GitHub/exynos8895/android_kernel_samsung_universal8895.git
18 years ago[PATCH] ata_piix: clean up port flags
Tejun Heo [Fri, 10 Nov 2006 09:08:10 +0000 (18:08 +0900)]
[PATCH] ata_piix: clean up port flags

* move common flags into PIIX_PATA_FLAGS and PIIX_SATA_FLAGS
* kill unnecessary ATA_FLAG_SRST

Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: use FLUSH_EXT only when driver is larger than LBA28 limit
Tejun Heo [Sat, 11 Nov 2006 11:10:45 +0000 (20:10 +0900)]
[PATCH] libata: use FLUSH_EXT only when driver is larger than LBA28 limit

Many drives support LBA48 even when its capacity is smaller than
1<<28, as LBA48 is required for many functionalities.  FLUSH_EXT is
mandatory for drives w/ LBA48 support.

Interestingly, at least one of such drives (ST960812A) has problems
dealing with FLUSH_EXT.  It eventually completes the command but takes
around 7 seconds to finish in many cases thus drastically slowing down
IO transactions.  This seems to be a firmware bug which sneaked into
production probably because no other ATA driver including linux IDE
issues FLUSH_EXT to drives which report support for LBA48 & FLUSH_EXT
but is smaller than 1<<28 blocks.

This patch adds ATA_DFLAG_FLUSH_EXT which is set iff the drive
supports LBA48 & FLUSH_EXT and is larger than LBA28 limit.  Both cache
flush paths are updated to issue FLUSH_EXT only when the flag is set.
Note that the changed behavior is more inline with the rest of libata.
libata prefers shorter commands whenever possible.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Danny Kukawka <dkukawka@novell.com>
Cc: Stefan Seyfried <seife@novell.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] ahci: honor PORTS_IMPL on ICH8s
Tejun Heo [Thu, 9 Nov 2006 06:08:40 +0000 (15:08 +0900)]
[PATCH] ahci: honor PORTS_IMPL on ICH8s

Some ICH8s use non-linear port mapping.  ahci driver didn't use to
honor PORTS_IMPL and this made ports after hole nonfunctional.  This
patch makes ahci mark those ports as dummy and properly initialize all
the implemented ports after the dummies.

As it's unknown whether other AHCIs implement PORTS_IMPL register
properly, new board id board_ahci_pi is added and selectively applied
to ICH8s.  All other AHCIs continue to use linear mapping regardless
of PORTS_IMPL value.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Robin H. Johnson <robbat2@gentoo.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[libata] ARM: add ixp4xx PATA driver
Alessandro Zummo [Tue, 14 Nov 2006 18:43:21 +0000 (13:43 -0500)]
[libata] ARM: add ixp4xx PATA driver

Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] sata_sis: slave support on SiS965
Uwe Koziolek [Wed, 8 Nov 2006 08:57:00 +0000 (09:57 +0100)]
[PATCH] sata_sis: slave support on SiS965

SiS965 and SiS180 chips must support slave mode,
SiS965L and SiS964 chips must not support slave mode.

SCR_STATUS for SATA ports in powerdown state fixed.
Now returning 0x0113 instead of 0x0117. Avoids problem
on detecting sata_sis controller.

Signed-off-by: Uwe Koziolek <uwe.koziolek@gmx.net>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] ahci: preserve PORTS_IMPL over host resets
Tejun Heo [Thu, 2 Nov 2006 03:17:23 +0000 (12:17 +0900)]
[PATCH] ahci: preserve PORTS_IMPL over host resets

Instead of writing 0xf blindly, preserve the content of write-once
PORTS_IMPL register over host resets.

This patch is taken from Jeff Garzik's AHCI init update patch.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] pdc202xx_old: Fix name clashes with PA-RISC
Alan Cox [Fri, 3 Nov 2006 13:18:06 +0000 (13:18 +0000)]
[PATCH] pdc202xx_old: Fix name clashes with PA-RISC

pdc_* functions are part of the global namespace for the PDC on PA-RISC
systems and this means our choice of pdc_ causes collisions between the
PDC globals and our static functions. Rename them to pdc202xx where they
are for both 2024x and 2026x.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[libata] ahci: Match PCI class code for AHCI
Jeff Garzik [Wed, 1 Nov 2006 10:10:42 +0000 (05:10 -0500)]
[libata] ahci: Match PCI class code for AHCI

Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: move dev->max_sectors configuration into ata_dev_configure()
Tejun Heo [Wed, 1 Nov 2006 09:39:55 +0000 (18:39 +0900)]
[PATCH] libata: move dev->max_sectors configuration into ata_dev_configure()

Move dev->max_sectors configuration from ata_scsi_dev_config() to
ata_dev_configure().

* more consistent.
* allows LLDs to peek at the default dev->max_sectors value.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: implement ATA_EHI_SETMODE and ATA_EHI_POST_SETMODE
Tejun Heo [Wed, 1 Nov 2006 09:39:27 +0000 (18:39 +0900)]
[PATCH] libata: implement ATA_EHI_SETMODE and ATA_EHI_POST_SETMODE

libata EH used to perform ata_set_mode() iff the EH session performed
reset as indicated by ATA_EHI_DID_RESET.  This is incorrect because
->dev_config() called by revalidation is allowed to modify transfer
mode which ata_set_mode() should take care of.  This patch implements
the following two flags.

* ATA_EHI_SETMODE: set during EH to schedule ata_set_mode().  Both new
  device attachment and revalidation set this flag.

* ATA_EHI_POST_SETMODE: set while the device is revalidated after
  ata_set_mode().  Post-setmode revalidation is different from initial
  configuaration and EH revalidation in that ->dev_config() is not
  allowed tune transfer mode.  LLD can use this flag to determine
  whether it's allowed to tune transfer mode.  Note that POST_SETMODE
  ->dev_config() is guaranteed to be preceded by non-POST_SETMODE
  ->dev_config().

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: implement ATA_EHI_PRINTINFO
Tejun Heo [Wed, 1 Nov 2006 09:38:52 +0000 (18:38 +0900)]
[PATCH] libata: implement ATA_EHI_PRINTINFO

Implement ehi flag ATA_EHI_PRINTINFO.  This flag is set when device
configuration needs to print out device info.  This used to be handled
by @print_info argument to ata_dev_configure() but LLDs also need to
know about it in ->dev_config() callback.

This patch replaces @print_info w/ ATA_EHI_PRINTINFO and make sata_sil
print workaround messages only on the initial configuration.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] ahci: update ahci-vt8251 reset sequence
Tejun Heo [Wed, 1 Nov 2006 09:00:24 +0000 (18:00 +0900)]
[PATCH] ahci: update ahci-vt8251 reset sequence

ahci-vt8251

* requires hardreset after PHY status change

* doesn't clear BSY on signature FIS after hardreset

* needs SError cleared for the port to operate after hardreset

This patch implements ahci_vt8251_hardreset() and sets
ATA_FLAG_HRST_TO_RESUME to handle the above behaviors.  This fixes EH
including hotplug on vt8251.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: separate out and export sata_port_hardreset()
Tejun Heo [Wed, 1 Nov 2006 08:59:53 +0000 (17:59 +0900)]
[PATCH] libata: separate out and export sata_port_hardreset()

Separate out sata_port_hardreset() from sata_std_hardreset().  This
will be used by LLD hardreset implementation and later by PMP.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] ahci: kill AHCI_FLAG_RESET_NEEDS_CLO
Tejun Heo [Wed, 1 Nov 2006 08:58:33 +0000 (17:58 +0900)]
[PATCH] ahci: kill AHCI_FLAG_RESET_NEEDS_CLO

Now that ahci_softreset() is fixed to automatically perform CLO if
BSY/DRQ is set on entry, AHCI_FLAG_RESET_NEEDS_CLO is redundant.  Kill
it.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] sata_sil24: add PMP related constants
Tejun Heo [Sun, 15 Oct 2006 23:47:18 +0000 (08:47 +0900)]
[PATCH] sata_sil24: add PMP related constants

Add PMP related constants.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] sata_sil24: rename PMP related constants
Tejun Heo [Sun, 15 Oct 2006 23:47:18 +0000 (08:47 +0900)]
[PATCH] sata_sil24: rename PMP related constants

Rename PMP related constants for consistency.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: make user scan wait for scan to complete
Tejun Heo [Sat, 30 Sep 2006 09:07:17 +0000 (18:07 +0900)]
[PATCH] libata: make user scan wait for scan to complete

Make user scan wait for scan to complete.  This way user can wait for
warm plug request to complete and is prevented from causing EH event
storm by repetitively issuing scan request while EH is in progress.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Fajun Chen <fajunchen@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] sata_nv: SRST sometimes fails after hotplug, use HRST_TO_RESUME
Tejun Heo [Thu, 28 Sep 2006 08:49:22 +0000 (17:49 +0900)]
[PATCH] sata_nv: SRST sometimes fails after hotplug, use HRST_TO_RESUME

NV controllers sometimes fail to perform softreset after hotplug.
Make it use hardreset to resume link.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: move ata_irq_on() into libata-sff.c
Tejun Heo [Mon, 9 Oct 2006 02:10:26 +0000 (11:10 +0900)]
[PATCH] libata: move ata_irq_on() into libata-sff.c

ata_irq_on() isn't used outside of libata core layer.  The function is
TF/SFF interface specific but currently used by core path with some
hack too.  Move it from include/linux/libata.h to
drivers/ata/libata-sff.c.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: handle 0xff status properly
Tejun Heo [Mon, 9 Oct 2006 09:32:15 +0000 (18:32 +0900)]
[PATCH] libata: handle 0xff status properly

libata waits for !BSY even when the status register reports 0xff.
This causes long boot delays when D8 isn't pulled down properly.  This
patch does the followings.

* don't wait if status register is 0xff in all wait functions

* make ata_busy_sleep() return 0 on success and -errno on failure.
  -ENODEV is returned on 0xff status and -EBUSY on other failures.

* make ata_bus_softreset() succeed on 0xff status.  0xff status is not
  reset failure.  It indicates no device.  This removes unnecessary
  retries on such ports.  Note that the code change assumes unoccupied
  port reporting 0xff status does not produce valid device signature.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Joe Jin <lkmaillist@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] ata: Generic platform_device libata driver
Paul Mundt [Sat, 28 Oct 2006 02:08:48 +0000 (19:08 -0700)]
[PATCH] ata: Generic platform_device libata driver

needs a changelog

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Tejun Heo <htejun@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: Revamp blacklist support to allow multiple kinds of blacklisting...
Alan Cox [Sat, 28 Oct 2006 02:08:46 +0000 (19:08 -0700)]
[PATCH] libata: Revamp blacklist support to allow multiple kinds of blacklisting flaws

Signed-off-by: Alan Cox <alan@redhat.com>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Tejun Heo <htejun@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] sata_nv ADMA/NCQ support for nForce4
Robert Hancock [Sat, 28 Oct 2006 02:08:41 +0000 (19:08 -0700)]
[PATCH] sata_nv ADMA/NCQ support for nForce4

This patch adds support for ADMA mode on NVIDIA nForce4 (CK804/MCP04) SATA
controllers to the sata_nv driver.  Benefits of ADMA mode include:

- NCQ support

- Reduced CPU overhead (controller DMAs command information from memory
  instead of them being pushed in by the CPU)

- Full 64-bit DMA support

ADMA mode is enabled by default in this version.  To disable it, set the
module parameter adma_enabled=0.

Signed-off-by: Robert Hancock <hancockr@shaw.ca>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Tejun Heo <htejun@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[libata] pata_marvell: minor and trivial cleanups
Jeff Garzik [Sat, 21 Oct 2006 19:54:13 +0000 (15:54 -0400)]
[libata] pata_marvell: minor and trivial cleanups

- use pci_iomap() [Alan, version 0.0.5t]
- fix Alan's version 0.0.5t change
- line length, trailing whitespace, case indentation cleanups
- don't use deprecated ->eng_timeout() in a driver that uses new EH

Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] pata_marvell: Marvell 6101/6145 PATA driver
Alan Cox [Mon, 16 Oct 2006 15:40:06 +0000 (16:40 +0100)]
[PATCH] pata_marvell: Marvell 6101/6145 PATA driver

This is a legacy mode PATA driver for the 6101/45 and will also drive
the SATA ports 1 & 2 in legacy mode as well if desired. Tested and
confirmed working by users. The chip supports AHCI type behaviour for
SATA and has a more advanced PATA interface as well so this driver will
get it working but not get best performance for now.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] pci_module_init() conversion for pata_pdc2027x
Henrik Kretzschmar [Tue, 10 Oct 2006 21:29:24 +0000 (14:29 -0700)]
[PATCH] pci_module_init() conversion for pata_pdc2027x

pci_module_init() conversion for pata_pdc2027x

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: add 40pin "short" cable support, honour drive side speed detection
Alan Cox [Tue, 10 Oct 2006 21:28:11 +0000 (14:28 -0700)]
[PATCH] libata: add 40pin "short" cable support, honour drive side speed detection

[deweerdt@free.fr: build fix]
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Frederik Deweerdt <deweerdt@free.fr>
Cc: Tejun Heo <htejun@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] pci_module_init-convertion-in-ata_genericc fix
Alan Cox [Tue, 10 Oct 2006 21:28:12 +0000 (14:28 -0700)]
[PATCH] pci_module_init-convertion-in-ata_genericc fix

Acked-by: Alan Cox <alan@redhat.com> but please apply as follows to keep
revision id straight

Cc: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] pci_module_init convertion in ata_generic.c
Henrik Kretzschmar [Tue, 10 Oct 2006 21:28:12 +0000 (14:28 -0700)]
[PATCH] pci_module_init convertion in ata_generic.c

pci_module_init convertion in ata_generic.c.

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years ago[PATCH] libata: Winbond support
Alan Cox [Fri, 29 Sep 2006 17:30:05 +0000 (18:30 +0100)]
[PATCH] libata: Winbond support

Winbond 83759A support in non-multichip mode (afaik nobody ever used
multichip mode anyway). The 83759 is not supported by this driver as it
is already handled elsewhere and doens't use the same interfaces.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog
Linus Torvalds [Sat, 2 Dec 2006 00:47:26 +0000 (16:47 -0800)]
Merge /linux/kernel/git/wim/linux-2.6-watchdog

* master.kernel.org:/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  [WATCHDOG] MIPS RM9000 on-chip watchdog device - patch 4
  [WATCHDOG] MIPS RM9000 on-chip watchdog device - patch 3
  [WATCHDOG] MIPS RM9000 on-chip watchdog device - patch 2
  [WATCHDOG] MIPS RM9000 on-chip watchdog device - patch 1
  [WATCHDOG] MIPS RM9000 on-chip watchdog device
  [WATCHDOG] Add iTCO vendor specific support
  [WATCHDOG] sc1200wdt.c pnp unregister fix.
  [WATCHDOG] config.h removal
  [WATCHDOG] NS pc87413-wdt Watchdog driver - fixes
  [WATCHDOG] NS pc87413-wdt Watchdog driver v1.1
  [WATCHDOG] NS pc87413-wdt Watchdog driver

18 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Sat, 2 Dec 2006 00:46:22 +0000 (16:46 -0800)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  [PATCH] libata: add missing sht->slave_destroy
  [PATCH] ahci: ignore PORT_IRQ_IF_ERR on JMB controllers

18 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Sat, 2 Dec 2006 00:44:02 +0000 (16:44 -0800)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus: (31 commits)
  [MIPS] Remove duplicate ISA DMA code for 0 DMA channel case.
  [MIPS] Remove unused definition of cpu_to_lelongp()
  [MIPS] Remove userspace proofing from <asm/bitops.h>.
  [MIPS] Remove old junk left from old atomic_lock.
  [MIPS] Use conditional traps for BUG_ON on MIPS II and better.
  [MIPS] mips HPT cleanup: make clocksource_mips public
  [MIPS] do_IRQ cleanup
  [MIPS] Avoid dupliate D-cache flush on R400C / R4400 SC and MC variants.
  [MIPS] Remove redundant r4k_blast_icache() calls
  [MIPS] Work around bogus gcc warnings.
  [MIPS] Fix double inclusions
  [MIPS] use generic_handle_irq, handle_level_irq, handle_percpu_irq
  [MIPS] IRQ cleanups
  [MIPS] mips hpt cleanup: get rid of mips_hpt_init
  [MIPS] PB1200: Remove duplicate definitions
  [MIPS] Fix alignment hole in struct cache_desc; shrink struct.
  [MIPS] Oprofile: kernel support for the R10000.
  [MIPS] Remove unused R10000 performance counter definitions.
  [MIPS] Add support for kexec
  [MIPS] Don't print presence of WAIT instruction on bootup.
  ...

18 years agoMerge branch 'for-2.6.20' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Sat, 2 Dec 2006 00:43:42 +0000 (16:43 -0800)]
Merge branch 'for-2.6.20' of git://git./linux/kernel/git/jmorris/selinux-2.6

* 'for-2.6.20' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/selinux-2.6:
  SELinux: validate kernel object classes and permissions
  SELinux: ensure keys constant in hashtab_search
  SELinux: export object class and permission definitions
  SELinux: remove current object class and permission validation mechanism

18 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Sat, 2 Dec 2006 00:43:06 +0000 (16:43 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (25 commits)
  IB/ucm: Fix deadlock in cleanup
  IB/cm: Fix automatic path migration support
  IPoIB: Fix skb leak when freeing neighbour
  IB/srp: Fix memory leak on reconnect
  RDMA/addr: list_move() cleanups
  RDMA/addr: Fix some cancellation problems in process_req()
  RDMA/amso1100: Prevent deadlock in destroy QP
  IB/mthca: Fix initial SRQ logsize for mem-free HCAs
  IB/ehca: Use WQE offset instead of WQE addr for pending work reqs
  RDMA/iwcm: Fix comment for iwcm_deref_id() to match code
  RDMA/iwcm: Remove unnecessary function argument
  RDMA/iwcm: Remove unnecessary initializations
  RDMA/iwcm: Fix memory leak
  RDMA/iwcm: Fix memory corruption bug in cm_work_handler()
  IB: Convert kmem_cache_t -> struct kmem_cache
  IB/ipath: Fix typo in pma_counter_select subscript
  RDMA/amso1100: Fix section mismatches
  IB/mthca: Fix section mismatches
  IB/srp: Increase supported CDB size
  RDMA/cm: Remove setting local write as part of QP access flags
  ...

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/gregkh/usb-2.6
Linus Torvalds [Sat, 2 Dec 2006 00:41:58 +0000 (16:41 -0800)]
Merge /pub/scm/linux/kernel/git/gregkh/usb-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/gregkh/usb-2.6: (103 commits)
  usbcore: remove unused argument in autosuspend
  USB: keep count of unsuspended children
  USB hub: simplify remote-wakeup handling
  USB: struct usb_device: change flag to bitflag
  OHCI: make autostop conditional on CONFIG_PM
  USB: Add autosuspend support to the hub driver
  EHCI: Fix root-hub and port suspend/resume problems
  USB: create a new thread for every USB device found during the probe sequence
  USB: add driver for the USB debug devices
  USB: added dynamic major number for USB endpoints
  USB: pegasus error path not resetting task's state
  USB: endianness fix for asix.c
  USB: build the appledisplay driver
  USB serial: replace kmalloc+memset with kzalloc
  USB: hid-core: canonical defines for Apple USB device IDs
  USB: idmouse cleanup
  USB: make drivers/usb/core/driver.c:usb_device_match() static
  USB: lh7a40x_udc remove double declaration
  USB: pxa2xx_udc recognizes ixp425 rev b0 chip
  usbtouchscreen: add support for DMC TSC-10/25 devices
  ...

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/gregkh/pci-2.6
Linus Torvalds [Sat, 2 Dec 2006 00:41:27 +0000 (16:41 -0800)]
Merge /pub/scm/linux/kernel/git/gregkh/pci-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/gregkh/pci-2.6: (28 commits)
  PCI: make arch/i386/pci/common.c:pci_bf_sort static
  PCI: ibmphp_pci.c: fix NULL dereference
  pciehp: remove unnecessary pci_disable_msi
  pciehp: remove unnecessary free_irq
  PCI: rpaphp: change device tree examination
  PCI: Change memory allocation for acpiphp slots
  i2c-i801: SMBus patch for Intel ICH9
  PCI: irq: irq and pci_ids patch for Intel ICH9
  PCI: pci_{enable,disable}_device() nestable ports
  PCI: switch pci_{enable,disable}_device() to be nestable
  PCI: arch/i386/kernel/pci-dma.c: ioremap balanced with iounmap
  pci/i386: style cleanups
  PCI: Block on access to temporarily unavailable pci device
  pci: fix __pci_register_driver error handling
  pci: clear osc support flags if no _OSC method
  acpiphp: fix missing acpiphp_glue_exit()
  acpiphp: fix use of list_for_each macro
  Altix: Initial ACPI support - ROM shadowing.
  Altix: SN ACPI hotplug support.
  Altix: Add initial ACPI IO support
  ...

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/gregkh/driver-2.6
Linus Torvalds [Sat, 2 Dec 2006 00:41:07 +0000 (16:41 -0800)]
Merge /pub/scm/linux/kernel/git/gregkh/driver-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/gregkh/driver-2.6: (36 commits)
  Driver core: show drivers in /sys/module/
  Documentation/driver-model/platform.txt update/rewrite
  Driver core: platform_driver_probe(), can save codespace
  driver core: Use klist_remove() in device_move()
  driver core: Introduce device_move(): move a device to a new parent.
  Driver core: make drivers/base/core.c:setup_parent() static
  driver core: Introduce device_find_child().
  sysfs: sysfs_write_file() writes zero terminated data
  cpu topology: consider sysfs_create_group return value
  Driver core: Call platform_notify_remove later
  ACPI: Change ACPI to use dev_archdata instead of firmware_data
  Driver core: add dev_archdata to struct device
  Driver core: convert sound core to use struct device
  Driver core: change mem class_devices to be real devices
  Driver core: convert fb code to use struct device
  Driver core: convert firmware code to use struct device
  Driver core: convert mmc code to use struct device
  Driver core: convert ppdev code to use struct device
  Driver core: convert PPP code to use struct device
  Driver core: convert cpuid code to use struct device
  ...

18 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Sat, 2 Dec 2006 00:40:13 +0000 (16:40 -0800)]
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  [CIFS] Fix timezone handling on stat to os/2
  [CIFS] Incorrect hardlink count when original file is cached (oplocked)

18 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bunk/trivial
Linus Torvalds [Sat, 2 Dec 2006 00:38:18 +0000 (16:38 -0800)]
Merge git://git./linux/kernel/git/bunk/trivial

* git://git.kernel.org/pub/scm/linux/kernel/git/bunk/trivial:
  Fix typos in drivers/isdn/hisax/isdnl2.c
  Fix typos in doc and comments
  BUG_ON conversion for fs/aio.c
  BUG_ON conversion for drivers/mmc/omap.c
  BUG_ON conversion for drivers/media/video/pwc/pwc-if.c
  Fix misc .c/.h comment typos
  Fix misc Kconfig typos
  Fix typos in /Documentation : Misc
  Fix typos in /Documentation : 'U-Z'
  Fix typos in /Documentation : 'T''
  Fix jiffies.h comment
  tabify MAINTAINERS
  fix spelling error in include/linux/kernel.h
  mqueue.h: don't include linux/types.h

18 years agoDriver core: show drivers in /sys/module/
Kay Sievers [Fri, 24 Nov 2006 11:15:25 +0000 (12:15 +0100)]
Driver core: show drivers in /sys/module/

Show the drivers, which belong to the module:
  $ ls -l /sys/module/usbcore/drivers/
  hub -> ../../../bus/usb/drivers/hub
  usb -> ../../../bus/usb/drivers/usb
  usbfs -> ../../../bus/usb/drivers/usbfs

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDocumentation/driver-model/platform.txt update/rewrite
David Brownell [Fri, 17 Nov 2006 07:30:14 +0000 (23:30 -0800)]
Documentation/driver-model/platform.txt update/rewrite

This is almost a rewrite of the driver-model/platform.txt documentation;
the previous text was obsolete (for several years), evidently it never
got updated to match the change from being a PC "legacy_bus" to the more
widely used core bus for most embedded systems.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: platform_driver_probe(), can save codespace
David Brownell [Fri, 17 Nov 2006 07:28:47 +0000 (23:28 -0800)]
Driver core: platform_driver_probe(), can save codespace

This defines a new platform_driver_probe() method allowing the driver's
probe() method, and its support code+data, to safely live in __init
sections for typical system configurations.

Many system-on-chip processors could benefit from this API, to the tune
of recovering hundreds to thousands of bytes per driver.  That's memory
which is currently wasted holding code which can never be called after
system startup, yet can not be removed.   It can't be removed because of
the linkage requirement that pointers to init section code (like, ideally,
probe support) must not live in other sections (like driver method tables)
after those pointers would be invalid.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agodriver core: Use klist_remove() in device_move()
Cornelia Huck [Wed, 22 Nov 2006 16:49:39 +0000 (17:49 +0100)]
driver core: Use klist_remove() in device_move()

As pointed out by Alan Stern, device_move needs to use klist_remove which waits
until removal is complete.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agodriver core: Introduce device_move(): move a device to a new parent.
Cornelia Huck [Mon, 20 Nov 2006 16:07:51 +0000 (17:07 +0100)]
driver core: Introduce device_move(): move a device to a new parent.

Provide a function device_move() to move a device to a new parent device. Add
auxilliary functions kobject_move() and sysfs_move_dir().
kobject_move() generates a new uevent of type KOBJ_MOVE, containing the
previous path (DEVPATH_OLD) in addition to the usual values. For this, a new
interface kobject_uevent_env() is created that allows to add further
environmental data to the uevent at the kobject layer.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Acked-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: make drivers/base/core.c:setup_parent() static
Adrian Bunk [Fri, 17 Nov 2006 01:19:44 +0000 (02:19 +0100)]
Driver core: make drivers/base/core.c:setup_parent() static

This patch makes the needlessly global setup_parent() static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agodriver core: Introduce device_find_child().
Cornelia Huck [Thu, 16 Nov 2006 14:42:07 +0000 (15:42 +0100)]
driver core: Introduce device_find_child().

Introduce device_find_child() to match device_for_each_child().

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agosysfs: sysfs_write_file() writes zero terminated data
Thomas Maier [Sun, 22 Oct 2006 17:17:47 +0000 (19:17 +0200)]
sysfs: sysfs_write_file() writes zero terminated data

since most of the files in sysfs are text files,
it would be nice, if the "store" function called
during sysfs_write_file() gets a zero terminated
string / data.
The current implementation seems not to ensure this.
(But only if it is the first time the zeroed buffer
page is allocated.)

So the buffer can be scanned by sscanf() easily,
for example.

This patch simply sets a \0 char behind the
data in buffer->page.

Signed-off-by: Thomas Maier <balagi@justmail.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agocpu topology: consider sysfs_create_group return value
Heiko Carstens [Thu, 9 Nov 2006 03:46:09 +0000 (19:46 -0800)]
cpu topology: consider sysfs_create_group return value

Take return value of sysfs_create_group() into account.  That function got
called in case of CPU_ONLINE notification.  Since callbacks are not allowed
to fail on CPU_ONLINE notification do the sysfs group creation on
CPU_UP_PREPARE notification.

Also remember if creation succeeded in a bitmask.  So it's possible to know
whether it's legal to call sysfs_remove_group or not.

In addition some other minor stuff:

- since CPU_UP_PREPARE might fail add CPU_UP_CANCELED handling as well.
- use hotcpu_notifier instead of register_hotcpu_notifier.
- #ifdef code that isn't needed in the !CONFIG_HOTPLUG_CPU case.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: Call platform_notify_remove later
Benjamin Herrenschmidt [Thu, 9 Nov 2006 03:46:14 +0000 (19:46 -0800)]
Driver core: Call platform_notify_remove later

Move the call to platform_notify_remove() to after the call to
bus_remove_device(), where it belongs.  It's bogus to notify the platform
of removal while drivers are still attached to the device and possibly
still operating since the platform might use this callback to tear down
some resources used by the driver (ACPI bits, iommu table, ...)

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "Brown, Len" <len.brown@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoACPI: Change ACPI to use dev_archdata instead of firmware_data
Benjamin Herrenschmidt [Sat, 11 Nov 2006 06:18:42 +0000 (17:18 +1100)]
ACPI: Change ACPI to use dev_archdata instead of firmware_data

Change ACPI to use dev_archdata instead of firmware_data

This patch changes ACPI to use the new dev_archdata on i386, x86_64
and ia64 (is there any other arch using ACPI ?) to store it's
acpi_handle.

It also removes the firmware_data field from struct device as this
was the only user.

Only build-tested on x86

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: add dev_archdata to struct device
Benjamin Herrenschmidt [Sat, 11 Nov 2006 06:18:39 +0000 (17:18 +1100)]
Driver core: add dev_archdata to struct device

Add arch specific dev_archdata to struct device

Adds an arch specific struct dev_arch to struct device. This enables
architecture to add specific fields to every device in the system, like
DMA operation pointers, NUMA node ID, firmware specific data, etc...

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Andi Kleen <ak@suse.de>
Acked-By: David Howells <dhowells@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert sound core to use struct device
Greg Kroah-Hartman [Tue, 8 Aug 2006 05:19:37 +0000 (22:19 -0700)]
Driver core: convert sound core to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

It also makes the struct sound_card to show up as a "real" device
where all the different sound class devices are placed as childs
and different card attribute files can hang off of. /sys/class/sound is
still a flat directory, but the symlink targets of all devices belonging
to the same card, point the the /sys/devices tree below the new card
device object.

Thanks to Kay for the updates to this patch.

Signed-off-by: Kay Sievers <kay.sievers@novell.com>
Acked-by: Jaroslav Kysela <perex@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: change mem class_devices to be real devices
Greg Kroah-Hartman [Wed, 26 Jul 2006 00:13:31 +0000 (17:13 -0700)]
Driver core: change mem class_devices to be real devices

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert fb code to use struct device
Greg Kroah-Hartman [Thu, 14 Sep 2006 14:30:59 +0000 (07:30 -0700)]
Driver core: convert fb code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert firmware code to use struct device
Greg Kroah-Hartman [Thu, 14 Sep 2006 14:30:59 +0000 (07:30 -0700)]
Driver core: convert firmware code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert mmc code to use struct device
Greg Kroah-Hartman [Tue, 12 Sep 2006 15:00:10 +0000 (17:00 +0200)]
Driver core: convert mmc code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert ppdev code to use struct device
Greg Kroah-Hartman [Tue, 12 Sep 2006 15:00:10 +0000 (17:00 +0200)]
Driver core: convert ppdev code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert PPP code to use struct device
Greg Kroah-Hartman [Tue, 12 Sep 2006 15:00:10 +0000 (17:00 +0200)]
Driver core: convert PPP code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert cpuid code to use struct device
Greg Kroah-Hartman [Tue, 8 Aug 2006 05:19:37 +0000 (22:19 -0700)]
Driver core: convert cpuid code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert msr code to use struct device
Greg Kroah-Hartman [Tue, 8 Aug 2006 05:19:37 +0000 (22:19 -0700)]
Driver core: convert msr code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoI2C: convert i2c-dev to use struct device instead of struct class_device
Greg Kroah-Hartman [Mon, 3 Jul 2006 20:46:24 +0000 (13:46 -0700)]
I2C: convert i2c-dev to use struct device instead of struct class_device

As class_device is going away eventually...

Cc: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert raw device code to use struct device
Greg Kroah-Hartman [Tue, 8 Aug 2006 05:19:37 +0000 (22:19 -0700)]
Driver core: convert raw device code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert tty core to use struct device
Greg Kroah-Hartman [Tue, 8 Aug 2006 05:19:37 +0000 (22:19 -0700)]
Driver core: convert tty core to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Also fixes up the isdn drivers that were putting something in the class
device's directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: change misc class_devices to be real devices
Greg Kroah-Hartman [Thu, 27 Jul 2006 23:16:04 +0000 (16:16 -0700)]
Driver core: change misc class_devices to be real devices

This also ment that some of the misc drivers had to also be fixed
up as they were assuming the device was a class_device.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert vc code to use struct device
Greg Kroah-Hartman [Tue, 8 Aug 2006 05:19:37 +0000 (22:19 -0700)]
Driver core: convert vc code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: convert vt code to use struct device
Greg Kroah-Hartman [Tue, 8 Aug 2006 05:19:37 +0000 (22:19 -0700)]
Driver core: convert vt code to use struct device

Converts from using struct "class_device" to "struct device" making
everything show up properly in /sys/devices/ with symlinks from the
/sys/class directory.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoCONFIG_SYSFS_DEPRECATED - class symlinks
Kay Sievers [Thu, 14 Sep 2006 09:23:28 +0000 (11:23 +0200)]
CONFIG_SYSFS_DEPRECATED - class symlinks

Turn off class symlinks CONFIG_SYSFS_DEPRECATED is enabled.

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoCONFIG_SYSFS_DEPRECATED - PHYSDEV* uevent variables
Kay Sievers [Thu, 14 Sep 2006 09:23:28 +0000 (11:23 +0200)]
CONFIG_SYSFS_DEPRECATED - PHYSDEV* uevent variables

Disable the PHYSDEV* uevent variables if CONFIG_SYSFS_DEPRECATED is
enabled.

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoCONFIG_SYSFS_DEPRECATED - device symlinks
Kay Sievers [Thu, 14 Sep 2006 09:23:28 +0000 (11:23 +0200)]
CONFIG_SYSFS_DEPRECATED - device symlinks

Turn off device symlinks CONFIG_SYSFS_DEPRECATED is enabled.

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoCONFIG_SYSFS_DEPRECATED - bus symlinks
Kay Sievers [Thu, 14 Sep 2006 09:23:28 +0000 (11:23 +0200)]
CONFIG_SYSFS_DEPRECATED - bus symlinks

Turn off the bus symlinks if CONFIG_SYSFS_DEPRECATED is enabled

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: make old versions of udev work properly
Greg Kroah-Hartman [Mon, 23 Oct 2006 23:37:58 +0000 (00:37 +0100)]
Driver core: make old versions of udev work properly

If CONFIG_SYSFS_DEPRECATED is enabled, old versions of udev will work
properly with devices that are associated with a class.

Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoCONFIG_SYSFS_DEPRECATED
Kay Sievers [Thu, 14 Sep 2006 09:23:28 +0000 (11:23 +0200)]
CONFIG_SYSFS_DEPRECATED

Provide a way to support older versions of udev that are shipped in
older distros.  If this option is disabled, it will also turn off the
compatible symlinks in sysfs that older programs might rely on.

When in doubt, or if running a distro older than 2006, say Yes here.

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver Core: Move virtual_device_parent() to core.c
Greg Kroah-Hartman [Mon, 23 Oct 2006 17:40:54 +0000 (10:40 -0700)]
Driver Core: Move virtual_device_parent() to core.c

It doesn't need to be global or in device.h

Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: fix "driver" symlink timing
Kay Sievers [Sat, 7 Oct 2006 19:55:55 +0000 (21:55 +0200)]
Driver core: fix "driver" symlink timing

Create the "driver" link before the child device may be created by
the probing logic. This makes it possible for userspace (udev), to
determine the driver property of the parent device, at the time the
child device is created.

Signed-off-by: Kay Sievers <kay.sievers@novell.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoDriver core: add notification of bus events
Benjamin Herrenschmidt [Wed, 25 Oct 2006 03:44:59 +0000 (13:44 +1000)]
Driver core: add notification of bus events

I finally did as you suggested and added the notifier to the struct
bus_type itself. There are still problems to be expected is something
attaches to a bus type where the code can hook in different struct
device sub-classes (which is imho a big bogosity but I won't even try to
argue that case now) but it will solve nicely a number of issues I've
had so far.

That also means that clients interested in registering for such
notifications have to do it before devices are added and after bus types
are registered. Fortunately, most bus types that matter for the various
usage scenarios I have in mind are registerd at postcore_initcall time,
which means I have a really nice spot at arch_initcall time to add my
notifiers.

There are 4 notifications provided. Device being added (before hooked to
the bus) and removed (failure of previous case or after being unhooked
from the bus), along with driver being bound to a device and about to be
unbound.

The usage I have for these are:

 - The 2 first ones are used to maintain a struct device_ext that is
hooked to struct device.firmware_data. This structure contains for now a
pointer to the Open Firmware node related to the device (if any), the
NUMA node ID (for quick access to it) and the DMA operations pointers &
iommu table instance for DMA to/from this device. For bus types I own
(like IBM VIO or EBUS), I just maintain that structure directly from the
bus code when creating the devices. But for bus types managed by generic
code like PCI or platform (actually, of_platform which is a variation of
platform linked to Open Firmware device-tree), I need this notifier.

 - The other two ones have a completely different usage scenario. I have
cases where multiple devices and their drivers depend on each other. For
example, the IBM EMAC network driver needs to attach to a MAL DMA engine
which is a separate device, and a PHY interface which is also a separate
device. They are all of_platform_device's (well, about to be with my
upcoming patches) but there is no say in what precise order the core
will "probe" them and instanciate the various modules. The solution I
found for that is to have the drivers for emac to use multithread_probe,
and wait for a driver to be bound to the target MAL and PHY control
devices (the device-tree contains reference to the MAL and PHY interface
nodes, which I can then match to of_platform_devices). Right now, I've
been polling, but with that notifier, I can more cleanly wait (with a
timeout of course).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: make arch/i386/pci/common.c:pci_bf_sort static
Adrian Bunk [Thu, 16 Nov 2006 12:16:23 +0000 (13:16 +0100)]
PCI: make arch/i386/pci/common.c:pci_bf_sort static

This patch makes the needlessly global pci_bf_sort static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: ibmphp_pci.c: fix NULL dereference
Adrian Bunk [Fri, 17 Nov 2006 01:19:25 +0000 (02:19 +0100)]
PCI: ibmphp_pci.c: fix NULL dereference

The correct order is: NULL check before dereference

Spotted by the Coverity checker.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agopciehp: remove unnecessary pci_disable_msi
Kenji Kaneshige [Mon, 13 Nov 2006 23:13:09 +0000 (15:13 -0800)]
pciehp: remove unnecessary pci_disable_msi

This patch fixes the problem that "irq XX: nobody cared" kernel oops
is reported when pciehp is once rmmoded and insmoded again. The cause
of this problem is pciehp driver calls pci_disable_msi() at controller
release time, even though it must be done by PCI Express Port Bus
driver. This patch removes unnecessary pci_disable_msi() call from
pciehp driver.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agopciehp: remove unnecessary free_irq
Kenji Kaneshige [Mon, 13 Nov 2006 23:13:00 +0000 (15:13 -0800)]
pciehp: remove unnecessary free_irq

This patch fixes the problem that the following error messages is
reported when pciehp driver is rmmoded.

Trying to free already-free IRQ XX

The cause of this problem is that pciehp driver is doing unknown 2nd
free_irq at driver unloading. This patch removes this unknown 2nd
free_irq call.

Note: The pciehp driver should be adapted to standard device driver
mode.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: rpaphp: change device tree examination
John Rose [Mon, 13 Nov 2006 23:12:52 +0000 (15:12 -0800)]
PCI: rpaphp: change device tree examination

Change the criterion that RPA PCI Hotplug and RPA DLPAR use when
determining the hotplug capabilities of a given device node.  The
"device_type" property is less consistent than "name" across PCI nodes
on newer hardware.

Signed-off-by: John Rose <johnrose@austin.ibm.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: Change memory allocation for acpiphp slots
Rolf Eike Beer [Mon, 13 Nov 2006 23:12:45 +0000 (15:12 -0800)]
PCI: Change memory allocation for acpiphp slots

Change memory allocation for acpiphp slots

Change the "struct slot" that acpiphp uses for managing it's slots to
directly contain the memory for the needed struct hotplug_slot_info and
the slot's name. This way we need only two memory allocations per slot
instead of four.

While we are at it: make_slot_name() is just a wrapper around snprintf()
knowing the right arguments to call it. Since the function makes just one
function call and is only called from one place I inlined it by hand.

Finally this fixes a possible bug waiting for someone to hit it. There were
two unused local variables in acpiphp_register_hotplug_slot(). gcc did not
find them because they were used in memory allocations with sizeof(*var).
They had the same types as the target of the allocation, but nevertheless
this was just weird.

Signed-off-by: Rolf Eike Beer <eike-hotplug@sf-tec.de>
Acked-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoi2c-i801: SMBus patch for Intel ICH9
Jason Gaston [Wed, 22 Nov 2006 23:19:12 +0000 (15:19 -0800)]
i2c-i801: SMBus patch for Intel ICH9

This updated patch adds the Intel ICH9 LPC and SMBus Controller DID's.  Thi=
s patch relies on the irq ICH9 patch to pci_ids.h.

Signed-off-by: Jason Gaston <jason.d.gaston@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: irq: irq and pci_ids patch for Intel ICH9
Jason Gaston [Wed, 22 Nov 2006 23:15:08 +0000 (15:15 -0800)]
PCI: irq: irq and pci_ids patch for Intel ICH9

This updated patch adds the Intel ICH9 LPC and SMBus Controller DID's.

Signed-off-by: Jason Gaston <jason.d.gaston@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: pci_{enable,disable}_device() nestable ports
Inaky Perez-Gonzalez [Wed, 22 Nov 2006 20:40:32 +0000 (12:40 -0800)]
PCI: pci_{enable,disable}_device() nestable ports

Change drivers/message/i20 pci driver to simply do a nestable
enable()/disable() instead of checking for it.

Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: switch pci_{enable,disable}_device() to be nestable
Inaky Perez-Gonzalez [Wed, 22 Nov 2006 20:40:31 +0000 (12:40 -0800)]
PCI: switch pci_{enable,disable}_device() to be nestable

Changes the pci_{enable,disable}_device() functions to work in a
nested basis, so that eg, three calls to enable_device() require three
calls to disable_device().

The reason for this is to simplify PCI drivers for
multi-interface/capability devices. These are devices that cram more
than one interface in a single function. A relevant example of that is
the Wireless [USB] Host Controller Interface (similar to EHCI) [see
http://www.intel.com/technology/comms/wusb/whci.htm].

In these kind of devices, multiple interfaces are accessed through a
single bar and IRQ line. For that, the drivers map only the smallest
area of the bar to access their register banks and use shared IRQ
handlers.

However, because the order at which those drivers load cannot be known
ahead of time, the sequence in which the calls to pci_enable_device()
and pci_disable_device() cannot be predicted. Thus:

1. driverA     starts     pci_enable_device()
2. driverB     starts     pci_enable_device()
3. driverA     shutdown   pci_disable_device()
4. driverB     shutdown   pci_disable_device()

between steps 3 and 4, driver B would loose access to it's device,
even if it didn't intend to.

By using this modification, the device won't be disabled until all the
callers to enable() have called disable().

This is implemented by replacing 'struct pci_dev->is_enabled' from a
bitfield to an atomic use count. Each caller to enable increments it,
each caller to disable decrements it. When the count increments from 0
to 1, __pci_enable_device() is called to actually enable the
device. When it drops to zero, pci_disable_device() actually does the
disabling.

We keep the backend __pci_enable_device() for pci_default_resume() to
use and also change the sysfs method implementation, so that userspace
enabling/disabling the device doesn't disable it one time too much.

Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: arch/i386/kernel/pci-dma.c: ioremap balanced with iounmap
Amol Lad [Tue, 17 Oct 2006 04:32:50 +0000 (10:02 +0530)]
PCI: arch/i386/kernel/pci-dma.c: ioremap balanced with iounmap

ioremap must be balanced by an iounmap and failing to do so can result
in a memory leak.

Tested (compilation only):
- using allmodconfig
- making sure the files are compiling without any warning/error due to
new changes

Signed-off-by: Amol Lad <amol@verismonetworks.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agopci/i386: style cleanups
Randy Dunlap [Tue, 17 Oct 2006 17:17:58 +0000 (10:17 -0700)]
pci/i386: style cleanups

Mostly CodingStyle cleanups for arch/i386/pci/i386.c:
- fit in 80 columns;
- use a #defined value instead of an inline constant;
Also change one resource_size_t (DBG) printk from %08lx to %lx since
it can be more than 32 bits (more than 8 hexits).

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: Block on access to temporarily unavailable pci device
Matthew Wilcox [Thu, 19 Oct 2006 15:41:28 +0000 (09:41 -0600)]
PCI: Block on access to temporarily unavailable pci device

The existing implementation of pci_block_user_cfg_access() was recently
criticised for providing out of date information and for returning errors
on write, which applications won't be expecting.

This reimplementation uses a global wait queue and a bit per device.
I've open-coded prepare_to_wait() / finish_wait() as I could optimise
it significantly by knowing that the pci_lock protected us at all points.

It looked a bit funny to be doing a spin_unlock_irqsave(); schedule(),
so I used spin_lock_irq() for the _user versions of pci_read_config and
pci_write_config.  Not carrying a flags pointer around made the code
much less nasty.

Attempts to block an already blocked device hit a BUG() and attempts to
unblock an already unblocked device hit a WARN().  If we need to block
access to a device from userspace, it's because it's unsafe for even
another bit of the kernel to access the device.  An attempt to block
a device for a second time means we're about to access the device to
perform some other operation, which could provoke undefined behaviour
from the device.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Acked-by: Adam Belay <abelay@novell.com>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agopci: fix __pci_register_driver error handling
Akinobu Mita [Thu, 9 Nov 2006 03:53:59 +0000 (19:53 -0800)]
pci: fix __pci_register_driver error handling

__pci_register_driver() error path forgot to unwind.
driver_unregister() needs to be called when pci_create_newid_file() failed.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agopci: clear osc support flags if no _OSC method
Kristen Carlson Accardi [Mon, 30 Oct 2006 21:08:12 +0000 (13:08 -0800)]
pci: clear osc support flags if no _OSC method

So it looks like pci aer code will call pci_osc_support_set to tell the
firmware about  OSC_EXT_PCI_CONFIG_SUPPORT flag.  that causes
ctrlset_buf[OSC_SUPPORT_TYPE] to evaluate to true when pciehp calls
pci_osc_control_set() is called (to attempt to use OSC to gain native
pcie control from firmware), regardless of whether or not _OSC was
actually successfully executed.  That causes this section of code:
 if (ctrlset_buf[OSC_SUPPORT_TYPE] &&
                ((global_ctrlsets & ctrlset) != ctrlset)) {
                return AE_SUPPORT;
        }
to be hit.

This patch will reset the OSC_SUPPORT_TYPE field if _OSC fails, and then
would allow pciehp to go ahead and try to run _OSC again.

Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoacpiphp: fix missing acpiphp_glue_exit()
Akinobu Mita [Mon, 30 Oct 2006 21:08:04 +0000 (13:08 -0800)]
acpiphp: fix missing acpiphp_glue_exit()

acpiphp_glue_exit() needs to be called to unwind when no slots found.
(It fixes data corruption when reloading acpiphp driver with no such devices)

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoacpiphp: fix use of list_for_each macro
Akinobu Mita [Mon, 30 Oct 2006 21:07:58 +0000 (13:07 -0800)]
acpiphp: fix use of list_for_each macro

This patch fixes invalid usage of list_for_each()

list_for_each (node, &bridge_list) {
bridge = (struct acpiphp_bridge *)node;
...
}

This code works while the member of list node is located at the
head of struct acpiphp_bridge.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoAltix: Initial ACPI support - ROM shadowing.
John Keller [Wed, 4 Oct 2006 21:49:52 +0000 (16:49 -0500)]
Altix: Initial ACPI support - ROM shadowing.

Support a shadowed ROM when running with an ACPI capable PROM.

Define a new dev.resource flag IORESOURCE_ROM_BIOS_COPY to
describe the case of a BIOS shadowed ROM, which can then
be used to avoid pci_map_rom() making an unneeded call to
pci_enable_rom().

Signed-off-by: John Keller <jpk@sgi.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoAltix: SN ACPI hotplug support.
John Keller [Wed, 4 Oct 2006 21:49:35 +0000 (16:49 -0500)]
Altix: SN ACPI hotplug support.

A few minor changes to the way slot/device fixup is done.

No need to be calling sn_pci_controller_fixup(), as
a root bus cannot be hotplugged.

Signed-off-by: John Keller <jpk@sgi.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoAltix: Add initial ACPI IO support
John Keller [Wed, 4 Oct 2006 21:49:25 +0000 (16:49 -0500)]
Altix: Add initial ACPI IO support

First phase in introducing ACPI support to SN.
In this phase, when running with an ACPI capable PROM,
the DSDT will define the root busses and all SN nodes
(SGIHUB, SGITIO). An ACPI bus driver will be registered
for the node devices, with the acpi_pci_root_driver being
used for the root busses. An ACPI vendor descriptor is
now used to pass platform specific information for both
nodes and busses, eliminating the need for the current
SAL calls. Also, with ACPI support, SN fixup code is no longer
needed to initiate the PCI bus scans, as the acpi_pci_root_driver
does that.

However, to maintain backward compatibility with non-ACPI capable
PROMs, none of the current 'fixup' code can been deleted, though
much restructuring has been done. For example, the bulk of the code
in io_common.c is relocated code that is now common regardless
of what PROM is running, while io_acpi_init.c and io_init.c contain
routines specific to an ACPI or non ACPI capable PROM respectively.

A new pci bus fixup platform vector has been created to provide
a hook for invoking platform specific bus fixup from pcibios_fixup_bus().

The size of io_space[] has been increased to support systems with
large IO configurations.

Signed-off-by: John Keller <jpk@sgi.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: Delete unused extern in powermac/pci.c
Matthew Wilcox [Tue, 10 Oct 2006 14:01:22 +0000 (08:01 -0600)]
PCI: Delete unused extern in powermac/pci.c

This file no longer uses pci_cache_line_size, so delete the declaration

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: Replace HAVE_ARCH_PCI_MWI with PCI_DISABLE_MWI
Matthew Wilcox [Tue, 10 Oct 2006 14:01:21 +0000 (08:01 -0600)]
PCI: Replace HAVE_ARCH_PCI_MWI with PCI_DISABLE_MWI

pSeries is the only architecture left using HAVE_ARCH_PCI_MWI and it's
really inappropriate for its needs.  It really wants to disable MWI
altogether.  So here are a pair of stub implementations for pci_set_mwi
and pci_clear_mwi.

Also rename pci_generic_prep_mwi to pci_set_cacheline_size since that
better reflects what it does.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Cc: Paul Mackerras <paulus@samba.org>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: Use pci_generic_prep_mwi on sparc64
Matthew Wilcox [Tue, 10 Oct 2006 14:01:20 +0000 (08:01 -0600)]
PCI: Use pci_generic_prep_mwi on sparc64

The setting of the CACHE_LINE_SIZE register in sparc64's pci
initialisation code isn't quite adequate as the device may have
incompatible requirements.  The generic code tests for this, so switch
sparc64 over to using it.

Since sparc64 has different L1 cache line size and PCI cache line size,
it would need to override the generic code like i386 and ia64 do.  We
know what the cache line size is at compile time though, so introduce a
new optional constant PCI_CACHE_LINE_BYTES.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: David Miller <davem@davemloft.net>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
18 years agoPCI: Use pci_generic_prep_mwi on ia64
Matthew Wilcox [Tue, 10 Oct 2006 14:01:19 +0000 (08:01 -0600)]
PCI: Use pci_generic_prep_mwi on ia64

The pci_generic_prep_mwi() code does everything that pcibios_prep_mwi()
does on ia64.  All we need to do is be sure that pci_cache_line_size
is set appropriately, and we can delete pcibios_prep_mwi().

Using SMP_CACHE_BYTES as the default was wrong on uniprocessor machines
as it is only 8 bytes.  The default in the generic code of L1_CACHE_BYTES
is at least as good.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>