Rafael J. Wysocki reported weird, multi-seconds delays during
suspend/resume and bisected it back to:

  commit 82a1fcb9
  Author: Ingo Molnar <mingo@elte.hu>
  Date:   Fri Jan 25 21:08:02 2008 +0100

      softlockup: automatically detect hung TASK_UNINTERRUPTIBLE tasks

fix it:

 - restore the old wakeup mechanism
 - fix break usage in do_each_thread() { } while_each_thread().
 - fix the hotplug switch stmt, a fall-through case was broken.
Bisected-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Tested-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fix bootup crash in native_read_tsc() that was reported on an Athlon-XP
and bisected. The correct feature boundary for X86_FEATURE_MFENCE_RDTSC
is not XMM but XMM2.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6: (100 commits)
  ide: move hwif_register() call out of ide_probe_port()
  ide: factor out code for tuning devices from ide_probe_port()
  ide: move handling of I/O resources out of ide_probe_port()
  ide: make probe_hwif() return an error value
  ide: use ide_remove_port_from_hwgroup in init_irq()
  ide: prepare init_irq() for using ide_remove_port_from_hwgroup()
  ide: factor out code removing port from hwgroup from ide_unregister()
  ide: I/O resources are released too early in ide_unregister()
  ide: cleanup ide_system_bus_speed()
  ide: remove needless zeroing of hwgroup fields from init_irq()
  ide: remove unused ide_hwgroup_t fields
  ide_platform: remove struct hwif_prop
  ide: remove hwif->present manipulations from hwif_init()
  ide: move wait_hwif_ready() documentation in the right place
  ide: fix handling of busy I/O resources in probe_hwif()
  <linux/hdsmart.h> is not used by kernel code
  ide: don't include <linux/hdsmart.h>
  ide-floppy: cleanup header
  ide: update/add my Copyrights
  ide: delete filenames/versions from comments
  ...

* Add BUG_ON(hwif->present) at the start of ide_probe_port().

* Move hwif_register() call (along with setting hwif->present) from
  ide_probe_port() to ide_device_add_all().

  As a result the port will be registered with the device tree _after_:
  - probing both devices (if both are present)
  - port reset (if hwif->reset is set)
  - restoring local IRQs state and re-enabling port IRQ

While at it:

* Rename hwif_register() to ide_register_port().
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Factor out code for tuning devices from ide_probe_port()
  to ide_port_tune_devices().

* Move ide_port_tune_devices() call from ide_probe_port()
  to ide_device_add_all().

There should be no functionality changes caused by this patch.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Reserve/release I/O resources in ide_device_add_all() instead of
ide_probe_port().
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Make probe_hwif() return an error value.

While at it:

* Remove comment about MAX_DRIVES == 2 limitation (it is not special to
  probe_hwif(), it is a general assumption taken by a lot of IDE code).

* Rename probe_hwif() to ide_probe_port().
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

There should be no functionality changes caused by this patch.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* BUG_ON() early if 'hwif->next == hwif' and there is a 'match' hwgroup.

* Remove printk() for impossible condition and add a BUG_ON(hwgroup->drive)
  to match code in ide_unregister().
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Release I/O resources after releasing DMA.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Use the facts that:

* When ide_init() is called 'system_bus_speed' always equals zero.

* system_bus_clock() is never called before ide_init().

and:

* Move printk() from ide_system_bus_speed() to ide_init().

* Don't assign 'system_bus_speed' in ide_system_bus_speed().

* Don't call ide_system_bus_speed() in system_bus_clock().
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Since kmalloc_node() is called with __GFP_ZERO flag there is no need to
explicitly zero hwgroup fields.

While at it:

* Use 'hwif' instead of 'hwif->drives[0].hwif' for kmalloc_node() call.

* Fix whitespace damage.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Remove 'if (hwif != hwif_prop.hwif)' check from plat_ide_remove()
  (pdev->dev.driver_data and hwif_prop.hwif are set at the same time
   in plat_ide_probe() and are identical).

* Use hwif->index for ide_unregister() in plat_ide_remove().

* Use local variables instead of hwif_prop.plat_ide[_alt]_mapbase in
  plat_ide_probe() and remove no longer needed struct hwif_prop.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Call hwif_init() only if hwif->present is set.

* If hwif_init() fails clear hwif->present.

* Remove hwif->present manipulations from hwif_init().
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Move wait_hwif_ready() documentation before this function.

* Fix trailing whitespaces.

* s/wait-for-busy/wait-for-non-busy/

* Remove no longer valid comment about the current code behavior.

* Docbook-ize it.

* Rename wait_hwif_ready() to ide_port_wait_ready().
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

It could be that I/O resources are busy because some other host driver
has already claimed them so don't unregister the devices.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

IDE doesn't need it.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Move ide-floppy historical changelog to
Documentation/ide/ChangeLog.ide-floppy.1996-2002
Signed-off-by: Borislav Petkov <bbpetkov@yahoo.de>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

The last years stuff and a trip down memory lane...
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Delete filenames/versions from comments.

I'm leaving decisions about adding DRV_VERSION defines and MODULE_VERSION()-s
to maintainers of the respective drivers.

While at it:

* Remove unused VERSION define from ide.c.

* Remove unused/stale DRV_VERSION define from au1xxx-ide.c.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Release DMA resource before zeroing pmif.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Add missing hw.dev setup (so hwif->gendev.parent will be set by
ide_init_port_hw() to point to the parent device).
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Make ide_build_sglist() and ide_destroy_dmatable() available also when
  CONFIG_BLK_DEV_IDEDMA_PCI=n.

* Use ide_build_sglist() and ide_destroy_dmatable() in {ics,au1xxx-}ide.c
  and remove no longer needed {ics,au}ide_build_sglist().

There should be no functionality changes caused by this patch.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* pci_map_sg() -> dma_map_sg() in ide_build_sglist().

* pci_unmap_sg() -> dma_unmap_sg() in ide_destroy_dmatable().

There should be no functionality changes caused by this patch except
for blackfin arch whose dma_[un]map_sg() implementation differs from
pci_[un]map_sg() one (on s390 arch there is no PCI, on avr32 and h8300
archs PCI is currently unsupported, on m32r arch PCI support depends
on BROKEN, on m68k arch PCI support depends on HADES which in turn
depends on BROKEN, on all other archs dma_[un]map_sg() functionality
matches with pci_[un]map_sg() one).

blackfin behavior change was ack-ed by Bryan Wu.

Cc: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Setup hwif->dev in icside_setup().

* Use hwif->dev instead of state->dev in icside_build_sglist(),
  icside_dma_end(), icside_dma_start() and icside_dma_setup().

* Remove no longer needed 'dev' field from struct icside_state.

Cc: Russell King <rmk@arm.linux.org.uk>
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Setup hwif->dev in au_ide_probe().

* Use hwif->dev instead of ahwif->dev in auide_build_sglist(),
  auide_build_dmatable(), auide_dma_end() and auide_ddma_init().

* Remove no longer needed 'dev' field from _auide_hwif type.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Keep pointer to struct device instead of struct pci_dev in ide_hwif_t.

While on it:
* Use *dev->dma_mask instead of pci_dev->dma_mask in ide_toggle_bounce().

There should be no functionality changes caused by this patch.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Use ide_destroy_dmatable() in:
* ide-dma.c::ide_build_dmatable()
* sgiioc4.c::sgiioc4_build_dma_table()
* pmac.c::pmac_ide_{build,destroy}_dmatable()

There should be no functionality changes caused by this patch.

v2:
* pmac.c build fix from Andrew Morton.

Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

do_rw_taskfile() has been fixed to check the return value
of ->dma_setup method so this BUG_ON() is no longer needed.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

'hwif->pci_dev && !hwif->pci_dev->vendor' condition is never true,
check for 'hwif->chipset != ide_pci' instead.
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Add IDE_HFLAG_NO_DSC host flag for hosts that doesn't support DSC overlap.

* Set it in aec62xx (for ATP850UF only) and hpt34x host drivers.

* Convert ide-tape device driver to check for IDE_HFLAG_NO_DSC flag.
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Rename 'simplex_stat' variable to 'dma_stat' in ide_get_or_set_dma_base().

* Factor out code for forcing host out of "simplex" mode from
  ide_get_or_set_dma_base() to ide_pci_clear_simplex() helper.

* Add IDE_HFLAG_CLEAR_SIMPLEX host flag and set it in alim15x3 (for M5229),
  amd74xx (for AMD 7409), cmd64x (for CMD643), generic (for Netcell) and
  serverworks (for CSB5) host drivers.

* Make ide_get_or_set_dma_base() test for IDE_HFLAG_CLEAR_SIMPLEX host flag
  instead of checking dev->device (BTW the code was buggy because it didn't
  check for dev->vendor, luckily none of these PCI Device IDs was used by
  some other vendor for PCI IDE controller).
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Remove no longer needed assertion from amd74xx_probe().

* Factor out cable detection for AMD7409 to amd7409_cable_detect() and for
  chipsets >= AMD7411 to amd7411_cable_detect().

* Use dev->vendor and dev->device instead of amd_config->udma_mask when
  selecting cable detection method and checking for broken FIFO support in
  init_chipset_amd74xx().

* Remove no longer needed AMD_BAD_FIFO define.

* Add 'swdma' parameter for setting .swdma_mask to DECLARE_AMD_DEV() macro.

* Add 'udma' parameter for setting .udma_mask to DECLARE_{AMD,NV}_DEV() macro.

* Keep a copy of a current amd74xx_chipsets[] entry in amd74xx_probe()
  in order to fix ->swdma_mask on early AMD7409 revisions and ->udma_mask
  on Serenade mainboards.

* Remove no longer needed fixups from init_chipset_amd74xx()
  and AMD_CHECK_{SWDMA,SERENADE} defines.

* Move printing banner message from init_chipset_amd74xx() to amd74xx_probe(),
  also remove incorrect comment while at it.

* Use hwif->ultra_mask instead of amd_config->udma_mask in amd_set_drive().

* Add 'udma_mask' argument to amd_set_speed() and pass UDMA mask from
  amd_set_drive() instead of using amd_config->udma_mask.

* Move amd_config->base from AMD_* defines to users of these defines and add
  0x40 the defined values.  Then add amd_offset() inline helper for selecting
  offset from 0x40 base (needed for nVidia controllers) and finally use it in
  amd_set_speed(), amd7411_cable_detect() and init_chipset_amd74xx() instead
  of amd_config->base.

* Remove no longer needed AMD_BAD_SWDMA define, ->{swdma,ultra}_mask setup
  from init_hwif_amd74xx(), amd_{config,chipset} variables and amd_ide_chips
  table.

* Fix init_chipset_amd74xx() comment.

* Bump driver version.

There should be no functionality changes caused by this patch.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

According to http://marc.info/?l=linux-ide&m=114346138611631, the drivers must
always register 8 DMA ports with ide_setup_dma(), so its last argument is not
needed. While at it, kill some useless parens in that function...
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

The TRM-290 chip is *not* SFF-8038i compatible and therefore can *not* call
ide_setup_dma() -- fix this and also cleanup the code a bit...
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>