search

OE4T / meta-tegra

BSP layer for NVIDIA Jetson platforms, based on L4T
MIT License
397 stars 220 forks source link

UEFI crashes for a headless boot (no display connected) #1417

Open picanteAvi opened 10 months ago

picanteAvi commented 10 months ago

If there is no HDMI cable connected, UEFI will crash in T194GraphicsOutputDxe driver init. Full log is attached but the relevant pieces are here:

NV-DISP snippet

[0003.840] I> regulator 'vdd-hdmi-5v0' already enabled
[0003.847] I> regulator 'vdd-hdmi-5v0' already enabled
[0003.847] E> tegrabl_display_init_regulator: hdmi cable is not connected
[0003.848] E> tegrabl_display_get_pdata, failed to parse dtb settings
[0003.849] E> cannot find any other nvdisp nodes
[0003.849] E> no valid display unit config found in dtb
[0003.856] W> display init failed
[0003.857] initializing target
[0003.859] calling apps_init()
[0003.862] starting app kernel_boot_app
[0003.866] I> Kernel type = Normal

UEFI snippet:

ERROR:   MPIDR 0x80000000: exception reason=0 syndrome=0xbe000000
ERROR:   **************************************
ERROR:   RAS Error in L2, ERRSELR_EL1=0x200:
ERROR:      Status = 0xfc006612
ERROR:      IERR = SCF to L2 Slave Error Read: 0x66
ERROR:   SERR = Error response from slave: 0x12
ERROR:      Overflow (there may be more errors) - Uncorrectable
ERROR:      Uncorrectable (this is fatal)
ERROR:      MISC0 = 0x80000000400000
ERROR:      MISC1 = 0x20240000000
ERROR:      ADDR = 0x800000001520101c
ERROR:   **************************************
ERROR:   **************************************

ERROR:   RAS Error in SCF_IOB, ERRSELR_EL1=0x401:
ERROR:      Status = 0xf4009604
ERROR:      IERR = CBB Interface Error: 0x96
ERROR:   SERR = Assertion failure: 0x4
ERROR:      Uncorrectable (this is fatal)
ERROR:      MISC0 = 0x40
ERROR:      MISC1 = 0x264a444461
ERROR:      ADDR = 0x800000001520101c
ERROR:   **************************************
ERROR:   RAS error handled!
ERROR:   sdei_dispatch_event returned -1
Head index=0 DC_DISP_SYNC_WIDTH  = 0x00000000

It seems like UEFI is trying to access some display control registers that are not inited after nvdisp hands over control? My workaround was to add a patch to the nvdisp binary. Basically, it will return a no-error when displays are not found, let the initialization complete, then fail when trying to access the controller. This gives the following log output for NV-DISP

[0003.763] I> regulator 'vdd-hdmi-5v0' already enabled
[0003.769] I> regulator 'vdd-hdmi-5v0' already enabled
[0003.770] E> tegrabl_display_init_regulator: hdmi cable is not connected
[0003.770] I> retrieved tmds range from prod_list_hdmi_soc
[0003.771] E> cannot find any other nvdisp nodes
[0003.777] E> I2C: slave not found in slaves.
[0003.778] E> I2C: Could not write 0 bytes to slave: 0x00a0 with repeat start true.
[0003.786] E> I2C_DEV: Failed to send register address 0x00000000.
[0003.791] E> I2C_DEV: Could not read 128 registers of size 1 from slave 0xa0 at 0x00000000 via instance 3.
[0003.801] E> could not read edid
[0003.809] I> hdmi_enable, starting HDMI initialisation
[0003.813] I> hdmi_enable, HDMI initialisation complete
[0003.822] initializing target
[0003.823] calling apps_init()
[0003.824] starting app kernel_boot_app
[0003.825] I> Kernel type = Normal

This seems hacky and there is probably a better way to fix it, whatever that might be.

0006-allow-headless-boot.patch uefi-crash-when-display-not-inited.log

dwalkes commented 10 months ago

Can you reproduce with https://github.com/OE4T/tegra-demo-distro ? If so, which MACHINE, branch, and image?

ichergui commented 3 months ago

Hi @picanteAvi Any updates on this issue ? Otherwise I will close it.