yenta_socket dies on rewriting configuration registers

Running 2.6.7 on a Thinkpad 240X, if I allow the PCI registers on my cardbus bridge to be rewritten after ACPI resume, the system dies with slightly strange symptoms. Every other character of kernel messages is written, with blank spaces in between. After a few lines of message, the system stops. If I do a resume with an unpatched kernel, it appears to get through to thawing processes before hanging - if I put debug statements in the resume path, it hangs within device_resume. I get this behaviour even if the yenta_socket driver is not loaded due to the generic 2.6.7 resume code. Commenting out both calls to pci_restore_state (from the driver itself and from the generic PCI code) results in suspend working correctly. The bridge is a Texas Instruments PCI1211.

The driver works correctly with my Thinkpad X40, which has a Ricoh RL5c476 II

Matthew Garrett | [email protected]