Hi, On Fri, Feb 16, 2024 at 10:33:00AM +0300, aigilea wrote: > Hello. > I've run into a problem with ACPI while trying to run any linux > distribution on "HP Spectre x360 14 inch 2-in-1 Laptop PC 14-eu0000" with > the latest BIOS "F.05 Rev.A" with kernel versions from 6.5 to 6.7. > During the boot or shortly after kernel spits out a crash backtrace > originating at "acpi_ps_get_arg" called from "acpi_ps_delete_parse_tree" > or at "acpi_ps_delete_parse_tree" itself. > It usually results in a panic and the system is halted but in some rare > cases it continues to work for some random period of time so I was able to > collect some debugging information. > I'm attaching parts of the log I've got booting Arch 2024.01 liveusb with > kernel 6.6.8 with the following command line: > acpi.debug_layer=0x30 acpi.debug_level=0x880640 log_buf_len=256M > dyndbg="file device_pm.c +p" > Layer is ACPI_PARSER | ACPI_NAMESPACE, level is ACPI_LV_PARSE_TREES | > ACPI_LV_EVALUATION | ACPI_LV_NAMES | ACPI_LV_EXEC | ACPI_LV_PARSE > I wasn't able to boot more recent kernels to a working state to collect > the log but they panic in the very same code. > Log part with crash starts at "ACPI: \_SB_.PC00.I2C3: PM: Power state > change: D0 -> D3cold" because I think this is what triggers the problem > and the crash happens after what looks like a double call of > "ps_delete_parse_tree" from "ps_pop_scope" with the same subtree_root arg > which should result in use-after-free and a double free. We have seen similar and I think this should be fixed in the BIOS side but I'm checking internally if this is the case or what was the conclusion. Temporary workaround is to blacklist intel_lpss_pci driver (although you may lose some of that functionality then but that's still better than completely non-functional system).
