linux-2.6-microblaze.git
9 years agommc: mmc: extend the mmc_send_tuning()
Chaotian Jing [Tue, 27 Oct 2015 06:24:28 +0000 (14:24 +0800)]
mmc: mmc: extend the mmc_send_tuning()

The mmc_execute_tuning() has already prepared the opcode,
there is no need to prepare it again at mmc_send_tuning(),
and, there is a BUG of mmc_send_tuning() to determine the opcode
by bus width, assume eMMC was running at HS200, 4bit mode,
then the mmc_send_tuning() will overwrite the opcode from CMD21
to CMD19, then got error.

in addition, extend an argument of "cmd_error" to allow getting
if there was cmd error when tune response.

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
[Ulf: Rebased patch]
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mediatek: add implement of ops->hw_reset()
Chaotian Jing [Tue, 27 Oct 2015 06:24:26 +0000 (14:24 +0800)]
mmc: mediatek: add implement of ops->hw_reset()

add implement of ops->hw_reset() for eMMC

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mediatek: fix got GPD checksum error interrupt when data transfer
Chaotian Jing [Tue, 27 Oct 2015 06:24:25 +0000 (14:24 +0800)]
mmc: mediatek: fix got GPD checksum error interrupt when data transfer

Even if we only use one gpd, we need alloc 2 gpd and make
the gpd->next pointer to the second gpd, or may get gpd checksum
error, this was checked by hardware

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mediatek: change the argument "ddr" to "timing"
Chaotian Jing [Tue, 27 Oct 2015 06:24:24 +0000 (14:24 +0800)]
mmc: mediatek: change the argument "ddr" to "timing"

use the ios->timing directly is better
It can reflect current timing and do settings by timing

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mediatek: make cmd_ints_mask to const
Chaotian Jing [Tue, 27 Oct 2015 06:24:23 +0000 (14:24 +0800)]
mmc: mediatek: make cmd_ints_mask to const

cmd_ints_mask and data_ints_mask are constant value,
so make it to const

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dt-bindings: update Mediatek MMC bindings
Chaotian Jing [Tue, 27 Oct 2015 06:24:22 +0000 (14:24 +0800)]
mmc: dt-bindings: update Mediatek MMC bindings

Add 400Mhz clock source for HS400 mode

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Add DT bindings for eMMC hardware reset support
Chaotian Jing [Tue, 27 Oct 2015 06:24:21 +0000 (14:24 +0800)]
mmc: core: Add DT bindings for eMMC hardware reset support

Sometime only need set MMC_CAP_HW_RESET for one of MMC hosts,
So set it in device tree is better.

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: Enable omap_hsmmc for Keystone 2
Lokesh Vutla [Tue, 20 Oct 2015 21:50:06 +0000 (16:50 -0500)]
mmc: omap_hsmmc: Enable omap_hsmmc for Keystone 2

Enable omap_hsmmc for Keystone 2 architecture which reuses the HSMMC
IP found on OMAP platforms.

Signed-off-by: Franklin S Cooper Jr <fcooper@ti.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-acpi: Add more ACPI HIDs for Intel controllers
Adrian Hunter [Wed, 21 Oct 2015 08:15:46 +0000 (11:15 +0300)]
mmc: sdhci-acpi: Add more ACPI HIDs for Intel controllers

Add ACPI HIDs for Intel host controllers including one
supporting HS400.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: Add more PCI IDs for Intel controllers
Adrian Hunter [Wed, 21 Oct 2015 08:15:45 +0000 (11:15 +0300)]
mmc: sdhci-pci: Add more PCI IDs for Intel controllers

Add PCI IDs for Intel host controllers

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoarm: lpc18xx_defconfig: remove CONFIG_MMC_DW_IDMAC
Shawn Lin [Wed, 16 Sep 2015 06:43:01 +0000 (14:43 +0800)]
arm: lpc18xx_defconfig: remove CONFIG_MMC_DW_IDMAC

DesignWare MMC Controller's transfer mode should be decided
at runtime instead of compile-time. So we remove this config
option and read dw_mmc's register to select DMA master.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoarm: hisi_defconfig: remove CONFIG_MMC_DW_IDMAC
Shawn Lin [Wed, 16 Sep 2015 06:42:49 +0000 (14:42 +0800)]
arm: hisi_defconfig: remove CONFIG_MMC_DW_IDMAC

DesignWare MMC Controller's transfer mode should be decided
at runtime instead of compile-time. So we remove this config
option and read dw_mmc's register to select DMA master.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Wei Xu <xuwei5@hisilicon.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoarm: exynos_defconfig: remove CONFIG_MMC_DW_IDMAC
Shawn Lin [Wed, 16 Sep 2015 06:42:37 +0000 (14:42 +0800)]
arm: exynos_defconfig: remove CONFIG_MMC_DW_IDMAC

DesignWare MMC Controller's transfer mode should be decided
at runtime instead of compile-time. So we remove this config
option and read dw_mmc's register to select DMA master.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoarc: axs10x_defconfig: remove CONFIG_MMC_DW_IDMAC
Shawn Lin [Wed, 16 Sep 2015 06:42:20 +0000 (14:42 +0800)]
arc: axs10x_defconfig: remove CONFIG_MMC_DW_IDMAC

DesignWare MMC Controller's transfer mode should be decided
at runtime instead of compile-time. So we remove this config
option and read dw_mmc's register to select DMA master.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agomips: pistachio_defconfig: remove CONFIG_MMC_DW_IDMAC
Shawn Lin [Wed, 16 Sep 2015 06:42:05 +0000 (14:42 +0800)]
mips: pistachio_defconfig: remove CONFIG_MMC_DW_IDMAC

DesignWare MMC Controller's transfer mode should be decided
at runtime instead of compile-time. So we remove this config
option and read dw_mmc's register to select DMA master.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Govindraj Raja <govindraj.raja@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoDocumentation: synopsys-dw-mshc: add bindings for idmac and edmac
Shawn Lin [Wed, 16 Sep 2015 06:41:50 +0000 (14:41 +0800)]
Documentation: synopsys-dw-mshc: add bindings for idmac and edmac

synopsys-dw-mshc supports three types of transfer mode. We add
bindings and description for how to use them at runtime.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: use macro for HCON register operations
Shawn Lin [Wed, 16 Sep 2015 06:41:37 +0000 (14:41 +0800)]
mmc: dw_mmc: use macro for HCON register operations

This patch add some macros for HCON register operations
to make code more readable.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Add external dma interface support
Shawn Lin [Wed, 16 Sep 2015 06:41:23 +0000 (14:41 +0800)]
mmc: dw_mmc: Add external dma interface support

DesignWare MMC Controller can supports two types of DMA
mode: external dma and internal dma. We get a RK312x platform
integrated dw_mmc and ARM pl330 dma controller. This patch add
edmac ops to support these platforms. I've tested it on RK31xx
platform with edmac mode and RK3288 platform with idmac mode.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-of-esdhc: avoid writing power control register
yangbo lu [Fri, 16 Oct 2015 07:44:03 +0000 (15:44 +0800)]
mmc: sdhci-of-esdhc: avoid writing power control register

The eSDHC doesn't have a standard power control register, so when
writing this register in stack we should do nothing to avoid
incorrect operation.

Signed-off-by: Yangbo Lu <yangbo.lu@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoDT: mmc: sh_mmcif: document R8A7794 support
Sergei Shtylyov [Thu, 15 Oct 2015 22:40:16 +0000 (01:40 +0300)]
DT: mmc: sh_mmcif: document R8A7794 support

Renesas R8A7794 SoC also has the MMCIF controller.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoDT: mmc: sh_mmcif: fix "compatible" property text
Sergei Shtylyov [Thu, 15 Oct 2015 22:39:39 +0000 (01:39 +0300)]
DT: mmc: sh_mmcif: fix "compatible" property text

The "compatible" property text contradicts even the example given in the MMCIF
binding document itself;  moreover, the Renesas MMCIF driver only matches  on
the generic "compatible" string and doesn't look for the SoC specific strings
at all. Thus describe "renesas,sh-mmcif" as a fallback value.

Fixes: b4c27763d749 ("mmc: sh_mmcif: Document DT bindings")
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: skip reclaiming host on mmc_add_card() error
Sergei Shtylyov [Wed, 14 Oct 2015 20:53:03 +0000 (23:53 +0300)]
mmc: skip reclaiming host on mmc_add_card() error

There's little sense in releasing the host on mmc_add_card() error
immediately after reclaiming  it, so reclaim the host  only in case
of success.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: android-goldfish: Allow compiling the driver with COMPILE_TEST
Luis de Bethencourt [Sat, 10 Oct 2015 16:08:32 +0000 (17:08 +0100)]
mmc: android-goldfish: Allow compiling the driver with COMPILE_TEST

The driver depends on GOLDFISH but there isn't a build dependency
so it's a good idea to allow the driver to always be built when the
COMPILE_TEST option is enabled.

That way, the driver can be built with a config generated by make
allyesconfig and check if a patch would break the build.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Reviewed-by: Jean Delvare <jdelvare@suse.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoARM: dts: rockchip: add tuning related settings to veyron devices
Heiko Stuebner [Mon, 12 Oct 2015 12:48:30 +0000 (14:48 +0200)]
ARM: dts: rockchip: add tuning related settings to veyron devices

This allows the tuning code to run and use higher speeds on capable cards.

Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoARM: dts: rockchip: Add drive/sample clocks for rk3288 dw_mmc devices
Alexandru M Stan [Mon, 12 Oct 2015 12:48:29 +0000 (14:48 +0200)]
ARM: dts: rockchip: Add drive/sample clocks for rk3288 dw_mmc devices

The drive/sample clocks can be phase shifted.  The drive clock
could be used in a future patch to adjust hold times.  The sample
clock is used for tuning.

Signed-off-by: Alexandru M Stan <amstan@chromium.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc-rockchip: MMC tuning with the clock phase framework
Alexandru M Stan [Mon, 12 Oct 2015 12:48:28 +0000 (14:48 +0200)]
mmc: dw_mmc-rockchip: MMC tuning with the clock phase framework

This algorithm will try 1 degree increments, since there's no way to tell
what resolution the underlying phase code uses. As an added bonus, doing
many tunings yields better results since some tests are run more than once
(ex: if the underlying driver uses 45 degree increments, the tuning code
will try the same angle more than once).

It will then construct a list of good phase ranges (even ranges that cross
360/0), will pick the biggest range then it will set the sample_clk to the
middle of that range.

We do not touch ciu_drive (and by extension define default-drive-phase).
Drive phase is mostly used to define minimum hold times, while one could
write some code to determine what phase meets the minimum hold time (ex 10
degrees) this will not work with the current clock phase framework (which
floors angles, so we'll get 0 deg, and there's no way to know what
resolution the floors happen at). We assume that the default drive angles
set by the hardware are good enough.

If a device has device specific code (like exynos) then that will still
take precedence, otherwise this new code will execute. If the device wants
to tune, but has no sample_clk defined we'll return EIO with an error
message.

Signed-off-by: Alexandru M Stan <amstan@chromium.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc-rockchip: dt-binding: Add tuning related things
Alexandru M Stan [Mon, 12 Oct 2015 12:48:27 +0000 (14:48 +0200)]
mmc: dw_mmc-rockchip: dt-binding: Add tuning related things

Add ciu_drive, ciu_sample clocks and default-sample-phase. This will later
be used by tuning code.

We do not touch ciu_drive (and by extension define default-drive-phase).
Drive phase is mostly used to define minimum hold times, while one could
write some code to determine what phase meets the minimum hold time
(ex 10 degrees) this will not work with the current clock phase framework
(which floors angles, so we'll get 0 deg, and there's no way to know what
resolution the floors happen at). We assume that the default drive angles
set by the hardware are good enough.

Signed-off-by: Alexandru M Stan <amstan@chromium.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Use mmc_regulator_set_vqmmc in start_signal_voltage_switch
Douglas Anderson [Mon, 12 Oct 2015 12:48:26 +0000 (14:48 +0200)]
mmc: dw_mmc: Use mmc_regulator_set_vqmmc in start_signal_voltage_switch

We've introduced a new helper in the MMC core:
mmc_regulator_set_vqmmc().  Let's use this in dw_mmc.  Using this new
helper has some advantages:

1. We get the mmc_regulator_set_vqmmc() behavior of trying to match
   VQMMC and VMMC when the signal voltage is 3.3V.  This ensures max
   compatibility.

2. We get rid of a few more warnings when probing unsupported
   voltages.

3. We get rid of some non-dw_mmc specific code in dw_mmc.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Add mmc_regulator_set_vqmmc()
Douglas Anderson [Mon, 12 Oct 2015 12:48:25 +0000 (14:48 +0200)]
mmc: core: Add mmc_regulator_set_vqmmc()

This adds logic to the MMC core to set VQMMC.  This is expected to be
called by MMC drivers like dw_mmc as part of (or instead of) their
start_signal_voltage_switch() callback.

A few notes:

* When setting the signal voltage to 3.3V we do our best to make VQMMC
  and VMMC match.  It's been reported that this makes some old cards
  happy since they were tested back in the day before UHS when VQMMC
  and VMMC were provided by the same regulator.  A nice side effect of
  this is that we don't end up on the hairy edge of VQMMC (2.7V),
  which some EEs claim is a little too close to the minimum for
  comfort.
  This is done in two steps. At first we try to find a VQMMC within
  a 0.3V tolerance of VMMC and if this is not supported by the
  supplying regulator we try to find a suitable voltage within the
  whole 2.7V-3.6V area of the spec.

* The two step approach is currently necessary, as the used
  regulator_set_voltage_triplet(min, target, max) uses a simple
  implementation that just tries two basic steps:
regulator_set_voltage(target, max);
regulator_set_voltage(min, target);
  So with only one step with 2.7-3.6V borders, if a suitable voltage
  is a bit below VMMC, we would directly get the lowest 2.7V
  which some boards (like Rockchips) don't like at all.

* When setting the signal voltage to 1.8V or 1.2V we aim for that
  specific voltage instead of picking the lowest one in the range.

* We very purposely don't print errors in mmc_regulator_set_vqmmc().
  There are cases where the MMC core will try several different
  voltages and we don't want to pollute the logs.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: move ocr-bit to voltage translation into separate function
Heiko Stuebner [Mon, 12 Oct 2015 16:00:54 +0000 (18:00 +0200)]
mmc: core: move ocr-bit to voltage translation into separate function

We will shortly need the calculation of an ocr-bit to the actual
voltage in a second place too, so move it from mmc_regulator_set_ocr
to a common function mmc_ocrbitnum_to_vdd to make that possible.

Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Remove MMC_CLKGATE
Ulf Hansson [Fri, 2 Oct 2015 08:56:11 +0000 (10:56 +0200)]
mmc: core: Remove MMC_CLKGATE

MMC_CLKGATE was once invented to save power by gating the bus clock at
request inactivity. At that time it served its purpose. The modern way to
deal with power saving for these scenarios, is by using runtime PM.

Nowadays, several host drivers have deployed runtime PM, but for those
that haven't and which still cares power saving at request inactivity,
it's certainly time to deploy runtime PM as it has been around for several
years now.

To simplify code to mmc core and thus decrease maintenance efforts, this
patch removes all code related to MMC_CLKGATE.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
9 years agommc: sdhci-of-esdhc: add quirk SDHCI_QUIRK_BROKEN_TIMEOUT_VAL for ls1021a
yangbo lu [Thu, 8 Oct 2015 10:36:57 +0000 (18:36 +0800)]
mmc: sdhci-of-esdhc: add quirk SDHCI_QUIRK_BROKEN_TIMEOUT_VAL for ls1021a

Signed-off-by: Yangbo Lu <yangbo.lu@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: kconfig: reconfigure MMC_SDHCI_OF_ESDHC option
yangbo lu [Thu, 8 Oct 2015 10:36:47 +0000 (18:36 +0800)]
mmc: kconfig: reconfigure MMC_SDHCI_OF_ESDHC option

Freescale eSDHC driver now supports both little-endian and
big-endian mode eSDHC IPs for ARM and PPC. So, MMC_SDHCI_OF_ESDHC
option needs to be reconfigured.

Signed-off-by: Yangbo Lu <yangbo.lu@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-of-esdhc: support both BE and LE host controller
yangbo lu [Thu, 8 Oct 2015 10:36:36 +0000 (18:36 +0800)]
mmc: sdhci-of-esdhc: support both BE and LE host controller

To support little endian eSDHC controller, we redefine both BE and
LE IO accessors. In the new accessors, use ioread*/iowrite* instead
of in_be32/out_be32 and will select accessors according to endian
mode in probe function.

Signed-off-by: Yangbo Lu <yangbo.lu@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dt: add DT binding for little-endian eSDHC host controller
yangbo lu [Thu, 8 Oct 2015 10:36:23 +0000 (18:36 +0800)]
mmc: dt: add DT binding for little-endian eSDHC host controller

Signed-off-by: Yangbo Lu <yangbo.lu@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: Add another PCI ID for an Intel eMMC host controller
Adrian Hunter [Tue, 6 Oct 2015 07:26:21 +0000 (10:26 +0300)]
mmc: sdhci-pci: Add another PCI ID for an Intel eMMC host controller

Add another PCI ID for an Intel eMMC host controller.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoclk: rockchip: Make calculations use rounding
Douglas Anderson [Wed, 30 Sep 2015 14:07:38 +0000 (16:07 +0200)]
clk: rockchip: Make calculations use rounding

Let's use DIV_ROUND_CLOSEST for rounding, not just truncating
division.  This lets us get closer to the right rate.

Before this:
  set_phase(86) delay_nums=26 reg[0xf000420c]=0x468 actual_degrees=83
  set_phase(89) delay_nums=27 reg[0xf000420c]=0x46c actual_degrees=86

After this:
  set_phase(86) delay_nums=27 reg[0xf000420c]=0x46c actual_degrees=86
  set_phase(89) delay_nums=28 reg[0xf000420c]=0x470 actual_degrees=90

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoclk: rockchip: Allow more precision for some mmc clock phases
Douglas Anderson [Wed, 30 Sep 2015 14:07:37 +0000 (16:07 +0200)]
clk: rockchip: Allow more precision for some mmc clock phases

Because of the inexact nature of the extra MMC delay elements (it's
not possible to keep the phase monotonic and to also make phases (mod
90) > 70), we previously only allowed phases (mod 90) of 22.5, 45,
and 67.5.

But it's not the end of the world if the MMC clock phase goes
non-monotonic.  At most we'll be 25 degrees off.  It's way better to
test more phases to look for bad ones than to be 25 degrees off, because
in the case of MMC really the point is to find bad phases and get as far
asway from the as possible.  If we get to test extra phases by going
slightly non-monotonic then that might be fine.  Worst case we would
end up at a phases that's slight differnt than the one we wanted, but
at least we'd still be quite far away from the a bad phase.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Fold in more precise variance-values of 44-77 instead of 40-80.
Fold in the actual removal of the monotonic requirement and adapt
patch message accordingly.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: fix simple_return.cocci warnings
kbuild test robot [Mon, 5 Oct 2015 20:01:04 +0000 (04:01 +0800)]
mmc: sdhci-pci: fix simple_return.cocci warnings

drivers/mmc/host/sdhci-pci-core.c:447:1-4: WARNING: end returns can be simpified

 Simplify a trivial if-return sequence.  Possibly combine with a
 preceding function call.

Generated by: scripts/coccinelle/misc/simple_return.cocci

CC: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: Make sdhci_pci_o2_fujin2_pci_init() static
Ben Hutchings [Mon, 5 Oct 2015 19:31:58 +0000 (20:31 +0100)]
mmc: sdhci-pci: Make sdhci_pci_o2_fujin2_pci_init() static

It has no external callers.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: Build o2micro support in the same module
Ben Hutchings [Mon, 5 Oct 2015 19:31:18 +0000 (20:31 +0100)]
mmc: sdhci-pci: Build o2micro support in the same module

sdhci-pci-o2micro.c contains no initialisation and its functions are only
called from shdci-pci.c, so there is no reason for it to be a separate
module, let alone or for it to always be built-in.

- Rename sdhci-pci.c to sdhci-pci-core.c so that the sdhci-pci module
  can be built from multiple source files
- Add sdhci-pci-o2micro.c to it
- Remove redundant exports

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sirf: fix the capbility to support DDR50
Weijun Yang [Sun, 4 Oct 2015 12:04:13 +0000 (12:04 +0000)]
mmc: sirf: fix the capbility to support DDR50

According to hardware spec, validate DDR50 mode
for SDXC.

Signed-off-by: Weijun Yang <york.yang@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: enable tuning for DDR50
Weijun Yang [Sun, 4 Oct 2015 12:04:12 +0000 (12:04 +0000)]
mmc: sdhci: enable tuning for DDR50

CMD19 tuning is also available for DDR50 mode.

Signed-off-by: Weijun Yang <york.yang@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: enable CMD19 tuning for DDR50 mode
Weijun Yang [Sun, 4 Oct 2015 12:04:11 +0000 (12:04 +0000)]
mmc: core: enable CMD19 tuning for DDR50 mode

As SD Specifications Part1 Physical Layer Specification Version
3.01 says, CMD19 tuning is available for unlocked cards in transfer
state of 1.8V signaling mode. The small difference between v3.00
and 3.01 spec means that CMD19 tuning is also available for DDR50
mode.

Signed-off-by: Weijun Yang <york.yang@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: block: copy resp[] data on err for MMC_IOC_MULTI_CMD
Grant Grundler [Thu, 24 Sep 2015 01:30:33 +0000 (18:30 -0700)]
mmc: block: copy resp[] data on err for MMC_IOC_MULTI_CMD

MMC_IOC_CMD and MMC_IOC_MULTI_CMD ioctl() code currently bails on
any eMMC errors. However, in case there is any resp[] data, we
should attempt to copy resp[] back to user space. The user app
can then determine which command(s) failed in the MMC_IOC_MULTI_CMD
case AND/OR report better diagnostics in both cases.

Gwendal Grignou provided the idea and it was previously implemented
and tested on v3.18 ChromeOS kernel:
    https://chromium-review.googlesource.com/#/c/299956

Signed-off-by: Grant Grundler <grundler@chromium.org>
Reviewed-by: Hyung Taek Ryoo <hryoo@nvidia.com>
Reviewed-by: Gwendal Grignou <gwendal@chromium.org>
Tested-by: David Riley <davidriley@chromium.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: debugfs: implement ios show for SDR12 and SDR25
Shawn Lin [Thu, 24 Sep 2015 00:13:43 +0000 (08:13 +0800)]
mmc: debugfs: implement ios show for SDR12 and SDR25

This patch add MMC_TIMING_UHS_SDR12 and MMC_TIMING_UHS_SDR25
for mmc_ios_show to show the ios->timing if mmc card runs under
these two modes.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sunxi: Add card busy detection
Hans de Goede [Tue, 22 Sep 2015 15:30:26 +0000 (17:30 +0200)]
mmc: sunxi: Add card busy detection

Some sdio wifi modules have not been working reliable with the sunxi-mmc
host code. This turns out to be caused by starting new io-rw commands while
the card signals that it is still busy processing a previous command.

This commit adds card-busy detection to the sunxi-mmc driver which together
with recent core changes to check card-busy before starting io-rw commands
fixes the wifi reliability issues on the Cubietruck and other sunxi boards
using sdio wifi.

Reported-by: Eugene K <sigintmailru@gmail.com>
Suggested-by: Eugene K <sigintmailru@gmail.com>
Cc: Eugene K <sigintmailru@gmail.com>
Cc: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: Wait for card_busy before starting sdio requests
Hans de Goede [Tue, 22 Sep 2015 15:30:25 +0000 (17:30 +0200)]
mmc: Wait for card_busy before starting sdio requests

Some sdio wifi chips will not work properly if we try to start new
sdio-rw requests while the device is signalling that it is busy.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: Add mmc_is_io_op helper function
Hans de Goede [Tue, 22 Sep 2015 15:30:24 +0000 (17:30 +0200)]
mmc: Add mmc_is_io_op helper function

Add a helper function to check if an opcode is a sd-io-rw-* opcode.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: block: Add new ioctl to send multi commands
Jon Hunter [Tue, 22 Sep 2015 09:27:53 +0000 (10:27 +0100)]
mmc: block: Add new ioctl to send multi commands

Certain eMMC devices allow vendor specific device information to be read
via a sequence of vendor commands. These vendor commands must be issued
in sequence and an atomic fashion. One way to support this would be to
add an ioctl function for sending a sequence of commands to the device
atomically as proposed here. These multi commands are simple array of
the existing mmc_ioc_cmd structure.

The structure passed via the ioctl uses a __u64 type to specify the number
of commands (so that the structure is aligned on a 64-bit boundary) and a
zero length array as a header for list of commands to be issued. The
maximum number of commands that can be sent is determined by
MMC_IOC_MAX_CMDS (which defaults to 255 and should be more than
sufficient).

This based upon work by Seshagiri Holi <sholi@nvidia.com>.

Signed-off-by: Seshagiri Holi <sholi@nvidia.com>
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pltfm: enable interrupt mode to detect card for ls1021a
Yangbo Lu [Wed, 16 Sep 2015 06:36:10 +0000 (14:36 +0800)]
mmc: sdhci-pltfm: enable interrupt mode to detect card for ls1021a

Enable interrupt mode to detect card instead of polling mode for
ls1021a by removing the quirk SDHCI_QUIRK_BROKEN_CARD_DETECTION.
This could improve data transferring performance and avoid the call
trace caused by polling card status sometime.

Signed-off-by: Yangbo Lu <yangbo.lu@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: pwrseq_simple: use GPIO descriptors array API
Javier Martinez Canillas [Mon, 21 Sep 2015 12:14:54 +0000 (14:14 +0200)]
mmc: pwrseq_simple: use GPIO descriptors array API

The simple power sequence provider sets a value for multiple GPIOs in one
go so it is better to use the API already provided by the GPIO descriptor
API instead of open coding the same logic.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mediatek: Add MMC_CAP_RUNTIME_RESUME support
Chaotian Jing [Tue, 22 Sep 2015 06:00:44 +0000 (14:00 +0800)]
mmc: mediatek: Add MMC_CAP_RUNTIME_RESUME support

Add MMC_CAP_RUNTIME_RESUME support to save resume time
Drop unnecessary SDC_ARG write

Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: debugfs: implement ios show for driver type
Shawn Lin [Tue, 22 Sep 2015 01:20:08 +0000 (09:20 +0800)]
mmc: debugfs: implement ios show for driver type

This patch add ios->drv_type for mmc_ios_show to show the
card's driver type.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap: Fix module autoload for OF platform driver
Luis de Bethencourt [Thu, 17 Sep 2015 21:50:25 +0000 (23:50 +0200)]
mmc: omap: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: moxart: Fix module autoload for OF platform driver
Luis de Bethencourt [Thu, 17 Sep 2015 21:49:59 +0000 (23:49 +0200)]
mmc: moxart: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pltfm: Use of_property_read_u32 instead of open-coding it
Tobias Klauser [Thu, 17 Sep 2015 13:09:54 +0000 (15:09 +0200)]
mmc: sdhci-pltfm: Use of_property_read_u32 instead of open-coding it

Use of_property_read_u32 instead of of_get_property with return value
checks and endianness conversion.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mmc_spi: Export OF module alias information
Javier Martinez Canillas [Wed, 16 Sep 2015 09:12:07 +0000 (11:12 +0200)]
mmc: mmc_spi: Export OF module alias information

Drivers needs to export the OF id table and this be built into
the module or udev won't have the necessary information to autoload
the driver module when the device is registered via OF.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-of-at91: remove a line of dead code
Dan Carpenter [Wed, 16 Sep 2015 06:19:49 +0000 (09:19 +0300)]
mmc: sdhci-of-at91: remove a line of dead code

The goto is correct and the unreachable "return -EINVAL" should be
removed.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: call sdhci_init() before request irq
Haibo Chen [Tue, 15 Sep 2015 10:32:58 +0000 (18:32 +0800)]
mmc: sdhci: call sdhci_init() before request irq

sdhci_init() will clear all irqs and set the needed irqs. So
logically sdhci_init() should be called before request irq.

If not, some irqs may be triggled and handled wrongly. Take
the following into consideration, after request irq, if
SDIO card interrupt enabled, a sd card in the sd slot will
trigger a mass of interrupt(SDHCI_INT_CARD_INT), because at
this time, the vmmc-regulator still not restore, no voltage
supply for the sd card, so the pin of data0~data3 change and
keep low, interrupt(SDHCI_INT_CARD_INT) will rise up ceaselessly.
Due to we already reguest irq, system will be busy in handling
this endless irq, can't response to other event.

So we should call sdhci_init() before request irq in sd resume.

Signed-off-by: Haibo Chen <haibo.chen@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-bcm-kona: fix logic to check for 8-bit data width
Javier Martinez Canillas [Wed, 16 Sep 2015 09:53:20 +0000 (11:53 +0200)]
mmc: sdhci-bcm-kona: fix logic to check for 8-bit data width

The driver prints if the data width is 8-bit but it's using a
binary OR instead of a binary AND so it will always report as
"is_8bit=Y" regardless of the flags in host->mmc->caps.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Ray Jui <rjui@broadcom.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: pwrseq: use gpiod_get() instead of index 0
Javier Martinez Canillas [Mon, 14 Sep 2015 12:00:35 +0000 (14:00 +0200)]
mmc: pwrseq: use gpiod_get() instead of index 0

The gpiod_get() function expands to gpiod_get_index() with index 0
so it's better to use it since is easier to read and more concise.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Convert __mmc_switch() into an internal core function
Ulf Hansson [Tue, 15 Sep 2015 13:19:45 +0000 (15:19 +0200)]
mmc: core: Convert __mmc_switch() into an internal core function

As there are no users of the __mmc_switch() API, except for the mmc core
itself, let's convert it from an exported function into an internal.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: vub300: Remove unneded semicolons
Javier Martinez Canillas [Wed, 16 Sep 2015 09:30:04 +0000 (11:30 +0200)]
mmc: vub300: Remove unneded semicolons

They aren't needed and are just creating null statements so remove it.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: wbsd: Remove unneded semicolon
Javier Martinez Canillas [Wed, 16 Sep 2015 09:30:03 +0000 (11:30 +0200)]
mmc: wbsd: Remove unneded semicolon

It's not needed an is just creating a null statement, so remove it.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sd: Remove superfluous error code assignment
Yousong Zhou [Wed, 9 Sep 2015 16:33:16 +0000 (00:33 +0800)]
mmc: sd: Remove superfluous error code assignment

Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Keep host claimed while invoking mmc_power_off|up()
Ulf Hansson [Fri, 11 Sep 2015 12:41:55 +0000 (14:41 +0200)]
mmc: core: Keep host claimed while invoking mmc_power_off|up()

As mmc_claim_host() invokes pm_runtime_get_sync() for the mmc host device,
it's important that the host is kept claimed for *all* accesses to it via
the host_ops callbacks.

In some code paths for SDIO, particularly related to the PM support,
mmc_power_off|up() is invoked without keeping the host claimed. Let's fix
these.

Moreover, mmc_start|stop_host() also invokes mmc_power_off|up() without
claiming the host, let's fix these as well.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
9 years agommc: sdhci-acpi: detect sd card reader on asus x205ta
Michele Curti [Sat, 5 Sep 2015 06:49:52 +0000 (08:49 +0200)]
mmc: sdhci-acpi: detect sd card reader on asus x205ta

Add an entry to the sdhci_acpi_uids list to detect the SD card
reader on the Asus X205Ta laptop.

dstd table:

Device (SDHC)
{
    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "PNP0FFF")  // _HID: Hardware ID
    Name (_CID, "PNP0D40" /* SDA Standard Compliant SD Host Controller */)
    Name (_DDN, "Intel(R) SD Card Controller - 80860F16")  // _DDN: DOS Dev
    Name (_UID, 0x03)  // _UID: Unique ID
    Name (RDEP, Package (0x02)

Signed-off-by: Michele Curti <michele.curti@gmail.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoLinux 4.3-rc7
Linus Torvalds [Sun, 25 Oct 2015 01:39:47 +0000 (10:39 +0900)]
Linux 4.3-rc7

9 years agoMerge tag 'usb-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Fri, 23 Oct 2015 22:52:59 +0000 (07:52 +0900)]
Merge tag 'usb-4.3-rc7' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are three xhci driver fixes for reported issues for 4.3-rc7

  All have been in linux-next for a while with no problems"

* tag 'usb-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  xhci: Add spurious wakeup quirk for LynxPoint-LP controllers
  xhci: handle no ping response error properly
  xhci: don't finish a TD if we get a short transfer event mid TD

9 years agoMerge tag 'tty-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Fri, 23 Oct 2015 22:52:09 +0000 (07:52 +0900)]
Merge tag 'tty-4.3-rc7' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial fixes from Greg KH:
 "Here are two fixes that resolve reported issues, one with the 8250
  driver, and the other with the generic fbcon driver.

  Both have been in linux-next for a while"

* tag 'tty-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  fbcon: initialize blink interval before calling fb_set_par
  Revert "serial: 8250_dma: don't bother DMA with small transfers"

9 years agoMerge tag 'staging-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Fri, 23 Oct 2015 22:51:13 +0000 (07:51 +0900)]
Merge tag 'staging-4.3-rc7' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fixes from Greg KH:
 "Here are four iio driver fixes for 4.3-rc7, fixing some reported
  issues.  All of these have been in linux-next for a while"

* tag 'staging-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  iio: mxs-lradc: Fix temperature offset
  iio: accel: sca3000: memory corruption in sca3000_read_first_n_hw_rb()
  iio: st_accel: fix interrupt handling on LIS3LV02
  iio: adc: twl4030: Fix ADC[3:6] readings

9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Linus Torvalds [Fri, 23 Oct 2015 22:28:05 +0000 (07:28 +0900)]
Merge tag 'for-linus' of git://git./linux/kernel/git/dledford/rdma

Pull infiniband fixes from Doug Ledford:
 "It's late in the game, I know, but these fixes seemed important enough
  to warrant a late pull request.  They all involve oopses or use after
  frees or corruptions.

  Six serious fixes:

   - Hold the mutex around the find and corresponding update of our gid

   - The ifa list is rcu protected, copy its contents under rcu to avoid
     using a freed structure

   - On error, netdev might be null, so check it before trying to
     release it

   - On init, if workqueue alloc fails, fail init

   - The new demux patches exposed a bug in mlx5 and ipath drivers, we
     need to use the payload P_Key to determine the P_Key the packet
     arrived on because the hardware doesn't tell us the truth

   - Due to a couple convoluted error flows, it is possible for the CM
     to trigger a use_after_free and a double_free of rb nodes.  Add two
     checks to prevent that.  This code has worked for 10+ years.  It is
     likely that some of the recent changes have caused this issue to
     surface.  The current patch will protect us from nasty events for
     now while we track down why this is just now showing up"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  IB/cm: Fix rb-tree duplicate free and use-after-free
  IB/cma: Use inner P_Key to determine netdev
  IB/ucma: check workqueue allocation before usage
  IB/cma: Potential NULL dereference in cma_id_from_event
  IB/core: Fix use after free of ifa
  IB/core: Fix memory corruption in ib_cache_gid_set_default_gid

9 years agoMerge tag 'dm-4.3-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Fri, 23 Oct 2015 22:23:52 +0000 (07:23 +0900)]
Merge tag 'dm-4.3-fixes-4' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:
 "Three stable fixes (two in btree code used by DM thinp and one to
  properly store flags in DM cache metadata's superblock)"

* tag 'dm-4.3-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm cache: the CLEAN_SHUTDOWN flag was not being set
  dm btree: fix leak of bufio-backed block in btree_split_beneath error path
  dm btree remove: fix a bug when rebalancing nodes after removal

9 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 23 Oct 2015 22:20:57 +0000 (07:20 +0900)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block layer fixes from Jens Axboe:
 "A final set of fixes for 4.3.

  It is (again) bigger than I would have liked, but it's all been
  through the testing mill and has been carefully reviewed by multiple
  parties.  Each fix is either a regression fix for this cycle, or is
  marked stable.  You can scold me at KS.  The pull request contains:

   - Three simple fixes for NVMe, fixing regressions since 4.3.  From
     Arnd, Christoph, and Keith.

   - A single xen-blkfront fix from Cathy, fixing a NULL dereference if
     an error is returned through the staste change callback.

   - Fixup for some bad/sloppy code in nbd that got introduced earlier
     in this cycle.  From Markus Pargmann.

   - A blk-mq tagset use-after-free fix from Junichi.

   - A backing device lifetime fix from Tejun, fixing a crash.

   - And finally, a set of regression/stable fixes for cgroup writeback
     from Tejun"

* 'for-linus' of git://git.kernel.dk/linux-block:
  writeback: remove broken rbtree_postorder_for_each_entry_safe() usage in cgwb_bdi_destroy()
  NVMe: Fix memory leak on retried commands
  block: don't release bdi while request_queue has live references
  nvme: use an integer value to Linux errno values
  blk-mq: fix use-after-free in blk_mq_free_tag_set()
  nvme: fix 32-bit build warning
  writeback: fix incorrect calculation of available memory for memcg domains
  writeback: memcg dirty_throttle_control should be initialized with wb->memcg_completions
  writeback: bdi_writeback iteration must not skip dying ones
  writeback: fix bdi_writeback iteration in wakeup_dirtytime_writeback()
  writeback: laptop_mode_timer_fn() needs rcu_read_lock() around bdi_writeback iteration
  nbd: Add locking for tasks
  xen-blkfront: check for null drvdata in blkback_changed (XenbusStateClosing)

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Fri, 23 Oct 2015 22:19:33 +0000 (07:19 +0900)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

Pull Ceph fixes from Sage Weil:
 "Two fixes.

  One is a stopgap to prevent a stack blowout when users have a deep
  chain of image clones.  (We'll rewrite this code to be non-recursive
  for the next window, but in the meantime this is a simple fix that
  avoids a crash.)

  The second fixes a refcount underflow"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: prevent kernel stack blow up on rbd map
  rbd: don't leak parent_spec in rbd_dev_probe_parent()

9 years agoMerge branch 'for-linus-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/mason...
Linus Torvalds [Fri, 23 Oct 2015 22:17:58 +0000 (07:17 +0900)]
Merge branch 'for-linus-4.3' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "I have two more small fixes this week:

  Qu's fix avoids unneeded COW during fallocate, and Christian found a
  memory leak in the error handling of an earlier fix"

* 'for-linus-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: fix possible leak in btrfs_ioctl_balance()
  btrfs: Avoid truncate tailing page if fallocate range doesn't exceed inode size

9 years agodm cache: the CLEAN_SHUTDOWN flag was not being set
Joe Thornber [Thu, 22 Oct 2015 17:10:55 +0000 (18:10 +0100)]
dm cache: the CLEAN_SHUTDOWN flag was not being set

If the CLEAN_SHUTDOWN flag is not set when a cache is loaded then all cache
blocks are marked as dirty and a full writeback occurs.

__commit_transaction() is responsible for setting/clearing
CLEAN_SHUTDOWN (based the flags_mutator that is passed in).

Fix this issue, of the cache's on-disk flags being wrong, by making sure
__commit_transaction() does not reset the flags after the mutator has
altered the flags in preparation for them being serialized to disk.

before:

sb_flags = mutator(le32_to_cpu(disk_super->flags));
disk_super->flags = cpu_to_le32(sb_flags);
disk_super->flags = cpu_to_le32(cmd->flags);

after:

disk_super->flags = cpu_to_le32(cmd->flags);
sb_flags = mutator(le32_to_cpu(disk_super->flags));
disk_super->flags = cpu_to_le32(sb_flags);

Reported-by: Bogdan Vasiliev <bogdan.vasiliev@gmail.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
9 years agodm btree: fix leak of bufio-backed block in btree_split_beneath error path
Mike Snitzer [Thu, 22 Oct 2015 14:56:40 +0000 (10:56 -0400)]
dm btree: fix leak of bufio-backed block in btree_split_beneath error path

btree_split_beneath()'s error path had an outstanding FIXME that speaks
directly to the potential for _not_ cleaning up a previously allocated
bufio-backed block.

Fix this by releasing the previously allocated bufio block using
unlock_block().

Reported-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <thornber@redhat.com>
Cc: stable@vger.kernel.org
9 years agodm btree remove: fix a bug when rebalancing nodes after removal
Joe Thornber [Wed, 21 Oct 2015 17:36:49 +0000 (18:36 +0100)]
dm btree remove: fix a bug when rebalancing nodes after removal

Commit 4c7e309340ff ("dm btree remove: fix bug in redistribute3") wasn't
a complete fix for redistribute3().

The redistribute3 function takes 3 btree nodes and shares out the entries
evenly between them.  If the three nodes in total contained
(MAX_ENTRIES * 3) - 1 entries between them then this was erroneously getting
rebalanced as (MAX_ENTRIES - 1) on the left and right, and (MAX_ENTRIES + 1) in
the center.

Fix this issue by being more careful about calculating the target number
of entries for the left and right nodes.

Unit tested in userspace using this program:
https://github.com/jthornber/redistribute3-test/blob/master/redistribute3_t.c

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
9 years agorbd: prevent kernel stack blow up on rbd map
Ilya Dryomov [Sun, 11 Oct 2015 17:38:00 +0000 (19:38 +0200)]
rbd: prevent kernel stack blow up on rbd map

Mapping an image with a long parent chain (e.g. image foo, whose parent
is bar, whose parent is baz, etc) currently leads to a kernel stack
overflow, due to the following recursion in the reply path:

  rbd_osd_req_callback()
    rbd_obj_request_complete()
      rbd_img_obj_callback()
        rbd_img_parent_read_callback()
          rbd_obj_request_complete()
            ...

Limit the parent chain to 16 images, which is ~5K worth of stack.  When
the above recursion is eliminated, this limit can be lifted.

Fixes: http://tracker.ceph.com/issues/12538

Cc: stable@vger.kernel.org # 3.10+, needs backporting for < 4.2
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
9 years agorbd: don't leak parent_spec in rbd_dev_probe_parent()
Ilya Dryomov [Sun, 11 Oct 2015 17:38:00 +0000 (19:38 +0200)]
rbd: don't leak parent_spec in rbd_dev_probe_parent()

Currently we leak parent_spec and trigger a "parent reference
underflow" warning if rbd_dev_create() in rbd_dev_probe_parent() fails.
The problem is we take the !parent out_err branch and that only drops
refcounts; parent_spec that would've been freed had we called
rbd_dev_unparent() remains and triggers rbd_warn() in
rbd_dev_parent_put() - at that point we have parent_spec != NULL and
parent_ref == 0, so counter ends up being -1 after the decrement.

Redo rbd_dev_probe_parent() to fix this.

Cc: stable@vger.kernel.org # 3.10+, needs backporting for < 4.2
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
9 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 23 Oct 2015 13:34:32 +0000 (22:34 +0900)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "Misc fixes: two KASAN fixes, two EFI boot fixes, two boot-delay
  optimization fixes, and a fix for a IRQ handling hang observed on
  virtual platforms"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mm, kasan: Silence KASAN warnings in get_wchan()
  compiler, atomics, kasan: Provide READ_ONCE_NOCHECK()
  x86, kasan: Fix build failure on KASAN=y && KMEMCHECK=y kernels
  x86/smpboot: Fix CPU #1 boot timeout
  x86/smpboot: Fix cpu_init_udelay=10000 corner case boot parameter misbehavior
  x86/ioapic: Disable interrupts when re-routing legacy IRQs
  x86/setup: Extend low identity map to cover whole kernel range
  x86/efi: Fix multiple GOP device support

9 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 23 Oct 2015 13:31:39 +0000 (22:31 +0900)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Misc fixes all around the map: an instrumentation fix, a nohz
  usability fix, a lockdep annotation fix and two task group scheduling
  fixes"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/core: Add missing lockdep_unpin() annotations
  sched/deadline: Fix migration of SCHED_DEADLINE tasks
  nohz: Revert "nohz: Set isolcpus when nohz_full is set"
  sched/fair: Update task group's load_avg after task migration
  sched/fair: Fix overly small weight for interactive group entities
  sched, tracing: Stop/start critical timings around the idle=poll idle loop

9 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 23 Oct 2015 13:10:51 +0000 (22:10 +0900)]
Merge branch 'akpm' (patches from Andrew)

Merge fixes from Andrew Morton:
 "9 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  ocfs2/dlm: unlock lockres spinlock before dlm_lockres_put
  fault-inject: fix inverted interval/probability values in printk
  lib/Kconfig.debug: disable -Wframe-larger-than warnings with KASAN=y
  mm: make sendfile(2) killable
  thp: use is_zero_pfn() only after pte_present() check
  mailmap: update Javier Martinez Canillas' email
  MAINTAINERS: add Sergey as zsmalloc reviewer
  mm: cma: fix incorrect type conversion for size during dma allocation
  kmod: don't run async usermode helper as a child of kworker thread

9 years agosched/core: Add missing lockdep_unpin() annotations
Peter Zijlstra [Fri, 23 Oct 2015 09:50:08 +0000 (11:50 +0200)]
sched/core: Add missing lockdep_unpin() annotations

Luca and Wanpeng reported two missing annotations that led to
false lockdep complaints. Add the missing annotations.

Reported-by: Luca Abeni <luca.abeni@unitn.it>
Reported-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: cbce1a686700 ("sched,lockdep: Employ lock pinning")
Link: http://lkml.kernel.org/r/20151023095008.GY17308@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'powerpc-4.3-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 23 Oct 2015 09:49:51 +0000 (18:49 +0900)]
Merge tag 'powerpc-4.3-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Revert "Use the POWER8 Micro Partition Prefetch Engine in KVM HV on
   POWER8" from Paul
 - Handle irq_happened flag correctly in off-line loop from Paul
 - Validate rtas.entry before calling enter_rtas() from Vasant

* tag 'powerpc-4.3-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/rtas: Validate rtas.entry before calling enter_rtas()
  powerpc/powernv: Handle irq_happened flag correctly in off-line loop
  powerpc: Revert "Use the POWER8 Micro Partition Prefetch Engine in KVM HV on POWER8"

9 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Fri, 23 Oct 2015 09:38:00 +0000 (18:38 +0900)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Arnd Bergmann:
 "Most of the changes this time are for incorrect device nodes in
  various ways, on on imx, berlin, exynos, ux500, uniphier, omap and
  meson.

  Chen-Yu Tsai now co-maintains mach-sunxi (Allwinner).

  Other bug fixes include
   - a partial revert of a broken tegra gpio patch
   - irq affinity for arm ccn
   - suspend on one Armada 385 machine
   - enable ZONE_DMA to avoid an OMAP crash for over 2GB RAM
   - turning on a regulator on beagleboard-x15 for HDMI
   - making the omap gpmc debug code visible
   - setup of orion network switch
   - a rare build regression for pxa"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (22 commits)
  ARM: OMAP2+: Fix imprecise external abort caused by bogus SRAM init
  thermal: exynos: Fix register read in TMU
  ARM: OMAP2+: Fix oops with LPAE and more than 2GB of memory
  ARM: tegra: Comment out gpio-ranges properties
  ARM: dts: uniphier: fix IRQ number for devices on PH1-LD6b ref board
  drivers/perf: arm_pmu: avoid CPU device_node reference leak
  bus: arm-ccn: Fix irq affinity setting on CPU migration
  bus: arm-ccn: Handle correctly no-more-cpus case
  ARM: mvebu: correct a385-db-ap compatible string
  ARM: meson6: DTS: Fix wrong reg mapping and IRQ numbers
  MAINTAINERS: Update Allwinner entry and add new maintainer
  ARM: ux500: modify initial levelshifter status
  ARM: pxa: fix pxa3xx DFI lockup hack
  Documentation: ARM: List new omap MMC requirements
  memory: omap-gpmc: dump "before" state before first modification
  memory: omap-gpmc: Fix unselectable debug option for GPMC
  ARM: dts: am57xx-beagle-x15: set VDD_SD to always-on
  ARM: dts: Fix audio card detection on Peach boards
  ARM: EXYNOS: Fix double of_node_put() when parsing child power domains
  ARM: orion: Fix DSA platform device after mvmdio conversion
  ...

9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 23 Oct 2015 09:32:29 +0000 (18:32 +0900)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM bugfixes from Paolo Bonzini:
 "Bug fixes for ARM, mostly 4.3 regressions related to virtual interrupt
  controller changes"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  arm/arm64: KVM: Fix disabled distributor operation
  arm/arm64: KVM: Clear map->active on pend/active clear
  arm/arm64: KVM: Fix arch timer behavior for disabled interrupts
  KVM: arm: use GIC support unconditionally
  KVM: arm/arm64: Fix memory leak if timer initialization fails
  KVM: arm/arm64: Do not inject spurious interrupts

9 years agoMerge tag 'trace-fixes-v4.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 23 Oct 2015 09:24:33 +0000 (18:24 +0900)]
Merge tag 'trace-fixes-v4.3-rc6' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "Running tests on other changes, the system locked up due to lots of
  warnings.  It was caused by the stack tracer triggering a warning
  about using rcu_dereference() when RCU was not watching.  This can
  happen due to the fact that the stack tracer uses the function tracer
  to check each function, and there are functions that may be called and
  traced when RCU stopped watching.  Namely when a function is called
  just before going idle or to userspace and after RCU stopped watching
  that current CPU.

  The first patch makes sure that RCU is watching when the stack tracer
  uses RCU.  The second patch is to make sure that the stack tracer does
  not get called by functions in NMI, as it's not NMI safe"

* tag 'trace-fixes-v4.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Do not allow stack_tracer to record stack in NMI
  tracing: Have stack tracer force RCU to be watching

9 years agoMerge tag 'sound-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 23 Oct 2015 09:21:55 +0000 (18:21 +0900)]
Merge tag 'sound-4.3-rc7' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "There is nothing to worry you much, only a few small & stable patches
  are found for usual stuff, HD-audio (a Lenovo laptop quirk, a fix for
  minor error handling) and ASoC (trivial fixes for RT298 and WM
  codecs).

  The only remaining major change is the fix for ASoC SX_TLV control
  that was overseen during refactoring, but the fix itself is trivial
  and safe"

* tag 'sound-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ASoC: wm8962: mark cache_dirty flag after software reset in pm_resume
  ASoC: rt298: fix wrong setting of gpio2_en
  ASoC: wm8904: Correct number of EQ registers
  ALSA: hda - Fix deadlock at error in building PCM
  ASoC: Add info callback for SX_TLV controls
  ASoC: rt298: correct index default value
  ALSA: hda - Fix inverted internal mic on Lenovo G50-80
  ALSA: hdac: Explicitly add io.h

9 years agoMerge tag 'media/v4.3-4' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Fri, 23 Oct 2015 09:15:20 +0000 (18:15 +0900)]
Merge tag 'media/v4.3-4' of git://git./linux/kernel/git/mchehab/linux-media

Pull media fixes from Mauro Carvalho Chehab:
 "Some regression fixes and potential security issues:

   - netup_unidvb: fix potential crash when spi is NULL
   - rtl28xxu: fix control message flaws
   - m88ds3103: fix a regression on Kernel 4.2
   - c8sectpfe: fix some issues on this new driver
   - v4l2-flash-led-class: fix a Kbuild dependency
   - si2157 and si2158: check for array boundary when uploading firmware
     files
   - horus3a and lnbh25: fix some building troubles when some options
     aren't selected
   - ir-hix5hd2: drop the use of IRQF_NO_SUSPEND"

* tag 'media/v4.3-4' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] m88ds3103: use own reg update_bits() implementation
  [media] rtl28xxu: fix control message flaws
  [media] v4l2-flash-led-class: Add missing VIDEO_V4L2 Kconfig dependency
  [media] netup_unidvb: fix potential crash when spi is NULL
  [media] si2168: Bounds check firmware
  [media] si2157: Bounds check firmware
  [media] ir-hix5hd2: drop the use of IRQF_NO_SUSPEND
  [media] c8sectpfe: fix return of garbage
  [media] c8sectpfe: fix ininitialized error return on firmware load failure
  [media] lnbh25: Fix lnbh25_attach() function return type
  [media] horus3a: Fix horus3a_attach() function parameters

9 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 23 Oct 2015 09:08:22 +0000 (18:08 +0900)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "I've been a bit slow gathering these:

   - drm/mst: one mutex leak in a fail path

   - radeon: two oops fixes, one dpm fix

   - i915: one messy set of fixes, where we revert the original fix, and
           pull back the proper set of fixes from -next on top.

   - nouveau: one fix for an illegal buffer placement.

  Doesn't look too bad, hopefully shouldn't be too much more"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/nouveau/gem: return only valid domain when there's only one
  drm: fix mutex leak in drm_dp_get_mst_branch_device
  drm/amdgpu: add missing dpm check for KV dpm late init
  drm/amdgpu/dpm: don't add pwm attributes if DPM is disabled
  drm/radeon/dpm: don't add pwm attributes if DPM is disabled
  drm/i915: Add primary plane to mask if it's visible
  drm/i915: Move sprite/cursor plane disable to intel_sanitize_crtc()
  drm/i915: Assign hwmode after encoder state readout
  Revert "drm/i915: Add primary plane to mask if it's visible"
  drm/i915: Deny wrapping an userptr into a framebuffer
  drm/i915: Enable DPLL VGA mode before P1/P2 divider write
  drm/i915: Restore lost DPLL register write on gen2-4
  drm/i915: Flush pipecontrol post-sync writes
  drm/i915: Fix kerneldoc for i915_gem_shrink_all

9 years agoocfs2/dlm: unlock lockres spinlock before dlm_lockres_put
Joseph Qi [Thu, 22 Oct 2015 20:32:29 +0000 (13:32 -0700)]
ocfs2/dlm: unlock lockres spinlock before dlm_lockres_put

dlm_lockres_put will call dlm_lockres_release if it is the last
reference, and then it may call dlm_print_one_lock_resource and
take lockres spinlock.

So unlock lockres spinlock before dlm_lockres_put to avoid deadlock.

Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofault-inject: fix inverted interval/probability values in printk
Florian Westphal [Thu, 22 Oct 2015 20:32:27 +0000 (13:32 -0700)]
fault-inject: fix inverted interval/probability values in printk

interval displays the probability and vice versa.

Fixes: 6adc4a22f20bb ("fault-inject: add ratelimit option")
Acked-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agolib/Kconfig.debug: disable -Wframe-larger-than warnings with KASAN=y
Andrey Ryabinin [Thu, 22 Oct 2015 20:32:24 +0000 (13:32 -0700)]
lib/Kconfig.debug: disable -Wframe-larger-than warnings with KASAN=y

When the kernel compiled with KASAN=y, GCC adds redzones for each
variable on stack.  This enlarges function's stack frame and causes:

'warning: the frame size of X bytes is larger than Y bytes'

The worst case I've seen for now is following:

   ../net/wireless/nl80211.c: In function `nl80211_send_wiphy':
   ../net/wireless/nl80211.c:1731:1: warning: the frame size of 5448 bytes is larger than 2048 bytes [-Wframe-larger-than=]

That kind of warning becomes useless with KASAN=y.  It doesn't
necessarily indicate that there is some problem in the code, thus we
should turn it off.

(The KASAN=y stack size in increased from 16k to 32k for this reason)

Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Acked-by: Abylay Ospan <aospan@netup.ru>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Mauro Carvalho Chehab <m.chehab@samsung.com>
Cc: Kozlov Sergey <serjk@netup.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomm: make sendfile(2) killable
Jan Kara [Thu, 22 Oct 2015 20:32:21 +0000 (13:32 -0700)]
mm: make sendfile(2) killable

Currently a simple program below issues a sendfile(2) system call which
takes about 62 days to complete in my test KVM instance.

        int fd;
        off_t off = 0;

        fd = open("file", O_RDWR | O_TRUNC | O_SYNC | O_CREAT, 0644);
        ftruncate(fd, 2);
        lseek(fd, 0, SEEK_END);
        sendfile(fd, fd, &off, 0xfffffff);

Now you should not ask kernel to do a stupid stuff like copying 256MB in
2-byte chunks and call fsync(2) after each chunk but if you do, sysadmin
should have a way to stop you.

We actually do have a check for fatal_signal_pending() in
generic_perform_write() which triggers in this path however because we
always succeed in writing something before the check is done, we return
value > 0 from generic_perform_write() and thus the information about
signal gets lost.

Fix the problem by doing the signal check before writing anything.  That
way generic_perform_write() returns -EINTR, the error gets propagated up
and the sendfile loop terminates early.

Signed-off-by: Jan Kara <jack@suse.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agothp: use is_zero_pfn() only after pte_present() check
Minchan Kim [Thu, 22 Oct 2015 20:32:19 +0000 (13:32 -0700)]
thp: use is_zero_pfn() only after pte_present() check

Use is_zero_pfn() on pteval only after pte_present() check on pteval
(It might be better idea to introduce is_zero_pte() which checks
pte_present() first).

Otherwise when working on a swap or migration entry and if pte_pfn's
result is equal to zero_pfn by chance, we lose user's data in
__collapse_huge_page_copy().  So if you're unlucky, the application
segfaults and finally you could see below message on exit:

BUG: Bad rss-counter state mm:ffff88007f099300 idx:2 val:3

Fixes: ca0984caa823 ("mm: incorporate zero pages into transparent huge pages")
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org> [4.1+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomailmap: update Javier Martinez Canillas' email
Javier Martinez Canillas [Thu, 22 Oct 2015 20:32:16 +0000 (13:32 -0700)]
mailmap: update Javier Martinez Canillas' email

The get_maintainer script still reports my old Collabora email based on
old commits but that address no longer exist so update mailmap to report
my current email and avoid people sending to the old address.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoMAINTAINERS: add Sergey as zsmalloc reviewer
Sergey Senozhatsky [Thu, 22 Oct 2015 20:32:13 +0000 (13:32 -0700)]
MAINTAINERS: add Sergey as zsmalloc reviewer

Nominate myself as a zsmalloc reviewer.

Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>