Re: Commit 4292a1e45fd4 ("PCI: Refactor distributing available memory to use loops") gives errors enumerating TBolt devices behind my TB dock

["Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>]

On Tue, 7 Oct 2025, Kenneth Crudup wrote:

> I'm running Linus' master (as of a8cdf51cda30).
> 
> I have a Thunderbolt dock (Amazon Basics generic thing, but it also happens on
> my CalDigit TB4). With the above commit (and aaae2863e731 ("PCI: Refactor
> remove_dev_resources() to use pbus_select_window()), so the Subject: commit
> would come out cleanly) if I plug in a TB device past my TB Dock, they don't
> fully enumerate (i.e., no DP tunneling, no partitions created, etc.)
> 
> I've attached the syslog from both bad runs, and good runs (i.e., after the
> reverts).
> 
> LMK if you need any further info,

Hi,

Thanks for the report.

Could you test if this patch helps:

https://lore.kernel.org/linux-pci/eb70b817-175c-7a34-d2bf-9472019afa47@linux.intel.com/

(I'm not sure if it's the same problem as these snippet around the problem 
spot seem to only show symptoms of an earlier problem which is not 
visible in those logs.)

If it doesn't help, I need more information:

- /proc/iomem

- dmesg that contains all PCI related messages, not just a snippet around 
the problem spot. Please enable dynamic debugging with
  dyndbg="file drivers/pci/*.c +p"
on the kernel command line so that the dmesg shows more information.

- lspci -vvv might be useful as well but often /proc/iomem is enough.

Preferrable from both good and bad cases so that it's easier to compare 
them efficiently.

-- 
 i.