NAK [D/E]: [SRU] [D/E/Unstable/OEM-B] [PATCH 0/7] Make hotplugging docking station to Thunderbolt port more reliable

[Kai-Heng Feng]

> On Nov 26, 2019, at 17:18, Thadeu Lima de Souza Cascardo <cascardo at canonical.com> wrote:
> 
> On Tue, Nov 26, 2019 at 04:51:42PM +0800, Kai-Heng Feng wrote:
>> BugLink: https://bugs.launchpad.net/bugs/1853991
>> 
>> [Impact]
>> Sometimes USB ports and USB ethernet on Thunderbolt Docking don't work
>> after hotplugging.
>> 
>> [Fix]
>> Add proper link delay when PCIe devices transit out from D3cold.
>> 
>> [Test]
>> Tested with Dell WD19DC docking station on Dell XPS 9380. After applying
>> the fix, the USB ports on docking always work after hotplugging.
>> 
>> [Regression Potential]
>> Low. This patch series adds proper delay required by PCIe spec, and only
>> affects hotplug ports, i.e. Thunderbolt.
> 
> That doesn't seem like a correct statement. The last commit mentions a previous
> attempt was reverted because of regressions. And it's backported from
> linux-next.

The patch author asked the user once had the regression to test it, and the result is no regression found.

> By reading the code, it seems it could at least cause some delays during
> resume, but I would like to see this better tested to show that's likely the
> most that could happen.

For systems use s2idle the delay won't be introduced because of direct-complete optimization.
For systems that use S3 there can be a delay, but at least I don't feel the delay on my tests.

> 
> Could you test other systems and look for problems during suspend-resume tests?

I don't find any problems on the three systems I've tested but I do understand changing PCI core implies higher risk.
As this issue isn't that severe (the docking will work if it's plugged again), I think we can skip D/E and focus on Focal.

Kai-Heng

> 
> Thanks.
> Cascardo.
> 
>> 
>> Felipe Balbi (1):
>>  PCI: Add support for Immediate Readiness
>> 
>> Keith Busch (1):
>>  PCI: Make link active reporting detection generic
>> 
>> Mika Westerberg (3):
>>  PCI: Make pcie_downstream_port() available outside of access.c
>>  PCI/PM: Add pcie_wait_for_link_delay()
>>  PCI/PM: Add missing link delays required by the PCIe spec
>> 
>> Oza Pawandeep (1):
>>  PCI/DPC: Clear interrupt status in interrupt handler top half
>> 
>> Rafael J. Wysocki (1):
>>  PCI: PM: Avoid skipping bus-level PM on platforms without ACPI
>> 
>> drivers/pci/access.c             |   9 --
>> drivers/pci/hotplug/pciehp_hpc.c |  22 +---
>> drivers/pci/pci-driver.c         |  17 ++-
>> drivers/pci/pci.c                | 180 +++++++++++++++++++++++++++++--
>> drivers/pci/pci.h                |  10 ++
>> drivers/pci/pcie/dpc.c           |   7 +-
>> drivers/pci/probe.c              |   1 +
>> include/linux/pci.h              |   2 +
>> include/linux/suspend.h          |  26 ++++-
>> include/uapi/linux/pci_regs.h    |   1 +
>> kernel/power/suspend.c           |   3 +
>> 11 files changed, 233 insertions(+), 45 deletions(-)
>> 
>> -- 
>> 2.17.1
>> 
>> 
>> -- 
>> kernel-team mailing list
>> kernel-team at lists.ubuntu.com
>> https://lists.ubuntu.com/mailman/listinfo/kernel-team