platform/kernel/linux-rpi.git
4 years agommc: sdhci-esdhc-imx: optimize the clock setting
Haibo Chen [Wed, 19 Feb 2020 08:25:50 +0000 (16:25 +0800)]
mmc: sdhci-esdhc-imx: optimize the clock setting

When force clock off, check the SDOFF of register PRSSTAT to make sure
the clock is gate off. Before force clock on, check the SDSTB of register
PRSSTAT to make sure the clock is stable, this will eliminate the clock
glitch.

Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1582100757-20683-2-git-send-email-haibo.chen@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci-esdhc-imx: add strobe-dll-delay-target support
Haibo Chen [Wed, 19 Feb 2020 08:25:49 +0000 (16:25 +0800)]
mmc: sdhci-esdhc-imx: add strobe-dll-delay-target support

strobe-dll-delay-target is the delay cell add on the strobe line.
Strobe line the the uSDHC loopback read clock which is use in HS400
mode. Different strobe-dll-delay-target may need to set for different
board/SoC. If this delay cell is not set to an appropriate value,
we may see some read operation meet CRC error after HS400 mode select
which already pass the tuning.

This patch add the strobe-dll-delay-target setting in driver, so that
user can easily config this delay cell in dts file.

Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1582100757-20683-1-git-send-email-haibo.chen@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agodoc: dt: fsl-imx-esdhc: add strobe-dll-delay-target binding
Haibo Chen [Wed, 19 Feb 2020 08:25:04 +0000 (16:25 +0800)]
doc: dt: fsl-imx-esdhc: add strobe-dll-delay-target binding

Add fsl,strobe-dll-delay-target binding.

Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Link: https://lore.kernel.org/r/1582100704-20601-1-git-send-email-haibo.chen@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci-esdhc-imx: restore the per_clk rate in PM_RUNTIME
Haibo Chen [Wed, 19 Feb 2020 08:22:43 +0000 (16:22 +0800)]
mmc: sdhci-esdhc-imx: restore the per_clk rate in PM_RUNTIME

When pm_runtime_suspend is run, a call to SCFW power off the SS (SS is a
power domain, usdhc belong to this SS power domain) in which the resource
resides is made. The SCFW can power off the SS if no other resource in
active in that SS. If so, all state associated with all the resources within
the SS that is powered off is lost, this includes the clock rates, clock state
etc. When pm_runtime_resume is called, the SS associated with that resource
is powered up. But the clocks are left in the default state.

This patch restore clock rate in pm_runtime_resume, make sure the
clock is right rather than depending on the default state setting
by SCFW.

Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1582100563-20555-5-git-send-email-haibo.chen@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci-esdhci-imx: retune needed for Mega/Mix enabled SoCs
Haibo Chen [Wed, 19 Feb 2020 08:22:42 +0000 (16:22 +0800)]
mmc: sdhci-esdhci-imx: retune needed for Mega/Mix enabled SoCs

For Mega/Mix enabled SoCs like MX7D and MX6SX, uSDHC will lost power in
LP mode no matter whether the MMC_KEEP_POWER flag is set or not.
This may cause state misalign between kernel and HW, especially for
SDIO3.0 WiFi cards.
e.g. SDIO WiFi driver usually will keep power during system suspend.
And after resume, no card re-enumeration called.
But the tuning state is lost due to Mega/Mix.
Then CRC error may happen during next data transfer.

So we should always fire a mmc_retune_needed() for such type SoC
to tell MMC core retuning is needed for next data transfer.
mmc: sdhci-esdhci-imx: retune needed for Mega/Mix enabled SoCs

Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1582100563-20555-4-git-send-email-haibo.chen@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci-esdhc-imx: no fail when no pinctrl available
Haibo Chen [Wed, 19 Feb 2020 08:22:41 +0000 (16:22 +0800)]
mmc: sdhci-esdhc-imx: no fail when no pinctrl available

When using jailhouse to support two Linux on i.MX8MQ EVK, we use the
1st Linux to configure pinctrl for the 2nd Linux. Then the 2nd Linux
could use the mmc without pinctrl driver.

So give a warning message when no pinctrl available, but no fail probe.

Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/1582100563-20555-3-git-send-email-haibo.chen@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci: do not enable card detect interrupt for gpio cd type
Haibo Chen [Wed, 19 Feb 2020 08:22:40 +0000 (16:22 +0800)]
mmc: sdhci: do not enable card detect interrupt for gpio cd type

Except SDHCI_QUIRK_BROKEN_CARD_DETECTION and MMC_CAP_NONREMOVABLE,
we also do not need to handle controller native card detect interrupt
for gpio cd type.
If we wrong enabled the card detect interrupt for gpio case, it will
cause a lot of unexpected card detect interrupts during data transfer
which should not happen.

Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/1582100563-20555-2-git-send-email-haibo.chen@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci_am654: Enable DLL only for some speed modes
Faiz Abbas [Wed, 8 Jan 2020 15:09:20 +0000 (20:39 +0530)]
mmc: sdhci_am654: Enable DLL only for some speed modes

Its recommended that DLL must only be enabled for SDR50, DDR50, DDR52,
SDR104, HS200 and HS400 speed modes. Move DLL configuration to its own
function and call it only in the above speed modes.

Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20200108150920.14547-4-faiz_abbas@ti.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci_am654: Update OTAPDLY writes
Faiz Abbas [Wed, 8 Jan 2020 15:09:19 +0000 (20:39 +0530)]
mmc: sdhci_am654: Update OTAPDLY writes

According to the latest AM65x Data Manual[1], a different output tap
delay value is optimum for a given speed mode. Therefore, deprecate the
ti,otap-del-sel binding and introduce a new binding for each of the
possible MMC/SD speed modes. If the legacy mode is not found, fall back
to old binding to maintain dts compatibility.

[1] http://www.ti.com/lit/gpn/am6526

Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/20200108150920.14547-3-faiz_abbas@ti.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agodt-bindings: mmc: sdhci-am654: Update Output tap delay binding
Faiz Abbas [Wed, 8 Jan 2020 15:09:18 +0000 (20:39 +0530)]
dt-bindings: mmc: sdhci-am654: Update Output tap delay binding

According to latest AM65x Data Manual[1], a different output tap delay
value is recommended for all speed modes. Therefore, replace the
ti,otap-del-sel binding with one ti,otap-del-sel- for each MMC/SD speed
mode.

[1] http://www.ti.com/lit/gpn/am6526

Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Link: https://lore.kernel.org/r/20200108150920.14547-2-faiz_abbas@ti.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: host: hsq: Add missing MODULE_LICENSE() and MODULE_DESCRIPTION()
Baolin Wang [Thu, 20 Feb 2020 01:30:01 +0000 (09:30 +0800)]
mmc: host: hsq: Add missing MODULE_LICENSE() and MODULE_DESCRIPTION()

Add missing MODULE_LICENSE() and MODULE_DESCRIPTION() in hsq driver to
fix below warning when compiling the hsq as a module.

"WARNING: modpost: missing MODULE_LICENSE() in drivers/mmc/host/mmc_hsq.o".

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/98ce471185f037fce57520763621590588766381.1582161803.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci: Add support for sdmmc variant revision 2.0
Ludovic Barre [Tue, 28 Jan 2020 09:06:36 +0000 (10:06 +0100)]
mmc: mmci: Add support for sdmmc variant revision 2.0

This patch adds a sdmmc variant revision 2.0. This revision is backward
compatible with 1.1, but adds DMA linked list support.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-10-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci_sdmmc: Implement signal voltage callbacks
Ludovic Barre [Tue, 28 Jan 2020 09:06:35 +0000 (10:06 +0100)]
mmc: mmci_sdmmc: Implement signal voltage callbacks

To prepare the voltage switch procedure, the VSWITCHEN bit must be set
before sending the CMD11. To confirm completion of voltage switch, the
VSWEND flag must be checked.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-9-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci: Add callbacks for to manage signal voltage switch
Ludovic Barre [Tue, 28 Jan 2020 09:06:34 +0000 (10:06 +0100)]
mmc: mmci: Add callbacks for to manage signal voltage switch

A variant may need to define some actions before and after a voltage
switch.

This patch adds 2 callbacks to manage signal voltage switch in the struct
mmci_host_ops. ->pre_sig_volt_switch() allows to prepare a signal voltage
switch before sending the SD_SWITCH_VOLTAGE command (CMD11).
->post_sig_volt_switch callback allows specific actions to be executed,
after the I/O signal voltage level has been changed.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-8-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci_sdmmc: Add execute tuning with delay block
Ludovic Barre [Tue, 28 Jan 2020 09:06:33 +0000 (10:06 +0100)]
mmc: mmci_sdmmc: Add execute tuning with delay block

The hardware delay block is used to align the sampling clock on the data
received by SDMMC. It is mandatory for SDMMC to support the SDR104 mode.
The delay block is used to generate an output clock which is dephased from
the input clock.  The phase of the output clock must be programmed by the
execute tuning interface.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-7-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agodt-bindings: mmc: mmci: add delay block base register for sdmmc
Ludovic Barre [Tue, 28 Jan 2020 09:06:32 +0000 (10:06 +0100)]
dt-bindings: mmc: mmci: add delay block base register for sdmmc

To support the sdr104 mode, the sdmmc variant has a hardware delay block to
manage the clock phase when sampling data received by the card.

This patch adds a second base register (optional) for sdmmc delay block.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Acked-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20200128090636.13689-6-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci: Add private pointer for variant
Ludovic Barre [Tue, 28 Jan 2020 09:06:31 +0000 (10:06 +0100)]
mmc: mmci: Add private pointer for variant

In variant init function, some references may be allocated for variant
specific usage. Add a private void* to mmci_host struct allows at variant
functions to access on this references by mmci_host structure.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-5-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci: Add a reference at mmc_host_ops in mmci struct
Ludovic Barre [Tue, 28 Jan 2020 09:06:30 +0000 (10:06 +0100)]
mmc: mmci: Add a reference at mmc_host_ops in mmci struct

The variant init function may need to add a mmc_host_ops, for example to
add the execute_tuning support if this feature is available. This patch
adds mmc_host_ops pointer in mmci struct.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-4-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci_sdmmc: Rename sdmmc_priv struct to sdmmc_idma
Ludovic Barre [Tue, 28 Jan 2020 09:06:29 +0000 (10:06 +0100)]
mmc: mmci_sdmmc: Rename sdmmc_priv struct to sdmmc_idma

This patch renames sdmmc_priv struct to sdmmc_idma which is assigned to
host->dma_priv.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-3-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: mmci_sdmmc: Replace sg_dma_xxx macros
Ludovic Barre [Tue, 28 Jan 2020 09:06:28 +0000 (10:06 +0100)]
mmc: mmci_sdmmc: Replace sg_dma_xxx macros

sg_dma_xxx should be used after a dma_map_sg call has been done to get bus
addresses of each of the SG entries and their lengths.  But mmci_host_ops
validate_data can be called before dma_map_sg.  This patch replaces theses
macros by sg->offset and sg->length which are always defined.

Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-2-ludovic.barre@st.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: core: Fixup support for HW busy detection for HPI commands
Ulf Hansson [Tue, 4 Feb 2020 08:54:48 +0000 (09:54 +0100)]
mmc: core: Fixup support for HW busy detection for HPI commands

In case the host specify a max_busy_timeout, we need to validate that the
needed timeout for the HPI command conforms to that requirement. If that's
not the case, let's convert from a R1B response to a R1 response, as to
instruct the host to avoid HW busy detection.

Additionally, when R1B is used we must also inform the host about the busy
timeout for the command, so let's do that via updating cmd.busy_timeout.

Finally, when R1B is used and in case the host supports HW busy detection,
there should be no need for doing polling, so then skip that.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-12-ulf.hansson@linaro.org
4 years agommc: core: Convert to mmc_poll_for_busy() for HPI commands
Ulf Hansson [Tue, 4 Feb 2020 08:54:47 +0000 (09:54 +0100)]
mmc: core: Convert to mmc_poll_for_busy() for HPI commands

Rather than open coding the polling loop in mmc_interrupt_hpi(), let's
convert to use mmc_poll_for_busy().

Note that, moving to mmc_poll_for_busy() for HPI also improves the
behaviour according to below.

- Adds support for polling via the optional ->card_busy() host ops.
- Require R1_READY_FOR_DATA to be set in the CMD13 response before exiting
  the polling loop.
- Adds a throttling mechanism to avoid CPU hogging when polling.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-11-ulf.hansson@linaro.org
4 years agommc: core: Drop redundant out-parameter to mmc_send_hpi_cmd()
Ulf Hansson [Tue, 4 Feb 2020 08:54:46 +0000 (09:54 +0100)]
mmc: core: Drop redundant out-parameter to mmc_send_hpi_cmd()

The 'u32 *status' is unused by the caller, so let's drop it.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-10-ulf.hansson@linaro.org
4 years agommc: core: Convert to mmc_poll_for_busy() for erase/trim/discard
Ulf Hansson [Tue, 4 Feb 2020 08:54:45 +0000 (09:54 +0100)]
mmc: core: Convert to mmc_poll_for_busy() for erase/trim/discard

Rather than open coding the polling loop in mmc_do_erase(), let's convert
to use mmc_poll_for_busy().

To allow a slightly different error parsing during polling, compared to the
__mmc_switch() case, a new in-parameter to mmc_poll_for_busy() is needed,
but other than that the conversion is straight forward.

Besides addressing the open coding issue, moving to mmc_poll_for_busy() for
erase/trim/discard improves the behaviour according to below.

- Adds support for polling via the optional ->card_busy() host ops.
- Returns zero to indicate success when the final polling attempt finds the
  card non-busy, even if the timeout expired.
- Exits the polling loop when state moves to R1_STATE_TRAN, rather than
  when leaving R1_STATE_PRG.
- Decreases the starting range for throttling to 32-64us.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-9-ulf.hansson@linaro.org
4 years agommc: core: Update CMD13 busy check for CMD6 commands
Ulf Hansson [Tue, 4 Feb 2020 08:54:44 +0000 (09:54 +0100)]
mmc: core: Update CMD13 busy check for CMD6 commands

Through mmc_poll_for_busy() a CMD13 may be sent to get the status of the
(e)MMC card. If the state of the card is R1_STATE_PRG, the card is
considered as being busy, which means we continue to poll with CMD13. This
seems to be sufficient, but it's also unnecessary fragile, as it means a
new command/request could potentially be sent to the card when it's in an
unknown state.

To try to improve the situation, but also to move towards a more consistent
CMD13 polling behaviour in the mmc core, let's deploy the same policy we
use for regular I/O write requests. In other words, let's check that card
returns to the R1_STATE_TRAN and that the R1_READY_FOR_DATA bit is set in
the CMD13 response, before exiting the polling loop.

Note that, potentially this changed behaviour could lead to unnecessary
waiting for the timeout to expire, if the card for some reason, moves to an
unexpected error state. However, as we bail out from the polling loop when
R1_SWITCH_ERROR bit is set or when the CMD13 fails, this shouldn't be an
issue.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-8-ulf.hansson@linaro.org
4 years agommc: core: Enable re-use of mmc_blk_in_tran_state()
Ulf Hansson [Tue, 4 Feb 2020 08:54:43 +0000 (09:54 +0100)]
mmc: core: Enable re-use of mmc_blk_in_tran_state()

To allow subsequent changes to re-use the code from the static function
mmc_blk_in_tran_state(), let's move it to a public header. While at it,
let's also rename it to mmc_ready_for_data(), as to try to better describe
its purpose.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-7-ulf.hansson@linaro.org
4 years agommc: core: Split up mmc_poll_for_busy()
Ulf Hansson [Tue, 4 Feb 2020 08:54:42 +0000 (09:54 +0100)]
mmc: core: Split up mmc_poll_for_busy()

To make the code more readable, move the part that gets the busy status of
the card out into a separate function, mmc_busy_status(). Then call it from
mmc_poll_for_busy().

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-6-ulf.hansson@linaro.org
4 years agommc: core: Drop redundant in-parameter to __mmc_switch()
Ulf Hansson [Tue, 4 Feb 2020 08:54:41 +0000 (09:54 +0100)]
mmc: core: Drop redundant in-parameter to __mmc_switch()

The use_busy_signal in-parameter is set true by all callers of
__mmc_switch(), hence it's redundant so drop it.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-5-ulf.hansson@linaro.org
4 years agommc: core: Extend mmc_switch_status() to rid of __mmc_switch_status()
Ulf Hansson [Tue, 4 Feb 2020 08:54:40 +0000 (09:54 +0100)]
mmc: core: Extend mmc_switch_status() to rid of __mmc_switch_status()

To simplify code, let's extend mmc_switch_status() to cope with needs
addressed in __mmc_switch_status(). Then move all users to the updated
mmc_switch_status() API and drop __mmc_switch_status() altogether.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-4-ulf.hansson@linaro.org
4 years agommc: core: Drop unused define
Ulf Hansson [Tue, 4 Feb 2020 08:54:39 +0000 (09:54 +0100)]
mmc: core: Drop unused define

The last user of MMC_OPS_TIMEOUT_MS was recently removed, however the
define stayed around. Let's remove it.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-3-ulf.hansson@linaro.org
4 years agommc: core: Throttle polling rate for CMD6
Ulf Hansson [Tue, 4 Feb 2020 08:54:38 +0000 (09:54 +0100)]
mmc: core: Throttle polling rate for CMD6

In mmc_poll_for_busy() we loop continuously, either by sending a CMD13 or
by invoking the ->card_busy() host ops, as to detect when the card stops
signaling busy. This behaviour is problematic as it may cause CPU hogging,
especially when the busy signal time reaches beyond a few ms.

Let's fix the issue by adding a throttling mechanism, that inserts a
usleep_range() in between the polling attempts. The sleep range starts at
32-64us, but increases for each loop by a factor of 2, up until the range
reaches ~32-64ms. In this way, we are able to keep the loop fine-grained
enough for short busy signaling times, while also not hogging the CPU for
longer times.

Note that, this change is inspired by the similar throttling mechanism that
we already use for mmc_do_erase().

Reported-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Baolin Wang <baolin.wang7@gmail.com>
Tested-by: Ludovic Barre <ludovic.barre@st.com>
Reviewed-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200204085449.32585-2-ulf.hansson@linaro.org
4 years agommc: host: sdhci-sprd: Add software queue support
Baolin Wang [Wed, 12 Feb 2020 04:13:00 +0000 (12:13 +0800)]
mmc: host: sdhci-sprd: Add software queue support

Add software queue support to improve the performance.

Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/f629b32943aae9e30ffa17acf4af06c270417001.1581478569.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: host: sdhci: Add a variable to defer to complete requests if needed
Baolin Wang [Wed, 12 Feb 2020 04:12:59 +0000 (12:12 +0800)]
mmc: host: sdhci: Add a variable to defer to complete requests if needed

When using the host software queue, it will trigger the next request in
irq handler without a context switch. But the sdhci_request() can not be
called in interrupt context when using host software queue for some host
drivers, due to the get_cd() ops can be sleepable.

But for some host drivers, such as Spreadtrum host driver, the card is
nonremovable, so the get_cd() ops is not sleepable, which means we can
complete the data request and trigger the next request in irq handler
to remove the context switch for the Spreadtrum host driver.

As suggested by Adrian, we should introduce a request_atomic() API to
indicate that a request can be called in interrupt context to remove
the context switch when using mmc host software queue. But this should
be done in another thread to convert the users of mmc host software queue.
Thus we can introduce a variable in struct sdhci_host to indicate that
we will always to defer to complete requests when using the host software
queue.

Suggested-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/e693e7a29beb3c1922b333f4603ea81f43d5c5b1.1581478568.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: host: sdhci: Add request_done ops for struct sdhci_ops
Baolin Wang [Wed, 12 Feb 2020 04:12:58 +0000 (12:12 +0800)]
mmc: host: sdhci: Add request_done ops for struct sdhci_ops

Add request_done ops for struct sdhci_ops as a preparation in case some
host controllers have different method to complete one request, such as
supporting request completion of MMC software queue.

Suggested-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/1539c801c8bbdbcd1d86f8c2dab375f5803c765a.1581478568.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: core: Enable the MMC host software queue for the SD card
Baolin Wang [Wed, 12 Feb 2020 04:12:57 +0000 (12:12 +0800)]
mmc: core: Enable the MMC host software queue for the SD card

Enable the MMC host software queue for the SD card if the host controller
supports the MMC host software queue.

On my Spreadtrum platform, I did not see any obvious performance changes
in 4K block size when changing to use hsq for the SD cards, I think the
reason is the SD card works at a low speed on my platform, and most of
time is spent in the hardware. But we can see some obvious improvements
when enabling the packed request based on hsq, that's why we still add hsq
support for the SD cards.

Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/0065b4631fef2d61c3b89d14a4ea4f2b7499ea56.1581478568.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: Add MMC host software queue support
Baolin Wang [Wed, 12 Feb 2020 04:12:56 +0000 (12:12 +0800)]
mmc: Add MMC host software queue support

Now the MMC read/write stack will always wait for previous request is
completed by mmc_blk_rw_wait(), before sending a new request to hardware,
or queue a work to complete request, that will bring context switching
overhead and spend some extra time to poll the card for busy completion
for I/O writes via sending CMD13, especially for high I/O per second
rates, to affect the IO performance.

Thus this patch introduces MMC software queue interface based on the
hardware command queue engine's interfaces, which is similar with the
hardware command queue engine's idea, that can remove the context
switching. Moreover we set the default queue depth as 64 for software
queue, which allows more requests to be prepared, merged and inserted
into IO scheduler to improve performance, but we only allow 2 requests
in flight, that is enough to let the irq handler always trigger the
next request without a context switch, as well as avoiding a long latency.

Moreover the host controller should support HW busy detection for I/O
operations when enabling the host software queue. That means, the host
controller must not complete a data transfer request, until after the
card stops signals busy.

From the fio testing data in cover letter, we can see the software
queue can improve some performance with 4K block size, increasing
about 16% for random read, increasing about 90% for random write,
though no obvious improvement for sequential read and write.

Moreover we can expand the software queue interface to support MMC
packed request or packed command in future.

Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/4409c1586a9b3ed20d57ad2faf6c262fc3ccb6e2.1581478568.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci-msm: Don't enable PWRSAVE_DLL for certain sdhc hosts
Ritesh Harjani [Fri, 7 Feb 2020 12:04:28 +0000 (17:34 +0530)]
mmc: sdhci-msm: Don't enable PWRSAVE_DLL for certain sdhc hosts

SDHC core with new 14lpp and later tech DLL should not enable
PWRSAVE_DLL since such controller's internal gating cannot meet
following MCLK requirement:
When MCLK is gated OFF, it is not gated for less than 0.5us and MCLK
must be switched on for at-least 1us before DATA starts coming.

Adding support for this requirement.

Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
Signed-off-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org>
Reviewed-by: Can Guo <cang@codeaurora.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1581077075-26011-1-git-send-email-vbadigan@codeaurora.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci-of-arasan: Remove quirk for broken base clock
Manish Narani [Tue, 21 Jan 2020 10:21:35 +0000 (15:51 +0530)]
mmc: sdhci-of-arasan: Remove quirk for broken base clock

This patch removes quirk which indicates a broken base clock. This was
making the kernel report wrong base clock of ~187MHz instead of 200MHz
even as the measurement on the hardware was showing 200MHz.

Signed-off-by: Manish Narani <manish.narani@xilinx.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1579602095-30060-5-git-send-email-manish.narani@xilinx.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: sdhci-of-arasan: Add support for DLL reset for ZynqMP platforms
Manish Narani [Tue, 21 Jan 2020 10:21:34 +0000 (15:51 +0530)]
mmc: sdhci-of-arasan: Add support for DLL reset for ZynqMP platforms

The DLL resets are required while executing the auto tuning procedure in
ZynqMP. This patch adds code to support the same.

Signed-off-by: Manish Narani <manish.narani@xilinx.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Link: https://lore.kernel.org/r/1579602095-30060-4-git-send-email-manish.narani@xilinx.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agofirmware: xilinx: Add DLL reset support
Manish Narani [Tue, 21 Jan 2020 10:21:33 +0000 (15:51 +0530)]
firmware: xilinx: Add DLL reset support

SD DLL resets are required for some of the operations on ZynqMP platform.
Add DLL reset support in ZynqMP firmware driver for SD DLL reset.

Signed-off-by: Manish Narani <manish.narani@xilinx.com>
Acked-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/1579602095-30060-3-git-send-email-manish.narani@xilinx.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agofirmware: xilinx: Add ZynqMP Tap Delay setup ioctl to the valid list
Manish Narani [Tue, 21 Jan 2020 10:21:32 +0000 (15:51 +0530)]
firmware: xilinx: Add ZynqMP Tap Delay setup ioctl to the valid list

The Tap Delay setup ioctl was not added to valid list due to which it
may fail to set Tap Delays for SD. This patch fixes the same.

Signed-off-by: Manish Narani <manish.narani@xilinx.com>
Acked-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/1579602095-30060-2-git-send-email-manish.narani@xilinx.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: tmio: remove superfluous callback wrappers
Wolfram Sang [Wed, 29 Jan 2020 20:37:09 +0000 (21:37 +0100)]
mmc: tmio: remove superfluous callback wrappers

After various refactoring, we can populate the mmc_ops callbacks
directly and don't need to have wrappers for them anymore.

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>
Link: https://lore.kernel.org/r/20200129203709.30493-7-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: tmio: factor out TAP usage
Wolfram Sang [Wed, 29 Jan 2020 20:37:08 +0000 (21:37 +0100)]
mmc: tmio: factor out TAP usage

TAPs are Renesas SDHI specific. Now that we moved all handling to the
SDHI core, we can also move the definitions from the TMIO struct to the
SDHI one.

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>
Link: https://lore.kernel.org/r/20200129203709.30493-6-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: tmio: enforce retune after runtime suspend
Wolfram Sang [Wed, 29 Jan 2020 20:37:07 +0000 (21:37 +0100)]
mmc: tmio: enforce retune after runtime suspend

Currently, select_tuning() is called after RPM resume. But
select_tuning() needs some additional function calls to work correctly.
Instead of reimplementing the whole postprocessing, just enforce
retuning.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/20200129203709.30493-5-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: tmio: give callback a generic name
Wolfram Sang [Wed, 29 Jan 2020 20:37:06 +0000 (21:37 +0100)]
mmc: tmio: give callback a generic name

check_scc_error() is too Renesas specific. Let's just call it
check_retune() to make it also easier understandable what it does.
Only a rename, no functional change.

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>
Link: https://lore.kernel.org/r/20200129203709.30493-4-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: renesas_sdhi: complain loudly if driver needs update
Wolfram Sang [Wed, 29 Jan 2020 20:37:05 +0000 (21:37 +0100)]
mmc: renesas_sdhi: complain loudly if driver needs update

When the tap array in the driver is too low, this is not a warning but
an error. Also _once is not helpful, we should make sure it is
prominently in the logs. It is safe to do this because this will only
show up during SoC enablement when we a new SoCs needs more taps (if
that ever will happen).

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>
Link: https://lore.kernel.org/r/20200129203709.30493-3-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: tmio: refactor tuning execution into SDHI driver
Wolfram Sang [Wed, 29 Jan 2020 20:37:04 +0000 (21:37 +0100)]
mmc: tmio: refactor tuning execution into SDHI driver

Move Renesas specific code for executing the tuning with a SCC into the
SDHI driver and leave only a generic call in the TMIO driver. Simplify
the code a little by removing init_tuning() and prepare_tuning()
callbacks. The latter is directly folded into the new execute_tuning()
callbacks.

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>
Link: https://lore.kernel.org/r/20200129203709.30493-2-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: renesas_sdhi: cleanup SCC defines
Wolfram Sang [Tue, 17 Dec 2019 11:40:34 +0000 (12:40 +0100)]
mmc: renesas_sdhi: cleanup SCC defines

Use increasing BIT numbers consistently and remove some superfluous
comments.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/20191217114034.13290-6-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: renesas_sdhi: enforce manual correction for Gen3
Wolfram Sang [Tue, 17 Dec 2019 11:40:33 +0000 (12:40 +0100)]
mmc: renesas_sdhi: enforce manual correction for Gen3

HW engineers say that automatic tap correction cannot be used for HS400
in all R-Car Gen3 SoCs. So, check for that SDHI variant and disable it
when HS400 is about to be enabled.

Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20191217114034.13290-5-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: renesas_sdhi: only check CMD status for HS400 manual correction
Takeshi Saito [Tue, 17 Dec 2019 11:40:32 +0000 (12:40 +0100)]
mmc: renesas_sdhi: only check CMD status for HS400 manual correction

R-Car Gen3 cannot use correction error status with HS400.
HS200: CMD and DAT signal timing are based on CLK signal.
HS400: CMD signal is based on CLK. DAT signal is based on DS signal.

In HS400, CMD signal is 200MHz(SDR). DAT signal is 200MHz(DDR).
Center position of signal is different between CMD and DAT.

TAP position should be adjusted to the center position of CMD signal.
DAT sampling timing is adjusted by HS400 calibration circuit regardless
of TAP position. Refer to renesas_sdhi_adjust_hs400mode_enable().

However, correction error status contains CMD and DAT status in HS400
(DAT signal is not masked in HS400). Therefore, correction error status
cannot use in HS400. It means that auto correction cannot be uses in
HS400. Manual correction can change to the correct TAP position by
ignoring DAT correction error status and using only CMD correction
status.

Signed-off-by: Takeshi Saito <takeshi.saito.xv@renesas.com>
[wsa: refactored patch from BSP]
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/20191217114034.13290-4-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: renesas_sdhi: Add manual correction
Takeshi Saito [Tue, 17 Dec 2019 11:40:31 +0000 (12:40 +0100)]
mmc: renesas_sdhi: Add manual correction

This patch adds a manual correction mechanism for SDHI. Currently, SDHI
uses automatic TAP position correction. However, TAP position can also
be corrected manually via correction error status flags.

Signed-off-by: Takeshi Saito <takeshi.saito.xv@renesas.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20191217114034.13290-3-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agommc: renesas_sdhi: remove double clear of automatic correction
Wolfram Sang [Tue, 17 Dec 2019 11:40:30 +0000 (12:40 +0100)]
mmc: renesas_sdhi: remove double clear of automatic correction

hw_reset() clears the automatic correction bit twice. I couldn't find
anything in the docs recommending that. Removing one of them didn't
cause any regressions here, so keep it simple.

Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20191217114034.13290-2-wsa+renesas@sang-engineering.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
4 years agoLinux 5.6-rc7
Linus Torvalds [Mon, 23 Mar 2020 01:31:56 +0000 (18:31 -0700)]
Linux 5.6-rc7

4 years agoMerge tag 'for-5.6-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Sun, 22 Mar 2020 18:35:33 +0000 (11:35 -0700)]
Merge tag 'for-5.6-rc6-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "Two fixes.

  The first is a regression: when dropping some incompat bits the
  conditions were reversed. The other is a fix for rename whiteout
  potentially leaving stack memory linked to a list"

* tag 'for-5.6-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: fix removal of raid[56|1c34} incompat flags after removing block group
  btrfs: fix log context list corruption after rename whiteout error

4 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sun, 22 Mar 2020 17:46:50 +0000 (10:46 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "10 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  x86/mm: split vmalloc_sync_all()
  mm, slub: prevent kmalloc_node crashes and memory leaks
  mm/mmu_notifier: silence PROVE_RCU_LIST warnings
  epoll: fix possible lost wakeup on epoll_ctl() path
  mm: do not allow MADV_PAGEOUT for CoW pages
  mm, memcg: throttle allocators based on ancestral memory.high
  mm, memcg: fix corruption on 64-bit divisor in memory.high throttling
  page-flags: fix a crash at SetPageError(THP_SWAP)
  mm/hotplug: fix hot remove failure in SPARSEMEM|!VMEMMAP case
  memcg: fix NULL pointer dereference in __mem_cgroup_usage_unregister_event

4 years agox86/mm: split vmalloc_sync_all()
Joerg Roedel [Sun, 22 Mar 2020 01:22:41 +0000 (18:22 -0700)]
x86/mm: split vmalloc_sync_all()

Commit 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in
__purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in
the vunmap() code-path.  While this change was necessary to maintain
correctness on x86-32-pae kernels, it also adds additional cycles for
architectures that don't need it.

Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported
severe performance regressions in micro-benchmarks because it now also
calls the x86-64 implementation of vmalloc_sync_all() on vunmap().  But
the vmalloc_sync_all() implementation on x86-64 is only needed for newly
created mappings.

To avoid the unnecessary work on x86-64 and to gain the performance
back, split up vmalloc_sync_all() into two functions:

* vmalloc_sync_mappings(), and
* vmalloc_sync_unmappings()

Most call-sites to vmalloc_sync_all() only care about new mappings being
synchronized.  The only exception is the new call-site added in the
above mentioned commit.

Shile Zhang directed us to a report of an 80% regression in reaim
throughput.

Fixes: 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()")
Reported-by: kernel test robot <oliver.sang@intel.com>
Reported-by: Shile Zhang <shile.zhang@linux.alibaba.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Borislav Petkov <bp@suse.de>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> [GHES]
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20191009124418.8286-1-joro@8bytes.org
Link: https://lists.01.org/hyperkitty/list/lkp@lists.01.org/thread/4D3JPPHBNOSPFK2KEPC6KGKS6J25AIDB/
Link: http://lkml.kernel.org/r/20191113095530.228959-1-shile.zhang@linux.alibaba.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agomm, slub: prevent kmalloc_node crashes and memory leaks
Vlastimil Babka [Sun, 22 Mar 2020 01:22:37 +0000 (18:22 -0700)]
mm, slub: prevent kmalloc_node crashes and memory leaks

Sachin reports [1] a crash in SLUB __slab_alloc():

  BUG: Kernel NULL pointer dereference on read at 0x000073b0
  Faulting instruction address: 0xc0000000003d55f4
  Oops: Kernel access of bad area, sig: 11 [#1]
  LE PAGE_SIZE=64K MMU=Hash SMP NR_CPUS=2048 NUMA pSeries
  Modules linked in:
  CPU: 19 PID: 1 Comm: systemd Not tainted 5.6.0-rc2-next-20200218-autotest #1
  NIP:  c0000000003d55f4 LR: c0000000003d5b94 CTR: 0000000000000000
  REGS: c0000008b37836d0 TRAP: 0300   Not tainted  (5.6.0-rc2-next-20200218-autotest)
  MSR:  8000000000009033 <SF,EE,ME,IR,DR,RI,LE>  CR: 24004844  XER: 00000000
  CFAR: c00000000000dec4 DAR: 00000000000073b0 DSISR: 40000000 IRQMASK: 1
  GPR00: c0000000003d5b94 c0000008b3783960 c00000000155d400 c0000008b301f500
  GPR04: 0000000000000dc0 0000000000000002 c0000000003443d8 c0000008bb398620
  GPR08: 00000008ba2f0000 0000000000000001 0000000000000000 0000000000000000
  GPR12: 0000000024004844 c00000001ec52a00 0000000000000000 0000000000000000
  GPR16: c0000008a1b20048 c000000001595898 c000000001750c18 0000000000000002
  GPR20: c000000001750c28 c000000001624470 0000000fffffffe0 5deadbeef0000122
  GPR24: 0000000000000001 0000000000000dc0 0000000000000002 c0000000003443d8
  GPR28: c0000008b301f500 c0000008bb398620 0000000000000000 c00c000002287180
  NIP ___slab_alloc+0x1f4/0x760
  LR __slab_alloc+0x34/0x60
  Call Trace:
    ___slab_alloc+0x334/0x760 (unreliable)
    __slab_alloc+0x34/0x60
    __kmalloc_node+0x110/0x490
    kvmalloc_node+0x58/0x110
    mem_cgroup_css_online+0x108/0x270
    online_css+0x48/0xd0
    cgroup_apply_control_enable+0x2ec/0x4d0
    cgroup_mkdir+0x228/0x5f0
    kernfs_iop_mkdir+0x90/0xf0
    vfs_mkdir+0x110/0x230
    do_mkdirat+0xb0/0x1a0
    system_call+0x5c/0x68

This is a PowerPC platform with following NUMA topology:

  available: 2 nodes (0-1)
  node 0 cpus:
  node 0 size: 0 MB
  node 0 free: 0 MB
  node 1 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
  node 1 size: 35247 MB
  node 1 free: 30907 MB
  node distances:
  node   0   1
    0:  10  40
    1:  40  10

  possible numa nodes: 0-31

This only happens with a mmotm patch "mm/memcontrol.c: allocate
shrinker_map on appropriate NUMA node" [2] which effectively calls
kmalloc_node for each possible node.  SLUB however only allocates
kmem_cache_node on online N_NORMAL_MEMORY nodes, and relies on
node_to_mem_node to return such valid node for other nodes since commit
a561ce00b09e ("slub: fall back to node_to_mem_node() node if allocating
on memoryless node").  This is however not true in this configuration
where the _node_numa_mem_ array is not initialized for nodes 0 and 2-31,
thus it contains zeroes and get_partial() ends up accessing
non-allocated kmem_cache_node.

A related issue was reported by Bharata (originally by Ramachandran) [3]
where a similar PowerPC configuration, but with mainline kernel without
patch [2] ends up allocating large amounts of pages by kmalloc-1k
kmalloc-512.  This seems to have the same underlying issue with
node_to_mem_node() not behaving as expected, and might probably also
lead to an infinite loop with CONFIG_SLUB_CPU_PARTIAL [4].

This patch should fix both issues by not relying on node_to_mem_node()
anymore and instead simply falling back to NUMA_NO_NODE, when
kmalloc_node(node) is attempted for a node that's not online, or has no
usable memory.  The "usable memory" condition is also changed from
node_present_pages() to N_NORMAL_MEMORY node state, as that is exactly
the condition that SLUB uses to allocate kmem_cache_node structures.
The check in get_partial() is removed completely, as the checks in
___slab_alloc() are now sufficient to prevent get_partial() being
reached with an invalid node.

[1] https://lore.kernel.org/linux-next/3381CD91-AB3D-4773-BA04-E7A072A63968@linux.vnet.ibm.com/
[2] https://lore.kernel.org/linux-mm/fff0e636-4c36-ed10-281c-8cdb0687c839@virtuozzo.com/
[3] https://lore.kernel.org/linux-mm/20200317092624.GB22538@in.ibm.com/
[4] https://lore.kernel.org/linux-mm/088b5996-faae-8a56-ef9c-5b567125ae54@suse.cz/

Fixes: a561ce00b09e ("slub: fall back to node_to_mem_node() node if allocating on memoryless node")
Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Reported-by: PUVICHAKRAVARTHY RAMACHANDRAN <puvichakravarthy@in.ibm.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Tested-by: Bharata B Rao <bharata@linux.ibm.com>
Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Christopher Lameter <cl@linux.com>
Cc: linuxppc-dev@lists.ozlabs.org
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Nathan Lynch <nathanl@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200320115533.9604-1-vbabka@suse.cz
Debugged-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agomm/mmu_notifier: silence PROVE_RCU_LIST warnings
Qian Cai [Sun, 22 Mar 2020 01:22:34 +0000 (18:22 -0700)]
mm/mmu_notifier: silence PROVE_RCU_LIST warnings

It is safe to traverse mm->notifier_subscriptions->list either under
SRCU read lock or mm->notifier_subscriptions->lock using
hlist_for_each_entry_rcu().  Silence the PROVE_RCU_LIST false positives,
for example,

  WARNING: suspicious RCU usage
  -----------------------------
  mm/mmu_notifier.c:484 RCU-list traversed in non-reader section!!

  other info that might help us debug this:

  rcu_scheduler_active = 2, debug_locks = 1
  3 locks held by libvirtd/802:
   #0: ffff9321e3f58148 (&mm->mmap_sem#2){++++}, at: do_mprotect_pkey+0xe1/0x3e0
   #1: ffffffff91ae6160 (mmu_notifier_invalidate_range_start){+.+.}, at: change_p4d_range+0x5fa/0x800
   #2: ffffffff91ae6e08 (srcu){....}, at: __mmu_notifier_invalidate_range_start+0x178/0x460

  stack backtrace:
  CPU: 7 PID: 802 Comm: libvirtd Tainted: G          I       5.6.0-rc6-next-20200317+ #2
  Hardware name: HP ProLiant BL460c Gen8, BIOS I31 11/02/2014
  Call Trace:
    dump_stack+0xa4/0xfe
    lockdep_rcu_suspicious+0xeb/0xf5
    __mmu_notifier_invalidate_range_start+0x3ff/0x460
    change_p4d_range+0x746/0x800
    change_protection+0x1df/0x300
    mprotect_fixup+0x245/0x3e0
    do_mprotect_pkey+0x23b/0x3e0
    __x64_sys_mprotect+0x51/0x70
    do_syscall_64+0x91/0xae8
    entry_SYSCALL_64_after_hwframe+0x49/0xb3

Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Link: http://lkml.kernel.org/r/20200317175640.2047-1-cai@lca.pw
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agoepoll: fix possible lost wakeup on epoll_ctl() path
Roman Penyaev [Sun, 22 Mar 2020 01:22:30 +0000 (18:22 -0700)]
epoll: fix possible lost wakeup on epoll_ctl() path

This fixes possible lost wakeup introduced by commit a218cc491420.
Originally modifications to ep->wq were serialized by ep->wq.lock, but
in commit a218cc491420 ("epoll: use rwlock in order to reduce
ep_poll_callback() contention") a new rw lock was introduced in order to
relax fd event path, i.e. callers of ep_poll_callback() function.

After the change ep_modify and ep_insert (both are called on epoll_ctl()
path) were switched to ep->lock, but ep_poll (epoll_wait) was using
ep->wq.lock on wqueue list modification.

The bug doesn't lead to any wqueue list corruptions, because wake up
path and list modifications were serialized by ep->wq.lock internally,
but actual waitqueue_active() check prior wake_up() call can be
reordered with modifications of ep ready list, thus wake up can be lost.

And yes, can be healed by explicit smp_mb():

  list_add_tail(&epi->rdlink, &ep->rdllist);
  smp_mb();
  if (waitqueue_active(&ep->wq))
wake_up(&ep->wp);

But let's make it simple, thus current patch replaces ep->wq.lock with
the ep->lock for wqueue modifications, thus wake up path always observes
activeness of the wqueue correcty.

Fixes: a218cc491420 ("epoll: use rwlock in order to reduce ep_poll_callback() contention")
Reported-by: Max Neunhoeffer <max@arangodb.com>
Signed-off-by: Roman Penyaev <rpenyaev@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Max Neunhoeffer <max@arangodb.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Christopher Kohlhoff <chris.kohlhoff@clearpool.io>
Cc: Davidlohr Bueso <dbueso@suse.de>
Cc: Jason Baron <jbaron@akamai.com>
Cc: Jes Sorensen <jes.sorensen@gmail.com>
Cc: <stable@vger.kernel.org> [5.1+]
Link: http://lkml.kernel.org/r/20200214170211.561524-1-rpenyaev@suse.de
References: https://bugzilla.kernel.org/show_bug.cgi?id=205933
Bisected-by: Max Neunhoeffer <max@arangodb.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agomm: do not allow MADV_PAGEOUT for CoW pages
Michal Hocko [Sun, 22 Mar 2020 01:22:26 +0000 (18:22 -0700)]
mm: do not allow MADV_PAGEOUT for CoW pages

Jann has brought up a very interesting point [1].  While shared pages
are excluded from MADV_PAGEOUT normally, CoW pages can be easily
reclaimed that way.  This can lead to all sorts of hard to debug
problems.  E.g.  performance problems outlined by Daniel [2].

There are runtime environments where there is a substantial memory
shared among security domains via CoW memory and a easy to reclaim way
of that memory, which MADV_{COLD,PAGEOUT} offers, can lead to either
performance degradation in for the parent process which might be more
privileged or even open side channel attacks.

The feasibility of the latter is not really clear to me TBH but there is
no real reason for exposure at this stage.  It seems there is no real
use case to depend on reclaiming CoW memory via madvise at this stage so
it is much easier to simply disallow it and this is what this patch
does.  Put it simply MADV_{PAGEOUT,COLD} can operate only on the
exclusively owned memory which is a straightforward semantic.

[1] http://lkml.kernel.org/r/CAG48ez0G3JkMq61gUmyQAaCq=_TwHbi1XKzWRooxZkv08PQKuw@mail.gmail.com
[2] http://lkml.kernel.org/r/CAKOZueua_v8jHCpmEtTB6f3i9e2YnmX4mqdYVWhV4E=Z-n+zRQ@mail.gmail.com

Fixes: 9c276cc65a58 ("mm: introduce MADV_COLD")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Daniel Colascione <dancol@google.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: "Joel Fernandes (Google)" <joel@joelfernandes.org>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200312082248.GS23944@dhcp22.suse.cz
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agomm, memcg: throttle allocators based on ancestral memory.high
Chris Down [Sun, 22 Mar 2020 01:22:23 +0000 (18:22 -0700)]
mm, memcg: throttle allocators based on ancestral memory.high

Prior to this commit, we only directly check the affected cgroup's
memory.high against its usage.  However, it's possible that we are being
reclaimed as a result of hitting an ancestor memory.high and should be
penalised based on that, instead.

This patch changes memory.high overage throttling to use the largest
overage in its ancestors when considering how many penalty jiffies to
charge.  This makes sure that we penalise poorly behaving cgroups in the
same way regardless of at what level of the hierarchy memory.high was
breached.

Fixes: 0e4b01df8659 ("mm, memcg: throttle allocators when failing reclaim over memory.high")
Reported-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Chris Down <chris@chrisdown.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nathan Chancellor <natechancellor@gmail.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: <stable@vger.kernel.org> [5.4.x+]
Link: http://lkml.kernel.org/r/8cd132f84bd7e16cdb8fde3378cdbf05ba00d387.1584036142.git.chris@chrisdown.name
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agomm, memcg: fix corruption on 64-bit divisor in memory.high throttling
Chris Down [Sun, 22 Mar 2020 01:22:20 +0000 (18:22 -0700)]
mm, memcg: fix corruption on 64-bit divisor in memory.high throttling

Commit 0e4b01df8659 had a bunch of fixups to use the right division
method.  However, it seems that after all that it still wasn't right --
div_u64 takes a 32-bit divisor.

The headroom is still large (2^32 pages), so on mundane systems you
won't hit this, but this should definitely be fixed.

Fixes: 0e4b01df8659 ("mm, memcg: throttle allocators when failing reclaim over memory.high")
Reported-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Chris Down <chris@chrisdown.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nathan Chancellor <natechancellor@gmail.com>
Cc: <stable@vger.kernel.org> [5.4.x+]
Link: http://lkml.kernel.org/r/80780887060514967d414b3cd91f9a316a16ab98.1584036142.git.chris@chrisdown.name
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agopage-flags: fix a crash at SetPageError(THP_SWAP)
Qian Cai [Sun, 22 Mar 2020 01:22:17 +0000 (18:22 -0700)]
page-flags: fix a crash at SetPageError(THP_SWAP)

Commit bd4c82c22c36 ("mm, THP, swap: delay splitting THP after swapped
out") supported writing THP to a swap device but forgot to upgrade an
older commit df8c94d13c7e ("page-flags: define behavior of FS/IO-related
flags on compound pages") which could trigger a crash during THP
swapping out with DEBUG_VM_PGFLAGS=y,

  kernel BUG at include/linux/page-flags.h:317!

  page dumped because: VM_BUG_ON_PAGE(1 && PageCompound(page))
  page:fffff3b2ec3a8000 refcount:512 mapcount:0 mapping:000000009eb0338c index:0x7f6e58200 head:fffff3b2ec3a8000 order:9 compound_mapcount:0 compound_pincount:0
  anon flags: 0x45fffe0000d8454(uptodate|lru|workingset|owner_priv_1|writeback|head|reclaim|swapbacked)

  end_swap_bio_write()
    SetPageError(page)
      VM_BUG_ON_PAGE(1 && PageCompound(page))

  <IRQ>
  bio_endio+0x297/0x560
  dec_pending+0x218/0x430 [dm_mod]
  clone_endio+0xe4/0x2c0 [dm_mod]
  bio_endio+0x297/0x560
  blk_update_request+0x201/0x920
  scsi_end_request+0x6b/0x4b0
  scsi_io_completion+0x509/0x7e0
  scsi_finish_command+0x1ed/0x2a0
  scsi_softirq_done+0x1c9/0x1d0
  __blk_mqnterrupt+0xf/0x20
  </IRQ>

Fix by checking PF_NO_TAIL in those places instead.

Fixes: bd4c82c22c36 ("mm, THP, swap: delay splitting THP after swapped out")
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Rafael Aquini <aquini@redhat.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200310235846.1319-1-cai@lca.pw
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agomm/hotplug: fix hot remove failure in SPARSEMEM|!VMEMMAP case
Baoquan He [Sun, 22 Mar 2020 01:22:13 +0000 (18:22 -0700)]
mm/hotplug: fix hot remove failure in SPARSEMEM|!VMEMMAP case

In section_deactivate(), pfn_to_page() doesn't work any more after
ms->section_mem_map is resetting to NULL in SPARSEMEM|!VMEMMAP case.  It
causes a hot remove failure:

  kernel BUG at mm/page_alloc.c:4806!
  invalid opcode: 0000 [#1] SMP PTI
  CPU: 3 PID: 8 Comm: kworker/u16:0 Tainted: G        W         5.5.0-next-20200205+ #340
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 0.0.0 02/06/2015
  Workqueue: kacpi_hotplug acpi_hotplug_work_fn
  RIP: 0010:free_pages+0x85/0xa0
  Call Trace:
   __remove_pages+0x99/0xc0
   arch_remove_memory+0x23/0x4d
   try_remove_memory+0xc8/0x130
   __remove_memory+0xa/0x11
   acpi_memory_device_remove+0x72/0x100
   acpi_bus_trim+0x55/0x90
   acpi_device_hotplug+0x2eb/0x3d0
   acpi_hotplug_work_fn+0x1a/0x30
   process_one_work+0x1a7/0x370
   worker_thread+0x30/0x380
   kthread+0x112/0x130
   ret_from_fork+0x35/0x40

Let's move the ->section_mem_map resetting after
depopulate_section_memmap() to fix it.

[akpm@linux-foundation.org: remove unneeded initialization, per David]
Fixes: ba72b4c8cf60 ("mm/sparsemem: support sub-section hotplug")
Signed-off-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200307084229.28251-2-bhe@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agomemcg: fix NULL pointer dereference in __mem_cgroup_usage_unregister_event
Chunguang Xu [Sun, 22 Mar 2020 01:22:10 +0000 (18:22 -0700)]
memcg: fix NULL pointer dereference in __mem_cgroup_usage_unregister_event

An eventfd monitors multiple memory thresholds of the cgroup, closes them,
the kernel deletes all events related to this eventfd.  Before all events
are deleted, another eventfd monitors the memory threshold of this cgroup,
leading to a crash:

  BUG: kernel NULL pointer dereference, address: 0000000000000004
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 800000033058e067 P4D 800000033058e067 PUD 3355ce067 PMD 0
  Oops: 0002 [#1] SMP PTI
  CPU: 2 PID: 14012 Comm: kworker/2:6 Kdump: loaded Not tainted 5.6.0-rc4 #3
  Hardware name: LENOVO 20AWS01K00/20AWS01K00, BIOS GLET70WW (2.24 ) 05/21/2014
  Workqueue: events memcg_event_remove
  RIP: 0010:__mem_cgroup_usage_unregister_event+0xb3/0x190
  RSP: 0018:ffffb47e01c4fe18 EFLAGS: 00010202
  RAX: 0000000000000001 RBX: ffff8bb223a8a000 RCX: 0000000000000001
  RDX: 0000000000000001 RSI: ffff8bb22fb83540 RDI: 0000000000000001
  RBP: ffffb47e01c4fe48 R08: 0000000000000000 R09: 0000000000000010
  R10: 000000000000000c R11: 071c71c71c71c71c R12: ffff8bb226aba880
  R13: ffff8bb223a8a480 R14: 0000000000000000 R15: 0000000000000000
  FS:  0000000000000000(0000) GS:ffff8bb242680000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000000000004 CR3: 000000032c29c003 CR4: 00000000001606e0
  Call Trace:
    memcg_event_remove+0x32/0x90
    process_one_work+0x172/0x380
    worker_thread+0x49/0x3f0
    kthread+0xf8/0x130
    ret_from_fork+0x35/0x40
  CR2: 0000000000000004

We can reproduce this problem in the following ways:

1. We create a new cgroup subdirectory and a new eventfd, and then we
   monitor multiple memory thresholds of the cgroup through this eventfd.

2.  closing this eventfd, and __mem_cgroup_usage_unregister_event ()
   will be called multiple times to delete all events related to this
   eventfd.

The first time __mem_cgroup_usage_unregister_event() is called, the
kernel will clear all items related to this eventfd in thresholds->
primary.

Since there is currently only one eventfd, thresholds-> primary becomes
empty, so the kernel will set thresholds-> primary and hresholds-> spare
to NULL.  If at this time, the user creates a new eventfd and monitor
the memory threshold of this cgroup, kernel will re-initialize
thresholds-> primary.

Then when __mem_cgroup_usage_unregister_event () is called for the
second time, because thresholds-> primary is not empty, the system will
access thresholds-> spare, but thresholds-> spare is NULL, which will
trigger a crash.

In general, the longer it takes to delete all events related to this
eventfd, the easier it is to trigger this problem.

The solution is to check whether the thresholds associated with the
eventfd has been cleared when deleting the event.  If so, we do nothing.

[akpm@linux-foundation.org: fix comment, per Kirill]
Fixes: 907860ed381a ("cgroups: make cftype.unregister_event() void-returning")
Signed-off-by: Chunguang Xu <brookxu@tencent.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/077a6f67-aefa-4591-efec-f2f3af2b0b02@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agoMerge tag 'block-5.6-20200320' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 21 Mar 2020 19:08:26 +0000 (12:08 -0700)]
Merge tag 'block-5.6-20200320' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "Just two NVMe fabrics fixes that should go into 5.6"

* tag 'block-5.6-20200320' of git://git.kernel.dk/linux-block:
  nvmet-tcp: set MSG_MORE only if we actually have more to send
  nvme-rdma: Avoid double freeing of async event data

4 years agoMerge tag 'io_uring-5.6-20200320' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 21 Mar 2020 18:54:47 +0000 (11:54 -0700)]
Merge tag 'io_uring-5.6-20200320' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "Two different fixes in here:

   - Fix for a potential NULL pointer deref for links with async or
     drain marked (Pavel)

   - Fix for not properly checking RLIMIT_NOFILE for async punted
     operations.

     This affects openat/openat2, which were added this cycle, and
     accept4. I did a full audit of other cases where we might check
     current->signal->rlim[] and found only RLIMIT_FSIZE for buffered
     writes and fallocate. That one is fixed and queued for 5.7 and
     marked stable"

* tag 'io_uring-5.6-20200320' of git://git.kernel.dk/linux-block:
  io_uring: make sure accept honor rlimit nofile
  io_uring: make sure openat/openat2 honor rlimit nofile
  io_uring: NULL-deref for IOSQE_{ASYNC,DRAIN}

4 years agoMerge branch 'turbostat' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux
Linus Torvalds [Sat, 21 Mar 2020 18:50:36 +0000 (11:50 -0700)]
Merge branch 'turbostat' of git://git./linux/kernel/git/lenb/linux

Pull turbostat updates from Len Brown:
 "Update to turbostat v20.03.20.

  These patches unlock the full turbostat features for some new
  machines, plus a couple other minor tweaks"

* 'turbostat' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
  tools/power turbostat: update version
  tools/power turbostat: Print cpuidle information
  tools/power turbostat: Fix 32-bit capabilities warning
  tools/power turbostat: Fix missing SYS_LPI counter on some Chromebooks
  tools/power turbostat: Support Elkhart Lake
  tools/power turbostat: Support Jasper Lake
  tools/power turbostat: Support Ice Lake server
  tools/power turbostat: Support Tiger Lake
  tools/power turbostat: Fix gcc build warnings
  tools/power turbostat: Support Cometlake

4 years agoMerge tag 'powerpc-5.6-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sat, 21 Mar 2020 15:51:45 +0000 (08:51 -0700)]
Merge tag 'powerpc-5.6-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "Two fixes for bugs introduced this cycle:

   - fix a crash when shutting down a KVM PR guest (our original style
     of KVM which doesn't use hypervisor mode)

   - fix for the recently added 32-bit KASAN_VMALLOC support

  Thanks to: Christophe Leroy, Greg Kurz, Sean Christopherson"

* tag 'powerpc-5.6-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  KVM: PPC: Fix kernel crash with PR KVM
  powerpc/kasan: Fix shadow memory protection with CONFIG_KASAN_VMALLOC

4 years agotools/power turbostat: update version
Len Brown [Mon, 18 Nov 2019 02:18:31 +0000 (21:18 -0500)]
tools/power turbostat: update version

A stitch in time saves nine.

Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Print cpuidle information
Len Brown [Sat, 21 Mar 2020 04:47:47 +0000 (00:47 -0400)]
tools/power turbostat: Print cpuidle information

Print cpuidle driver and governor.

Originally-by: Antti Laakso <antti.laakso@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
4 years agoMerge branch 'nvme-5.6-rc6' of git://git.infradead.org/nvme into block-5.6
Jens Axboe [Sat, 21 Mar 2020 01:14:16 +0000 (19:14 -0600)]
Merge branch 'nvme-5.6-rc6' of git://git.infradead.org/nvme into block-5.6

Pull NVMe fixes from Keith:

"Two late nvme fabrics fixes for 5.6: a double free with the rdma
 transport, and a regression fix for tcp; please pull."

* 'nvme-5.6-rc6' of git://git.infradead.org/nvme:
  nvmet-tcp: set MSG_MORE only if we actually have more to send
  nvme-rdma: Avoid double freeing of async event data

4 years agobtrfs: fix removal of raid[56|1c34} incompat flags after removing block group
Filipe Manana [Fri, 20 Mar 2020 18:43:48 +0000 (18:43 +0000)]
btrfs: fix removal of raid[56|1c34} incompat flags after removing block group

We are incorrectly dropping the raid56 and raid1c34 incompat flags when
there are still raid56 and raid1c34 block groups, not when we do not any
of those anymore. The logic just got unintentionally broken after adding
the support for the raid1c34 modes.

Fix this by clear the flags only if we do not have block groups with the
respective profiles.

Fixes: 9c907446dce3 ("btrfs: drop incompat bit for raid1c34 after last block group is gone")
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
4 years agonvmet-tcp: set MSG_MORE only if we actually have more to send
Sagi Grimberg [Thu, 12 Mar 2020 23:06:38 +0000 (16:06 -0700)]
nvmet-tcp: set MSG_MORE only if we actually have more to send

When we send PDU data, we want to optimize the tcp stack
operation if we have more data to send. So when we set MSG_MORE
when:
- We have more fragments coming in the batch, or
- We have a more data to send in this PDU
- We don't have a data digest trailer
- We optimize with the SUCCESS flag and omit the NVMe completion
  (used if sq_head pointer update is disabled)

This addresses a regression in QD=1 with SUCCESS flag optimization
as we unconditionally set MSG_MORE when we didn't actually have
more data to send.

Fixes: 70583295388a ("nvmet-tcp: implement C2HData SUCCESS optimization")
Reported-by: Mark Wunderlich <mark.wunderlich@intel.com>
Tested-by: Mark Wunderlich <mark.wunderlich@intel.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Keith Busch <kbusch@kernel.org>
4 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 20 Mar 2020 16:28:25 +0000 (09:28 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:

 - Fix panic() when it occurs during secondary CPU startup

 - Fix "kpti=off" when KASLR is enabled

 - Fix howler in compat syscall table for vDSO clock_getres() fallback

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: compat: Fix syscall number of compat_clock_getres
  arm64: kpti: Fix "kpti=off" when KASLR is enabled
  arm64: smp: fix crash_smp_send_stop() behaviour
  arm64: smp: fix smp_send_stop() behaviour

4 years agoMerge tag 'char-misc-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Fri, 20 Mar 2020 16:24:22 +0000 (09:24 -0700)]
Merge tag 'char-misc-5.6-rc7' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are some small different driver fixes for 5.6-rc7:

   - binderfs fix, yet again

   - slimbus new device id added

   - hwtracing bugfixes for reported issues and a new device id

  All of these have been in linux-next with no reported issues"

* tag 'char-misc-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  intel_th: pci: Add Elkhart Lake CPU support
  intel_th: Fix user-visible error codes
  intel_th: msu: Fix the unexpected state warning
  stm class: sys-t: Fix the use of time_after()
  slimbus: ngd: add v2.1.0 compatible
  binderfs: use refcount for binder control devices too

4 years agoMerge tag 'staging-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Fri, 20 Mar 2020 16:20:38 +0000 (09:20 -0700)]
Merge tag 'staging-5.6-rc7' of git://git./linux/kernel/git/gregkh/staging

Pull staging/IIO fixes from Greg KH:
 "Here are a number of small staging and IIO driver fixes for 5.6-rc7

  Nothing major here, just resolutions for some reported problems:
   - iio bugfixes for a number of different drivers
   - greybus loopback_test fixes
   - wfx driver fixes

  All of these have been in linux-next with no reported issues"

* tag 'staging-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: rtl8188eu: Add device id for MERCUSYS MW150US v2
  staging: greybus: loopback_test: fix potential path truncations
  staging: greybus: loopback_test: fix potential path truncation
  staging: greybus: loopback_test: fix poll-mask build breakage
  staging: wfx: fix RCU usage between hif_join() and ieee80211_bss_get_ie()
  staging: wfx: fix RCU usage in wfx_join_finalize()
  staging: wfx: make warning about pending frame less scary
  staging: wfx: fix lines ending with a comma instead of a semicolon
  staging: wfx: fix warning about freeing in-use mutex during device unregister
  staging/speakup: fix get_word non-space look-ahead
  iio: ping: set pa_laser_ping_cfg in of_ping_match
  iio: chemical: sps30: fix missing triggered buffer dependency
  iio: st_sensors: remap SMO8840 to LIS2DH12
  iio: light: vcnl4000: update sampling periods for vcnl4040
  iio: light: vcnl4000: update sampling periods for vcnl4200
  iio: accel: adxl372: Set iio_chan BE
  iio: magnetometer: ak8974: Fix negative raw values in sysfs
  iio: trigger: stm32-timer: disable master mode when stopping
  iio: adc: stm32-dfsdm: fix sleep in atomic context
  iio: adc: at91-sama5d2_adc: fix differential channels in triggered mode

4 years agoMerge tag 'usb-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Fri, 20 Mar 2020 16:16:35 +0000 (09:16 -0700)]
Merge tag 'usb-5.6-rc7' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some small USB fixes for 5.6-rc7. And there's a thunderbolt
  driver fix thrown in for good measure as well.

  These fixes are:
   - new device ids for usb-serial drivers
   - thunderbolt error code fix
   - xhci driver fixes
   - typec fixes
   - cdc-acm driver fixes
   - chipidea driver fix
   - more USB quirks added for devices that need them.

  All of these have been in linux-next with no reported issues"

* tag 'usb-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  USB: cdc-acm: fix rounding error in TIOCSSERIAL
  USB: cdc-acm: fix close_delay and closing_wait units in TIOCSSERIAL
  usb: quirks: add NO_LPM quirk for RTL8153 based ethernet adapters
  usb: chipidea: udc: fix sleeping function called from invalid context
  USB: serial: pl2303: add device-id for HP LD381
  USB: serial: option: add ME910G1 ECM composition 0x110b
  usb: host: xhci-plat: add a shutdown
  usb: typec: ucsi: displayport: Fix a potential race during registration
  usb: typec: ucsi: displayport: Fix NULL pointer dereference
  USB: Disable LPM on WD19's Realtek Hub
  usb: xhci: apply XHCI_SUSPEND_DELAY to AMD XHCI controller 1022:145c
  xhci: Do not open code __print_symbolic() in xhci trace events
  thunderbolt: Fix error code in tb_port_is_width_supported()

4 years agoMerge tag 'tty-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Fri, 20 Mar 2020 16:13:35 +0000 (09:13 -0700)]
Merge tag 'tty-5.6-rc7' of git://git./linux/kernel/git/gregkh/tty

Pull tty fixes from Greg KH:
 "Here are three small tty_io bugfixes for reported issues that Eric has
  resolved for 5.6-rc7

  All of these have been in linux-next with no reported issues"

* tag 'tty-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty: fix compat TIOCGSERIAL checking wrong function ptr
  tty: fix compat TIOCGSERIAL leaking uninitialized memory
  tty: drop outdated comments about release_tty() locking

4 years agoMerge tag 'sound-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 20 Mar 2020 16:10:29 +0000 (09:10 -0700)]
Merge tag 'sound-5.6-rc7' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A few fixes covering the issues reported by syzkaller, a couple of
  fixes for the MIDI decoding bug, and a few usual HD-audio quirks.

  Some of them are about ALSA core stuff, but they are small fixes just
  for corner cases, and nothing thrilling"

* tag 'sound-5.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda/realtek - Enable the headset of Acer N50-600 with ALC662
  ALSA: hda/realtek - Enable headset mic of Acer X2660G with ALC662
  ALSA: seq: oss: Fix running status after receiving sysex
  ALSA: seq: virmidi: Fix running status after receiving sysex
  ALSA: pcm: oss: Remove WARNING from snd_pcm_plug_alloc() checks
  ALSA: hda/realtek: Fix pop noise on ALC225
  ALSA: line6: Fix endless MIDI read loop
  ALSA: pcm: oss: Avoid plugin buffer overflow

4 years agoMerge tag 'drm-fixes-2020-03-20' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 20 Mar 2020 16:03:54 +0000 (09:03 -0700)]
Merge tag 'drm-fixes-2020-03-20' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Hope you are well hiding out above the garage. A few amdgpu changes
  but nothing too major. I've had a wisdom tooth out this week so
  haven't been to on top of things, but all seems good.

  core:
   - fix lease warning

  i915:
   - Track active elements during dequeue
   - Fix failure to handle all MCR ranges
   - Revert unnecessary workaround

  amdgpu:
   - Pageflip fix
   - VCN clockgating fixes
   - GPR debugfs fix for umr
   - GPU reset fix
   - eDP fix for MBP
   - DCN2.x fix

  dw-hdmi:
   - fix AVI frame colorimetry

  komeda:
   - fix compiler warning

  bochs:
   - downgrade a binding failure to a warning"

* tag 'drm-fixes-2020-03-20' of git://anongit.freedesktop.org/drm/drm:
  drm/amd/display: Fix pageflip event race condition for DCN.
  drm/amdgpu: fix typo for vcn2.5/jpeg2.5 idle check
  drm/amdgpu: fix typo for vcn2/jpeg2 idle check
  drm/amdgpu: fix typo for vcn1 idle check
  drm/lease: fix WARNING in idr_destroy
  drm/i915: Handle all MCR ranges
  Revert "drm/i915/tgl: Add extra hdc flush workaround"
  drm/i915/execlists: Track active elements during dequeue
  drm/bochs: downgrade pci_request_region failure from error to warning
  drm/amd/display: Add link_rate quirk for Apple 15" MBP 2017
  drm/amdgpu: add fbdev suspend/resume on gpu reset
  drm/amd/amdgpu: Fix GPR read from debugfs (v2)
  drm/amd/display: fix typos for dcn20_funcs and dcn21_funcs struct
  drm/komeda: mark PM functions as __maybe_unused
  drm/bridge: dw-hdmi: fix AVI frame colorimetry

4 years agoio_uring: make sure accept honor rlimit nofile
Jens Axboe [Fri, 20 Mar 2020 02:16:56 +0000 (20:16 -0600)]
io_uring: make sure accept honor rlimit nofile

Just like commit 4022e7af86be, this fixes the fact that
IORING_OP_ACCEPT ends up using get_unused_fd_flags(), which checks
current->signal->rlim[] for limits.

Add an extra argument to __sys_accept4_file() that allows us to pass
in the proper nofile limit, and grab it at request prep time.

Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: make sure openat/openat2 honor rlimit nofile
Jens Axboe [Fri, 20 Mar 2020 01:23:18 +0000 (19:23 -0600)]
io_uring: make sure openat/openat2 honor rlimit nofile

Dmitry reports that a test case shows that io_uring isn't honoring a
modified rlimit nofile setting. get_unused_fd_flags() checks the task
signal->rlimi[] for the limits. As this isn't easily inheritable,
provide a __get_unused_fd_flags() that takes the value instead. Then we
can grab it when the request is prepared (from the original task), and
pass that in when we do the async part part of the open.

Reported-by: Dmitry Kadashev <dkadashev@gmail.com>
Tested-by: Dmitry Kadashev <dkadashev@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agotools/power turbostat: Fix 32-bit capabilities warning
Len Brown [Fri, 20 Mar 2020 03:24:17 +0000 (23:24 -0400)]
tools/power turbostat: Fix 32-bit capabilities warning

warning: `turbostat' uses 32-bit capabilities (legacy support in use)

Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Fix missing SYS_LPI counter on some Chromebooks
Len Brown [Thu, 19 Mar 2020 22:26:05 +0000 (18:26 -0400)]
tools/power turbostat: Fix missing SYS_LPI counter on some Chromebooks

Some Chromebook BIOS' do not export an ACPI LPIT, which is how
Linux finds the residency counter for CPU and SYSTEM low power states,
that is exports in /sys/devices/system/cpu/cpuidle/*residency_us

When these sysfs attributes are missing, check the debugfs attrubte
from the pmc_core driver, which accesses the same counter value.

Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Support Elkhart Lake
Chen Yu [Wed, 18 Mar 2020 10:55:48 +0000 (18:55 +0800)]
tools/power turbostat: Support Elkhart Lake

From a turbostat point of view the Tremont-based Elkhart Lake
is very similar to Goldmont, reuse the code of Goldmont.

Elkhart Lake does not support 'group turbo limit counter'
nor C3, adjust the code accordingly.

Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Support Jasper Lake
Chen Yu [Tue, 14 Jan 2020 04:09:45 +0000 (12:09 +0800)]
tools/power turbostat: Support Jasper Lake

Jasper Lake, like Elkhart Lake, uses a Tremont CPU.
So reuse the code.

Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Support Ice Lake server
Chen Yu [Tue, 14 Jan 2020 04:06:49 +0000 (12:06 +0800)]
tools/power turbostat: Support Ice Lake server

From a turbostat point of view, Ice Lake server looks like Sky Lake server.

Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Support Tiger Lake
Chen Yu [Tue, 14 Jan 2020 04:04:55 +0000 (12:04 +0800)]
tools/power turbostat: Support Tiger Lake

From a turbostat point of view, Tiger Lake looks like Ice Lake.

Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Fix gcc build warnings
Len Brown [Thu, 19 Mar 2020 22:33:12 +0000 (18:33 -0400)]
tools/power turbostat: Fix gcc build warnings

Warning: ‘__builtin_strncpy’ specified bound 20 equals destination size
[-Wstringop-truncation]

reduce param to strncpy, to guarantee that a null byte is always copied
into destination buffer.

Signed-off-by: Len Brown <len.brown@intel.com>
4 years agotools/power turbostat: Support Cometlake
Chen Yu [Thu, 31 Oct 2019 13:14:22 +0000 (21:14 +0800)]
tools/power turbostat: Support Cometlake

From a turbostat point of view, Cometlake is like Kabylake.

Suggested-by: Rui Zhang <rui.zhang@intel.com>
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
4 years agoMerge tag 'drm-intel-fixes-2020-03-19' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Fri, 20 Mar 2020 02:51:34 +0000 (12:51 +1000)]
Merge tag 'drm-intel-fixes-2020-03-19' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

drm/i915 fixes for v5.6-rc7:
- Track active elements during dequeue
- Fix failure to handle all MCR ranges
- Revert unnecessary workaround

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/877dzgepvu.fsf@intel.com
4 years agoMerge tag 'amd-drm-fixes-5.6-2020-03-19' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Fri, 20 Mar 2020 02:48:10 +0000 (12:48 +1000)]
Merge tag 'amd-drm-fixes-5.6-2020-03-19' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

amd-drm-fixes-5.6-2020-03-19:

amdgpu:
- Pageflip fix
- VCN clockgating fixes
- GPR debugfs fix for umr
- GPU reset fix
- eDP fix for MBP
- DCN2.x fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200319204054.1036478-1-alexander.deucher@amd.com
4 years agoKVM: PPC: Fix kernel crash with PR KVM
Greg Kurz [Wed, 18 Mar 2020 17:43:30 +0000 (18:43 +0100)]
KVM: PPC: Fix kernel crash with PR KVM

With PR KVM, shutting down a VM causes the host kernel to crash:

[  314.219284] BUG: Unable to handle kernel data access on read at 0xc00800000176c638
[  314.219299] Faulting instruction address: 0xc008000000d4ddb0
cpu 0x0: Vector: 300 (Data Access) at [c00000036da077a0]
    pc: c008000000d4ddb0: kvmppc_mmu_pte_flush_all+0x68/0xd0 [kvm_pr]
    lr: c008000000d4dd94: kvmppc_mmu_pte_flush_all+0x4c/0xd0 [kvm_pr]
    sp: c00000036da07a30
   msr: 900000010280b033
   dar: c00800000176c638
 dsisr: 40000000
  current = 0xc00000036d4c0000
  paca    = 0xc000000001a00000   irqmask: 0x03   irq_happened: 0x01
    pid   = 1992, comm = qemu-system-ppc
Linux version 5.6.0-master-gku+ (greg@palmb) (gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)) #17 SMP Wed Mar 18 13:49:29 CET 2020
enter ? for help
[c00000036da07ab0c008000000d4fbe0 kvmppc_mmu_destroy_pr+0x28/0x60 [kvm_pr]
[c00000036da07ae0c0080000009eab8c kvmppc_mmu_destroy+0x34/0x50 [kvm]
[c00000036da07b00c0080000009e50c0 kvm_arch_vcpu_destroy+0x108/0x140 [kvm]
[c00000036da07b30c0080000009d1b50 kvm_vcpu_destroy+0x28/0x80 [kvm]
[c00000036da07b60c0080000009e4434 kvm_arch_destroy_vm+0xbc/0x190 [kvm]
[c00000036da07ba0c0080000009d9c2c kvm_put_kvm+0x1d4/0x3f0 [kvm]
[c00000036da07c00c0080000009da760 kvm_vm_release+0x38/0x60 [kvm]
[c00000036da07c30c000000000420be0 __fput+0xe0/0x310
[c00000036da07c90c0000000001747a0 task_work_run+0x150/0x1c0
[c00000036da07cf0c00000000014896c do_exit+0x44c/0xd00
[c00000036da07dc0c0000000001492f4 do_group_exit+0x64/0xd0
[c00000036da07e00c000000000149384 sys_exit_group+0x24/0x30
[c00000036da07e20c00000000000b9d0 system_call+0x5c/0x68

This is caused by a use-after-free in kvmppc_mmu_pte_flush_all()
which dereferences vcpu->arch.book3s which was previously freed by
kvmppc_core_vcpu_free_pr(). This happens because kvmppc_mmu_destroy()
is called after kvmppc_core_vcpu_free() since commit ff030fdf5573
("KVM: PPC: Move kvm_vcpu_init() invocation to common code").

The kvmppc_mmu_destroy() helper calls one of the following depending
on the KVM backend:

- kvmppc_mmu_destroy_hv() which does nothing (Book3s HV)

- kvmppc_mmu_destroy_pr() which undoes the effects of
  kvmppc_mmu_init() (Book3s PR 32-bit)

- kvmppc_mmu_destroy_pr() which undoes the effects of
  kvmppc_mmu_init() (Book3s PR 64-bit)

- kvmppc_mmu_destroy_e500() which does nothing (BookE e500/e500mc)

It turns out that this is only relevant to PR KVM actually. And both
32 and 64 backends need vcpu->arch.book3s to be valid when calling
kvmppc_mmu_destroy_pr(). So instead of calling kvmppc_mmu_destroy()
from kvm_arch_vcpu_destroy(), call kvmppc_mmu_destroy_pr() at the
beginning of kvmppc_core_vcpu_free_pr(). This is consistent with
kvmppc_mmu_init() being the last call in kvmppc_core_vcpu_create_pr().

For the same reason, if kvmppc_core_vcpu_create_pr() returns an
error then this means that kvmppc_mmu_init() was either not called
or failed, in which case kvmppc_mmu_destroy() should not be called.
Drop the line in the error path of kvm_arch_vcpu_create().

Fixes: ff030fdf5573 ("KVM: PPC: Move kvm_vcpu_init() invocation to common code")
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/158455341029.178873.15248663726399374882.stgit@bahia.lan
4 years agodrm/amd/display: Fix pageflip event race condition for DCN.
Mario Kleiner [Mon, 2 Mar 2020 06:17:32 +0000 (07:17 +0100)]
drm/amd/display: Fix pageflip event race condition for DCN.

Commit '16f17eda8bad ("drm/amd/display: Send vblank and user
events at vsartup for DCN")' introduces a new way of pageflip
completion handling for DCN, and some trouble.

The current implementation introduces a race condition, which
can cause pageflip completion events to be sent out one vblank
too early, thereby confusing userspace and causing flicker:

prepare_flip_isr():

1. Pageflip programming takes the ddev->event_lock.
2. Sets acrtc->pflip_status == AMDGPU_FLIP_SUBMITTED
3. Releases ddev->event_lock.

--> Deadline for surface address regs double-buffering passes on
    target pipe.

4. dc_commit_updates_for_stream() MMIO programs the new pageflip
   into hw, but too late for current vblank.

=> pflip_status == AMDGPU_FLIP_SUBMITTED, but flip won't complete
   in current vblank due to missing the double-buffering deadline
   by a tiny bit.

5. VSTARTUP trigger point in vblank is reached, VSTARTUP irq fires,
   dm_dcn_crtc_high_irq() gets called.

6. Detects pflip_status == AMDGPU_FLIP_SUBMITTED and assumes the
   pageflip has been completed/will complete in this vblank and
   sends out pageflip completion event to userspace and resets
   pflip_status = AMDGPU_FLIP_NONE.

=> Flip completion event sent out one vblank too early.

This behaviour has been observed during my testing with measurement
hardware a couple of time.

The commit message says that the extra flip event code was added to
dm_dcn_crtc_high_irq() to prevent missing to send out pageflip events
in case the pflip irq doesn't fire, because the "DCH HUBP" component
is clock gated and doesn't fire pflip irqs in that state. Also that
this clock gating may happen if no planes are active. This suggests
that the problem addressed by that commit can't happen if planes
are active.

The proposed solution is therefore to only execute the extra pflip
completion code iff the count of active planes is zero and otherwise
leave pflip completion handling to the pflip irq handler, for a
more race-free experience.

Note that i don't know if this fixes the problem the original commit
tried to address, as i don't know what the test scenario was. It
does fix the observed too early pageflip events though and points
out the problem introduced.

Fixes: 16f17eda8bad ("drm/amd/display: Send vblank and user events at vsartup for DCN")
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agoMerge tag 'mmc-v5.6-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Thu, 19 Mar 2020 19:45:14 +0000 (12:45 -0700)]
Merge tag 'mmc-v5.6-rc6' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:

 - rtsx_pci: Fix support for some various speed modes

 - sdhci-of-at91: Fix support for GPIO card detect on SAMA5D2

 - sdhci-cadence: Fix support for DDR52 speed mode for eMMC on UniPhier

 - sdhci-acpi: Fix broken WP support on Acer Aspire Switch 10

 - sdhci-acpi: Workaround FW bug for suspend on Lenovo Miix 320

* tag 'mmc-v5.6-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: rtsx_pci: Fix support for speed-modes that relies on tuning
  mmc: sdhci-of-at91: fix cd-gpios for SAMA5D2
  mmc: sdhci-cadence: set SDHCI_QUIRK2_PRESET_VALUE_BROKEN for UniPhier
  mmc: sdhci-acpi: Disable write protect detection on Acer Aspire Switch 10 (SW5-012)
  mmc: sdhci-acpi: Switch signal voltage back to 3.3V on suspend on external microSD on Lenovo Miix 320

4 years agoarm64: compat: Fix syscall number of compat_clock_getres
Vincenzo Frascino [Thu, 19 Mar 2020 14:11:38 +0000 (14:11 +0000)]
arm64: compat: Fix syscall number of compat_clock_getres

The syscall number of compat_clock_getres was erroneously set to 247
(__NR_io_cancel!) instead of 264. This causes the vDSO fallback of
clock_getres() to land on the wrong syscall for compat tasks.

Fix the numbering.

Cc: <stable@vger.kernel.org>
Fixes: 53c489e1dfeb6 ("arm64: compat: Add missing syscall numbers")
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
4 years agoMerge tag '5.6-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Thu, 19 Mar 2020 17:19:11 +0000 (10:19 -0700)]
Merge tag '5.6-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Three small smb3 fixes, two for stable"

* tag '5.6-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  CIFS: fiemap: do not return EINVAL if get nothing
  CIFS: Increment num_remote_opens stats counter even in case of smb2_query_dir_first
  cifs: potential unintitliazed error code in cifs_getattr()

4 years agoMerge tag 'kbuild-fixes-v5.6-3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 19 Mar 2020 16:57:48 +0000 (09:57 -0700)]
Merge tag 'kbuild-fixes-v5.6-3' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - fix __uint128_t capability test in Kconfig when GCC that defaults to
   32-bit is used to build the 64-bit kernel

 - suppress new noisy Clang warnings -Wpointer-to-enum-cast

 - move the namespace field in Module.symvers for the backward
   compatibility reason for the depmod tool

 - use available compression for initramdisk when INTRAMFS_SOURCE is
   defined, which was the original behavior

 - fix modpost to handle correct large section numbers when it refers to
   modversion CRCs and module namespaces

 - fix comments and documents

* tag 'kbuild-fixes-v5.6-3' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  scripts/kallsyms: fix wrong kallsyms_relative_base
  modpost: Get proper section index by get_secindex() instead of st_shndx
  initramfs: restore default compression behavior
  modpost: move the namespace field in Module.symvers last
  kbuild: Disable -Wpointer-to-enum-cast
  kbuild: doc: fix references to other documents
  int128: fix __uint128_t compiler test in Kconfig
  kconfig: introduce m32-flag and m64-flag
  kbuild: Fix inconsistent comment

4 years agortc: max8907: add missing select REGMAP_IRQ
Corentin Labbe [Wed, 18 Mar 2020 15:26:49 +0000 (15:26 +0000)]
rtc: max8907: add missing select REGMAP_IRQ

I have hit the following build error:

  armv7a-hardfloat-linux-gnueabi-ld: drivers/rtc/rtc-max8907.o: in function `max8907_rtc_probe':
  rtc-max8907.c:(.text+0x400): undefined reference to `regmap_irq_get_virq'

max8907 should select REGMAP_IRQ

Fixes: 94c01ab6d7544 ("rtc: add MAX8907 RTC driver")
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>