[SCSI] qla2xxx: Code changes to support new dynamic logging infrastructure.

The code is changed to support the new dynamic logging infrastructure.
Following are the levels added.
Default is 0 - no logging.  0x40000000 - Module Init & Probe.
0x20000000 - Mailbox Cmnds. 0x10000000 - Device Discovery.
0x08000000 - IO tracing.    0x04000000 - DPC Thread.
0x02000000 - Async events.  0x01000000 - Timer routines.
0x00800000 - User space.    0x00400000 - Task Management.
0x00200000 - AER/EEH.       0x00100000 - Multi Q.
0x00080000 - P3P Specific.  0x00040000 - Virtual Port.
0x00020000 - Buffer Dump.   0x00010000 - Misc.
0x7fffffff - For enabling all logs, can be too many logs.

Setting ql2xextended_error_logging module parameter to any of the above
value, will enable the debug for that particular level.
Do LOGICAL OR of the value to enable more than one level.

Signed-off-by: Saurav Kashyap <saurav.kashyap@qlogic.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: Joe Carnuccio <joe.carnuccio@qlogic.com>
Signed-off-by: Chad Dupuis <chad.dupuis@qlogic.com>
Signed-off-by: Madhuranath Iyengar <Madhu.Iyengar@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] qla2xxx: Basic infrastructure for dynamic logging.

This patch adds the dynamic logging framework to the qla2xxx driver.
The user will be able to change the logging levels on the fly i.e.
without load/unload of the driver. This also enables logging to be
enabled for a particular section of the driver such as initialization,
device discovery etc.

Signed-off-by: Saurav Kashyap <saurav.kashyap@qlogic.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] scsi_lib: pause between error retries

During cable pull tests on our 16G FC adapter, we are seeing errors,
typically reads to close targets, which fail due to CRC or framing
errors caused by the cable being pull (return status DID_ERROR).
The adapter detects the error on one of the first frames received,
marks the FC exchange as dead (further frames go to bit bucket) and
signals the host of the error. This action is so quick, and coupled
with fast host CPUs, creates a scenario in which the midlayer sees
the failure and retries the io almost immediately. We've seen link
traces with the retry on the link while the original i/o is still
being processed by the target. We're also seeing the time window
for the "link to pull-apart" and the physical interface to report
disconnected to be in the few millisecond range. Which means, we're
encountering scenarios where the full retry count is exhausted
(all with error) by the midlayer before the link disconnect state
is detected.

We looked at 8G FC behavior and occasionally see the same behavior,
but as the link was slower, it rarely could exhaust all retries
before the link reported disconnect.

What is needed is a slight delay between io retries due to DID_ERROR
to cover this error.  It is inappropriate to put this delay in the
driver, as the error is indistinguishable from other link-related errors,
nor does the driver track whether the io is a retry or not. This is also
easier than tracking between-io-error bursts that are seen in this
scenario.

The patch below updates the retry path so that it inserts a delay as
if the target was busy.  The busy delay is on the order of 6ms. This
delay is sufficient to ensure the link down condition is reported
before the retry count is exhausted (at most 1 retry is seen).

Signed-off-by: Alex Iannicelli <alex.iannicelli@emulex.com>
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mpt2sas: WarpDrive Infinite command retries due to wrong scsi command entry in MPI message

Issue:

This issue is seen on LSI H/W WarpDrive SSS6200 When filed direct I/O
is tried as volume I/O the scmd field in internal lookup table get
cleared and because of that the retried volume I/O never gets reported
as completed to SML.

Result:

I/O timeout and Error handling thread will kicking off

Fix:

Setting back the scmd in the lookup table before retrying the failed
direct i/o

Signed-off-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] bnx2fc: Replace printks with KERN_ALERT to KERN_ERR/KERN_INFO

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] scsi_transport_spi: Export host width and HBA id

Currently it's impossible to find out if the host supports
wide SCSI unless you're committed to trawl through syslog.
And it's near impossible to find the actual HBA id, which
is settable for some SCSI HBAs (like aic7xxx).
So export them via sysfs.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Ankit Jain <jankit@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: Add support for interrupt tasklet

Add support for interrupt tasklet, which will improve performance.
Correct spelling of "20011"

[jejb: simplified ifdefs and fixed unused variable problem]
Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: update comments

Remove obsolete comments and add new comments

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: misc improvements

Change code to match HBA datasheet.
Change code to make it readable.
Add support big endian for mvs_prd_imt.
Add cpu_to_le32 and cpu_to_le64 to use on addr.
Add scan_finished for structure mvs_prv_info.

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: Add new macros and functions

Add new macros: MVS_SOFT_RESET, MVS_HARD_RESET, MVS_PHY_TUNE,
MVS_COMMAND_ACTIVE, EXP_BRCT_CHG, MVS_MAX_SG
Add new member sg_width in struct mvs_chip_info
Use macros rather than magic number
Add new functions: mvs_fill_ssp_resp_iu, mvs_set_sense,
mvs_94xx_clear_srs_irq, mvs_94xx_phy_set_link_rate

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: Remove unused macros, variables and functions

Remove unused macros: VSR_PHY_VS0, VSR_PHY_VS1, MVS_SLOTS,
MVS_CAN_QUEUE, MVS_MSI, SG_MX, _MV_DUMP, MV_DISABLE_NCQ
Remove unused variables for mvs_info: irq, exp_req, cmd_size
Remove unused functions: mvs_get_sas_addr, mvs_hexdump,
mvs_hba_sb_dump, mvs_hab_memory_dump, mvs_hba_cq_dump

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: fix 94xx hotplug issue

Fix 94xx A0/B0 revision hotplug issue.
Remove unused macro: DISABLE_HOTPLUG_DMA_FIX

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: Add driver version and interrupt coalescing to device attributes in sysfs

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: add support for 94xx phy tuning and multiple revisions

Add 94xx phy tuning to aid manufacturing.
Add support for 94xx multiple revisions: A0, B0, C0, C1, C2.

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

[SCSI] mvsas: Add support for Non specific NCQ error interrupt

Signed-off-by: Xiangliang Yu <yuxiangl@marvell.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Merge 'akpm' patch series

* Merge akpm patch series: (122 commits)
  drivers/connector/cn_proc.c: remove unused local
  Documentation/SubmitChecklist: add RCU debug config options
  reiserfs: use hweight_long()
  reiserfs: use proper little-endian bitops
  pnpacpi: register disabled resources
  drivers/rtc/rtc-tegra.c: properly initialize spinlock
  drivers/rtc/rtc-twl.c: check return value of twl_rtc_write_u8() in twl_rtc_set_time()
  drivers/rtc: add support for Qualcomm PMIC8xxx RTC
  drivers/rtc/rtc-s3c.c: support clock gating
  drivers/rtc/rtc-mpc5121.c: add support for RTC on MPC5200
  init: skip calibration delay if previously done
  misc/eeprom: add eeprom access driver for digsy_mtc board
  misc/eeprom: add driver for microwire 93xx46 EEPROMs
  checkpatch.pl: update $logFunctions
  checkpatch: make utf-8 test --strict
  checkpatch.pl: add ability to ignore various messages
  checkpatch: add a "prefer __aligned" check
  checkpatch: validate signature styles and To: and Cc: lines
  checkpatch: add __rcu as a sparse modifier
  checkpatch: suggest using min_t or max_t
  ...

Did this as a merge because of (trivial) conflicts in
 - Documentation/feature-removal-schedule.txt
 - arch/xtensa/include/asm/uaccess.h
that were just easier to fix up in the merge than in the patch series.

drivers/connector/cn_proc.c: remove unused local

Fix the warning

  drivers/connector/cn_proc.c: In function 'proc_ptrace_connector':
  drivers/connector/cn_proc.c:176: warning: unused variable 'tracer'

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation/SubmitChecklist: add RCU debug config options

There have been persistent lockdep RCU splats, indicating that submitters
are not testing with CONFIG_PROVE_RCU.  Add this config option to the list
in Documentation/SubmitChecklist.  Also add CONFIG_DEBUG_OBJECTS_RCU_HEAD
for good measure.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

reiserfs: use hweight_long()

Use hweight_long() to count free bits in the bitmap.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

reiserfs: use proper little-endian bitops

Using __test_and_{set,clear}_bit_le() with ignoring its return value can
be replaced with __{set,clear}_bit_le().

This introduces reiserfs_{set,clear}_le_bit for __{set,clear}_bit_le and
does the above change with them.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pnpacpi: register disabled resources

When parsing PnP ACPI resource structures, it may happen that some of
the resources are disabled (in which case "the size" of the resource
equals zero).

The current solution is to skip these resources completely - with the
unfortunate side effect that they are not registered despite the fact
that they exist, after all.  (The downside of this approach is that
these resources cannot be used as templates for setting the actual
device's resources because they are missing from the template.) The
kernel's APM implementation does not suffer from this problem and
registers all resources regardless of "their size".

This patch fixes a problem with (at least) the vintage IBM ThinkPad 600E
(and most likely also with the 600, 600X, and 770X which have a very
similar layout) where some of its PnP devices support options where
either an IRQ, a DMA, or an IO port is disabled.  Without this patch,
the devices can not be configured using the
"/sys/bus/pnp/devices/*/resources" interface.

The manipulation of these resources is important because the 600E has
very demanding requirements.  For instance, the number of IRQs is not
sufficient to support all devices of the 600E.  Fortunately, some of the
devices, like the sound card's MPU-401 UART, can be configured to not
use any IRQ, hence freeing an IRQ for a device that requires one.
(Still, the device's "ResourceTemplate" requires an IRQ resource
descriptor which cannot be created if the resource has not been
registered in the first place.)

As an example, the dependent sets of the 600E's CSC0103 device (the
MPU-401 UART) are listed, with the patch applied, as:

  Dependent: 00 - Priority preferred
    port 0x300-0x330, align 0xf, size 0x4, 16-bit address decoding
    irq <none> High-Edge
  Dependent: 01 - Priority acceptable
    port 0x300-0x330, align 0xf, size 0x4, 16-bit address decoding
    irq 5,7,2/9,10,11,15 High-Edge

(The same result is obtained when PNPBIOS is used instead of PnP ACPI.)
Without the patch, the IRQ resource in the preferred option is not
listed at all:

  Dependent: 00 - Priority preferred
    port 0x300-0x330, align 0xf, size 0x4, 16-bit address decoding
  Dependent: 01 - Priority acceptable
    port 0x300-0x330, align 0xf, size 0x4, 16-bit address decoding
    irq 5,7,2/9,10,11,15 High-Edge

And in fact, the 600E's DSDT lists the disabled IRQ as an option, as can
be seen from the following excerpt from the DSDT:

Name (_PRS, ResourceTemplate ()
{
        StartDependentFn (0x00, 0x00)
        {
            IO (Decode16, 0x0300, 0x0330, 0x10, 0x04)
            IRQNoFlags () {}
        }
        StartDependentFn (0x01, 0x00)
        {
            IO (Decode16, 0x0300, 0x0330, 0x10, 0x04)
            IRQNoFlags () {5,7,9,10,11,15}
        }
        EndDependentFn ()
})

With this patch applied, a user space program - or maybe even the kernel
- can allocate all devices' resources optimally.  For the 600E, this
means to find optimal resources for (at least) the serial port, the
parallel port, the infrared port, the MWAVE modem, the sound card, and
the MPU-401 UART.

The patch applies the idea to register disabled resources to all types
of resources, not just to IRQs, DMAs, and IO ports.  At the same time,
it mimics the behavior of the "pnp_assign_xxx" functions from
"drivers/pnp/manager.c" where resources with "no size" are considered
disabled.

No regressions were observed on hardware that does not require this
patch.

The patch is applied against 2.6.39.

NB: The kernel's current PnP interface does not allow for disabling individual
resources using the "/sys/bus/pnp/devices/$device/resources" file.  Assuming
this could be done, a device could be configured to use a disabled resource
using a simple series of calls:

  echo disable > /sys/bus/pnp/devices/$device/resources
  echo clear > /sys/bus/pnp/devices/$device/resources
  echo set irq disabled > /sys/bus/pnp/devices/$device/resources
  echo fill > /sys/bus/pnp/devices/$device/resources
  echo activate > /sys/bus/pnp/devices/$device/resources

This patch addresses only the parsing of PnP ACPI devices.

ChangeLog (v1 -> v2):
 - extend patch description
 - fix typo in patch itself

Signed-off-by: Witold Szczeponik <Witold.Szczeponik@gmx.net>
Cc: Len Brown <lenb@kernel.org>
Cc: Adam Belay <abelay@mit.edu>
Cc: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Cc: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/rtc/rtc-tegra.c: properly initialize spinlock

Using __SPIN_LOCK_UNLOCKED for a dynamically allocated lock is wrong and
breaks the build with PREEMPT_RT_FULL.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: Andrew Chew <achew@nvidia.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/rtc/rtc-twl.c: check return value of twl_rtc_write_u8() in twl_rtc_set_time()

We forget to save the return value of the call to
twl_rtc_write_u8(save_control, REG_RTC_CTRL_REG); in 'ret', making the
test of 'ret < 0' dead code since 'ret' then couldn't possibly have
changed since the last test just a few lines above.  It also makes us not
detect failures from that specific twl_rtc_write_u8() call.

Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Alexandre Rusev <source@mvista.com>
Cc: "George G. Davis" <gdavis@mvista.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/rtc: add support for Qualcomm PMIC8xxx RTC

Add support for PMIC8xxx based RTC.  PMIC8xxx is Qualcomm's power
management IC that internally houses an RTC module.  This driver
communicates with the PMIC module over SSBI bus.

[akpm@linux-foundation.org: cosmetic tweaks]
Acked-by: Wan ZongShun <mcuos.com@gmail.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Anirudh Ghayal <aghayal@codeaurora.org>
Signed-off-by: Ashay Jaiswal <ashayj@codeaurora.org>
Cc: Samuel Ortiz <sameo@linux.intel.com>
Cc: Wan ZongShun <mcuos.com@gmail.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/rtc/rtc-s3c.c: support clock gating

Add support for clock gating.  Power consumption can be reduced by setting
rtc_clk disabled state except for when RTC related registers are accessed.

Signed-off-by: Donggeun Kim <dg77.kim@samsung.com>
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Signed-off-by: KyungMin Park <kyungmin.park@samsung.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Ben Dooks <ben@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/rtc/rtc-mpc5121.c: add support for RTC on MPC5200

MPC5200B contains a limited version of RTC from MPC5121.  Add support for
the RTC on that CPU.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Anatolij Gustschin <agust@denx.de>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

init: skip calibration delay if previously done

For each CPU, do the calibration delay only once.  For subsequent calls,
use the cached per-CPU value of loops_per_jiffy.

This saves about 200ms of resume time on dual core Intel Atom N5xx based
systems.  This helps bring down the kernel resume time on such systems
from about 500ms to about 300ms.

[akpm@linux-foundation.org: make cpu_loops_per_jiffy static]
[akpm@linux-foundation.org: clean up message text]
[akpm@linux-foundation.org: fix things up after upstream rmk changes]
Signed-off-by: Sameer Nanda <snanda@chromium.org>
Cc: Phil Carmody <ext-phil.2.carmody@nokia.com>
Cc: Andrew Worsley <amworsley@gmail.com>
Cc: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

misc/eeprom: add eeprom access driver for digsy_mtc board

Both displays on digsy_mtc board obtain their configuration from microwire
EEPROMs which are connected to the SoC over GPIO lines.  We need an easy
way to access the EEPROMs to write the needed display configuration or to
read out the currently programmed configuration.  The generic
eeprom_93xx46 SPI driver added by previous patch allows EEPROM access over
sysfs.  Using the simple driver added by this patch we provide used GPIO
interface and access control description on the board for generic
eeprom_93xx46 driver and spi_gpio driver.

Signed-off-by: Anatolij Gustschin <agust@denx.de>
Cc: Jonathan Cameron <jic23@cam.ac.uk>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

misc/eeprom: add driver for microwire 93xx46 EEPROMs

Add EEPROM driver for 93xx46 chips.  It can also be used with spi_gpio
driver to access 93xx46 EEPROMs connected over GPIO lines.  This driver
supports read/write/erase access to the EEPROM chips over sysfs files.

[rdunlap@xenotime.net: fix printk format]
Signed-off-by: Anatolij Gustschin <agust@denx.de>
Cc: Jonathan Cameron <jic23@cam.ac.uk>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

checkpatch.pl: update $logFunctions

Previous behavior allowed only alphabetic prefixes like pr_info to exceed
the 80 column line length limit.

ath6kl wants to add a digit into the prefix, so allow numbers as well as
digits in the <prefix>_<level> printks.

<prefix>_<level>_ratelimited and <prefix>_<level>_once and WARN_RATELIMIT
and WARN_ONCE may now exceed 80 cols.

Add missing <prefix>_printk type for completeness.

Signed-off-by: Joe Perches <joe@perches.com>
Cc: Andy Whitcroft <apw@canonical.com>
Cc: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

checkpatch: make utf-8 test --strict

Some patches are sent in using ISO-8859 or even Windows codepage 1252.

Make checkpatch accept these by default and only emit the "Invalid UTF-8"
message when using --strict.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

checkpatch.pl: add ability to ignore various messages

Some users would like the ability to not emit some of the messages that
checkpatch produces.  This can make it easier to use checkpatch in other
projects and integrate into scm hook scripts.

Add command line option to "--ignore" various message types.  Add option
--show-types to emit the "type" of each message.  Categorize all ERROR,
WARN and CHK messages with types.

Add optional .checkpatch.conf file to store default options.
3 paths are searched for .checkpatch.conf
    .             customized per-tree configurations
    $HOME         user global configuration when per-tree configs don't exist
    ./scripts     lk defaults to override script
The .conf file can contain any valid command-line argument and
the contents are prepended to any additional command line arguments.
Multiple lines may be used, blank lines are ignored, # is a comment.

Update "false positive" output for readability.

Update version to 0.32

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Cc: Andy Whitcroft <apw@canonical.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

checkpatch: add a "prefer __aligned" check

Prefer the use of __aligned(size) over __attribute__((__aligned___(size)))

Link: http://lkml.kernel.org/r/20110609094526.1571774c.akpm@linux-foundation.org
Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

checkpatch: validate signature styles and To: and Cc: lines

Signatures have many forms and can sometimes cause problems if not in the
correct format when using git send-email or quilt.

Try to verify the signature tags and email addresses to use the generally
accepted "Signed-off-by: Full Name <email@domain.tld>" form.

Original idea by Anish Kumar <anish198519851985@gmail.com>

Signed-off-by: Joe Perches <joe@perches.com>
Cc: Anish Kumar <anish198519851985@gmail.com>
Cc: Nick Bowler <nbowler@elliptictech.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

checkpatch: add __rcu as a sparse modifier

Fix "need consistent spacing around '*'" error after a __rcu sparse
annotation which was caused by the missing __rcu entry in the
checkpatch.pl internal list of sparse keywords.

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Cc: Andy Whitcroft <apw@canonical.com>
Cc: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

checkpatch: suggest using min_t or max_t

A common issue with min() or max() is using a cast on one or both of the
arguments when using min_t/max_t could be better.

Add cast detection to uses of min/max and suggest an appropriate use of
min_t or max_t instead.

Caveat:  This only works for min() or max() on a single line.
         It does not find min() or max() split across multiple lines.

This does find:
min((u32)foo, bar);
But it does not find:
max((unsigned long)foo,
    bar);

Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/firmware/sigma.c needs MODULE_LICENSE

Fix module tainting message:

  sigma: module license 'unspecified' taints kernel.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib: make _tolower() public

This function is required by *printf and kstrto* functions that are
located in the different modules.  This patch makes _tolower() public.
However, it's good idea to not use the helper outside of mentioned
functions.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/lcm.c: quiet sparse noise

The symbol 'lcm' is exported to the kernel (EXPORT_SYMBOL_GPL).
Pick up it's definition in <linux/lcm.h> to quiet the sparse noise:

  warning: symbol 'lcm' was not declared. Should it be static?

Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mach-ux500: add lm3530 ALS platform data for U5500

From: Shreshtha Kumar Sahu <shreshthakumar.sahu@stericsson.com>

platform data for simple backlight driver for LM3530
in the u5500 platform

Signed-off-by: Shreshtha Kumar Sahu <shreshthakumar.sahu@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arch/arm/mach-ux500/board-u5500.c: calibrate ALS input voltage

Provide the support for auto calibration of ALS Zone boundaries based on
min/max ALS input voltage.

Signed-off-by: Shreshtha Kumar Sahu <shreshthakumar.sahu@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/leds/leds-netxbig: make LEDS_NETXBIG depend on LEDS_CLASS

We call led_classdev_register/led_classdev_unregister in
create_netxbig_led/delete_netxbig_led, thus make LEDS_NETXBIG depend on
LEDS_CLASS.

This patch fixes below build error if LEDS_CLASS is not configured.

    LD      .tmp_vmlinux1
  drivers/built-in.o: In function `create_netxbig_led':
  drivers/leds/leds-netxbig.c:350: undefined reference to `led_classdev_register'
  drivers/leds/leds-netxbig.c:361: undefined reference to `led_classdev_unregister'
  drivers/built-in.o: In function `delete_netxbig_led':
  drivers/leds/leds-netxbig.c:313: undefined reference to `led_classdev_unregister'

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Acked-by: Simon Guinot <sguinot@lacie.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/leds/leds-sunfire.c: fix sunfire_led_generic_probe() error handling

- return -ENOMEM if kzalloc fails, rather than the current -EINVAL

- fix a memory leak in the case of goto out_unregister_led_cdevs

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/leds/leds-lp5521.c: provide section tagging

Tag the and remove() function as __devexit respectively.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MAINTAINERS: update HIGH RESOLUTION TIMERS patterns

clockchips.h was typoed as clockevents.h

Signed-off-by: Joe Perches <joe@perches.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

get_maintainers.pl: improve .mailmap parsing

Entries that used formats other than "Proper Name <commit@email.xx>"
were not parsed properly.

Try to improve the parsing so that the entries in the forms of:
    Proper Name <proper@email.xx> <commit@email.xx>
and
    Proper Name <proper@email.xx> Commit Name <commit@email.xx>
are transformed correctly.

Signed-off-by: Joe Perches <joe@perches.com>
Reviewed-by: Florian Mickler <florian@mickler.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/configs.c: include MODULE_*() when CONFIG_IKCONFIG_PROC=n

If CONFIG_IKCONFIG=m but CONFIG_IKCONFIG_PROC=n we get a module that has
no MODULE_LICENSE definition.  Move the MODULE_*() definitions outside the
CONFIG_IKCONFIG_PROC #ifdef to prevent this configuration from tainting
the kernel.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
Acked-by: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

notifiers: vt: move vt notifiers into vt.h

It is not necessary to share the same notifier.h.

Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: David Miller <davem@davemloft.net>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

notifiers: pm: move pm notifiers into suspend.h

It is not necessary to share the same notifier.h.

Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: David Miller <davem@davemloft.net>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

notifiers: sys: move reboot notifiers into reboot.h

It is not necessary to share the same notifier.h.

This patch already moves register_reboot_notifier() and
unregister_reboot_notifier() from kernel/notifier.c to kernel/sys.c.

[amwang@redhat.com: make allyesconfig succeed on ppc64]
Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: David Miller <davem@davemloft.net>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: WANG Cong <amwang@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

notifiers: net: move netdevice notifiers into netdevice.h

It is not necessary to share the same notifier.h.

Signed-off-by: WANG Cong <amwang@redhat.com>
Acked-by: David Miller <davem@davemloft.net>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

notifiers: cpu: move cpu notifiers into cpu.h

We presently define all kinds of notifiers in notifier.h.  This is not
necessary at all, since different subsystems use different notifiers, they
are almost non-related with each other.

This can also save much build time.  Suppose I add a new netdevice event,
really I don't have to recompile all the source, just network related.
Without this patch, all the source will be recompiled.

I move the notify events near to their subsystem notifier registers, so
that they can be found more easily.

This patch:

It is not necessary to share the same notifier.h.

Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: David Miller <davem@davemloft.net>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/misc: add support the FSA9480 USB Switch

The FSA9480 is a USB port accessory detector and switch.  This patch adds
support the FSA9480 USB Switch.

[akpm@linux-foundation.org: make a couple of things static]
Signed-off-by: Donggeun Kim <dg77.kim@samsung.com>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

geode: reflect mfgpt dependency on mfd

As stated in drivers/mfd/cs5535-mfd.c, the mfd driver exposes the BARs
which then make the GPIO, MFGPT, ACPI, etc.  all visible to the system.

So the dependencies of the MFGPT stuff have changed, and most people
expect Kconfig to bring in the necessary dependencies.  Without them, the
module fails to load and most people don't understand why because the
details of the rewrite aren't captured anywhere most people who know to
look.

This dependency needs to be reflected in Kconfig.

Signed-off-by: Philip A. Prindeville <philipp@redfish-solutions.com>
Acked-by: Alexandros C. Couloumbis <alex@ozo.com>
Acked-by: Andres Salomon <dilinger@queued.net>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

eisa/pci_eisa.c: fix section mismatch

Fixes

  WARNING: vmlinux.o(.data+0x15d3ac): Section mismatch in reference from the variable pci_eisa_driver to the function .init.text:pci_eisa_init()
  The variable pci_eisa_driver references the function __init pci_eisa_init()
  If the reference is valid then annotate the variable with __init* or __refdata (see linux/init.h) or name the variable:
  *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console

Signed-off-by: Arnaud Lacombe <lacombar@gmail.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/kernel.h: hide internal macros from userspace

Unexpose to userland the following macros
 - __FUNCTION__
 - NUMA_BUILD
 - COMPACTION_BUILD
 - REBUILD_DUE_TO_FTRACE_MCOUNT_RECORD

Signed-off-by: Arnaud Lacombe <lacombar@gmail.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Acked-by: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Michal Marek <mmarek@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/kernel.h: fix a headers_check warning

Fix the warning:

  usr/include/linux/kernel.h:65: userspace cannot reference function or variable defined in the kernel

As Michal noted, BUILD_BUG_ON stuffs should be moved
under #ifdef __KERNEL__.

Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Michal Marek <mmarek@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/ioport.h: new helper to define common struct resource constructs

Resource definitions that just define start, end and flags =
IORESOURCE_MEM or IORESOURCE_IRQ (with start=end) are quite common.  So
introduce a shortcut for them.  For completeness add macros for
IORESOURCE_DMA and IORESOURCE_IO, too and also make available a set of
macros to specify named resources of all types which are less common.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

devres: fix possible use after free

devres uses the pointer value as key after it's freed, which is safe but
triggers spurious use-after-free warnings on some static analysis tools.
Rearrange code to avoid such warnings.

Signed-off-by: Maxin B. John <maxin.john@gmail.com>
Reviewed-by: Rolf Eike Beer <eike-kernel@sf-tec.de>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

asm-generic/system.h: drop useless __KERNEL__

This header isn't exported to user-space, and even if it was, the
__KERNEL__ check covers the entire file, so we'd get a useless stub in the
first place.  So punt it.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers: use kzalloc/kcalloc instead of 'kmalloc+memset', where possible

Signed-off-by: Rakib Mullick <rakib.mullick@gmail.com>
Cc: Jeff Garzik <jgarzik@pobox.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Tejun Heo <tj@kernel.org>
Cc: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: remove dead code

GCC 4.6's -Wunused-but-set-variable found some dead code.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: adjust size of pid_buf

Linux can have pids up to 4*1024*1024.  To handle such huge numbers
pid_buf needs to be larger.

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: implement a x86_64 vDSO

Until now UML had no x86_64 vDSO.  So glibc always used the vsyscall page
for gettimeday() and friends.  Calls to gettimeday() returned falsely the
host time and confused some programs.

This patch adds a vDSO which turns all __vdso_* calls into a system call
so that UML can trap them.

As glibc still uses the vsyscall page for static binaries this patch
improves the situation only for dynamic binaries.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: set __HAVE_ARCH_GATE_AREA for x86_64

Implement arch_vma_name() and make get_gate_vma(), in_gate_area() and
in_gate_area_no_mm() a nop.

We need arch_vma_name() to support vDSO.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: Set __HAVE_ARCH_GATE_AREA for i386

When UML is unable to reuse the host's vDSO FIXADDR_USER_START is zero.
To handle this special case correclty we have to implement custom gate
area helper methods.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: disable scan_elf_aux() on x86_64

Reusing the host's vDSO makes only sense on x86_32.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: free resources

When creating the temp file there's a memory and file descriptor leak upon
error.

Signed-off-by: Davidlohr Bueso <dave@gnu.org>
Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Vitaliy Ivanov <vitalivanov@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: drivers/slip_user.c memory leak fix

Do not free memory when you failed to allocate it.

Signed-off-by: Vitaliy Ivanov <vitalivanov@gmail.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: helper.c warning corrections

Fix this warning:

  arch/um/os-Linux/helper.c: In function `helper_child':
  arch/um/os-Linux/helper.c:38:7: warning: ignoring return value of `write', declared with attribute warn_unused_result

[richard@nod.at: happens only with -D_FORTIFY_SOURCE=2]
Signed-off-by: Vitaliy Ivanov <vitalivanov@gmail.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: cow_user.c warning corrections

Fix this warning:

  arch/um/drivers/cow_user.c: In function `absolutize':
  arch/um/drivers/cow_user.c:189:7: warning: ignoring return value of `chdir', declared with attribute warn_unused_result

[richard@nod.at: happens only with -D_FORTIFY_SOURCE=2]
Signed-off-by: Vitaliy Ivanov <vitalivanov@gmail.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: drivers/net_user.c memory leak fix

Perform memory cleanup on exit.  On receiving invalid 'pid' we still
should clean 'output' variable.

Signed-off-by: Vitaliy Ivanov <vitalivanov@gmail.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: reinstate kernel version in generated .config

Commit 0954828fcbf3 ("kconfig: replace KERNELVERSION usage by the
mainmenu's prompt") made the kernel version disappear from the generated
.config file when configuring for UML.  As UML's Kconfig doesn't have a
mainmenu, kconfig falls back to the default string "Linux Kernel
Configuration".

Add a suitable mainmenu to the main UML Kconfig file to fix this.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: add netpoll support

To make netconsole usable on UML, its ethernet driver needs netpoll
support.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: fix _FORTIFY_SOURCE=2 support for kernel modules

When UML is compiled with _FORTIFY_SOURCE we have to export all _chk()
functions which are used in modules.  For now it's only the case for
__sprintf_chk().

Tested-by: Florian Fainelli <florian@openwrt.org>
Reported-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Richard Weinberger <richard@nod.at>
Acked-by: Vitaliy Ivanov <vitalivanov@gmail.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: clean up delay functions

Both sys-i386 and sys-x86_64 support now ndelay().  The delay functions
are based on arch/x86/lib/delay.c.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um, exec: remove redundant set_fs(USER_DS)

The address limit is already set in flush_old_exec() so this
set_fs(USER_DS) is redundant.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

um: clean up vm-flags.h

There is no need to define VM_{STACK,DATA}_DEFAULT_FLAGS as variable.
It's also useless to test for TIF_IA32 as 64bit UML has no IA32 emulation.

Signed-off-by: Richard Weinberger <richard@nod.at>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris, exec: remove redundant set_fs(USER_DS)

The address limit is already set in flush_old_exec() so those calls to
set_fs(USER_DS) are redundant.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Jesper Nilsson <jesper.nilsson@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris: fix some build warnings in pinmux.c

Fix some harmless warnings such as

  arch/cris/arch-v32/mach-a3/pinmux.c:273: warning: ISO C90 forbids mixed declarations and code:

Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Jesper Nilsson <jesper.nilsson@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

m68k, exec: remove redundant set_fs(USER_DS)

The address limit is already set in flush_old_exec() so those calls to
set_fs(USER_DS) are redundant.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Greg Ungerer <gerg@uclinux.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

m32r, exec: remove redundant set_fs(USER_DS)

The address limit is already set in flush_old_exec() so this
set_fs(USER_DS) is redundant.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alpha, exec: remove redundant set_fs(USER_DS)

The address limit is already set in flush_old_exec() so this
set_fs(USER_DS) is redundant.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

h8300, exec: remove redundant set_fs(USER_DS)

The address limit is already set in flush_old_exec() so those calls to
set_fs(USER_DS) are redundant.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

writeback: account NR_WRITTEN at IO completion time

NR_WRITTEN is now accounted at block IO enqueue time, which is not very
accurate as to common understanding.  This moves NR_WRITTEN accounting to
the IO completion time and makes it more consistent with BDI_WRITTEN,
which is used for bandwidth estimation.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: Michael Rubin <mrubin@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: simplify unuse and writepage

shmem_unuse_inode() and shmem_writepage() contain a little code to cope
with pages inserted independently into the filecache, probably by a
filesystem stacked on top of tmpfs, then fed to its ->readpage() or
->writepage().

Unionfs was indeed experimenting with working in that way three years ago,
but I find no current examples: nowadays the stacking filesystems use vfs
interfaces to the lower filesystem.

It's now illegal: remove most of that code, adding some WARN_ON_ONCEs.

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Erez Zadok <ezk@fsl.cs.sunysb.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: simplify filepage/swappage

We can now simplify shmem_getpage_gfp(): there is no longer a dilemma of
filepage passed in via shmem_readpage(), then swappage found, which must
then be copied over to it.

Although at first it's tempting to replace the **pagep arg by returning
struct page *, that makes a mess of IS_ERR_OR_NULL(page)s in all the
callers, so leave as is.

Insert BUG_ON(!PageUptodate) when we find and lock page: some of the
complication came from uninitialized pages inserted into filecache prior
to readpage; but now we're in control, and only release pagelock on
filecache once it's uptodate (if an error occurs in reading back from
swap, the page remains in swapcache, never moved to filecache).

Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: simplify prealloc_page

The prealloc_page handling in shmem_getpage_gfp() is unnecessarily
complicated: first simplify that before going on to filepage/swappage.

That's right, don't report ENOMEM when the preallocation fails: we may or
may not need the page.  But simply report ENOMEM once we find we do need
it, instead of dropping lock, repeating allocation, unwinding on failure
etc.  And leave the out label on the fast path, don't goto.

Fix something that looks like a bug but turns out not to be: set
PageSwapBacked on prealloc_page before its mem_cgroup_cache_charge(), as
the removed case was doing.  That's important before adding to LRU
(determines which LRU the page goes on), and does affect which path it
takes through memcontrol.c, but in the end MEM_CGROUP_CHANGE_TYPE_ SHMEM
is handled no differently from CACHE.

Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Shaohua Li <shaohua.li@intel.com>
Cc: "Zhang, Yanmin" <yanmin.zhang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: remove_shmem_readpage

Remove that pernicious shmem_readpage() at last: the things we needed it
for (splice, loop, sendfile, i915 GEM) are now fully taken care of by
shmem_file_splice_read() and shmem_read_mapping_page_gfp().

This removal clears the way for a simpler shmem_getpage_gfp(), since page
is never passed in; but leave most of that cleanup until after.

sys_readahead() and sys_fadvise(POSIX_FADV_WILLNEED) will now EINVAL,
instead of unexpectedly trying to read ahead on tmpfs: if that proves to
be an issue for someone, then we can either arrange for them to return
success instead, or try to implement async readahead on tmpfs.

Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: pass gfp to shmem_getpage_gfp

Make shmem_getpage() a wrapper, passing mapping_gfp_mask() down to
shmem_getpage_gfp(), which in turn passes gfp down to shmem_swp_alloc().

Change shmem_read_mapping_page_gfp() to use shmem_getpage_gfp() in the
CONFIG_SHMEM case; but leave tiny !SHMEM using read_cache_page_gfp().

Add a BUG_ON() in case anyone happens to call this on a non-shmem mapping;
though we might later want to let that case route to read_cache_page_gfp().

It annoys me to have these two almost-redundant args, gfp and fault_type:
I can't find a better way; but initialize fault_type only in shmem_fault().

Note that before, read_cache_page_gfp() was allocating i915_gem's pages
with __GFP_NORETRY as intended; but the corresponding swap vector pages
got allocated without it, leaving a small possibility of OOM.

Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: refine shmem_file_splice_read

Tidy up shmem_file_splice_read():

Remove readahead: okay, we could implement shmem readahead on swap,
but have never done so before, swap being the slow exceptional path.

Use shmem_getpage() instead of find_or_create_page() plus ->readpage().

Remove several comments: sorry, I found them more distracting than
helpful, and this will not be the reference version of splice_read().

Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: clone shmem_file_splice_read()

Copy __generic_file_splice_read() and generic_file_splice_read() from
fs/splice.c to shmem_file_splice_read() in mm/shmem.c.  Make
page_cache_pipe_buf_ops and spd_release_page() accessible to it.

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Jens Axboe <jaxboe@fusionio.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/futex: fix futex writes on archs with SW tracking of dirty & young

I haven't reproduced it myself but the fail scenario is that on such
machines (notably ARM and some embedded powerpc), if you manage to hit
that futex path on a writable page whose dirty bit has gone from the PTE,
you'll livelock inside the kernel from what I can tell.

It will go in a loop of trying the atomic access, failing, trying gup to
"fix it up", getting succcess from gup, go back to the atomic access,
failing again because dirty wasn't fixed etc...

So I think you essentially hang in the kernel.

The scenario is probably rare'ish because affected architecture are
embedded and tend to not swap much (if at all) so we probably rarely hit
the case where dirty is missing or young is missing, but I think Shan has
a piece of SW that can reliably reproduce it using a shared writable
mapping & fork or something like that.

On archs who use SW tracking of dirty & young, a page without dirty is
effectively mapped read-only and a page without young unaccessible in the
PTE.

Additionally, some architectures might lazily flush the TLB when relaxing
write protection (by doing only a local flush), and expect a fault to
invalidate the stale entry if it's still present on another processor.

The futex code assumes that if the "in_atomic()" access -EFAULT's, it can
"fix it up" by causing get_user_pages() which would then be equivalent to
taking the fault.

However that isn't the case.  get_user_pages() will not call
handle_mm_fault() in the case where the PTE seems to have the right
permissions, regardless of the dirty and young state.  It will eventually
update those bits ...  in the struct page, but not in the PTE.

Additionally, it will not handle the lazy TLB flushing that can be
required by some architectures in the fault case.

Basically, gup is the wrong interface for the job.  The patch provides a
more appropriate one which boils down to just calling handle_mm_fault()
since what we are trying to do is simulate a real page fault.

The futex code currently attempts to write to user memory within a
pagefault disabled section, and if that fails, tries to fix it up using
get_user_pages().

This doesn't work on archs where the dirty and young bits are maintained
by software, since they will gate access permission in the TLB, and will
not be updated by gup().

In addition, there's an expectation on some archs that a spurious write
fault triggers a local TLB flush, and that is missing from the picture as
well.

I decided that adding those "features" to gup() would be too much for this
already too complex function, and instead added a new simpler
fixup_user_fault() which is essentially a wrapper around handle_mm_fault()
which the futex code can call.

[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix some nits Darren saw, fiddle comment layout]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reported-by: Shan Hai <haishan.bai@gmail.com>
Tested-by: Shan Hai <haishan.bai@gmail.com>
Cc: David Laight <David.Laight@ACULAB.COM>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Darren Hart <darren.hart@intel.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove useless rcu lock-unlock from mapping_tagged()

radix_tree_tagged() is lockless - it reads from a member of the raid-tree
root node.  It does not require any protection.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: page allocator: reconsider zones for allocation after direct reclaim

With zone_reclaim_mode enabled, it's possible for zones to be considered
full in the zonelist_cache so they are skipped in the future.  If the
process enters direct reclaim, the ZLC may still consider zones to be full
even after reclaiming pages.  Reconsider all zones for allocation if
direct reclaim returns successfully.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: page allocator: initialise ZLC for first zone eligible for zone_reclaim

There have been a small number of complaints about significant stalls
while copying large amounts of data on NUMA machines reported on a
distribution bugzilla.  In these cases, zone_reclaim was enabled by
default due to large NUMA distances.  In general, the complaints have not
been about the workload itself unless it was a file server (in which case
the recommendation was disable zone_reclaim).

The stalls are mostly due to significant amounts of time spent scanning
the preferred zone for pages to free.  After a failure, it might fallback
to another node (as zonelists are often node-ordered rather than
zone-ordered) but stall quickly again when the next allocation attempt
occurs.  In bad cases, each page allocated results in a full scan of the
preferred zone.

Patch 1 checks the preferred zone for recent allocation failure
        which is particularly important if zone_reclaim has failed
        recently.  This avoids rescanning the zone in the near future and
        instead falling back to another node.  This may hurt node locality
        in some cases but a failure to zone_reclaim is more expensive than
        a remote access.

Patch 2 clears the zlc information after direct reclaim.
        Otherwise, zone_reclaim can mark zones full, direct reclaim can
        reclaim enough pages but the zone is still not considered for
        allocation.

This was tested on a 24-thread 2-node x86_64 machine.  The tests were
focused on large amounts of IO.  All tests were bound to the CPUs on
node-0 to avoid disturbances due to processes being scheduled on different
nodes.  The kernels tested are

3.0-rc6-vanilla Vanilla 3.0-rc6
zlcfirst Patch 1 applied
zlcreconsider Patches 1+2 applied

FS-Mark
./fs_mark  -d  /tmp/fsmark-10813  -D  100  -N  5000  -n  208  -L  35  -t  24  -S0  -s  524288
                fsmark-3.0-rc6       3.0-rc6        3.0-rc6
                   vanilla  zlcfirs  zlcreconsider
Files/s  min          54.90 ( 0.00%)       49.80 (-10.24%)       49.10 (-11.81%)
Files/s  mean        100.11 ( 0.00%)      135.17 (25.94%)      146.93 (31.87%)
Files/s  stddev       57.51 ( 0.00%)      138.97 (58.62%)      158.69 (63.76%)
Files/s  max         361.10 ( 0.00%)      834.40 (56.72%)      802.40 (55.00%)
Overhead min       76704.00 ( 0.00%)    76501.00 ( 0.27%)    77784.00 (-1.39%)
Overhead mean    1485356.51 ( 0.00%)  1035797.83 (43.40%)  1594680.26 (-6.86%)
Overhead stddev  1848122.53 ( 0.00%)   881489.88 (109.66%)  1772354.90 ( 4.27%)
Overhead max     7989060.00 ( 0.00%)  3369118.00 (137.13%) 10135324.00 (-21.18%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds)        501.49    493.91    499.93
Total Elapsed Time (seconds)               2451.57   2257.48   2215.92

MMTests Statistics: vmstat
Page Ins                                       46268       63840       66008
Page Outs                                   90821596    90671128    88043732
Swap Ins                                           0           0           0
Swap Outs                                          0           0           0
Direct pages scanned                        13091697     8966863     8971790
Kswapd pages scanned                               0     1830011     1831116
Kswapd pages reclaimed                             0     1829068     1829930
Direct pages reclaimed                      13037777     8956828     8648314
Kswapd efficiency                               100%         99%         99%
Kswapd velocity                                0.000     810.643     826.346
Direct efficiency                                99%         99%         96%
Direct velocity                             5340.128    3972.068    4048.788
Percentage direct scans                         100%         83%         83%
Page writes by reclaim                             0           3           0
Slabs scanned                                 796672      720640      720256
Direct inode steals                          7422667     7160012     7088638
Kswapd inode steals                                0     1736840     2021238

Test completes far faster with a large increase in the number of files
created per second.  Standard deviation is high as a small number of
iterations were much higher than the mean.  The number of pages scanned by
zone_reclaim is reduced and kswapd is used for more work.

LARGE DD
                3.0-rc6       3.0-rc6       3.0-rc6
                    vanilla     zlcfirst     zlcreconsider
download tar           59 ( 0.00%)   59 ( 0.00%)   55 ( 7.27%)
dd source files       527 ( 0.00%)  296 (78.04%)  320 (64.69%)
delete source          36 ( 0.00%)   19 (89.47%)   20 (80.00%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds)        125.03    118.98    122.01
Total Elapsed Time (seconds)                624.56    375.02    398.06

MMTests Statistics: vmstat
Page Ins                                     3594216      439368      407032
Page Outs                                   23380832    23380488    23377444
Swap Ins                                           0           0           0
Swap Outs                                          0         436         287
Direct pages scanned                        17482342    69315973    82864918
Kswapd pages scanned                               0      519123      575425
Kswapd pages reclaimed                             0      466501      522487
Direct pages reclaimed                       5858054     2732949     2712547
Kswapd efficiency                               100%         89%         90%
Kswapd velocity                                0.000    1384.254    1445.574
Direct efficiency                                33%          3%          3%
Direct velocity                            27991.453  184832.737  208171.929
Percentage direct scans                         100%         99%         99%
Page writes by reclaim                             0        5082       13917
Slabs scanned                                  17280       29952       35328
Direct inode steals                           115257     1431122      332201
Kswapd inode steals                                0           0      979532

This test downloads a large tarfile and copies it with dd a number of
times - similar to the most recent bug report I've dealt with.  Time to
completion is reduced.  The number of pages scanned directly is still
disturbingly high with a low efficiency but this is likely due to the
number of dirty pages encountered.  The figures could probably be improved
with more work around how kswapd is used and how dirty pages are handled
but that is separate work and this result is significant on its own.

Streaming Mapped Writer
MMTests Statistics: duration
User/Sys Time Running Test (seconds)        124.47    111.67    112.64
Total Elapsed Time (seconds)               2138.14   1816.30   1867.56

MMTests Statistics: vmstat
Page Ins                                       90760       89124       89516
Page Outs                                  121028340   120199524   120736696
Swap Ins                                           0          86          55
Swap Outs                                          0           0           0
Direct pages scanned                       114989363    96461439    96330619
Kswapd pages scanned                        56430948    56965763    57075875
Kswapd pages reclaimed                      27743219    27752044    27766606
Direct pages reclaimed                         49777       46884       36655
Kswapd efficiency                                49%         48%         48%
Kswapd velocity                            26392.541   31363.631   30561.736
Direct efficiency                                 0%          0%          0%
Direct velocity                            53780.091   53108.759   51581.004
Percentage direct scans                          67%         62%         62%
Page writes by reclaim                           385         122        1513
Slabs scanned                                  43008       39040       42112
Direct inode steals                                0          10           8
Kswapd inode steals                              733         534         477

This test just creates a large file mapping and writes to it linearly.
Time to completion is again reduced.

The gains are mostly down to two things.  In many cases, there is less
scanning as zone_reclaim simply gives up faster due to recent failures.
The second reason is that memory is used more efficiently.  Instead of
scanning the preferred zone every time, the allocator falls back to
another zone and uses it instead improving overall memory utilisation.

This patch: initialise ZLC for first zone eligible for zone_reclaim.

The zonelist cache (ZLC) is used among other things to record if
zone_reclaim() failed for a particular zone recently.  The intention is to
avoid a high cost scanning extremely long zonelists or scanning within the
zone uselessly.

Currently the zonelist cache is setup only after the first zone has been
considered and zone_reclaim() has been called.  The objective was to avoid
a costly setup but zone_reclaim is itself quite expensive.  If it is
failing regularly such as the first eligible zone having mostly mapped
pages, the cost in scanning and allocation stalls is far higher than the
ZLC initialisation step.

This patch initialises ZLC before the first eligible zone calls
zone_reclaim().  Once initialised, it is checked whether the zone failed
zone_reclaim recently.  If it has, the zone is skipped.  As the first zone
is now being checked, additional care has to be taken about zones marked
full.  A zone can be marked "full" because it should not have enough
unmapped pages for zone_reclaim but this is excessive as direct reclaim or
kswapd may succeed where zone_reclaim fails.  Only mark zones "full" after
zone_reclaim fails if it failed to reclaim enough pages after scanning.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: preallocate page before lock_page() at filemap COW

Currently we are keeping faulted page locked throughout whole __do_fault
call (except for page_mkwrite code path) after calling file system's fault
code.  If we do early COW, we allocate a new page which has to be charged
for a memcg (mem_cgroup_newpage_charge).

This function, however, might block for unbounded amount of time if memcg
oom killer is disabled or fork-bomb is running because the only way out of
the OOM situation is either an external event or OOM-situation fix.

In the end we are keeping the faulted page locked and blocking other
processes from faulting it in which is not good at all because we are
basically punishing potentially an unrelated process for OOM condition in
a different group (I have seen stuck system because of ld-2.11.1.so being
locked).

We can do test easily.

 % cgcreate -g memory:A
 % cgset -r memory.limit_in_bytes=64M A
 % cgset -r memory.memsw.limit_in_bytes=64M A
 % cd kernel_dir; cgexec -g memory:A make -j

Then, the whole system will live-locked until you kill 'make -j'
by hands (or push reboot...) This is because some important page in a
a shared library are locked.

Considering again, the new page is not necessary to be allocated
with lock_page() held. And usual page allocation may dive into
long memory reclaim loop with holding lock_page() and can cause
very long latency.

There are 3 ways.
  1. do allocation/charge before lock_page()
     Pros. - simple and can handle page allocation in the same manner.
             This will reduce holding time of lock_page() in general.
     Cons. - we do page allocation even if ->fault() returns error.

  2. do charge after unlock_page(). Even if charge fails, it's just OOM.
     Pros. - no impact to non-memcg path.
     Cons. - implemenation requires special cares of LRU and we need to modify
             page_add_new_anon_rmap()...

  3. do unlock->charge->lock again method.
     Pros. - no impact to non-memcg path.
     Cons. - This may kill LOCK_PAGE_RETRY optimization. We need to release
             lock and get it again...

This patch moves "charge" and memory allocation for COW page
before lock_page(). Then, we can avoid scanning LRU with holding
a lock on a page and latency under lock_page() will be reduced.

Then, above livelock disappears.

[akpm@linux-foundation.org: fix code layout]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reported-by: Lutz Vieweg <lvml@5t9.de>
Original-idea-by: Michal Hocko <mhocko@suse.cz>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Ying Han <yinghan@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tmpfs: no need to use i_lock

2.6.36's 7e496299d4d2 ("tmpfs: make tmpfs scalable with percpu_counter for
used blocks") to make tmpfs scalable with percpu_counter used
inode->i_lock in place of sbinfo->stat_lock around i_blocks updates; but
that was adverse to scalability, and unnecessary, since info->lock is
already held there in the fast paths.

Remove those uses of i_lock, and add info->lock in the three error paths
where it's then needed across shmem_free_blocks().  It's not actually
needed across shmem_unacct_blocks(), but they're so often paired that it
looks wrong to split them apart.

Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: pincer in truncate_inode_pages_range

truncate_inode_pages_range()'s final loop has a nice pincer property,
bringing start and end together, squeezing out the last pages.  But the
range handling missed out on that, just sliding up the range, perhaps
letting pages come in behind it.  Add one more test to give it the same
pincer effect.

Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: consistent truncate and invalidate loops

Make the pagevec_lookup loops in truncate_inode_pages_range(),
invalidate_mapping_pages() and invalidate_inode_pages2_range() more
consistent with each other.

They were relying upon page->index of an unlocked page, but apologizing
for it: accept it, embrace it, add comments and WARN_ONs, and simplify the
index handling.

invalidate_inode_pages2_range() had special handling for a wrapped
page->index + 1 = 0 case; but MAX_LFS_FILESIZE doesn't let us anywhere
near there, and a corrupt page->index in the radix_tree could cause more
trouble than that would catch.  Remove that wrapped handling.

invalidate_inode_pages2_range() uses min() to limit the pagevec_lookup
when near the end of the range: copy that into the other two, although
it's less useful than you might think (it limits the use of the buffer,
rather than the indices looked up).

Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>