Re: [PATCH RFC] thunderbolt: Add JHL9580 Barlow Ridge (TB5) ICM support — host_reset timeout blocking PCIe tunneling

[Alan Borzeszkowski <alan.borzeszkowski@linux.intel.com>]

Hi,

Please send full dmesg with 'thunderbolt.dyndbg=+p' parameter in linux
command line.

Firmware Connection Manager (ICM) is not used with Barlow Ridge, only
software CM.

On Fri, 2026-07-03 at 11:49 +0800, Neo wrote:
> Hi maintainers,
> 
> 
> JHL9580 Barlow Ridge (8086:5780 bridge, 8086:5781 NHI), Intel's first
> Thunderbolt 5 / USB4 v2 host controller, has incomplete support in
> the
> kernel through v7.2-rc1. The host router reset times out and PCIe
> tunneling is non-functional, making eGPU use impossible.
> 
> 
> === Problem Summary ===
> 
> 
> 1. get_upstream_port() in icm.c does not recognize Barlow Ridge
> bridge
>    (0x5780). Neither does icm_probe() handle the NHI (0x5781).
> 
> 
> 2. With these fixed, icm_firmware_reset() → pcie2cio_write() times
> out
>    (-110) on Barlow Ridge. Neither icm_ar_cio_reset (0x50/BIT(9)) nor
>    icm_tr_cio_reset (0x777/BIT(1)) work.
> 
> 
> 3. Without ICM firmware running, REG_FW_STS (0x39944) is read-only
> and
>    writes are silently ignored. USB3 tunneling works (XHCI pass-
> through)
>    but PCIe tunnels are never established.
> 
> 
> 4. Even with software CM (tb_acpi_is_native()=1), USB4 ports show
>    "link: none" and the eGPU GPU never appears in lspci.
> 
> 
> === Hardware ===
> 
> 
>   Host: MSI MS-15M3 (E15M3IMS.117 BIOS, 2026-04-17)
>   CPU: Intel Arrow Lake-HX Ultra 9 275HX
>   TB5: Intel JHL9580 Barlow Ridge [8086:5780]/[8086:5781] (rev 84)
>   eGPU: GIGABYTE AORUS RTX5090 AI BOX (USB4)
>   OS: Ubuntu 26.04, kernel 7.0.0-27 (same issue on 7.0.0-22)
> 
> 
> === Patch Attempted (works up to firmware start) ===
> 
> 
> --- a/drivers/thunderbolt/icm.c
> +++ b/drivers/thunderbolt/icm.c
> @@ get_upstream_port():
> +     case 0x5780: /* JHL9580 Barlow Ridge Host Bridge */
> 
> 
> @@ icm_probe():
> +     case 0x5781: /* JHL9580 Barlow Ridge NHI */
> +           icm->max_boot_acl = ICM_AR_PREBOOT_ACL_ENTRIES;
> +           icm->can_upgrade_nvm = !x86_apple_machine;
> +           icm->is_supported = icm_ar_is_supported;
> +           icm->cio_reset = icm_ar_cio_reset;
> +           icm->get_mode = icm_ar_get_mode;
> +           icm->get_route = icm_ar_get_route;
> +           icm->save_devices = icm_fr_save_devices;
> +           icm->driver_ready = icm_ar_driver_ready;
> +           icm->device_connected = icm_fr_device_connected;
> +           icm->device_disconnected = icm_fr_device_disconnected;
> +           icm->xdomain_connected = icm_fr_xdomain_connected;
> +           icm->xdomain_disconnected = icm_fr_xdomain_disconnected;
> +           tb->cm_ops = &icm_ar_ops;
> +           break;
> 
> 
> Result: icm_probe succeeds, firmware_start begins, but
> icm_firmware_reset → pcie2cio_write times out.
> 
> 
> === dmesg ===
> 
> 
>   thunderbolt 0000:04:00.0: timeout resetting host router
>   (REG_RESET HRR remains 0x01 indefinitely; extended to 5000ms, no
> change)
> 
> 
> === Questions ===
> 
> 
> 1. What is the correct CIO reset sequence for Barlow Ridge?
> 2. Is the ICM firmware interface different from Alpine/Titan Ridge?
> 3. Are there plans to add native Barlow Ridge support to icm.c?
> 4. Should Barlow Ridge use software CM exclusively (skip ICM)?
> 
> 
> Full diagnostic logs, lspci, and ACPI DSDT excerpt available on
> request.
> Test hardware available for any proposed patches.
> 
> 
> Thank you,
> 	
> Neo
> liqys@vip.qq.com