Skip to contentSkip to conversation

New: PTMControl Enabled blocks Low Power Idle in 5% of s2idle attempts - Dell XPS-13 9300 Ice Lake]

[email protected]: Bjorn Helgaas 1 participant 1 message Sep 23, 2020 PCI

lore
Share & tools
lore
#0 lore raw

----- Forwarded message from [email protected] -----

Date: Wed, 23 Sep 2020 15:22:47 +0000
From: [email protected]
To: [email protected]
Subject: [Bug 209361] New: PTMControl Enabled blocks Low Power Idle in 5% of
s2idle attempts - Dell XPS-13 9300 Ice Lake
Message-ID: <[email protected]/>

https://bugzilla.kernel.org/show_bug.cgi?id=209361

            Bug ID: 209361
           Summary: PTMControl Enabled blocks Low Power Idle in 5% of
                    s2idle attempts - Dell XPS-13 9300 Ice Lake
           Product: Drivers
           Version: 2.5
    Kernel Version: 5.9-rc5
          Hardware: All
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: PCI
          Assignee: [email protected]
          Reporter: [email protected]
        Regression: No

System: Dell XPS-13 9300, Intel(R) Core(TM) i7-1065G7

Up through Linux 5.9-rc5, this system will typically get very good residency in Low-Power-Idle upon s2idle. But on 5% of attempts, it gets ZERO residency in low-power-idle.

I have confirmed that this does not change if the time sleeping is extended, or the time between experiments is extended.

David Box suggested disabling PTM.
Doing so resulted -- for the first time --
the system entering LPI on s2idle 100% of s2idle attempts (770 attempts over 8 hours)

~/bin/pci_write8 0 0x1d 0 0x158 0
~/bin/pci_write8 0 0x1d 7 0x158 0

is what I used to disable PTM on my system
(write a 0, where pci_read8 previously showed the value 3)

This change is visible in the output of lspci as for both bridges:

< PTMControl: Enabled:+ RootSelected:+

Quoted 1 line

PTMControl: Enabled:- RootSelected:-

Signature · 4 lines
You are receiving this mail because:
You are watching the assignee of the bug.

----- End forwarded message -----