Register interconnect links between IP blocks, rather than the IP
blocks themselves.  (The IP blocks will be registered as a side-effect
of registering the links.)

The objective is to reduce the number of lines of static data and
facilitate the sharing of IP block data between different SoCs.  These
objectives come at the penalty of increased boot time due to increased
computation.

While here, fix a few whitespace problems and inaccurate variable names.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Add support for direct IP block interconnect ("link") registration to
the hwmod code via a new function, omap_hwmod_register_links().  This
will replace direct registration of hwmods, and a subsequent patch
will remove omap_hwmod_register().

This change will allow a subsequent patch to remove the hwmod data
link arrays.  This will reduce the size of the hwmod static data and
also make it easier to generate the data files.  It will also make it
possible to share some of the struct omap_hwmod records across
multiple SoCs, since the link array pointers will be removed from the
struct omap_hwmod.

The downside is that boot time will increase.  Minimizing boot time
was the reason why the link arrays were originally introduced.
Removing them will require extra computation during boot to allocate
memory and associate IP blocks with their interconnects.  However,
since the current kernel development focus is on reducing the number
of lines in arch/arm/mach-omap2/, boot time impact is now seemingly
considered a lower priority.

This patch contains additional complexity to reduce the number of
memory allocations required for this change.  This reduces the boot
time impact: total hwmod link registration time was ~ 2655
microseconds with a simple allocation strategy, but is now ~ 549
microseconds[1] with the approach taken by this patch.

1. Measured on a BeagleBoard 35xx @ 500MHz MPU/333 MHz CORE, average
   of 7 samples.  Total uncertainty is +/- 61 microseconds.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

An IP block's MPU interface port only needs to be found once.  The result
can be cached to speed further lookups.  This patch consolidates these
two steps into a single function.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

To reduce the number of lines of data in the OMAP portion of the Linux
code base, subsequent patches will remove the lists of hwmod
interconnect links from the static hwmod data.  These lists will be
built dynamically during boot.  To ease this transition, this patch
centralizes the way that interconnect links are iterated into a single
function, _fetch_next_ocp_if().
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Most IP blocks on the OMAP SoC have an interconnect link that is
intended to be used by the MPU to communicate with the IP block.
Several parts of the hwmod code need to be able to identify this link.
Currently, this is open-coded.  However, future patches will change
the way that interconnect links are represented and will make
identifying the link more complex.  So to avoid code duplication, this
patch centralizes the MPU port link identification code into a new
function, _find_mpu_rt_port().
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Extend the OCP_* register offsets in the struct
omap_hwmod_class_sysconfig to 32 bits.  This is required to add the
OMAP4+ GPU hwmod, which uses OCP_* register offsets larger than 16
bits.

Another possible solution may be to simply add a single 16 bit offset
field in this structure, and to add code to factor that offset into
all OCP_* register accesses.  This would save some memory, since
almost no modules need 32 bit offsets.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Some hwmods were commented out from the OMAP4 data, under the theory
that they shouldn't be added until drivers were ready.  But part of
the utility of the hwmod code is that it can reset and properly
initialize IP blocks that have no drivers associated with them.
Rather than commenting the links in the future hwmod data conversion
patches, discussing this with Benoit, it seems best to simply
uncomment them now.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

One of the OMAP4 links was missing OCP_USER flags, since it was only
used by the DSP initiator, and we did not have an OCP_USER_DSP flag.
Future patches will switch the hwmod code and data to register
interfaces, rather than hwmods, and it will be mandatory for all
interfaces to have at least one user bit set.  This patch resolves the
issue by adding OCP_USER_DSP and marking the DSP-IVA interface
appropriately.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Commit 407a6888 ("OMAP4: hwmod data:
Add AESS, McPDM, bandgap, counter_32k, MMC, KBD, ISS & IPU") adds a
hwmod for the bandgap die temperature sensor IP block.  This IP block
has no interconnect port or firewall region, nor does it have an
independent register space or OCP control registers.  Its registers
are embedded in the System Control Module (SCM) IP block.  So it
appears that the bandgap device should be created by the SCM driver.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

GPTIMER12 is attached to the L4 SEC interconnect, not directly to L4 WKUP.
Add the L4 SEC interconnect and attach GPTIMER12 to it.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

The OMAP3 hwmod data was missing a DSS->L3 interconnect link for the
OMAP3430 ES1 DSS hwmod.  Since the hwmod code and data is being modified
to register interfaces rather than hwmods, this would result in the DSS hwmod
not being registered correctly on OMAP3430ES1.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

Commit a52e2ab6 ("ARM: OMAP3: hwmod
data: disable multiblock reads on MMC1/2 on OMAP34xx/35xx <= ES2.1")
didn't link the MMC hwmods to the interconnects correctly.  Future
patches will register hwmods by interface, so if this is not fixed,
the MMC IP blocks won't be registered.  Update the interface data
records to point to the correct IP blocks.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

Some of the 2xxx and 3xxx hwmod data files use the old naming style
for hwmods, ending in a "_hwmod".  These names are used by the OMAP
integration code to map hwmods to platform_devices, so they need to be
consistent, or the platform_devices won't be created.  Remove the
_hwmod suffix to conform with the rest of the OMAP SoC data.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

arch/arm/mach-omap2/timer.c pokes around inside the hwmod data
structures.  Since the hwmod data structures are about to change, this
code will break.  This patch modifies the timer code to use
recently-added hwmod functions instead.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Tony Lindgren <tony@atomide.com>

The timer integration code pokes around in hwmod data structures.
Those data structures are about to change.  Define a function,
omap_hwmod_get_resource_byname(), for the timer integration code to
use instead.

The original patch has been changed to use struct resource by Tony's
request, although the caller of this function should not be a driver._
Platform drivers should get their data through the regular platform_*
functions; DT drivers through the appropriate of_* functions.  This a
function is only for use by OMAP core code in arch/arm/*omap*.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>
Cc: Tony Lindgren <tony@atomide.com>

A subsequent patch will need to know the struct omap_hwmod_addr_space
record corresponding to the module's register target, used by the MPU.
So, convert _find_mpu_rt_base() into _find_mpu_rt_addr_space().  Then
modify its sole current user, _populate_mpu_rt_base(), to extract the
MPU RT base address itself from the struct omap_hwmod_addr_space record.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Change the way that hardreset lines are handled by the hwmod code.
Hardreset lines are generally associated with initiator IP blocks.
Prior to this change, the hwmod code expected to control hardreset
lines itself, asserting them on shutdown and deasserting them upon
enable.  But driver authors inside TI have commented to us that their
drivers require direct control over these lines.  Unfortunately, these
drivers haven't been posted publicly yet, so it's hard to determine
exactly what is needed, a priori.  This change attempts to set forth
some reasonable semantics that should be an improvement over the
current code.

The semantics implemented by this patch are as follows:

- If the hwmod is not marked with HWMOD_INIT_NO_RESET, then assert all
  associated hardreset lines during IP block setup.  This is intended
  to place the IP blocks into a known state that will not interfere
  with other devices during kernel boot.

- IP blocks with hardreset lines will not be automatically enabled or
  idled during setup.  Instead, they will be left in the INITIALIZED
  state.

- When the hwmod code is asked to enable, idle, or shutdown an IP
  block with asserted hardreset lines, the hwmod code will do nothing.
  The driver integration code must do the remaining work needed to
  control these IP blocks.  Once this driver integration code is posted
  to the lists, hopefully we can consolidate it and move it inside the
  hwmod code.

Custom reset functions for IP blocks with hardreset lines still should
be supported and are strongly endorsed.  It is intended that every
subsystem with hardreset lines should have a custom reset function
that can place their subsystem into quiescent idle with the hardreset
lines deasserted.

This reverts most of commit 5365efbe
("OMAP: hwmod: Add hardreset management support").  Later code
reorganizations caused the sequencing of the code from this patch to
be changed, anyway.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Reorganize the code involved in resetting and configuring an IP block
to make it easier to read and maintain.  This involves improving
documentation, splitting some large functions up into smaller ones to
better conform with Documentation/CodingStyle, and removing some
unnecessary code.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Reorganize the code involved in initializing the internal data for
each hwmod to make it easier to read and maintain.  This involves
improving documentation and removing some duplicated and unnecessary
code.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Revise the IP block reset process.  This patch ensures that the
OCP_SYSCONFIG registers are reloaded after a custom reset.  Since
OCP_SYSCONFIG bits are cleared during reset, they should be
reprogrammed unless the IP block is being left in reset.  (The only IP
blocks that are left in reset are IP blocks with hardreset lines and
no custom reset function.)  If the IP block is left in reset, then it
is inaccessible to the MPU, and an access to the OCP_SYSCONFIG
register will cause an abort.

This version incorporates comments from Omar Ramirez Luna
<omar.ramirez@ti.com> to skip the OCP_SYSCONFIG access after asserting
hardreset lines.  This allows the MMU (IOMMU) IP block, which has
both hardreset lines and an OCP_SYSCONFIG register.

Also, ignore _ocp_softreset() errors if the IP block doesn't include a
softreset bit.  This is needed since a subsequent patch will start
taking the return value of the _reset() function seriously.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>
Cc: Omar Ramirez Luna <omar.ramirez@ti.com>

Remove the pseudo-hwmods associated with hardreset lines from the
OMAP4 data file.  Future patches will convert this data to register
hwmods by interfaces, rather than registering hwmods directly.  The
pseudo-hwmods aren't associated with any interfaces, so this will
create a problem.

After this change, the hwmod code will reset processor IPs at the
hwmod level, rather than by individual hardreset lines.  So, for
example, if the IVA device driver code wishes to place one of the
sequencer cores into reset, while leaving the other active, it must do
so itself by calling the appropriate PRM functions.

This patch will cause a change in the initialization behavior of the
DSP, IVA, and IPU.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

Parts of the hwmod code test to see if a module has one and only one
hardreset line before taking an action.  It seems more appropriate
to control all hardreset lines associated with a hwmod, not just one.

It so happens that with the current hwmod data, this patch will not
change any behavior, since hwmods with hardreset lines have only one
hardreset line associated with them, and 'pseudo-hwmods' are used to
handle the other hardreset lines.  But future hwmod data patches to
remove the pseudo-hwmods will change this.
Signed-off-by: NPaul Walmsley <paul@pwsan.com>
Cc: Benoît Cousson <b-cousson@ti.com>

The clocks for all DSS slave interfaces were recently changed to "dss_ick" on
OMAP2 and OMAP3, this clock can be autoidled by PRCM. The VENC interface
previously had "dss_54m_fck" as it's clock which couldn't be autoidled, and
hence the OCPIF_SWSUP_IDLE flag was needed.

Remove the OCPIF_SWSUP_IDLE flag from VENC interfaces as it's clock is
now "dss_ick".  This allows the PRCM hardware to autoidle the VENC
interface clocks when they are not active, rather than relying on the
software to do it, which can keep the interface clocks active
unnecessarily.
Signed-off-by: NArchit Taneja <archit@ti.com>
[paul@pwsan.com: add a short description of the fix to the commit log]
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

This reverts commit f9a2f9c3.  This
commit caused a regression in the I2C hwmod reset on OMAP2/3/4,
logging messages similar to these during boot:

[    0.200378] omap_hwmod: i2c1: softreset failed (waited 10000 usec)
[    0.222076] omap_hwmod: i2c2: softreset failed (waited 10000 usec)

While the original patch was intended to fix some reset-related timing
issues, it's believed that these problems were actually fixed by
commit 2800852a ("ARM: OMAP2+: hwmod:
Restore sysc after a reset"):

    http://marc.info/?l=linux-arm-kernel&m=133410322617245&w=2

Cc: Rajendra Nayak <rnayak@ti.com>
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

Due to HW limitation, some IPs should not be accessed just after a
softreset. Since the current hwmod sequence is accessing the sysconfig
register just after the reset, it might lead to OCP bus error in
that case.

Add a new field in the sysconfig structure to specify a delay in usecs
needed after doing a softreset.

In the case of the ISS and FDIF modules, the L3 OCP port will be
disconnected upon a SW reset. That issue was confirmed with HW simulation
and an errata should be available soon. The HW recommendation to avoid
that is to wait for 100 OCP clk cycles, before accessing the IP.

Considering the worse case (OPP50), the L3 bus will run at 100 MHz,
so a 1 usec delay is needed. Add an x2 margin to be safe.
Acked-by: NBenoit Cousson <b-cousson@ti.com>
Signed-off-by: NFernando Guzman Lugo <fernando.lugo@ti.com>
[paul@pwsan.com: dropped FDIF change for now since the hwmod data is not
 yet upstream; the FDIF change will need to be added later once the FDIF
 data is merged]
Signed-off-by: NPaul Walmsley <paul@pwsan.com>

Pull two more small regmap fixes from Mark Brown:
 - Now we have users for it that aren't running Android it turns out
   that regcache_sync_region() is much more useful to drivers if it's
   exported for use by modules.  Who knew?
 - Make sure we don't divide by zero when doing debugfs dumps of
   rbtrees, not visible up until now because everything was providing at
   least some cache on startup.

* tag 'regmap-3.4-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
  regmap: prevent division by zero in rbtree_show
  regmap: Export regcache_sync_region()

Pull a few KVM fixes from Avi Kivity:
 "A bunch of powerpc KVM fixes, a guest and a host RCU fix (unrelated),
  and a small build fix."

* 'kvm-updates/3.4' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Resolve RCU vs. async page fault problem
  KVM: VMX: vmx_set_cr0 expects kvm->srcu locked
  KVM: PMU: Fix integer constant is too large warning in kvm_pmu_set_msr()
  KVM: PPC: Book3S: PR: Fix preemption
  KVM: PPC: Save/Restore CR over vcpu_run
  KVM: PPC: Book3S HV: Save and restore CR in __kvmppc_vcore_entry
  KVM: PPC: Book3S HV: Fix kvm_alloc_linear in case where no linears exist
  KVM: PPC: Book3S: Compile fix for ppc32 in HIOR access code

Pull SuperH fixes from Paul Mundt.

* tag 'sh-for-linus' of git://github.com/pmundt/linux-sh:
  sh: fix clock-sh7757 for the latest sh_mobile_sdhi driver
  serial: sh-sci: use serial_port_in/out vs sci_in/out.
  sh: vsyscall: Fix up .eh_frame generation.
  sh: dma: Fix up device attribute mismatch from sysdev fallout.
  sh: dwarf unwinder depends on SHcompact.
  sh: fix up fallout from system.h disintegration.

Pull security layer fixlet from James Morris.

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  sysctl: fix write access to dmesg_restrict/kptr_restrict

Pull ACPI & Power Management patches from Len Brown:
 "Two fixes for cpuidle merge-window changes, plus a URL fix in
  MAINTAINERS"

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
  MAINTAINERS: Update git url for ACPI
  cpuidle: Fix panic in CPU off-lining with no idle driver
  ACPI processor: Use safe_halt() rather than halt() in acpi_idle_play_dead()

Pull target fixes from Nicholas Bellinger:
 "Pull two tcm_fc fabric related fixes for -rc2:

  Note that both have been CC'ed to stable, and patch #1 is the
  important one that addresses a memory corruption bug related to FC
  exchange timeouts + command abort.

  Thanks again to MDR for tracking down this issue!"

* '3.4-rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  tcm_fc: Do not free tpg structure during wq allocation failure
  tcm_fc: Add abort flag for gracefully handling exchange timeout

Avoid freeing a registered tpg structure if an alloc_workqueue call
fails.  This fixes a bug where the failure was leaking memory associated
with se_portal_group setup during the original core_tpg_register() call.
Signed-off-by: NMark Rustad <mark.d.rustad@intel.com>
Acked-by: NKiran Patil <Kiran.patil@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Add abort flag and use it to terminate processing when an exchange
is timed out or is reset. The abort flag is used in place of the
transport_generic_free_cmd function call in the reset and timeout
cases, because calling that function in that context would free
memory that was in use. The aborted flag allows the lifetime to
be managed in a more normal way, while truncating the processing.

This change eliminates a source of memory corruption which
manifested in a variety of ugly ways.

(nab: Drop unused struct fc_exch *ep in ft_recv_seq)
Signed-off-by: NMark Rustad <mark.d.rustad@intel.com>
Acked-by: NKiran Patil <Kiran.patil@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Signed-off-by: NIgor Murzov <e-mail@date.by>
Signed-off-by: NLen Brown <len.brown@intel.com>

Pull arch/tile bug fixes from Chris Metcalf:
 "This includes Paul Gortmaker's change to fix the <asm/system.h>
  disintegration issues on tile, a fix to unbreak the tilepro ethernet
  driver, and a backlog of bugfix-only changes from internal Tilera
  development over the last few months.

  They have all been to LKML and on linux-next for the last few days.
  The EDAC change to MAINTAINERS is an oddity but discussion on the
  linux-edac list suggested I ask you to pull that change through my
  tree since they don't have a tree to pull edac changes from at the
  moment."

* 'stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile: (39 commits)
  drivers/net/ethernet/tile: fix netdev_alloc_skb() bombing
  MAINTAINERS: update EDAC information
  tilepro ethernet driver: fix a few minor issues
  tile-srom.c driver: minor code cleanup
  edac: say "TILEGx" not "TILEPro" for the tilegx edac driver
  arch/tile: avoid accidentally unmasking NMI-type interrupt accidentally
  arch/tile: remove bogus performance optimization
  arch/tile: return SIGBUS for addresses that are unaligned AND invalid
  arch/tile: fix finv_buffer_remote() for tilegx
  arch/tile: use atomic exchange in arch_write_unlock()
  arch/tile: stop mentioning the "kvm" subdirectory
  arch/tile: export the page_home() function.
  arch/tile: fix pointer cast in cacheflush.c
  arch/tile: fix single-stepping over swint1 instructions on tilegx
  arch/tile: implement panic_smp_self_stop()
  arch/tile: add "nop" after "nap" to help GX idle power draw
  arch/tile: use proper memparse() for "maxmem" options
  arch/tile: fix up locking in pgtable.c slightly
  arch/tile: don't leak kernel memory when we unload modules
  arch/tile: fix bug in delay_backoff()
  ...

Merge tag 'stable/for-linus-3.4-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen

Pull xen fixes from Konrad Rzeszutek Wilk:
 "Two fixes for regressions:
   * one is a workaround that will be removed in v3.5 with proper fix in
     the tip/x86 tree,
   * the other is to fix drivers to load on PV (a previous patch made
     them only load in PVonHVM mode).

  The rest are just minor fixes in the various drivers and some cleanup
  in the core code."

* tag 'stable/for-linus-3.4-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
  xen/pcifront: avoid pci_frontend_enable_msix() falsely returning success
  xen/pciback: fix XEN_PCI_OP_enable_msix result
  xen/smp: Remove unnecessary call to smp_processor_id()
  xen/x86: Workaround 'x86/ioapic: Add register level checks to detect bogus io-apic entries'
  xen: only check xen_platform_pci_unplug if hvm

Pull MMC fixes from Chris Ball:
 - Disable use of MSI in sdhci-pci, which caused multiple chipsets to
   stop working in 3.4-rc1.  I'll wait to turn this on again until we
   have a chipset whitelist for it.
 - Fix a libertas SDIO powered-resume regression introduced in 3.3;
   thanks to Neil Brown and Rafael Wysocki for this fix.
 - Fix module reloading on omap_hsmmc.
 - Stop trusting the spec/card's specified maximum data timeout length,
   and use three seconds instead.  Previously we used 300ms.

Also cleanups and fixes for s3c, atmel, sh_mmcif and omap_hsmmc.

* tag 'mmc-fixes-for-3.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc: (28 commits)
  mmc: use really long write timeout to deal with crappy cards
  mmc: sdhci-dove: Fix compile error by including module.h
  mmc: Prevent 1.8V switch for SD hosts that don't support UHS modes.
  Revert "mmc: sdhci-pci: Add MSI support"
  Revert "mmc: sdhci-pci: add quirks for broken MSI on O2Micro controllers"
  mmc: core: fix power class selection
  mmc: omap_hsmmc: fix module re-insertion
  mmc: omap_hsmmc: convert to module_platform_driver
  mmc: omap_hsmmc: make it behave well as a module
  mmc: omap_hsmmc: trivial cleanups
  mmc: omap_hsmmc: context save after enabling runtime pm
  mmc: omap_hsmmc: use runtime put sync in probe error patch
  mmc: sdio: Use empty system suspend/resume callbacks at the bus level
  mmc: bus: print bus speed mode of UHS-I card
  mmc: sdhci-pci: add quirks for broken MSI on O2Micro controllers
  mmc: sh_mmcif: Simplify calculation of mmc->f_min
  mmc: sh_mmcif: mmc->f_max should be half of the bus clock
  mmc: sh_mmcif: double clock speed
  mmc: block: Remove use of mmc_blk_set_blksize
  mmc: atmel-mci: add support for odd clock dividers
  ...

I have a new optimized x86 "strncpy_from_user()" that will use these
same helper functions for all the same reasons the name lookup code uses
them.  This is preparation for that.

This moves them into an architecture-specific header file.  It's
architecture-specific for two reasons:

 - some of the functions are likely to want architecture-specific
   implementations.  Even if the current code happens to be "generic" in
   the sense that it should work on any little-endian machine, it's
   likely that the "multiply by a big constant and shift" implementation
   is less than optimal for an architecture that has a guaranteed fast
   bit count instruction, for example.

 - I expect that if architectures like sparc want to start playing
   around with this, we'll need to abstract out a few more details (in
   particular the actual unaligned accesses).  So we're likely to have
   more architecture-specific stuff if non-x86 architectures start using
   this.

   (and if it turns out that non-x86 architectures don't start using
   this, then having it in an architecture-specific header is still the
   right thing to do, of course)
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>