Merge branch 'fixes' into next

mmc: meson-gx: replace WARN_ONCE with dev_warn_once about scatterlist size alignment in block mode

Since commit e085b51c74cc ("mmc: meson-gx: check for scatterlist size alignment in block mode"),
support for SDIO SD_IO_RW_EXTENDED transferts are properly filtered but some driver
like brcmfmac still gives a block sg buffer size not aligned with SDIO block,
triggerring a WARN_ONCE() with scary stacktrace even if the transfer works fine
but with possible degraded performances.

Simply replace with dev_warn_once() to inform user this should be fixed to avoid
degraded performance.

This should be ultimately fixed in brcmfmac, but since it's only a performance issue
the warning should be removed.

Fixes: e085b51c74cc ("mmc: meson-gx: check for scatterlist size alignment in block mode")
Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Link: https://lore.kernel.org/r/20210416094347.2015896-1-narmstrong@baylibre.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: dw_mmc-rockchip: Just set default sample value for legacy mode

.set_ios() is called from .resume() as well. For SDIO device which sets
keep-power-in-suspend, nothing should be changed after resuming, as well
as sample tuning value, since this value is tuned already. So we should
not overwrite it with the default value.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Link: https://lore.kernel.org/r/1618539454-182170-1-git-send-email-shawn.lin@rock-chips.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-s3c: constify uses of driver/match data

The driver data (struct sdhci_s3c_drv_data) stored in of_device_id
table is allocated as const and used only in const-way.  Skip
unnecessary const-away casts and convert all users to work with pointer
to const.  This is both more logical and safer.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Reviewed-by: Sylwester Nawrocki <snawrocki@kernel.org>
Link: https://lore.kernel.org/r/20210415084412.51125-3-krzysztof.kozlowski@canonical.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-s3c: correct kerneldoc of sdhci_s3c_drv_data

Correct the name of sdhci_s3c_drv_data structure in kerneldoc:

  drivers/mmc/host/sdhci-s3c.c:143: warning:
    expecting prototype for struct sdhci_s3c_driver_data. Prototype was for struct sdhci_s3c_drv_data instead

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Reviewed-by: Sylwester Nawrocki <snawrocki@kernel.org>
Link: https://lore.kernel.org/r/20210415084412.51125-2-krzysztof.kozlowski@canonical.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-s3c: simplify getting of_device_id match data

Use of_device_get_match_data() to make the code slightly smaller and to
remove the of_device_id table forward declaration.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Reviewed-by: Sylwester Nawrocki <snawrocki@kernel.org>
Link: https://lore.kernel.org/r/20210415084412.51125-1-krzysztof.kozlowski@canonical.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: tmio: always restore irq register

Currently, only SDHI on R-Car Gen2+ reinitializes the irq register
during reset but it should be done on all instances. We can move it from
the SDHI driver to the TMIO core, because we now have the
'sd_irq_mask_all' variable which carries the proper value to use. That
also means we can remove the initialization from tmio_mmc_probe()
because it calls tmio_mmc_reset(), too. We only move that
tmio_mmc_reset() call there a little to ensure 'sd_irq_mask_all' is
properly set.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/20210413083137.11171-1-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-pci-gli: Enlarge ASPM L1 entry delay of GL975x

GL975x enters ASPM L1 state after a short idle in default.
Enlarge the idle period to 7.9us for improving the R/W performance.

Signed-off-by: Ben Chuang <benchuanggli@gmail.com>
Link: https://lore.kernel.org/r/20210415032637.5219-1-benchuanggli@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Let eMMC sanitize not retry in case of timeout/failure

Not all commands need to be retried in case of a timeout/failure. This is
the case for the eMMC sanitize command, for example, which is issued
through the ioctl interface. More precisely, in case of timeout, retrying
could make the user wait for a very long time as each retry loop could last
for a couple of minutes.

Therefore, let's not use retries for the eMMC sanitize command.

Signed-off-by: Bean Huo <beanhuo@micron.com>
Link: https://lore.kernel.org/r/20210414212236.346813-3-huobean@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Add a retries parameter to __mmc_switch function

Add command retries parameter to __mmc_switch(), let caller
pass retries according to the caller's condition.

Signed-off-by: Bean Huo <beanhuo@micron.com>
Link: https://lore.kernel.org/r/20210414212236.346813-2-huobean@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

memstick: r592: remove unused variable

Fix the following clang warning:

drivers/memstick/host/r592.c:363:6: warning: variable ‘len’ set but not
used [-Wunused-but-set-variable].

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Link: https://lore.kernel.org/r/1618366903-94346-1-git-send-email-jiapeng.chong@linux.alibaba.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-st: Remove unnecessary error log

devm_ioremap_resource() has recorded error log, so it's
unnecessary to record log again.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Laibin Qiu <qiulaibin@huawei.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Link: https://lore.kernel.org/r/20210409023801.3326572-1-qiulaibin@huawei.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-msm: Remove unnecessary error log

devm_ioremap_resource() has recorded error log, so it's
unnecessary to record log again.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jia Yang <jiayang5@huawei.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20210409015424.3277212-1-jiayang5@huawei.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: owl-mmc: Remove unnecessary error log

devm_ioremap_resource() has recorded error log, so it's
unnecessary to record log again.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Laibin Qiu <qiulaibin@huawei.com>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Link: https://lore.kernel.org/r/20210409023349.3325907-1-qiulaibin@huawei.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: renesas_sdhi: enable WAIT_WHILE_BUSY

Now that we got the timeout handling in the driver correct, we can use
this capability to avoid polling via the MMC core.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/20210409094606.4317-1-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-tegra: Add required callbacks to set/clear CQE_EN bit

CMD8 is not supported with Command Queue Enabled. Add required callback
to clear CQE_EN and CQE_INTR fields in the host controller register
before sending CMD8. Add corresponding callback in the CQHCI resume path
to re-enable CQE_EN and CQE_INTR fields.

Reported-by: Kamal Mostafa <kamal@canonical.com>
Tested-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Aniruddha Tvs Rao <anrao@nvidia.com>
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20210407094617.770495-1-jonathanh@nvidia.com
Cc: stable@vger.kernel.org # v5.10+
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-pci-gli: Improve GL9763E L1 entry delay to increase battery life

For GL9763E, although there is the best performance at the maximum delay.
Change the value to 20us in order to have better power consumption.
This change may reduce the maximum performance by 10%.

Signed-off-by: Ben Chuang <ben.chuang@genesyslogic.com.tw>
Link: https://lore.kernel.org/r/20210407093816.8863-1-benchuanggli@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

dt-bindings: mmc: iproc-sdhci: Convert to json-schema

Convert the brcm,iproc-sdhci binding to DT schema format using json-schema

Signed-off-by: Nicolas Saenz Julienne <nsaenz@kernel.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20210406104802.20898-2-nsaenz@kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: moxart: Remove unused variable 'dma_time' and 'pio_time'

Fixes the following W=1 kernel build warning(s):

drivers/mmc/host/moxart-mmc.c:257:7: warning: variable ‘dma_time’ set
but not used
drivers/mmc/host/moxart-mmc.c:395:7: warning: variable ‘pio_time’ set
but not used

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Link: https://lore.kernel.org/r/1617357231-93064-1-git-send-email-yang.lee@linux.alibaba.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Use userland specified timeout value for eMMC sanitize

As the density increases, the 4-minute timeout value for sanitize is no
longer feasible. At the same time, devices of different densities have
different timeout values, which makes it difficult to use a common timeout
value. Therefore, let's pass down the userland-specified sanitize timeout
value so it can be used.

Signed-off-by: Bean Huo <beanhuo@micron.com>
Link: https://lore.kernel.org/r/20210402092432.25069-2-huobean@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-pci: Fix initialization of some SD cards for Intel BYT-based controllers

Bus power may control card power, but the full reset done by SDHCI at
initialization still may not reset the power, whereas a direct write to
SDHCI_POWER_CONTROL can. That might be needed to initialize correctly, if
the card was left powered on previously.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210331081752.23621-1-adrian.hunter@intel.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: dw_mmc-k3: use the correct HiSilicon copyright

s/Hisilicon/HiSilicon/g.
It should use capital S, according to
https://www.hisilicon.com/en/terms-of-use.

Signed-off-by: Hao Fang <fanghao11@huawei.com>
Acked-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Link: https://lore.kernel.org/r/1617086448-51054-1-git-send-email-fanghao11@huawei.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-esdhc-imx: separate 100/200 MHz pinctrl states check

As indicated by function esdhc_change_pinstate(), SDR50 and DDR50
require pins_100mhz, while SDR104 and HS400 require pins_200mhz.  Some
system design may support SDR50 and DDR50 with 100mhz pin state only
(without 200mhz one).  Currently the combined 100/200 MHz pinctrl state
check prevents such system from running SDR50 and DDR50.  Separate the
check to support such system design.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Reviewed-by: Haibo Chen <haibo.chen@nxp.com>
Link: https://lore.kernel.org/r/20210326110214.28416-1-shawnguo@kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-brcmstb: Remove CQE quirk

Remove the CQHCI_QUIRK_SHORT_TXFR_DESC_SZ quirk because the
latest chips have this fixed and earlier chips have other
CQE problems that prevent the feature from being enabled.

Signed-off-by: Al Cooper <alcooperx@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20210325192834.42955-1-alcooperx@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci: replace mmc->parent with mmc_dev() for consistency

As pointed out by Ulf, "both "mmc->parent" and mmc_dev(mmc) are being
used in the entire c-file". Convert all the mmc->parent usage in all
sdhci host driver to mmc_dev() for consistency.

Suggested-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Link: https://lore.kernel.org/r/20210324155013.1e5faa3c@xhacker.debian
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-of-dwcmshc: set MMC_CAP_WAIT_WHILE_BUSY

The host supports HW busy detection of the device busy signaling over
dat0 line. Set MMC_CAP_wAIT_WHILE_BUSY host capability.

Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Link: https://lore.kernel.org/r/20210324154703.69f97fde@xhacker.debian
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-of-dwcmshc: fix error return code in dwcmshc_probe()

Fix to return negative error code -ENOMEM from the error handling
case instead of 0, as done elsewhere in this function.

Fixes: c2c4da37837e ("mmc: sdhci-of-dwcmshc: add rockchip platform support")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Link: https://lore.kernel.org/r/20210323112956.1016884-1-weiyongjun1@huawei.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-of-dwcmshc: add ACPI support for BlueField-3 SoC

This commit adds ACPI support in the sdhci-of-dwcmshc driver for
BlueField-3 SoC. It has changes to only use the clock hierarchy
for Deviec Tree since the clk is not supported by ACPI. Instead,
ACPI can define 'clock-frequency' which is parsed by existing
sdhci_get_property(). This clock value will be returned in function
dwcmshc_get_max_clock().

Signed-off-by: Liming Sun <limings@nvidia.com>
Reviewed-by: Khalil Blaiech <kblaiech@nvidia.com>
Link: https://lore.kernel.org/r/1616453211-275165-1-git-send-email-limings@nvidia.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-pci: Add PCI IDs for Intel LKF

Add PCI IDs for Intel LKF eMMC and SD card host controllers.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20210322055356.24923-1-adrian.hunter@intel.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: tmio: always flag retune when resetting and a card is present

After reset, we manually flagged retune in runtime resume, but missed it
in the workqueue. To fix that and avoid the problem in the future, let's
flag retune in the reset handler directly whenever a card is present.

Reported-by: Takeshi Saito <takeshi.saito.xv@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20210316085717.7276-3-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: tmio: restore bus width when resetting

Resetting the IP core will lose the bus width information and not all
code paths recover it. So, make sure the latest bus width gets restored
in the reset routine. For that, tmio_mmc_set_bus_width() is moved, but
not modified.

Signed-off-by: Takeshi Saito <takeshi.saito.xv@renesas.com>
[wsa: reworded commit message]
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20210316085717.7276-2-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: cavium: Use '"%s...", __func__' to print function name

Use preferred way '"%s...", __func__' to print function name.

Signed-off-by: Bean Huo <beanhuo@micron.com>
Link: https://lore.kernel.org/r/20210319121357.255176-2-huobean@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-acpi: Add device ID for the AMDI0041 variant of the AMD eMMC controller.

This variant is present on a Lenovo IdeaPad Slim 1, which uses an AMD Dali/Athlon Silver 3050e.
The Windows AMD SD Host Controller driver also lists this as a valid device ID.

Adding this device ID makes the internal eMMC storage on the Lenovo accessible.
Consequently this makes Linux installable and usable on it as well.

Signed-off-by: James Young <james@pocketfluff.org>
Link: https://lore.kernel.org/r/20210318124025.3002861-1-james@pocketfluff.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

dt-bindings: mmc: Add compatible for Mediatek MT8195

This commit adds dt-binding documentation of mmc for Mediatek MT8195 SoC
Platform.

Signed-off-by: Seiya Wang <seiya.wang@mediatek.com>
Link: https://lore.kernel.org/r/20210319023427.16711-6-seiya.wang@mediatek.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: renesas_sdhi: do hard reset if possible

All recent SDHI instances can be reset via the reset controller. If one
is found, use it instead of the open coded reset. This is to get a
future-proof sane reset state.

Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20210317091622.31890-4-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: renesas_sdhi: break SCC reset into own function

renesas_sdhi_reset used to mainly reset the SCC but is now doing more
and even more will be added. So, factor out SCC reset to have a clear
distinction when we want to reset either SCC or SDHI+SCC.

Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20210317091622.31890-3-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: tmio: abort DMA before reset

We will soon allow resetting the whole IP core via a reset controller.
For this case, DMA must be terminated before the actual reset. For the
other cases, it is probably better, too.

Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20210317091622.31890-2-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-of-dwcmshc: add rockchip platform support

sdhci based synopsys MMC IP is also used on some rockchip platforms,
so add a basic support here.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Link: https://lore.kernel.org/r/1615879102-45919-3-git-send-email-shawn.lin@rock-chips.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

dt-bindings: mmc: sdhci-of-dwcmhsc: Add rockchip support

This patch adds rockchip support in sdhci-of-dwcmhsc.yaml

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Link: https://lore.kernel.org/r/1615879102-45919-2-git-send-email-shawn.lin@rock-chips.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

dt-bindings: mmc: sdhci-of-dwcmhsc: Convert to yaml file

This patch converts sdhci-of-dwcmshc.txt to sdhci-of-dwcmshc.yaml

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/1615879102-45919-1-git-send-email-shawn.lin@rock-chips.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-esdhc-imx: Use device_get_match_data()

The retrieval of driver data can be a bit simplified by using
device_get_match_data(), so switch to it.

Signed-off-by: Fabio Estevam <festevam@gmail.com>
Link: https://lore.kernel.org/r/20210314230824.148969-2-festevam@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-esdhc-imx: Remove non-DT stub

i.MX has been converted to a devicetree only platform, so remove
the non-DT stub for sdhci_esdhc_imx_probe_dt().

Also, make the driver depend on OF now.

Signed-off-by: Fabio Estevam <festevam@gmail.com>
Link: https://lore.kernel.org/r/20210314230824.148969-1-festevam@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

memstick: Remove useless else branch

Remove else branch on line 334 of memstick.c, after the return of the
previous branch. Found by checkpatch.

Signed-off-by: Joey Pabalan <jpabalanb@gmail.com>
Link: https://lore.kernel.org/r/20210313173740.GA580681@joeylaptop
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Fix hanging on I/O during system suspend for removable cards

The mmc core uses a PM notifier to temporarily during system suspend, turn
off the card detection mechanism for removal/insertion of (e)MMC/SD/SDIO
cards. Additionally, the notifier may be used to remove an SDIO card
entirely, if a corresponding SDIO functional driver don't have the system
suspend/resume callbacks assigned. This behaviour has been around for a
very long time.

However, a recent bug report tells us there are problems with this
approach. More precisely, when receiving the PM_SUSPEND_PREPARE
notification, we may end up hanging on I/O to be completed, thus also
preventing the system from getting suspended.

In the end what happens, is that the cancel_delayed_work_sync() in
mmc_pm_notify() ends up waiting for mmc_rescan() to complete - and since
mmc_rescan() wants to claim the host, it needs to wait for the I/O to be
completed first.

Typically, this problem is triggered in Android, if there is ongoing I/O
while the user decides to suspend, resume and then suspend the system
again. This due to that after the resume, an mmc_rescan() work gets punted
to the workqueue, which job is to verify that the card remains inserted
after the system has resumed.

To fix this problem, userspace needs to become frozen to suspend the I/O,
prior to turning off the card detection mechanism. Therefore, let's drop
the PM notifiers for mmc subsystem altogether and rely on the card
detection to be turned off/on as a part of the system_freezable_wq, that we
are already using.

Moreover, to allow and SDIO card to be removed during system suspend, let's
manage this from a ->prepare() callback, assigned at the mmc_host_class
level. In this way, we can use the parent device (the mmc_host_class
device), to remove the card device that is the child, in the
device_prepare() phase.

Reported-by: Kiwoong Kim <kwmad.kim@samsung.com>
Cc: stable@vger.kernel.org # v4.5+
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/20210310152900.149380-1-ulf.hansson@linaro.org
Reviewed-by: Kiwoong Kim <kwmad.kim@samsung.com>

mmc: sdio: fix a typo in the comment of SDIO_SD_REV_3_00

I believe "Spev" is typo, should be "Spec".

Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Link: https://lore.kernel.org/r/20210311174157.561dada9@xhacker.debian
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci: Use "mmc" directly rather than "host->mmc"

Clean up the code to use the "mmc" directly instead of "host->mmc".
If the code sits in hot code path, this clean up also brings trvial
performance improvement. Take the sdhci_post_req() for example:

before the patch:
     ...
     8d0: a9be7bfd  stp x29, x30, [sp, #-32]!
     8d4: 910003fd  mov x29, sp
     8d8: f9000bf3  str x19, [sp, #16]
     8dc: f9400833  ldr x19, [x1, #16]
     8e0: b9404261  ldr w1, [x19, #64]
     8e4: 34000161  cbz w1, 910 <sdhci_post_req+0x50>
     8e8: f9424400  ldr x0, [x0, #1160]
     8ec: d2800004  mov x4, #0x0                    // #0
     8f0: b9401a61  ldr w1, [x19, #24]
     8f4: b9403262  ldr w2, [x19, #48]
     8f8: f9400000  ldr x0, [x0]
     8fc: f278003f  tst x1, #0x100
     900: f9401e61  ldr x1, [x19, #56]
     904: 1a9f17e3  cset w3, eq  // eq = none
     908: 11000463  add w3, w3, #0x1
     90c: 94000000  bl 0 <dma_unmap_sg_attrs>
     ...

After the patch:
     ...
     8d0: a9be7bfd  stp x29, x30, [sp, #-32]!
     8d4: 910003fd  mov x29, sp
     8d8: f9000bf3  str x19, [sp, #16]
     8dc: f9400833  ldr x19, [x1, #16]
     8e0: b9404261  ldr w1, [x19, #64]
     8e4: 34000141  cbz w1, 90c <sdhci_post_req+0x4c>
     8e8: b9401a61  ldr w1, [x19, #24]
     8ec: d2800004  mov x4, #0x0                    // #0
     8f0: b9403262  ldr w2, [x19, #48]
     8f4: f9400000  ldr x0, [x0]
     8f8: f278003f  tst x1, #0x100
     8fc: f9401e61  ldr x1, [x19, #56]
     900: 1a9f17e3  cset w3, eq  // eq = none
     904: 11000463  add w3, w3, #0x1
     908: 94000000  bl 0 <dma_unmap_sg_attrs>
     ...

We saved one ldr instruction: "ldr     x0, [x0, #1160]"

Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20210311174046.597d1951@xhacker.debian
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-pci: Avoid comma separated statements

Use semicolons.

Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20210311181432.6385cd2b@xhacker.debian
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: via-sdmmc: remove unneeded variable 'ret'

Fix the following coccicheck warning:
./drivers/mmc/host/via-sdmmc.c:1274:5-8: Unneeded variable: "ret".
Return "0" on line 1295

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Link: https://lore.kernel.org/r/1615272478-52458-1-git-send-email-yang.lee@linux.alibaba.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: dw_mmc: simplify optional reset handling

As of commit bb475230b8e5 ("reset: make optional functions really
optional"), the reset framework API calls use NULL pointers to describe
optional, non-present reset controls.

This allows to return errors from devm_reset_control_get_optional and to
call reset_control_(de)assert unconditionally.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://lore.kernel.org/r/20210305090724.18832-1-p.zabel@pengutronix.de
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-st: simplify optional reset handling

As of commit bb475230b8e5 ("reset: make optional functions really
optional"), the reset framework API calls use NULL pointers to describe
optional, non-present reset controls.

This allows to return errors from devm_reset_control_get_optional and to
call reset_control_(de)assert unconditionally.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://lore.kernel.org/r/20210305090827.19124-1-p.zabel@pengutronix.de
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

memstick: core: Assign error code of mspro_block_resume()

When mspro_block_init_card() fails, no error return code of
mspro_block_resume() is assigned/propagated. Let's fix this.

Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Link: https://lore.kernel.org/r/20210305021445.3435-1-baijiaju1990@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Remove mq->use_cqe from the struct mmc_queue

The host->cqe_enabled is already containing the needed information about
whether the CQE is enabled or not, hence there is no need to keep another
copy of it around.

Signed-off-by: Luca Porzio <lporzio@micron.com>
Signed-off-by: Zhan Liu <zliua@micron.com>
Link: https://lore.kernel.org/r/20210215003217.GA12240@lupo-laptop
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: block: Fix error path in mmc_blk_probe()

Returning zero to indicate success, when we actually have failed to probe
is wrong. As a matter of fact, it leads to that mmc_blk_remove() gets
called at a card removal and then triggers "NULL pointer dereference"
splats. This is because mmc_blk_remove() relies on data structures and
pointers to be setup from mmc_blk_probe(), of course.

There have been no errors reported about this, which is most likely because
mmc_blk_probe() never fails like this. Nevertheless, let's fix the code by
propagating the error codes correctly and prevent us from leaking memory by
calling also destroy_workqueue() in the error path.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Link: https://lore.kernel.org/r/20210303122049.151986-4-ulf.hansson@linaro.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: block: Simplify logging during probe about added partitions

To simplify the code, move the logging into the common mmc_blk_alloc_req()
and drop the rather useless information about the partition type/id.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Link: https://lore.kernel.org/r/20210303122049.151986-3-ulf.hansson@linaro.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: block: Drop use of unlikely() in mmc_blk_probe()

mmc_blk_probe() isn't a hotpath, which makes it's questionable to use
unlikely(). Therefore let's simply drop it.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Link: https://lore.kernel.org/r/20210303122049.151986-2-ulf.hansson@linaro.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: renesas_sdhi: use custom mask for TMIO_MASK_ALL

Populate the new member for custom mask values to make sure this value
is applied whenever needed. Also, rename the define holding the value
because this is not only about initialization anymore.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/20210304092903.8534-1-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci: Check for reset prior to DMA address unmap

For data read commands, SDHC may initiate data transfers even before it
completely process the command response. In case command itself fails,
driver un-maps the memory associated with data transfer but this memory
can still be accessed by SDHC for the already initiated data transfer.
This scenario can lead to un-mapped memory access error.

To avoid this scenario, reset SDHC (when command fails) prior to
un-mapping memory. Resetting SDHC ensures that all in-flight data
transfers are either aborted or completed. So we don't run into this
scenario.

Swap the reset, un-map steps sequence in sdhci_request_done().

Suggested-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org>
Signed-off-by: Pradeep P V K <pragalla@codeaurora.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1614760331-43499-1-git-send-email-pragalla@qti.qualcomm.com
Cc: stable@vger.kernel.org # v4.9+
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-pci-o2micro: Add missing checks in sdhci_pci_o2_probe

It's odd to adopt different error handling on failure of
pci_read_config_dword(). Check the return value and terminate
execution flow on failure of all pci_read_config_dword() calls
in this function.

Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Link: https://lore.kernel.org/r/20210226084146.29095-1-dinghao.liu@zju.edu.cn
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-pci-gli: Enable short circuit protection mechanism of GL9755

Short circuit protection mechanism of GL9755 is disabled by HW default
setting. Enable short circuit protection to prevent GL9755 from being
damaged by short circuit or over current.

Signed-off-by: Renius Chen <reniuschengl@gmail.com>
Link: https://lore.kernel.org/r/20210225111307.62975-1-reniuschengl@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: sdhci-esdhc-imx: validate pinctrl before use it

When imx_data->pinctrl is not a valid pointer, pinctrl_lookup_state
will trigger kernel panic.

When we boot Dual OS on Jailhouse hypervisor, we let the 1st Linux to
configure pinmux ready for the 2nd OS, so the 2nd OS not have pinctrl
settings.

Similar to this commit b62eee9f804e ("mmc: sdhci-esdhc-imx: no fail when no pinctrl available").

Reviewed-by: Bough Chen <haobo.chen@nxp.com>
Reviewed-by: Alice Guo <alice.guo@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Link: https://lore.kernel.org/r/1614222604-27066-6-git-send-email-peng.fan@oss.nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

dt-bindings: mmc: fsl-imx-esdhc: add clock bindings

Add clock bindings for fsl-imx-esdhc yaml

Signed-off-by: Peng Fan <peng.fan@nxp.com>
Link: https://lore.kernel.org/r/1614222604-27066-5-git-send-email-peng.fan@oss.nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

dt-bindings: mmc: fsl-imx-esdhc: add pinctrl bindings

Add pinctrl bindings for fsl-imx-esdhc yaml

Signed-off-by: Peng Fan <peng.fan@nxp.com>
Link: https://lore.kernel.org/r/1614222604-27066-2-git-send-email-peng.fan@oss.nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: tmio: support custom irq masks

SDHI Gen2+ has a different value for TMIO_MASK_ALL, so add a member to
support that. If the member is not used, the previous default value is
applied.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/20210223100830.25125-2-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Set read only for SD cards with permanent write protect bit

Some of SD cards sets permanent write protection bit in their CSD register,
due to lifespan or internal problem. To avoid unnecessary I/O write
operations, let's parse the bits in the CSD during initialization and mark
the card as read only for this case.

Signed-off-by: Seunghui Lee <sh043.lee@samsung.com>
Link: https://lore.kernel.org/r/20210222083156.19158-1-sh043.lee@samsung.com
Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: uniphier-sd: Fix a resource leak in the remove function

A 'tmio_mmc_host_free()' call is missing in the remove function, in order
to balance a 'tmio_mmc_host_alloc()' call in the probe.
This is done in the error handling path of the probe, but not in the remove
function.

Add the missing call.

Fixes: 3fd784f745dd ("mmc: uniphier-sd: add UniPhier SD/eMMC controller driver")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Masahiro Yamada <masahiroy@kernel.org>
Link: https://lore.kernel.org/r/20210220142953.918608-1-christophe.jaillet@wanadoo.fr
Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: uniphier-sd: Fix an error handling path in uniphier_sd_probe()

A 'uniphier_sd_clk_enable()' call should be balanced by a corresponding
'uniphier_sd_clk_disable()' call.
This is done in the remove function, but not in the error handling path of
the probe.

Add the missing call.

Fixes: 3fd784f745dd ("mmc: uniphier-sd: add UniPhier SD/eMMC controller driver")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Masahiro Yamada <masahiroy@kernel.org>
Link: https://lore.kernel.org/r/20210220142935.918554-1-christophe.jaillet@wanadoo.fr
Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: tmio: remove workaround for NON_REMOVABLE

RPM handling has been improved twice since this comment, and also SCC
handling has been improved a lot. All the testing we did (Geert's and
Niklas' and Wolfram's board farms) with the workaround removed did not
lead to problems, so it is time to get rid of it to the best of our
knowledge.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Link: https://lore.kernel.org/r/20210218110224.6910-1-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Reduce code duplication to mmc_spi_send_{csd|cid}

mmc_spi_send_{csd|cid} have similar function body. Let's remove the
duplicated part to simplify the code, just add opcode to distinguish
them in changed mmc_spi_send_cxd().

Signed-off-by: Yue Hu <huyue2@yulong.com>
Link: https://lore.kernel.org/r/20210215015116.1355-1-zbestahu@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: dw_mmc: Drop redundant call to ->card_event callback

The ->card_event callback isn't being used by any of the dw_mmc variants.
It's likely a leftover from an earlier change, hence let's just drop the
redundant call to it.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Jaehoon Chung <jh80.chung@samsung.com>

mmc: core: Drop reference counting of the bus_ops

When the mmc_rescan work is enabled for execution (host->rescan_disable),
it's the only instance per mmc host that is allowed to set/clear the
host->bus_ops pointer.

Besides the mmc_rescan work, there are a couple of scenarios when the
host->bus_ops pointer may be accessed. Typically, those can be described as
as below:

*)
Upper mmc driver layers (like the mmc block device driver or an SDIO
functional driver) needs to execute a host->bus_ops callback. This can be
considered as safe without having to use some special locking mechanism,
because they operate on top of the struct mmc_card. As long as there is a
card to operate upon, the mmc core guarantees that there is a host->bus_ops
assigned as well. Note that, upper layer mmc drivers are of course
responsible to clean up from themselves from their ->remove() callbacks,
otherwise things would fall apart anyways.

**)
Via the mmc host instance, we may need to force a removal of an inserted
mmc card. This happens when a mmc host driver gets unbind, for example. In
this case, we protect the host->bus_ops pointer from concurrent accesses,
by disabling the mmc_rescan work upfront (host->rescan_disable). See
mmc_stop_host() for example.

This said, it seems like the reference counting of the host->bus_ops
pointer at some point have become superfluous. As this is an old mechanism
of the mmc core, it a bit difficult to digest the history of when that
could have happened. However, let's drop the reference counting to avoid
unnecessary code-paths and lockings.

Cc: Pierre Ossman <pierre@ossman.eu>
Cc: Russell King <linux@armlinux.org.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20210212131610.236843-1-ulf.hansson@linaro.org

mmc: core: Drop superfluous validations in mmc_hw|sw_reset()

The mmc_hw|sw_reset() APIs are designed to be called solely from upper
layers, which means drivers that operates on top of the struct mmc_card,
like the mmc block device driver and an SDIO functional driver.

Additionally, as long as the struct mmc_host has a valid pointer to a
struct mmc_card, the corresponding host->bus_ops pointer stays valid and
assigned.

For these reasons, let's drop the superfluous reference counting and the
redundant validations in mmc_hw|sw_reset().

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20210212131532.236775-1-ulf.hansson@linaro.org

mmc: core: Do a power cycle when the CMD11 fails

A CMD11 is sent to the SD/SDIO card to start the voltage switch procedure
into 1.8V I/O. According to the SD spec a power cycle is needed of the
card, if it turns out that the CMD11 fails. Let's fix this, to allow a
retry of the initialization without the voltage switch, to succeed.

Note that, whether it makes sense to also retry with the voltage switch
after the power cycle is a bit more difficult to know. At this point, we
treat it like the CMD11 isn't supported and therefore we skip it when
retrying.

Signed-off-by: DooHyun Hwang <dh0421.hwang@samsung.com>
Link: https://lore.kernel.org/r/20210210045936.7809-1-dh0421.hwang@samsung.com
Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

Linux 5.12-rc5

Merge tag 'perf-tools-fixes-for-v5.12-2020-03-28' of git://git./linux/kernel/git/acme/linux

Pull perf tooling fixes from Arnaldo Carvalho de Melo:

 - Avoid write of uninitialized memory when generating PERF_RECORD_MMAP*
   records.

 - Fix 'perf top' BPF support related crash with perf_event_paranoid=3 +
   kptr_restrict.

 - Validate raw event with sysfs exported format bits.

 - Fix waipid on SIGCHLD delivery bugs in 'perf daemon'.

 - Change to use bash for daemon test on Debian, where the default is
   dash and thus fails for use of bashisms in this test.

 - Fix memory leak in vDSO found using ASAN.

 - Remove now useless (due to the fact that BPF now supports static
   vars) failing sub test "BPF relocation checker".

 - Fix auxtrace queue conflict.

 - Sync linux/kvm.h with the kernel sources.

* tag 'perf-tools-fixes-for-v5.12-2020-03-28' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  perf test: Change to use bash for daemon test
  perf record: Fix memory leak in vDSO found using ASAN
  perf test: Remove now useless failing sub test "BPF relocation checker"
  perf daemon: Return from kill functions
  perf daemon: Force waipid for all session on SIGCHLD delivery
  perf top: Fix BPF support related crash with perf_event_paranoid=3 + kptr_restrict
  perf pmu: Validate raw event with sysfs exported format bits
  perf synthetic events: Avoid write of uninitialized memory when generating PERF_RECORD_MMAP* records
  tools headers UAPI: Sync linux/kvm.h with the kernel sources
  perf synthetic-events: Fix uninitialized 'kernel_thread' variable
  perf auxtrace: Fix auxtrace queue conflict

Merge tag 'auxdisplay-for-linus-v5.12-rc6' of git://github.com/ojeda/linux

Pull auxdisplay fix from Miguel Ojeda:
 "Remove in_interrupt() usage (Sebastian Andrzej Siewior)"

* tag 'auxdisplay-for-linus-v5.12-rc6' of git://github.com/ojeda/linux:
  auxdisplay: Remove in_interrupt() usage.

Merge tag 'x86-urgent-2021-03-28' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "Two fixes:

   - Fix build failure on Ubuntu with new GCC packages that turn
     on -fcf-protection

   - Fix SME memory encryption PTE encoding bug - AFAICT the code
     worked on 4K page sizes (level 1) but had the wrong shift at
     higher page level orders (level 2 and higher)"

* tag 'x86-urgent-2021-03-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/build: Turn off -fcf-protection for realmode targets
  x86/mem_encrypt: Correct physical address calculation in __set_clr_pte_enc()

Merge tag 'locking-urgent-2021-03-28' of git://git./linux/kernel/git/tip/tip

Pull locking fix from Ingo Molnar:
 "Fix the non-debug mutex_lock_io_nested() method to map to
  mutex_lock_io() instead of mutex_lock().

  Right now nothing uses this API explicitly, but this is an
  accident waiting to happen"

* tag 'locking-urgent-2021-03-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  locking/mutex: Fix non debug version of mutex_lock_io_nested()

Merge tag '5.12-rc4-smb3' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Five cifs/smb3 fixes, two for stable.

  Includes an important fix for encryption and an ACL fix, as well as a
  fix for possible reflink data corruption"

* tag '5.12-rc4-smb3' of git://git.samba.org/sfrench/cifs-2.6:
  smb3: fix cached file size problems in duplicate extents (reflink)
  cifs: Silently ignore unknown oplock break handle
  cifs: revalidate mapping when we open files for SMB1 POSIX
  cifs: Fix chmod with modefromsid when an older ACE already exists.
  cifs: Adjust key sizes and key generation routines for AES256 encryption

Merge tag 'io_uring-5.12-2021-03-27' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:

 - Use thread info versions of flag testing, as discussed last week.

 - The series enabling PF_IO_WORKER to just take signals, instead of
   needing to special case that they do not in a bunch of places. Ends
   up being pretty trivial to do, and then we can revert all the special
   casing we're currently doing.

 - Kill dead pointer assignment

 - Fix hashed part of async work queue trace

 - Fix sign extension issue for IORING_OP_PROVIDE_BUFFERS

 - Fix a link completion ordering regression in this merge window

 - Cancellation fixes

* tag 'io_uring-5.12-2021-03-27' of git://git.kernel.dk/linux-block:
  io_uring: remove unsued assignment to pointer io
  io_uring: don't cancel extra on files match
  io_uring: don't cancel-track common timeouts
  io_uring: do post-completion chore on t-out cancel
  io_uring: fix timeout cancel return code
  Revert "signal: don't allow STOP on PF_IO_WORKER threads"
  Revert "kernel: freezer should treat PF_IO_WORKER like PF_KTHREAD for freezing"
  Revert "kernel: treat PF_IO_WORKER like PF_KTHREAD for ptrace/signals"
  Revert "signal: don't allow sending any signals to PF_IO_WORKER threads"
  kernel: stop masking signals in create_io_thread()
  io_uring: handle signals for IO threads like a normal thread
  kernel: don't call do_exit() for PF_IO_WORKER threads
  io_uring: maintain CQE order of a failed link
  io-wq: fix race around pending work on teardown
  io_uring: do ctx sqd ejection in a clear context
  io_uring: fix provide_buffers sign extension
  io_uring: don't skip file_end_write() on reissue
  io_uring: correct io_queue_async_work() traces
  io_uring: don't use {test,clear}_tsk_thread_flag() for current

Merge tag 'block-5.12-2021-03-27' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - Fix regression from this merge window with the xarray partition
   change, which allowed partition counts that overflow the u8 that
   holds the partition number (Ming)

 - Fix zone append warning (Johannes)

 - Segmentation count fix for multipage bvecs (David)

 - Partition scan fix (Chris)

* tag 'block-5.12-2021-03-27' of git://git.kernel.dk/linux-block:
  block: don't create too many partitions
  block: support zone append bvecs
  block: recalculate segment count for multi-segment discards correctly
  block: clear GD_NEED_PART_SCAN later in bdev_disk_changed

Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Seven fixes, all in drivers (qla2xxx, mkt3sas, qedi, target,
  ibmvscsi).

  The most serious are the target pscsi oom and the qla2xxx revert which
  can otherwise cause a use after free"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: target: pscsi: Clean up after failure in pscsi_map_sg()
  scsi: target: pscsi: Avoid OOM in pscsi_map_sg()
  scsi: mpt3sas: Fix error return code of mpt3sas_base_attach()
  scsi: qedi: Fix error return code of qedi_alloc_global_queues()
  scsi: Revert "qla2xxx: Make sure that aborted commands are freed"
  scsi: ibmvfc: Make ibmvfc_wait_for_ops() MQ aware
  scsi: ibmvfc: Fix potential race in ibmvfc_wait_for_ops()

io_uring: remove unsued assignment to pointer io

There is an assignment to io that is never read after the assignment,
the assignment is redundant and can be removed.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: don't cancel extra on files match

As tasks always wait and kill their io-wq on exec/exit, files are of no
more concern to us, so we don't need to specifically cancel them by hand
in those cases. Moreover we should not, because io_match_task() looks at
req->task->files now, which is always true and so leads to extra
cancellations, that wasn't a case before per-task io-wq.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/0566c1de9b9dd417f5de345c817ca953580e0e2e.1616696997.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: don't cancel-track common timeouts

Don't account usual timeouts (i.e. not linked) as REQ_F_INFLIGHT but
keep behaviour prior to dd59a3d595cc1 ("io_uring: reliably cancel linked
timeouts").

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/104441ef5d97e3932113d44501fda0df88656b83.1616696997.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: do post-completion chore on t-out cancel

Don't forget about io_commit_cqring() + io_cqring_ev_posted() after
exit/exec cancelling timeouts. Both functions declared only after
io_kill_timeouts(), so to avoid tons of forward declarations move
it down.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/72ace588772c0f14834a6a4185d56c445a366fb4.1616696997.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: fix timeout cancel return code

When we cancel a timeout we should emit a sensible return code, like
-ECANCELED but not 0, otherwise it may trick users.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/7b0ad1065e3bd1994722702bd0ba9e7bc9b0683b.1616696997.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Revert "signal: don't allow STOP on PF_IO_WORKER threads"

This reverts commit 4db4b1a0d1779dc159f7b87feb97030ec0b12597.

The IO threads allow and handle SIGSTOP now, so don't special case them
anymore in task_set_jobctl_pending().

Signed-off-by: Jens Axboe <axboe@kernel.dk>

Revert "kernel: freezer should treat PF_IO_WORKER like PF_KTHREAD for freezing"

This reverts commit 15b2219facadec583c24523eed40fa45865f859f.

Before IO threads accepted signals, the freezer using take signals to wake
up an IO thread would cause them to loop without any way to clear the
pending signal. That is no longer the case, so stop special casing
PF_IO_WORKER in the freezer.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

Revert "kernel: treat PF_IO_WORKER like PF_KTHREAD for ptrace/signals"

This reverts commit 6fb8f43cede0e4bd3ead847de78d531424a96be9.

The IO threads do allow signals now, including SIGSTOP, and we can allow
ptrace attach. Attaching won't reveal anything interesting for the IO
threads, but it will allow eg gdb to attach to a task with io_urings
and IO threads without complaining. And once attached, it will allow
the usual introspection into regular threads.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

Revert "signal: don't allow sending any signals to PF_IO_WORKER threads"

This reverts commit 5be28c8f85ce99ed2d329d2ad8bdd18ea19473a5.

IO threads now take signals just fine, so there's no reason to limit them
specifically. Revert the change that prevented that from happening.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

kernel: stop masking signals in create_io_thread()

This is racy - move the blocking into when the task is created and
we're marking it as PF_IO_WORKER anyway. The IO threads are now
prepared to handle signals like SIGSTOP as well, so clear that from
the mask to allow proper stopping of IO threads.

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Reported-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: handle signals for IO threads like a normal thread

We go through various hoops to disallow signals for the IO threads, but
there's really no reason why we cannot just allow them. The IO threads
never return to userspace like a normal thread, and hence don't go through
normal signal processing. Instead, just check for a pending signal as part
of the work loop, and call get_signal() to handle it for us if anything
is pending.

With that, we can support receiving signals, including special ones like
SIGSTOP.

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: don't create too many partitions

Commit a33df75c6328 ("block: use an xarray for disk->part_tbl") drops the
check on max supported number of partitionsr, and allows partition with
bigger partition numbers to be added. However, ->bd_partno is defined as
u8, so partition index of xarray table may not match with ->bd_partno.
Then delete_partition() may delete one unmatched partition, and caused
use-after-free.

Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reported-by: syzbot+8fede7e30c7cee0de139@syzkaller.appspotmail.com
Fixes: a33df75c6328 ("block: use an xarray for disk->part_tbl")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

smb3: fix cached file size problems in duplicate extents (reflink)

There were two problems (one of which could cause data corruption)
that were noticed with duplicate extents (ie reflink)
when debugging why various xfstests were being incorrectly skipped
(e.g. generic/138, generic/140, generic/142). First, we were not
updating the file size locally in the cache when extending a
file due to reflink (it would refresh after actimeo expires)
but xfstest was checking the size immediately which was still
0 so caused the test to be skipped.  Second, we were setting
the target file size (which could shrink the file) in all cases
to the end of the reflinked range rather than only setting the
target file size when reflink would extend the file.

CC: <stable@vger.kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>

cifs: Silently ignore unknown oplock break handle

Make SMB2 not print out an error when an oplock break is received for an
unknown handle, similar to SMB1.  The debug message which is printed for
these unknown handles may also be misleading, so fix that too.

The SMB2 lease break path is not affected by this patch.

Without this, a program which writes to a file from one thread, and
opens, reads, and writes the same file from another thread triggers the
below errors several times a minute when run against a Samba server
configured with "smb2 leases = no".

 CIFS: VFS: \\192.168.0.1 No task to wake, unknown frame received! NumMids 2
 00000000: 424d53fe 00000040 00000000 00000012  .SMB@...........
 00000010: 00000001 00000000 ffffffff ffffffff  ................
 00000020: 00000000 00000000 00000000 00000000  ................
 00000030: 00000000 00000000 00000000 00000000  ................

Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
Reviewed-by: Tom Talpey <tom@talpey.com>
Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>

cifs: revalidate mapping when we open files for SMB1 POSIX

RHBZ: 1933527

Under SMB1 + POSIX, if an inode is reused on a server after we have read and
cached a part of a file, when we then open the new file with the
re-cycled inode there is a chance that we may serve the old data out of cache
to the application.
This only happens for SMB1 (deprecated) and when posix are used.
The simplest solution to avoid this race is to force a revalidate
on smb1-posix open.

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>

cifs: Fix chmod with modefromsid when an older ACE already exists.

My recent fixes to cifsacl to maintain inherited ACEs had
regressed modefromsid when an older ACL already exists.

Found testing xfstest 495 with modefromsid mount option

Fixes: f5065508897a ("cifs: Retain old ACEs when converting between mode bits and ACL")

Signed-off-by: Shyam Prasad N <sprasad@microsoft.com>
Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>

kernel: don't call do_exit() for PF_IO_WORKER threads

Right now we're never calling get_signal() from PF_IO_WORKER threads, but
in preparation for doing so, don't handle a fatal signal for them. The
workers have state they need to cleanup when exiting, so just return
instead of calling do_exit() on their behalf. The threads themselves will
detect a fatal signal and do proper shutdown.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge tag 'for-5.12/dm-fixes-2' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - Fix DM verity target's optional argument processing.

 - Fix DM core's zoned model and zone sectors checks.

 - Fix spurious "detected capacity change" pr_info() when creating new
   DM device.

 - Fix DM ioctl out of bounds array access in handling of
   DM_LIST_DEVICES_CMD when no devices exist.

* tag 'for-5.12/dm-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm ioctl: fix out of bounds array access when no devices
  dm: don't report "detected capacity change" on device creation
  dm table: Fix zoned model check and zone sectors check
  dm verity: fix DM_VERITY_OPTS_MAX value

dm ioctl: fix out of bounds array access when no devices

If there are not any dm devices, we need to zero the "dev" argument in
the first structure dm_name_list. However, this can cause out of
bounds write, because the "needed" variable is zero and len may be
less than eight.

Fix this bug by reporting DM_BUFFER_FULL_FLAG if the result buffer is
too small to hold the "nl->dev" value.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: Mike Snitzer <snitzer@redhat.com>