BSP Layers and Reference Images for Yocto Project Issue Tracker

Description: U-boot build fails when more than one configuration is listed in UBOOT_CONFIG variable.

Description: Booting a FIT image that includes a RAM disk is not working. Boot fails with the following errors: `sha256Error: Address arguments are not aligned CAAM was not setup properly or it is faulty error!`, `sha256dev_get_priv: null device CAAM was not setup properly or it is faulty`.

Description: USB OTG Is Not Working In U-Boot

Description: Playing high resolution video (anything that is full-hd or beyond) is not working correctly because there is not enough CMA memory available.

Workaround: Increase CMA memory size in the device tree.

Description: PCIe devices, including Wi-FI, are not always detected at boot, failure rate is about 5%.

Description: USB Host doesn't work on Colibri i.MX8X with Aster device tree

Workaround: Use evaluation board device tree file

Description: U-Boot custom built may not work if size of SPL binary (u-boot-spl.bin) is bigger that 111236 bytes. In that case the SPL stack could override the DDR firmware leading to a failure during memory training and a boot failure.

Workaround: Use Toradex provided binaries, reduce generate binary size removing functionality (e.g. set CONFIG_FSL_CAAM=n) or use the kirkstone Toradex OE generated toolchain that is known to create working binaries.

Description: Playtime on analog audio is faster (2x), this happen around 15% of the time after a reboot.

Workaround: Set `assigned-clock-rates = <786432000>, <49152000>, <12288000>;` in DTS `spdif0` node.

Description: The wakealarm feature of RTC1 does not work when using the `sysfs` interface. Thus one can do a timebased wakeup from suspend from any source provided by the SoM.

Workaround: Use `rtcwake` tool.

Description: Bluetooth interface might not be properly initialized at boot, the following errors might happen: `Bluetooth: hci0: command 0x1003 tx timeout`, `Failed to start discovery: org.bluez.Error.NotReady`

Description: There is a performance and boot time regression on initial BSP 6 releases compared to BSP 5, e.g. boot time on Apalis iMX6 increased from ~18seconds to ~30seconds, the reason is that some unwanted debug kernel configuration was enabled.

Workaround: Set `CONFIG_DEBUG_WW_MUTEX_SLOWPATH=n` or `CONFIG_PROVE_LOCKING=n` in the Linux kernel configuration

Description: The NAU8822 audio codec available on the Verdin development board cannot be configured in bridge-tied load

Description: Colibri iMX6ULL doesn't wake up from suspend if the ethernet interface is disabled when the system is put to sleep. With the ethernet interface enabled, suspend / resume works.

Workaround: Don't disable the ethernet interface if you need to sleep/resume. Otherwise, use BSP 6.1 or later, this issue is already fixed there.

Description: The bootaux command is not enabled on the Verdin iMX8MM u-boot default configuration.

Workaround: If the bootaux command is needed, build a custom u-boot with CONFIG_IMX_BOOTAUX set.

Description: When no display is connected the graphical demo application is crashing

Description: Some OE overrides in our layers are using a no longer supported syntax, because of that Bluetooth does not work out-of-the-box on Verdin iMX8M Plus >= V1.1A.

Workaround: Manually execute hciattach: ``` /usr/bin/hciattach /dev/ttymxc3 any 115200 flow /usr/sbin/rfkill unblock bluetooth /usr/bin/hciconfig hci0 up # Change BT UART baudrate to a higher speed, for example 921600 /usr/bin/hcitool -i hci0 cmd 0x3F 0x09 0x00 0x10 0x0e 0x00 /bin/stty -F /dev/ttymxc3 921600 ```

Description: The default audio interface is not the analogue one

Workaround: Explicitly specify the required audio interface

Description: Bluealsa is not starting/working

Description: The Wi-Fi regulatory DB is not automatically loaded, because of that a limited number of Wi-Fi channels is available and the sample Wi-Fi AP configuration is not working out of the box.

Workaround: Manually reload the regulatory db: `iw reg reload` and `iw reg set `. Verify with `iw reg get`

Description: Wi-Fi connection is not stable, it stops working after less than a minute of stress testing.

Description: CAAM driver has some failure warning during boot (`caam_jr 30901000.jr: failed to flush job ring 0`, `caam_jr: probe of 30901000.jr failed with error -5`). JR0 is reserved for secure world usage and Linux kernel cannot use it.

Workaround: The error message is just cosmetic, no known issues.

Description: DSI-to-HDMI does not implement hot plug detect on event, a 10 seconds polling is used instead. Because of that timing could be incorrect when a different monitor is plugged in.

Description: USB Port works only in host mode.

Description: USB OTG device/host role switching is not working

Description: The U-Boot mmc info command may truncate the eMMC name by one character at the end.

Workaround: Use BSP6 or backport patch from upstream U-Boot where this got already fixed. https://source.denx.de/u-boot/u-boot/-/commit/84191f73847c351019bb50728d28220d91b0aee4

Description: The SLEEP_MOCI# is not properly configured on the Verdin iMX8M Plus, because of that any carrier board relying on this signal could have malfunctions. CAN/PCIe and USB are not working correctly on Dahlia because of that.

Description: Hot-plugging USB devices on the USB host port USB_2 does not work on a custom carrier board without a USB hub in-between. Note that current Toradex carrier boards (Dahlia and the Verdin Development Board) do have such a USB hub and therefore do NOT exhibit this issue as the USB hub itself is handling the hot-plugging. This issue is related to USB runtime power-management.

Workaround: Add `power-domains = <&usb_otg2_pd>;` property to `&usbphynop2` node in the device tree.

Description: USB Host doesn't work on Colibri i.MX8X with Aster device tree

Workaround: Use evaluation board device tree file

Description: Under very rare timing circumstances, a data corruption might occur on a dirty cache line that is evicted from the L1 Data Cache due to another cache line being fully written, see ARM Cortex-A9 Errata 845369 for details. This could happen because `CONFIG_ARM_ERRATA_845369` is not enabled in U-Boot configuration. Enabling this errata has some effect on memory benchmark results, but it is not expected to have a significant impact on real applications.

Workaround: Enable `CONFIG_ARM_ERRATA_845369` in U-Boot configuration.

Description: Hot-plugging USB devices on the USB host port USB_2 does not work on a custom carrier board without a USB hub in-between. Note that current Toradex carrier boards (Dahlia and the Verdin Development Board) do have such a USB hub and therefore do NOT exhibit this issue as the USB hub itself is handling the hot-plugging. This issue is related to USB runtime power-management.

Workaround: Add `power-domains = <&usb_otg2_pd>;` property to `&usbphynop2` node in the device tree.

Description: It is not possible to control the built-int Ixora V1.2 LEDs using the Apalis IMX8.

Description: Print systemd boot messages to the uart, the default console output

Description: MDIO read timeout is randomly popping up during link-up and causing ethernet to stop working

Workaround: Setting the interface down and up again fixes the problem

Description: MDIO read timeout can prevent proper ethernet initialization at boot

Workaround: Bring the interface down and up again

Description: Spi CAN Controller is receiving wrong data when frames of different length are sent

Description: The Verdin iMX8M Mini may crash during boot with the default configuration if the DSI to HDMI display adapter is NOT plugged in.

Workaround: Remove DSI to HDMI adapter overlay in case you do not have the adapter present (see also `skip_fdt_overlays`)

Description: Bring-up of PCIe fails with a PCIe PLL lock timeout at boot.

Workaround: Use BSP 5.2.0.

Description: In U-Boot the NAND storage is much slower compared to Linux (1MiB/s vs 30MiB/s)

Description: Suspend/Resume sometimes freezes when going to suspend. This happens about every 5 to 10 times when going to suspend.

Workaround: None

Description: The 0028 Apalis iMX6 Quad 2GB IT SoM may sporadically fail to boot in case the ambient temperature exceeds 70 degrees Celsius. The issue is likely caused by sub-optimum RAM timings. The issue is going to be resolved by a future software update (delivering updated RAM timings).

Workaround: Updating to BSP 5.4.0 or any newer release fixes the issue.

Description: The interactive U-Boot cfgblock create command wrongly always creates a ConfigBlock for the NAND rather than the eMMC version of the module.

Description: The following messages are shown upon resume:
[ 1912.202108] fec 30be0000.ethernet eth0: MDIO write timeout
[ 1912.202119] Micrel KSZ8041 30be0000.ethernet-1:00: failed to set led mode

Workaround: Cosmetic problem. Nothing needs to be done.

Description: SATA is not working stable.

Workaround: None

Description: When trying to resume an apalis-imx8 with WAKE1_MICO# signal it does not wake up and the module freezes. It seems to crash on resume.

Description: We may violate the on-module PHY reset timing and attempt communication with the PHY after resume while the PHY is still held in reset.

Workaround: In device tree delay the voltage regulator coupled to the fec/phy

Description: Software reboot may stall 1.5 minutes stopping Weston Wayland compositor.
Stopping Weston Wayland Compositor...
[* ] (1 of 2) A stop job is running for …ayland Compositor (13s / 1min 31s)
This does not always happen. Not sure whether it depends on how long the session was already active.

Workaround: Just be patient (;-p).

Description: cfgblock create <16-digit-serial> modifies the 7th digit to be always 0. This creates a wrong version information for modules with an assembly version 'K' or higher (e.g. Apalis iMX6 V1.1K)

Workaround: Start 'cfgblock create' without parameters to enter the data in the interactive mode.

Description: As version X1 of the camera adapter board uses different GPIOs as compared to the initial E1 version, the apalis-imx8_ar0521_overlay for the e-con Systems e-CAM55_CUMI0521_MOD camera does not work on BSP 5.1.0 Quarterly.

Workaround: You may find the following patch fixing the pin muxing for the Rev. X1 camera adapters:
https://git.toradex.com/cgit/device-tree-overlays.git/commit/?h=toradex_5.4-2.3.x-imx&id=08808b502c224e6143ae19518cf4f844ca2ee03c

Description: The following crash can be observed during system reboot:To reproduce please type, reboot command and check for trace from weston ........ [ 54.358713] weston: unhandled page fault (11) at 0x636f6c33, code 0x005 [ 54.365339] pgd = d33edf7e [ 54.368125] [636f6c33] *pgd=00000000 [ 54.371718] CPU: 0 PID: 495 Comm: weston Not tainted 5.4.91-5.1.0-devel+git.6ad99da2c675 #1 [ 54.380101] Hardware name: Freescale i.MX6 Ultralite (Device Tree) [ 54.386291] PC is at 0x76d5cc88 [ 54.389459] LR is at 0x76d5cc79 [ 54.392604] pc : [<76d5cc88>] lr : [<76d5cc79>] psr: 600b0030 [ 54.398895] sp : 7e84c698 ip : 76eb1bfc fp : 7e84c810 [ 54.404121] r10: 01cf4cd8 r9 : 7e84c9fc r8 : 01cf9988 [ 54.409371] r7 : 01cf5308 r6 : 01cf9988 r5 : 01d51d78 r4 : 636f6c2f [ 54.415901] r3 : 01d22b40 r2 : 01cf305c r1 : 02000000 r0 : 636f6c2f [ 54.422467] Flags: nZCv IRQs on FIQs on Mode USER_32 ISA Thumb Segment user [ 54.429907] Control: 10c5387d Table: 9530406a DAC: 00000055 [ 54.435659] CPU: 0 PID: 495 Comm: weston Not tainted 5.4.91-5.1.0-devel+git.6ad99da2c675 #1 [ 54.444013] Hardware name: Freescale i.MX6 Ultralite (Device Tree) [ 54.450222] [<8010db6c>] (unwind_backtrace) from [<8010ae84>] (show_stack+0x10/0x14) [ 54.457980] [<8010ae84>] (show_stack) from [<808c9514>] (dump_stack+0x90/0xa4) [ 54.465218] [<808c9514>] (dump_stack) from [<8011194c>] (__do_user_fault+0xfc/0x100) [ 54.472972] [<8011194c>] (__do_user_fault) from [<80111d2c>] (do_page_fault+0x354/0x394) [ 54.481074] [<80111d2c>] (do_page_fault) from [<80111ed4>] (do_DataAbort+0x3c/0xc0) [ 54.488741] [<80111ed4>] (do_DataAbort) from [<80101ddc>] (__dabt_usr+0x3c/0x40) [ 54.496138] Exception stack(0x952fdfb0 to 0x952fdff8) [ 54.501199] dfa0: 636f6c2f 02000000 01cf305c 01d22b40 [ 54.509385] dfc0: 636f6c2f 01d51d78 01cf9988 01cf5308 01cf9988 7e84c9fc 01cf4cd8 7e84c810 [ 54.517567] dfe0: 76eb1bfc 7e84c698 76d5cc79 76d5cc88 600b0030 ffffffff\ ........

Description: There is no output on the parallel RGB port which affects the operation of 7 inch capacitive and resistive touch panels.

Description: The XDG_RUNTIME_DIR variable required for graphical Wayland/Weston output is not set in an SSH shell.

Workaround: Use the local serial console instead.

Description: USB switches its role-state from client to host, client works but host does not enumerate plugged in devices.

Workaround: None

Description: HDMI driver is sometimes crashing when booted.

Workaround: Do not use CONFIG_DRM_CDNS_HDMI_CEC

Description: On certain modules USB Host does not work correctly and outputs an endless error "usb 3-1: device descriptor read/all, error -71"

Workaround: Use the adjusted pinmuxing for USB-HSIC &usbh1 node in devicetree from most recent images. http://git.toradex.com/cgit/linux-toradex.git/tree/arch/arm64/boot/dts/freescale/imx8-apalis-v1.1.dtsi?h=toradex_5.4-2.3.x-imx#n1192

Description: Add support for USB Host in U-Boot.

Description: Add STMPE ADC driver back that is missing.

Description: Adjust developer website to new paths used on BSP 5.

Description: The Apalis iMX8 HDMI has problems with some display resolutions. This ticket solves the issues.

Description: NXP i.MX 8QXP/8QM-based modules the buildinfo shows some garbage instead of IMX-MKIMAGE version used.

Description: WEIM pins missing in device tree.

Workaround: Copy the device tree setting from kernel branch 4.14-2.3.x-imx.

Description: Enabling fec2 in device tree will lead to eth0 and eth1 interfaces stop working.

Description: USB not working properly with multiple errors looking like
[ 12.054106] usb 1-1.1: new high-speed USB device number 7 using cdns-usb3
[ 12.199669] usb 1-1.1: unable to read config index 0 descriptor/all
[ 12.206008] usb 1-1.1: can't read configurations, error -71
[ 12.298077] usb 1-1.1: new high-speed USB device number 8 using cdns-usb3
[ 12.455115] usb 1-1.1: unable to read config index 0 descriptor/all
[ 12.461442] usb 1-1.1: can't read configurations, error -71

Description: Add usbgadget and libcomposite to upstream images.

Description: NXP issued an urgent notice about an issue they found in the PMIC driver which may damage i.MX 8M Mini and Nano application processors.

Workaround: Update to a later BSP version which has this addressed. You are encouraged to switch to BSP 5 since BSP 4 has been deferred (https://www.toradex.com/blog/torizon-yocto-project-long-term-support-alignment).

Description: If the HDMI firmware gets loaded in U-Boot, LVDS and HDMI won't work concurrently.

Workaround: Load HDMI firmware in linux from /lib/firmware like the BSP does it by default.

Description: Reset command doesn't work on Verdin iMX8M Mini.

Description: cfgblock create <16-digit-serial> modifies the 7th digit to be always 0. This creates a wrong version information for modules with an assembly version 'K' or higher (e.g. Apalis iMX6 V1.1K)

Workaround: Start 'cfgblock create' without parameters to enter the data in the interactive mode.

Description: The mwifiex_sdio driver goes into an unrecoverable state after the firmware of the SDIO Wi-Fi module has crashed. Note: this only covers the Wi-Fi only use case while NXP is still working on the concurrent Wi-Fi Bluetooth use case (see ELB-3039).

Workaround: Use the mwifiex_sdio driver from backports-5.4 for robust Wi-Fi functionality. Note: this only covers the Wi-Fi only use case while NXP is still working on the concurrent Wi-Fi Bluetooth use case (see ELB-3039).

Description: The Pin Settings in the Colibri iMX 6ULL for GPIOs are wrong.

Workaround: Use correct Pin Settings as documented in the Device Tree Customization article.

Description: Amend the following page with instructions on how to build for Apalis iMX8 and Colibri iMX8X, including the toolchain to use: https://developer.toradex.com/knowledge-base/build-u-boot-and-linux-kernel-from-source-code

Description: pixelclk-active = 1 still drives data at CLK in active low

Workaround: Use display resolution timings in kernel-parameters set in u-boot.

Description: While V1.0B modules with their bring-up BSP are fused and do boot from the primary eMMC hardware boot area partition their U-Boot environment still gets stored to the regular eMMC hardware user area partition which may corrupt the primary boot FAT partition.

Workaround: Do NOT save your U-Boot environment or your primary boot FAT partition will get corrupted.

Description: Move to a newer mainline U-Boot for future-prove support.

Description: Timeouts and reads from false channels may be observed when using STMPE811 ADC.

Description: During bring-up we disabled HS200 as it proved not to be stable but maybe that is just due to the SDHCI stack assuming a wrong signalling voltage of 3.3 volts: root@apalis-tk1:~# cat /sys/kernel/debug/mmc0/ios clock: 52000000 Hz vdd: 7 (1.65 - 1.95 V) bus mode: 2 (push-pull) chip select: 0 (don't care) power mode: 2 (on) bus width: 3 (8 bits) timing spec: 7 (sd uhs DDR50) signal voltage: 0 (3.30 V)

Description: U-Boot hangs early if certain UHS-I SD cards are used.

Description: The following pinmuxing error may be seen during boot: [ 20.255028] vf610-pinctrl 40048000.iomuxc: pin VF610_PAD_PTB21 already requested by 40048000.iomuxc; cannot claim for spi1.0 [ 20.281777] vf610-pinctrl 40048000.iomuxc: pin-43 (spi1.0) status -22 [ 20.296277] vf610-pinctrl 40048000.iomuxc: could not request pin 43 (VF610_PAD_PTB21) from group can_int on device 40048000.iomuxc [ 20.324217] mcp251x spi1.0: Error applying setting, reverse things back

Workaround: Luckily, due to the pin already being muxed as a GPIO by means of hogging it still actually functions properly.

Description: The on-module Wi-Fi might not resuming after suspend.

Workaround: Re-loading all Wi-Fi related kernel modules may help.

Description: Apalis TK1, Apalis/Colibri iMX6*/7: SGTL5000 audio codec may produce clack/pop sounds while playing.

Workaround: Fix removing all found clacks/pops produced by SGTL5000 audio codec available in -next branches of toradex_tk1_l4t_r21.7 and toradex_4.9-2.3.x-imx.

Description: The MCP2515 SPI CAN controller as found on the Colibri Evaluation Board is disabled by default in favour of toradex,evalspi SPIDEV. Let us change this like on the other Colibri modules.

Workaround: Just add Status = "disabled" to the toradex,evalspi device tree node and okay to the MCP2515 one.

Description: Resuming after suspend may not work.

Description: The following temperature trip points are in effect now:
Throttling: 95°C
Shutdown: 105°C

Description: Apalis TK1 running BSP 2.8b3 may experience issues connecting USB 2.0 devices to the OTG USB port.

Workaround: Modify the U-Boot defargs environmental variable from "usb_port_owner_info=3" to "usb_port_owner_info=2" to disable USB 3.0 for stable OTG operation. This is now the default as of BSP 2.8b6.

Description: Doing ifdown/ifup may leave Ethernet PHY in an unknown state just permanently blinking preventing any communication. This is caused by an Ethernet PHY power-down without physically powering down resp. rail leaving the PHY in an unknown state which later upon powering it back up fails locking PLL. We may just disable PHY power-down bit (already implemented on Colibri VF50/61) or where possible properly implement powering down resp. rail.

Description: Back feeding on Ethernet rail when Ethernet is powered down.

Description: Colibi imx6ull v1.0 supports BT 4.0 and BLE 4.1 Starting with imx6ull v1.1 we support BT\BLE 5.0

Description: The alternate function of the pins in the IOMUXC SNVS is wrongly displayed. These pins are always GPIO. (TAMPER[0-9], BOOT_MODE[0-1]. Fix this.

Description: The CAN muxing ALT2 (can1.RX, can1.TX) for these pins are labelled ftm1.RX, ftm1.TX. Fix this.

Description: The firmware might behave erratic/crash when loading using U-Boot bootaux command with an elf firmware.

Description: The Linux kernel 4.1.44-2.7.4 may report a wrong i.MX 6 SoC version upon boot:
[ 0.109969] CPU identified as i.MX6Q, silicon rev 0.6

Workaround: The later Linux kernel 4.9.87-2.8.3 does report it correctly:
[ 0.069884] CPU identified as i.MX6Q, silicon rev 1.6
You may find a fix on the toradex_4.1-2.0.x-imx-next branch.

Description: The Colibri iMX6ULL V1.0 and V1.1 HW has no discrete pull ups on the I2C bus to the touch controller. This may lead to unreliable communication with the internal 100k pull ups and the used 80kHz I2C clock.

Description: Apalis TK1 LVDS drive strength settings are hard-coded. Idea is to move those to the device tree as well.

Workaround: For now, settings can be changes in drivers/video/tegra/dc/sor.c file. Function tegra_dc_sor_enable_lvds contains writes to NV_SOR_LANE_DRIVE_CURRENT and NV_SOR_LANE4_DRIVE_CURRENT registers. Registers definitions can be found in the "Tegra K1 Reference Manual".

Description: Forward ported commit 7ad9771527d2 ("apalis-tk1: fix PCIe reset for reliable gigabit Ethernet operation") was never validated. Turns out it does not work and may lead to PCIe gigabit bring-up issues.

Description: Enforcing to build 32-bit binaries for native and portions thereof using -m32 in EXTRA_OEMAKE_class-native may lead to the following error when building e.g. u-boot-toradex:
cbootimage: Accessing a corrupted shared library

Workaround: The -m32 as specified in EXTRA_OEMAKE_class-native in recipes-bsp/cbootimage/cbootimage.bb of meta-toradex-tegra may be overridden to your liking. The same applies to the following as well:
meta-toradex-demos: recipes-devtools/mtd/mtd-utils_%.bbappend
meta-toradex-nxp: recipes-devtools/imx-loader/imx-loader_git.bb
meta-toradex-tegra: recipes-bsp/tegrarcm/tegrarcm.bb
meta-toradex-tegra: recipes-support/libcryptopp/libcryptopp_5.6.4.bb

Description: Allow using MXM or SODIMM pin numbers to get or set GPIOs. This also allows for easier code porting from one to another Toradex module.

Description: The apalis-tk1-mainline machine fails poweroff/shutdown and instead just reboots.

Workaround: The rocko-next branch of meta-toradex-tegra features a patch which disables the AMS AS3722 PMIC's auto power on when AC OK feature otherwise enabled by default and causing a reboot rather than full poweroff.

Description: SD card is not detected in U-Boot even after using "mmc rescan".

Workaround: Fix already available in 2016.11-toradex-next U-Boot branch.

Description: OpenCV4Tegra currently fails to build due to a missing dependency on libpng12.

Workaround: The libpng12 recipe has been re-added to the rocko-next branch of meta-toradex-tegra so OpenCV4Tegra is at least buildable again. However, for full OpenCV4Tegra support, we recommend using NVIDIA's JetPack resp. Linux for Tegra aka L4T as explained in the following article on our developer website:
https://developer.toradex.com/knowledge-base/installing-nvidia-jetpack-with-l4t-on-apalis-tk1

Description: Building 2.8b2 and earlier fails on Ubuntu 18.04 LTS.

Workaround: It works on Ubuntu 16.04 LTS.

Description: While BSP 2.8b1 enumerated PCIe on the Apalis evaluation board just fine BSP 2.8b2 may fail with the following error message:
busn_res: can not insert [bus 02-01] under [bus 01] (conflicts with (null) [bus 01])

Workaround: As a workaround you may just revert the following offending commit for now:
http://git.toradex.com/cgit/linux-toradex.git/commit/?h=toradex_4.9-1.0.x-imx&id=9a4bf05126f42c2632729ab0da503021d74ed454

Description: RS485 with using RTS to control the transceiver is not working. The RTS signal is not asserted during transmission. This is a regression in 2.8 BSPs (Linux kernel 4.9).

Workaround: Use ioctl to configure RS-485 and set the flag SER_RS485_RX_DURING_TX and SER_RS485_RTS_AFTER_SEND.

Description: When the UART with DMA (default) receives a character with framing errors (e.g. due to a break signal) the FIFO can get out of sync. Once out of sync, characters appear delayed only after new characters have been received.

Workaround: Reopen UART in case overruns have been detected (see /proc/tty/driver/fsl-lpuart). Use UART without DMA which handles Framing Errors correctly.

Description: While all other eMMC based Toradex SoMs do support the fatwrite U-Boot command the Colibri iMX7 Dual 1GB currently does not. This is especially sad should anybody try using any legacy update procedure e.g. just for flashing the M4 firmware.

Workaround: For now just use UMS, do it from Embedded Linux user space or use the Toradex Easy Installer with a custom image JSON.

Description: Installing a single backport kernel module to an image at build time fails (e.g. by using IMAGE_INSTALL_append in local.conf).

Workaround: Install all backport kernel modules by using the "backports" package.

Description: OE-Core layer did backport more package splits in the linux-firmware recipe. Due to this the HEAD of the morty branch clashes now with a similar split we did in meta-toradex-bsp-common resulting in a fatal QA error. (Compare with commit 0eaef98) ERROR: linux-firmware-1_0.0+gitAUTOINC+bf04291309-r0 do_package: QA Issue: linux-firmware-sd8887 is listed in PACKAGES multiple times, this leads to packaging errors. [packages-list]

Workaround: Delete the lines PACKAGES_prepend = "\ ${PN}-sd8887 \ " from meta-toradex-bsp-common/recipes-kernel/linux-firmware/linux-firmware_%.bbappend

Description: PCIe enumeration aka "pci enum" U-Boot command may fail in U-Boot due to some I2C/Power management issue.

Description: SPDIF is not working in the V2.6 beta 2 release. This is a regression against older releases.

Workaround: Fixes found in the -next branches of our 3.14, 4.1 and 4.9 based Linux kernels.

Description: The update to rocko (BSP 2.8b1) made automount fail for systems which do not trigger mounting through dbus/udisk2, i.e. which rely on udev rules.

Description: When building 2.8b1 building stress fails with "ERROR: stress-1.0.4-r0 do_fetch: Fetcher failure for URL: 'http://people.seas.harvard.edu/~apw/stress/stress-1.0.4.tar.gz'. Checksum mismatch!".

Workaround: Use 2.8b2 or remove the stress recipe under "meta-toradex-demos/recipes-support/stress/stress_1.0.4.bb" (this will use a fixed version of stress in openembedded-core/meta/recipes-extended/).

Description: In the demo images for BSP 2.8b1 the image viewer gpicview is not installed.

Workaround: Remove adwaita-icon-theme from BAD_RECOMMENDATIONS when building the image with openembedded.

Description: The thermal sensor driver on i.MX 7Solo does not probe successfully. There is no thermal sensor data available.

Description: During boot-up of Colibri iMX7 Dual 1GB (with eMMC flash) sometimes the system freezes right after "Starting File System Check on Root Device".

Description: Target is to achieve >2k msg/s (1Mbit/s) on both CAN channels.

Description: Pads not used in U-Boot are accessible via GPIO command. However, SODIMM 169 is currently not available and is driven high at the moment in U-Boot.

Description: Apalis TK1 did not wake from sleep if using WAKE1_MICO#.

Description: On Apalis/Colibri T20/T30 both fchdir() and fstat() fail with EBADF when the file was previously opened with the O_PATH flag.

Workaround: Required patches have been backported to our downstream Linux kernel: http://git.toradex.com/cgit/linux-toradex.git/log/?h=tegra-next

Description: Use ISO 8601 date format in the release_date property of the Toradex Easy Installer image JSON descriptor file.

Description: While BSP 2.8b1 does no longer require any special handling older BSPs apalis-tk1-mainline Mesa needed DRI_PRIME=1 to be defined as environment variable in order for OpenGL ES applications like glmark2-es to work right.

Workaround: In older BSPs adding DRI_PRIME=1 environment variable is required for OpenGL ES.

Description: DDC/EDID detection was failing using Apalis TK1 V1.2A with 2.7b4 on Ixora.

Description: Camera support has been greatly improved by an update to the OV5640 driver from our camera partner Antmicro as well as NVIDIA's latest L4T R21.6 update.

Description: While the Apalis PWM3 pin is shared with the Apalis BKL1_PWM pin currently the former does not have the PWM pin muxing activated. Note that exporting PWM3 for raw sysfs access will therefore fail and one has to revert to using the pwm backlight.

Workaround: Change resp. pin muxing in the device tree. You may find a patch on our toradex_tk1_l4t_r21.5-next branch.

Description: Target is to achieve >2k msg/s on both CAN channels.

Description: The kernel does not know about rw,noatime. The string rw by itself is valid, and to pass additional arguments rootflags=noatime can be used. The kernel mounts ext3/ext4 file systems by default ro, hence so far the kernel mounted the rootfs read-only. The init manager systemd then remounts the rootfs after file system checking. This makes sense and does not add measurable boot time. Let the kernel explicitly mount ro on block device based rootfs. The root mount flag noatime is only relevant when re-mounting read-write, hence it needs to be present in /etc/fstab. UBIFS used on raw NAND devices has no user space file checking, hence there is no value to mount it in read-only first. Also mount time for UBIFS is larger, and mounting first read-only and then read-write has higher boot time (300ms, likely due to missing block cache). Let the kernel explicitly mount rw on raw NAND based rootfs. UBIFS has no atime support, hence no root mount flags needed.

Workaround: The change to initially read-only mounted root file system on Apalis TK1 requires the U-Boot emmcargs environment variable to be adjusted to run NVIDIA's JetPack resp. L4T as documented here: http://developer.toradex.com/knowledge-base/installing-nvidia-jetpack-with-l4t-on-apalis-tk1#Flashing_Latest_Boot_Loader

Description: If an image creates a rootfs which is exactly a multiple of 64 MByte and then if one executes in U-Boot 'run update' the detection that all 64 MByte chunks have been flashed fails and the update never completes. Note that the Toradex Easy Installer is not affected.

Workaround: Check the last rootfs chunk on the update media and create a zero length chunk with incremented number, e.g.: $ ls -l //colibri-imx6/ ... -rw-r--r--. 1 root root 67108864 Jul 12 15:13 root.ext3-106 -rw-r--r--. 1 root root 67108864 Jul 12 15:13 root.ext3-107 -rw-r--r--. 1 root root 48128 Jul 12 15:13 SPL ... $ touch //colibri-imx6/root.ext3-108

Description: Note: This is only an issue for Colibri iMX7S 256MB V1.1C and Colibri iMX7D 512MB V1.1D and later modules.
As it turns out not using LDO2 anywhere at all in the regular case is a bad idea leading to the following:
[ 2.317688] LDO2: disabling
Of course display and SD card still worked on the EvalBoard due to back feeding through some pull-ups however on Viola with just one pull-up it failed once both display and SD card were active...
Currently without an SD card inserted the regulator framework actually powers the LDO2 off however e.g. the BL_ON pin is also on that same rail!

Workaround: The following fixes were integrated:
http://git.toradex.com/cgit/linux-toradex.git/commit/?h=toradex_4.1-2.0.x-imx&id=0284f779bba00b930bd9bb12703e2092478531c7
http://git.toradex.com/cgit/linux-toradex.git/commit/?h=toradex_4.1-2.0.x-imx&id=ac8dcd0d7b7d17ad8ac42b42cb1f51d4417f4950
Fix this by specifying LDO2 to be always on and use the regular 3.3V supply from the carrier board as the vmmc-supply.
Specify that carrier board 3.3V supply to be always on as well to avoid the following misleading message during boot:
[ 2.299654] 3.3V: disabling
One remaining issue is that now in suspend LDO2 is still powered and even adding the following did not help:
regulator-state-mem {
    regulator-off-in-suspend;
};
Further investigation pending.

Description: Unfortunately our current Angstrom/OpenEmbedded/Yocto BSP exclusively uses gstreamer 1.0 which we learned from multiple sources is not really production ready for TK1. We may consider switching to or at least including gstreamer 0.10 in the future as well.

Workaround: Use JetPack for now which allows installing both gstreamer 0.10 as well as 1.0. http://developer.toradex.com/knowledge-base/installing-nvidia-jetpack-with-l4t-on-apalis-tk1

Description: Currently various things may fail or behave strangely on certain modules if the config block is missing. Unify that behaviour by e.g. rather not allowing use of any features (e.g. Ethernet, RNDIS) at all in this case.

Workaround: Use the 'cfgblock create' U-Boot command to restore the config block. Alternatively we implemented similar functionality into our Toradex Easy Installer.

Description: Our 2016.11 U-Boot recipe in meta-toradex-nxp fails to build. - The RPM packaging errors out because the generated spec file contains a dash in the Release variable: Release: 2.7.2-gitr60021a4

Workaround: - Edit recipes-bsp/u-boot/u-boot-toradex_2016.11.bbappend so that the PR variable does not contain a dash ('-').

Description: USB Gadget schema fails to load on Apalis iMX6 Quad 2GB IT and Apalis iMX6 Dual 512MB modules which leads to RNDIS not working.

Description: In the 3.14 based kernels the device tree property "linux,rs485-enabled-at-boot-time" allowed to enable RS485 at boot time. This device tree property so far had no effect on 4.1 based releases.

Workaround: This is now implemented in our toradex_4.1-2.0.x-imx-next branch.

Description: Specifying display timings in device tree and only using minimal boot arguments (e.g. video=mxcfb3:dev=ldb,bpp=32) leads to a kernel crash.

Description: Using ACL with ubifs seems to increase the likelihood of filesystem corruption. Since ACL is an optional features and unlikely necessary on an embedded system switch ACL off.

Description: Depending on its DVFS state right before hardware reset the Apalis/Colibri T30 module may hang during a subsequent boot. Software reset seems not affected.

Workaround: Either power-cycle or make sure you are running at a higher DVFS state upon hardware reset. A fix for this called 'apalis/colibri_t30: improve vdd core pmic handling' may be found on our U-Boot 2016.11-toradex-next branch.

Description: Apalis iMX6: The gpio-tool can not read or change the current pinmuxing of the NANDF_DATA4 pin but rather the SD4_DAT6 pin got changed instead. Apalis TK1: Some of the Apalis GPIOs were not accessible with the gpio-tool.

Workaround: Don't use the gpio-tool for pin NANDF_DATA4 on Apalis iMX6.

Description: Systemd may hang upon first boot due to certain services like run-postinsts and opkg-configure triggering a configuration reload later causing dynamic device services like the dev-ttyS0 one to hang.

Workaround: A fix has been implemented in our -next branch.

Description: Colibri T20 modules may draw more power be it during regular operation or in suspend.

Workaround: This issue is related to the 'Module May Hang During Boot if HDMI is not Connected' one whose fix can be found on the tegra-next branch of our linux-toradex Linux kernel repository: http://git.toradex.com/cgit/linux-toradex.git/log/?h=tegra-next

Description: The wrap-up script in the Toradex Easy Installer Colibri iMX6 V2.7 Beta 1 image does not allow to flash Colibri iMX6 modules.

Description: In the U-Boot version 2016.11 as shipped with our BSP V2.7 beta 1 the USB host mode initialisation (usb start) returns "Port not available." on both ports. This issue has been fixed and is available on our -next branch.

Description: After the RTC has been under-powered, some RTC deliver invalid date/time. The external RTC currently shipped on Toradex evaluation boards has a oscillator fail bit which indicates when RTC does not contain a valid date/time.

Description: On Colibri VF50/VF61 in situations with frequent unexpected power-cuts there is a possibility of filesystem getting corrupted.

Workaround: Use a Read-Only Root Filesystem.

Description: The elf file format is supported by most commercial and free development environments and provides a standardized header for crucial information such as load address and entry point.

Description: Testing various SDIO based Wi-Fi solutions we observed it to completely lock-up the system.

Workaround: Find the following commit fixing this: http://git.toradex.com/cgit/linux-toradex.git/commit/?h=toradex_tk1_l4t_r21.5&id=20e255fde2d32d8bccd9e589fad4a2f663fcfb3a

Description: Rarely a timeout waiting for dev-ttys0.device on first boot has been observed.

Workaround: Just power-cycle again.

Description: With the move to morty all SysV initscripts have been installed which caused interference. This got fixed by finally getting rid of all SysV remnants.

Description: Using keypad driver causes kernel freeze on iMX7

Workaround: Use the generic GPIO driven matrix keyboard driver

Description: The SD card 1-bit operation mode, enabled by seeting the 'bus-width' property of the device tree 'esdhc' node to <1>, not work while using SD card. It was tested with Colibri VFxx.

Description: Our current PEX 8605 errata workaround violates the PCIe clock and reset specification should no PCIe switch be present as e.g. the case on Ixora.

Description: Our current update procedure reads the rootfs payload in pre split files of 64 MB size. The number of chunks is limited to 89 resulting in a maximum size of 5.5 GB.

Workaround: - use the ums command in U-Boot to export the modules eMMC as a mass storage device. - deploy a minimum rootfs and install the 'big' payload after booting into Linux

Description: Using rpmsg and exercising the memory subsystem leads to unrecoverable kernel crashes.

Workaround: Use "linux,usable-memory = <0x80000000 0xff00000>;" for iMX 7Solo or "linux,usable-memory = <0x80000000 0x1ff00000>;" for i.MX 7Dual in the device tree memory node.

Description: In later OE versions the default SSH config changed so that X11 forwarding is not allowed. $ ssh -X root@192.168.10.185 X11 forwarding request failed on channel 0

Workaround: Edit /etc/ssh/sshd_config to include X11Forwarding yes

Description: When connecting the module using USB device connection, the module provides an IP address via it's own DHCP server. The DHCP server uses DHCP Option 3 (Router) to announce itself as router which can lead to internet connection interruption on the host.

Description: OE migrated the sato image to GTK+ 3. This includes pcman-fm with its libfm. As lxpanel is using libfm we have to follow here or provide our own pcman-fm/libfm recipes. If we move to GTK+ 3 I guess we should do it with all GTK+ based software. e.g all of lxde/lxdm (which may have some issues), the GPIOconfig tool and tegrastats-gtk.

Description: As that whole source mirror stuff has been introduced rather haphazardly whoever did that also failed maintaining/updating it since V2.5 e.g. http://sources.toradex.com/V2.5/. The wohle process around this is currently unclear.

Workaround: http://git.toradex.com/cgit/meta-toradex-demos.git/commit/?h=jethro&id=a6c6fa422e9cf9586e0eb8aba40e0460513edf58

Description: VisionWorks demo achieves 10FPS on Apalis TK1 & 30FPS on Jetson TK1. Please identify the cause and resolve it.

Workaround: Fixed by a combination of Norrin soctherm and Laguna cl-dvfs just like used on Jetson TK1: http://git.toradex.com/cgit/linux-toradex.git/commit/?h=toradex_tk1_l4t_r21.5&id=3c204d1590b9fcead7169c7f066f6918bcbde55f

Description: Changing the default settings of multiplexed pins of the camera input allows saving up to 100mW of power.

Description: Using Colibri iMX7S freezes when using Viola. Could be a boot rom issue.

Workaround: Disable SNVS RTC in the device tree.

Description: When booting the M4 core using U-Boot (m4boot) rpmsg seems not to work (the kernel module loads but the initial test message does not seem to arrive/be recognized at the FreeRTOS side).

Workaround: Use the remoteproc method to boot the M4 core.

Description: Even if a USB controller is in peripheral mode, loading the USB Gadget schema fails with "Failed to start Load default USB gadget schema g1.schema.".

Workaround: Remove the command (ExecStartPost) copying the serial-number to the USB Gadget configuration in /etc/systemd/system/multi-user.target.wants/usbg.service.

Description: Our latest kernel is missing RS485 handling e.g. to use the RTS (CTS) output to control a RS485 transmitter.

Workaround: You may find the following commit from our older Linux kernel 3.10.17 as deployed in our embedded Linux BSP V2.4 Beta 1 (May 18, 2015) useful: http://git.toradex.com/cgit/linux-toradex.git/commit?h=toradex_imx_3.10.17_1.0.0_ga&id=82cf64948ddab0360fab876ee321b6191ed7258a

Description: Unfortunately, the older Linux kernel 3.14.52 as used in V2.6 does disable the SD card power rail VGEN3 upon boot which is reported as follows:
VGEN3: disabling
Subsequent SD card operation fails as follows:
mmcblk1: error -110 sending status command, aborting

Workaround: Please find an updated device tree as per the following commit:
http://git.toradex.com/cgit/linux-toradex.git/commit/?h=toradex_imx_3.14.52_1.1.0_ga&id=7659fe414d37861050ec909bb2d27c4bfca87d27

Description: Despite removing the conflicting UART2 Linux freezes on startup.

Workaround: None.

Description: The labels on special keys such as caps lock, enter or tab are missing.

Workaround: A fix is already available on the V2.6-next Meta-Toradex branch.

Description: Gstreamer may freeze when generating its initial database on some modules resulting in the following error message: (gst-plugin-scanner:683): GLib-GObject-WARNING **: cannot register existing type 'GstImxCompositorPad'

Workaround: The following gets you around this: mv /usr/lib/libgstimxcompositor.so.0.12.0 /usr/lib/libgstimxcompositor.so.0.12.0.orig gst-inspect-1.0 mv /usr/lib/libgstimxcompositor.so.0.12.0.orig /usr/lib/libgstimxcompositor.so.0.12.0 Alternatively one may uninstall libgstimxcompositor0 as follows: opkg remove --force-depends libgstimxcompositor0

Description: A low percentage of modules don't boot reliably because of the bootrom not handling the eMMC correctly. This leads to non booting modules. Using reset solves the problem, a power-cycle mostly works too.

Workaround: Using different i.MX6 fusing options and eMMC configurations seems to solve the problem. For the time being you may evaluate this as described in the following article on our developer website: http://developer.toradex.com/software-resources/arm-family/linux/linux-booting#eMMC_Fast_Boot_Mode_ApalisColibri_iMX6

Description: When rebooting the system using software reboot Linux can leave the PHY in an undefined state (e.g. down) in which case Ethernet may not work in U-Boot unless a hardware reset is performed.

Description: The PMIC RTC does not meet the expected accuracy.

Workaround: If possible use the external RTC provided by the carrier board.

Description: USB detection may fail especially on the OTG port plus the transfer speed may be rather low.

Workaround: Reverting http://git.toradex.com/cgit/linux-toradex.git/commit/?h=tegra&id=629b91fa046df5213ed3312d2e2b8683e6bdd70c helps but reintroduces issue #6422.

Description: Very rare occurrence of eMMC issues as shown below in combination with extensively heavy SATA and/or SD card stress testing: I/O error, dev mmcblk0

Workaround: Avoid extensively heavy SATA and/or SD card stress testing. Usually power-cycling brings the system back to life.

Description: Rarely after about 500 software reboot cycles the module locks up during boot.

Workaround: Just power-cycle again.

Description: The U-Boot boot loader does not switch on the power on the USB host port. This makes the USB host port not working by default on the Viola and Iris Carrier Boards.

Workaround: Use U-Boot's GPIO functionality to switch on the USB host power (GPIO 83).

Description: Rare occurrence of eMMC issues as shown below can lead to read-only root file system or even non-booting systems: I/O error, dev mmcblk0

Workaround: Usually power-cycling brings the system back to life. The tegra-next branch of our Linux kernel source repository contains a fix for this.

Description: Using UART on Vybrid without DMA can lead to boot freezes when characters have been received during startup.

Workaround: Enable DMA mode to avoid this problem completely or avoid character transmission to the module during startup.

Description: SPI by using Vybrid's DSPI peripheral does not work as intended.

Description: XRandR based rotation as described on the following article on our developer website is broken since our V2.1 releases: http://developer.toradex.com/knowledge-base/x-server-%28linux%29#Rotation

Workaround: Please revert to one of our older V2.0 based releases for now.

Description: Several hours of gstreamer streaming aborts due to memory leak in nvvidconv gstreamer plugin.

Workaround: Under certain circumstances using an older NVIDIA L4T R16.3 based BSP might be feasible until NVIDIA releases a fix.

Description: Analogue audio codec disappearing during reboot particularly if previously used for audio recording.

Workaround: Power-cycling should help. Alternatively wait for us to backport respective Linux kernel driver fixes.

Description: The PWM other than PWM do not work.

Description: Vybrid's internal RTC is not as accurate as expected (~2% drift).

Workaround: Update time regularly using NTP or use external RTC (/dev/rtc1).

Description: Currently if doing SD boot the U-Boot environment is not preserved on the SD card.

Workaround: Build latest U-Boot sources from git.toradex.com.

Description: Audio capture can stall causing stutter in captured stream.

Workaround: None. Await for update to driver.

Description: The main scheduling clock is non-monotonic increasing, which leads to time gaps in kernel log output.

Workaround: So far we haven't noticed any issues because of that. The real-time clock works as expected.