Skip to contentSkip to conversation

btintel_pcie: Intel BE211 8086:a876 requests unavailable ibt-0190-01a1 firmware

BUG Sergey Lebedev 1 participant 1 message Sep 8 Bluetooth drivers

lore
Share & tools
lore
#0 lore raw

Mark,

The driver is asking for the right file. That file has never existed in linux-firmware, so no update and no other distribution will produce it.

The name is built in btintel.c from two hardware fields, and the source says which:

/* ibt-<cnvi_top type+cnvi_top step>-<cnvr_top type+cnvr_top step-fw_id> */ format = "intel/ibt-%04x-%04x-iml.%s";
snprintf(fw_name, len, format, cnvi, cnvr, suffix);

Both paths that can produce an "-iml" name build it the same way. First field is cnvi_top, the connectivity part in the SoC; second is cnvr_top, the radio. So yours is cnvi 0190 with cnvr 01a1, and the request matches your silicon.

I have the same SoC side. Surface Pro 11 (Intel), Lunar Lake, and the 8086:a876 rev 10 you report - but subsystem 8086:000e where yours is 8086:0011. It works here, and loads:

Found device firmware: intel/ibt-0190-0291-iml.sfi Found device firmware: intel/ibt-0190-0291-pci.sfi Found Intel DDC parameters: intel/ibt-0190-0291-pci.ddc

Same 0190, and 0291 against your 01a1. Two machines on the same connectivity part carrying different radios.

I checked linux-firmware.git itself rather than any distribution package, since package version strings do not compare across distributions - a frozen LTS string can carry backported content and say nothing about it. The tree listing and WHENCE agree on what is published:

ibt-0190-* with 0041 and 0291
*-01a1-* with 00a0 and 1190

And a path history query - run with two known-present files as controls, because an empty result from a broken query looks identical to an empty result from a true absence - returns 11 commits for ibt-0190-0291-pci.sfi, 6 for ibt-1190-01a1-pci.sfi, and none at all for ibt-0190-01a1-pci.sfi. The file has never been in the tree, not merely dropped from it.

If it helps in asking Intel, their commits name each pair as a core:

  0190-0291   BlazarI          (this machine)
  0190-0041   BlazarIGfP
  1190-01a1   BlazarIW
  00a0-01a1   Scorpius

There is no commit for 0190-01a1 under any name.

One thing not to do: rename a neighbouring blob. ibt-1190-01a1 is your radio against a different connectivity part and ibt-0190-0291 the reverse, and the driver will not stop you - it validates the CSS header version and the SBE type against hw_variant, but nothing checks that the image belongs to your radio, so a renamed file is handed to the controller rather than rejected.

So this is a firmware publication request and not a kernel bug, and it needs Intel. I have put Kiran and Chandrashekar on Cc, who work on btintel_pcie and are better placed than I am to say whether 0190-01a1 is coming, or whether the module in your machine is not yet enabled on Linux.

Your report: [email protected]

Sergey