openwrt/staging/blogic.git
15 years agoBlackfin arch: fix bug - traps test case 19 for exception 0x2d fails
Bernd Schmidt [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: fix bug - traps test case 19 for exception 0x2d fails

Enable null pointer checking for ICPLBs. The code was there but for
some reason I had commented it out at some stage during development.

Should restrict this to 1K since atomic ops start there.

Signed-off-by: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: add platform device bfin_mii-bus and KSZ8893M switch driver platform...
Graf Yang [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: add platform device bfin_mii-bus and KSZ8893M switch driver platform resources to board files

Signed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: build jtag tty driver as a module by default
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: build jtag tty driver as a module by default

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: fix 2 bugs related to debug
Jie Zhang [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: fix 2 bugs related to debug

 - unable to single step over emuexcpt instruction
 - gdbproxy goes into infinite loop when doing gdb does "next" over
   "emuexcpt"

Don't decrement PC after software breakpoint.

Signed-off-by: Jie Zhang <jie.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Add ANOMALY_05000380 to BF54x to kill the compile warning
Bryan Wu [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Add ANOMALY_05000380 to BF54x to kill the compile warning

Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Fix bug - 561 SMP kernel can't boot from jffs2
Graf Yang [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Fix bug - 561 SMP kernel can't boot from jffs2

bss_l2 section is garbage when the data in this section is used by
_bfin_relocate_l1_mem, so move the zero out function ahead.

Signed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: base SIC_IWR# programming on whether the MMR exists
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: base SIC_IWR# programming on whether the MMR exists

base SIC_IWR# programming on whether the MMR exists
rather than having to maintain another list of processors

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: read SYSCR on newer parts that mirror the bits of SWRST in it
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: read SYSCR on newer parts that mirror the bits of SWRST in it

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: fixup board init function name
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: fixup board init function name

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: drop CONFIG_I2C_BOARDINFO ifdefs
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: drop CONFIG_I2C_BOARDINFO ifdefs

Drop CONFIG_I2C_BOARDINFO ifdefs as the common i2c header handles this
already by stubbing things out

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: bfin_reset->_bfin_reset redirection no longer needed
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: bfin_reset->_bfin_reset redirection no longer needed

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: sync reboot handler with version in u-boot
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: sync reboot handler with version in u-boot

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Faster Implementation of csum_tcpudp_nofold()
Michael Hennerich [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Faster Implementation of csum_tcpudp_nofold()

Avoid conditional branch instructions during carry bit additions.
Special thanks to Bernd.
Simplify: Use ((len + proto) << 8) like every other __LITTLE_ENDIAN__ machine

Cc: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Fix bug - BF518 port F, G, and H have different mux offset compare...
Graf Yang [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Fix bug - BF518 port F, G, and H have different mux offset compare to BF527

[Mike Frysinger <vapier.adi@gmail.com>: keep the ifdef nest down]

Signed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Add in cflag to support mlong-calls for kgdb_test
Grace Pan [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Add in cflag to support mlong-calls for kgdb_test

Signed-off-by: Grace Pan <grace.pan@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Fix bug - Run "reboot" hangs bf518-ezbrd
Sonic Zhang [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Fix bug - Run "reboot" hangs bf518-ezbrd

[Mike Frysinger <vapier.adi@gmail.com>:
 - setup P_DEFAULT_BOOT_SPI_CS  for every arch based on
   the default bootrom behavior and convert all our boards
   to it
 - revert previous anomaly change ... bf51x is not affected
   by anomaly 05000353]

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoMAINTIANERS: Blackfin: remove subscribers-only marking
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
MAINTIANERS: Blackfin: remove subscribers-only marking

remove subscribers-only marking as the list is
automatically & silently moderated for people

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Add ability to count and display number of NMI interrupts
Robin Getz [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Add ability to count and display number of NMI interrupts

Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Add a few more instructions that can cause the trace buffer to be...
Robin Getz [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Add a few more instructions that can cause the trace buffer to be discontiguous

Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Fix URL
Robin Getz [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Fix URL

Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: cleanup bf54x ifdef mess in gpio code
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: cleanup bf54x ifdef mess in gpio code

merge more of the bf54x and !bf54x gpio code together to
cut down on #ifdef mess

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Add one more check on `fp' to prevent double fault
Jie Zhang [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Add one more check on `fp' to prevent double fault

Signed-off-by: Jie Zhang <jie.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: explicit add a might sleep to gpio_free
Uwe Kleine-Koenig [Wed, 4 Feb 2009 09:02:30 +0000 (17:02 +0800)]
Blackfin arch: explicit add a might sleep to gpio_free

According to the documentation gpio_free should only be called from task
context only.  To make this more explicit add a might sleep to all
implementations.

This patch changes the gpio_free implementations for the blackfin
architecture.

Signed-off-by: Uwe Kleine-Koenig <ukleinek@strlen.de>
Cc: David Brownell <david-b@pacbell.net>
Acked-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
15 years agoBlackfin arch: don't accidently re-enable interrupts
Robin Getz [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: don't accidently re-enable interrupts

Make sure we don't accidently re-enable interrupts if we are being
called in atomic context

Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Print out where the bootmode is coming from (for easier debugging).
Robin Getz [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Print out where the bootmode is coming from (for easier debugging).

Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: reset POLAR setting when acquiring a gpio for the first time
Michael Hennerich [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: reset POLAR setting when acquiring a gpio for the first time

when requesting a GPIO for the first time, the POLAR setting is not
set to a sane state.  this can lead to indeterminate behavior that
cannot be resolved without an explicit write to the Blackfin port POLAR
register.

when requesting a GPIO for the first time via gpio_request(), the POLAR
setting for the GPIO in question should be set to sane state.  this
should occur if the GPIO has not been allocated in any other way.

some examples:
 - when doing something like "request_irq(); gpio_request();" on the
   same GPIO, the POLAR setting should not be reset.
 - when doing "gpio_request(); gpio_request();" on the same GPIO, the
   POLAR setting should be reset only the first time and not the second.

Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Fix Bug - request lines with peripheral_request_list, but don't get...
Michael Hennerich [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Fix Bug - request lines with peripheral_request_list, but don't get freed with peripheral_free_list

Remove erroneous check_gpio(ident) in peripheral_free()

Reported-by: Michael McTernan <mmcternan@airvana.com>
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: Fix Bug - Kernel does not boot if re-program clocks
Michael Hennerich [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: Fix Bug - Kernel does not boot if re-program clocks

On BF561 EBIU_SDGCTL bit 31 controls the SDRAM external data
path width, typically set 0 for a 32-bit bus width. On other
Blackfin derivatives this bit should be set by default.

Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: fix bug - Dmacopy failed in BF537-STAMP
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: fix bug - Dmacopy failed in BF537-STAMP

Dmacopy failed in BF537-STAMP when copy from SRAM to SDRAM and kernel
will reboot automatically.

Fixing by doing a SSYNC before mucking with DMA registers

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: enable bfin_eth in bf51x by default
Sonic Zhang [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: enable bfin_eth in bf51x by default

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: line up machine-/cpu- vars after BF54xM addition
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: line up machine-/cpu- vars after BF54xM addition

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: add support for mobile ddr bf54x parts
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: add support for mobile ddr bf54x parts

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agoBlackfin arch: scrub comments/whitespace/cvs keywords
Mike Frysinger [Wed, 4 Feb 2009 08:49:45 +0000 (16:49 +0800)]
Blackfin arch: scrub comments/whitespace/cvs keywords

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
15 years agowrite-back: fix nr_to_write counter
Artem Bityutskiy [Mon, 2 Feb 2009 16:33:49 +0000 (18:33 +0200)]
write-back: fix nr_to_write counter

Commit 05fe478dd04e02fa230c305ab9b5616669821dd3 introduced some
@wbc->nr_to_write breakage.

It made the following changes:
 1. Decrement wbc->nr_to_write instead of nr_to_write
 2. Decrement wbc->nr_to_write _only_ if wbc->sync_mode == WB_SYNC_NONE
 3. If synced nr_to_write pages, stop only if if wbc->sync_mode ==
    WB_SYNC_NONE, otherwise keep going.

However, according to the commit message, the intention was to only make
change 3.  Change 1 is a bug.  Change 2 does not seem to be necessary,
and it breaks UBIFS expectations, so if needed, it should be done
separately later.  And change 2 does not seem to be documented in the
commit message.

This patch does the following:
 1. Undo changes 1 and 2
 2. Add a comment explaining change 3 (it very useful to have comments
    in _code_, not only in the commit).

Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoFix my email address in qd65xx.[ch]/pata_qdi.c
Samuel Thibault [Tue, 3 Feb 2009 12:12:58 +0000 (13:12 +0100)]
Fix my email address in qd65xx.[ch]/pata_qdi.c

The @fnac.net will be shut down within a couple of months, so fix my
email address.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'linux-next' of git://git.infradead.org/ubifs-2.6
Linus Torvalds [Wed, 4 Feb 2009 00:52:44 +0000 (16:52 -0800)]
Merge branch 'linux-next' of git://git.infradead.org/ubifs-2.6

* 'linux-next' of git://git.infradead.org/ubifs-2.6:
  UBIFS: remove fast unmounting
  UBIFS: return sensible error codes
  UBIFS: remount ro fixes
  UBIFS: spelling fix 'date' -> 'data'
  UBIFS: sync wbufs after syncing inodes and pages
  UBIFS: fix LPT out-of-space bug (again)
  UBIFS: fix no_chk_data_crc
  UBIFS: fix assertions
  UBIFS: ensure orphan area head is initialized
  UBIFS: always clean up GC LEB space
  UBIFS: add re-mount debugging checks
  UBIFS: fix LEB list freeing
  UBIFS: simplify locking
  UBIFS: document dark_wm and dead_wm better
  UBIFS: do not treat all data as short term
  UBIFS: constify operations
  UBIFS: do not commit twice

15 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Wed, 4 Feb 2009 00:52:10 +0000 (16:52 -0800)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  NVRAM depends on RTC_DRV_CMOS
  rename platform_driver name "flash" to "sa1100-mtd"
  annotate that [fp, #-4] is the saved lr
  Use __SPIN_LOCK_UNLOCKED to initialize bad_irq_desc.lock
  ARM: OMAP: fix fault in enter_full_retention()
  ARM: OMAP: Mask interrupts when disabling interrupts, v2
  ARM: OMAP: gptimer min_delta_ns corrected
  ARM: OMAP: Fix hsmmc init, v2
  ARM: OMAP: Fix omap34xx revision detection for ES3.1
  ARM: OMAP: DMA: Fix uninitialized channel flags
  ARM: OMAP: Fix race in OMAP2/3 DMA IRQ handling
  ARM: OMAP: Fix McBSP spin_lock deadlock
  [ARM] 5366/1: fix shared memory coherency with VIVT L1 + L2 caches
  [ARM] call undefined instruction exception handler with irqs enabled
  [ARM] msm: fix build errors
  [ARM] etherh: continue fixing build failure

15 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfashe...
Linus Torvalds [Wed, 4 Feb 2009 00:50:20 +0000 (16:50 -0800)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/mfasheh/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2:
  ocfs2: add quota call to ocfs2_remove_btree_range()
  ocfs2: Wakeup the downconvert thread after a successful cancel convert
  ocfs2: Access the xattr bucket only before modifying it.
  configfs: Silence lockdep on mkdir(), rmdir() and configfs_depend_item()
  ocfs2: Fix possible deadlock in ocfs2_write_dquot()
  ocfs2: Push out dropping of dentry lock to ocfs2_wq

15 years agoMerge branch 'for-linus' of git://oss.sgi.com/xfs/xfs
Linus Torvalds [Wed, 4 Feb 2009 00:49:54 +0000 (16:49 -0800)]
Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  [XFS] Warn on transaction in flight on read-only remount
  xfs: Check buffer lengths in log recovery
  don't reallocate sxp variable passed into xfs_swapext

15 years ago[XFS] Warn on transaction in flight on read-only remount
Felix Blyakher [Fri, 23 Jan 2009 03:34:05 +0000 (21:34 -0600)]
[XFS] Warn on transaction in flight on read-only remount

Till VFS can correctly support read-only remount without racing,
use WARN_ON instead of BUG_ON on detecting transaction in flight
after quiescing filesystem.

Signed-off-by: Felix Blyakher <felixb@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
15 years agoxfs: Check buffer lengths in log recovery
Dave Chinner [Thu, 22 Jan 2009 04:37:47 +0000 (15:37 +1100)]
xfs: Check buffer lengths in log recovery

Before trying to obtain, read or write a buffer,
check that the buffer length is actually valid. If
it is not valid, then something read in the recovery
process has been corrupted and we should abort
recovery.

Reported-by: Eric Sesterhenn <snakebyte@gmx.de>
Tested-by: Eric Sesterhenn <snakebyte@gmx.de>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Felix Blyakher <felixb@sgi.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Felix Blyakher <felixb@sgi.com>
15 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux...
Felix Blyakher [Tue, 3 Feb 2009 16:38:41 +0000 (10:38 -0600)]
Merge branch 'master' of git://git./linux/kernel/git/torvalds/linux-2.6 into for-linus

15 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Tue, 3 Feb 2009 15:39:55 +0000 (07:39 -0800)]
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:
  libata: implement HORKAGE_1_5_GBPS and apply it to WD My Book
  libata: add no penalty retry request for EH device handling routines
  libata: improve probe failure handling
  libata: add @spd_limit to sata_down_spd_limit()
  libata: clear dev->ering in smarter way
  libata: check onlineness before using SPD in sata_down_spd_limit()
  libata: move ata_dev_disable() to libata-eh.c
  libata: fix EH device failure handling
  sata_nv: ck804 has borked hardreset too
  ide/libata: fix ata_id_is_cfa() (take 4)
  libata: fix kernel-doc warnings
  ahci: add a module parameter to ignore the SSS flags for async scanning
  sata_mv: Fix chip type for Hightpoint RocketRaid 1740/1742
  [libata] sata_sil: Fix compilation error with libata debugging enabled

15 years agolibata: implement HORKAGE_1_5_GBPS and apply it to WD My Book
Tejun Heo [Thu, 29 Jan 2009 11:31:36 +0000 (20:31 +0900)]
libata: implement HORKAGE_1_5_GBPS and apply it to WD My Book

3Gbps is often much more prone to transmission failures.  It's usually
okay to let EH handle speed down after transmission failures but some
WD My Book drives completely shutdown after certain transmission
failures and after it only power cycling can revive them.  Combined
with the fact that external drives often end up with cable assembly
which is longer than usual and more likely to have intervening gender,
this makes these drives very likely to shutdown under certain
configurations virtually rendering them unusable.

This patch implements HOARKGE_1_5_GBPS and applies it to WD My Book
such that 1.5Gbps is forced once the device is identified.

Please take a look at the following bz for related reports.

  http://bugzilla.kernel.org/show_bug.cgi?id=9913

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: add no penalty retry request for EH device handling routines
Tejun Heo [Thu, 29 Jan 2009 11:31:35 +0000 (20:31 +0900)]
libata: add no penalty retry request for EH device handling routines

Let -EAGAIN from EH device handling routines trigger EH retry without
consuming its tries count.  This will be used to implement link SPD
horkage which requires hardreset to adjust SPD without affecting other
EH decisions.  As it bypasses the forward progress guarantee provided
by the tries count, the requester is responsible for ensuring forward
progress.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: improve probe failure handling
Tejun Heo [Thu, 29 Jan 2009 11:31:34 +0000 (20:31 +0900)]
libata: improve probe failure handling

When link is flaky at high speed, it isn't uncommon for a device to
repeatedly fail probing sequence early after successfully negotiating
high link speed.  This often leads to consecutive hotplug events
without successful probing.

This patch improves libata EH such that it remembers probing trials
and if there have been more than two unsuccessful trials in the past
60 seconds, slows down link speed to 1.5Gbps.

As link speed negotiation is the duty of the PHY layer proper, the
goal of this fallback mechanism is to provide the last resort when
everything else fails, which unfortunately happens not too
infrequently, so no fancy 6->3->1.5 speeding down or highest
successful transmission speed seen kind of logics (yet).

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: add @spd_limit to sata_down_spd_limit()
Tejun Heo [Thu, 29 Jan 2009 11:31:33 +0000 (20:31 +0900)]
libata: add @spd_limit to sata_down_spd_limit()

Add @spd_limit to sata_down_spd_limit() so that the caller can specify
the SPD limit it wants.  This parameter doesn't get in the way even
when it's too low.  The closest possible limit is applied.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: clear dev->ering in smarter way
Tejun Heo [Thu, 29 Jan 2009 11:31:32 +0000 (20:31 +0900)]
libata: clear dev->ering in smarter way

dev->ering used to be cleared together with the rest of ata_device in
ata_dev_init() which is called whenever a probing event occurs.
dev->ering is about to be used to track probing failures so it needs
to remain persistent over multiple porbing events.  This patch
achieves this by doing the following.

* Instead of CLEAR_OFFSET, define CLEAR_BEGIN and CLEAR_END and only
  clear between BEGIN and END.  ering is moved after END.  The split
  of persistent area is to allow hotter items remain at the head.

* ering is explicitly cleared on ata_dev_disable() and when device
  attach succeeds.  So, ering is persistent throug a device's life
  time (unless explicitly cleared of course) and also through periods
  inbetween disablement of an attached device and successful detection
  of the next one.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: check onlineness before using SPD in sata_down_spd_limit()
Tejun Heo [Thu, 29 Jan 2009 11:31:31 +0000 (20:31 +0900)]
libata: check onlineness before using SPD in sata_down_spd_limit()

sata_down_spd_limit() should check whether the link is online before
using the SPD value to determine how to limit the link speed.  Factor
out onlineness test and test it from sata_down_spd_limit().

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: move ata_dev_disable() to libata-eh.c
Tejun Heo [Thu, 29 Jan 2009 11:31:30 +0000 (20:31 +0900)]
libata: move ata_dev_disable() to libata-eh.c

ata_dev_disable() is about to be more tightly integrated into EH
logic.  Move it to libata-eh.c.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: fix EH device failure handling
Tejun Heo [Thu, 29 Jan 2009 11:31:29 +0000 (20:31 +0900)]
libata: fix EH device failure handling

The dev->pio_mode > XFER_PIO_0 test is there to avoid unnecessary
speed down warning messages but it accidentally disabled SATA link spd
down during configuration phase after reset where PIO mode is always
zero.

This patch fixes the problem by moving the test where it belongs.
This makes libata probing sequence behave better when the connection
is flaky at higher link speeds which isn't too uncommon for eSATA
devices.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agosata_nv: ck804 has borked hardreset too
Tejun Heo [Sun, 1 Feb 2009 01:56:31 +0000 (10:56 +0900)]
sata_nv: ck804 has borked hardreset too

While playing with nvraid, I found out that rmmoding and insmoding
often trigger hardreset failure on the first port (the second one was
always okay).  Seriously, how diverse can you get with hardreset
behaviors?  Anyways, make ck804 use noclassify variant too.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agoide/libata: fix ata_id_is_cfa() (take 4)
Sergei Shtylyov [Sun, 1 Feb 2009 16:46:39 +0000 (20:46 +0400)]
ide/libata: fix ata_id_is_cfa() (take 4)

When checking for the CFA feature set support, ata_id_is_cfa() tests bit 2 in
word 82 of the identify data instead the word 83;  it also checks the ATA/PI
version support in the word 80 (which the CompactFlash specifications have as
reserved), this having no slightest chance to work on the modern CF cards that
don't have 0x848A in the word 0...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: fix kernel-doc warnings
Randy Dunlap [Wed, 21 Jan 2009 00:28:59 +0000 (16:28 -0800)]
libata: fix kernel-doc warnings

Fix libata kernel-doc warnings:

Warning(linux-next-20090120//drivers/ata/libata-core.c:4720): Excess function parameter 'dev' description in 'ata_qc_new'
Warning(linux-next-20090120//drivers/ata/libata-scsi.c:428): No description found for parameter 'ap'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agoahci: add a module parameter to ignore the SSS flags for async scanning
Arjan van de Ven [Mon, 26 Jan 2009 10:05:44 +0000 (02:05 -0800)]
ahci: add a module parameter to ignore the SSS flags for async scanning

The SSS flag, which directs the OS to spin up one disk at a time
to not have the PSU blow out, sometimes gets set even when not needed.
The effect of this is a longer-than-needed boot time.

This patch adds a module parameter that makes the driver ignore SSS
at least as far as the parallel scan during boot is concerned...

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agosata_mv: Fix chip type for Hightpoint RocketRaid 1740/1742
Mark Lord [Tue, 27 Jan 2009 21:33:13 +0000 (16:33 -0500)]
sata_mv: Fix chip type for Hightpoint RocketRaid 1740/1742

Fix chip type for the Highpoint RocketRAID 1740 and 1742 PCI cards.
These really do have Marvell 6042 chips on them, rather than the 5081 chip.

Confirmed by multiple (two) users (for the 1740), and by examining
the product photographs from Highpoint's web site.

Signed-off-by: Mark Lord <mlord@pobox.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years ago[libata] sata_sil: Fix compilation error with libata debugging enabled
Pasi Kärkkäinen [Mon, 2 Feb 2009 19:47:14 +0000 (21:47 +0200)]
[libata] sata_sil: Fix compilation error with libata debugging enabled

I tried compiling 2.6.29-rc1 and 2.6.29-rc3 with libata debugging enabled
and got the following error:

  CC [M]  drivers/ata/sata_sil.o
drivers/ata/sata_sil.c: In function 'sil_fill_sg':
drivers/ata/sata_sil.c:327: error: 'pi' undeclared (first use in this function)
drivers/ata/sata_sil.c:327: error: (Each undeclared identifier is reported only once
drivers/ata/sata_sil.c:327: error: for each function it appears in.)
make[2]: *** [drivers/ata/sata_sil.o] Error 1
make[1]: *** [drivers/ata] Error 2
make: *** [drivers] Error 2

include/linux/libata.h has the following enabled:

#define ATA_DEBUG
#define ATA_VERBOSE_DEBUG
#define ATA_IRQ_TRAP

This fixes the compilation.

Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
Linus Torvalds [Tue, 3 Feb 2009 03:28:58 +0000 (19:28 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
  PCI hotplug: Change link order of pciehp & acpiphp
  PCI hotplug: fakephp: Allocate PCI resources before adding the device
  PCI MSI: Fix undefined shift by 32
  PCI PM: Do not wait for buses in B2 or B3 during resume
  PCI PM: Power up devices before restoring their state
  PCI PM: Fix hibernation breakage on EeePC 701
  PCI: irq and pci_ids patch for Intel Tigerpoint DeviceIDs
  PCI PM: Fix suspend error paths and testing facility breakage

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg...
Linus Torvalds [Tue, 3 Feb 2009 03:27:00 +0000 (19:27 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/penberg/slab-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/slab-2.6:
  slub: fix per cpu kmem_cache_cpu array memory leak
  kmalloc: return NULL instead of link failure

15 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Tue, 3 Feb 2009 03:26:44 +0000 (19:26 -0800)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  fbdev/atyfb: Fix DSP config on some PowerMacs & PowerBooks
  powerpc: Fix oops on some machines due to incorrect pr_debug()
  powerpc/ps3: Printing fixups for l64 to ll64 convserion drivers/net
  powerpc/5200: update device tree binding documentation
  powerpc/5200: Bugfix for PCI mapping of memory and IMMR
  powerpc/5200: update defconfigs

15 years agoMerge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 3 Feb 2009 03:26:29 +0000 (19:26 -0800)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched_rt: don't use first_cpu on cpumask created with cpumask_and
  sched: fix buddie group latency
  sched: clear buddies more aggressively
  sched: symmetric sync vs avg_overlap
  sched: fix sync wakeups
  cpuset: fix possible deadlock in async_rebuild_sched_domains

15 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Tue, 3 Feb 2009 03:26:06 +0000 (19:26 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6: (45 commits)
  V4L/DVB (10411): s5h1409: Perform s5h1409 soft reset after tuning
  V4L/DVB (10403): saa7134-alsa: saa7130 doesn't support digital audio
  V4L/DVB (10229): ivtv: fix memory leak
  V4L/DVB (10385): gspca - main: Fix memory leak when USB disconnection while streaming.
  V4L/DVB (10325): em28xx: Fix for fail to submit URB with IRQs and Pre-emption Disabled
  V4L/DVB (10317): radio-mr800: fix radio->muted and radio->stereo
  V4L/DVB (10314): cx25840: ignore TUNER_SET_CONFIG in the command callback.
  V4L/DVB (10288): af9015: bug fix: stick does not work always when plugged
  V4L/DVB (10287): af9015: fix second FE
  V4L/DVB (10270): saa7146: fix unbalanced mutex_lock/unlock
  V4L/DVB (10265): budget.c driver: Kernel oops: "BUG: unable to handle kernel paging request at ffffffff
  V4L/DVB (10261): em28xx: fix kernel panic on audio shutdown
  V4L/DVB (10257): em28xx: Fix for KWorld 330U Board
  V4L/DVB (10256): em28xx: Fix for KWorld 330U AC97
  V4L/DVB (10254): em28xx: Fix audio URB transfer buffer race condition
  V4L/DVB (10250): cx25840: fix regression: fw not loaded on first use
  V4L/DVB (10248): v4l-dvb: fix a bunch of compile warnings.
  V4L/DVB (10243): em28xx: fix compile warning
  V4L/DVB (10240): Fix obvious swapped names in v4l2_subdev logic
  V4L/DVB (10233): [PATCH] Terratec Cinergy DT XS Diversity new USB ID (0ccd:0081)
  ...

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
Linus Torvalds [Tue, 3 Feb 2009 03:24:14 +0000 (19:24 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/drzeus/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
  pxamci: enable DMA for write ops after CMD/RESP
  pxamci: replace #ifdef CONFIG_PXA27x with if (cpu_is_pxa27x())
  ricoh_mmc: Use suspend_late/resume_early
  mmci: Add support for ST Micro derivate
  mmc: Add a MX2/MX3 specific SDHC driver

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
Linus Torvalds [Tue, 3 Feb 2009 03:23:49 +0000 (19:23 -0800)]
Merge git://git./linux/kernel/git/bart/ide-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6:
  icside: fix PCB version 6 support (v2)
  tx4939ide: typo fix and minor cleanup
  ide: add CS5536 host driver (v3)
  ide: Force VIA IDE legacy interrupts for AmigaOne boards
  IDE: Unregister and disable devices if initialization fails.
  ide: fix ide_register_port() failure handling
  ide: struct device - replace bus_id with dev_name(), dev_set_name()
  ide-cd: fix DMA for non bio-backed requests

15 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/dvrabel/uwb
Linus Torvalds [Tue, 3 Feb 2009 03:20:17 +0000 (19:20 -0800)]
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:
  uwb: lock rc->rsvs_lock with spin_lock_bh()
  wusb: timeout when waiting for ASL/PZL updates in whci-hcd
  uwb: remove unused #include <version.h>'s
  wusb: return -ENOTCONN when resetting a port with no connected device
  uwb: safely remove all reservations

15 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Tue, 3 Feb 2009 03:19:50 +0000 (19:19 -0800)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: add text file detailing queue/ sysfs files
  bio.h: If they MUST be inlined, then use __always_inline
  Fix misleading comment in bio.h
  block: fix inconsistent parenthesisation of QUEUE_FLAG_DEFAULT
  block: fix oops in blk_queue_io_stat()

15 years agovirtio-pci: do not oops on config change if driver not loaded
Mark McLoughlin [Tue, 3 Feb 2009 03:03:53 +0000 (13:33 +1030)]
virtio-pci: do not oops on config change if driver not loaded

The host really shouldn't be notifying us of config changes
before the device status is VIRTIO_CONFIG_S_DRIVER or
VIRTIO_CONFIG_S_DRIVER_OK.

However, if we do happen to be interrupted while we're not
attached to a driver, we really shouldn't oops. Prevent
this simply by checking that device->driver is non-NULL
before trying to notify the driver of config changes.

Problem observed by doing a "set_link virtio.0 down" with
QEMU before the net driver had been loaded.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomodules: Use a better scheme for refcounting
Eric Dumazet [Tue, 3 Feb 2009 03:01:36 +0000 (13:31 +1030)]
modules: Use a better scheme for refcounting

Current refcounting for modules (done if CONFIG_MODULE_UNLOAD=y) is
using a lot of memory.

Each 'struct module' contains an [NR_CPUS] array of full cache lines.

This patch uses existing infrastructure (percpu_modalloc() &
percpu_modfree()) to allocate percpu space for the refcount storage.

Instead of wasting NR_CPUS*128 bytes (on i386), we now use
nr_cpu_ids*sizeof(local_t) bytes.

On a typical distro, where NR_CPUS=8, shiping 2000 modules, we reduce
size of module files by about 2 Mbytes. (1Kb per module)

Instead of having all refcounters in the same memory node - with TLB misses
because of vmalloc() - this new implementation permits to have better
NUMA properties, since each  CPU will use storage on its preferred node,
thanks to percpu storage.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoocfs2: add quota call to ocfs2_remove_btree_range()
Mark Fasheh [Thu, 29 Jan 2009 23:06:21 +0000 (15:06 -0800)]
ocfs2: add quota call to ocfs2_remove_btree_range()

We weren't reclaiming the clusters which get free'd from this function,
so any user punching holes in a file would still have those bytes accounted
against him/her. Add the call to vfs_dq_free_space_nodirty() to fix this.
Interestingly enough, the journal credits calculation already took this into
account.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Acked-by: Jan Kara <jack@suse.cz>
15 years agoocfs2: Wakeup the downconvert thread after a successful cancel convert
Sunil Mushran [Fri, 30 Jan 2009 01:12:31 +0000 (17:12 -0800)]
ocfs2: Wakeup the downconvert thread after a successful cancel convert

When two nodes holding PR locks on a resource concurrently attempt to
upconvert the locks to EX, the master sends a BAST to one of the nodes. This
message tells that node to first cancel convert the upconvert request,
followed by downconvert to a NL. Only when this lock is downconverted to NL,
can the master upconvert the first node's lock to EX.

While the fs was doing the cancel convert, it was forgetting to wake up the
dc thread after a successful cancel, leading to a deadlock.

Reported-and-Tested-by: David Teigland <teigland@redhat.com>
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
15 years agoocfs2: Access the xattr bucket only before modifying it.
Tao Ma [Thu, 8 Jan 2009 00:21:43 +0000 (08:21 +0800)]
ocfs2: Access the xattr bucket only before modifying it.

In ocfs2_xattr_value_truncate, we may call b-tree codes which will
extend the journal transaction. It has a potential problem that it
may let the already-accessed-but-not-dirtied buffers gone. So we'd
better access the bucket after we call ocfs2_xattr_value_truncate.
And as for the root buffer for the xattr value, b-tree code will
acess and dirty it, so we don't need to worry about it.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
15 years agoconfigfs: Silence lockdep on mkdir(), rmdir() and configfs_depend_item()
Joel Becker [Wed, 17 Dec 2008 22:23:52 +0000 (14:23 -0800)]
configfs: Silence lockdep on mkdir(), rmdir() and configfs_depend_item()

When attaching default groups (subdirs) of a new group (in mkdir() or
in configfs_register()), configfs recursively takes inode's mutexes
along the path from the parent of the new group to the default
subdirs. This is needed to ensure that the VFS will not race with
operations on these sub-dirs. This is safe for the following reasons:

- the VFS allows one to lock first an inode and second one of its
  children (The lock subclasses for this pattern are respectively
  I_MUTEX_PARENT and I_MUTEX_CHILD);
- from this rule any inode path can be recursively locked in
  descending order as long as it stays under a single mountpoint and
  does not follow symlinks.

Unfortunately lockdep does not know (yet?) how to handle such
recursion.

I've tried to use Peter Zijlstra's lock_set_subclass() helper to
upgrade i_mutexes from I_MUTEX_CHILD to I_MUTEX_PARENT when we know
that we might recursively lock some of their descendant, but this
usage does not seem to fit the purpose of lock_set_subclass() because
it leads to several i_mutex locked with subclass I_MUTEX_PARENT by
the same task.

>From inside configfs it is not possible to serialize those recursive
locking with a top-level one, because mkdir() and rmdir() are already
called with inodes locked by the VFS. So using some
mutex_lock_nest_lock() is not an option.

I am proposing two solutions:
1) one that wraps recursive mutex_lock()s with
   lockdep_off()/lockdep_on().
2) (as suggested earlier by Peter Zijlstra) one that puts the
   i_mutexes recursively locked in different classes based on their
   depth from the top-level config_group created. This
   induces an arbitrary limit (MAX_LOCK_DEPTH - 2 == 46) on the
   nesting of configfs default groups whenever lockdep is activated
   but this limit looks reasonably high. Unfortunately, this alos
   isolates VFS operations on configfs default groups from the others
   and thus lowers the chances to detect locking issues.

This patch implements solution 1).

Solution 2) looks better from lockdep's point of view, but fails with
configfs_depend_item(). This needs to rework the locking
scheme of configfs_depend_item() by removing the variable lock recursion
depth, and I think that it's doable thanks to the configfs_dirent_lock.
For now, let's stick to solution 1).

Signed-off-by: Louis Rilling <louis.rilling@kerlabs.com>
Acked-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
15 years agoocfs2: Fix possible deadlock in ocfs2_write_dquot()
Jan Kara [Mon, 12 Jan 2009 22:20:32 +0000 (23:20 +0100)]
ocfs2: Fix possible deadlock in ocfs2_write_dquot()

It could happen that some limit has been set via quotactl() and in parallel
->mark_dirty() is called from another thread doing e.g. dquot_alloc_space(). In
such case ocfs2_write_dquot() must not try to sync the dquot because that needs
global quota lock but that ranks above transaction start.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
15 years agoocfs2: Push out dropping of dentry lock to ocfs2_wq
Jan Kara [Mon, 12 Jan 2009 22:20:31 +0000 (23:20 +0100)]
ocfs2: Push out dropping of dentry lock to ocfs2_wq

Dropping of last reference to dentry lock is a complicated operation involving
dropping of reference to inode. This can get complicated and quota code in
particular needs to obtain some quota locks which leads to potential deadlock.
Thus we defer dropping of inode reference to ocfs2_wq.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
15 years agopxamci: enable DMA for write ops after CMD/RESP
Cliff Brake [Thu, 22 Jan 2009 22:07:03 +0000 (17:07 -0500)]
pxamci: enable DMA for write ops after CMD/RESP

With the PXA270 MMC hardware, there seems to be an issue of
data corruption on writes where a 4KB data block is offset
by one byte.

If we delay enabling the DMA for writes until after the CMD/RESP
has finished, the problem seems to be fixed.

related to PXA270 Erratum #91

Tested-by: Vernon Sauder <VernonInHand@gmail.com>
Signed-off-by: Cliff Brake <cbrake@bec-systems.com>
Acked-by: Eric Miao <eric.miao@marvell.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
15 years agopxamci: replace #ifdef CONFIG_PXA27x with if (cpu_is_pxa27x())
Cliff Brake [Thu, 22 Jan 2009 21:58:58 +0000 (16:58 -0500)]
pxamci: replace #ifdef CONFIG_PXA27x with if (cpu_is_pxa27x())

Signed-off-by: Cliff Brake <cbrake@bec-systems.com>
Acked-by: Eric Miao <eric.miao@marvell.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
15 years agoricoh_mmc: Use suspend_late/resume_early
philipl@overt.org [Sun, 18 Jan 2009 19:11:20 +0000 (14:11 -0500)]
ricoh_mmc: Use suspend_late/resume_early

If ricoh_mmc suspends before sdhci_pci, it will pull the card
out from under the controller, which could leave the system in
a very confused state.

Using suspend_late/resume_early ensures that sdhci_pci suspends first
and resumes second.

Signed-off-by: Philip Langdale <philipl@overt.org>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
15 years agommci: Add support for ST Micro derivate
Linus Walleij [Sun, 4 Jan 2009 14:18:54 +0000 (15:18 +0100)]
mmci: Add support for ST Micro derivate

This patch adds support for the ST Microelectronics version of
the PL180 PrimeCell. They use designer ID 0x80 and have a few
alterations/bugfixes related to open drain and HW flow control.
They also add some SDIO registers, I am unsure if these are
in ST HW only or if this is things also added in later ARM
revisions, but they are included in the mmci.h file for
completeness.

Signed-off-by: Linus Walleij <linus.walleij@ericsson.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
15 years agommc: Add a MX2/MX3 specific SDHC driver
Sascha Hauer [Tue, 6 Jan 2009 16:04:14 +0000 (17:04 +0100)]
mmc: Add a MX2/MX3 specific SDHC driver

This patch adds a MX2/MX3 specific SDHC driver. The hardware is basically
the same as in the MX1, but unlike the MX1 controller the MX2
controller just works as expected. Since the MX1 driver has more
workarounds for bugs than anything else I had no success with supporting
MX1 and MX2 in a sane way in one driver.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
15 years agoicside: fix PCB version 6 support (v2)
Bartlomiej Zolnierkiewicz [Mon, 2 Feb 2009 19:12:23 +0000 (20:12 +0100)]
icside: fix PCB version 6 support (v2)

We need to pass struct ide_port_info also to ide_host_register().

v2:
Fix v5/v6 mismatch noticed by Russell.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agotx4939ide: typo fix and minor cleanup
Atsushi Nemoto [Mon, 2 Feb 2009 19:12:23 +0000 (20:12 +0100)]
tx4939ide: typo fix and minor cleanup

The bcount is greater than 0 and less than or equal to 0x10000.
Thus '(bcount & 0xffff) == 0x0000' can be simplified as 'bcount == 0x10000'.

Suggested-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agoide: add CS5536 host driver (v3)
Bartlomiej Zolnierkiewicz [Mon, 2 Feb 2009 19:12:23 +0000 (20:12 +0100)]
ide: add CS5536 host driver (v3)

This is a port of libata's pata_cs5536.c (written by Martin K. Petersen)
to IDE subsystem.

Changes done while at it:

* Reprogram PIO/MWDMA timings if needed before and after DMA transfer
  (chipset uses shared PIO/MWDMA timings).

* Fix cable detection to report 80-wires cable if BIOS set it for any
  device on a port (IDE core will do drive-side cable detection later).

* Don't disable UDMA while programming PIO timings.

* Simplify PCI/MSR support.

Pros of having IDE host driver in addition to libata's one:

* IDE is much lighter than SCSI+libata, the host driver itself is also
  a bit smaller:

   text    data     bss     dec     hex filename
   1261     496       4    1761     6e1 drivers/ata/pata_cs5536.o
   1242     128       4    1374     55e drivers/ide/cs5536.o

* This allows use of IDE features which are unavailable under libata.

v2:
* Fixes per review from Sergei:
  - simplify dependency check in Kconfig
  - use IDE_DRV_MASK also for ->drive_data
  - disable UDMA when programming MWDMA
  - program new DTC timings only when necessary
  - fix printk() level in cs5536_init_one()

* Fix patch description according to comments from Alan and Sergei.

v3:
* Smarter masking of UDMA bits per Sergei's suggestion.

Cc: Martin K. Petersen <mkp@mkp.net>
Cc: Karl Auerbach <karl@iwl.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agoide: Force VIA IDE legacy interrupts for AmigaOne boards
Gerhard Pircher [Mon, 2 Feb 2009 19:12:22 +0000 (20:12 +0100)]
ide: Force VIA IDE legacy interrupts for AmigaOne boards

The AmigaOne uses the onboard VIA IDE controller in legacy mode (like the
Pegasos).

Signed-off-by: Gerhard Pircher <gerhard_pircher@gmx.net>
Cc: "Grant Likely" <grant.likely@secretlab.ca>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agoIDE: Unregister and disable devices if initialization fails.
Ian Campbell [Mon, 2 Feb 2009 19:12:22 +0000 (20:12 +0100)]
IDE: Unregister and disable devices if initialization fails.

On reboot the loop in device_shutdown gets confused by these partially
initialized devices and goes into an infinite loop. Therefore unregister
and disable these devices.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
[bart: remove leftover hwif->present clearing + update patch description]
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agoide: fix ide_register_port() failure handling
Bartlomiej Zolnierkiewicz [Mon, 2 Feb 2009 19:12:21 +0000 (20:12 +0100)]
ide: fix ide_register_port() failure handling

* Factor out port freeing from ide_host_free() to ide_free_port().

* Add ide_disable_port() and use it on ide_register_port() failure.

Cc: Ian Campbell <Ian.Campbell@citrix.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agoide: struct device - replace bus_id with dev_name(), dev_set_name()
Kay Sievers [Mon, 2 Feb 2009 19:12:21 +0000 (20:12 +0100)]
ide: struct device - replace bus_id with dev_name(), dev_set_name()

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Cc: linux-ide@vger.kernel.org
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agoide-cd: fix DMA for non bio-backed requests
Borislav Petkov [Mon, 2 Feb 2009 19:12:21 +0000 (20:12 +0100)]
ide-cd: fix DMA for non bio-backed requests

This one fixes http://bugzilla.kernel.org/show_bug.cgi?id=12320.

Signed-off-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
15 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux...
David Vrabel [Mon, 2 Feb 2009 17:52:39 +0000 (17:52 +0000)]
Merge branch 'master' of git://git./linux/kernel/git/torvalds/linux-2.6 into for-upstream

15 years agoblock: add text file detailing queue/ sysfs files
Jens Axboe [Mon, 2 Feb 2009 12:02:31 +0000 (13:02 +0100)]
block: add text file detailing queue/ sysfs files

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
15 years agobio.h: If they MUST be inlined, then use __always_inline
Alberto Bertogli [Mon, 2 Feb 2009 11:41:07 +0000 (12:41 +0100)]
bio.h: If they MUST be inlined, then use __always_inline

bvec_kmap_irq() and bvec_kunmap_irq() comments say they MUST be inlined,
so mark them as __always_inline.

Signed-off-by: Alberto Bertogli <albertito@blitiri.com.ar>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
15 years agoFix misleading comment in bio.h
Alberto Bertogli [Mon, 2 Feb 2009 11:41:07 +0000 (12:41 +0100)]
Fix misleading comment in bio.h

The comment says "remember to add offset!", but the function already adds
it.

Signed-off-by: Alberto Bertogli <albertito@blitiri.com.ar>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
15 years agoMerge branches 'topic/slab/fixes' and 'topic/slub/fixes' into for-linus
Pekka Enberg [Mon, 2 Feb 2009 09:18:49 +0000 (11:18 +0200)]
Merge branches 'topic/slab/fixes' and 'topic/slub/fixes' into for-linus

15 years agoblock: fix inconsistent parenthesisation of QUEUE_FLAG_DEFAULT
Jens Axboe [Mon, 2 Feb 2009 07:43:48 +0000 (08:43 +0100)]
block: fix inconsistent parenthesisation of QUEUE_FLAG_DEFAULT

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
15 years agoblock: fix oops in blk_queue_io_stat()
Jens Axboe [Mon, 2 Feb 2009 07:42:32 +0000 (08:42 +0100)]
block: fix oops in blk_queue_io_stat()

Some initial probe requests don't have disk->queue mapped yet, so we
can't rely on a non-NULL queue in blk_queue_io_stat(). Wrap it in
blk_do_io_stat().

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
15 years agofbdev/atyfb: Fix DSP config on some PowerMacs & PowerBooks
Risto Suominen [Tue, 13 Jan 2009 20:09:30 +0000 (20:09 +0000)]
fbdev/atyfb: Fix DSP config on some PowerMacs & PowerBooks

Since the complete re-write in 2.6.10, some PowerMacs (At least PowerMac 5500
and PowerMac G3 Beige rev A) with ATI Mach64 chip have suffered from unstable
columns in their framebuffer image. This seems to depend on a value (4) read
from PLL_EXT_CNTL register, which leads to incorrect DSP config parameters to
be written to the chip. This patch uses a value calculated by aty_init_pll_ct
instead, as a starting point.

There are questions as to whether this should be extended to other platforms
or maybe made dependent on specific chip types, but in the meantime, this has
been tested on various powermacs and works for them so let's commit it.

Signed-off-by: Risto Suominen <Risto.Suominen@gmail.com>
Tested-by: Michael Pettersson <mike@it.uu.se>
Cc: <stable@kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Fix oops on some machines due to incorrect pr_debug()
Benjamin Herrenschmidt [Sun, 1 Feb 2009 17:03:59 +0000 (17:03 +0000)]
powerpc: Fix oops on some machines due to incorrect pr_debug()

Recently, a patch left DEBUG enabled in the powerpc common PCI code,
resulting in an old bug in a pr_debug() statement to show up and cause
a NULL dereference on some machines.

This fixes the pr_debug() statement and reverts to DEBUG not being
force-enabled in that file.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/ps3: Printing fixups for l64 to ll64 convserion drivers/net
Stephen Rothwell [Tue, 13 Jan 2009 20:09:30 +0000 (20:09 +0000)]
powerpc/ps3: Printing fixups for l64 to ll64 convserion drivers/net

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Geoff Levand <geoffrey.levand@am.sony.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agoManually revert "mlock: downgrade mmap sem while populating mlocked regions"
Linus Torvalds [Sun, 1 Feb 2009 19:00:16 +0000 (11:00 -0800)]
Manually revert "mlock: downgrade mmap sem while populating mlocked regions"

This essentially reverts commit 8edb08caf68184fb170f4f69c7445929e199eaea.

It downgraded our mmap semaphore to a read-lock while mlocking pages, in
order to allow other threads (and external accesses like "ps" et al) to
walk the vma lists and take page faults etc.  Which is a nice idea, but
the implementation does not work.

Because we cannot upgrade the lock back to a write lock without
releasing the mmap semaphore, the code had to release the lock entirely
and then re-take it as a writelock.  However, that meant that the caller
possibly lost the vma chain that it was following, since now another
thread could come in and mmap/munmap the range.

The code tried to work around that by just looking up the vma again and
erroring out if that happened, but quite frankly, that was just a buggy
hack that doesn't actually protect against anything (the other thread
could just have replaced the vma with another one instead of totally
unmapping it).

The only way to downgrade to a read map _reliably_ is to do it at the
end, which is likely the right thing to do: do all the 'vma' operations
with the write-lock held, then downgrade to a read after completing them
all, and then do the "populate the newly mlocked regions" while holding
just the read lock.  And then just drop the read-lock and return to user
space.

The (perhaps somewhat simpler) alternative is to just make all the
callers of mlock_vma_pages_range() know that the mmap lock got dropped,
and just re-grab the mmap semaphore if it needs to mlock more than one
vma region.

So we can do this "downgrade mmap sem while populating mlocked regions"
thing right, but the way it was done here was absolutely not correct.
Thus the revert, in the expectation that we will do it all correctly
some day.

Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'omap-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind...
Russell King [Sun, 1 Feb 2009 17:53:26 +0000 (17:53 +0000)]
Merge branch 'omap-fixes' of git://git./linux/kernel/git/tmlind/linux-omap-2.6

15 years agoV4L/DVB (10411): s5h1409: Perform s5h1409 soft reset after tuning
Devin Heitmueller [Mon, 26 Jan 2009 06:07:59 +0000 (03:07 -0300)]
V4L/DVB (10411): s5h1409: Perform s5h1409 soft reset after tuning

Just like with the s5h1411, the s5h1409 needs a soft-reset in order for it
to know that the tuner has been told to change frequencies.  This change
changes the behavior from "random tuning times between 500ms to complete
tuning lock failures" to "tuning lock consistently within 700ms".

Thanks to Robert Krakora <rob.krakora@messagenetsystems.com> for doing
initial testing of the patch on the KWorld 330U.

Thanks to Andy Walls <awalls@radix.net> for doing testing of the patch on
the HVR-1600.

Thanks to Michael Krufky <mkrufky@linuxtv.org> for doing additional testing.

Signed-off-by: Devin Heitmueller <dheitmueller@linuxtv.org>
Signed-off-by: Michael Krufky <mkrufky@linuxtv.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>