Add Wildcat Lake to supported CPU model table

[karanshukla]

Wildcat Lake (family 6, model 0xd5) is missing from intel_id_table[], so check_cpu_id() fails to match and cthd_engine_adaptive::thd_engine_init() exits with "Unsupported cpu model or platform". Because that path calls exit(EXIT_SUCCESS), systemd records a clean exit and Restart=on-failure never fires, so the unit stays enabled while doing nothing.

Marked adaptive_only, consistent with the other recent client platforms (Lunar Lake, Arrow Lake, Panther Lake, Nova Lake). That third field is what makes this safe to add without any silicon-specific values: thermald carries no thermal constants for these platforms and reads GDDV out of firmware at runtime, so the entry asserts only that this platform is driven by its own firmware tables.

Platform

• Dell XPS 13 DX13260, BIOS 1.6.0
• Intel Core 5 320, family 6 model 213 (0xD5) stepping 1
• Kernel 7.1.6-201.fc44.x86_64, Fedora 44
• INTC10FC:00 INT3400 with a populated 1208-byte data_vault, \_SB_.IETM
• RAPL powercap MSR + MMIO both present

Affects both Fedora's 2.5.9 package and master at 2d93d94 (2.5.12-rc1).

Testing

Built from master 2d93d94, installed to /usr/local/bin and run via a systemd drop-in.

Before: thermald.service exits immediately at every boot.

thermald[955]: 40 CPUID levels; family:model:stepping 0x6:d5:1 (6:213:1)
thermald[955]:  Need Linux PowerCap sysfs
thermald[955]: Unsupported cpu model or platform

After: the daemon stays running, parses this machine's GDDV, selects target Balance Mode-28C, and drives rapl_controller_mmio between the firmware's own PL1MIN 15000 / PL1MAX 25000. Measured on intel-rapl-mmio:0/constraint_0_power_limit_uw: 15000000 before, 25000000 after.

Nothing else in the adaptive path needed changes. thermald discovers INT3400 by scanning for a driver directory entry starting with INT, so the newer INTC10FC ACPI ID is picked up as-is. available_uuids reads UNKNOWN and current_uuid reads INVALID, which matches other recent client platforms where policy comes from GDDV rather than UUID selection.

One limit worth stating: the passive throttle-down branch was not exercised. Under six cores of openssl speed -evp aes-256-cbc for four minutes, SEN1 (the sensor GDDV binds the 54C passive trip to) peaked at 50C while the package reached 76C, so the trip was never crossed. That is a property of which sensor the platform's firmware selects, not of this change. Verified up to the trip, not through it.

Note, not part of this patch

thd_platform_intel.cpp:157 prints " Need Linux PowerCap sysfs" unconditionally on any table miss, without checking whether powercap is present. On this machine powercap is fully populated, so the message points at the wrong thing and is easy to chase for a while. Happy to send that separately if it is worth fixing.

Unrelated

No overlap with #597, which touches the Qualcomm SPEL path in thd_engine_default.cpp.

[spandruvada]

Already a commit in master for this.
dd0eb7e

[karanshukla]

Confirmed, thanks. dd0eb7e adds the identical entry in the same position, my branch was based on 2d93d94 and predates it. Should have rebased before sending.

Closing. I'll send the misleading powercap log message noted at the bottom of the description as a separate one-liner.