firefly-linux-kernel-4.4.55.git
9 years agommc: sdhci-st: Add macros for register offsets and bitfields for mmcss glue regs
Peter Griffin [Fri, 10 Apr 2015 09:40:23 +0000 (10:40 +0100)]
mmc: sdhci-st: Add macros for register offsets and bitfields for mmcss glue regs

The stih407 family SoC's have additional glue registers in the flashSS which
are used to configure the Arasan controller.

This patch adds macros for the register offsets and bitfields which will be
used by subsequent patches to support stih407 family SoC's.

Signed-off-by: Peter Griffin <peter.griffin@linaro.org>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Acked-by: Maxime Coquelin <maxime.coquelin@st.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: Call mmc_of_parse()
Fabio Estevam [Wed, 8 Apr 2015 13:17:44 +0000 (10:17 -0300)]
mmc: sdhci-esdhc-imx: Call mmc_of_parse()

Currently it is not possible to use 'mmc-pwrseq-simple' property with this
driver because mmc_of_parse() is never called.

mmc_of_parse() calls mmc_pwrseq_alloc() that manages MMC power sequence and
allows passing GPIOs in the devicetree to properly power/reset the Wifi
chipset.

When using mmc_of_parse() we no longer need to have custom code to request
card-detect and write-protect pins, as this can now be handled by the mmc
core.

Tested on a imx6sl-warp board where BT/Wifi is functional and also on a
imx6q-sabresd.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Add locking around cmd11 timer
Doug Anderson [Fri, 3 Apr 2015 18:13:07 +0000 (11:13 -0700)]
mmc: dw_mmc: Add locking around cmd11 timer

It is possible for the cmd11 interrupt to fire and delete the
cmd11_timer before the cmd11_timer was actually setup.  Let's fix this
race by adding a few spinlocks.  Note that the race wasn't seen in
practice without adding some printk statements, but it still seems
wise to fix.

Fixes: 5c935165da79 ("mmc: dw_mmc: Add a timeout for sending CMD11")
Signed-off-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: Add a return in an unexpected cmd11 timeout
Doug Anderson [Fri, 3 Apr 2015 18:13:06 +0000 (11:13 -0700)]
mmc: dw_mmc: Add a return in an unexpected cmd11 timeout

If we get an unexpected cmd11 timeout we shouldn't actually treat it
as a timeout (not that we really expect to get an unexpected cmd11
timeout, but still).

Fixes: 5c935165da79 ("mmc: dw_mmc: Add a timeout for sending CMD11")
Reported-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-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: Increase cmd11 timeout to 500ms
Doug Anderson [Fri, 3 Apr 2015 18:13:05 +0000 (11:13 -0700)]
mmc: dw_mmc: Increase cmd11 timeout to 500ms

Although the cmd11 interrupt should come within 2ms, that's a very
short time.  Let's increase the timeout to be really sure that we
don't get an accidnetal timeout.  One case in particular this is
useful is if you've got a serial console and printk in just the right
places.  Under that scenario I've seen delays of up to 130ms before
the interrupt fired.

CMD11 is only sent during card insertion, so this extra timeout
shouldn't be terrible.

Fixes: 5c935165da79 ("mmc: dw_mmc: Add a timeout for sending CMD11")
Signed-off-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: fix fifo ordering in big endian
Ben Dooks [Wed, 25 Mar 2015 11:27:52 +0000 (11:27 +0000)]
mmc: dw_mmc: fix fifo ordering in big endian

The dw_mmc driver changes to make the IO accesors endian agnostic did not
take into account the fifo accesses do not need to be swapped. To fix this
add a mmci_fifo_read/write wrapper to allow these to be passed through the
IO without being swapped.

Since these are now specific functions, it would be easier just to store
the pointer to the fifo registers in the host block instead of the offset
to them. So change the host->data_offset to host->fifo_reg (which also
means we catch all the places this is read or written).

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: change idmac descriptor files to __le32
Ben Dooks [Wed, 25 Mar 2015 11:27:51 +0000 (11:27 +0000)]
mmc: dw_mmc: change idmac descriptor files to __le32

The dw_mmc driver does not take into account the processor may be in
big endian when writing the descriptors. Change the descriptors for
the 32bit IDMA to use __le32 and ensure they are suitably swapped
before writing.

Note, this has not been tested as the socfpga driver does not try to
use idma.

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: make IO accessors endian agnostic
Ben Dooks [Wed, 25 Mar 2015 11:27:50 +0000 (11:27 +0000)]
mmc: dw_mmc: make IO accessors endian agnostic

The dw_mmc driver does not use endian agnostic IO accessors, so fix
the use of __raw reads and writes to be the relaxed versions.

This fixes the dw_mmc driver initialisation on Altera socfpga in big endian.

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Convert the error field in struct mmc_command|data into an int
Ulf Hansson [Tue, 31 Mar 2015 10:41:55 +0000 (12:41 +0200)]
mmc: core: Convert the error field in struct mmc_command|data into an int

Everybody expects the error field in the struct mmc_command|data to be
and int but it's actually an unsigned int. Let's convert it into an int
to meet the expectations.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-of-arasan: Call OF parsing for MMC
Michal Simek [Tue, 7 Apr 2015 05:57:32 +0000 (07:57 +0200)]
mmc: sdhci-of-arasan: Call OF parsing for MMC

Also check MMC OF properties. The controller supports MMC too.

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pci: fix 64 BIT DMA quirks for rtsx
Micky Ching [Tue, 7 Apr 2015 03:32:01 +0000 (11:32 +0800)]
mmc: sdhci-pci: fix 64 BIT DMA quirks for rtsx

rts5250 chip failed handle 64 bit ADMA for address below 4G.
Add 64 BIT quirks to disable this feature.

Signed-off-by: Micky Ching <micky_ching@realsil.com.cn>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: Add support for marking hpi as broken through devicetree
Hans de Goede [Wed, 1 Apr 2015 15:26:23 +0000 (17:26 +0200)]
mmc: Add support for marking hpi as broken through devicetree

The eMMC on a tablet I've will stop working / communicating as soon as
the kernel executes:

mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
  EXT_CSD_HPI_MGMT, 1,
  card->ext_csd.generic_cmd6_time);

There seems to be no way to reliable identify eMMC-s which have a broken
hpi implementation, but at least for eMMC's which are soldered onto a board
we can work around this by specifying that hpi is broken in devicetree.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-tegra: convert to use GPIO descriptors
Mylene JOSSERAND [Mon, 30 Mar 2015 21:39:25 +0000 (23:39 +0200)]
mmc: sdhci-tegra: convert to use GPIO descriptors

Modify the driver to handle GPIOs using the descriptor API.

Signed-off-by: Mylene JOSSERAND <josserand.mylene@gmail.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Tested-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: use generic slot-gpio isr to manage card detect pin
Andreas Fenkart [Tue, 3 Mar 2015 12:28:18 +0000 (13:28 +0100)]
mmc: omap_hsmmc: use generic slot-gpio isr to manage card detect pin

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: simplify card/cover detect isr
Andreas Fenkart [Tue, 3 Mar 2015 12:28:17 +0000 (13:28 +0100)]
mmc: omap_hsmmc: simplify card/cover detect isr

strip the card dectet logic from cover detect isr and vice versa
the generic mmc_gpio_cd_irqt isr, uses 200ms on removal/insertion,
hence that should be fine here as well

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Don't try to enable the CD until we're sure we're not deferring
Doug Anderson [Wed, 11 Mar 2015 22:15:14 +0000 (15:15 -0700)]
mmc: dw_mmc: Don't try to enable the CD until we're sure we're not deferring

If dw_mci_init_slot() returns that we got a probe deferral then it may
leave slot->mmc as NULL.  That will cause dw_mci_enable_cd() to crash
when it calls mmc_gpio_get_cd().

Fix this by moving the call of dw_mci_enable_cd() until we're sure
that we're good.  Note that if we have more than one slot and one
defers (but the others don't) things won't work so well.  ...but
that's not a new thing and everyone has already agreed that multislot
support ought to be removed from dw_mmc eventually anyway since it is
unused, untested, and you can see several bugs like this by inspecting
the code.

Fixes: bcafaf5470f0 ("mmc: dw_mmc: Only enable CD after setup and only if needed")
Signed-off-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: Add a timeout for sending CMD11
Doug Anderson [Mon, 9 Mar 2015 23:18:21 +0000 (16:18 -0700)]
mmc: dw_mmc: Add a timeout for sending CMD11

In the Designware databook's description of the "Voltage Switch Normal
Scenario" it instructs us to set a timer and fail the voltage change
if we don't see the voltage change interrupt within 2ms.  Let's
implement that.  Without implementing this I have often been able to
reproduce a hang while trying to send CMD11 on an rk3288-based board
while constantly ejecting and inserting UHS cards.

Signed-off-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: enable card read threshold when mode is HS400
Jaehoon Chung [Thu, 5 Mar 2015 10:45:21 +0000 (19:45 +0900)]
mmc: dw_mmc: enable card read threshold when mode is HS400

Enable card-read-threshold, when eMMC mode is HS400.
Refer to f1d2736c8156 (mmc: dw_mmc: control card read threshold)

Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Enable runtime PM management of host devices
Ulf Hansson [Fri, 27 Mar 2015 11:15:15 +0000 (12:15 +0100)]
mmc: core: Enable runtime PM management of host devices

Currently those host drivers which have deployed runtime PM, deals with
the runtime PM reference counting entirely by themselves.

Since host drivers don't know when the core will send the next request
through some of the host_ops callbacks, they need to handle runtime PM
get/put between each an every request.

In quite many cases this has some negative effects, since it leads to a
high frequency of scheduled runtime PM suspend operations. That due to
the runtime PM reference count will normally reach zero in-between
every request.

We can decrease that frequency, by enabling the core to deal with
runtime PM reference counting of the host device. Since the core often
knows that it will send a seqeunce of requests, it makes sense for it
to keep a runtime PM reference count during these periods.

More exactly, let's increase the runtime PM reference count by invoking
pm_runtime_get_sync() from __mmc_claim_host(). Restore that action by
invoking pm_runtime_mark_last_busy() and pm_runtime_put_autosuspend()
in mmc_release_host(). In this way a runtime PM reference count will be
kept during the complete cycle of a claim -> release host.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Konstantin Dorfman <kdorfman@codeaurora.org>
9 years agoARM: OMAP2: HSMMC: explicit fields to declare cover/card detect pin
Andreas Fenkart [Fri, 20 Mar 2015 14:53:54 +0000 (15:53 +0100)]
ARM: OMAP2: HSMMC: explicit fields to declare cover/card detect pin

board-rx51 has no card detect pin in the mmc slot, but can detect that
the (cell-phone) cover has been removed and the card is accessible.
The semantics between cover/card detect differ, the gpio on the slot
informs you after the card has been removed, cover removal does not
necessarily mean that the card has been removed.
This means different code paths are necessary. To complete this we
also want different fields in the platform data for cover and card
detect. This separation is not pushed all the way down into struct
omap2_hsmmc_info which is used to initialize the platform data.
If we did that we had to go over all board files and set the new
gpio_cod pin to -EINVAL. If we forget one board or some out-of-tree
archicture forgets that the default '0' is used which is a valid pin
number.

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: Remove the ->enable|disable() callbacks
NeilBrown [Wed, 25 Mar 2015 21:43:37 +0000 (08:43 +1100)]
mmc: core: Remove the ->enable|disable() callbacks

These callbacks have been set to deprecated for some time. The last
user (omap_hsmmc) has moved away from using them, which thus enables
us to completely remove them.

Signed-off-by: NeilBrown <neil@brown.name>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: stop using ->enable|disable() callbacks
NeilBrown [Thu, 26 Mar 2015 01:18:23 +0000 (12:18 +1100)]
mmc: omap_hsmmc: stop using ->enable|disable() callbacks

The ->enable|disable() callbacks are only used to get and put runtime
PM references. Currently omap_hsmmc's ->set_ios() already does this
itself.

Other host drivers deals with runtime PM without using the
->enable|disable() callbacks and thus do the runtime PM reference
counting themselves. Apply that approach for omap_hsmmc as well and
then discard the ->enable|disable() callbacks.

Signed-off-by: NeilBrown <neil@brown.name>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-msm: Add support for vendor capabilities registers
Georgi Djakov [Mon, 23 Mar 2015 16:47:29 +0000 (18:47 +0200)]
mmc: sdhci-msm: Add support for vendor capabilities registers

Some versions of this controller do not advertise their 3.0v and
8bit bus-width support capabilities. It is required to explicitly
set these capabilities for the specific controller versions.

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: support voltage-range property
Sascha Hauer [Tue, 24 Mar 2015 13:45:04 +0000 (14:45 +0100)]
mmc: sdhci-esdhc-imx: support voltage-range property

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: mmci: Cascade EPROBE_DEFER from regulators.
Bjorn Andersson [Wed, 25 Mar 2015 01:39:49 +0000 (18:39 -0700)]
mmc: mmci: Cascade EPROBE_DEFER from regulators.

Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoMAINTAINERS: mmc: Add Jaehoon as co-maintainer for SDHCI SAMSUNG DRIVER
Ulf Hansson [Fri, 20 Mar 2015 09:15:29 +0000 (10:15 +0100)]
MAINTAINERS: mmc: Add Jaehoon as co-maintainer for SDHCI SAMSUNG DRIVER

Jaehoon has volunteered in maintaining the SDHCI SAMSUNG DRIVER so
add him.

Since we are updating this section let's also correct file path to
cover all files for sdhci-s3c.

Cc: Ben Dooks <ben-linux@fluff.org>
Cc: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: atmel-mci: use endian agnostic IO
Ben Dooks [Wed, 18 Mar 2015 15:53:11 +0000 (15:53 +0000)]
mmc: atmel-mci: use endian agnostic IO

Change the __raw IO functions to endian agnostic relaxed ones to allow
the driver to function on big endian ARM systems.

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Acked-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: constify of_device_id array
Fabian Frederick [Mon, 16 Mar 2015 19:59:07 +0000 (20:59 +0100)]
mmc: constify of_device_id array

of_device_id is always used as const.
(See driver.of_match_table and open firmware functions)

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: core: fold mmc_set_bus_width calls into sdio_enable_4bit_bus.
NeilBrown [Tue, 24 Feb 2015 02:42:23 +0000 (13:42 +1100)]
mmc: core: fold mmc_set_bus_width calls into sdio_enable_4bit_bus.

Every call to sdio_enable_4bit_bus is followed (on success) by a call
to mmc_set_bus_width().

To simplify the code, include those calls directly in
sdio_enable_4bit_bus().

Signed-off-by: NeilBrown <neil@brown.name>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoMAINTAINERS: mmc: Cleanup MMC/SD/SDIO section and SDHCI driver section
Ulf Hansson [Fri, 13 Mar 2015 12:39:03 +0000 (13:39 +0100)]
MAINTAINERS: mmc: Cleanup MMC/SD/SDIO section and SDHCI driver section

As Chris Ball has moved on to other assignments, he's no longer able to
help me maintain MMC. Let's remove him from the MMC sections in
MAINTAINERS and add him to CREDIT file.

This also affects the SDHCI DRIVER section, since its state now becomes
orphan.

Cc: Chris Ball <chris@printf.net>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Chris Ball <chris@printf.net>
9 years agoMAINTAINERS: mmc: Remove the SDHCI-OF section
Ulf Hansson [Fri, 13 Mar 2015 12:25:25 +0000 (13:25 +0100)]
MAINTAINERS: mmc: Remove the SDHCI-OF section

Anton told me that he isn't maintaing the SDHCI-OF parts anymore, so
let's remove him from this section to avoid confusion.

Morover, since the SDHCI-OF section overlaps with the SDHCI DRIVER
section, let's just remove it completely.

Cc: Anton Vorontsov <anton@enomsg.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sunxi: add MMC_CAP_SDIO_IRQ capability
Hans de Goede [Tue, 10 Mar 2015 15:36:36 +0000 (16:36 +0100)]
mmc: sunxi: add MMC_CAP_SDIO_IRQ capability

When the sunxi mmc-controller code was initially merged MMC_CAP_SDIO_IRQ was
not added to the host caps because of issues with some sdio wifi modules.

It turns out that these issues have nothing to do with using sdio-irq support,
they also happen with oob interrupts. Since the hardware supports sdio-irq
everywhere, and since the one reason to not claim the capability is gone,
add MMC_CAP_SDIO_IRQ to the default host caps.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-iproc: fix oops in sdhci_iproc_writew
Dmitry Torokhov [Fri, 13 Mar 2015 01:11:01 +0000 (18:11 -0700)]
mmc: sdhci-iproc: fix oops in sdhci_iproc_writew

The driver co-allocates sdhci_iproc_host with sdhci_pltfm_host and so to
access it we need to use sdhci_pltfm_priv() and not pltfm_host->priv.

Signed-off-by: Dmitry Torokhov <dtor@chromium.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: fix card presence logic in sdhci_request function
Scott Branden [Tue, 10 Mar 2015 18:35:10 +0000 (11:35 -0700)]
mmc: sdhci: fix card presence logic in sdhci_request function

The sdhci_request function should consider a non-removable device
always present.
Call the correct logic already available in sdhci_do_get_cd function.

This fixes some logic paths where MMC requests are being made to
non-removable devices that do not have the card detect pin connected
on the hardware as it is non-removable.

Signed-off-by: Scott Branden <sbranden@broadcom.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-spear: Remove exported header
Ulf Hansson [Wed, 4 Mar 2015 15:30:07 +0000 (16:30 +0100)]
mmc: sdhci-spear: Remove exported header

Move the member for card_int_gpio into the struct spear_sdhci.

In this way we eliminate the last user of the struct sdhci_plat_data,
which enables us to remove the exported header for sdhci-spear.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-spear: Simplify by adding build dependency to CONFIG_OF
Ulf Hansson [Wed, 4 Mar 2015 13:57:44 +0000 (14:57 +0100)]
mmc: sdhci-spear: Simplify by adding build dependency to CONFIG_OF

This driver is used on SoCs which are using CONFIG_OF. By adding a
compile dependency in the Kconfig, it enables us to simplify some code.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: Remove the sdhci exported header file
Ulf Hansson [Wed, 4 Mar 2015 09:19:14 +0000 (10:19 +0100)]
mmc: sdhci: Remove the sdhci exported header file

Since there no users of the struct sdhci_host, but the shdci host
drivers themselves, let's move the definition of it to the local sdhci
header.

The exported sdhci header then becomes empty, so let's remove it.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-pltfm: remove the unneeded check of disabled device
Kevin Hao [Wed, 4 Feb 2015 06:33:53 +0000 (14:33 +0800)]
mmc: sdhci-pltfm: remove the unneeded check of disabled device

Since commit cd1e65044d44 ("of/device: Don't register disabled
devices"), the disabled device will not be registered at all. So
we don't need to do the check again in the platform device driver.

And the check in the current code is useless even if we really
run into a disabled device. In this case, it just doesn't parse
the dtb for the infos such as quirks or clock, but it will continue
to try to init the disabled device after that check. So just remove it.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: tegra: Optimize write_w path for tegra114 and later
Rhyland Klein [Wed, 11 Feb 2015 17:55:51 +0000 (12:55 -0500)]
mmc: tegra: Optimize write_w path for tegra114 and later

Setup a different set of sdhci_ops for tegra114 and later so that
the write_w callback is only used on tegra114. This allows us to
remove the NVQUIRK_SHADOW_XFER_MODE_REG and simply the logic
in tegra_sdhci_writew.

Suggested-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Rhyland Klein <rklein@nvidia.com>
Acked-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: Add exclusion between cmd and interrupt
Kouichi Tomita [Sun, 15 Feb 2015 14:46:46 +0000 (23:46 +0900)]
mmc: sh_mmcif: Add exclusion between cmd and interrupt

A command end interrupt should not be processed between command issue
and setting of wait_for flag. It expects already the flag to be set.
Therefore the exclusive control was added.

Signed-off-by: Kouichi Tomita <kouichi.tomita.yn@renesas.com>
Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sh_mmcif: Move dev_err() of mmcif_timeout_work()
Kouichi Tomita [Sun, 15 Feb 2015 14:46:47 +0000 (23:46 +0900)]
mmc: sh_mmcif: Move dev_err() of mmcif_timeout_work()

If interruption of command already occurred, mrq pointer in dev_err()
would refer to NULL, because the host-state is changed to STATE_IDLE
and mrq pointer is changed to NULL by interrupt handler.
Therefore dev_err is moved after checking STATE_IDLE.

Signed-off-by: Kouichi Tomita <kouichi.tomita.yn@renesas.com>
Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: Fix hardware dependencies for sdhci-pxav3
Jean Delvare [Mon, 23 Feb 2015 10:30:40 +0000 (11:30 +0100)]
mmc: Fix hardware dependencies for sdhci-pxav3

It was brought to my attention that the sdhci-pxav3 driver is needed
on a few more ARM machines than I initially thought. Add the missing
architectures to the dependency list.

Credits to Peter Robinson for noticing my mistake and reporting.

Reported-by: Peter Robinson <pbrobinson@gmail.com>
Signed-off-by: Jean Delvare <jdelvare@suse.de>
Cc: Chris Ball <chris@printf.net>
Cc: Ulf Hansson <ulf.hansson@linaro.org>
Cc: Eric Miao <eric.y.miao@gmail.com>
Cc: Haojian Zhuang <haojian.zhuang@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: pwrseq: simplify alloc/free hooks
Alexandre Courbot [Thu, 12 Feb 2015 04:36:11 +0000 (13:36 +0900)]
mmc: pwrseq: simplify alloc/free hooks

The alloc() and free() hooks required each pwrseq implementation to set
host->pwrseq themselves. This is error-prone and could be done at a
higher level if alloc() was changed to return a pointer to a struct
mmc_pwrseq instead of an error code.

This patch performs this change and moves the burden of maintaining
host->pwrseq from the power sequence hooks to the pwrseq code.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sunxi: avoid invalid pointer calculation
Arnd Bergmann [Tue, 24 Feb 2015 09:47:27 +0000 (10:47 +0100)]
mmc: sunxi: avoid invalid pointer calculation

The sunxi mmc driver tries to calculate a dma address by using pointer
arithmetic, which causes a warning when dma_addr_t is wider than a pointer:

drivers/mmc/host/sunxi-mmc.c: In function 'sunxi_mmc_init_idma_des':
drivers/mmc/host/sunxi-mmc.c:296:35: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
  struct sunxi_idma_des *pdes_pa = (struct sunxi_idma_des *)host->sg_dma;
                                   ^

To avoid this warning and to simplify the logic, this changes
the code to avoid the cast and calculate the correct address
manually. The behavior should be unchanged.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: David Lanzendörfer <david.lanzendoerfer@o2s.ch>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: kconfig: replace PPC_OF with PPC
Kevin Hao [Thu, 26 Feb 2015 12:08:21 +0000 (20:08 +0800)]
mmc: kconfig: replace PPC_OF with PPC

The PPC_OF is a ppc specific option which is used to mean that the
firmware device tree access functions are available. Since all the
ppc platforms have a device tree, it is aways set to 'y' for ppc.
So it makes no sense to keep a such option in the current kernel.
Replace it with PPC.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-iproc: add IPROC SDHCI driver
Scott Branden [Tue, 10 Feb 2015 00:06:30 +0000 (16:06 -0800)]
mmc: sdhci-iproc: add IPROC SDHCI driver

Add IPROC SDHCI driver for IPROC family of Broadcom devices.

Acked-by: Ray Jui <rjui@broadcom.com>
Signed-off-by: Corneliu Doban <cdoban@broadcom.com>
Signed-off-by: Scott Branden <sbranden@broadcom.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-iproc: add device tree bindings
Scott Branden [Tue, 10 Feb 2015 00:06:31 +0000 (16:06 -0800)]
mmc: sdhci-iproc: add device tree bindings

Add device tree binding documentation for IPROC SDHCI driver.

Acked-by: Ray Jui <rjui@broadcom.com>
Signed-off-by: Corneliu Doban <cdoban@broadcom.com>
Signed-off-by: Scott Branden <sbranden@broadcom.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: do not set AUTO_CMD12 for multi-block CMD53
Corneliu Doban [Tue, 10 Feb 2015 00:06:29 +0000 (16:06 -0800)]
mmc: sdhci: do not set AUTO_CMD12 for multi-block CMD53

For CMD53 in block mode, the host does not need to stop the transfer,
as it stops when the block count (present in CMD53) is reached.

Signed-off-by: Corneliu Doban <cdoban@broadcom.com>
Signed-off-by: Scott Branden <sbranden@broadcom.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: add quirk for ACMD23 broken
Scott Branden [Tue, 10 Feb 2015 00:06:28 +0000 (16:06 -0800)]
mmc: sdhci: add quirk for ACMD23 broken

Add quirk to handle broken auto-CMD23.
Some controllers do not respond after the first auto-CMD23 is issued.

This allows CMD23 to still work (mandatory for the faster UHS-I mode)
rather than disabling CMD23 entirely via SDHCI_QUIRK2_HOST_NO_CMD23.

Signed-off by: Corneliu Doban <cdoban@broadcom.com>
Signed-off-by: Scott Branden <sbranden@broadcom.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-esdhc-imx: silence a false curly braces warning
Dan Carpenter [Thu, 26 Feb 2015 20:37:55 +0000 (23:37 +0300)]
mmc: sdhci-esdhc-imx: silence a false curly braces warning

Static checkers suggest that probably we intended to put curly braces
around the writel() to make it part of the else path.  But, I think
actually the indenting is off and the code works fine as is.

The stray tab was introduced in 0322191e6298 ('mmc: sdhci-esdhc-imx: add
sd3.0 SDR clock tuning support')

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Dong Aisheng <b29396@freescale.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: tmio: Remove bogus un-initialization in tmio_mmc_host_free()
Geert Uytterhoeven [Wed, 18 Feb 2015 16:34:59 +0000 (17:34 +0100)]
mmc: tmio: Remove bogus un-initialization in tmio_mmc_host_free()

If CONFIG_DEBUG_SLAB=y:

    sh_mobile_sdhi ee100000.sd: Got CD GPIO
    sh_mobile_sdhi ee100000.sd: Got WP GPIO
    platform ee100000.sd: Driver sh_mobile_sdhi requests probe deferral
    ...
    Slab corruption (Not tainted): kmalloc-1024 start=ed8b3c00, len=1024
    2d0: 00 00 00 00 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  ....kkkkkkkkkkkk
    Prev obj: start=ed8b3800, len=1024
    000: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk
    010: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Struct tmio_mmc_host is embedded inside struct mmc_host, and thus is
freed by the call to mmc_free_host(). Hence it must not be written to
afterwards, as that will corrupt freed (and perhaps already reused)
memory.

Fixes: 94b110aff8679b14 ("mmc: tmio: add tmio_mmc_host_alloc/free()")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: exynos: dw_mci_exynos_prepare_hs400_tuning() can be static
Wu Fengguang [Thu, 5 Mar 2015 10:02:54 +0000 (18:02 +0800)]
mmc: dw_mmc: exynos: dw_mci_exynos_prepare_hs400_tuning() can be static

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: add hibernation support
Russ Dill [Fri, 27 Feb 2015 11:24:34 +0000 (13:24 +0200)]
mmc: omap_hsmmc: add hibernation support

Setting a dev_pm_ops suspend/resume pair but not a set of
hibernation functions means those pm functions will not be
called upon hibernation.
Fix this by using SET_SYSTEM_SLEEP_PM_OPS, which appropriately
assigns the suspend and hibernation handlers and move
omap_hsmmc_x callbacks under CONFIG_PM_SLEEP to avoid build warnings.

Signed-off-by: Russ Dill <Russ.Dill@ti.com>
[Grygorii.Strashko@linaro.org: rebased on top of K4.0]
Signed-off-by: Grygorii Strashko <grygorii.strashko@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: use distinctive code paths for cover / card detect logic
Andreas Fenkart [Tue, 3 Mar 2015 12:28:15 +0000 (13:28 +0100)]
mmc: omap_hsmmc: use distinctive code paths for cover / card detect logic

Mobile phones (some) have no card detect pin, but can detect if the
cover is removed. The purpose is the same; detect if card is being
added/removed, but the details differ.
When the cover is removed, it does not mean the card is gone. But it
might, since it is accessible now. It's like a warning. All the driver
does is to limit write access to the card, see protect_card flag.
In contrast, card detect notifies us after the fact, e.g.
card is gone, card is inserted. We can't take precautions, but we can
rely on those events, -- the card is really gone, or do scan the card.
To summarize there is not much code sharing between cover and card
detect, it only increases confusion. By splitting, both will be
simplified in a followup patch.

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: use slot-gpio functions to manage read-only pin directly
Andreas Fenkart [Tue, 3 Mar 2015 12:28:14 +0000 (13:28 +0100)]
mmc: omap_hsmmc: use slot-gpio functions to manage read-only pin directly

The indirection via omap_hsmmc_get_ro and omap_hsmmc_get_wp is
redundant. Also dropped setting gpio_wp to EINVAL since platform date
is read-only
Untested: no device with ro pin was available, but change is fairly
simple

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: omap_hsmmc: remove unused fields from struct omap_hsmmc_host
Andreas Fenkart [Tue, 3 Mar 2015 12:28:13 +0000 (13:28 +0100)]
mmc: omap_hsmmc: remove unused fields from struct omap_hsmmc_host

addon to: 09108968b7b72b6083a3bfc8f8259a74ed57255e
    mmc: omap_hsmmc: remove prepare/complete system suspend support

Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sunxi: Use devm_reset_control_get_optional() for reset control
Chen-Yu Tsai [Tue, 3 Mar 2015 01:44:40 +0000 (09:44 +0800)]
mmc: sunxi: Use devm_reset_control_get_optional() for reset control

The reset control for the sunxi mmc controller is optional. Some
newer platforms (sun6i, sun8i, sun9i) have it, while older ones
(sun4i, sun5i, sun7i) don't.

Use the properly stubbed _optional version so the driver does not
fail to compile when RESET_CONTROLLER=n.

This patch also adds a check for deferred probing on the reset
control.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Cc: <stable@vger.kernel.org> # 3.16+
Acked-by: David Lanzendörfer <david.lanzendoerfer@o2s.ch>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: set the .remove to sdhci_pltfm_unregister()
Kevin Hao [Fri, 27 Feb 2015 07:47:31 +0000 (15:47 +0800)]
mmc: sdhci: set the .remove to sdhci_pltfm_unregister()

In these drivers, the driver specific .remove function just a simple
wrapper of function sdhci_pltfm_unregister(). So remove these wrappers
and just set .remove to sdhci_pltfm_unregister().

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci: disable the clock in sdhci_pltfm_unregister()
Kevin Hao [Fri, 27 Feb 2015 07:47:30 +0000 (15:47 +0800)]
mmc: sdhci: disable the clock in sdhci_pltfm_unregister()

So we can avoid to sprinkle the clk_disable_unprepare() in many
drivers.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-bcm-kona: kill the "external_clk" member in driver private struct
Kevin Hao [Fri, 27 Feb 2015 07:47:29 +0000 (15:47 +0800)]
mmc: sdhci-bcm-kona: kill the "external_clk" member in driver private struct

Actually we can use the "clk" in the struct sdhci_pltfm_host. Also
change the "external clock" to "core clock" and kill two redundant
private functions in this driver as suggested by Ray Jui.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Reviewed-by: Ray Jui <rjui@broadcom.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-sirf: kill the "clk" member in driver private struct
Kevin Hao [Fri, 27 Feb 2015 07:47:28 +0000 (15:47 +0800)]
mmc: sdhci-sirf: kill the "clk" member in driver private struct

Actually we can use the "clk" in the struct sdhci_pltfm_host.
With this change we can also kill the private function for get
max clock in this driver.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: tegra: use devm help functions to get the clk and gpio
Kevin Hao [Fri, 27 Feb 2015 07:47:27 +0000 (15:47 +0800)]
mmc: tegra: use devm help functions to get the clk and gpio

Simplify the error and remove path.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Acked-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-dove: kill the driver specific private struct
Kevin Hao [Fri, 27 Feb 2015 07:47:26 +0000 (15:47 +0800)]
mmc: sdhci-dove: kill the driver specific private struct

There is only one "clk" member in this driver specific private struct.
Actually we can use the "clk" member in the struct sdhci_pltfm_host,
and then kill this struct completely.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sdhci-dove: remove the unneeded error check
Kevin Hao [Fri, 27 Feb 2015 07:47:25 +0000 (15:47 +0800)]
mmc: sdhci-dove: remove the unneeded error check

The function clk_disable_unprepare() already take care of either error
or null cases.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: sirf: update sdhci_sirf_execute_tuning procedure
weijun yang [Sun, 15 Feb 2015 15:43:51 +0000 (23:43 +0800)]
mmc: sirf: update sdhci_sirf_execute_tuning procedure

For the original tuning code, delay value is set to SD Bus Clock Delay
Register (SD_CLK_DELAY_SETTING) as (val | (Val << 7) | (val << 16)),
which means CLK_DELAY_IN1, CLK_DELAY_IN2 and CLK_DELAY_OUT are the
same and with 128 steps. This is doubtful. In CSR design specification
documents CS-304575-DR-3H, this issue is clarified, the delay[13:0] in
SD_CLK_DELAY_SETTING is simplied to the concatenation of {CLK_DELAY_IN2,
CLK_DELAY_IN1}.
Besides, for CMD19 tuning, no need to set CLK_DELAY_OUT([22,16]
of SD_CLK_DELAY_SETTING).

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: dw_mmc: Don't crash if we get an interrupt before slot has initted
Doug Anderson [Wed, 25 Feb 2015 18:11:52 +0000 (10:11 -0800)]
mmc: dw_mmc: Don't crash if we get an interrupt before slot has initted

It's unlikely that this is really needed on any single-slot systems
where we disable card detects until the end of probe, but it still
seems safer to check to make sure that a slot has been initted before
we try to dereference it to find the SDIO interrupt mask.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Tested-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Only enable CD after setup and only if needed
Doug Anderson [Wed, 25 Feb 2015 18:11:51 +0000 (10:11 -0800)]
mmc: dw_mmc: Only enable CD after setup and only if needed

We really don't want to get a card detect interrupt during probe time
since it can confuse things.  Let's disable the card detect interrupt
until we're in a really good place: the end of probe.  Let's also
simply avoid enabling the card detect interrupt if it's not used.

It appears that (at least on rk3288) when vqmmc is turned on it can
cause a bogus "card detect" interrupt.  That meant that we were
getting a predictable card detect interrupt while we were in
mmc_add_host().  On the version of the kernel I'm working with at
least (3.14), this is not a great time to get a card detect interrupt
since I think that we don't grab all the needed locks in
mmc_add_host() and children.  I put stack dumps in dw_mci_setup_bus()
and found that I could see two distinct stack crawls that looked like:

Caller one:
* dw_mci_setup_bus
* dw_mci_set_ios
* mmc_power_up
* mmc_start_host
* mmc_add_host

Caller two:
* dw_mci_setup_bus
* dw_mci_set_ios
* mmc_set_chip_select
* mmc_go_idle
* mmc_rescan
* process_one_work
* worker_thread
* kthread

Signed-off-by: Doug Anderson <dianders@chromium.org>
Tested-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Don't start commands while busy
Doug Anderson [Fri, 20 Feb 2015 20:31:56 +0000 (12:31 -0800)]
mmc: dw_mmc: Don't start commands while busy

We've seen problems on some WiFi modules where we seem to send a CMD53
(which requires the data lines) while the module is asserting busy.
We shouldn't do that.

The Designware Databook says that before issuing a new data transfer
command we should check for busy, so that's what we'll do.

We'll leverage the existing dw_mmc knowledge about whether it should
wait for the previous command to finish to know whether we should
check for busy before sending the command.  This means we won't end up
incorrectly waiting for things like CMD52 (SDIO) or CMD13 (SD) which
don't use the data line.

Note that this also has the advantage of making sure that we don't
change the clock while the card is busy, too.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Tested-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Give a good reset after we give power
Doug Anderson [Fri, 20 Feb 2015 18:57:19 +0000 (10:57 -0800)]
mmc: dw_mmc: Give a good reset after we give power

We should give dw_mmc a good reset after we apply power.  On some
boards vqmmc may actually be connected to the IP block in the SoC so
it's good to reset after power comes in.

Without this we sometimes see failures enumerating cards on rk3288.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Tested-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: Make sure we only adjust the clock when power is on
Doug Anderson [Fri, 20 Feb 2015 18:57:18 +0000 (10:57 -0800)]
mmc: dw_mmc: Make sure we only adjust the clock when power is on

It appears that we can confuse things if we try to turn on the MMC
clock when the power is off.  Adjust is so that we turn the clock on
(using dw_mci_setup_bus) after power is all the way on and we turn the
clock off before the power goes off.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Tested-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: fix mmc_test by not sending abort for DRTO/EBE errors
addy ke [Fri, 20 Feb 2015 02:55:25 +0000 (10:55 +0800)]
mmc: dw_mmc: fix mmc_test by not sending abort for DRTO/EBE errors

The STOP command can terminate a data transfer between a memory card and
mmc controller.

As show in Synopsys DesignWare Cores Mobile Storage Host Databook:
Data timeout and Data end-bit error will terminate further data transfer
by mmc controller. So we should not send abort command to terminate a
data transfer again if we got DRTO and EBE interrupt.

After this patch, all mmc_test cases can pass on RK3288-Pink2 board.

Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
Reviewed-by: Doug Anderson <dianders@chromium.org>
Tested-by: Doug Anderson <dianders@chromium.org>
Tested-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agommc: dw_mmc: rockchip: add support MMC_CAP_RUNTIME_RESUME capability
addy ke [Fri, 20 Feb 2015 02:37:40 +0000 (10:37 +0800)]
mmc: dw_mmc: rockchip: add support MMC_CAP_RUNTIME_RESUME capability

To support HS200 and UHS mode, mmc core will call init_card() to
execute tuning:
- sdio: init_card can be executed at runtime resume.
- sd and mmc: init_card can be executed at resume or runtime resume,
  which depends on MMC_CAP_RUNTIME_RESUME capability.

On rk3288 SoC, host will get DRTO interrupt when host send command
to read tuning data. This will spend more than 111ms:
drto_ms = drto_clks * 1000 / bus_hz = 111ms.

And the total tuning time will be more than 400ms.

So we should add MMC_CAP_RUNTIME_RESUME capability to execute tuning
at runtime resume. Only if we do so, can we pass resume test.

Reviewed-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Addy Ke <addy.ke@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: exynos: Support eMMC's HS400 mode
Seungwon Jeon [Thu, 29 Jan 2015 02:41:57 +0000 (08:11 +0530)]
mmc: dw_mmc: exynos: Support eMMC's HS400 mode

Implements HS400 mode support for exynos host driver.
This also include some updates as new mode is added.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Alim Akhtar <alim.akhtar@samsung.com>
[Alim: addressed review comments]
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
9 years agoLinux 4.0-rc5
Linus Torvalds [Sun, 22 Mar 2015 23:50:21 +0000 (16:50 -0700)]
Linux 4.0-rc5

9 years agoMerge tag 'md/4.0-rc4-fix' of git://neil.brown.name/md
Linus Torvalds [Sun, 22 Mar 2015 23:38:19 +0000 (16:38 -0700)]
Merge tag 'md/4.0-rc4-fix' of git://neil.brown.name/md

Pull bugfix for md from Neil Brown:
 "One fix for md in 4.0-rc4

  Regression in recent patch causes crash on error path"

* tag 'md/4.0-rc4-fix' of git://neil.brown.name/md:
  md: fix problems with freeing private data after ->run failure.

9 years agoMerge tag 'driver-core-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 22 Mar 2015 19:07:47 +0000 (12:07 -0700)]
Merge tag 'driver-core-4.0-rc5' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core fixes from Greg KH:
 "Here are two bugfixes for things reported.  One regression in kernfs,
  and another issue fixed in the LZ4 code that was fixed in the
  "upstream" codebase that solves a reported kernel crash

  Both have been in linux-next for a while"

* tag 'driver-core-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  LZ4 : fix the data abort issue
  kernfs: handle poll correctly on 'direct_read' files.

9 years agoMerge tag 'char-misc-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sun, 22 Mar 2015 19:03:14 +0000 (12:03 -0700)]
Merge tag 'char-misc-4.0-rc5' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc fixes from Greg KH:
 "Here are three fixes for 4.0-rc5 that revert 3 PCMCIA patches that
  were merged in 4.0-rc1 that cause regressions.  So let's revert them
  for now and they will be reworked and resent sometime in the future.

  All have been tested in linux-next for a while"

* tag 'char-misc-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  Revert "pcmcia: add a new resource manager for non ISA systems"
  Revert "pcmcia: fix incorrect bracketing on a test"
  Revert "pcmcia: add missing include for new pci resource handler"

9 years agoMerge tag 'staging-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sun, 22 Mar 2015 18:59:02 +0000 (11:59 -0700)]
Merge tag 'staging-4.0-rc5' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fixes from Greg KH:
 "Here are four small staging driver fixes, all for the vt6656 and
  vt6655 drivers, that resolve some reported issues with them.

  All of these patches have been in linux next for a while"

* tag 'staging-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  vt6655: Fix late setting of byRFType.
  vt6655: RFbSetPower fix missing rate RATE_12M
  staging: vt6656: vnt_rf_setpower: fix missing rate RATE_12M
  staging: vt6655: vnt_tx_packet fix dma_idx selection.

9 years agoMerge tag 'tty-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sun, 22 Mar 2015 18:54:29 +0000 (11:54 -0700)]
Merge tag 'tty-4.0-rc5' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial driver fix from Greg KH:
 "Here's a single 8250 serial driver that fixes a reported deadlock with
  the serial console and the tty driver.

  It's been in linux-next for a while now"

* tag 'tty-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: 8250_dw: Fix deadlock in LCR workaround

9 years agoMerge tag 'usb-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 22 Mar 2015 18:33:55 +0000 (11:33 -0700)]
Merge tag 'usb-4.0-rc5' of git://git./linux/kernel/git/gregkh/usb

Pull USB / PHY driver fixes from Greg KH:
 "Here's a number of USB and PHY driver fixes for 4.0-rc5.

  The largest thing here is a revert of a gadget function driver patch
  that removes 500 lines of code.  Other than that, it's a number of
  reported bugs fixes and new quirk/id entries.

  All have been in linux-next for a while"

* tag 'usb-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (33 commits)
  usb: common: otg-fsm: only signal connect after switching to peripheral
  uas: Add US_FL_NO_ATA_1X for Initio Corporation controllers / devices
  USB: ehci-atmel: rework clk handling
  MAINTAINERS: add entry for USB OTG FSM
  usb: chipidea: otg: add a_alt_hnp_support response for B device
  phy: omap-usb2: Fix missing clk_prepare call when using old dt name
  phy: ti/omap: Fix modalias
  phy: core: Fixup return value of phy_exit when !pm_runtime_enabled
  phy: miphy28lp: Convert to devm_kcalloc and fix wrong sizof
  phy: miphy365x: Convert to devm_kcalloc and fix wrong sizeof
  phy: twl4030-usb: Remove redundant assignment for twl->linkstat
  phy: exynos5-usbdrd: Fix off-by-one valid value checking for args->args[0]
  phy: Find the right match in devm_phy_destroy()
  phy: rockchip-usb: Fixup rockchip_usb_phy_power_on failure path
  phy: ti-pipe3: Simplify ti_pipe3_dpll_wait_lock implementation
  phy: samsung-usb2: Remove NULL terminating entry from phys array
  phy: hix5hd2-sata: Check return value of platform_get_resource
  phy: exynos-dp-video: Kill exynos_dp_video_phy_pwr_isol function
  Revert "usb: gadget: zero: Add support for interrupt EP"
  Revert "xhci: Clear the host side toggle manually when endpoint is 'soft reset'"
  ...

9 years agoMerge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sat, 21 Mar 2015 20:05:37 +0000 (13:05 -0700)]
Merge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma

Pull slave dmaengine fixes from Vinod Koul:
 "Four fixes for dw, pl08x, imx-sdma and at_hdmac driver.  Nothing
  unusual here, simple fixes to these drivers"

* 'fixes' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: pl08x: Define capabilities for generic capabilities reporting
  dmaengine: dw: append MODULE_ALIAS for platform driver
  dmaengine: imx-sdma: switch to dynamic context mode after script loaded
  dmaengine: at_hdmac: Fix calculation of the residual bytes

9 years agoMerge tag 'pm+acpi-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Sat, 21 Mar 2015 19:51:36 +0000 (12:51 -0700)]
Merge tag 'pm+acpi-4.0-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management and ACPI fixes from Rafael Wysocki:
 "These are fixes for recent regressions (PCI/ACPI resources and at91
  RTC locking), a stable-candidate powercap RAPL driver fix and two ARM
  cpuidle fixes (one stable-candidate too).

  Specifics:

   - Revert a recent PCI commit related to IRQ resources management that
     introduced a regression for drivers attempting to bind to devices
     whose previous drivers did not balance pci_enable_device() and
     pci_disable_device() as expected (Rafael J Wysocki).

   - Fix a deadlock in at91_rtc_interrupt() introduced by a typo in a
     recent commit related to wakeup interrupt handling (Dan Carpenter).

   - Allow the power capping RAPL (Running-Average Power Limit) driver
     to use different energy units for domains within one CPU package
     which is necessary to handle Intel Haswell EP processors correctly
     (Jacob Pan).

   - Improve the cpuidle mvebu driver's handling of Armada XP SoCs by
     updating the target residency and exit latency numbers for those
     chips (Sebastien Rannou).

   - Prevent the cpuidle mvebu driver from calling cpu_pm_enter() twice
     in a row before cpu_pm_exit() is called on the same CPU which
     breaks the core's assumptions regarding the usage of those
     functions (Gregory Clement)"

* tag 'pm+acpi-4.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  Revert "x86/PCI: Refine the way to release PCI IRQ resources"
  rtc: at91rm9200: double locking bug in at91_rtc_interrupt()
  powercap / RAPL: handle domains with different energy units
  cpuidle: mvebu: Update cpuidle thresholds for Armada XP SOCs
  cpuidle: mvebu: Fix the CPU PM notifier usage

9 years agoMerge git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 21 Mar 2015 19:41:50 +0000 (12:41 -0700)]
Merge git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "A bunch of fixes across drivers:

  radeon:
     disable two ended allocation for now, it breaks some stuff

  amdkfd:
     misc fixes

  nouveau:
     fix irq loop problem, add basic support for GM206 (new hw)

  i915:
     fix some WARNs people were seeing

  exynos:
     fix some iommu interactions causing boot failures"

* git://people.freedesktop.org/~airlied/linux:
  drm/radeon: drop ttm two ended allocation
  drm/exynos: fix the initialization order in FIMD
  drm/exynos: fix typo config name correctly.
  drm/exynos: Check for NULL dereference of crtc
  drm/exynos: IS_ERR() vs NULL bug
  drm/exynos: remove unused files
  drm/i915: Make sure the primary plane is enabled before reading out the fb state
  drm/nouveau/bios: fix i2c table parsing for dcb 4.1
  drm/nouveau/device/gm100: Basic GM206 bring up (as copy of GM204)
  drm/nouveau/device: post write to NV_PMC_BOOT_1 when flipping endian switch
  drm/nouveau/gr/gf100: fix some accidental or'ing of buffer addresses
  drm/nouveau/fifo/nv04: remove the loop from the interrupt handler
  drm/radeon: Changing number of compute pipe lines
  drm/amdkfd: Fix SDMA queue init. in non-HWS mode
  drm/amdkfd: destroy mqd when destroying kernel queue
  drm/i915: Ensure plane->state->fb stays in sync with plane->fb

9 years agoMerge tag 'devicetree-fixes-for-4.0-part2' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Sat, 21 Mar 2015 19:33:01 +0000 (12:33 -0700)]
Merge tag 'devicetree-fixes-for-4.0-part2' of git://git./linux/kernel/git/robh/linux

Pull more DeviceTree fixes vfom Rob Herring:

 - revert setting stdout-path as preferred console.  This caused
   regressions in PowerMACs and other systems.

 - yet another fix for stdout-path option parsing.

 - fix error path handling in of_irq_parse_one

* tag 'devicetree-fixes-for-4.0-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  Revert "of: Fix premature bootconsole disable with 'stdout-path'"
  of: handle both '/' and ':' in path strings
  of: unittest: Add option string test case with longer path
  of/irq: Fix of_irq_parse_one() returned error codes

9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending
Linus Torvalds [Sat, 21 Mar 2015 18:24:38 +0000 (11:24 -0700)]
Merge git://git./linux/kernel/git/nab/target-pending

Pull SCSI target fixes from Nicholas Bellinger:
 "Here are current target-pending fixes for v4.0-rc5 code that have made
  their way into the queue over the last weeks.

  The fixes this round include:

   - Fix long-standing iser-target logout bug related to early
     conn_logout_comp completion, resulting in iscsi_conn use-after-tree
     OOpsen.  (Sagi + nab)

   - Fix long-standing tcm_fc bug in ft_invl_hw_context() failure
     handing for DDP hw offload.  (DanC)

   - Fix incorrect use of unprotected __transport_register_session() in
     tcm_qla2xxx + other single local se_node_acl fabrics.  (Bart)

   - Fix reference leak in target_submit_cmd() -> target_get_sess_cmd()
     for ack_kref=1 failure path.  (Bart)

   - Fix pSCSI backend ->get_device_type() statistics OOPs with
     un-configured device.  (Olaf + nab)

   - Fix virtual LUN=0 target_configure_device failure OOPs at modprobe
     time.  (Claudio + nab)

   - Fix FUA write false positive failure regression in v4.0-rc1 code.
     (Christophe Vu-Brugier + HCH)"

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  target: do not reject FUA CDBs when write cache is enabled but emulate_write_cache is 0
  target: Fix virtual LUN=0 target_configure_device failure OOPs
  target/pscsi: Fix NULL pointer dereference in get_device_type
  tcm_fc: missing curly braces in ft_invl_hw_context()
  target: Fix reference leak in target_get_sess_cmd() error path
  loop/usb/vhost-scsi/xen-scsiback: Fix use of __transport_register_session
  tcm_qla2xxx: Fix incorrect use of __transport_register_session
  iscsi-target: Avoid early conn_logout_comp for iser connections
  Revert "iscsi-target: Avoid IN_LOGOUT failure case for iser-target"
  target: Disallow changing of WRITE cache/FUA attrs after export

9 years agoMerge tag 'dm-4.0-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Sat, 21 Mar 2015 18:15:13 +0000 (11:15 -0700)]
Merge tag 'dm-4.0-fixes' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull devicemapper fixes from Mike Snitzer:
 "A handful of stable fixes for DM:
   - fix thin target to always zero-fill reads to unprovisioned blocks
   - fix to interlock device destruction's suspend from internal
     suspends
   - fix 2 snapshot exception store handover bugs
   - fix dm-io to cope with DISCARD and WRITE_SAME capabilities changing"

* tag 'dm-4.0-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm io: deal with wandering queue limits when handling REQ_DISCARD and REQ_WRITE_SAME
  dm snapshot: suspend merging snapshot when doing exception handover
  dm snapshot: suspend origin when doing exception handover
  dm: hold suspend_lock while suspending device during device deletion
  dm thin: fix to consistently zero-fill reads to unprovisioned blocks

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Sat, 21 Mar 2015 17:53:37 +0000 (10:53 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "Most of these are fixing extent reservation accounting, or corners
  with tree writeback during commit.

  Josef's set does add a test, which isn't strictly a fix, but it'll
  keep us from making this same mistake again"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: fix outstanding_extents accounting in DIO
  Btrfs: add sanity test for outstanding_extents accounting
  Btrfs: just free dummy extent buffers
  Btrfs: account merges/splits properly
  Btrfs: prepare block group cache before writing
  Btrfs: fix ASSERT(list_empty(&cur_trans->dirty_bgs_list)
  Btrfs: account for the correct number of extents for delalloc reservations
  Btrfs: fix merge delalloc logic
  Btrfs: fix comp_oper to get right order
  Btrfs: catch transaction abortion after waiting for it
  btrfs: fix sizeof format specifier in btrfs_check_super_valid()

9 years agoMerge branch 'for-4.0' of git://linux-nfs.org/~bfields/linux
Linus Torvalds [Sat, 21 Mar 2015 17:41:15 +0000 (10:41 -0700)]
Merge branch 'for-4.0' of git://linux-nfs.org/~bfields/linux

Pull nfsd bufix from Bruce Fields:
 "This is a fix for a crash easily triggered by 4.1 activity to a server
  built with CONFIG_NFSD_PNFS.

  There are some more bugfixes queued up that I intend to pass along
  next week, but this is the most critical"

* 'for-4.0' of git://linux-nfs.org/~bfields/linux:
  Subject: nfsd: don't recursively call nfsd4_cb_layout_fail

9 years agoMerge tag 'upstream-4.0-rc5' of git://git.infradead.org/linux-ubifs
Linus Torvalds [Sat, 21 Mar 2015 17:36:44 +0000 (10:36 -0700)]
Merge tag 'upstream-4.0-rc5' of git://git.infradead.org/linux-ubifs

Pull UBI fix from Artem Bityutskiy:
 "This fixes a bug introduced during the v4.0 merge window where we
  forgot to put braces where they should be"

* tag 'upstream-4.0-rc5' of git://git.infradead.org/linux-ubifs:
  UBI: fix missing brace control flow

9 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Sat, 21 Mar 2015 17:24:10 +0000 (10:24 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:

 - mm switching fix where the kernel pgd ends up in the user TTBR0 after
   returning from an EFI run-time services call

 - fix __GFP_ZERO handling for atomic pool and CMA DMA allocations (the
   generic code does get the gfp flags, so it's left with the arch code
   to memzero accordingly)

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: Honor __GFP_ZERO in dma allocations
  arm64: efi: don't restore TTBR0 if active_mm points at init_mm

9 years agoMerge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm
Linus Torvalds [Sat, 21 Mar 2015 17:03:22 +0000 (10:03 -0700)]
Merge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm

Pull ARM fixes from Russell King:
 "Another few ARM fixes.  Fabrice fixed the L2 cache DT parsing to allow
  prefetch configuration to be specified even when the cache size
  parsing fails.

  Laura noticed that the setting of page attributes wasn't working for
  modules due to is_module_addr() always returning false.

  Marc Gonzalez (aka Mason) noticed a potential latent bug with the way
  we read one of the CPUID registers (where we could attempt to read a
  non-present CPUID register which may fault.)

  I've fixed an issue where 32-bit DMA masks were failing with memory
  which extended to the top of physical address space, and I've also
  added debugging output of the page tables when we hit a data access
  exception which we don't specifically handle - prompted by the lack of
  information in a bug report"

* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  ARM: 8313/1: Use read_cpuid_ext() macro instead of inline asm
  ARM: 8311/1: Don't use is_module_addr in setting page attributes
  ARM: 8310/1: l2c: Fix prefetch settings dt parsing
  ARM: dump pgd, pmd and pte states on unhandled data abort faults
  ARM: dma-api: fix off-by-one error in __dma_supported()

9 years agoMerge branches 'pm-cpuidle', 'powercap', 'irq-pm' and 'acpi-resources'
Rafael J. Wysocki [Fri, 20 Mar 2015 23:39:12 +0000 (00:39 +0100)]
Merge branches 'pm-cpuidle', 'powercap', 'irq-pm' and 'acpi-resources'

* pm-cpuidle:
  cpuidle: mvebu: Update cpuidle thresholds for Armada XP SOCs
  cpuidle: mvebu: Fix the CPU PM notifier usage

* powercap:
  powercap / RAPL: handle domains with different energy units

* irq-pm:
  rtc: at91rm9200: double locking bug in at91_rtc_interrupt()

* acpi-resources:
  Revert "x86/PCI: Refine the way to release PCI IRQ resources"

9 years agomd: fix problems with freeing private data after ->run failure.
NeilBrown [Fri, 13 Mar 2015 00:51:18 +0000 (11:51 +1100)]
md: fix problems with freeing private data after ->run failure.

If ->run() fails, it can either free the data structures it
allocated, or leave that task to ->free() which will be called
on failures.

However:
  md.c calls ->free() even if ->private_data is NULL, which
     causes problems in some personalities.
  raid0.c frees the data, but doesn't clear ->private_data,
     which will become a problem when we fix md.c

So better fix both these issues at once.

Reported-by: Richard W.M. Jones <rjones@redhat.com>
Fixes: 5aa61f427e4979be733e4847b9199ff9cc48a47e
URL: https://bugzilla.kernel.org/show_bug.cgi?id=94381
Signed-off-by: NeilBrown <neilb@suse.de>
9 years agoarm64: Honor __GFP_ZERO in dma allocations
Suzuki K. Poulose [Thu, 19 Mar 2015 18:17:09 +0000 (18:17 +0000)]
arm64: Honor __GFP_ZERO in dma allocations

Current implementation doesn't zero out the pages allocated.
Honor the __GFP_ZERO flag and zero out if set.

Cc: <stable@vger.kernel.org> # v3.14+
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
9 years agoarm64: efi: don't restore TTBR0 if active_mm points at init_mm
Will Deacon [Thu, 19 Mar 2015 15:43:00 +0000 (15:43 +0000)]
arm64: efi: don't restore TTBR0 if active_mm points at init_mm

init_mm isn't a normal mm: it has swapper_pg_dir as its pgd (which
contains kernel mappings) and is used as the active_mm for the idle
thread.

When restoring the pgd after an EFI call, we write current->active_mm
into TTBR0. If the current task is actually the idle thread (e.g. when
initialising the EFI RTC before entering userspace), then the TLB can
erroneously populate itself with junk global entries as a result of
speculative table walks.

When we do eventually return to userspace, the task can end up hitting
these junk mappings leading to lockups, corruption or crashes.

This patch fixes the problem in the same way as the CPU suspend code by
ensuring that we never switch to the init_mm in efi_set_pgd and instead
point TTBR0 at the zero page. A check is also added to cpu_switch_mm to
BUG if we get passed swapper_pg_dir.

Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Fixes: f3cdfd239da5 ("arm64/efi: move SetVirtualAddressMap() to UEFI stub")
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
9 years agoRevert "x86/PCI: Refine the way to release PCI IRQ resources"
Rafael J. Wysocki [Fri, 20 Mar 2015 13:56:19 +0000 (14:56 +0100)]
Revert "x86/PCI: Refine the way to release PCI IRQ resources"

Commit b4b55cda5874 (Refine the way to release PCI IRQ resources)
introduced a regression in the PCI IRQ resource management by causing
the IRQ resource of a device, established when pci_enabled_device()
is called on a fully disabled device, to be released when the driver
is unbound from the device, regardless of the enable_cnt.

This leads to the situation that an ill-behaved driver can now make a
device unusable to subsequent drivers by an imbalance in their use of
pci_enable/disable_device().  That is a serious problem for secondary
drivers like vfio-pci, which are innocent of the transgressions of
the previous driver.

Since the solution of this problem is not immediate and requires
further discussion, revert commit b4b55cda5874 and the issue it was
supposed to address (a bug related to xen-pciback) will be taken
care of in a different way going forward.

Reported-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
9 years agoMerge tag 'drm-intel-fixes-2015-03-19' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Fri, 20 Mar 2015 07:32:21 +0000 (17:32 +1000)]
Merge tag 'drm-intel-fixes-2015-03-19' of git://anongit.freedesktop.org/drm-intel into drm-fixes

Backporting a couple of plane related fixes from drm-next to v4.0.

* tag 'drm-intel-fixes-2015-03-19' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: Make sure the primary plane is enabled before reading out the fb state
  drm/i915: Ensure plane->state->fb stays in sync with plane->fb

9 years agoMerge tag 'drm-amdkfd-fixes-2015-03-19' of git://people.freedesktop.org/~gabbayo...
Dave Airlie [Fri, 20 Mar 2015 07:32:01 +0000 (17:32 +1000)]
Merge tag 'drm-amdkfd-fixes-2015-03-19' of git://people.freedesktop.org/~gabbayo/linux into drm-fixes

- Fixing SDMA initialization when in non-HWS mode (debug mode)
- Memory leak fix when destroying kernel queue
- Fix number of available compute pipelines according to new firmware

* tag 'drm-amdkfd-fixes-2015-03-19' of git://people.freedesktop.org/~gabbayo/linux:
  drm/radeon: Changing number of compute pipe lines
  drm/amdkfd: Fix SDMA queue init. in non-HWS mode
  drm/amdkfd: destroy mqd when destroying kernel queue

9 years agotarget: do not reject FUA CDBs when write cache is enabled but emulate_write_cache...
Christophe Vu-Brugier [Thu, 19 Mar 2015 13:30:13 +0000 (14:30 +0100)]
target: do not reject FUA CDBs when write cache is enabled but emulate_write_cache is 0

A check that rejects a CDB with FUA bit set if no write cache is
emulated was added by the following commit:

  fde9f50 target: Add sanity checks for DPO/FUA bit usage

The condition is as follows:

  if (!dev->dev_attrib.emulate_fua_write ||
      !dev->dev_attrib.emulate_write_cache)

However, this check is wrong if the backend device supports WCE but
"emulate_write_cache" is disabled.

This patch uses se_dev_check_wce() (previously named
spc_check_dev_wce) to invoke transport->get_write_cache() if the
device has a write cache or check the "emulate_write_cache" attribute
otherwise.

Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christophe Vu-Brugier <cvubrugier@fastmail.fm>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
9 years agotarget: Fix virtual LUN=0 target_configure_device failure OOPs
Nicholas Bellinger [Thu, 5 Mar 2015 03:28:24 +0000 (03:28 +0000)]
target: Fix virtual LUN=0 target_configure_device failure OOPs

This patch fixes a NULL pointer dereference triggered by a late
target_configure_device() -> alloc_workqueue() failure that results
in target_free_device() being called with DF_CONFIGURED already set,
which subsequently OOPses in destroy_workqueue() code.

Currently this only happens at modprobe target_core_mod time when
core_dev_setup_virtual_lun0() -> target_configure_device() fails,
and the explicit target_free_device() gets called.

To address this bug originally introduced by commit 0fd97ccf45, go
ahead and move DF_CONFIGURED to end of target_configure_device()
code to handle this special failure case.

Reported-by: Claudio Fleiner <cmf@daterainc.com>
Cc: Claudio Fleiner <cmf@daterainc.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: <stable@vger.kernel.org> # v3.7+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>