GitHub/MotorolaMobilityLLC/kernel-slsi.git
11 years agommc: omap_hsmmc: context save and restore for DT
Tony Lindgren [Sun, 20 Oct 2013 18:55:19 +0000 (00:25 +0530)]
mmc: omap_hsmmc: context save and restore for DT

We want to get rid of the omap specific platform init code
callbacks as they don't play nice with device tree.

Convert the context loss check to be based on a
register state detection instead.

Cc: Andreas Fenkart <afenkart@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
[add check for CON and HCTL in context save and restore function]
Signed-off-by: Balaji T K <balajitk@ti.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: omap_hsmmc: Fix pbias_disable for omap4
Balaji T K [Sun, 20 Oct 2013 18:55:18 +0000 (00:25 +0530)]
mmc: omap_hsmmc: Fix pbias_disable for omap4

pbias_disable is set to protect the mmc pbias i/o cells in DT boot
by preventing voltage switch. Currently pbias_disable is enabled only
for omap3 and not for omap4 due to reg_offset difference of 0x100.
Enable pbias_disable for omap4+ too by using res->start
which does not include the reg_offset.

Signed-off-by: Balaji T K <balajitk@ti.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci: remove unneeded call when have preset value quirk
Dong Aisheng [Fri, 18 Oct 2013 11:48:50 +0000 (19:48 +0800)]
mmc: sdhci: remove unneeded call when have preset value quirk

Remove unneeded call of call sdhci_enable_preset_value when having
SDHCI_QUIRK2_PRESET_VALUE_BROKEN.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: add preset value quirk for mx6
Dong Aisheng [Fri, 18 Oct 2013 11:48:49 +0000 (19:48 +0800)]
mmc: sdhci-esdhc-imx: add preset value quirk for mx6

The i.MX6 does not support preset value feature.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: enable SDR50 tuning for imx6q/dl
Dong Aisheng [Fri, 18 Oct 2013 11:48:48 +0000 (19:48 +0800)]
mmc: sdhci-esdhc-imx: enable SDR50 tuning for imx6q/dl

The imx6q/dl supports SDR50 tunning, enable it for a better timing
on SDR50 mode.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: add delay line setting support
Dong Aisheng [Fri, 18 Oct 2013 11:48:47 +0000 (19:48 +0800)]
mmc: sdhci-esdhc-imx: add delay line setting support

The DLL(Delay Line) is newly added to assist in sampling read data.
The DLL provides the ability to programmatically select a quantized
delay (in fractions of the clock period) regardless of on-chip variations
such as process, voltage and temperature (PVT).

This patch adds a user interface to set slave delay line via device tree.
It's usually used in high speed mode like mmc DDR mode when the signal
quality is not good caused by board design, e.g. the signal path is too
long.  User can manually set delay line to find a suitable data sampling
window for card to work properly.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: add DDR mode support for mx6
Dong Aisheng [Fri, 18 Oct 2013 11:48:46 +0000 (19:48 +0800)]
mmc: sdhci-esdhc-imx: add DDR mode support for mx6

When DDR mode is enabled, the initial pre_div should be 2.
And the pre_div value should be changed accordingly
from
...
02h) Base clock divided by 4
01h) Base clock divided by 2
00h) Base clock divided by 1
to
..
02h) Base clock divided by 8
01h) Base clock divided by 4
00h) Base clock divided by 2

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci: report error once the maximum tuning loops exhausted or timeout
Dong Aisheng [Fri, 18 Oct 2013 11:48:45 +0000 (19:48 +0800)]
mmc: sdhci: report error once the maximum tuning loops exhausted or timeout

The original code missed to report an error when the maximum tuning
loops exhausted or timeout, it will cause the upper layer to wrongly
think the tuning process is passed.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: fix reading cap_1 register value for mx6sl
Dong Aisheng [Fri, 18 Oct 2013 11:48:44 +0000 (19:48 +0800)]
mmc: sdhci-esdhc-imx: fix reading cap_1 register value for mx6sl

When reading CAP_1 register for mx6sl, ignore bit[0-15] as it stores
CAP_2 register value which is new introduced in mx6sl.

Without this fix, the max clock for mx6sl may not be correct since
it's wrongly calculated by reading CAP_1 register.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: add std tuning support for mx6sl
Dong Aisheng [Fri, 18 Oct 2013 11:48:43 +0000 (19:48 +0800)]
mmc: sdhci-esdhc-imx: add std tuning support for mx6sl

The mx6sl supports standard sdhci tuning, then esdhc_executing_tuning
is only needed for mx6q/dl. We introduce is_imx6_usdhc() and
is_imx6sl_usdhc() to handle the difference.

The standard tuning is enabled by setting ESDHC_TUNE_CTRL_STD_TUNING_EN bit
in new register ESDHC_TUNE_CTRL and operates with new tuning bits
defined in SDHCI_ACMD12_ERR register.

Note: mx6sl can also work on the old manually tuning mode as mx6q/dl if
not enable standard tuning mode.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: create struct esdhc_soc_data
Shawn Guo [Thu, 17 Oct 2013 07:19:47 +0000 (15:19 +0800)]
mmc: sdhci-esdhc-imx: create struct esdhc_soc_data

Create a struct esdhc_soc_data with moving 'flags' field from
pltfm_imx_data into it, and pass the pointer of this SoC specific data
structure through of_device_id.data directly, so that the translation
from enum imx_esdhc_type to flags can be saved.

With the change, enum imx_esdhc_type can be eliminated, since we can
implement the is_imx*_esdhc() by checking the esdhc_soc_data pointer.
The unused is_imx35_esdhc() and is_imx51_esdhc() are also removed, and the
others are kept there as we will need to use them to handle some small
register differences later, where use of new flags might be a little
overkilled.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Dong Aisheng <b29396@freescale.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: pdev->id_entry should be immutable
Shawn Guo [Thu, 17 Oct 2013 07:19:46 +0000 (15:19 +0800)]
mmc: sdhci-esdhc-imx: pdev->id_entry should be immutable

As a good practice, device driver should not modify pdev->id_entry but
keep it immutable.  Let's assign of_device_id.data with imx_esdhc_type
constants directly, so that we do not have to manipulate pdev->id_entry
in .probe().

As the result, sdhci-esdhc-imx53 and sdhci-usdhc-imx6q can be removed
from platform_device_id table now, since they will only probe from
device tree.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Dong Aisheng <b29396@freescale.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: add flag ESDHC_FLAG_USDHC
Shawn Guo [Thu, 17 Oct 2013 07:19:45 +0000 (15:19 +0800)]
mmc: sdhci-esdhc-imx: add flag ESDHC_FLAG_USDHC

Add flag ESDHC_FLAG_USDHC to tell that the ESDHC is actually an USDHC
block, and replace the is_imx6q_usdhc() occurrences with inline function
esdhc_is_usdhc() which checks the flag.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Dong Aisheng <b29396@freescale.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: add flag ESDHC_FLAG_ENGCM07207
Shawn Guo [Thu, 17 Oct 2013 07:19:44 +0000 (15:19 +0800)]
mmc: sdhci-esdhc-imx: add flag ESDHC_FLAG_ENGCM07207

Just like the use of the flag ESDHC_FLAG_MULTIBLK_NO_INT, let's add
another flag ESDHC_FLAG_ENGCM07207 to enable the workaround for errata
ENGcm07207 and set the flag for i.MX25 and i.MX35 ESDHC.

While at it, let's use BIT() macro for ESDHC_FLAG_MULTIBLK_NO_INT as
well.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Dong Aisheng <b29396@freescale.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: bfin_sdh: Forgot to write SDH_CFG register in function set_ios.
Sonic Zhang [Wed, 16 Oct 2013 05:42:05 +0000 (13:42 +0800)]
mmc: bfin_sdh: Forgot to write SDH_CFG register in function set_ios.

The DMA operation in SDIO CMD53 may fail because the 4 bit SDIO mode
is not set up correctly on BF548.
Remove 2 blank lines as well.

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: wmt-sdmmc: Simplify wmt_set_sd_power implementation
Axel Lin [Thu, 3 Oct 2013 09:04:40 +0000 (17:04 +0800)]
mmc: wmt-sdmmc: Simplify wmt_set_sd_power implementation

Simplify the code to set/clear BM_SD_OFF bit:

          | power_inverted: 0  | power_inverted: 1
-------------------------------------------------
enable: 0 | SET BM_SD_OFF      | CLEAR BM_SD_OFF
-------------------------------------------------
enable: 1 | CLEAR BM_SD_OFF    | SET BM_SD_OFF

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Acked-by: Tony Prisk <linux@prisktech.co.nz>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-pci: add Intel Merrifield support
David Cohen [Tue, 1 Oct 2013 20:18:15 +0000 (13:18 -0700)]
mmc: sdhci-pci: add Intel Merrifield support

Implement initial SDHCI Intel Merrifield support.  This patch is based
on previous one from Yunpeng Gao <yunpeng.gao@intel.com>.

Signed-off-by: David Cohen <david.a.cohen@linux.intel.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: set actual_clock in clock setting
Dong Aisheng [Fri, 13 Sep 2013 11:11:37 +0000 (19:11 +0800)]
mmc: sdhci-esdhc-imx: set actual_clock in clock setting

This enables access the actual_clock via sys.
root@imx6qsabreauto:~# cat /sys/kernel/debug/mmc0/ios
clock:          198000000 Hz
actual clock:   198000000 Hz
vdd:            17 (2.9 ~ 3.0 V)
bus mode:       2 (push-pull)
chip select:    0 (don't care)
power mode:     2 (on)
bus width:      2 (4 bits)
timing spec:    6 (sd uhs SDR104)
signal voltage: 0 (1.80 V)

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: correct pre_div for imx6q
Dong Aisheng [Fri, 13 Sep 2013 11:11:36 +0000 (19:11 +0800)]
mmc: sdhci-esdhc-imx: correct pre_div for imx6q

According to spec, the pre_div for imx6q should be 1, or the biggest
clock rate we can get is a half of host clock rate.  This may cause
we can not get the proper clock rate as we want.  e.g. if the desired
clock is 200Mhz, however, the host clock is 200Mhz too, then it causes
the actual clock we get is 100Mhz due to pre_div is 2.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: change pinctrl state according to uhs mode
Dong Aisheng [Fri, 13 Sep 2013 11:11:35 +0000 (19:11 +0800)]
mmc: sdhci-esdhc-imx: change pinctrl state according to uhs mode

Without proper pinctrl state, the card may not be able to work
on high speed stablely. e.g. SDR104.

This patch add pinctrl state switch code according to different
uhs mode include 100mhz sate, 200mhz sate and normal state
(50Mhz and below).

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: add sd3.0 SDR clock tuning support
Dong Aisheng [Fri, 13 Sep 2013 11:11:34 +0000 (19:11 +0800)]
mmc: sdhci-esdhc-imx: add sd3.0 SDR clock tuning support

Freescale i.MX6Q/DL uSDHC clock tuning progress is a little different from
the standard tuning process defined in host controller spec v3.0.
Thus we use platform_execute_tuning instead of standard sdhci tuning.

The main difference are:
1) not only generate Buffer Read Ready interrupt when tuning is performing.
   It generates all other DATA interrupts like the normal data command.
2) SDHCI_CTRL_EXEC_TUNING is not automatically cleared by HW,
   instead it's controlled by SW.
3) SDHCI_CTRL_TUNED_CLK is not automatically set by HW,
   it's controlled by SW.
4) the clock delay for every tuning is set by SW.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc-imx: support real clock on and off for imx6q
Dong Aisheng [Fri, 13 Sep 2013 11:11:33 +0000 (19:11 +0800)]
mmc: sdhci-esdhc-imx: support real clock on and off for imx6q

The signal voltage switch flow requires to shutdown and output
clock in a specific sequence according to standard host controller
v3.0 spec. In that timing, the card must really receive clock or not.

However, for i.MX6Q, the uSDHC will not output clock even the clock
is enabled until there is command or data in transfer on the bus,
which will then cause singal voltage switch always to fail.

For i.MX6Q, we clear ESDHC_VENDOR_SPEC_FRC_SDCLK_ON bit to let
controller to gate off clock automatically and set that bit
to force clock output if clock is on.

This is required by SD3.0 support.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-esdhc: move common esdhc_set_clock to platform driver
Dong Aisheng [Fri, 13 Sep 2013 11:11:32 +0000 (19:11 +0800)]
mmc: sdhci-esdhc: move common esdhc_set_clock to platform driver

We need a lot of imx6 specific things into common esdhc_set_clock
for support SD3.0 and eMMC DDR mode which is not needed for power pc
platforms, so esdhc_set_clock seems not so common anymore.

Instead of keeping add platform specfics things into this common API,
we choose to move that code into platform driver itself to handle.
This can also exclude the dependency between imx and power pc on this
headfile and is easy for maintain in the future.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci: allow platform access of sdhci_send_command
Dong Aisheng [Fri, 13 Sep 2013 11:11:31 +0000 (19:11 +0800)]
mmc: sdhci: allow platform access of sdhci_send_command

It helps for platform code to use it send tuning commands.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci: add hooks for platform specific tuning
Dong Aisheng [Fri, 13 Sep 2013 11:11:30 +0000 (19:11 +0800)]
mmc: sdhci: add hooks for platform specific tuning

The tuning of some platforms may not follow the standard host control
spec v3.0, e.g. Freescale uSDHC on i.MX6Q/DL.
Add a hook here to allow execute platform specific tuning instead of
standard host controller tuning.

The hook only replaces the tuning process, so it's placed after tuning
checking and before the real tuning process.

Some notes for the tuning hook:
1) it needs handle lock itself if it wants to access host controller
according platform specific implementation.
2) do not need to handle runtime pm since it executes with runtime pm
get already.

Signed-off-by: Dong Aisheng <b29396@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-bcm2835: Use sdhci_pltfm_unregister instead of open coded
Axel Lin [Tue, 17 Sep 2013 08:00:47 +0000 (16:00 +0800)]
mmc: sdhci-bcm2835: Use sdhci_pltfm_unregister instead of open coded

This avoid duplicated implementation.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Acked-by: Stephen Warren <swarren@wwwdotorg.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: sdhci-bcm-kona: Use sdhci_pltfm_unregister instead of open coded
Axel Lin [Tue, 17 Sep 2013 07:59:32 +0000 (15:59 +0800)]
mmc: sdhci-bcm-kona: Use sdhci_pltfm_unregister instead of open coded

This avoid duplicated implementation and also fixes missing iounmap() and
release_mem_region() calls in sdhci_bcm_kona_remove().  sdhci_pltfm_init()
calls request_mem_region() and ioremap(), thus we need to call the
corresponding iounmap() and release_mem_region() calls in
sdhci_bcm_kona_remove().

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc-socfpga: Staticize dw_mci_socfpga_probe
Sachin Kamat [Thu, 12 Sep 2013 06:46:36 +0000 (12:16 +0530)]
mmc: dw_mmc-socfpga: Staticize dw_mci_socfpga_probe

'dw_mci_socfpga_probe' is used only in this file. Make it static.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc-socfpga: Remove redundant of_match_ptr
Sachin Kamat [Thu, 12 Sep 2013 06:46:35 +0000 (12:16 +0530)]
mmc: dw_mmc-socfpga: Remove redundant of_match_ptr

'dw_mci_socfpga_match' is always compiled in.
Hence of_match_ptr is not necessary.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: mvsdio: Convert to devm_ioremap_resource
Sachin Kamat [Thu, 12 Sep 2013 06:46:34 +0000 (12:16 +0530)]
mmc: mvsdio: Convert to devm_ioremap_resource

devm_request_and_ioremap() is deprecated. Use devm_ioremap_resource()
instead.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: core: clean up duplicate macros
Jackey Shen [Fri, 17 May 2013 09:17:43 +0000 (17:17 +0800)]
mmc: core: clean up duplicate macros

Clean up the duplicate macros:
mmc_sd_card_uhs -> mmc_card_uhs
mmc_sd_card_set_uhs -> mmc_card_set_uhs

Signed-off-by: Jackey Shen <jackey.shen@amd.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: atmel-mci: fix oops in atmci_tasklet_func
Rodolfo Giometti [Mon, 9 Sep 2013 15:31:59 +0000 (17:31 +0200)]
mmc: atmel-mci: fix oops in atmci_tasklet_func

In some cases, a NULL pointer dereference happens because data is NULL when
STATE_END_REQUEST case is reached in atmci_tasklet_func.

Cc: <stable@vger.kernel.org> # 3.9+
Signed-off-by: Rodolfo Giometti <giometti@enneenne.com>
Acked-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: atmel-mci: abort transfer on timeout error
Ludovic Desroches [Mon, 9 Sep 2013 15:29:56 +0000 (17:29 +0200)]
mmc: atmel-mci: abort transfer on timeout error

When a software timeout occurs, the transfer is not stopped. In DMA case,
it causes DMA channel to be stuck because the transfer is still active
causing following transfers to be queued but not computed.

Cc: <stable@vger.kernel.org> # 3.9+
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Reported-by: Alexander Morozov <etesial@gmail.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: add ignorance case for CMD13 CRC error
Seungwon Jeon [Wed, 4 Sep 2013 12:21:05 +0000 (21:21 +0900)]
mmc: add ignorance case for CMD13 CRC error

While speed mode is changed, CMD13 cannot be guaranteed.
According to the spec., it is not recommended to use CMD13
to check the busy completion of the timing change.
If CMD13 is used in this case, CRC error must be ignored.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: fix the transfer termination in IDMAC mode
Seungwon Jeon [Fri, 30 Aug 2013 15:14:38 +0000 (00:14 +0900)]
mmc: dw_mmc: fix the transfer termination in IDMAC mode

In IDMAC mode EVENT_XFER_COMPLETE is set when RI/TI of last descriptor
is done. So if errors are happened in the middle of data transfers,
'dw_mci_stop_dma' during error handing can be called and eventually
prevents this flag to be set.  This results in permanent wait for
EVENT_XFER_COMPLETE in 'dw_mci_tasklet_func'.  Therefore, if dma
running is stopped forcibly, EVENT_XFER_COMPLETE should be set.

Reported-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: amend use of idmac sw reset
Seungwon Jeon [Fri, 30 Aug 2013 15:14:33 +0000 (00:14 +0900)]
mmc: dw_mmc: amend use of idmac sw reset

First, compiling warning along with previous change is removed.
[drivers/mmc/host/dw_mmc.c:1890:7: warning: unused variable 'ctrl']
And with the recommendation in manual, IDMAC software reset is followed
by dma-reset of the CTRL register in order to terminate the transfer.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: gather each reset code into functions
Seungwon Jeon [Fri, 30 Aug 2013 15:14:23 +0000 (00:14 +0900)]
mmc: dw_mmc: gather each reset code into functions

There are three resets in CTRL register.  FIFO reset is especially used
in several points with the same routine. It could be replaced with one
function and the others may be applied similarly if needed.  So,
mci_wait_reset() is modified to allow various bit field of reset.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: rework the code related to cmd/data completion
Seungwon Jeon [Fri, 30 Aug 2013 15:14:17 +0000 (00:14 +0900)]
mmc: dw_mmc: rework the code related to cmd/data completion

Main change corresponds to dw_mci_command_complete().  And EBE is
divided into read and write.  Some minor changes for code readability.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: guarantee stop-abort cmd in data errors
Seungwon Jeon [Fri, 30 Aug 2013 15:14:05 +0000 (00:14 +0900)]
mmc: dw_mmc: guarantee stop-abort cmd in data errors

In error cases, DTO interrupt may or may not be generated depending
on remained data. Stop/Abort command ensures DTO generation for that
situation. Currently if 'stop' field of data is empty, there is no
stop/abort command. So, it could hang waiting DTO. This change
reinforces these cases.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: fix error handling on response error
Seungwon Jeon [Fri, 30 Aug 2013 15:13:59 +0000 (00:13 +0900)]
mmc: dw_mmc: fix error handling on response error

Even if response error is detected in case data command, data transfer
is continued. It means that data can live in FIFO. Current handling
just breaks out the request when seeing the command error. This causes
kernel panic in dw_mci_read_data_pio() [host->data = NULL]. And also,
FIFO should be guaranteed to be empty.

Unable to handle kernel NULL pointer dereference at virtual address 00000018
<...>
[<c02af814>] (dw_mci_read_data_pio+0x68/0x198) from [<c02b04b4>] (dw_mci_interrupt+0x374/0x3a0)
[<c02b04b4>] (dw_mci_interrupt+0x374/0x3a0) from [<c006b094>] (handle_irq_event_percpu+0x50/0x194)
[<c006b094>] (handle_irq_event_percpu+0x50/0x194) from [<c006b214>] (handle_irq_event+0x3c/0x5c)
[<c006b214>] (handle_irq_event+0x3c/0x5c) from [<c006de1c>] (handle_fasteoi_irq+0xa4/0x148)
[<c006de1c>] (handle_fasteoi_irq+0xa4/0x148) from [<c006aa88>] (generic_handle_irq+0x20/0x30)
[<c006aa88>] (generic_handle_irq+0x20/0x30) from [<c000f154>] (handle_IRQ+0x38/0x90)
[<c000f154>] (handle_IRQ+0x38/0x90) from [<c00085bc>] (gic_handle_irq+0x34/0x68)
[<c00085bc>] (gic_handle_irq+0x34/0x68) from [<c0011f40>] (__irq_svc+0x40/0x70)
Exception stack(0xef0b1c00 to 0xef0b1c48)
1c00: 000eb0cf ffffffff 00001300 c01a7738 ef295e10 0000000a c04df298 ef0b1dc0
1c20: ef295ec0 00000000 00000000 00000006 00000000 ef0b1c48 c02b1274 c01a7764
1c40: 20000113 ffffffff
[<c0011f40>] (__irq_svc+0x40/0x70) from  [<c01a7764>] (__loop_delay+0x0/0xc)
Code: e1a00005 e0891006 e0662004 e12fff33 (e59a3018)
---[ end trace a7043b9ba9aed1db ]---
Kernel panic - not syncing: Fatal exception in interrupt

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: control card read threshold
Seungwon Jeon [Fri, 30 Aug 2013 15:13:55 +0000 (00:13 +0900)]
mmc: dw_mmc: control card read threshold

Card Read Threshold should be ensured that the card clock does not stop
in the middle of a block of data being transferred from the card to the
Host. Specially, clock stop is allowed in fast transfer such as HS200
or SDR104 mode. And so, it should be enabled.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: adjust the fifoth with block size
Seungwon Jeon [Fri, 30 Aug 2013 15:13:42 +0000 (00:13 +0900)]
mmc: dw_mmc: adjust the fifoth with block size

This change helps to choose msize, rx_watermark and tx_watermark
depending on block size for IDMAC mode.  For SDIO block size can be
variable, so if these values are set incorrectly, card clock may stop.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: set the supported max/min frequency
Seungwon Jeon [Fri, 30 Aug 2013 15:13:31 +0000 (00:13 +0900)]
mmc: dw_mmc: set the supported max/min frequency

Both f_max and f_min will be informed for core layer to request
valid clock rate. But current setting from 'host->bus_hz' may
not represent the max/min frequency properly. Even if host can
actually support high speed than bus_hz, core layer will not
request clock rate over bus_hz. Basically, f_max/f_min can be set
with the values according to spec. And then host will make its best
effort to meet the rate.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: move supports-highspeed of quirks to caps
Seungwon Jeon [Fri, 30 Aug 2013 15:13:22 +0000 (00:13 +0900)]
mmc: dw_mmc: move supports-highspeed of quirks to caps

'supports-highspeed' is not one of the quirks but is a capability.
So, it's removed from quirks.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: add the capability to support hs200 mode
Seungwon Jeon [Fri, 30 Aug 2013 15:13:09 +0000 (00:13 +0900)]
mmc: dw_mmc: add the capability to support hs200 mode

As host controller can support eMMC's HS200 mode at 1.8V or 1.2V,
these capability will be added.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: exynos: adjust the clock rate with speed mode
Seungwon Jeon [Fri, 30 Aug 2013 15:13:03 +0000 (00:13 +0900)]
mmc: dw_mmc: exynos: adjust the clock rate with speed mode

Exynos's host has divider logic before 'cclk_in' to controller core.
It means that actual clock rate of ciu clock comes from this divider
value. So, source clock should be adjusted along with 'ciu_div' which
indicates the host's divider ratio. Setting clock rate basically fits
the required speed. Specially, 'cclk_in' should have double rate of
target speed in case of DDR 8-bit mode.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: exynos: add variable delay tuning sequence
Seungwon Jeon [Fri, 30 Aug 2013 15:12:50 +0000 (00:12 +0900)]
mmc: dw_mmc: exynos: add variable delay tuning sequence

Implements variable delay tuning. In this change, exynos host can
determine the correct sampling point for the HS200 and SDR104 speed mode.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: add support tuning scheme
Seungwon Jeon [Fri, 30 Aug 2013 15:12:42 +0000 (00:12 +0900)]
mmc: dw_mmc: add support tuning scheme

For the speed modes HS200 and SDR104, tuning is needed to determine the
correct sampling point. Actual tuning procedure is provided by specific
host controller driver.  This patch defines the tuning command and
tuning data.  Additionally, 'struct dw_mci_slot' is moved to header
file to consider the extensive usages in driver.

Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: exynos: configure SMU in exynos5420
Yuvaraj Kumar C D [Fri, 30 Aug 2013 15:12:35 +0000 (00:12 +0900)]
mmc: dw_mmc: exynos: configure SMU in exynos5420

Exynos5420 Mobile Storage Host controller has Security Management
Unit (SMU) for channel 0 and channel 1 (mainly for eMMC).
This time, SMU configuration is set for non-encryption mode.

Signed-off-by: Yuvaraj Kumar C D <yuvaraj.cd@samsung.com>
Signed-off-by: Alim Akhtar <alim.akhtar@samsung.com>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: move the platform specific init call
Yuvaraj Kumar C D [Fri, 30 Aug 2013 15:12:19 +0000 (00:12 +0900)]
mmc: dw_mmc: move the platform specific init call

Current platform specific private data initialization call
dw_mci_exynos_priv_init() can be used to do platform specific
initialization of SMU and others in future. So the drv_data->init
call has moved to dw_mci_probe().

Signed-off-by: Yuvaraj Kumar C D <yuvaraj.cd@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: socfpga: move socfpga private init
Yuvaraj Kumar C D [Fri, 30 Aug 2013 15:12:10 +0000 (00:12 +0900)]
mmc: dw_mmc: socfpga: move socfpga private init

Currently platform specific private data initialization is done by
dw_mci_socfpga_priv_init and dw_mci_socfpga_parse_dt. As we already have
separate platform specific device tree parser dw_mci_socfpga_parse_dt,
move the dw_mci_socfpga_priv_init code to dw_mci_socfpga_parse_dt.
We can use the dw_mci_socfpga_priv_init to do some actual platform
specific initialization.

Signed-off-by: Yuvaraj Kumar C D <yuvaraj.cd@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: exynos: move the exynos private init
Yuvaraj Kumar C D [Fri, 30 Aug 2013 15:11:57 +0000 (00:11 +0900)]
mmc: dw_mmc: exynos: move the exynos private init

Currently platform specific private data initialization is done by
dw_mci_exynos_priv_init and dw_mci_exynos_parse_dt. As we already have
separate platform specific device tree parser dw_mci_exynos_parse_dt,
move the dw_mci_exynos_priv_init code to dw_mci_exynos_parse_dt.
We can use the dw_mci_exynos_priv_init to do some actual platform
specific initialization of SMU and etc.

Signed-off-by: Yuvaraj Kumar C D <yuvaraj.cd@samsung.com>
Tested-by: Alim Akhtar <alim.akhtar@samsung.com>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: Set timeout to max upon resume
Doug Anderson [Fri, 30 Aug 2013 15:11:49 +0000 (00:11 +0900)]
mmc: dw_mmc: Set timeout to max upon resume

The TMOUT register is set to 0xffffffff at probe time but isn't
set after suspend/resume.  Add an init of this value.

No problems were observed without this (it will also be set in
__dw_mci_start_request if there is data to send), but it makes the
register dump before and after suspend cleaner.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: Honor requests to set the clock to 0
Doug Anderson [Fri, 30 Aug 2013 15:11:43 +0000 (00:11 +0900)]
mmc: dw_mmc: Honor requests to set the clock to 0

Previously the dw_mmc driver would ignore any requests to disable the
card's clock.  This doesn't seem like a good thing in general, but had
one extra bad side effect in the following situation:
* mmc core would set clk to 400kHz at boot time while scanning
* mmc core would set clk to 0 since no card, but it would be ignored.
* suspend to ram and resume; clocks in the dw_mmc IP block are now 0
  but dw_mmc thinks that they're 400kHz (it ignored the set to 0).
* insert card
* mmc core would set clk to 400kHz which would be considered a no-op.

Note that if there is no card in the slot and we do a suspend/resume
cycle, we _do_ still end up with differences in a dw_mmc register
dump, but the differences are clock related and we've got the clock
disabled both before and after, so this should be OK.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: Add exynos resume_noirq callback to clear WAKEUP_INT
Doug Anderson [Fri, 30 Aug 2013 15:11:21 +0000 (00:11 +0900)]
mmc: dw_mmc: Add exynos resume_noirq callback to clear WAKEUP_INT

If the WAKEUP_INT is asserted at wakeup and not cleared, we'll end up
looping around forever.  This has been seen to happen on exynos5420
silicon despite the fact that we haven't enabled any wakeup events due
to a silicon errata.  It is safe to do on all exynos variants.

Signed-off-by: Doug Anderson <dianders@chromium.org>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: dw_mmc: don't queue up a card detect at slot startup
Doug Anderson [Fri, 30 Aug 2013 15:11:06 +0000 (00:11 +0900)]
mmc: dw_mmc: don't queue up a card detect at slot startup

The MMC subsystem handles looking for a card at probe time. Queuing up our
own can race with the rest of the MMC subsystem and cause problems if we
get unlucky with timing. Just remove driver own detection triggering.  While
progressing the request from 'mmc_rescan', if 'dw_mci_work_routine_card'
routine is activated, it will cancel the current request. The problem case
is that 'mmc_rescan' is prior to 'dw_mci_work_routine_card' from host own.
Specifically, the following message shows the detection problem in driver's
probing. It would get an err -123 (-ENOMEDIUM) during probe.

[    4.216595] dwmmc_exynos 12210000.dwmmc1: Using internal DMA controller.
[    4.395935] dwmmc_exynos 12210000.dwmmc1: Version ID is 250a
[    4.401948] dwmmc_exynos 12210000.dwmmc1: DW MMC controller at irq 108, 64 bit host data width, 64 deep fifo
[    4.424430] dwmmc_exynos 12210000.dwmmc1: sdr0 mode (irq=108, width=0)
[    4.453975] dwmmc_exynos 12210000.dwmmc1: sdr0 mode (irq=108, width=0)
[    4.459592] mmc_host mmc1: Bus speed (slot 0) = 100000000Hz (slot req 400000Hz, actual 400000HZ div = 125)
[    4.484258] dwmmc_exynos 12210000.dwmmc1: 1 slots initialized
[    4.485406] dwmmc_exynos 12210000.dwmmc1: sdr0 mode (irq=108, width=0)
[    4.487606] dwmmc_exynos 12210000.dwmmc1: sdr0 mode (irq=108, width=0)
[    4.489794] dwmmc_exynos 12210000.dwmmc1: sdr0 mode (irq=108, width=0)
[    4.509757] mmc1: error -123 whilst initialising SDIO card

Signed-off-by: Doug Anderson <dianders@chromium.org>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agosh: ecovec: fixup compile error on sdhi
Kuninori Morimoto [Wed, 25 Sep 2013 01:55:22 +0000 (18:55 -0700)]
sh: ecovec: fixup compile error on sdhi

afa2c9407f8908 ("sh: ecovec24: Use MMC/SDHI CD and RO GPIO") added
.tmio_flags = TMIO_MMC_USE_GPIO_CD on sh_mobile_sdhi_info, but it needs
<linux/mfd/tmio.h> header.  This patch adds it.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Reviewed-by: Yusuke Goda <yusuke.goda.sx@renesas.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agommc: core: remove dead function mmc_try_claim_host
Grant Grundler [Fri, 20 Sep 2013 01:21:36 +0000 (18:21 -0700)]
mmc: core: remove dead function mmc_try_claim_host

cscope says there are no callers for mmc_try_claim_host in the kernel.
No reason to keep it.

Signed-off-by: Grant Grundler <grundler@chromium.org>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
11 years agoMerge tag 'stable/for-linus-3.12-rc2-tag' of git://git.kernel.org/pub/scm/linux/kerne...
Linus Torvalds [Wed, 25 Sep 2013 22:50:53 +0000 (15:50 -0700)]
Merge tag 'stable/for-linus-3.12-rc2-tag' of git://git./linux/kernel/git/xen/tip

Pull Xen fixes from Konrad Rzeszutek Wilk:
 "Bug-fixes and one update to the kernel-paramters.txt documentation.

   - Fix PV spinlocks triggering jump_label code bug
   - Remove extraneous code in the tpm front driver
   - Fix ballooning out of pages when non-preemptible
   - Fix deadlock when using a 32-bit initial domain with large amount
     of memory
   - Add xen_nopvpsin parameter to the documentation"

* tag 'stable/for-linus-3.12-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/spinlock: Document the xen_nopvspin parameter.
  xen/p2m: check MFN is in range before using the m2p table
  xen/balloon: don't alloc page while non-preemptible
  xen: Do not enable spinlocks before jump_label_init() has executed
  tpm: xen-tpmfront: Remove the locality sysfs attribute
  tpm: xen-tpmfront: Fix default durations

11 years agoMerge tag 'dm-3.12-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Wed, 25 Sep 2013 22:12:46 +0000 (15:12 -0700)]
Merge tag 'dm-3.12-fixes' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device-mapper fixes from Mike Snitzer:
 "A few fixes for dm-snapshot, a 32 bit fix for dm-stats, a couple error
  handling fixes for dm-multipath.  A fix for the thin provisioning
  target to not expose non-zero discard limits if discards are disabled.

  Lastly, add two DM module parameters which allow users to tune the
  emergency memory reserves that DM mainatins per device -- this helps
  fix a long-standing issue for dm-multipath.  The conservative default
  reserve for request-based dm-multipath devices (256) has proven
  problematic for users with many multipathed SCSI devices but
  relatively little memory.  To responsibly select a smaller value users
  should use the new nr_bios tracepoint info (via commit 75afb352
  "block: Add nr_bios to block_rq_remap tracepoint") to determine the
  peak number of bios their workloads create"

* tag 'dm-3.12-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm: add reserved_bio_based_ios module parameter
  dm: add reserved_rq_based_ios module parameter
  dm: lower bio-based mempool reservation
  dm thin: do not expose non-zero discard limits if discards disabled
  dm mpath: disable WRITE SAME if it fails
  dm-snapshot: fix performance degradation due to small hash size
  dm snapshot: workaround for a false positive lockdep warning
  dm stats: fix possible counter corruption on 32-bit systems
  dm mpath: do not fail path on -ENOSPC

11 years agoMerge git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Wed, 25 Sep 2013 21:56:39 +0000 (14:56 -0700)]
Merge git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fixes from Ralf Baechle:

 - Fix a comment

 - A small cleanup the main purpose of which is to work around an
   internal compiler error bug in certain Codesource toolchains.

* git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: mm: Move some checks out of 'for' loop in DMA operations
  MIPS: cpu-features.h: s/MIPS53/MIPS64/

11 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Wed, 25 Sep 2013 21:53:55 +0000 (14:53 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

Pull powerpc fixes from Ben Herrenschmidt:
 "Here are a few things for -rc2, this time it's all written by me so it
  can only be perfect .... right ? :)

  So we have the fix to call irq_enter/exit on the irq stack we've been
  discussing, plus a cleanup on top to remove an unused (and broken)
  stack limit tracking feature (well, make it 32-bit only in fact where
  it is used and works properly).

  Then we have two things that I wrote over the last couple of days and
  made the executive decision to include just because I can (and I'm
  sure you won't object .... right ?).

  They fix a couple of annoying and long standing "issues":

   - We had separate zImages for when booting via Open Firmware vs.
     booting via a flat device-tree, while it's trivial to make one that
     deals with both

   - We wasted a ton of cycles spinning secondary CPUs uselessly at boot
     instead of starting them when needed on pseries, thus contributing
     significantly to global warming"

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc/pseries: Do not start secondaries in Open Firmware
  powerpc/zImage: make the "OF" wrapper support ePAPR boot
  powerpc: Remove ksp_limit on ppc64
  powerpc/irq: Run softirqs off the top of the irq stack

11 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 25 Sep 2013 20:29:18 +0000 (13:29 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "An EFI fix and two reboot-quirk fixes"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/reboot: Fix apparent cut-n-paste mistake in Dell reboot workaround
  x86/reboot: Add quirk to make Dell C6100 use reboot=pci automatically
  x86, efi: Don't map Boot Services on i386

11 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 25 Sep 2013 20:28:45 +0000 (13:28 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Three small fixes"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/balancing: Fix cfs_rq->task_h_load calculation
  sched/balancing: Fix 'local->avg_load > busiest->avg_load' case in fix_small_imbalance()
  sched/balancing: Fix 'local->avg_load > sds->avg_load' case in calculate_imbalance()

11 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 25 Sep 2013 20:28:08 +0000 (13:28 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Assorted standalone fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/intel: Add model number for Avoton Silvermont
  perf: Fix capabilities bitfield compatibility in 'struct perf_event_mmap_page'
  perf/x86/intel/uncore: Don't use smp_processor_id() in validate_group()
  perf: Update ABI comment
  tools lib lk: Uninclude linux/magic.h in debugfs.c
  perf tools: Fix old GCC build error in trace-event-parse.c:parse_proc_kallsyms()
  perf probe: Fix finder to find lines of given function
  perf session: Check for SIGINT in more loops
  perf tools: Fix compile with libelf without get_phdrnum
  perf tools: Fix buildid cache handling of kallsyms with kcore
  perf annotate: Fix objdump line parsing offset validation
  perf tools: Fill in new definitions for madvise()/mmap() flags
  perf tools: Sharpen the libaudit dependencies test

11 years agoupdate contact information for Mikael Pettersson
Mikael Pettersson [Wed, 25 Sep 2013 20:21:55 +0000 (22:21 +0200)]
update contact information for Mikael Pettersson

My old @it.uu.se email address is going away, so update relevant
files to point to my @gmail.com address instead.  In sata_promise.c
just delete the address, people can get it from MAINTAINERS.

Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMIPS: mm: Move some checks out of 'for' loop in DMA operations
Jayachandran C [Wed, 25 Sep 2013 13:01:05 +0000 (18:31 +0530)]
MIPS: mm: Move some checks out of 'for' loop in DMA operations

The check cpu_needs_post_dma_flush() in mips_dma_sync_sg_for_cpu() and
the check !plat_device_is_coherent() in mips_dma_sync_sg_for_device()
can be moved outside the for loop.

As a side effect, this also avoids a GCC bug that caused kernel compile
to fail with the error:

arch/mips/mm/dma-default.c: In function 'mips_dma_sync_sg_for_cpu':
arch/mips/mm/dma-default.c:316:1: internal compiler error: in add_insn_before, at emit-rtl.c:3852

This gcc failure is seen in Code Sourcery toolchains [e.g. gcc version
4.7.2 (Sourcery CodeBench Lite 2012.09-99)] after commit "MIPS: Optimize
current_cpu_type() for better code."

Signed-off-by: Jayachandran C <jchandra@broadcom.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/5907/
Reviewed-by: Markos Chandras <markos.chandras@imgtec.com>
Tested-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
11 years agoxen/spinlock: Document the xen_nopvspin parameter.
Konrad Rzeszutek Wilk [Wed, 25 Sep 2013 14:07:20 +0000 (10:07 -0400)]
xen/spinlock: Document the xen_nopvspin parameter.

Which disables in the ticketlock slowpath the Xen PV optimization's.
Useful for diagnosing issues and comparing benchmarks in
over-commit CPU scenarios.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
11 years agoxen/p2m: check MFN is in range before using the m2p table
David Vrabel [Fri, 13 Sep 2013 14:13:30 +0000 (15:13 +0100)]
xen/p2m: check MFN is in range before using the m2p table

On hosts with more than 168 GB of memory, a 32-bit guest may attempt
to grant map an MFN that is error cannot lookup in its mapping of the
m2p table.  There is an m2p lookup as part of m2p_add_override() and
m2p_remove_override().  The lookup falls off the end of the mapped
portion of the m2p and (because the mapping is at the highest virtual
address) wraps around and the lookup causes a fault on what appears to
be a user space address.

do_page_fault() (thinking it's a fault to a userspace address), tries
to lock mm->mmap_sem.  If the gntdev device is used for the grant map,
m2p_add_override() is called from from gnttab_mmap() with mm->mmap_sem
already locked.  do_page_fault() then deadlocks.

The deadlock would most commonly occur when a 64-bit guest is started
and xenconsoled attempts to grant map its console ring.

Introduce mfn_to_pfn_no_overrides() which checks the MFN is within the
mapped portion of the m2p table before accessing the table and use
this in m2p_add_override(), m2p_remove_override(), and mfn_to_pfn()
(which already had the correct range check).

All faults caused by accessing the non-existant parts of the m2p are
thus within the kernel address space and exception_fixup() is called
without trying to lock mm->mmap_sem.

This means that for MFNs that are outside the mapped range of the m2p
then mfn_to_pfn() will always look in the m2p overrides.  This is
correct because it must be a foreign MFN (and the PFN in the m2p in
this case is only relevant for the other domain).

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Cc: Stefano Stabellini <stefano.stabellini@citrix.com>
Cc: Jan Beulich <JBeulich@suse.com>
--
v3: check for auto_translated_physmap in mfn_to_pfn_no_overrides()
v2: in mfn_to_pfn() look in m2p_overrides if the MFN is out of
    range as it's probably foreign.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
11 years agox86/reboot: Fix apparent cut-n-paste mistake in Dell reboot workaround
Dave Jones [Wed, 25 Sep 2013 00:13:44 +0000 (20:13 -0400)]
x86/reboot: Fix apparent cut-n-paste mistake in Dell reboot workaround

This seems to have been copied from the Optiplex 990 entry
above, but somoene forgot to change the ident text.

Signed-off-by: Dave Jones <davej@fedoraproject.org>
Link: http://lkml.kernel.org/r/20130925001344.GA13554@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agopowerpc/pseries: Do not start secondaries in Open Firmware
Benjamin Herrenschmidt [Wed, 25 Sep 2013 04:02:50 +0000 (14:02 +1000)]
powerpc/pseries: Do not start secondaries in Open Firmware

Starting secondary CPUs early on from Open Firmware and placing them
in a holding spin loop slows down the boot process significantly under
some hypervisors such as KVM.

This is also unnecessary when RTAS supports querying the CPU state

So let's not do it.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
11 years agopowerpc/zImage: make the "OF" wrapper support ePAPR boot
Benjamin Herrenschmidt [Tue, 24 Sep 2013 06:10:38 +0000 (16:10 +1000)]
powerpc/zImage: make the "OF" wrapper support ePAPR boot

This makes the "OF" zImage wrapper (zImage.pseries, zImage.pmac,
zImage.maple) work if booted via a flat device-tree (ePAPR boot
mode), and thus potentially usable with kexec.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
11 years agopowerpc: Remove ksp_limit on ppc64
Benjamin Herrenschmidt [Tue, 24 Sep 2013 05:17:21 +0000 (15:17 +1000)]
powerpc: Remove ksp_limit on ppc64

We've been keeping that field in thread_struct for a while, it contains
the "limit" of the current stack pointer and is meant to be used for
detecting stack overflows.

It has a few problems however:

 - First, it was never actually *used* on 64-bit. Set and updated but
not actually exploited

 - When switching stack to/from irq and softirq stacks, it's update
is racy unless we hard disable interrupts, which is costly. This
is fine on 32-bit as we don't soft-disable there but not on 64-bit.

Thus rather than fixing 2 in order to implement 1 in some hypothetical
future, let's remove the code completely from 64-bit. In order to avoid
a clutter of ifdef's, we remove the updates from C code completely
during interrupt stack switching, and instead maintain it from the
asm helper that is used to do the stack switching in the first place.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
11 years agopowerpc/irq: Run softirqs off the top of the irq stack
Benjamin Herrenschmidt [Mon, 23 Sep 2013 04:29:11 +0000 (14:29 +1000)]
powerpc/irq: Run softirqs off the top of the irq stack

Nowadays, irq_exit() calls __do_softirq() pretty much directly
instead of calling do_softirq() which switches to the decicated
softirq stack.

This has lead to observed stack overflows on powerpc since we call
irq_enter() and irq_exit() outside of the scope that switches to
the irq stack.

This fixes it by moving the stack switching up a level, making
irq_enter() and irq_exit() run off the irq stack.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
11 years agomm: Place preemption point in do_mlockall() loop
Paul E. McKenney [Wed, 25 Sep 2013 01:29:11 +0000 (18:29 -0700)]
mm: Place preemption point in do_mlockall() loop

There is a loop in do_mlockall() that lacks a preemption point, which
means that the following can happen on non-preemptible builds of the
kernel. Dave Jones reports:

 "My fuzz tester keeps hitting this.  Every instance shows the non-irq
  stack came in from mlockall.  I'm only seeing this on one box, but
  that has more ram (8gb) than my other machines, which might explain
  it.

    INFO: rcu_preempt self-detected stall on CPU { 3}  (t=6500 jiffies g=470344 c=470343 q=0)
    sending NMI to all CPUs:
    NMI backtrace for cpu 3
    CPU: 3 PID: 29664 Comm: trinity-child2 Not tainted 3.11.0-rc1+ #32
    Call Trace:
      lru_add_drain_all+0x15/0x20
      SyS_mlockall+0xa5/0x1a0
      tracesys+0xdd/0xe2"

This commit addresses this problem by inserting the required preemption
point.

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge branch 'akpm' (patches from Andrew Morton)
Linus Torvalds [Wed, 25 Sep 2013 00:00:35 +0000 (17:00 -0700)]
Merge branch 'akpm' (patches from Andrew Morton)

Merge fixes from Andrew Morton:
 "Bunch of fixes.

  And a reversion of mhocko's "Soft limit rework" patch series.  This is
  actually your fault for opening the merge window when I was off racing ;)

  I didn't read the email thread before sending everything off.
  Johannes Weiner raised significant issues:

    http://www.spinics.net/lists/cgroups/msg08813.html

  and we agreed to back it all out"

I clearly need to be more aware of Andrew's racing schedule.

* akpm:
  MAINTAINERS: update mach-bcm related email address
  checkpatch: make extern in .h prototypes quieter
  cciss: fix info leak in cciss_ioctl32_passthru()
  cpqarray: fix info leak in ida_locked_ioctl()
  kernel/reboot.c: re-enable the function of variable reboot_default
  audit: fix endless wait in audit_log_start()
  revert "memcg, vmscan: integrate soft reclaim tighter with zone shrinking code"
  revert "memcg: get rid of soft-limit tree infrastructure"
  revert "vmscan, memcg: do softlimit reclaim also for targeted reclaim"
  revert "memcg: enhance memcg iterator to support predicates"
  revert "memcg: track children in soft limit excess to improve soft limit"
  revert "memcg, vmscan: do not attempt soft limit reclaim if it would not scan anything"
  revert "memcg: track all children over limit in the root"
  revert "memcg, vmscan: do not fall into reclaim-all pass too quickly"
  fs/ocfs2/super.c: use a bigger nodestr in ocfs2_dismount_volume
  watchdog: update watchdog_thresh properly
  watchdog: update watchdog attributes atomically

11 years agoMAINTAINERS: update mach-bcm related email address
Christian Daudt [Tue, 24 Sep 2013 22:27:47 +0000 (15:27 -0700)]
MAINTAINERS: update mach-bcm related email address

Update email address on mach-bcm + drivers for Broadcom mobile SoCs.

Signed-off-by: Christian Daudt <csd@broadcom.com>
Cc: Olof Johansson <olof@lixom.net>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Stephen Warren <swarren@wwwdotorg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agocheckpatch: make extern in .h prototypes quieter
Joe Perches [Tue, 24 Sep 2013 22:27:46 +0000 (15:27 -0700)]
checkpatch: make extern in .h prototypes quieter

The use of extern in .h files is a bit contentious.

Make the warning be emitted only when --strict is used on the command
line.

Signed-off-by: Joe Perches <joe@perches.com>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agocciss: fix info leak in cciss_ioctl32_passthru()
Dan Carpenter [Tue, 24 Sep 2013 22:27:45 +0000 (15:27 -0700)]
cciss: fix info leak in cciss_ioctl32_passthru()

The arg64 struct has a hole after ->buf_size which isn't cleared.  Or if
any of the calls to copy_from_user() fail then that would cause an
information leak as well.

This was assigned CVE-2013-2147.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Mike Miller <mike.miller@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agocpqarray: fix info leak in ida_locked_ioctl()
Dan Carpenter [Tue, 24 Sep 2013 22:27:44 +0000 (15:27 -0700)]
cpqarray: fix info leak in ida_locked_ioctl()

The pciinfo struct has a two byte hole after ->dev_fn so stack
information could be leaked to the user.

This was assigned CVE-2013-2147.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Mike Miller <mike.miller@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/reboot.c: re-enable the function of variable reboot_default
Chuansheng Liu [Tue, 24 Sep 2013 22:27:43 +0000 (15:27 -0700)]
kernel/reboot.c: re-enable the function of variable reboot_default

Commit 1b3a5d02ee07 ("reboot: move arch/x86 reboot= handling to generic
kernel") did some cleanup for reboot= command line, but it made the
reboot_default inoperative.

The default value of variable reboot_default should be 1, and if command
line reboot= is not set, system will use the default reboot mode.

[akpm@linux-foundation.org: fix comment layout]
Signed-off-by: Li Fei <fei.li@intel.com>
Signed-off-by: liu chuansheng <chuansheng.liu@intel.com>
Acked-by: Robin Holt <robinmholt@linux.com>
Cc: <stable@vger.kernel.org> [3.11.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoaudit: fix endless wait in audit_log_start()
Konstantin Khlebnikov [Tue, 24 Sep 2013 22:27:42 +0000 (15:27 -0700)]
audit: fix endless wait in audit_log_start()

After commit 829199197a43 ("kernel/audit.c: avoid negative sleep
durations") audit emitters will block forever if userspace daemon cannot
handle backlog.

After the timeout the waiting loop turns into busy loop and runs until
daemon dies or returns back to work.  This is a minimal patch for that
bug.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Richard Guy Briggs <rgb@redhat.com>
Cc: Eric Paris <eparis@redhat.com>
Cc: Chuck Anderson <chuck.anderson@oracle.com>
Cc: Dan Duval <dan.duval@oracle.com>
Cc: Dave Kleikamp <dave.kleikamp@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "memcg, vmscan: integrate soft reclaim tighter with zone shrinking code"
Andrew Morton [Tue, 24 Sep 2013 22:27:41 +0000 (15:27 -0700)]
revert "memcg, vmscan: integrate soft reclaim tighter with zone shrinking code"

Revert commit 3b38722efd9f ("memcg, vmscan: integrate soft reclaim
tighter with zone shrinking code")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "memcg: get rid of soft-limit tree infrastructure"
Andrew Morton [Tue, 24 Sep 2013 22:27:40 +0000 (15:27 -0700)]
revert "memcg: get rid of soft-limit tree infrastructure"

Revert commit e883110aad71 ("memcg: get rid of soft-limit tree
infrastructure")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "vmscan, memcg: do softlimit reclaim also for targeted reclaim"
Andrew Morton [Tue, 24 Sep 2013 22:27:38 +0000 (15:27 -0700)]
revert "vmscan, memcg: do softlimit reclaim also for targeted reclaim"

Revert commit a5b7c87f9207 ("vmscan, memcg: do softlimit reclaim also
for targeted reclaim")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "memcg: enhance memcg iterator to support predicates"
Andrew Morton [Tue, 24 Sep 2013 22:27:37 +0000 (15:27 -0700)]
revert "memcg: enhance memcg iterator to support predicates"

Revert commit de57780dc659 ("memcg: enhance memcg iterator to support
predicates")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "memcg: track children in soft limit excess to improve soft limit"
Andrew Morton [Tue, 24 Sep 2013 22:27:36 +0000 (15:27 -0700)]
revert "memcg: track children in soft limit excess to improve soft limit"

Revert commit 7d910c054be4 ("memcg: track children in soft limit excess
to improve soft limit")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "memcg, vmscan: do not attempt soft limit reclaim if it would not scan anything"
Andrew Morton [Tue, 24 Sep 2013 22:27:35 +0000 (15:27 -0700)]
revert "memcg, vmscan: do not attempt soft limit reclaim if it would not scan anything"

Revert commit e839b6a1c8d0 ("memcg, vmscan: do not attempt soft limit
reclaim if it would not scan anything")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "memcg: track all children over limit in the root"
Andrew Morton [Tue, 24 Sep 2013 22:27:34 +0000 (15:27 -0700)]
revert "memcg: track all children over limit in the root"

Revert commit 1be171d60bdd ("memcg: track all children over limit in the
root")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "memcg, vmscan: do not fall into reclaim-all pass too quickly"
Andrew Morton [Tue, 24 Sep 2013 22:27:33 +0000 (15:27 -0700)]
revert "memcg, vmscan: do not fall into reclaim-all pass too quickly"

Revert commit e975de998b96 ("memcg, vmscan: do not fall into reclaim-all
pass too quickly")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agofs/ocfs2/super.c: use a bigger nodestr in ocfs2_dismount_volume
Goldwyn Rodrigues [Tue, 24 Sep 2013 22:27:32 +0000 (15:27 -0700)]
fs/ocfs2/super.c: use a bigger nodestr in ocfs2_dismount_volume

While printing 32-bit node numbers, an 8-byte string is not enough.
Increase the size of the string to 12 chars.

This got left out in commit 49fa8140e487 ("fs/ocfs2/super.c: Use bigger
nodestr to accomodate 32-bit node numbers").

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agowatchdog: update watchdog_thresh properly
Michal Hocko [Tue, 24 Sep 2013 22:27:30 +0000 (15:27 -0700)]
watchdog: update watchdog_thresh properly

watchdog_tresh controls how often nmi perf event counter checks per-cpu
hrtimer_interrupts counter and blows up if the counter hasn't changed
since the last check.  The counter is updated by per-cpu
watchdog_hrtimer hrtimer which is scheduled with 2/5 watchdog_thresh
period which guarantees that hrtimer is scheduled 2 times per the main
period.  Both hrtimer and perf event are started together when the
watchdog is enabled.

So far so good.  But...

But what happens when watchdog_thresh is updated from sysctl handler?

proc_dowatchdog will set a new sampling period and hrtimer callback
(watchdog_timer_fn) will use the new value in the next round.  The
problem, however, is that nobody tells the perf event that the sampling
period has changed so it is ticking with the period configured when it
has been set up.

This might result in an ear ripping dissonance between perf and hrtimer
parts if the watchdog_thresh is increased.  And even worse it might lead
to KABOOM if the watchdog is configured to panic on such a spurious
lockup.

This patch fixes the issue by updating both nmi perf even counter and
hrtimers if the threshold value has changed.

The nmi one is disabled and then reinitialized from scratch.  This has
an unpleasant side effect that the allocation of the new event might
fail theoretically so the hard lockup detector would be disabled for
such cpus.  On the other hand such a memory allocation failure is very
unlikely because the original event is deallocated right before.

It would be much nicer if we just changed perf event period but there
doesn't seem to be any API to do that right now.  It is also unfortunate
that perf_event_alloc uses GFP_KERNEL allocation unconditionally so we
cannot use on_each_cpu() and do the same thing from the per-cpu context.
The update from the current CPU should be safe because
perf_event_disable removes the event atomically before it clears the
per-cpu watchdog_ev so it cannot change anything under running handler
feet.

The hrtimer is simply restarted (thanks to Don Zickus who has pointed
this out) if it is queued because we cannot rely it will fire&adopt to
the new sampling period before a new nmi event triggers (when the
treshold is decreased).

[akpm@linux-foundation.org: the UP version of __smp_call_function_single ended up in the wrong place]
Signed-off-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agowatchdog: update watchdog attributes atomically
Michal Hocko [Tue, 24 Sep 2013 22:27:29 +0000 (15:27 -0700)]
watchdog: update watchdog attributes atomically

proc_dowatchdog doesn't synchronize multiple callers which might lead to
confusion when two parallel callers might confuse watchdog_enable_all_cpus
resp watchdog_disable_all_cpus (eg watchdog gets enabled even if
watchdog_thresh was set to 0 already).

This patch adds a local mutex which synchronizes callers to the sysctl
handler.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Don Zickus <dzickus@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge branch 'bcache' (bcache fixes from Kent Overstreet)
Linus Torvalds [Tue, 24 Sep 2013 21:42:03 +0000 (14:42 -0700)]
Merge branch 'bcache' (bcache fixes from Kent Overstreet)

Merge bcache fixes from Kent Overstreet:
 "There's fixes for _three_ different data corruption bugs, all of which
  were found by users hitting them in the wild.

  The first one isn't bcache specific - in 3.11 bcache was switched to
  the bio_copy_data in fs/bio.c, and that's when the bug in that code
  was discovered, but it's also used by raid1 and pktcdvd.  (That was my
  code too, so the bug's doubly embarassing given that it was or
  should've been just a cut and paste from bcache code.  Dunno what
  happened there).

  Most of these (all the non data corruption bugs, actually) were ready
  before the merge window and have been sitting in Jens' tree, but I
  don't know what's been up with him lately..."

* emailed patches from Kent Overstreet <kmo@daterainc.com>:
  bcache: Fix flushes in writeback mode
  bcache: Fix for handling overlapping extents when reading in a btree node
  bcache: Fix a shrinker deadlock
  bcache: Fix a dumb CPU spinning bug in writeback
  bcache: Fix a flush/fua performance bug
  bcache: Fix a writeback performance regression
  bcache: Correct printf()-style format length modifier
  bcache: Fix for when no journal entries are found
  bcache: Strip endline when writing the label through sysfs
  bcache: Fix a dumb journal discard bug
  block: Fix bio_copy_data()

11 years agobcache: Fix flushes in writeback mode
Kent Overstreet [Tue, 24 Sep 2013 06:17:36 +0000 (23:17 -0700)]
bcache: Fix flushes in writeback mode

In writeback mode, when we get a cache flush we need to make sure we
issue a flush to the backing device.

The code for sending down an extra flush was wrong - by cloning the bio
we were probably getting flags that didn't make sense for a bare flush,
and also the old code was firing for FUA bios, for which we don't need
to send a flush to the backing device.

This was causing data corruption somehow - the mechanism was never
determined, but this patch fixes it for the users that were seeing it.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agobcache: Fix for handling overlapping extents when reading in a btree node
Kent Overstreet [Tue, 24 Sep 2013 06:17:35 +0000 (23:17 -0700)]
bcache: Fix for handling overlapping extents when reading in a btree node

btree_sort_fixup() was overly clever, because it was trying to avoid
pulling a key off the btree iterator in more than one place.

This led to a really obscure bug where we'd break early from the loop in
btree_sort_fixup() if the current key overlapped with keys in more than
one older set, and the next key it overlapped with was zero size.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agobcache: Fix a shrinker deadlock
Kent Overstreet [Tue, 24 Sep 2013 06:17:34 +0000 (23:17 -0700)]
bcache: Fix a shrinker deadlock

GFP_NOIO means we could be getting called recursively - mca_alloc() ->
mca_data_alloc() - definitely can't use mutex_lock(bucket_lock) then.
Whoops.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agobcache: Fix a dumb CPU spinning bug in writeback
Kent Overstreet [Tue, 24 Sep 2013 06:17:33 +0000 (23:17 -0700)]
bcache: Fix a dumb CPU spinning bug in writeback

schedule_timeout() != schedule_timeout_uninterruptible()

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agobcache: Fix a flush/fua performance bug
Kent Overstreet [Tue, 24 Sep 2013 06:17:32 +0000 (23:17 -0700)]
bcache: Fix a flush/fua performance bug

bch_journal_meta() was missing the flush to make the journal write
actually go down (instead of waiting up to journal_delay_ms)...

Whoops

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agobcache: Fix a writeback performance regression
Kent Overstreet [Tue, 24 Sep 2013 06:17:31 +0000 (23:17 -0700)]
bcache: Fix a writeback performance regression

Background writeback works by scanning the btree for dirty data and
adding those keys into a fixed size buffer, then for each dirty key in
the keybuf writing it to the backing device.

When read_dirty() finishes and it's time to scan for more dirty data, we
need to wait for the outstanding writeback IO to finish - they still
take up slots in the keybuf (so that foreground writes can check for
them to avoid races) - without that wait, we'll continually rescan when
we'll be able to add at most a key or two to the keybuf, and that takes
locks that starves foreground IO.  Doh.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>