firefly-linux-kernel-4.4.55.git
9 years agommc: sdhci-esdhc-imx: using specific compatible string in binding doc
Dong Aisheng [Wed, 27 May 2015 10:13:29 +0000 (18:13 +0800)]
mmc: sdhci-esdhc-imx: using specific compatible string in binding doc

Using specific compatible string in binding doc to make the binding
more clear.
It's also used to avoid checkpatch warning in the future like follows:
WARNING: DT compatible string "fsl,imx6sx-usdhc" appears un-documented --
check ./Documentation/devicetree/bindings/
+       { .compatible = "fsl,imx6sx-usdhc", .data = &usdhc_imx6sx_data, },

total: 0 errors, 1 warnings, 18 lines checked

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: add ADMA Length Mismatch errata fix
Dong Aisheng [Wed, 27 May 2015 10:13:28 +0000 (18:13 +0800)]
mmc: sdhci-esdhc-imx: add ADMA Length Mismatch errata fix

The uSDHC has an ADMA Length Mismatch errata ERR004536 which may
cause ADMA work abnormally. The errata has already been fixed for
i.MX6Q TO1.2 and i.MX6DL TO1.1 by enable the bit 7 in 0x6c register.
Unfortunately this fix is not included in i.MX6SL.
So we disable ADMA for i.MX6SL and use SDMA instead.

Signed-off-by: Dong Aisheng <aisheng.dong@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: usdhc does not have missing card interrupt issue
Dong Aisheng [Wed, 27 May 2015 10:13:27 +0000 (18:13 +0800)]
mmc: sdhci-esdhc-imx: usdhc does not have missing card interrupt issue

The usdhc does not have missing card interrupt issue, so don't execute
workaround for usdhc.

Signed-off-by: Dong Aisheng <aisheng.dong@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: merge the same register check into one place
Dong Aisheng [Wed, 27 May 2015 10:13:26 +0000 (18:13 +0800)]
mmc: sdhci-esdhc-imx: merge the same register check into one place

In esdhc_writel_le() function, there's duplicated checking of the same
register as follows:
"if (unlikely(reg == SDHCI_INT_ENABLE || reg == SDHCI_SIGNAL_ENABLE))".
Merge them into one and remove the duplicated one.

Signed-off-by: Dong Aisheng <aisheng.dong@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: Enable HS400 for some Intel host controllers
Adrian Hunter [Fri, 6 Feb 2015 12:13:01 +0000 (14:13 +0200)]
mmc: sdhci-pci: Enable HS400 for some Intel host controllers

Enable detection of HS400 support via capability bit-63
for some Intel host controllers.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: Add support for drive strength selection for SPT
Adrian Hunter [Fri, 6 Feb 2015 12:13:00 +0000 (14:13 +0200)]
mmc: sdhci-pci: Add support for drive strength selection for SPT

Implement the select_drive_strength callback to provide
drive strength selection for Intel SPT.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: Add a callback to select drive strength
Adrian Hunter [Fri, 6 Feb 2015 12:12:59 +0000 (14:12 +0200)]
mmc: sdhci: Add a callback to select drive strength

Add a callbak to let host drivers select drive
strength.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mmc: Add driver strength selection
Adrian Hunter [Fri, 6 Feb 2015 12:12:58 +0000 (14:12 +0200)]
mmc: mmc: Add driver strength selection

Add the ability to set eMMC driver strength
for HS200 and HS400.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mmc: Read card's valid driver strength mask
Adrian Hunter [Fri, 6 Feb 2015 12:12:57 +0000 (14:12 +0200)]
mmc: mmc: Read card's valid driver strength mask

In preparation for supporing drive strength selection
for eMMC, read the card's valid driver strengths.

Note that though the SD spec uses the term "drive strength",
the JEDEC eMMC spec uses the term "driver strength".

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Record card drive strength
Adrian Hunter [Fri, 6 Feb 2015 12:12:56 +0000 (14:12 +0200)]
mmc: core: Record card drive strength

In preparation for adding drive strength support
for eMMC, add drive_strength to struct mmc_card
to record the card drive strength for UHS-I modes
and HS200 / HS400. For eMMC this will be needed
when switching between HS200 and HS400.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Factor out common code in drive strength selection
Adrian Hunter [Fri, 6 Feb 2015 12:12:55 +0000 (14:12 +0200)]
mmc: core: Factor out common code in drive strength selection

Make a new function out of common code used for drive
strength selection.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Add 'card' to drive strength selection callback
Adrian Hunter [Fri, 6 Feb 2015 12:12:54 +0000 (14:12 +0200)]
mmc: core: Add 'card' to drive strength selection callback

In preparation for supporting also eMMC drive strength,
add the 'card' as a parameter so that the callback can
distinguish different types of cards if necessary.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Simplify card drive strength mask
Adrian Hunter [Fri, 6 Feb 2015 12:12:53 +0000 (14:12 +0200)]
mmc: core: Simplify card drive strength mask

Card drive strength selection uses a callback to
which a mask of supported drive strengths is passed.
Currently, the bits are checked against the values
in the SD specifications. That is not necessary
because the callback will anyway match the mask
against a valid value. Simplify by taking the mask
as is but still ensuring that the default mandatory
value (type B) is always supported.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Allow card drive strength to be different to host
Adrian Hunter [Fri, 6 Feb 2015 12:12:52 +0000 (14:12 +0200)]
mmc: core: Allow card drive strength to be different to host

Initialization of UHS-I modes for SD and SDIO cards
employs a callback to allow the host driver to
choose a drive strength value. Currently that
assumes the card drive strength and host driver
type must be the same value. Change to let the
callback make that decision and return both the
card drive strength and host driver type.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Reset driver type to default
Adrian Hunter [Fri, 6 Feb 2015 12:12:51 +0000 (14:12 +0200)]
mmc: core: Reset driver type to default

IO state variable drv_type could be set during card
initialization. Consequently, it must be reset to the
default value when setting the initial state.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: calculate best clock with parent clock
Kuninori Morimoto [Thu, 14 May 2015 07:22:46 +0000 (07:22 +0000)]
mmc: sh_mmcif: calculate best clock with parent clock

MMCIF IP on R-Car series has parent clock which can be set several rate,
and it was not implemented on old SH-Mobile series (= SH-Mobile series
parent clock was fixed rate) R-Car series MMCIF can use more high speed
access if it setups parent clock. This patch adds parent clock setup
method. It will be used if DT has "max-frequency", and then, this driver
assumes it is booted on R-Car Gen2 or later SoC. Because SH-Mobile series
(which doesn't boot from DT) and R-Car series (which boots from DT) have
different divider.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
[Ulf: Silence compiler warning]

9 years agommc: sh_mmcif: use sh_mmcif_xxx prefix for all functions
Kuninori Morimoto [Thu, 14 May 2015 07:21:36 +0000 (07:21 +0000)]
mmc: sh_mmcif: use sh_mmcif_xxx prefix for all functions

Current sh_mmcif driver is using sh_mmcif_xxx and mmcif_xxx
for functions. This patch used sh_mmcif_xxx for all functions.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: add sh_mmcif_host_to_dev() macro and use it.
Kuninori Morimoto [Thu, 14 May 2015 07:21:18 +0000 (07:21 +0000)]
mmc: sh_mmcif: add sh_mmcif_host_to_dev() macro and use it.

Current sh_mmcif driver is directly using &host->pd->dev in all place.
It is not big problem, but it is unreadable, and it can be cause of
future bug. This patch adds new sh_mmcif_host_to_dev() and use it.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: fix driver type B and D handling in sdhci_do_set_ios()
Petri Gynther [Wed, 20 May 2015 21:35:00 +0000 (14:35 -0700)]
mmc: sdhci: fix driver type B and D handling in sdhci_do_set_ios()

sdhci_do_set_ios() doesn't currently program SDHCI_HOST_CONTROL2
register correctly when host->preset_enabled == false.

Add code to handle the missing cases MMC_SET_DRIVER_TYPE_B and
MMC_SET_DRIVER_TYPE_D.

Signed-off-by: Petri Gynther <pgynther@google.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agopowerpc/dts: add eSDHC compatible list
Yangbo Lu [Fri, 15 May 2015 02:46:02 +0000 (10:46 +0800)]
powerpc/dts: add eSDHC compatible list

Add eSDHC compatible list for P2041/P3041/P4080/P5020/P5040.

Signed-off-by: Yangbo Lu <yangbo.lu@freescale.com>
Acked-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mmc-test: use swap() in mmc_test_nonblock_transfer()
Fabian Frederick [Mon, 18 May 2015 17:52:58 +0000 (19:52 +0200)]
mmc: mmc-test: use swap() in mmc_test_nonblock_transfer()

Use kernel.h macro definition.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Increase delay for voltage to stabilize from 3.3V to 1.8V
Doug Anderson [Tue, 12 May 2015 21:46:11 +0000 (14:46 -0700)]
mmc: core: Increase delay for voltage to stabilize from 3.3V to 1.8V

Since the regulator used for the SDMMC IO voltage is not expected to
draw a lot of current, most systems will probably use an inexpensive
LDO for it.  LDO regulators apparently have the feature that they
don't actively drive the voltage down--they wait for other components
in the system to drag the voltage down.  Thus they will transition
faster under heavy loads and slower under light loads.

During an SDMMC voltage change from 3.3V to 1.8V, we are almost
certainly under a light load.  To be specific:
* The regulator is hooked through pulls to CMD0-3 and DAT.  Probably
  the CMD pulls are something like 47K and the DAT is something like
  10K.
* The card is supposed to be driving DAT0-3 low during voltage change
  which will draw _some_ current, but not a lot.
* The regulator is also provided to the SDMMC host controller, but the
  SDMMC host controller is in open drain mode during the voltage
  change and so shouldn't be drawing much current.

In order to keep the SDMMC host working properly (or for noise
reasons), there might also be a capacitor attached to the SDMMC IO
regulator.  This also will have the effect of slowing down transitions
of the regulator, especially under light loads.

From experimental evidence, we've seen the voltage change fail if the
card doesn't detect that the voltage fell to less than about 2.3V when
we turn on the clock.  On one device (that admittedly had a 47K CMD
pullup instead of a 10K CMD pullup) we saw that the voltage was just
about 2.3V after 5ms and thus the voltage change would sometimes fail.
Doubling the delay gave margin and made the voltage change work 100%
of the time, despite the slightly weaker CMD pull.

At the moment submitting this as an RFC patch since my problem _could_
be fixed by increasing the pull strength (or using a smaller
capacitor).  However being a little bit more lenient to strange
hardware could also be a good thing.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Acked-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-st: Fix modalias to make module auto-loading work
Zhangfei Gao [Thu, 14 May 2015 08:59:47 +0000 (16:59 +0800)]
mmc: sdhci-st: Fix modalias to make module auto-loading work

Make the modalias match driver name, this is required to make module
auto-loading work.

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
CC: Peter Griffin <peter.griffin@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: rockchip: Fix modalias to make module auto-loading work
Zhangfei Gao [Thu, 14 May 2015 08:59:46 +0000 (16:59 +0800)]
mmc: dw_mmc: rockchip: Fix modalias to make module auto-loading work

Make the modalias match driver name, this is required to make module
auto-loading work.

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
CC: Doug Anderson <dianders@chromium.org>
Reviewed-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: exynos: Fix modalias to make module auto-loading work
Zhangfei Gao [Thu, 14 May 2015 08:59:45 +0000 (16:59 +0800)]
mmc: dw_mmc: exynos: Fix modalias to make module auto-loading work

Make the modalias match driver name, this is required to make module
auto-loading work.

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: k3: Fix modalias to make module auto-loading work
Zhangfei Gao [Thu, 14 May 2015 08:59:44 +0000 (16:59 +0800)]
mmc: dw_mmc: k3: Fix modalias to make module auto-loading work

Make the modalias match driver name, this is required to make module
auto-loading work.

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Use core to handle absent write protect line
Lars-Peter Clausen [Wed, 6 May 2015 18:31:22 +0000 (20:31 +0200)]
mmc: dw_mmc: Use core to handle absent write protect line

Use the new MMC_CAP2_NO_WRITE_PROTECT to let the core handle the case where
no write protect line is present instead of having custom driver code to
handle it.

dw_mci_of_get_slot_quirks() is slightly refactored to directly modify the
mmc_host capabilities instead of returning a quirk mask.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: k3: support hi6220
Zhangfei Gao [Thu, 14 May 2015 08:45:19 +0000 (16:45 +0800)]
mmc: dw_mmc: k3: support hi6220

Support hi6220, tested on hikey board
emmc: support hs
sd: support hs, sdr12, sdr25

Signed-off-by: Jorge A. Ramirez-Ortiz <jorge.ramirez-ortiz@linaro.org>
Signed-off-by: Dan Yuan <yuandan@hisilicon.com>
Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: add switch_voltage
Zhangfei Gao [Thu, 14 May 2015 08:45:18 +0000 (16:45 +0800)]
mmc: dw_mmc: add switch_voltage

switch_voltage is required on some platform since special register accessing

Signed-off-by: Jorge A. Ramirez-Ortiz <jorge.ramirez-ortiz@linaro.org>
Signed-off-by: Dan Yuan <yuandan@hisilicon.com>
Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoDocument: dw_mmc-k3: add document of hi6220 mmc
Zhangfei Gao [Thu, 14 May 2015 08:45:17 +0000 (16:45 +0800)]
Document: dw_mmc-k3: add document of hi6220 mmc

Add bindings for hi6220 mmc support

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: cast unsigned int to typeof(sector_t) to avoid unexpected error
Kuninori Morimoto [Mon, 11 May 2015 07:35:28 +0000 (07:35 +0000)]
mmc: cast unsigned int to typeof(sector_t) to avoid unexpected error

card->csd.capacity is defined as "unsigned int", and sector_t is defined as
"u64" or "unsigned long" (depends on CONFIG_LBDAF). Thus, sector_t data
might have strange data (see below). This patch cast it to typeof(sector_t)
Special thanks to coverity <http://www.coverity.com>

ex) if sector_t was u64

        unsigned int data;
        sector_t sector;

        data = 0x800000;
        sector = (data << 8); // 0xffffffff80000000
        sector = (((typeof(sector_t))data) << 8); // 0x80000000

or

        data = 0x80000000;
        sector = (data << 8); // 0x0
        sector = (((typeof(sector_t))data) << 8); // 0x8000000000

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: cast u8 to unsigned long long to avoid unexpected error
Kuninori Morimoto [Mon, 11 May 2015 07:34:53 +0000 (07:34 +0000)]
mmc: cast u8 to unsigned long long to avoid unexpected error

card->ext_csd.enhanced_area_offset is defined as "unsigned long long",
and, ext_csd[] is defined as u8.
unsigned long long data might have strange data if first bit of ext_csd[]
was 1. this patch cast it to (unsigned long long)
Special thanks to coverity <http://www.coverity.com>

ex)
        u8  data8;
        u64 data64;

        data8 = 0x80;
        data64 = (data8 << 24); // 0xffffffff80000000
        data64 = (((unsigned long long)data8) << 24); // 0x80000000;

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: host: sdhci: Use BUG_ON()
Fabio Estevam [Sat, 9 May 2015 21:44:51 +0000 (18:44 -0300)]
mmc: host: sdhci: Use BUG_ON()

Use BUG_ON() instead of an 'if' condition followed by BUG().

The semantic patch that makes this change is available
in scripts/coccinelle/misc/bugon.cocci.

More information about semantic patching is available at
http://coccinelle.lip6.fr/

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: host: mxcmmc: Simplify a trivial if-return sequence
Fabio Estevam [Sat, 9 May 2015 21:44:50 +0000 (18:44 -0300)]
mmc: host: mxcmmc: Simplify a trivial if-return sequence

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

The semantic patch that makes this change is available
in scripts/coccinelle/misc/simple_return.cocci.

More information about semantic patching is available at
http://coccinelle.lip6.fr/

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: card: mmc_test: Simplify a trivial if-return sequence
Fabio Estevam [Sat, 9 May 2015 21:44:49 +0000 (18:44 -0300)]
mmc: card: mmc_test: Simplify a trivial if-return sequence

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

The semantic patch that makes this change is available
in scripts/coccinelle/misc/simple_return.cocci.

More information about semantic patching is available at
http://coccinelle.lip6.fr/

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: Do not break platform data boards
Fabio Estevam [Sat, 9 May 2015 12:57:09 +0000 (09:57 -0300)]
mmc: sdhci-esdhc-imx: Do not break platform data boards

The only user of this driver that has not been converted to fully
device tree is the i.MX35 SoC.

There is a i.MX35-based board (mach-pcm043.c) that uses platform data
to pass wp_gpio and cd_gpio information.

Commit 8d86e4fcccf61ba ("mmc: sdhci-esdhc-imx: Call mmc_of_parse()")
broke the platform data case by removing mmc_gpio_request_ro() and
mmc_gpio_request_cd(), so restore the functionality for the non-dt
case.

Also, restore the check for ESDHC_CD_CONTROLLER so that we can still
support the "fsl,cd-controller" property.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: Move mmc_of_parse() to the dt probe
Fabio Estevam [Sat, 9 May 2015 12:57:08 +0000 (09:57 -0300)]
mmc: sdhci-esdhc-imx: Move mmc_of_parse() to the dt probe

mmc_of_parse() should be placed inside sdhci_esdhc_imx_probe_dt() as it
suits only for the dt case.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dt: Allow to specify that no write protect signal is present
Lars-Peter Clausen [Wed, 6 May 2015 18:31:20 +0000 (20:31 +0200)]
mmc: dt: Allow to specify that no write protect signal is present

Allow to specify in the device-tree that no physical write-protect signal
is connected to a particular instance of a MMC controller. Setting the
property will cause the core will assume that the SD card is always
read-write.

The name for the new property is 'disable-wp' and was chosen based on the
property with the same function from the Synopsys designware mobile storage
host controller DT bindings specification.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Ian Campbell <ijc+devicetree@hellion.org.uk>
Cc: Kumar Gala <galak@codeaurora.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: Add support for disabling write-protect detection
Lars-Peter Clausen [Wed, 6 May 2015 18:31:19 +0000 (20:31 +0200)]
mmc: Add support for disabling write-protect detection

It is not uncommon to see systems where there is no physical write-protect
signal (e.g. when using eMMC or microSD card slots). For some controllers,
which have a dedicated write-protection detection logic (like SDHCI
controllers), the get_ro() callback can return bogus data in such a case.

Instead of handling this on a per controller basis this patch adds a new
capability flag to the MMC core that can be set to specify that the result
of get_ro() is invalid. When the flag is set the core will not call
get_ro() and assume that the card is always read-write.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Don't print reset warning if reset is not supported
Adrian Hunter [Thu, 7 May 2015 10:10:25 +0000 (13:10 +0300)]
mmc: core: Don't print reset warning if reset is not supported

Check the error code for EOPNOTSUPP and do not print
reset warning in that case.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: block: Retry errored data requests when re-tuning is needed
Adrian Hunter [Thu, 7 May 2015 10:10:24 +0000 (13:10 +0300)]
mmc: block: Retry errored data requests when re-tuning is needed

Retry errored data requests when re-tuning is needed and
add a flag to struct mmc_blk_request so that the retry
is only done once.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: block: Check re-tuning in the recovery path
Adrian Hunter [Thu, 7 May 2015 10:10:23 +0000 (13:10 +0300)]
mmc: block: Check re-tuning in the recovery path

If re-tuning is needed, do it in the recovery path to
give recovery commands a better chance of success.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Flag re-tuning is needed on CRC errors
Adrian Hunter [Thu, 7 May 2015 10:10:22 +0000 (13:10 +0300)]
mmc: core: Flag re-tuning is needed on CRC errors

CRC errors could possibly be alleviated by
re-tuning so flag re-tuning needed in those cases.
Note this has no effect if re-tuning has not been
enabled.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: Change to new way of doing re-tuning
Adrian Hunter [Thu, 7 May 2015 10:10:21 +0000 (13:10 +0300)]
mmc: sdhci: Change to new way of doing re-tuning

Make use of mmc core support for re-tuning instead
of doing it all in the sdhci driver.

This patch also changes to flag the need for re-tuning
always after runtime suspend when tuning has been used
at initialization. Previously it was only done if
the re-tuning timer was in use.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Add support for HS400 re-tuning
Adrian Hunter [Thu, 7 May 2015 10:10:20 +0000 (13:10 +0300)]
mmc: core: Add support for HS400 re-tuning

HS400 re-tuning must be done in HS200 mode. Add
the ability to switch from HS400 mode to HS200
mode before re-tuning and switch back to HS400
after re-tuning.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Separate out the mmc_switch status check so it can be re-used
Adrian Hunter [Thu, 7 May 2015 10:10:19 +0000 (13:10 +0300)]
mmc: core: Separate out the mmc_switch status check so it can be re-used

Make a separate function to do the mmc_switch status check
so it can be re-used. This is preparation for adding support
for HS400 re-tuning.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mmc: Hold re-tuning in mmc_sleep()
Adrian Hunter [Thu, 7 May 2015 10:10:18 +0000 (13:10 +0300)]
mmc: mmc: Hold re-tuning in mmc_sleep()

The sleep command is issued after deselecting the
card, but re-tuning won't work on a deselected card
so re-tuning must be held.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Hold re-tuning while bkops ongoing
Adrian Hunter [Thu, 7 May 2015 10:10:17 +0000 (13:10 +0300)]
mmc: core: Hold re-tuning while bkops ongoing

Hold re-tuning during bkops to prevent
it from conflicting with the busy state.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Hold re-tuning during erase commands
Adrian Hunter [Thu, 7 May 2015 10:10:16 +0000 (13:10 +0300)]
mmc: core: Hold re-tuning during erase commands

Hold re-tuning during erase commands to prevent
it from conflicting with the sequence of commands.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Hold re-tuning during switch commands
Adrian Hunter [Thu, 7 May 2015 10:10:15 +0000 (13:10 +0300)]
mmc: core: Hold re-tuning during switch commands

Hold re-tuning during switch commands to prevent
it from conflicting with the busy state or the CMD13
verification.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Add support for re-tuning before each request
Adrian Hunter [Thu, 7 May 2015 10:10:14 +0000 (13:10 +0300)]
mmc: core: Add support for re-tuning before each request

At the start of each request, re-tune if needed and
then hold off re-tuning again until the request is done.

Note that though there is one function that starts
requests (mmc_start_request) there are two that wait for
the request to be done (mmc_wait_for_req_done and
mmc_wait_for_data_req_done).  Also note that
mmc_wait_for_data_req_done can return even when the
request is not done (which allows the block driver
to prepare a newly arrived request while still
waiting for the previous request).

This patch ensures re-tuning is held for the duration
of a request.  Subsequent patches will also hold
re-tuning at other times when it might cause a
conflict.

In addition, possibly a command is failing because
re-tuning is needed. Use mmc_retune_recheck() to check
re-tuning. At that point re-tuning is held, at least by
the request, so mmc_retune_recheck() flags host->retune_now
if the hold count is 1.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Enable / disable re-tuning
Adrian Hunter [Thu, 7 May 2015 10:10:13 +0000 (13:10 +0300)]
mmc: core: Enable / disable re-tuning

Enable re-tuning when tuning is executed and
disable re-tuning when card is no longer initialized.

In the case of SDIO suspend, the card can keep power.
In that case, re-tuning need not be disabled, but, if
a re-tuning timer is being used, ensure it is disabled
and assume that re-tuning will be needed upon resume
since it is not known how long the suspend will last.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: host: Add facility to support re-tuning
Adrian Hunter [Thu, 7 May 2015 10:10:12 +0000 (13:10 +0300)]
mmc: host: Add facility to support re-tuning

Currently, there is core support for tuning during
initialization. There can also be a need to re-tune
periodically (e.g. sdhci) or to re-tune after the
host controller is powered off (e.g. after PM
runtime suspend / resume) or to re-tune in response
to CRC errors.

The main requirements for re-tuning are:
  - ability to enable / disable re-tuning
  - ability to flag that re-tuning is needed
  - ability to re-tune before any request
  - ability to hold off re-tuning if the card is busy
  - ability to hold off re-tuning if re-tuning is in
  progress
  - ability to run a re-tuning timer

To support those requirements 7 members are added to struct
mmc_host:

  unsigned int can_retune:1; /* re-tuning can be used */
  unsigned int doing_retune:1; /* re-tuning in progress */
  unsigned int retune_now:1;   /* do re-tuning at next req */
  int need_retune; /* re-tuning is needed */
  int hold_retune; /* hold off re-tuning */
  unsigned int retune_period;  /* re-tuning period in secs */
  struct timer_list retune_timer; /* for periodic re-tuning */

need_retune is an integer so it can be set without needing
synchronization. hold_retune is a integer to allow nesting.

Various simple functions are provided to set / clear those
variables.

Subsequent patches take those functions into use.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-{pxav2,pxav3}: Use of_match_ptr() macro
Axel Lin [Tue, 5 May 2015 09:11:54 +0000 (17:11 +0800)]
mmc: sdhci-{pxav2,pxav3}: Use of_match_ptr() macro

This eliminates having an #ifdef returning NULL for the case
when OF is disabled.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Acked-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-of-arasan: Add the support for sdhci-arasan4.9a
Suman Tripathi [Mon, 4 May 2015 13:39:51 +0000 (19:09 +0530)]
mmc: sdhci-of-arasan: Add the support for sdhci-arasan4.9a

This patch adds the quirks and compatible string in sdhci-of-arasan.c
to support sdhci-arasan4.9a version of controller.

Signed-off-by: Suman Tripathi <stripathi@apm.com>
Reviewed-by: Michal Simek <michal.simek@xilinx.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-s3c: Constify platform_device_id
Krzysztof Kozlowski [Fri, 1 May 2015 15:49:22 +0000 (00:49 +0900)]
mmc: sdhci-s3c: Constify platform_device_id

The platform_device_id is not modified by the driver and core uses it as
const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski.k@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-imx: Constify platform_device_id
Krzysztof Kozlowski [Fri, 1 May 2015 15:49:21 +0000 (00:49 +0900)]
mmc: sdhci-imx: Constify platform_device_id

The platform_device_id is not modified by the driver and core uses it as
const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski.k@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: s3cmci: Constify platform_device_id
Krzysztof Kozlowski [Fri, 1 May 2015 15:49:20 +0000 (00:49 +0900)]
mmc: s3cmci: Constify platform_device_id

The platform_device_id is not modified by the driver and core uses it as
const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski.k@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: rtsx: Constify platform_device_id
Krzysztof Kozlowski [Fri, 1 May 2015 15:49:19 +0000 (00:49 +0900)]
mmc: rtsx: Constify platform_device_id

The platform_device_id is not modified by the driver and core uses it as
const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski.k@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mxs: Constify platform_device_id
Krzysztof Kozlowski [Fri, 1 May 2015 15:49:18 +0000 (00:49 +0900)]
mmc: mxs: Constify platform_device_id

The platform_device_id is not modified by the driver and core uses it as
const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski.k@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: davinci: Constify platform_device_id
Krzysztof Kozlowski [Fri, 1 May 2015 15:49:17 +0000 (00:49 +0900)]
mmc: davinci: Constify platform_device_id

The platform_device_id is not modified by the driver and core uses it as
const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski.k@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-sirf: fake version and capbility registers
Weijun Yang [Mon, 27 Apr 2015 08:15:14 +0000 (08:15 +0000)]
mmc: sdhci-sirf: fake version and capbility registers

chips have some issues for version and capbility registers, here we fake
them.

Signed-off-by: Weijun Yang <Weijun.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-sirf: fix the tuning count in platform_execute_tuning
Weijun Yang [Mon, 27 Apr 2015 08:15:13 +0000 (08:15 +0000)]
mmc: sdhci-sirf: fix the tuning count in platform_execute_tuning

hardware has 16bit to record the tuning count, so fix it to 16384.
at the same time, tuned_phases[SIRF_TUNING_COUNT] is useless as the
array is never used, so move it to a variant.

Signed-off-by: Weijun Yang <Weijun.Yang@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoMAINTAINERS: mmc: Update Ian Molton's address for tmio_mmc driver
Ben Hutchings [Sun, 26 Apr 2015 23:01:53 +0000 (00:01 +0100)]
MAINTAINERS: mmc: Update Ian Molton's address for tmio_mmc driver

Ian no longer works for Codethink so the current address doesn't work.

Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: TMIO: Use devm_request_irq()
Ian Molton [Sun, 26 Apr 2015 23:01:36 +0000 (00:01 +0100)]
mmc: TMIO: Use devm_request_irq()

Clean up resource allocation and freeing.

Signed-off-by: Ian Molton <ian.molton@codethink.co.uk>
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: TMIO: Fix I/O mapping leak on error using devm_ioremap()
Ian Molton [Sun, 26 Apr 2015 23:01:17 +0000 (00:01 +0100)]
mmc: TMIO: Fix I/O mapping leak on error using devm_ioremap()

Signed-off-by: Ian Molton <ian.molton@codethink.co.uk>
[bwh: Forward-ported to 4.0]
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: TMIO: Ensure MFD cell is disabled on probe error path
Ian Molton [Sun, 26 Apr 2015 23:01:06 +0000 (00:01 +0100)]
mmc: TMIO: Ensure MFD cell is disabled on probe error path

Signed-off-by: Ian Molton <ian.molton@codethink.co.uk>
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci_f_sdh30: Fix the size passed to sdhci_alloc_host
Axel Lin [Fri, 24 Apr 2015 07:02:00 +0000 (15:02 +0800)]
mmc: sdhci_f_sdh30: Fix the size passed to sdhci_alloc_host

sdhci_alloc_host() takes priv_size rather than
sizeof(struct sdhci_host) + priv_size.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci_f_sdh30: Staticize local functions
Axel Lin [Fri, 24 Apr 2015 07:00:29 +0000 (15:00 +0800)]
mmc: sdhci_f_sdh30: Staticize local functions

Make local functions static.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-of-esdhc: Pre divider starts at 1
Joakim Tjernlund [Mon, 20 Apr 2015 21:12:13 +0000 (23:12 +0200)]
mmc: sdhci-of-esdhc: Pre divider starts at 1

For PowerPC esdhc pre divider starts at 1, fixing the increases
the actual clock from 40KHz to 50 KHz.

Signed-off-by: Joakim Tjernlund <joakim.tjernlund@transmode.se>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: host: sdhci-esdhc-imx: fix broken email address
Wolfram Sang [Mon, 20 Apr 2015 13:51:42 +0000 (15:51 +0200)]
mmc: host: sdhci-esdhc-imx: fix broken email address

My Pengutronix address is not valid anymore, redirect people to the Pengutronix
kernel team.

Reported-by: Harald Geyer <harald@ccbib.org>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Acked-by: Robert Schwebel <r.schwebel@pengutronix.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: pwrseq: dt: example with reset clock and active low pin
Andreas Fenkart [Thu, 16 Apr 2015 08:03:14 +0000 (10:03 +0200)]
mmc: pwrseq: dt: example with reset clock and active low pin

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: separate sh_mmcif_clk_update() into setup and prepare
Kuninori Morimoto [Thu, 23 Apr 2015 08:16:04 +0000 (08:16 +0000)]
mmc: sh_mmcif: separate sh_mmcif_clk_update() into setup and prepare

Current sh_mmcif_clk_update() is called from probe() and set_ios(),
but, the purpose of later one is just clk_prepare_enable().
No need to setup mmc->f_max/f_min in many times.
This patch separe sh_mmcif_clk_update() into clk_prepare_enable()
and mmc->f_max/f_min setup.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: remove unnecessary int clk from struct sh_mmcif_host
Kuninori Morimoto [Thu, 23 Apr 2015 08:15:08 +0000 (08:15 +0000)]
mmc: sh_mmcif: remove unnecessary int clk from struct sh_mmcif_host

struct sh_mmcif_host has 1) int clk, 2) struct clock *hclk,
and host->clk = clk_get_rate(host->hclk).
This int clk is not necessary. Let's remove it.
And, current hclk is confusable naming. Let's rename it to clk.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: cleanup to use dev instead of &pdev->dev
Kuninori Morimoto [Thu, 23 Apr 2015 08:14:12 +0000 (08:14 +0000)]
mmc: sh_mmcif: cleanup to use dev instead of &pdev->dev

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: move mmcif_of_match to upside
Kuninori Morimoto [Thu, 23 Apr 2015 08:13:25 +0000 (08:13 +0000)]
mmc: sh_mmcif: move mmcif_of_match to upside

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Tested-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdio: add reset callback to bus operations
Andreas Fenkart [Thu, 23 Apr 2015 11:15:42 +0000 (13:15 +0200)]
mmc: sdio: add reset callback to bus operations

Some drivers schedule automatic hw resets. An example is mwifiex,
which schedules a card reset if the command handler between driver
and card firmware becomes out of sync

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoLinux 4.1-rc5
Linus Torvalds [Mon, 25 May 2015 01:22:35 +0000 (18:22 -0700)]
Linux 4.1-rc5

9 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sun, 24 May 2015 18:15:28 +0000 (11:15 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "This is a set of five fixes: Two MAINTAINER email updates (urgent
  because the non-avagotech emails will start bouncing) an lpfc big
  endian oops fix, a 256 byte sector hang fix (to eliminate 256 byte
  sectors) and a storvsc fix which could cause test unit ready failures
  on bringup"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  MAINTAINERS: Revise lpfc maintainers for Avago Technologies ownership of Emulex
  MAINTAINERS, be2iscsi: change email domain
  sd: Disable support for 256 byte/sector disks
  lpfc: Fix breakage on big endian kernels
  storvsc: Set the SRB flags correctly when no data transfer is needed

9 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 24 May 2015 00:57:40 +0000 (17:57 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Thomas Gleixner:
 "One more fix from the timer departement:

    - Handle division of negative nanosecond values proper on 32bit.

      A recent cleanup wrecked the sign handling of the dividend and
      dropped the check for negative divisors"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  ktime: Fix ktime_divns to do signed division

9 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 24 May 2015 00:33:10 +0000 (17:33 -0700)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irqchip fix from Thomas Gleixner:
 "A fix for a GIC-V3 irqchip regression which prevents some systems from
  booting"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/gicv3-its: ITS table size should not be smaller than PSZ

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Sat, 23 May 2015 18:28:25 +0000 (11:28 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

Pull two Ceph fixes from Sage Weil:
 "These fix an issue with the RBD notifications when there are topology
  changes in the cluster"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  Revert "libceph: clear r_req_lru_item in __unregister_linger_request()"
  libceph: request a new osdmap if lingering request maps to no osd

9 years agoMerge branch 'for-linus-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/mason...
Linus Torvalds [Sat, 23 May 2015 18:14:10 +0000 (11:14 -0700)]
Merge branch 'for-linus-4.1' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "I fixed up a regression from 4.0 where conversion between different
  raid levels would sometimes bail out without converting.

  Filipe tracked down a race where it was possible to double allocate
  chunks on the drive.

  Mark has a fix for fiemap.  All three will get bundled off for stable
  as well"

* 'for-linus-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: fix regression in raid level conversion
  Btrfs: fix racy system chunk allocation when setting block group ro
  btrfs: clear 'ret' in btrfs_check_shared() loop

9 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 23 May 2015 00:34:24 +0000 (17:34 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Radeon has two displayport fixes, one for a regression.

  i915 regression flicker fix needed so 4.0 can get fixed.

  A bunch of msm fixes and a bunch of exynos fixes, these two are
  probably a bit larger than I'd like, but most of them seems pretty
  good"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (29 commits)
  drm/radeon: fix error flag checking in native aux path
  drm/radeon: retry dcpd fetch
  drm/msm/mdp5: fix incorrect parameter for msm_framebuffer_iova()
  drm/exynos: dp: Lower level of EDID read success message
  drm/exynos: cleanup exynos_drm_plane
  drm/exynos: 'win' is always unsigned
  drm/exynos: mixer: don't dump registers under spinlock
  drm/exynos: Consolidate return statements in fimd_bind()
  drm/exynos: Constify exynos_drm_crtc_ops
  drm/exynos: Fix build breakage on !DRM_EXYNOS_FIMD
  drm/exynos: mixer: Constify platform_device_id
  drm/exynos: mixer: cleanup pixelformat handling
  drm/exynos: mixer: also allow NV21 for the video processor
  drm/exynos: mixer: remove buffer count handling in vp_video_buffer()
  drm/exynos: plane: honor buffer offset for dma_addr
  drm/exynos: fb: use drm_format_num_planes to get buffer count
  drm/i915: fix screen flickering
  drm/msm: fix locking inconsistencies in gpu->destroy()
  drm/msm/dsi: Simplify the code to get the number of read byte
  drm/msm: Attach assigned encoder to eDP and DSI connectors
  ...

9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 22 May 2015 22:44:50 +0000 (15:44 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Don't leak ipvs->sysctl_tbl, from Tommi Rentala.

 2) Fix neighbour table entry leak in rocker driver, from Ying Xue.

 3) Do not emit bonding notifications for unregistered interfaces, from
    Nicolas Dichtel.

 4) Set ipv6 flow label properly when in TIME_WAIT state, from Florent
    Fourcot.

 5) Fix regression in ipv6 multicast filter test, from Henning Rogge.

 6) do_replace() in various footables netfilter modules is missing a
    check for 0 counters in the datastructure provided by the user.  Fix
    from Dave Jones, and found with trinity.

 7) Fix RCU bug in packet scheduler classifier module unloads, from
    Daniel Borkmann.

 8) Avoid deadlock in tcp_get_info() by using u64_sync.  From Eric
    Dumzaet.

 9) Input packet processing can race with inetdev_destroy() teardown,
    fix potential OOPS in ip_error() by explicitly testing whether the
    inetdev is still attached.  From Eric W Biederman.

10) MLDv2 parser in bridge multicast code breaks too early while
    parsing.  Fix from Thadeu Lima de Souza Cascardo.

11) Asking for settings on non-zero PHYID doesn't work because we do not
    import the command structure from the user and use the PHYID
    provided there.  Fix from Arun Parameswaran.

12) Fix UDP checksums with IPV6 RAW sockets, from Vlad Yasevich.

13) Missing NF_TABLES depends for TPROXY etc can cause build failures,
    fix from Florian Westphal.

14) Fix netfilter conntrack to handle RFC5961 challenge ACKs properly,
    from Jesper Dangaard Brouer.

15) If netlink autobind retry fails, we have to reset the sockets portid
    back to zero.  From Herbert Xu.

16) VXLAN netns exit code unregisters using wrong device, from John W
    Linville.

17) Add some USB device IDs to ath3k and btusb bluetooth drivers, from
    Dmitry Tunin and Wen-chien Jesse Sung.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (44 commits)
  bridge: fix lockdep splat
  net: core: 'ethtool' issue with querying phy settings
  bridge: fix parsing of MLDv2 reports
  ARM: zynq: DT: Use the zynq binding with macb
  net: macb: Disable half duplex gigabit on Zynq
  net: macb: Document zynq gem dt binding
  ipv4: fill in table id when replacing a route
  cdc_ncm: Fix tx_bytes statistics
  ipv4: Avoid crashing in ip_error
  tcp: fix a potential deadlock in tcp_get_info()
  net: sched: fix call_rcu() race on classifier module unloads
  net: phy: Make sure phy_start() always re-enables the phy interrupts
  ipv6: fix ECMP route replacement
  ipv6: do not delete previously existing ECMP routes if add fails
  Revert "netfilter: bridge: query conntrack about skb dnat"
  netfilter: ensure number of counters is >0 in do_replace()
  netfilter: nfnetlink_{log,queue}: Register pernet in first place
  tcp: don't over-send F-RTO probes
  tcp: only undo on partial ACKs in CA_Loss
  net/ipv6/udp: Fix ipv6 multicast socket filter regression
  ...

9 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 22 May 2015 22:15:30 +0000 (15:15 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "Three small fixes that have been picked up the last few weeks.
  Specifically:

   - Fix a memory corruption issue in NVMe with malignant user
     constructed request.  From Christoph.

   - Kill (now) unused blk_queue_bio(), dm was changed to not need this
     anymore.  From Mike Snitzer.

   - Always use blk_schedule_flush_plug() from the io_schedule() path
     when flushing a plug, fixing a !TASK_RUNNING warning with md.  From
     Shaohua"

* 'for-linus' of git://git.kernel.dk/linux-block:
  sched: always use blk_schedule_flush_plug in io_schedule_out
  nvme: fix kernel memory corruption with short INQUIRY buffers
  block: remove export for blk_queue_bio

9 years agoMerge tag 'md/4.1-rc4-fixes' of git://neil.brown.name/md
Linus Torvalds [Fri, 22 May 2015 22:10:07 +0000 (15:10 -0700)]
Merge tag 'md/4.1-rc4-fixes' of git://neil.brown.name/md

Pull md bugfixes from Neil Brown:
 "I have a few more raid5 bugfixes pending, but I want them to get a bit
  more review first.  In the meantime:

   - one serious RAID0 data corruption - caused by recent bugfix that
     wasn't reviewed properly.

   - one raid5 fix in new code (a couple more of those to come).

   - one little fix to stop static analysis complaining about silly rcu
     annotation"

* tag 'md/4.1-rc4-fixes' of git://neil.brown.name/md:
  md/bitmap: remove rcu annotation from pointer arithmetic.
  md/raid0: fix restore to sector variable in raid0_make_request
  raid5: fix broken async operation chain

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Fri, 22 May 2015 21:49:55 +0000 (14:49 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Updates for the input subsystem.

  The main change is that we tell joydev not to touch "absolute mice",
  such as VMware virtual mouse, as that produced bad result (cursor
  stuck in upper right corner) with games"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: smtpe-ts - wait 50mS until polling for pen-up
  Input: smtpe-ts - use msecs_to_jiffies() instead of HZ
  Input: joydev - don't classify the vmmouse as a joystick
  Input: vmmouse - do not reference non-existing version of X driver
  Input: alps - fix finger jumps on lifting 2 fingers on v7 touchpad
  Input: elantech - fix semi-mt protocol for v3 HW
  Input: sx8654 - fix memory allocation check

9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 22 May 2015 21:26:36 +0000 (14:26 -0700)]
Merge git://git./linux/kernel/git/herbert/crypto-2.6

Pull another crypto fix from Herbert Xu:
 "Fix ICV corruption in s390/ghash when the same tfm is used by more
  than one thread"

* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: s390/ghash - Fix incorrect ghash icv buffer handling.

9 years agobridge: fix lockdep splat
Eric Dumazet [Thu, 21 May 2015 20:28:29 +0000 (13:28 -0700)]
bridge: fix lockdep splat

Following lockdep splat was reported :

[   29.382286] ===============================
[   29.382315] [ INFO: suspicious RCU usage. ]
[   29.382344] 4.1.0-0.rc0.git11.1.fc23.x86_64 #1 Not tainted
[   29.382380] -------------------------------
[   29.382409] net/bridge/br_private.h:626 suspicious
rcu_dereference_check() usage!
[   29.382455]
               other info that might help us debug this:

[   29.382507]
               rcu_scheduler_active = 1, debug_locks = 0
[   29.382549] 2 locks held by swapper/0/0:
[   29.382576]  #0:  (((&p->forward_delay_timer))){+.-...}, at:
[<ffffffff81139f75>] call_timer_fn+0x5/0x4f0
[   29.382660]  #1:  (&(&br->lock)->rlock){+.-...}, at:
[<ffffffffa0450dc1>] br_forward_delay_timer_expired+0x31/0x140
[bridge]
[   29.382754]
               stack backtrace:
[   29.382787] CPU: 0 PID: 0 Comm: swapper/0 Not tainted
4.1.0-0.rc0.git11.1.fc23.x86_64 #1
[   29.382838] Hardware name: LENOVO 422916G/LENOVO, BIOS A1KT53AUS 04/07/2015
[   29.382882]  0000000000000000 3ebfc20364115825 ffff880666603c48
ffffffff81892d4b
[   29.382943]  0000000000000000 ffffffff81e124e0 ffff880666603c78
ffffffff8110bcd7
[   29.383004]  ffff8800785c9d00 ffff88065485ac58 ffff880c62002800
ffff880c5fc88ac0
[   29.383065] Call Trace:
[   29.383084]  <IRQ>  [<ffffffff81892d4b>] dump_stack+0x4c/0x65
[   29.383130]  [<ffffffff8110bcd7>] lockdep_rcu_suspicious+0xe7/0x120
[   29.383178]  [<ffffffffa04520f9>] br_fill_ifinfo+0x4a9/0x6a0 [bridge]
[   29.383225]  [<ffffffffa045266b>] br_ifinfo_notify+0x11b/0x4b0 [bridge]
[   29.383271]  [<ffffffffa0450d90>] ? br_hold_timer_expired+0x70/0x70 [bridge]
[   29.383320]  [<ffffffffa0450de8>]
br_forward_delay_timer_expired+0x58/0x140 [bridge]
[   29.383371]  [<ffffffffa0450d90>] ? br_hold_timer_expired+0x70/0x70 [bridge]
[   29.383416]  [<ffffffff8113a033>] call_timer_fn+0xc3/0x4f0
[   29.383454]  [<ffffffff81139f75>] ? call_timer_fn+0x5/0x4f0
[   29.383493]  [<ffffffff8110a90f>] ? lock_release_holdtime.part.29+0xf/0x200
[   29.383541]  [<ffffffffa0450d90>] ? br_hold_timer_expired+0x70/0x70 [bridge]
[   29.383587]  [<ffffffff8113a6a4>] run_timer_softirq+0x244/0x490
[   29.383629]  [<ffffffff810b68cc>] __do_softirq+0xec/0x670
[   29.383666]  [<ffffffff810b70d5>] irq_exit+0x145/0x150
[   29.383703]  [<ffffffff8189f506>] smp_apic_timer_interrupt+0x46/0x60
[   29.383744]  [<ffffffff8189d523>] apic_timer_interrupt+0x73/0x80
[   29.383782]  <EOI>  [<ffffffff816f131f>] ? cpuidle_enter_state+0x5f/0x2f0
[   29.383832]  [<ffffffff816f131b>] ? cpuidle_enter_state+0x5b/0x2f0

Problem here is that br_forward_delay_timer_expired() is a timer
handler, calling br_ifinfo_notify() which assumes either rcu_read_lock()
or RTNL are held.

Simplest fix seems to add rcu read lock section.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Josh Boyer <jwboyer@fedoraproject.org>
Reported-by: Dominick Grift <dac.override@gmail.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: core: 'ethtool' issue with querying phy settings
Arun Parameswaran [Wed, 20 May 2015 21:35:30 +0000 (14:35 -0700)]
net: core: 'ethtool' issue with querying phy settings

When trying to configure the settings for PHY1, using commands
like 'ethtool -s eth0 phyad 1 speed 100', the 'ethtool' seems to
modify other settings apart from the speed of the PHY1, in the
above case.

The ethtool seems to query the settings for PHY0, and use this
as the base to apply the new settings to the PHY1. This is
causing the other settings of the PHY 1 to be wrongly
configured.

The issue is caused by the '_ethtool_get_settings()' API, which
gets called because of the 'ETHTOOL_GSET' command, is clearing
the 'cmd' pointer (of type 'struct ethtool_cmd') by calling
memset. This clears all the parameters (if any) passed for the
'ETHTOOL_GSET' cmd. So the driver's callback is always invoked
with 'cmd->phy_address' as '0'.

The '_ethtool_get_settings()' is called from other files in the
'net/core'. So the fix is applied to the 'ethtool_get_settings()'
which is only called in the context of the 'ethtool'.

Signed-off-by: Arun Parameswaran <aparames@broadcom.com>
Reviewed-by: Ray Jui <rjui@broadcom.com>
Reviewed-by: Scott Branden <sbranden@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobridge: fix parsing of MLDv2 reports
Thadeu Lima de Souza Cascardo [Fri, 22 May 2015 15:18:59 +0000 (12:18 -0300)]
bridge: fix parsing of MLDv2 reports

When more than a multicast address is present in a MLDv2 report, all but
the first address is ignored, because the code breaks out of the loop if
there has not been an error adding that address.

This has caused failures when two guests connected through the bridge
tried to communicate using IPv6. Neighbor discoveries would not be
transmitted to the other guest when both used a link-local address and a
static address.

This only happens when there is a MLDv2 querier in the network.

The fix will only break out of the loop when there is a failure adding a
multicast address.

The mdb before the patch:

dev ovirtmgmt port vnet0 grp ff02::1:ff7d:6603 temp
dev ovirtmgmt port vnet1 grp ff02::1:ff7d:6604 temp
dev ovirtmgmt port bond0.86 grp ff02::2 temp

After the patch:

dev ovirtmgmt port vnet0 grp ff02::1:ff7d:6603 temp
dev ovirtmgmt port vnet1 grp ff02::1:ff7d:6604 temp
dev ovirtmgmt port bond0.86 grp ff02::fb temp
dev ovirtmgmt port bond0.86 grp ff02::2 temp
dev ovirtmgmt port bond0.86 grp ff02::d temp
dev ovirtmgmt port vnet0 grp ff02::1:ff00:76 temp
dev ovirtmgmt port bond0.86 grp ff02::16 temp
dev ovirtmgmt port vnet1 grp ff02::1:ff00:77 temp
dev ovirtmgmt port bond0.86 grp ff02::1:ff00:def temp
dev ovirtmgmt port bond0.86 grp ff02::1:ffa1:40bf temp

Fixes: 08b202b67264 ("bridge br_multicast: IPv6 MLD support.")
Reported-by: Rik Theys <Rik.Theys@esat.kuleuven.be>
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Tested-by: Rik Theys <Rik.Theys@esat.kuleuven.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoARM: zynq: DT: Use the zynq binding with macb
Nathan Sullivan [Fri, 22 May 2015 14:22:11 +0000 (09:22 -0500)]
ARM: zynq: DT: Use the zynq binding with macb

Use the new zynq binding for macb ethernet, since it will disable half
duplex gigabit like the Zynq TRM says to do.

Signed-off-by: Nathan Sullivan <nathan.sullivan@ni.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: macb: Disable half duplex gigabit on Zynq
Nathan Sullivan [Fri, 22 May 2015 14:22:10 +0000 (09:22 -0500)]
net: macb: Disable half duplex gigabit on Zynq

According to the Zynq TRM, gigabit half duplex is not supported.  Add a
new cap and compatible string so Zynq can avoid advertising that mode.

Signed-off-by: Nathan Sullivan <nathan.sullivan@ni.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: macb: Document zynq gem dt binding
Nathan Sullivan [Fri, 22 May 2015 14:22:09 +0000 (09:22 -0500)]
net: macb: Document zynq gem dt binding

Signed-off-by: Nathan Sullivan <nathan.sullivan@ni.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv4: fill in table id when replacing a route
Michal Kubeček [Fri, 22 May 2015 11:40:09 +0000 (13:40 +0200)]
ipv4: fill in table id when replacing a route

When replacing an IPv4 route, tb_id member of the new fib_alias
structure is not set in the replace code path so that the new route is
ignored.

Fixes: 0ddcf43d5d4a ("ipv4: FIB Local/MAIN table collapse")
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocdc_ncm: Fix tx_bytes statistics
Bjørn Mork [Fri, 22 May 2015 11:15:22 +0000 (13:15 +0200)]
cdc_ncm: Fix tx_bytes statistics

The tx_curr_frame_payload field is u32. When we try to calculate a
small negative delta based on it, we end up with a positive integer
close to 2^32 instead.  So the tx_bytes pointer increases by about
2^32 for every transmitted frame.

Fix by calculating the delta as a signed long.

Cc: Ben Hutchings <ben.hutchings@codethink.co.uk>
Reported-by: Florian Bruhin <me@the-compiler.org>
Fixes: 7a1e890e2168 ("usbnet: Fix tx_bytes statistic running backward in cdc_ncm")
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Fri, 22 May 2015 18:25:45 +0000 (14:25 -0400)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contain Netfilter fixes for your net tree, they are:

1) Fix a race in nfnetlink_log and nfnetlink_queue that can lead to a crash.
   This problem is due to wrong order in the per-net registration and netlink
   socket events. Patch from Francesco Ruggeri.

2) Make sure that counters that userspace pass us are higher than 0 in all the
   x_tables frontends. Discovered via Trinity, patch from Dave Jones.

3) Revert a patch for br_netfilter to rely on the conntrack status bits. This
   breaks stateless IPv6 NAT transformations. Patch from Florian Westphal.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv4: Avoid crashing in ip_error
Eric W. Biederman [Fri, 22 May 2015 09:58:12 +0000 (04:58 -0500)]
ipv4: Avoid crashing in ip_error

ip_error does not check if in_dev is NULL before dereferencing it.

IThe following sequence of calls is possible:
CPU A                          CPU B
ip_rcv_finish
    ip_route_input_noref()
        ip_route_input_slow()
                               inetdev_destroy()
    dst_input()

With the result that a network device can be destroyed while processing
an input packet.

A crash was triggered with only unicast packets in flight, and
forwarding enabled on the only network device.   The error condition
was created by the removal of the network device.

As such it is likely the that error code was -EHOSTUNREACH, and the
action taken by ip_error (if in_dev had been accessible) would have
been to not increment any counters and to have tried and likely failed
to send an icmp error as the network device is going away.

Therefore handle this weird case by just dropping the packet if
!in_dev.  It will result in dropping the packet sooner, and will not
result in an actual change of behavior.

Fixes: 251da4130115b ("ipv4: Cache ip_error() routes even when not forwarding.")
Reported-by: Vittorio Gambaletta <linuxbugs@vittgam.net>
Tested-by: Vittorio Gambaletta <linuxbugs@vittgam.net>
Signed-off-by: Vittorio Gambaletta <linuxbugs@vittgam.net>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotcp: fix a potential deadlock in tcp_get_info()
Eric Dumazet [Fri, 22 May 2015 04:51:19 +0000 (21:51 -0700)]
tcp: fix a potential deadlock in tcp_get_info()

Taking socket spinlock in tcp_get_info() can deadlock, as
inet_diag_dump_icsk() holds the &hashinfo->ehash_locks[i],
while packet processing can use the reverse locking order.

We could avoid this locking for TCP_LISTEN states, but lockdep would
certainly get confused as all TCP sockets share same lockdep classes.

[  523.722504] ======================================================
[  523.728706] [ INFO: possible circular locking dependency detected ]
[  523.734990] 4.1.0-dbg-DEV #1676 Not tainted
[  523.739202] -------------------------------------------------------
[  523.745474] ss/18032 is trying to acquire lock:
[  523.750002]  (slock-AF_INET){+.-...}, at: [<ffffffff81669d44>] tcp_get_info+0x2c4/0x360
[  523.758129]
[  523.758129] but task is already holding lock:
[  523.763968]  (&(&hashinfo->ehash_locks[i])->rlock){+.-...}, at: [<ffffffff816bcb75>] inet_diag_dump_icsk+0x1d5/0x6c0
[  523.774661]
[  523.774661] which lock already depends on the new lock.
[  523.774661]
[  523.782850]
[  523.782850] the existing dependency chain (in reverse order) is:
[  523.790326]
-> #1 (&(&hashinfo->ehash_locks[i])->rlock){+.-...}:
[  523.796599]        [<ffffffff811126bb>] lock_acquire+0xbb/0x270
[  523.802565]        [<ffffffff816f5868>] _raw_spin_lock+0x38/0x50
[  523.808628]        [<ffffffff81665af8>] __inet_hash_nolisten+0x78/0x110
[  523.815273]        [<ffffffff816819db>] tcp_v4_syn_recv_sock+0x24b/0x350
[  523.822067]        [<ffffffff81684d41>] tcp_check_req+0x3c1/0x500
[  523.828199]        [<ffffffff81682d09>] tcp_v4_do_rcv+0x239/0x3d0
[  523.834331]        [<ffffffff816842fe>] tcp_v4_rcv+0xa8e/0xc10
[  523.840202]        [<ffffffff81658fa3>] ip_local_deliver_finish+0x133/0x3e0
[  523.847214]        [<ffffffff81659a9a>] ip_local_deliver+0xaa/0xc0
[  523.853440]        [<ffffffff816593b8>] ip_rcv_finish+0x168/0x5c0
[  523.859624]        [<ffffffff81659db7>] ip_rcv+0x307/0x420

Lets use u64_sync infrastructure instead. As a bonus, 64bit
arches get optimized, as these are nop for them.

Fixes: 0df48c26d841 ("tcp: add tcpi_bytes_acked to tcp_info")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>