** Description changed:
- This is actually a duplicate to
- https://bugzilla.kernel.org/show_bug.cgi?id=203183 that is for internal
- tracking.
+ === SRU Justification ===
+ [Impact]
+ TSC marked unstable by skewed HPET after Intel SoC reached PC10 state.
+
+ [Fix]
+ Disable HPET on affected SoC generations, currently happens to CFL and
+ ICL.
+
+ [Test]
+ I can confirm disabling HPET can avoid this issue.
+
+ [Regression Potential]
+ Low. TSC on recent Intel SoC is now an invariant clocksource, so it's
+ harmless to disable HPET on these platforms.
+
+ === Original Bug Report ===
+ This is actually a duplicate to
https://bugzilla.kernel.org/show_bug.cgi?id=203183 that is for internal
tracking.
[Overview]
On Intel CoffeeLake it's observed tsc is always marked unstable after
turning off screen and let cpu enter idle state Package C10(PC10), and
then clock source is switched to hpet. Following messages are printed in
dmesg when this occurs:
- [ 115.454117] clocksource: timekeeping watchdog on CPU2: Marking
clocksource 'tsc' as unstable because the skew is too large:
- [ 115.454146] clocksource: 'hpet' wd_now: a44e6960 wd_last: a42fe23b mask:
ffffffff
- [ 115.454151] clocksource: 'tsc' cs_now: 69ee1f8d9f cs_last: 699580f6fd
mask: ffffffffffffffff
- [ 115.454160] tsc: Marking TSC unstable due to clocksource watchdog
- [ 115.454675] TSC found unstable after boot, most likely due to broken
BIOS. Use 'tsc=unstable'.
- [ 115.454679] sched_clock: Marking unstable (115712298090,
-257671127)<-(115460381128, -5706726)
- [ 115.455825] clocksource: Switched to clocksource hpet
+ [ 115.454117] clocksource: timekeeping watchdog on CPU2: Marking
clocksource 'tsc' as unstable because the skew is too large:
+ [ 115.454146] clocksource: 'hpet' wd_now: a44e6960 wd_last: a42fe23b mask:
ffffffff
+ [ 115.454151] clocksource: 'tsc' cs_now: 69ee1f8d9f cs_last: 699580f6fd
mask: ffffffffffffffff
+ [ 115.454160] tsc: Marking TSC unstable due to clocksource watchdog
+ [ 115.454675] TSC found unstable after boot, most likely due to broken
BIOS. Use 'tsc=unstable'.
+ [ 115.454679] sched_clock: Marking unstable (115712298090,
-257671127)<-(115460381128, -5706726)
+ [ 115.455825] clocksource: Switched to clocksource hpet
[Step to reproduce]
1. use `powertop` to observe if CPU has ever enter PC10,
2. issue `xset dpms force off` to turn off screen and idle for some minutes,
3. check dmesg for above tsc marked unstable messages
[Actual Results]
TSC marked unstable and clock source switched to HPET.
[Expected Results]
TSC remains the clock source across the PC10 event.
[Build Date & Hardware]
This is reproducible on kernel version 4.14 through 5.1-rc2. Order
kernel versions are not tested.
[Additional Information]
This was first found on Ubuntu (all versions) that system doesn't go
suspend after stayed idle for 20 minutes. Clock source was switched to
HPET, the system time is no correctly synced with reality so that the
system doesn't consider it has stayed idle long enough to begin suspend
process.
** Also affects: linux-oem (Ubuntu)
Importance: Undecided
Status: New
** Also affects: linux-oem-osp1 (Ubuntu)
Importance: Undecided
Status: New
** Also affects: linux (Ubuntu Eoan)
Importance: Undecided
Status: New
** Also affects: linux-oem (Ubuntu Eoan)
Importance: Undecided
Status: New
** Also affects: linux-oem-osp1 (Ubuntu Eoan)
Importance: Undecided
Status: New
** Also affects: linux (Ubuntu Bionic)
Importance: Undecided
Status: New
** Also affects: linux-oem (Ubuntu Bionic)
Importance: Undecided
Status: New
** Also affects: linux-oem-osp1 (Ubuntu Bionic)
Importance: Undecided
Status: New
** Also affects: linux (Ubuntu Disco)
Importance: Undecided
Status: New
** Also affects: linux-oem (Ubuntu Disco)
Importance: Undecided
Status: New
** Also affects: linux-oem-osp1 (Ubuntu Disco)
Importance: Undecided
Status: New
** Changed in: linux-oem-osp1 (Ubuntu Disco)
Status: New => Won't Fix
** Changed in: linux-oem-osp1 (Ubuntu Eoan)
Status: New => Won't Fix
** Changed in: linux-oem (Ubuntu Eoan)
Status: New => Won't Fix
** Changed in: linux-oem (Ubuntu Disco)
Status: New => Won't Fix
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1840239
Title:
tsc marked unstable after entered PC10 on Intel CoffeeLake
Status in HWE Next:
Triaged
Status in linux package in Ubuntu:
Triaged
Status in linux-oem package in Ubuntu:
New
Status in linux-oem-osp1 package in Ubuntu:
New
Status in linux source package in Bionic:
New
Status in linux-oem source package in Bionic:
New
Status in linux-oem-osp1 source package in Bionic:
New
Status in linux source package in Disco:
New
Status in linux-oem source package in Disco:
Won't Fix
Status in linux-oem-osp1 source package in Disco:
Won't Fix
Status in linux source package in Eoan:
New
Status in linux-oem source package in Eoan:
Won't Fix
Status in linux-oem-osp1 source package in Eoan:
Won't Fix
Bug description:
=== SRU Justification ===
[Impact]
TSC marked unstable by skewed HPET after Intel SoC reached PC10 state.
[Fix]
Disable HPET on affected SoC generations, currently happens to CFL and
ICL.
[Test]
I can confirm disabling HPET can avoid this issue.
[Regression Potential]
Low. TSC on recent Intel SoC is now an invariant clocksource, so it's
harmless to disable HPET on these platforms.
=== Original Bug Report ===
This is actually a duplicate to
https://bugzilla.kernel.org/show_bug.cgi?id=203183 that is for internal
tracking.
[Overview]
On Intel CoffeeLake it's observed tsc is always marked unstable after
turning off screen and let cpu enter idle state Package C10(PC10), and
then clock source is switched to hpet. Following messages are printed
in dmesg when this occurs:
[ 115.454117] clocksource: timekeeping watchdog on CPU2: Marking
clocksource 'tsc' as unstable because the skew is too large:
[ 115.454146] clocksource: 'hpet' wd_now: a44e6960 wd_last: a42fe23b mask:
ffffffff
[ 115.454151] clocksource: 'tsc' cs_now: 69ee1f8d9f cs_last: 699580f6fd
mask: ffffffffffffffff
[ 115.454160] tsc: Marking TSC unstable due to clocksource watchdog
[ 115.454675] TSC found unstable after boot, most likely due to broken
BIOS. Use 'tsc=unstable'.
[ 115.454679] sched_clock: Marking unstable (115712298090,
-257671127)<-(115460381128, -5706726)
[ 115.455825] clocksource: Switched to clocksource hpet
[Step to reproduce]
1. use `powertop` to observe if CPU has ever enter PC10,
2. issue `xset dpms force off` to turn off screen and idle for some minutes,
3. check dmesg for above tsc marked unstable messages
[Actual Results]
TSC marked unstable and clock source switched to HPET.
[Expected Results]
TSC remains the clock source across the PC10 event.
[Build Date & Hardware]
This is reproducible on kernel version 4.14 through 5.1-rc2. Order
kernel versions are not tested.
[Additional Information]
This was first found on Ubuntu (all versions) that system doesn't go
suspend after stayed idle for 20 minutes. Clock source was switched to
HPET, the system time is no correctly synced with reality so that the
system doesn't consider it has stayed idle long enough to begin
suspend process.
To manage notifications about this bug go to:
https://bugs.launchpad.net/hwe-next/+bug/1840239/+subscriptions
--
Mailing list: https://launchpad.net/~kernel-packages
Post to : kernel-packages@lists.launchpad.net
Unsubscribe : https://launchpad.net/~kernel-packages
More help : https://help.launchpad.net/ListHelp
