New: pci=assign-busses breaks power-off of T4215 laptop
A message that is not in the archive (sent off-list or not yet received): [email protected]/
Bjorn Helgaas
maintainer[+cc linux-pci, Andreas, switching to email for discussion]
Quoted 18 lines · On Sat, Sep 13, 2014 at 10:08 AM, <[email protected]> wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=84471
Bug ID: 84471 Summary: pci=assign-busses breaks power-off of T4215 laptop Product: Drivers Version: 2.5 Kernel Version: 3.17.0-rc4 Hardware: x86-64 OS: Linux Tree: Mainline Status: NEW Severity: normal Priority: P1 Component: PCI Assignee: [email protected] Reporter: [email protected] Regression: No
Quoted 5 lines
If the kernel argument "pci=assign-busses" is specified as a kernel boot parameter, power-off does no longer work with the T4215 laptop. The machine goes through the power-down cycle, but then sits idle with a blank screen after reporting "powering down". Without the kernel parameter, power-off works correctly.
Thank you very much for the report!
It sounds like this is a regression, i.e., power-off used to work correctly with "pci=assign-busses", but it doesn't any more. If that's the case, what's the most recent kernel you know of where it worked correctly?
Can you attach complete dmesg and "lspci -vv" output (as root) both with and without "pci=assign-busses" to the bugzilla? If you have an older kernel where power-off works correctly with "pci=assign-busses", please attach the same information for that kernel, too.
Why do you use "pci=assign-busses" in the first place? This is likely a different issue, but if "pci=assign-busses" fixes something, I consider that a bug in itself, because Linux should be smart enough to do the right thing even without "pci=assign-busses". So if something on your machine requires "pci=assign-busses" in order to work, please open a second bugzilla for that.
It's possible this is the same problem reported at
https://bugzilla.kernel.org/show_bug.cgi?id=83441. We think that is related to fc1b253141b3 ("PCI: Don't scan random busses in pci_scan_bridge()"), which appeared in v3.15. I attached a test patch to that bugzilla to revert that commit. If you think it looks like the same problem, you might try testing that patch.
Thanks again,
Bjorn