arm64: defconfig: imx8mmn_somdevices_defconfig: Avoid firmware load error and sysfs fallback.

As explained here:
https://patchwork.kernel.org/project/linux-arm-kernel/patch/20220107180314.1816515-6-marcel@ziswiler.com/
https://git.toradex.com/cgit/linux-toradex.git/commit/?h=toradex_4.14-2.0.x-imx&id=23abfc519183c2e6bc5644206177ef80b79f023e

Signed-off-by: Josep Orga <jorga@somdevices.com>

LF-5479: dmaengine: imx-sdma: sdma memcpy test fail

The sdma memcpy test fail due to the channel priority not configured.
The sdma_config_write-> sdma_config_channel-> sdma_set_channel_priority
didn't be called in memcpy, So there are only memcpy test fail.

The test fail log :

root@imx8mnevk:~# echo 2000 > /sys/module/dmatest/parameters/timeout
root@imx8mnevk:~# echo 1 > /sys/module/dmatest/parameters/iterations
root@imx8mnevk:~# echo dma0chan1 > /sys/module/dmatest/parameters/channel                                                                                                                                [  248.264090] dmatest: Added 1 threads using dma0chan1
root@imx8mnevk:~# echo 1 > /sys/module/dmatest/parameters/run
[  253.961791] dmatest: Started 1 threads using dma0chan1
root@imx8mnevk:~# [  255.967043] dmatest: dma0chan1-copy0: result #1: 'test timed out' with src_off=0xc4 dst_off=0x3c len=0x3ec0 (0)
[  255.977236] dmatest: dma0chan1-copy0: summary 1 tests, 1 failures 0.49 iops 7 KB/s (0)

This patch adds sdma_set_channel_priority for memcpy in transfer_init.

Fixes: 111097595c1c ("MLK-25465-1 dmaengine: imx-sdma: move runtime behind to transfer phase")
Signed-off-by: Joy Zou <joy.zou@nxp.com>
Reviewed by: Dong Aisheng <aisheng.dong@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-5353: dmaengine: imx-sdma: optimize confused runtime related code

In order to call sdma_runtime_resume, the driver call pm_runtime_get_sync
and pm_runtime_enable when the pm_runtime is false, it may be confused.
The handler of true and false pm_runtime are mixed in sdma_remove function.

This patch calls directly sdma_runtime_resume and remove pm_runtime_enable
when the pm_runtime is false, and separate the handler of true and false
pm_runtime.

Signed-off-by: Joy Zou <joy.zou@nxp.com>
Reviewed-by: Dong Aisheng <aisheng.dong@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-5353: dmaengine: imx-sdma: optimize external firmware load check

If external firmware has not been loaded, the sdma driver should return
as soon as possible for client using ram script.

The original firmware load check in sdma_transfer_init, if firmware
is not ready, the sdma_config_write also will run, it is unnecessary.

This patch checks firmware load status earlier and return fail immediately
in sdma_config_write if firmware not ready and use ram script.

Signed-off-by: Joy Zou <joy.zou@nxp.com>
Reviewed-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Reviewed-by: Dong Aisheng <aisheng.dong@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-5352: dmaengine: imx-sdma: fix the potential access registers without clock

The issue can be triggered with the sdma pm_runtime false.

If the dma client try to use sdma in boot phase, it may cause
system hang. The sdma driver really enable clk after sdma load
firmware, the clk_disable_unused will disable the SDMA1_ROOT_CLK
before the sdma driver enable clk. This issue will comes if access
the sdma register when the SDMA1_ROOT_CLK is disable.

This issue is very easy to reproduce with hdmi connected displayer
on imx7d-sbd board which the i2c will use sdma early before sdma
firmware loaded and clock enabled.

The client i2c calls dmaengine_prep_slave_single->device_prep_slave_sg
->sdma_prep_slave_sg->sdma_config_write->sdma_config_channel to get
descriptor in boot phase. Then, the function sdma_config_channel will
access sdma registers. If clk is disable will cause system hang.

This patch adds clk_enable and clk_disable to make sure the clk
enable before access sdma hardware.

Signed-off-by: Joy Zou <joy.zou@nxp.com>
Reviewed-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Reviewed-by: Dong Aisheng <aisheng.dong@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

MMFMWK-9038: dmaengine: imx-sdma: kernel dump met in boot

The issue can be triggered with the sdma pm_runtime true.

The dma clinet pcm calls snd_dmaengine_pcm_trigger->device_prep_dma_cyclic
->sdma_prep_dma_cyclic->pm_runtime_get_sync->sdma_transfer_init
->sdma_load_context to get descriptor. The sdma_buffer_descriptor
bd0 is allocated by sdma_runtime_resume. The sdma_runtime_resume
function will not be called because the runtime_status is
RPM_ACTIVE. The sdma_load_context accessing bd0 will cause dump.

The fail log as follow:

[   13.613406] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
[   13.622200] Mem abort info:
[   13.625010]   ESR = 0x96000044
[   13.628078]   EC = 0x25: DABT (current EL), IL = 32 bits
[   13.633402]   SET = 0, FnV = 0
[   13.636469]   EA = 0, S1PTW = 0
[   13.639620] Data abort info:
[   13.642515]   ISV = 0, ISS = 0x00000044
[   13.646364]   CM = 0, WnR = 1
[   13.649347] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000105ac5000
[   13.655800] [0000000000000000] pgd=0000000000000000, p4d=0000000000000000
[   13.662605] Internal error: Oops: 96000044 [#1] PREEMPT SMP
[   13.668186] Modules linked in:
[   13.671259] CPU: 3 PID: 695 Comm: alsa-sink-30c20 Not tainted 5.10.72-00003-g5f44cf7c854d-dirty #14
[   13.680315] Hardware name: NXP i.MX8MPlus EVK board (DT)
[   13.685646] pstate: 80000085 (Nzcv daIf -PAN -UAO -TCO BTYPE=--)
[   13.691671] pc : sdma_transfer_init+0x21c/0x26c
[   13.696223] lr : sdma_transfer_init+0x1bc/0x26c
[   13.700767] sp : ffff800012d9bb80
[   13.704095] x29: ffff800012d9bb80 x28: 0000000000000003
[   13.709425] x27: ffff0000c19c46a8 x26: 0000000000000000
[   13.714754] x25: 0000000000000080 x24: ffff0000c19c0080
[   13.720083] x23: 00000000000003c2 x22: 0000000000000020
[   13.725412] x21: ffff800011f4d200 x20: ffff0000d499a900
[   13.730741] x19: ffff0000c19c0298 x18: ffffffffffffffff
[   13.736069] x17: 0000000000000000 x16: 0000000000000000
[   13.741398] x15: ffff800092d9b7e7 x14: 3030303165303320
[   13.746726] x13: ffff800011b517b0 x12: 0000000000001000
[   13.752055] x11: 000000000000037f x10: ffff800011b517b0
[   13.757384] x9 : 0000000000000000 x8 : ffff800011f4d280
[   13.762712] x7 : 0000000000000000 x6 : 000000000000003f
[   13.768041] x5 : 0000000000000040 x4 : 0000000000000000
[   13.773369] x3 : 0000000000000004 x2 : 0000000000000000
[   13.778697] x1 : 0000000000000000 x0 : 0000000001830020
[   13.784025] Call trace:
[   13.786493]  sdma_transfer_init+0x21c/0x26c
[   13.790696]  sdma_prep_dma_cyclic+0x88/0x2e4
[   13.794984]  snd_dmaengine_pcm_trigger+0xec/0x1c0
[   13.799706]  dmaengine_pcm_trigger+0x18/0x24
[   13.803995]  snd_soc_pcm_component_trigger+0x48/0xf0
[   13.808977]  soc_pcm_trigger+0xa8/0xd0
[   13.812744]  snd_pcm_do_start+0x38/0x44
[   13.816597]  snd_pcm_action_single+0x48/0xac
[   13.820885]  snd_pcm_action+0x7c/0x9c
[   13.824566]  snd_pcm_action_lock_irq+0x8c/0xbc
[   13.829026]  snd_pcm_common_ioctl+0xfa4/0x11c0
[   13.833487]  snd_pcm_ioctl+0x34/0x50
[   13.837083]  __arm64_sys_ioctl+0xa8/0xf0
[   13.841024]  el0_svc_common.constprop.0+0x78/0x1c4
[   13.845829]  do_el0_svc+0x28/0x9c
[   13.849163]  el0_svc+0x14/0x20
[   13.852235]  el0_sync_handler+0xa4/0x130
[   13.856174]  el0_sync+0x180/0x1c0
[   13.859507] Code: b90026a0 52800400 531b6ad6 72a03060 (b9000340)
[   13.865605] ---[ end trace 0f90c4454359a810 ]---
[   13.870233] note: alsa-sink-30c20[695] exited with preempt_count 2

This patch removes pm_runtime_set_active in sdma_probe. The clk is
disable and the power is on when sdma probe, but the genpd_dev_pm_sync
will sync power. So we think the runtime_status is RPM_SUSPENDED. The
default runtime_status is RPM_SUSPENDED. This patch adds
pm_runtime_set_suspended in order to improve code readability. The
sdma_runtime_resume will be called.

This issue is very hard to reproduce with yocto rootfs.
If it uses auto login with ubuntu rootfs which is more easy to reproduce
on kernel 5.10, but not easy on kernel 5.15.

Reviewed-by: Dong Aisheng <aisheng.dong@nxp.com>
Signed-off-by: Joy Zou <joy.zou@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-4858: dmaengine: imx-sdma: retry load sdma firmware fail

If there is more than one sdma controller, the first controller
load sdma firmware using fallback failed in first time, it will
retry load sdma firmware again. But the second sdma controller
error state clean up may be still not finished (the firmware
request priv data has not been released). So the retry will fail
immediately due to get the old firmware priv data.

The fail log as follow:

[   62.505753] imx-sdma 30e10000.dma-controller: external firmware not found, using ROM firmware
[   62.599868] imx-sdma 30bd0000.dma-controller: firmware found.
[   62.605808] imx-sdma 30bd0000.dma-controller: loaded firmware 4.6

This patch adds a bit delay to wait for the second controller
firmware priv data released.

This issue is very hard to reproduce with yocto rootfs.
The 20ms delay is tested value with ubuntu rootfs which is more easy
to reproduce.

Signed-off-by: Joy Zou <joy.zou@nxp.com>
Reviewed-by: Dong Aisheng <aisheng.dong@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-4435: dmaengine: imx-sdma: fix kernel dump during MP3 decoding

update base_bd_ptr for channel0 after bd0 descriptor allocated again.
Otherwise, sdma may hang if the stale/wrong bd0 data touched by channel0
as below:

[  221.334360] imx-sdma 30e10000.dma-controller: Timeout waiting for CH0 ready
[  221.341333] imx-sdma 30e10000.dma-controller: save context error!
[  221.347437] PM: dpm_run_callback(): sdma_suspend+0x0/0x160 returns -110
[  221.354054] PM: Device 30e10000.dma-controller failed to suspend late: error -110
[  221.361937] PM: late suspend of devices failed
[  221.370127] ------------[ cut here ]------------
[  221.374745] lcdif_pixel_clk already disabled
[  221.379050] WARNING: CPU: 0 PID: 4510 at drivers/clk/clk.c:952 clk_core_disable+0xa4/0xb0
[  221.387222] Modules linked in: snvs_ui(O) fsl_jr_uio caam_jr caamkeyblob_desc caamhash_desc caamalg_desc crypto_engine rng_core authenc libdes crct10dif_ce imx8_media_dev(C) flexcan can_dev caam secvio error fuse [last unloaded: snvs_ui]
[  221.408283] CPU: 0 PID: 4510 Comm: rtc_wakeup.sh Tainted: G         C O      5.10.52-lts-5.10.y+g5788c4507376 #1
[  221.418451] Hardware name: NXP i.MX8MPlus EVK board (DT)
[  221.423761] pstate: 40000085 (nZcv daIf -PAN -UAO -TCO BTYPE=--)
[  221.429764] pc : clk_core_disable+0xa4/0xb0
[  221.433943] lr : clk_core_disable+0xa4/0xb0
[  221.438122] sp : ffff800012fdb820
[  221.441433] x29: ffff800012fdb820 x28: 0000000000000000
[  221.446743] x27: 0000000000000008 x26: ffff800011621f60
[  221.452053] x25: ffff0000c4b7f000 x24: ffff80001128d138
[  221.457364] x23: 0000000000000038 x22: ffff0000c18fea00
[  221.462674] x21: ffff800011d0ba70 x20: ffff0000c1873a00
[  221.467984] x19: ffff0000c1873a00 x18: 0000000000000030
[  221.473294] x17: 0000000000000000 x16: 0000000000000000
[  221.478604] x15: ffff0000c48d12b8 x14: ffffffffffffffff
[  221.483913] x13: ffff800011b51780 x12: 000000000000095a
[  221.489223] x11: 000000000000031e x10: ffff800011ba9780
[  221.494534] x9 : 00000000fffff000 x8 : ffff800011b51780
[  221.499844] x7 : ffff800011ba9780 x6 : 0000000000000000
[  221.505154] x5 : ffff00017f3ee900 x4 : 0000000000000000
[  221.510464] x3 : 0000000000000027 x2 : 0000000000000023
[  221.515774] x1 : f7f28654584c0c00 x0 : 0000000000000000
[  221.521085] Call trace:
[  221.523529]  clk_core_disable+0xa4/0xb0
[  221.527363]  clk_disable+0x34/0x50
[  221.530764]  lcdifv3_set_mode+0x40/0x2f0
[  221.534685]  lcdifv3_crtc_atomic_enable+0x88/0xcc
[  221.539387]  drm_atomic_helper_commit_modeset_enables+0x200/0x250
[  221.545478]  lcdifv3_drm_atomic_commit_tail+0x30/0x70
[  221.550526]  commit_tail+0xa0/0x180
[  221.554012]  drm_atomic_helper_commit+0x160/0x390
[  221.558714]  drm_atomic_commit+0x4c/0x60
[  221.562633]  drm_atomic_helper_commit_duplicated_state+0xf0/0x10c
[  221.568723]  drm_atomic_helper_resume+0x94/0x170
[  221.573338]  drm_mode_config_helper_resume+0x24/0x90
[  221.578299]  imx_drm_resume+0x14/0x20
[  221.581959]  platform_pm_resume+0x30/0x70
[  221.585968]  dpm_run_callback.constprop.0+0x3c/0xe4
[  221.590842]  device_resume+0x88/0x180
[  221.594501]  dpm_resume+0xe8/0x220
[  221.597900]  dpm_resume_end+0x18/0x30
[  221.601562]  suspend_devices_and_enter+0x1a4/0x5a0
[  221.606349]  pm_suspend+0x2e0/0x34c
[  221.609835]  state_store+0x8c/0x110
[  221.613323]  kobj_attr_store+0x1c/0x30
[  221.617070]  sysfs_kf_write+0x48/0x60
[  221.620730]  kernfs_fop_write_iter+0x118/0x1ac
[  221.625171]  new_sync_write+0xe8/0x180
[  221.628917]  vfs_write+0x244/0x2a4
[  221.632316]  ksys_write+0x6c/0x100
[  221.635714]  __arm64_sys_write+0x20/0x30
[  221.639636]  el0_svc_common.constprop.0+0x78/0x1a0
[  221.644423]  do_el0_svc+0x24/0x90
[  221.647737]  el0_svc+0x14/0x20
[  221.650789]  el0_sync_handler+0x1a4/0x1b0
[  221.654796]  el0_sync+0x180/0x1c0
[  221.658107] ---[ end trace c48b0b0e987f7565 ]---
[  221.662752] ------------[ cut here ]------------

Signed-off-by: Robin Gong <yibin.gong@nxp.com>
Reviewed-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

MLK-17285 dma: imx-sdma: fix sound mute in high loading

SDMA channel will stop in case of cpu high loading. That caused by
sdma ISR can't be handled in time while all cyclic bds consumed by
sdma side, then sdma channel stopped. Re-enable channel in this case
to workaound the issue.

Signed-off-by: Robin Gong <yibin.gong@nxp.com>
Reviewed-by: Shengjiu Wang <shengjiu.wang@nxp.com>
(cherry picked from commit 98c43d24d19d28ae364e7fedf48d61e98a2ec9e4)
(cherry picked from commit 65f426ddd092175602993cf4494e75f84f50b58a)

LF-3941: dmaengine: imx-sdma: fix audio noise with multi-fifo script

Correct chn_real_count for multi-fifo script.

Signed-off-by: Robin Gong <yibin.gong@nxp.com>
Reviewed-by: Shengjiu Wang <shengjiu.wang@nxp.com>
(cherry picked from commit 7a5f2c72fc2954853bb090af115555630ebd5116)

arm64: dts: imx8mn-somdevices.dtsi: Disable usbotg1 over current.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Disconnect over current USB signal.
· As GPIO1_IO13 is used as ecspi2 CS1 signal.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Add CS1 to ecspi2 and flexspi.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Disable BT from UART1.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: arm64: dts: Change old I2C4_SCL pin:
· PMIC INT changed to SD1_STROBE_GPIO2_IO11.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Change old UART4 pins:
· Ethernet phy reset changed to SAI2_MCLK_GPIO4_IO27.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Add UART4 peripheral.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Disable usbotg1 over current.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Disconnect over current USB signals.
· As GPIO1_IO13 is used as ecspi2 CS1 signal.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Add CS1 to ecspi2 and flexspi.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Disable BT from UART1.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Change old I2C4_SCL pin:
· PMIC INT changed to SD1_STROBE_GPIO2_IO11.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Change old UART4 pins:
· Ethernet phy reset changed to SAI2_MCLK_GPIO4_IO27.
· Pcie clkreq changed to I2C4_SCL_PCIE1_CLKREQ_B.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Add UART4 peripheral.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Enable PCIe and set PCIe reference clock as internal PLL.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Enable pca9450 level shifter.

Signed-off-by: Josep Orga <jorga@somdevices.com>

regulator: pca9450: Make I2C Level Translator configurable

Make the I2C Level Translator included in PCA9450 configurable from devicetree. The reset state is off. By setting nxp,i2c-lt-enable, the I2C Level Translator will be enabled while in STANDBY or RUN state.

Signed-off-by: Per-Daniel Olsson <perdo@axis.com>
Signed-off-by: Rickard x Andersson <rickaran@axis.com>
Link: https://lore.kernel.org/r/20220429072211.24957-2-rickaran@axis.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Josep Orga <jorga@somdevices.com>

regulator: Add property for I2C level shifter

By setting nxp,i2c-lt-enable the I2C level translator is enabled.

Signed-off-by: Per-Daniel Olsson <perdo@axis.com>
Signed-off-by: Rickard x Andersson <rickaran@axis.com>
Link: https://lore.kernel.org/r/20220429072211.24957-1-rickaran@axis.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: defconfig: imx8mmn_somdevices_defconfig: Set CONFIG_CRC_CCITT as y to match CONFIG_PPP_ASYNC configuration.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: defconfig: imx8mmn_somdevices_defconfig: Add modem detection and PPP configuration.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mmn somdevices dts: Add phy-reset-in-suspend to somdevices dtsi files fec.

Signed-off-by: Josep Orga <jorga@somdevices.com>

net: fec: Add phy-reset-in-suspend functionality.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices-C0P1.dts: Changed leds to gpio-leds.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Add flexspi device.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Add gpio pinctrl to i2c1 pins.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Changes to decrease temperature.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Add SN65DSI83 support and eeprom.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Add SPI and SPI CAN controller.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Set proper configuration to usb1 (otg).

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Set proper pins: · Set proper pins to pmic, usdhc2, mipi_csi, sai5 and uart2. · Added SMARC gpios.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Configure WiFi/BT pins.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Configure memory and allocate less cma memory.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Disabled not used peripherals.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices.dtsi: Removed not used peripherals.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: defconfig: Renamed defconfig to imx8mmn_somdevices_defconfig as somdevices imx8mm and imx8mn boards will use same defconfig.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices: Add support for AR8035 phy.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mn-somdevices: Changed heartbeat gpio led.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Added imx8mn-somdevices-C0P1.dts.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Added imx8mn-evk.dts features to imx8mn-somdevices.dtsi.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Added imx8mn-somdevices.dtsi copying imx8mn-evk.dtsi.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: video: Added somdevices fullscreen logo [1024x600]

Signed-off-by: Alejandro Benete <abenete@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Changes to decrease temperature.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: defconfig: imx8mm_somdevices_defconfig: Add CAN and SN65DSI83 support.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Add SN65DSI83 support and eeprom.

Signed-off-by: Josep Orga <jorga@somdevices.com>

drm: bridge: add support for TI SN65DSI83 MIPI DSI -> LVDS bridge.
Based on https://github.com/karo-electronics/karo-tx-linux/commits/imx_5.10.9_1.0.0/drivers/gpu/drm/bridge/sn65dsi83
Commits 8ddea41926c08f5afdec3589f57a63d1d28b7aec and 9512d3fdb4b00b7b575e7349b3cfcdc3b20a7306.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Add SPI CAN controller.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Set proper configuration to usb1 (otg) and add usb2 (host).

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Set proper pins:
· Set proper pins to pmic, pcie, usdhc2, mipi_csi, sai5 and uart2.
· Added SMARC gpios.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Configure WiFi/BT pins.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Configure memory and allocate less cma memory.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Disabled not used peripherals:
· Disabled mipi_csi and pcie.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: imx8mm-somdevices.dtsi: Removed not used peripherals.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: defconfig: Add imx8mm_somdevices_defconfig copying imx_v8_defconfig.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Add support for AR8035 phy.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Changed heartbeat gpio led.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Added imx8mm-somdevices-C0P1.dts.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Added imx8mm-evk.dts features to imx8mm-somdevices.dtsi.

Signed-off-by: Josep Orga <jorga@somdevices.com>

arm64: dts: Added imx8mm-somdevices.dtsi copying imx8mm-evk.dtsi.

Signed-off-by: Josep Orga <jorga@somdevices.com>

sdk_dpaa: rework private bpool seeding per CPU per interface

The dpa_bp introduction has modified the existing design of buffer
pool seeding. The most notable (and unintended) change was that
instead of placing in the buffer pool an allotment of buffers for
each CPU, for each interface, only an allotment for each CPU was
performed. With this, the possibility that in certain scenarios,
each interface can draw from the buffer pool enough buffers to
drain it completely, but not enough to trigger a refill. This may
result in a complete stall of the reception on the interface.

The current patch reintroduces the initial design, that makes sure
each probed interface contributes to the common buffer pool an
allotment of buffers for each CPU, removing the above mentioned
issue.

In consequence, the per buffer pool seed_cb() callback is removed.
The seeding is now done at net device probe.

Reported-by: Radu Bulie <radu-andrei.bulie@nxp.com>
Signed-off-by: Madalin Bucur <madalin.bucur@nxp.com>
Signed-off-by: Camelia Groza <camelia.groza@nxp.com>

Revert "swiotlb: don't modify orig_addr in swiotlb_tbl_sync_single"

This reverts commit 85a5a6875ca93dc4efbf20df942ba41d27a917e3.

The tlb_addr might not be aligned with tlb bucket size when call
swiotlb_tbl_sync_single, which will result in data corruption.
The issue was seen when do the tcrypt test: modprobe tcrypt mode=211 sec=1
The call stack as the followings: the tlb_addr might be: 0x00000000c280f9c0
for example, in one of the test:
testing speed of rfc4106(gcm(aes)) (rfc4106-gcm-aes-caam) decryption

[   64.985483] Hardware name: NXP i.MX8MPlus EVK board (DT)
[   64.990793] Call trace:
[   64.993245]  dump_backtrace+0x0/0x1a0
[   64.996909]  show_stack+0x18/0x70
[   65.000226]  dump_stack+0xd0/0x128
[   65.003629]  swiotlb_tbl_sync_single+0x16c/0x1bc
[   65.008248]  dma_sync_single_for_device+0xf4/0x12c
[   65.013054]  rfc4106_setkey+0x68/0x9c [caam_jr]
[   65.017586]  crypto_aead_setkey+0x2c/0x60
[   65.021600]  test_aead_speed.constprop.0+0x264/0x660 [tcrypt]
[   65.027346]  do_test+0x37fc/0x5c48 [tcrypt]
[   65.031531]  tcrypt_mod_init+0x58/0x1000 [tcrypt]
[   65.036235]  do_one_initcall+0x54/0x1c0
[   65.040071]  do_init_module+0x54/0x250
[   65.043819]  load_module+0x6d0/0x78c
[   65.047395]  __do_sys_finit_module+0xb8/0x100
[   65.051752]  __arm64_sys_finit_module+0x24/0x30
[   65.056283]  el0_svc_common.constprop.0+0x78/0x1a0
[   65.061073]  do_el0_svc+0x24/0x90
[   65.064389]  el0_svc+0x14/0x20
[   65.067444]  el0_sync_handler+0x1a4/0x1b0
[   65.071453]  el0_sync+0x178/0x180

The fix was to adjust orig_addr by adding the offset back by revert the patch.

Fixes:85a5a6875ca9("swiotlb: don't modify orig_addr in swiotlb_tbl_sync_single)
Signed-off-by: Jason Liu <jason.hui.liu@nxp.com>
Cc: Horia Geanta <horia.geanta@nxp.com>
Cc: Dong Aisheng <aisheng.dong@nxp.com>
Reviewed-by: Dong Aisheng <aisheng.dong@nxp.com>

net: phy: mscc: configure in-band auto-negotiation for VSC8514

Add the in-band configuration knob for the VSC8514 quad PHY. Tested with
QSGMII in-band AN both on and off on NXP LS1028A-RDB and T1040-RDB.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>

net: phylink: explicitly configure in-band autoneg for PHYs that support it

Currently Linux has no control over whether a MAC-to-PHY interface uses
in-band signaling or not, even though phylink has the
managed = "in-band-status";
property which denotes that the MAC expects in-band signaling to be used.

The problem is really that if the in-band signaling is configurable in
both the PHY and the MAC, there is a risk that they are out of sync
unless phylink manages them both. Most if not all in-band autoneg state
machines follow IEEE 802.3 clause 37, which means that they will not
change the operating mode of the SERDES lane from control to data mode
unless in-band AN completed successfully. Therefore traffic will not
work.

It is particularly unpleasant that currently, we assume that PHYs which
have configurable in-band AN come pre-configured from a prior boot stage
such as U-Boot, because once the bootloader changes, all bets are off.

Let's introduce a new PHY driver method for configuring in-band autoneg,
and make phylink be its first user. The main PHY library does not call
phy_config_inband_autoneg, because it does not know what to configure it
to. Presumably, non-phylink drivers can also call phy_config_inband_autoneg
individually.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>

sdk_dpaa: memac: fix AQR autoneg to lower than 10G partners

Stop removing unsupported system side modes with AQR PHYs. Originally,
their removal didn't impact the modes advertised by the PHY. Since
auto-negotiation support was added to the aquantia driver, the
unsupported modes were no longer advertised. This prohibited the
PHY from negotiating with 1G link partners on LS1046ARDB.

Patch adapted from similar fix for the upstream DPAA1 Ethernet driver:
commit fddb5a50801a ("dpaa_eth: support all modes with rate adapting PHYs")

Aquantia auto-negotiation:
commit 09c4c57f7bc4 ("net: phy: aquantia: add support for auto-negotiation
configuration")

Signed-off-by: Camelia Groza <camelia.groza@nxp.com>

fsl_qbman: remove reference to archdata.iommu_domain for v5.9

The dev->archdata.iommu_domain was removed in upstream kernel v5.9
in commit 6255c8c8d256 ("powerpc/dma: Remove dev->archdata.iommu_domain").

Signed-off-by: Camelia Groza <camelia.groza@nxp.com>

powerpc/configs/sdk_dpaa: update targets to corenet_base.config for v5.5

The base config was renamed in upstream commit 58b12eb28e34
("powerpc/configs: Rename foo_basic_defconfig to foo_base.config").

Signed-off-by: Camelia Groza <camelia.groza@nxp.com>

LF-3296-1 arm64:move elfcorehdr reservation early for crash dump kernel

elfcorehdr_addr is fixed address passed to 2nd kernel which may be conflicted
with potential reserved memory in 2nd kernel,so reserve_elfcorehdr() ahead of
early_init_fdt_scan_reserved_mem() can relieve this situation.

Signed-off-by: Nikhil Gupta <nikhil.gupta@nxp.com>
Signed-off-by: Poonam Aggrwal <poonam.aggrwal@nxp.com>
Reviewed-by: Priyanka Jain <priyanka.jain@nxp.com>
Reviewed-by: Aisheng Dong <aisheng.dong@nxp.com>

LF-3296-2 watchdog: sbsa_gwdt: add shutdown hook to driver

Kexec maybe need more time than timeout of wdt(sbsa)
which will reset the system.
So it is necessary to add shutdown hook to disable the
wdt when run kexec.

Signed-off-by: Zhao Qiang <qiang.zhao@nxp.com>
Reviewed-by: Priyanka Jain <priyanka.jain@nxp.com>

MLK-25508: dmaengine: imx-sdma: fix kernel dump in audio record

Check whether audiomix really off or not in runtime resume, if not, do
nothing at HW level including script load to avoid the below crash:

[   40.348730] kernel BUG at mm/vmalloc.c:2064!
[   40.353001] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   40.358486] Modules linked in:
[   40.361547] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.10.35-05035-g08680b128d38-dirty #490
[   40.369984] Hardware name: NXP i.MX8MPlus EVK board (DT)
[   40.375297] pstate: 00000005 (nzcv daif -PAN -UAO -TCO BTYPE=--)
[   40.381310] pc : __get_vm_area_node+0x164/0x188
[   40.385842] lr : vmap+0x84/0x100
[   40.389069] sp : ffff800011bdba80
[   40.392384] x29: ffff800011bdba80 x28: 0000000000000000
[   40.397700] x27: ffff0000c0447548 x26: ffff0000c69b80d4
[   40.403015] x25: 0000000000001000 x24: 006800000000070b
[   40.408333] x23: ffff0000c7399300 x22: 0000000000001000
[   40.413651] x21: 0000000000000010 x20: 0000000000000001
[   40.418972] x19: fffffdfffff01540 x18: 00000000000000c0
[   40.424290] x17: 0000000000000000 x16: 0000000000000000
[   40.429605] x15: fffffdfffff024c0 x14: fffffdfffff024c8
[   40.434921] x13: dead000000000122 x12: 0000000000001000
[   40.440236] x11: ffff00017f3fba08 x10: ffff80016dc2b000
[   40.445552] x9 : ffffffffffffffff x8 : 0000000000010001
[   40.450870] x7 : ffff8000100f51a0 x6 : 0000000000000cc0
[   40.456187] x5 : 00000000ffffffff x4 : fffffdffbfff0000
[   40.461505] x3 : ffff800010000000 x2 : 0000000000000010
[   40.466821] x1 : 0000000000000001 x0 : 0000000000001000
[   40.472137] Call trace:
[   40.474588]  __get_vm_area_node+0x164/0x188
[   40.478775]  dma_common_contiguous_remap+0xa8/0xe8
[   40.483568]  dma_direct_alloc+0x194/0x320
[   40.487581]  dma_alloc_attrs+0x7c/0xe0
[   40.491333]  sdma_load_script+0x64/0x170
[   40.495260]  sdma_runtime_resume+0x1cc/0x1e8
[   40.499533]  pm_generic_runtime_resume+0x28/0x40
[   40.504153]  __genpd_runtime_resume+0x2c/0xa0
[   40.508512]  genpd_runtime_resume+0x88/0x208
[   40.512784]  __rpm_callback+0xd4/0x148
[   40.516535]  rpm_callback+0x20/0x80
[   40.520025]  rpm_resume+0x39c/0x580
[   40.523515]  __pm_runtime_resume+0x38/0x80
[   40.527614]  sdma_int_handler+0x40/0x380
[   40.531539]  __handle_irq_event_percpu+0x4c/0x160
[   40.536243]  handle_irq_event_percpu+0x30/0x88
[   40.540686]  handle_irq_event+0x44/0xd8
[   40.544525]  handle_fasteoi_irq+0xb4/0x160
[   40.548621]  generic_handle_irq+0x2c/0x40
[   40.552632]  __handle_domain_irq+0x60/0xb8
[   40.556730]  gic_handle_irq+0x54/0x118
[   40.560482]  el1_irq+0xc0/0x180
[   40.563626]  cpuidle_enter_state+0x134/0x2e8
[   40.567898]  cpuidle_enter+0x34/0x48
[   40.571476]  call_cpuidle+0x18/0x38
[   40.574963]  do_idle+0x1f4/0x258
[   40.578190]  cpu_startup_entry+0x20/0x68
[   40.582116]  rest_init+0xd4/0xe0
[   40.585347]  arch_call_rest_init+0xc/0x14
[   40.589359]  start_kernel+0x470/0x4a4
[   40.593027] Code: aa1403e0 a94153f3 a8c67bfd d65f03c0 (d4210000)
[   40.599126] ---[ end trace 1207e0499c4741de ]---

Signed-off-by: Robin Gong <yibin.gong@nxp.com>
Reviewed-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-1723: spi: lpspi: run transfer speed_hz sanity check

Coverity reports Division or modulo by zero potential issue(CWE-369)
about the provided transfer speed_hz. Run sanity check first to avoid this
problem.

Signed-off-by: Han Xu <han.xu@nxp.com>
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Reviewed-by: Jun Li <jun.li@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

dpaa2-mac: workaround for zero advertising on the MAC

It seems there are some circumstances in the latest MC firmware versions
where an empty advertising field is returned by the dpmac_get_link_cfg()
command. This will eventually lead to the link not going up.
Work around this issue by skipping PHY reconfiguration if this happens.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>

dpaa2-eth: Set congestion group before the Rx queues become operational

Setting  the CGID for the first time or dynamically changing it
with a different one  on a FQ that has been in service (has frames in it)
is not supported, the byte/frame counts in both the old and new CGRs may be
corrupted if this is done.

This patch adds the configuration of the CGR(congestion group record)
for the network interface RX queues, in the probe function before the queues
become operational,  to prevent the corruption of the aforementioned counter.

Signed-off-by: Radu Bulie <radu-andrei.bulie@nxp.com>

dpaa2-console.c: free buffer before returning from dpaa2_console_read

Dpaa2_console_read function frees kbuf buffer before returning.
The variable no longer goes out of scope, leaking the storage it
points to.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>

dpaa2-eth: fix a possible null pointer dereference

In case the FD that we are cleaning up on the Tx confirmation path
is corrupted (carries a wrong SWA type), there will be no skb
to work with.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>

LF-3799 dpaa2-switch: convert BR_STATE enum to DPSW_STP_STATE enum

Starting with mc 10.28.0, the firmware returns an error any time
an invalid STP state is configured.
An explicit function converts from BR_STATE_* kernel enums to
DPSW_STP_STATE_* mc enums.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>

LF-3798: dprc-driver.c: resource pools are no longer initialized

dprc_scan_container() function is no longer called on a device
endpoint lookup, which, in turn, called fsl_mc_init_all_resource_pools.
Scanning for objects is done using the dprc_scan_objects()
function directly.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>

LF-3715-2 arm64: dts: imx8mp-evk: add "rtl821x,clkout-disable" property for PHY

Add "rtl821x,clkout-disable" property to disable CLKOUT clock for
Ethernet PHY.

Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Reviewed-by: Jun Li <jun.li@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3715-1 net: phy: realtek: change "rtl821x,clkout_en" to "rtl821x,clkout-disable"

For PHY RTL8211FDI-CG, CLKOUT clock default is enabled by hardware, use
"rtl821x,clkout-disable" property to disable CLKOUT if user required. If
use "rtl821x,clkout_en" property to enable CLKOUT clock, it will block existing
cases which need this CLKOUT for MAC but doesn't add this property. It's
safe to change "rtl821x,clkout_en" to "rtl821x,clkout-disable", as to be
removed property("rtl821x,clkout_en") never been used in the kernel
tree.

Camelia Groza validated on LS1043ARDB.

Tested-by: Camelia Groza <camelia.groza@nxp.com>
Reviewed-by: Camelia Groza <camelia.groza@nxp.com>
Reviewed-by: Jun Li <jun.li@nxp.com>
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3886 [#imx-2532] Coverity Issue:15625673 Uninitialized scalar variable

Signed-off-by: Nicușor Cîțu <nicusor.citu@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3811 net: fec: fix TX Bandwidth Fluctuations

MAC-NET IP with AVB support, queue 0 for best-effor, and queue 1&2 for AVB.
The default bandwidth configured of queue 1&2 is 50% in the driver, which
leads to TX bandwidth fluctuations when selecting queue 1&2 for transmitting
packets. Select queue based on VLAN info when AVB enabled.

Commit da722186f654 ("net: fec: set GPR bit on suspend by DT configuration.")
refactor the fec_devtype, need adjust downstream code accordingly.

Reviewed-by: Jun Li <jun.li@nxp.com>
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3831 net: stmmac: fix system hang if change mac address after interface ifdown

Fix system hang with below sequences:
~# ifconfig ethx down
~# ifconfig ethx hw ether xx:xx:xx:xx:xx:xx

After ethx down, stmmac all clocks gated off and then register access
causes system hang.

Reviewed-by: Jun Li <jun.li@nxp.com>
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3861 arm64: dts: imx8mm/n-evk: disable over current for usb1

Since imx8mm and imx8mn evk base board does not support over current
detection for usb1, so disable it.

Reviewed-by: Haibo Chen <haibo.chen@nxp.com>
Signed-off-by: Li Jun <jun.li@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3803: drm: imx: mhdp: Move phy_power function to encoder_enable

Split power up function to arc and phy only power up functions.
Move phy power up function to encoder_enable.

Signed-off-by: Sandor Yu <Sandor.yu@nxp.com>
Reviewed-by: Robby Cai <robby.cai@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3830 Revert "MLK-21499 arm: imx: put pads into OFF state before entering VLLS"

This reverts commit 6bec1553aaf630b80c9239d8895d423843f31801.

Put all the pads into off mode have side effect if the pad used for
GPIO interrupt. system suspend/resume will be broken, We can not
add it in our generic release.

Signed-off-by: Jacky Bai <ping.bai@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3794: ASoC: fsl_xcvr: Fix hang issue after suspend with RX recording

There is a unhandled interrupt after suspend, which cause endless
interrupt when system resume, so system will hang.

Disable all interrupt in runtime suspend callback.

Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Reviewed-by: Viorel Suman <viorel.suman@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3819: ASoC: fsl_spdif: Fix unexpected interrupt after suspend

When system enter suspend,  the machine driver suspend callback
function will be called, then the cpu driver trigger callback
(SNDRV_PCM_TRIGGER_SUSPEND) be called, the interrupt should be
disabled

But the machine driver suspend and cpu dai driver suspend order
maybe changed, the cpu dai driver's suspend callback is called before
machine driver's suppend callback, then the interrupt is not cleared
successfully in trigger callback.

So need to clear interrupts in cpu dai driver's suspend callback
to avoid such issue.

Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Reviewed-by: Viorel Suman <viorel.suman@nxp.com>
Acked-by: Jason Liu <jason.hui.liu@nxp.com>

LF-3779 remoteproc: imx_rproc: fix earlyboot detection

Here the earlyboot detection updated to use 0xB for the low four bits.
The previous detection not work, because with BIT0 set, BIT1 will
be automatically set, and BIT3 are also set by us.

So we detect BIT[3,1,0] here.

Reviewed-by: Ye Li <ye.li@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>

MLK-25413: fsl_xcvr: Fix channel swap issue with ARC

With pause and resume test for ARC, there is occassionally
channel swap. Change the clearing dpath operation after
the dma enablement to fix this issue.

Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Reviewed-by: Viorel Suman <viorel.suman@nxp.com>

LF-3773 media: isi: m2m: only fill colorspace when userspace don't touch them

Only fill colorspace, ycbcr_enc, xfer_func and quantization when userspace
don't touch them.

Signed-off-by: Bing Song <bing.song@nxp.com>
Signed-off-by: Guoniu.zhou <guoniu.zhou@nxp.com>
Reviewed-by: Robby Cai <robby.cai@nxp.com>

LF-3774-3 ASoC: SOF: Fix sof_ipc_msg_data for compress streams

sof_ipc_msg_data relies on sof_pcm_stream to retrieve the Host position
offset. This doesn't work for compress audio.

For compressed audio we need to retrieve position offset from
sof_compr_stream.

Signed-off-by: Daniel Baluta <daniel.baluta@nxp.com>

LF-3774-2 ASoC: SOF: imx: Use newly introduced generic IPC stream ops

This makes IMX use the newly introduced generic IPC ops
instead of imx specific ones.

Signed-off-by: Daniel Baluta <daniel.baluta@nxp.com>