safe/jmp/linux-2.6
15 years agomutex: speed up generic mutex implementations
Nick Piggin [Tue, 21 Oct 2008 08:59:15 +0000 (10:59 +0200)]
mutex: speed up generic mutex implementations

- atomic operations which both modify the variable and return something imply
  full smp memory barriers before and after the memory operations involved
  (failing atomic_cmpxchg, atomic_add_unless, etc don't imply a barrier because
  they don't modify the target). See Documentation/atomic_ops.txt.
  So remove extra barriers and branches.

- All architectures support atomic_cmpxchg. This has no relation to
  __HAVE_ARCH_CMPXCHG. We can just take the atomic_cmpxchg path unconditionally

This reduces a simple single threaded fastpath lock+unlock test from 590 cycles
to 203 cycles on a ppc970 system.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/czankel/xtensa-2.6
Linus Torvalds [Thu, 23 Oct 2008 16:16:56 +0000 (09:16 -0700)]
Merge git://git./linux/kernel/git/czankel/xtensa-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/czankel/xtensa-2.6:
  xtensa: Add config files for Diamond 232L - Rev B processor variant
  xtensa: Fix io regions
  xtensa: Add support for the Sonic Ethernet device for the XT2000 board.
  xtensa: replace remaining __FUNCTION__ occurrences
  xtensa: use newer __SPIN_LOCK_UNLOCKED macro
  XTENSA: warn about including <asm/rwsem.h> directly.

15 years agomemcg: fix page_cgroup allocation
KAMEZAWA Hiroyuki [Wed, 22 Oct 2008 21:15:05 +0000 (14:15 -0700)]
memcg: fix page_cgroup allocation

page_cgroup_init() is called from mem_cgroup_init(). But at this
point, we cannot call alloc_bootmem().
(and this caused panic at boot.)

This patch moves page_cgroup_init() to init/main.c.

Time table is following:
==
  parse_args(). # we can trust mem_cgroup_subsys.disabled bit after this.
  ....
  cgroup_init_early()  # "early" init of cgroup.
  ....
  setup_arch()         # memmap is allocated.
  ...
  page_cgroup_init();
  mem_init();   # we cannot call alloc_bootmem after this.
  ....
  cgroup_init() # mem_cgroup is initialized.
==

Before page_cgroup_init(), mem_map must be initialized. So,
I added page_cgroup_init() to init/main.c directly.

(*) maybe this is not very clean but
    - cgroup_init_early() is too early
    - in cgroup_init(), we have to use vmalloc instead of alloc_bootmem().
    use of vmalloc area in x86-32 is important and we should avoid very large
    vmalloc() in x86-32. So, we want to use alloc_bootmem() and added page_cgroup_init()
    directly to init/main.c

[akpm@linux-foundation.org: remove unneeded/bad mem_cgroup_subsys declaration]
[akpm@linux-foundation.org: fix build]
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Tested-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agojbd: abort instead of waiting for nonexistent transactions
Duane Griffin [Wed, 22 Oct 2008 21:15:03 +0000 (14:15 -0700)]
jbd: abort instead of waiting for nonexistent transactions

The __log_wait_for_space function sits in a loop checkpointing
transactions until there is sufficient space free in the journal.
However, if there are no transactions to be processed (e.g.  because the
free space calculation is wrong due to a corrupted filesystem) it will
never progress.

Check for space being required when no transactions are outstanding and
abort the journal instead of endlessly looping.

This patch fixes the bug reported by Sami Liedes at:
http://bugzilla.kernel.org/show_bug.cgi?id=10976

Signed-off-by: Duane Griffin <duaneg@dghda.com>
Tested-by: Sami Liedes <sliedes@cc.hut.fi>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agojbd: test BH_Write_EIO to detect errors on metadata buffers
Hidehiro Kawai [Wed, 22 Oct 2008 21:15:02 +0000 (14:15 -0700)]
jbd: test BH_Write_EIO to detect errors on metadata buffers

__try_to_free_cp_buf(), __process_buffer(), and __wait_cp_io() test
BH_Uptodate flag to detect write I/O errors on metadata buffers.  But by
commit 95450f5a7e53d5752ce1a0d0b8282e10fe745ae0 "ext3: don't read inode
block if the buffer has a write error"(*), BH_Uptodate flag can be set to
inode buffers with BH_Write_EIO in order to avoid reading old inode data.
So now, we have to test BH_Write_EIO flag of checkpointing inode buffers
instead of BH_Uptodate.  This patch does it.

Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Acked-by: Jan Kara <jack@suse.cz>
Acked-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoext3: add checks for errors from jbd
Hidehiro Kawai [Wed, 22 Oct 2008 21:15:01 +0000 (14:15 -0700)]
ext3: add checks for errors from jbd

If the journal has aborted due to a checkpointing failure, we have to
keep the contents of the journal space.  Otherwise, the filesystem will
lose uncheckpointed metadata completely and become inconsistent.  To
avoid this, we need to keep needs_recovery flag if checkpoint has
failed.

With this patch, ext3_put_super() detects a checkpointing failure from
the return value of journal_destroy(), then it invokes ext3_abort() to
make the filesystem read only and keep needs_recovery flag.  Errors
from journal_flush() are also handled by this patch in some places.

Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Cc: Jan Kara <jack@ucw.cz>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agojbd: fix error handling for checkpoint io
Hidehiro Kawai [Wed, 22 Oct 2008 21:15:00 +0000 (14:15 -0700)]
jbd: fix error handling for checkpoint io

When a checkpointing IO fails, current JBD code doesn't check the error
and continue journaling.  This means latest metadata can be lost from both
the journal and filesystem.

This patch leaves the failed metadata blocks in the journal space and
aborts journaling in the case of log_do_checkpoint().  To achieve this, we
need to do:

1. don't remove the failed buffer from the checkpoint list where in
   the case of __try_to_free_cp_buf() because it may be released or
   overwritten by a later transaction
2. log_do_checkpoint() is the last chance, remove the failed buffer
   from the checkpoint list and abort the journal
3. when checkpointing fails, don't update the journal super block to
   prevent the journaled contents from being cleaned.  For safety,
   don't update j_tail and j_tail_sequence either
4. when checkpointing fails, notify this error to the ext3 layer so
   that ext3 don't clear the needs_recovery flag, otherwise the
   journaled contents are ignored and cleaned in the recovery phase
5. if the recovery fails, keep the needs_recovery flag
6. prevent cleanup_journal_tail() from being called between
   __journal_drop_transaction() and journal_abort() (a race issue
   between journal_flush() and __log_wait_for_space()

Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Acked-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoprofiling: fix up CONFIG_PROC_FS=n build
Paul Mundt [Wed, 22 Oct 2008 21:14:59 +0000 (14:14 -0700)]
profiling: fix up CONFIG_PROC_FS=n build

In the case where procfs is disabled, create_proc_profile() does not
exist. Stub it in with the others.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: page_cgroup needs linux/vmalloc.h for vmalloc_node()/vfree().
Paul Mundt [Wed, 22 Oct 2008 21:14:58 +0000 (14:14 -0700)]
mm: page_cgroup needs linux/vmalloc.h for vmalloc_node()/vfree().

mm/page_cgroup.c: In function 'init_section_page_cgroup':
mm/page_cgroup.c:111: error: implicit declaration of function 'vmalloc_node'
mm/page_cgroup.c:111: warning: assignment makes pointer from integer without a cast
mm/page_cgroup.c: In function '__free_page_cgroup':
mm/page_cgroup.c:140: error: implicit declaration of function 'vfree'

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Thu, 23 Oct 2008 15:28:25 +0000 (08:28 -0700)]
Merge git://git./linux/kernel/git/benh/powerpc

* git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc: (53 commits)
  powerpc: Support for relocatable kdump kernel
  powerpc: Don't use a 16G page if beyond mem= limits
  powerpc: Add del_node() for early boot code to prune inapplicable devices.
  powerpc: Further compile fixup for STRICT_MM_TYPECHECKS
  powerpc: Remove empty #else from signal_64.c
  powerpc: Move memory size print into common show_cpuinfo for 32-bit
  hvc_console: Remove __devexit annotation of hvc_remove()
  hvc_console: Add support for tty window resizing
  hvc_console: Fix loop if put_char() returns 0
  hvc_console: Add tty driver flag TTY_DRIVER_RESET_TERMIOS
  hvc_console: Add a hangup notifier for backends
  powerpc/83xx: Add DS1339 RTC support for MPC8349E-mITX boards .dts
  powerpc/83xx: Add support for MCU microcontroller in .dts files
  powerpc/85xx: Move mpc8572ds.dts to address-cells/size-cells = <2>
  of/spi: Support specifying chip select as active high via device tree
  powerpc: Remove device_type = "board_control" properties in .dts files
  i2c-cpm: Suppress autoprobing for devices
  powerpc/85xx: Fix mpc8536ds dma interrupt numbers
  powerpc/85xx: Enable enhanced functions for 8536 TSEC
  powerpc: Delete unused prom_strtoul and prom_memparse
  ...

15 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/dvrabel/uwb
Linus Torvalds [Thu, 23 Oct 2008 15:20:34 +0000 (08:20 -0700)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/dvrabel/uwb

* 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/dvrabel/uwb: (47 commits)
  uwb: wrong sizeof argument in mac address compare
  uwb: don't use printk_ratelimit() so often
  uwb: use kcalloc where appropriate
  uwb: use time_after() when purging stale beacons
  uwb: add credits for the original developers of the UWB/WUSB/WLP subsystems
  uwb: add entries in the MAINTAINERS file
  uwb: depend on EXPERIMENTAL
  wusb: wusb-cbaf (CBA driver) sysfs ABI simplification
  uwb: document UWB and WUSB sysfs files
  uwb: add symlinks in sysfs between radio controllers and PALs
  uwb: dont tranmit identification IEs
  uwb: i1480/GUWA100U: fix firmware download issues
  uwb: i1480: remove MAC/PHY information checking function
  uwb: add Intel i1480 HWA to the UWB RC quirk table
  uwb: disable command/event filtering for D-Link DUB-1210
  uwb: initialize the debug sub-system
  uwb: Fix handling IEs with empty IE data in uwb_est_get_size()
  wusb: fix bmRequestType for Abort RPipe request
  wusb: fix error path for wusb_set_dev_addr()
  wusb: add HWA host controller driver
  ...

15 years agoMerge branch 'for-next' of git://git.o-hand.com/linux-mfd
Linus Torvalds [Thu, 23 Oct 2008 15:18:33 +0000 (08:18 -0700)]
Merge branch 'for-next' of git://git.o-hand.com/linux-mfd

* 'for-next' of git://git.o-hand.com/linux-mfd:
  mfd: check for platform_get_irq() return value in sm501
  mfd: use pci_ioremap_bar() in sm501
  mfd: Don't store volatile bits in WM8350 register cache
  mfd: don't export wm3850 static functions
  mfd: twl4030-gpio driver
  mfd: rtc-twl4030 driver
  mfd: twl4030 IRQ handling update

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Thu, 23 Oct 2008 15:16:03 +0000 (08:16 -0700)]
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:
  IB/ehca: Reject dynamic memory add/remove when ehca adapter is present
  IB/ehca: Fix reported max number of QPs and CQs in systems with >1 adapter
  IPoIB: Set netdev offload features properly for child (VLAN) interfaces
  IPoIB: Clean up ethtool support
  mlx4_core: Add Ethernet PCI device IDs
  mlx4_en: Add driver for Mellanox ConnectX 10GbE NIC
  mlx4_core: Multiple port type support
  mlx4_core: Ethernet MAC/VLAN management
  mlx4_core: Get ethernet MTU and default address from firmware
  mlx4_core: Support multiple pre-reserved QP regions
  Update NetEffect maintainer emails to Intel emails
  RDMA/cxgb3: Remove cmid reference on tid allocation failures
  IB/mad: Use krealloc() to resize snoop table
  IPoIB: Always initialize poll_timer to avoid crash on unload
  IB/ehca: Don't allow creating UC QP with SRQ
  mlx4_core: Add QP range reservation support
  RDMA/ucma: Test ucma_alloc_multicast() return against NULL, not with IS_ERR()

15 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Thu, 23 Oct 2008 15:15:29 +0000 (08:15 -0700)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  sata_via: load DEVICE register when CTL changes
  libata: set device class to NONE if phys_offline
  libata-eh: fix slave link EH action mask handling
  libata: transfer EHI control flags to slave ehc.i
  libata-sff: fix ata_sff_post_internal_cmd()
  libata: initialize port_task when !CONFIG_ATA_SFF

15 years agoMerge branch 'for-linus' of master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Thu, 23 Oct 2008 15:12:21 +0000 (08:12 -0700)]
Merge branch 'for-linus' of /home/rmk/linux-2.6-arm

* 'for-linus' of master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] clps711x: add sparsemem definitions
  [ARM] 5315/1: Fix section mismatch warning (sa1111)
  [ARM] Orion: activate workaround for 88f6183 SPI clock erratum
  [ARM] Orion: instantiate the dsa switch driver
  [ARM] mv78xx0: force link speed/duplex on eth2/eth3
  [ARM] remove extra brace in arch/arm/mach-pxa/trizeps4.c
  [ARM] balance parenthesis in header file
  [ARM] pxa: fix trizeps PCMCIA build
  [ARM] pxa: fix trizeps defconfig
  [ARM] dmabounce requires ZONE_DMA
  [ARM] 5303/1: period_cycles should be greater than 1
  [ARM] 5310/1: Fix cache flush functions for ARMv4
  [ARM] pxa: fix 3bca103a1e658d23737d20e1989139d9ca8973bf
  [ARM] pxa: fix redefinition of NR_IRQS
  [ARM] S3C24XX: Fix redefine of DEFINE_TIMER() in s3c24xx pwm-clock.c
  [ARM] S3C2443: Fix HCLK rate
  [ARM] S3C24XX: Serial driver debug depends on DEBUG_LL
  [ARM] S3C24XX: pwm-clock set_parent mask fix

15 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Thu, 23 Oct 2008 15:07:35 +0000 (08:07 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6: (41 commits)
  [IA64] Fix annoying IA64_TR_ALLOC_MAX message.
  [IA64] kill sys32_pipe
  [IA64] remove sys32_pause
  [IA64] Add Variable Page Size and IA64 Support in Intel IOMMU
  ia64/pv_ops: paravirtualized instruction checker.
  ia64/xen: a recipe for using xen/ia64 with pv_ops.
  ia64/pv_ops: update Kconfig for paravirtualized guest and xen.
  ia64/xen: preliminary support for save/restore.
  ia64/xen: define xen machine vector for domU.
  ia64/pv_ops/xen: implement xen pv_time_ops.
  ia64/pv_ops/xen: implement xen pv_irq_ops.
  ia64/pv_ops/xen: define the nubmer of irqs which xen needs.
  ia64/pv_ops/xen: implement xen pv_iosapic_ops.
  ia64/pv_ops/xen: paravirtualize entry.S for ia64/xen.
  ia64/pv_ops/xen: paravirtualize ivt.S for xen.
  ia64/pv_ops/xen: paravirtualize DO_SAVE_MIN for xen.
  ia64/pv_ops/xen: define xen paravirtualized instructions for hand written assembly code
  ia64/pv_ops/xen: define xen pv_cpu_ops.
  ia64/pv_ops/xen: define xen pv_init_ops for various xen initialization.
  ia64/pv_ops/xen: elf note based xen startup.
  ...

15 years agosata_via: load DEVICE register when CTL changes
Tejun Heo [Tue, 21 Oct 2008 15:45:57 +0000 (00:45 +0900)]
sata_via: load DEVICE register when CTL changes

VIA controllers clear DEVICE register when IEN changes.  Make sure
DEVICE is updated along with CTL.

This change is separated from Joseph Chan's larger patch.

  http://thread.gmane.org/gmane.linux.kernel.commits.mm/40640

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Joseph Chan <JosephChan@via.com.tw>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: set device class to NONE if phys_offline
Tejun Heo [Tue, 21 Oct 2008 15:31:34 +0000 (00:31 +0900)]
libata: set device class to NONE if phys_offline

Reset methods don't have access to phys link status for slave links
and may incorrectly indicate device presence causing unnecessary probe
failures for unoccupied links.  This patch clears device class to NONE
during post-reset processing if phys link is offline.

As on/offlineness semantics is strictly defined and used in multiple
places by the core layer, this won't change behavior for drivers which
don't use slave links.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata-eh: fix slave link EH action mask handling
Tejun Heo [Tue, 21 Oct 2008 11:37:21 +0000 (20:37 +0900)]
libata-eh: fix slave link EH action mask handling

Slave link action mask is transferred to master link and all the EH
actions are taken by the master link.  ata_eh_about_to_do() and
ata_eh_done() are called with ATA_EH_ALL_ACTIONS to clear the slave
link actions during transfer.  This always sets ATA_PFLAG_RECOVERED
flag causing spurious "EH complete" messages.

Don't set ATA_PFLAG_RECOVERED for slave link actions.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: transfer EHI control flags to slave ehc.i
Tejun Heo [Tue, 21 Oct 2008 05:26:39 +0000 (14:26 +0900)]
libata: transfer EHI control flags to slave ehc.i

ATA_EHI_NO_AUTOPSY and ATA_EHI_QUIET are used to control the behavior
of EH.  As only the master link is visible outside EH, these flags are
set only for the master link although they should also apply to the
slave link, which causes spurious EH messages during probe and
suspend/resume.

This patch transfers those two flags to slave ehc.i before performing
slave autopsy and reporting.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata-sff: fix ata_sff_post_internal_cmd()
Tejun Heo [Mon, 20 Oct 2008 04:10:21 +0000 (13:10 +0900)]
libata-sff: fix ata_sff_post_internal_cmd()

ata_sff_post_internal_cmd() needs to grab port lock before calling
ata_bmdma_stop() and also need to clear hsm_task_state.  Fix it.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: initialize port_task when !CONFIG_ATA_SFF
Tejun Heo [Mon, 20 Oct 2008 04:11:56 +0000 (13:11 +0900)]
libata: initialize port_task when !CONFIG_ATA_SFF

ap->port_task was not initialized if !CONFIG_ATA_SFF later triggering
lockdep warning.  Make sure it's initialized.

Reported by Larry Finger.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agoMerge branches 'cma', 'cxgb3', 'ehca', 'ipoib', 'mad', 'mlx4' and 'nes' into for...
Roland Dreier [Wed, 22 Oct 2008 22:56:41 +0000 (15:56 -0700)]
Merge branches 'cma', 'cxgb3', 'ehca', 'ipoib', 'mad', 'mlx4' and 'nes' into for-next

15 years agoIB/ehca: Reject dynamic memory add/remove when ehca adapter is present
Stefan Roscher [Wed, 22 Oct 2008 22:54:38 +0000 (15:54 -0700)]
IB/ehca: Reject dynamic memory add/remove when ehca adapter is present

Since the ehca device driver does not support dynamic memory add and
remove operations, the driver must explicitly reject such requests in
order to prevent unpredictable behaviors related to existing memory
regions that cover all of memory being used by InfiniBand protocols in
the kernel.

The solution (for now at least) is to add a memory notifier to the
ehca device driver and if a request for dynamic memory add or remove
comes in, ehca will always reject it.  The user can add or remove
memory by hot-removing the ehca adapter, performing the memory
operation, and then hot-adding the ehca adapter back.

Signed-off-by: Stefan Roscher <stefan.roscher@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agoIB/ehca: Fix reported max number of QPs and CQs in systems with >1 adapter
Stefan Roscher [Wed, 22 Oct 2008 22:52:31 +0000 (15:52 -0700)]
IB/ehca: Fix reported max number of QPs and CQs in systems with >1 adapter

Because ehca adapters can differ in the maximum number of QPs and CQs
we have to save the maximum number of these ressources per adapter and
not globally per ehca driver. This fix introduces 2 new members to the
shca structure to store the maximum value for QPs and CQs per adapter.

The module parameters are now used as initial values for those
variables.  If a user selects an invalid number of CQs or QPs we don't
print an error any longer, instead we will inform the user with a
warning and set the values to the respective maximum supported by the
HW.

Signed-off-by: Stefan Roscher <stefan.roscher@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agoIPoIB: Set netdev offload features properly for child (VLAN) interfaces
Or Gerlitz [Wed, 22 Oct 2008 22:49:49 +0000 (15:49 -0700)]
IPoIB: Set netdev offload features properly for child (VLAN) interfaces

Child devices were created without any offload features set, fix this by
moving the code that computes the features into generic function which is
now called through non-child and child device creation.

Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com>
-- v1 has a bug where the 'result' flag in ipoib_vlan_add may be used uninitialized
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agoIPoIB: Clean up ethtool support
Or Gerlitz [Wed, 22 Oct 2008 22:49:29 +0000 (15:49 -0700)]
IPoIB: Clean up ethtool support

Add a get_rx_csum method.  Remove the driver's own get_tso method, as
the ethtool kernel code uses the default one if nothing is provided.

Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agomlx4_core: Add Ethernet PCI device IDs
Yevgeny Petrilin [Wed, 22 Oct 2008 22:48:03 +0000 (15:48 -0700)]
mlx4_core: Add Ethernet PCI device IDs

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agomlx4_en: Add driver for Mellanox ConnectX 10GbE NIC
Yevgeny Petrilin [Wed, 22 Oct 2008 22:47:49 +0000 (15:47 -0700)]
mlx4_en: Add driver for Mellanox ConnectX 10GbE NIC

The Mellanox ConnectX can operate as an InfiniBand adapter, as an
Ethernet NIC, or as a Fibre Channel (FC) HBA.  The kernel has a
low-level driver, mlx4_core, which handles multiplexing access to the
device, and there is also already an InfiniBad driver, mlx4_ib.

This patch adds a new driver, mlx4_en, which implements a standard
Ethernet NIC driver.

Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agomlx4_core: Multiple port type support
Yevgeny Petrilin [Wed, 22 Oct 2008 22:38:42 +0000 (15:38 -0700)]
mlx4_core: Multiple port type support

Multi-protocol adapters support different port types.  Each consumer
of mlx4_core queries for supported port types; in particular mlx4_ib
can no longer assume that all physical ports belong to it.  Port type
is configured through a sysfs interface.  When the type of a port is
changed, all mlx4 interfaces are unregistered, and then registered
again with the new port types.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agomlx4_core: Ethernet MAC/VLAN management
Yevgeny Petrilin [Wed, 22 Oct 2008 18:44:46 +0000 (11:44 -0700)]
mlx4_core: Ethernet MAC/VLAN management

Add support for managing MAC and VLAN filters for each port.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Oren Duer <oren@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agomlx4_core: Get ethernet MTU and default address from firmware
Yevgeny Petrilin [Wed, 22 Oct 2008 17:56:48 +0000 (10:56 -0700)]
mlx4_core: Get ethernet MTU and default address from firmware

Get maximum ethernet MTU and default MAC address from the firmware
QUERY_DEV_CAP command.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agomlx4_core: Support multiple pre-reserved QP regions
Yevgeny Petrilin [Wed, 22 Oct 2008 17:25:29 +0000 (10:25 -0700)]
mlx4_core: Support multiple pre-reserved QP regions

For ethernet support, we need to reserve QPs for the ethernet and
fibre channel driver.  The QPs are reserved at the end of the QP
table.  (This way we assure that they are aligned to their size)

We need to consider these reserved ranges in bitmap creation, so we
extend the mlx4 bitmap utility functions to allow reserved ranges at
both the bottom and the top of the range.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
15 years agopowerpc: Support for relocatable kdump kernel
Mohan Kumar M [Tue, 21 Oct 2008 17:38:10 +0000 (17:38 +0000)]
powerpc: Support for relocatable kdump kernel

This adds relocatable kernel support for kdump. With this one can
use the same regular kernel to capture the kdump. A signature (0xfeed1234)
is passed in r6 from panic code to the next kernel through kexec_sequence
and purgatory code. The signature is used to differentiate between
kdump kernel and non-kdump kernels.

The purgatory code compares the signature and sets the __kdump_flag in
head_64.S.  During the boot up, kernel code checks __kdump_flag and if it
is set, the kernel will behave as relocatable kdump kernel. This kernel
will boot at the address where it was loaded by kexec-tools ie. at the
address reserved through crashkernel boot parameter.

CONFIG_CRASH_DUMP depends on CONFIG_RELOCATABLE option to build kdump
kernel as relocatable. So the same kernel can be used as production and
kdump kernel.

This patch incorporates the changes suggested by Paul Mackerras to avoid
GOT use and to avoid two copies of the code.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Mohan Kumar M <mohan@in.ibm.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Don't use a 16G page if beyond mem= limits
Jon Tollefson [Tue, 21 Oct 2008 15:27:36 +0000 (15:27 +0000)]
powerpc: Don't use a 16G page if beyond mem= limits

If mem= is used on the boot command line to limit memory then the memory block where a 16G page resides may not be available.

Thanks to Michael Ellerman for finding the problem.

Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Add del_node() for early boot code to prune inapplicable devices.
Mike Ditto [Tue, 21 Oct 2008 11:32:29 +0000 (11:32 +0000)]
powerpc: Add del_node() for early boot code to prune inapplicable devices.

Some platforms have variants that can share most of a flat device tree but need
a few devices selectively pruned at boot time.  This adds del_node() to ops.h
to allow access to the existing fdt_del_node().

Signed-off-by: Mike Ditto <mditto@consentry.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Further compile fixup for STRICT_MM_TYPECHECKS
David Gibson [Mon, 20 Oct 2008 17:55:29 +0000 (17:55 +0000)]
powerpc: Further compile fixup for STRICT_MM_TYPECHECKS

A patch of mine was recently committed to fix up STRICT_MM_TYPECHECKS
behaviour on powerpc (f5ea64dcbad89875d130596df14c9b25d994a737).
However, something which breaks it again seems to have slipped in
afterwards.  So, here's another small fix.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Remove empty #else from signal_64.c
Michael Neuling [Wed, 15 Oct 2008 18:21:07 +0000 (18:21 +0000)]
powerpc: Remove empty #else from signal_64.c

Remove empty/bogus #else from signal_64.c

Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Move memory size print into common show_cpuinfo for 32-bit
Becky Bruce [Wed, 15 Oct 2008 08:25:28 +0000 (08:25 +0000)]
powerpc: Move memory size print into common show_cpuinfo for 32-bit

Most of the platforms were printing the size of the memory
in their show_cpuinfo implementations. This moves that to
the common show_cpuinfo, so that all 32-bit platforms will
now print the size of memory.  I also update the code
to deal with the fact that total_memory is now a phys_addr_t.

Signed-off-by: Becky Bruce <becky.bruce@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agohvc_console: Remove __devexit annotation of hvc_remove()
Hendrik Brueckner [Mon, 13 Oct 2008 23:12:52 +0000 (23:12 +0000)]
hvc_console: Remove __devexit annotation of hvc_remove()

Removed __devexit annotation of hvc_remove() to avoid a section mismatch
if the backend initialization fails and hvc_remove() must be used to
clean up allocated hvc structs (called in section __init or __devinit).

Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agohvc_console: Add support for tty window resizing
Hendrik Brueckner [Tue, 14 Oct 2008 05:02:09 +0000 (05:02 +0000)]
hvc_console: Add support for tty window resizing

The patch provides the hvc_resize() function to update the terminal
window dimensions (struct winsize) for a specified hvc console.
The function stores the new window size and schedules a function
that finally updates the tty winsize and signals the change to
user space (SIGWINCH).
Because the winsize update must acquire a mutex and might sleep,
the function is scheduled instead of being called from hvc_poll()
or khvcd.

This patch uses the tty_do_resize() routine from the tty layer.
A pending resize work is canceled in hvc_close() and hvc_hangup().

Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agohvc_console: Fix loop if put_char() returns 0
Hendrik Brueckner [Mon, 13 Oct 2008 23:12:50 +0000 (23:12 +0000)]
hvc_console: Fix loop if put_char() returns 0

If put_char() routine of a hvc console backend returns 0, then the
hvc console starts looping in the following scenarios:

1. hvc_console_print()
If put_char() returns 0 then the while loop may loop forever.
I have added the missing check for 0 to throw away console messages.

2. khvcd may loop:
The thread calls hvc_poll() --> hvc_push()... if there are still
buffered data then the HVC_POLL_WRITE bit is set and causes the
khvcd thread to loop (if yield() returns immediately).

However, instead of looping, the khvcd thread could sleep for
MIN_TIMEOUT (doing the same as for get_chars()).
The MIN_TIMEOUT is set if hvc_push() was not able to write
data to the backend. If data has been written, the timeout is
set to 0 to immediately re-schedule hvc_poll().

Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Tested-by: Christian Borntraeger <borntraeger@de.ibm.com> (virtio_console)
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agohvc_console: Add tty driver flag TTY_DRIVER_RESET_TERMIOS
Hendrik Brueckner [Mon, 13 Oct 2008 23:12:49 +0000 (23:12 +0000)]
hvc_console: Add tty driver flag TTY_DRIVER_RESET_TERMIOS

After a tty hangup() or close() operation, processes might not reset the
termio settings to a sane state. In order to reset the termios to its
default settings the tty driver flag TTY_DRIVER_RESET_TERMIOS has been added.

TTY driver flag description from include/linux/tty_driver.h:
TTY_DRIVER_RESET_TERMIOS --- requests the tty layer to reset the
  termios setting when the last process has closed the device.
  Used for PTY's, in particular.

Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agohvc_console: Add a hangup notifier for backends
Hendrik Brueckner [Mon, 13 Oct 2008 23:12:48 +0000 (23:12 +0000)]
hvc_console: Add a hangup notifier for backends

I have added a hangup notifier that can be used by hvc console
backends to handle a tty hangup. The default irq hangup notifier
calls the notifier_del_irq() for compatibility.

Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agomfd: check for platform_get_irq() return value in sm501
Roel Kluin [Tue, 21 Oct 2008 23:39:55 +0000 (01:39 +0200)]
mfd: check for platform_get_irq() return value in sm501

sm501_devdata->irq is unsigned, while platform_get_irq() returns a
signed int.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
15 years agomfd: use pci_ioremap_bar() in sm501
Arjan van de Ven [Sun, 28 Sep 2008 23:14:52 +0000 (16:14 -0700)]
mfd: use pci_ioremap_bar() in sm501

Use the newly introduced pci_ioremap_bar() function in drivers/mfd.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
15 years agomfd: Don't store volatile bits in WM8350 register cache
Mark Brown [Mon, 20 Oct 2008 21:58:50 +0000 (23:58 +0200)]
mfd: Don't store volatile bits in WM8350 register cache

This makes the contents of the cache clearer and fixes incorrect
initialisation of the cache for partially volatile registers.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
15 years agomfd: don't export wm3850 static functions
Stephen Rothwell [Mon, 20 Oct 2008 21:55:30 +0000 (23:55 +0200)]
mfd: don't export wm3850 static functions

October 10th linux-next build (powerpc allyesconfig) failed like this:

drivers/mfd/wm8350-core.c:1131: error: __ksymtab_wm8350_create_cache causes a section type conflict

Caused by commit 89b4012befb1abca5e86d232bc0e2a797b0d9825 ("mfd: Core
support for the WM8350 AudioPlus PMIC"). wm8350_create_cache is not used
elsewhere, so remove the EXPORT_SYMBOL.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
15 years agomfd: twl4030-gpio driver
David Brownell [Mon, 20 Oct 2008 21:51:46 +0000 (23:51 +0200)]
mfd: twl4030-gpio driver

This adds basic support for the GPIOs in the twl4030 power management
chip.  That includes two open drain LED drivers, and the use of GPIO-0
(and GPIO-1) as MMC/SD card detect switches which can control whether
the VMMC1 (and VMMC2) regulators are active.

This version of the code has a debounce call that will probably be
replaced before long, when a more generic interface exists.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
15 years agomfd: rtc-twl4030 driver
David Brownell [Mon, 20 Oct 2008 21:50:05 +0000 (23:50 +0200)]
mfd: rtc-twl4030 driver

This adds a driver for the RTC inside the TWL4030 multi-function device.
It's a fairly basic RTC, with a wake-capable alarm.

Note that many of the pre-release Overo boards now in circulation can't
effectively use this RTC, because of a wiring error that puts its TWL
chip into "secure" mode.  (As in "secure yourself against tampering".)
This isn't an issue on other OMAP3 boards now supported in mainline,
such as Beagle and Labrador.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
15 years agomfd: twl4030 IRQ handling update
David Brownell [Mon, 20 Oct 2008 21:46:28 +0000 (23:46 +0200)]
mfd: twl4030 IRQ handling update

- Move it into a separate file; clean and streamline it
 - Restructure the init code for reuse during secondary dispatch
 - Support both levels (primary, secondary) of IRQ dispatch
 - Use a workqueue for irq mask/unmask and trigger configuration

Code for two subchips currently share that secondary handler code.
One is the power subchip; its IRQs are now handled by this core,
courtesy of this patch.  The other is the GPIO module, which will
be supported through a later patch.

There are also minor changes to the header file, mostly related
to GPIO support; nothing yet in mainline cares about those.  A
few references to OMAP-specific symbols are disabled; when they
can all be removed, the TWL4030 support ceases being OMAP-specific.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
15 years ago[ARM] clps711x: add sparsemem definitions
Russell King [Tue, 21 Oct 2008 22:29:56 +0000 (23:29 +0100)]
[ARM] clps711x: add sparsemem definitions

Fix the edb7211_defconfig build errors.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
15 years ago[ARM] 5315/1: Fix section mismatch warning (sa1111)
Kristoffer Ericson [Tue, 21 Oct 2008 18:47:22 +0000 (19:47 +0100)]
[ARM] 5315/1: Fix section mismatch warning (sa1111)

This patch fixes the section mismatch warning from
sa1111.o at buildtime.

  CC      arch/arm/common/sa1111.o
  LD      arch/arm/common/built-in.o
  LD      vmlinux.o
  MODPOST vmlinux.o
WARNING: vmlinux.o(.text+0x87f4): Section mismatch in reference from the function sa1111_probe() to the function .devinit.text:sa1110_mb_enable()
The function sa1111_probe() references
the function __devinit sa1110_mb_enable().
This is often because sa1111_probe lacks a __devinit
annotation or the annotation of sa1110_mb_enable is wrong.

Signed-off-by: Kristoffer Ericson <kristoffer.ericson@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
15 years agoxtensa: Add config files for Diamond 232L - Rev B processor variant
Chris Zankel [Tue, 21 Oct 2008 16:11:43 +0000 (09:11 -0700)]
xtensa: Add config files for Diamond 232L - Rev B processor variant

The Diamond 232L processor is a pre-configured Xtensa processor tailored
for Linux application.

Signed-off-by: Chris Zankel <chris@zankel.net>
15 years agoxtensa: Fix io regions
Chris Zankel [Tue, 21 Oct 2008 16:08:31 +0000 (09:08 -0700)]
xtensa: Fix io regions

The uncached area starts at e000.0000 and spans 1GB. Also add an IOADDR
macro to determine the bypass region to access the IO space.

Signed-off-by: Chris Zankel <chris@zankel.net>
15 years agoxtensa: Add support for the Sonic Ethernet device for the XT2000 board.
Chris Zankel [Tue, 6 May 2008 06:36:35 +0000 (23:36 -0700)]
xtensa: Add support for the Sonic Ethernet device for the XT2000 board.

Add support for the on-board Sonic Ethernet device for the XT2000
evaluation board.

Signed-off-by: Chris Zankel <chris@zankel.net>
15 years agoMerge commit 'origin' into master
Benjamin Herrenschmidt [Tue, 21 Oct 2008 04:52:04 +0000 (15:52 +1100)]
Merge commit 'origin' into master

Manual merge of:

arch/powerpc/Kconfig
arch/powerpc/include/asm/page.h

15 years agoMerge commit 'kumar/kumar-for-2.6.28'
Benjamin Herrenschmidt [Tue, 21 Oct 2008 04:49:55 +0000 (15:49 +1100)]
Merge commit 'kumar/kumar-for-2.6.28'

15 years agopowerpc/83xx: Add DS1339 RTC support for MPC8349E-mITX boards .dts
Anton Vorontsov [Fri, 17 Oct 2008 18:56:59 +0000 (22:56 +0400)]
powerpc/83xx: Add DS1339 RTC support for MPC8349E-mITX boards .dts

The RTC is sitting on the I2C2 bus at address 0x68. RTC interrupt signal
is connected to the IPIC's EXT2 interrupt line, the line is shared with
Vitesse 8201 Ethernet PHY.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agopowerpc/83xx: Add support for MCU microcontroller in .dts files
Anton Vorontsov [Fri, 17 Oct 2008 18:57:09 +0000 (22:57 +0400)]
powerpc/83xx: Add support for MCU microcontroller in .dts files

MCU is an external Freescale MC9S08QG8 microcontroller, mainly used to
provide soft power-off function, but also exports two GPIOs (wired to
the LEDs and also available from the external headers).

Added the MCU on mpc8349emitx, mpc837xrdb and mpc8315erdb boards.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agopowerpc/85xx: Move mpc8572ds.dts to address-cells/size-cells = <2>
Kumar Gala [Tue, 21 Oct 2008 04:02:26 +0000 (23:02 -0500)]
powerpc/85xx: Move mpc8572ds.dts to address-cells/size-cells = <2>

Change the top-level #address-cells and #size-cells to <2> so the
mpc8572ds.dts is easier to deal with both a true 32-bit physical
or 36-bit physical address space.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agoof/spi: Support specifying chip select as active high via device tree
Wolfgang Ocker [Wed, 15 Oct 2008 13:00:47 +0000 (15:00 +0200)]
of/spi: Support specifying chip select as active high via device tree

The patch allows to specify that an SPI device needs an active high chip
select.

Signed-off-by: Wolfgang Ocker <weo@reccoware.de>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agopowerpc: Remove device_type = "board_control" properties in .dts files
Anton Vorontsov [Sat, 18 Oct 2008 00:23:52 +0000 (04:23 +0400)]
powerpc: Remove device_type = "board_control" properties in .dts files

We don't want to encourage the bogus device_type usage.

The device type isn't used in the code, so we can simply remove it from
the documentation and dts files.

Boards should specify proper compatible entries instead.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agoi2c-cpm: Suppress autoprobing for devices
Wolfram Sang [Thu, 16 Oct 2008 19:12:00 +0000 (21:12 +0200)]
i2c-cpm: Suppress autoprobing for devices

Similar to commit 618b26d52843c0f85b8eb143cf2695d7f6fd072d, also remove
automatic probing for this i2c controller. Might need updates to dts files
using it.

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Acked-by: Jochen Friedrich <jochen@scram.de>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agopowerpc/85xx: Fix mpc8536ds dma interrupt numbers
Ed Swarthout [Fri, 17 Oct 2008 05:41:32 +0000 (00:41 -0500)]
powerpc/85xx: Fix mpc8536ds dma interrupt numbers

Signed-off-by: Ed Swarthout <Ed.Swarthout@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agopowerpc/85xx: Enable enhanced functions for 8536 TSEC
Jason Jin [Thu, 16 Oct 2008 09:31:32 +0000 (17:31 +0800)]
powerpc/85xx: Enable enhanced functions for 8536 TSEC

Signed-off-by: Jason Jin <Jason.jin@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
15 years agopowerpc: Delete unused prom_strtoul and prom_memparse
Milton Miller [Mon, 20 Oct 2008 18:03:18 +0000 (18:03 +0000)]
powerpc: Delete unused prom_strtoul and prom_memparse

These functions should have been static, and inspection shows they
are no longer used.   (We used to parse mem= but we now defer that
to early_param).

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Revert CHRP boot wrapper to real-base = 12MB on 32-bit
Paul Mackerras [Mon, 20 Oct 2008 17:42:42 +0000 (17:42 +0000)]
powerpc: Revert CHRP boot wrapper to real-base = 12MB on 32-bit

Commit 9b09c6d909dfd8de96b99b9b9c808b94b0a71614 ("powerpc: Change the
default link address for pSeries zImage kernels") changed the
real-base value in the CHRP note added by addnote to the zImage from
12MB to 32MB.  It turns out that this causes unnecessary extra reboots
on old 32-bit CHRP machines.  This therefore adds a -r flag to addnote
to allow us to specify what real-base value it should put in the CHRP
note, and adjusts the wrapper script to pass -r c00000 to addnote when
making a zImage for a CHRP machine.  Also, CHRP machines ignore the
RPA note, so we don't need to arrange for it to be the same as the
kernel's.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Always trim numa memory to lmb_end_of_DRAM()
Milton Miller [Mon, 20 Oct 2008 15:37:04 +0000 (15:37 +0000)]
powerpc: Always trim numa memory to lmb_end_of_DRAM()

numa_enforce_memory_limit tried to be smart and only call lmb_end_of_DRAM
when a memory limit was set via mem= on the command line.  However,
the early boot code will also limit memory added to the lmb system
when iommu=off is specified.  When this happens, the page allocator
is given pages not in the linear mapping and this results in a fatal
data reference to the unmapped page.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Use cpu_thread_in_core in smp_init for of_spin_map
Milton Miller [Mon, 20 Oct 2008 15:37:03 +0000 (15:37 +0000)]
powerpc: Use cpu_thread_in_core in smp_init for of_spin_map

We used to assume that even numbered threads were the primary
threads, ie those that would be listed and started as a cpu from
open firmware.  Replace a left over is even (% 2) check with a check
for it being a primary thread and update the comments.

Tested with a debug print on pseries, identical code found for cell.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Find and destroy possible stale kernel added properties
Milton Miller [Mon, 20 Oct 2008 15:37:03 +0000 (15:37 +0000)]
powerpc: Find and destroy possible stale kernel added properties

64 bit powerpc requires the kexec user space tools avoid overwriting
the static kernel image and translation hash table when choosing
where to put memory image data because it copies the data into place
using the kernels virtual memory system.  Kexec userspace determines
these and other areas blocked by reading properties the kernel adds,
but does not filter these properties when creating the device tree
for the next kernel.

When the second kernel tries to add its values for these properties,
the export via /proc/device-tree is hidden by the pre-existing but
stale values from the flat tree.  Kexec userspace reads the old
property, allocates the new kernel at the old kernel's end, and
gets rejected by the overlap check.

Search and remove these stale properties before adding the new values.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Remove Kconfig support for PPC_MERGE
Kumar Gala [Mon, 20 Oct 2008 08:03:46 +0000 (08:03 +0000)]
powerpc: Remove Kconfig support for PPC_MERGE

There are no users of PPC_MERGE in tree so we can get rid of it.
It was a hold over from the arch/ppc days.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Fix build issue with CONFIG_RELOCATABLE=y
Kumar Gala [Mon, 20 Oct 2008 03:16:55 +0000 (03:16 +0000)]
powerpc: Fix build issue with CONFIG_RELOCATABLE=y

There are two issues when we enable CONFIG_RELOCATABLE.  The first is due
to the fact that phys_addr_t is now defined in linux/types.h.  The second
is due to the fact that the DMA code changes expose memstart_addr to
prom_init.c

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agoof: Format string bug in drivers/of/of_i2c.c
Jon Smirl [Mon, 20 Oct 2008 02:13:15 +0000 (02:13 +0000)]
of: Format string bug in drivers/of/of_i2c.c

Format string bug.  Not exploitable, as this is only writable by root,
but worth fixing all the same.

See 326f6a5c9c9e1a62aec37bdc0c3f8d53adabe77b

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Add missing cuImage.initrd.% target
Grant Likely [Fri, 17 Oct 2008 19:12:07 +0000 (19:12 +0000)]
powerpc: Add missing cuImage.initrd.% target

This target is needed to build cuImages with an embedded ramdisk image.

Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Fix boot wrapper memcmp() called with zero length argument
Mike Ditto [Fri, 17 Oct 2008 09:27:51 +0000 (09:27 +0000)]
powerpc: Fix boot wrapper memcmp() called with zero length argument

I noticed, when trying to use, e.g.,
node = find_node_by_prop_value(prev, "booleanprop", "", 0))
to search for all nodes with a certain boolean property, that memcmp()
returns garbage when comparing zero bytes.  It should return zero.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/numa: Make memory reserve code more robust
Jon Tollefson [Thu, 16 Oct 2008 18:59:43 +0000 (18:59 +0000)]
powerpc/numa: Make memory reserve code more robust

Adjust amount to reserve based on previous nodes for reserves spanning
multiple nodes. Check if the node active range is empty before attempting
to pass the reserve to bootmem.  In practice the range shouldn't be empty,
but to be sure we check.

Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/oprofile: Fix mutex locking for cell spu-oprofile
Carl Love [Tue, 14 Oct 2008 23:37:01 +0000 (23:37 +0000)]
powerpc/oprofile: Fix mutex locking for cell spu-oprofile

The issue is the SPU code is not holding the kernel mutex lock while
adding samples to the kernel buffer.

This patch creates per SPU buffers to hold the data.  Data
is added to the buffers from in interrupt context.  The data
is periodically pushed to the kernel buffer via a new Oprofile
function oprofile_put_buff(). The oprofile_put_buff() function
is called via a work queue enabling the funtion to acquire the
mutex lock.

The existing user controls for adjusting the per CPU buffer
size is used to control the size of the per SPU buffers.
Similarly, overflows of the SPU buffers are reported by
incrementing the per CPU buffer stats.  This eliminates the
need to have architecture specific controls for the per SPU
buffers which is not acceptable to the OProfile user tool
maintainer.

The export of the oprofile add_event_entry() is removed as it
is no longer needed given this patch.

Note, this patch has not addressed the issue of indexing arrays
by the spu number.  This still needs to be fixed as the spu
numbering is not guarenteed to be 0 to max_num_spus-1.

Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Maynard Johnson <maynardj@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Acked-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Unsigned speed cannot be negative in udbg_16559.c
roel kluin [Tue, 14 Oct 2008 14:36:31 +0000 (14:36 +0000)]
powerpc: Unsigned speed cannot be negative in udbg_16559.c

"unsigned int" speed cannot be negative, it's thus pointless
to test if it is.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/pseries: Validate PFN in pseries_remove_lmb()
Nathan Fontenot [Mon, 13 Oct 2008 08:42:00 +0000 (08:42 +0000)]
powerpc/pseries: Validate PFN in pseries_remove_lmb()

The pfn of the memory to be removed should be validated prior to
attempting to remove the memory.  In cases where the probe of a
memory section fails during hotplug add, the pfn for the lmb may
not be valid.

Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/boot: Compare _start against ei.loadsize instead ei.memsize
Sebastian Siewior [Sun, 12 Oct 2008 23:15:26 +0000 (23:15 +0000)]
powerpc/boot: Compare _start against ei.loadsize instead ei.memsize

If the vmlinux binary in memory is larger than 4 MiB than it collides
with the initial boot code which is linked at 4 MiB in case of cuBoot.
If the the uncompressed image size (on disk size) is less than 4 MiB
then it would fit. The difference between those two sizes is the bss
section. In cuBoot we have the dtb embedded right after the data
section so it is very likely that the reset of the bss section (in
kernel's start up code) will overwrite the dtb blob. Therefore we
reallocate the dtb. Something similar is allready done to the initrd.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agobinfmt_elf_fdpic: Update for cputime changes.
Paul Mundt [Tue, 21 Oct 2008 03:07:40 +0000 (12:07 +0900)]
binfmt_elf_fdpic: Update for cputime changes.

Commit f06febc96ba8e0af80bcc3eaec0a109e88275fac ("timers: fix itimer/
many thread hang") introduced a new task_cputime interface and
subsequently only converted binfmt_elf over to it.  This results in the
build for binfmt_elf_fdpic blowing up given that p->signal->{u,s}time
have disappeared from underneath us.

Apply the same trivial fix from binfmt_elf to binfmt_elf_fdpic.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge commit 'gcl/gcl-next'
Benjamin Herrenschmidt [Tue, 21 Oct 2008 02:14:33 +0000 (13:14 +1100)]
Merge commit 'gcl/gcl-next'

15 years agoMerge commit 'jwb/jwb-next'
Benjamin Herrenschmidt [Tue, 21 Oct 2008 02:14:22 +0000 (13:14 +1100)]
Merge commit 'jwb/jwb-next'

15 years agopowerpc/spufs: Explain conditional decrement of aff_sched_count
Andre Detsch [Tue, 21 Oct 2008 00:16:09 +0000 (11:16 +1100)]
powerpc/spufs: Explain conditional decrement of aff_sched_count

This patch adds a comment to clarify why atomic_dec_if_positive is being used
to decrement gang's aff_sched_count on SPU context unbind.

Signed-off-by: Andre Detsch <adetsch@br.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agopowerpc/spufs: Improve search of node for contexts with SPU affinity
Andre Detsch [Tue, 21 Oct 2008 00:15:23 +0000 (11:15 +1100)]
powerpc/spufs: Improve search of node for contexts with SPU affinity

This patch improves redability of the code responsible for trying to find
a node with enough SPUs not committed to other affinity gangs.

An additional check is also added, to avoid taking into account gangs that
have no SPU affinity.

Signed-off-by: Andre Detsch <adetsch@br.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agopowerpc/spufs: Use kmalloc rather than kzalloc for switch log buffer
Jeremy Kerr [Fri, 17 Oct 2008 01:02:31 +0000 (12:02 +1100)]
powerpc/spufs: Use kmalloc rather than kzalloc for switch log buffer

No need to zero the entire buffer, just the head and tail indices.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agopowerpc/spufs: Don't spu_acquire_saved unnecessarily in regs read
Jeremy Kerr [Thu, 16 Oct 2008 00:11:12 +0000 (11:11 +1100)]
powerpc/spufs: Don't spu_acquire_saved unnecessarily in regs read

With most file readers (eg cat, dd), reading a context's regs file will
result in two reads: the first to read the data, and the second to
return EOF. Because each read performs a spu_acquire_saved, we end up
descheduling and re-scheduling the context twice.

This change does a simple check to see if we'd return EOF before
calling spu_acquire_saved(), saving the extra schedule operation.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agopowerpc/spufs: Don't require full buffer in switch_log read
Jeremy Kerr [Wed, 15 Oct 2008 23:51:46 +0000 (10:51 +1100)]
powerpc/spufs: Don't require full buffer in switch_log read

Currently, read() on the sputrace log will block until the read buffer
is full. This makes it difficult to retrieve the end of the buffer, as
the user will need to read with the right-sized buffer.

In a similar method as 91553a1b5e0df006a3573a88d98ee7cd48a3818a, this
change makes the switch_log return if there has already been data
read.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agopowerpc/spufs: Use state_mutex for switch_log locking, and prevent multiple openers
Jeremy Kerr [Wed, 15 Oct 2008 23:03:46 +0000 (10:03 +1100)]
powerpc/spufs: Use state_mutex for switch_log locking, and prevent multiple openers

Currently, we use ctx->mapping_lock and ctx->switch_log->lock for the
context switch log. The mapping lock only prevents concurrent open()s,
so we require the switch_lock->lock for reads.

Since writes to the switch log buffer occur on context switches, we're
better off synchronising with the state_mutex, which is held during a
switch. Since we're serialised througout the buffer reads and writes,
we can use the state mutex to protect open and release too, and
can now kfree() the log buffer on release. This allows us to perform
the switch log notify without taking any extra locks.

Because the buffer is only present while the file is open, we can use
it to prevent multiple simultaneous openers.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agopowerpc/spufs: sputrace: Don't block until the read buffer is full
Jeremy Kerr [Wed, 15 Oct 2008 02:20:42 +0000 (13:20 +1100)]
powerpc/spufs: sputrace: Don't block until the read buffer is full

Currently, read() on the sputrace buffer will only return data when
the user buffer is exhausted. This may mean that we never see the
end of the event log, unless we read() with exactly the right-sized
buffer.

This change makes sputrace_read not block if we have data ready to
return.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agopowerpc/spufs: sputrace: Only enable logging on open(), prevent multiple openers
Jeremy Kerr [Wed, 15 Oct 2008 02:16:18 +0000 (13:16 +1100)]
powerpc/spufs: sputrace: Only enable logging on open(), prevent multiple openers

Currently, sputrace will start logging to the event buffer before the
log buffer has been open()ed. This results in a heap of "lost samples"
warnings if the sputrace file hasn't yet been opened.

Since the buffer is reset on open() anyway, there's no need to enable
logging when no-one has opened the log.

Because open clears the log, make it return EBUSY for mutliple open
calls.

Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
15 years agoibm_newemac: Fix new MAL feature handling
Josh Boyer [Thu, 16 Oct 2008 04:38:40 +0000 (04:38 +0000)]
ibm_newemac: Fix new MAL feature handling

Support for new features needed by the PPC 405EZ boards
introduced some errors in the MAL and EMAC feature handling.
This broke 'allmodconfig' builds as CONFIG_PPC_DCR_NATIVE is
not set for those.

This patch fixes these errors by wrapping the code in the
appropriate #ifdefs.

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
15 years agomm: remove duplicated #include's
Huang Weiyi [Mon, 20 Oct 2008 22:43:33 +0000 (06:43 +0800)]
mm: remove duplicated #include's

Removed duplicated #include <linux/vmalloc.h> in mm/vmalloc.c and
"internal.h" in mm/memory.c.

Signed-off-by: Huang Weiyi <weiyi.huang@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoExport tiny shmem_file_setup for DRM-GEM
Hugh Dickins [Mon, 20 Oct 2008 23:04:04 +0000 (00:04 +0100)]
Export tiny shmem_file_setup for DRM-GEM

We're trying to keep the !CONFIG_SHMEM tiny-shmem.c (using ramfs without
swap) in synch with CONFIG_SHMEM shmem.c (and mpm is preparing patches
to combine them).  I was glad to see EXPORT_SYMBOL_GPL(shmem_file_setup)
go into shmem.c, but why not support DRM-GEM when !CONFIG_SHMEM too?
But caution says still depend on MMU, since !CONFIG_MMU is.. different.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Matt Mackall <mpm@selenic.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomisc: replace remaining __FUNCTION__ with __func__
Harvey Harrison [Mon, 20 Oct 2008 23:00:08 +0000 (16:00 -0700)]
misc: replace remaining __FUNCTION__ with __func__

__FUNCTION__ is gcc-specific, use __func__

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofix CONFIG_HIGHMEM compile error in drivers/gpu/drm/i915/i915_gem.c
Thomas Gleixner [Mon, 20 Oct 2008 22:47:45 +0000 (00:47 +0200)]
fix CONFIG_HIGHMEM compile error in drivers/gpu/drm/i915/i915_gem.c

commit 9b7530cc329eb036cfa589930c270e85031f554c ("i915: cleanup coding
horrors in i915_gem_gtt_pwrite()")

broke the i386 build for CONFIG_HIGHMEM=y.

Caught by automatic testing http://www.tglx.de/autoqa-logs/000137-0006-0001.log

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
[ My bad. It's the same patch I sent out earlier, nobody noticed then either.. ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofix WARN() for PPC
Arjan van de Ven [Mon, 20 Oct 2008 21:41:03 +0000 (14:41 -0700)]
fix WARN() for PPC

powerpc doesn't use the generic WARN_ON infrastructure.  The newly
introduced WARN() as a result didn't print the message, this patch adds
the printk for this specific case.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agokexec: fix crash_save_vmcoreinfo_init build problem
Luck, Tony [Mon, 20 Oct 2008 22:23:40 +0000 (15:23 -0700)]
kexec: fix crash_save_vmcoreinfo_init build problem

This fixes

  kernel/kexec.c: In function 'crash_save_vmcoreinfo_init':
  kernel/kexec.c:1374: error: 'vmlist' undeclared (first use in this function)
  kernel/kexec.c:1374: error: (Each undeclared identifier is reported only once
  kernel/kexec.c:1374: error: for each function it appears in.)
  kernel/kexec.c:1410: error: invalid use of undefined type 'struct vm_struct'
  make[1]: *** [kernel/kexec.o] Error 1

Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-rmk' of git://git.marvell.com/orion into devel
Russell King [Mon, 20 Oct 2008 22:06:15 +0000 (23:06 +0100)]
Merge branch 'for-rmk' of git://git.marvell.com/orion into devel