upcall races in the legacy NFSv4 idmapper
 - Remove an O_DIRECT refcounting issue
 - Fix a pNFS refcounting bug when the file layout metadata server is
   also acting as a data server
 - Fix a pNFS module loading race.

* tag 'nfs-for-3.5-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFS: Force the legacy idmapper to be single threaded
  NFS: Initialise commit_info.rpc_out when !defined(CONFIG_NFS_V4)
  NFS: Fix a refcounting issue in O_DIRECT
  NFSv4.1: Fix a race in set_pnfs_layoutdriver
  NFSv4.1: Fix umount when filelayout DS is also the MDS

Pull btrfs fixes from Chris Mason:
 "This is a small pull with btrfs fixes.  The biggest of the bunch is
  another fix for the new backref walking code.

  We're still hammering out one btrfs dio vs buffered reads problem, but
  that one will have to wait for the next rc."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: delay iput with async extents
  Btrfs: add a missing spin_lock
  Btrfs: don't assume to be on the correct extent in add_all_parents
  Btrfs: introduce btrfs_next_old_item

Pull hwmon fixes from Guenter Roeck:
 "Two minor fixes in emc2103 and applesmc drivers."

* tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (emc2103) Fix use of an uninitilized variable in error case
  hwmon: (applesmc) Limit key length in warning messages

Rename the XFS log structure to xlog to help crash distinquish it from the
other logs in Linux.
Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>

Revert commit 1307bbd2, which uses the s_umount semaphore to provide
exclusion between xfs_sync_worker and unmount, in favor of shutting down
the sync worker before freeing the log in xfs_log_unmount.  This is a
cleaner way of resolving the race between xfs_sync_worker and unmount
than using s_umount.
Signed-off-by: Ben Myers <bpm@sgi.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>

Commit de1cbee4 which removed b_file_offset in favor of b_bn introduced a bug
causing xfs_buf_allocate_memory() to overestimate the number of necessary
pages. The problem is that xfs_buf_alloc() sets b_bn to -1 and thus effectively
every buffer is straddling a page boundary which causes
xfs_buf_allocate_memory() to allocate two pages and use vmalloc() for access
which is unnecessary.

Dave says xfs_buf_alloc() doesn't need to set b_bn to -1 anymore since the
buffer is inserted into the cache only after being fully initialized now.
So just make xfs_buf_alloc() fill in proper block number from the beginning.

CC: David Chinner <dchinner@redhat.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>

When we fail to find an matching extent near the requested extent
specification during a left-right distance search in
xfs_alloc_ag_vextent_near, we fail to free the original cursor that
we used to look up the XFS_BTNUM_CNT tree and hence leak it.
Reported-by: Chris J Arges <chris.j.arges@canonical.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>

An inode in the AIL can be flush locked and marked stale if
a cluster free transaction occurs at the right time. The
inode item is then marked as flushing, which causes xfsaild
to spin and leaves the filesystem stalled. This is
reproduced by running xfstests 273 in a loop for an
extended period of time.

Check for stale inodes before the flush lock. This marks
the inode as pinned, leads to a log flush and allows the
filesystem to proceed.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>