Netdev List
 help / color / mirror / Atom feed
* Re: [PATCH] e1000e: add Dell Pro 5 Micro P5M1260 to disable K1 list
       [not found] <20261119001818.513503-1-acelan.kao@canonical.com>
@ 2026-11-20  1:54 ` Ye Shu
  0 siblings, 0 replies; only message in thread
From: Ye Shu @ 2026-11-20  1:54 UTC (permalink / raw)
  To: Chia-Lin Kao (AceLan), Tony Nguyen, Przemek Kitszel, Andrew Lunn,
	David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni
  Cc: Dima Ruinskiy, Vitaly Lifshits, intel-wired-lan, netdev,
	linux-kernel

Hi Chia-Lin and others,

On 11/18/26 4:18 PM, Chia-Lin Kao (AceLan) wrote:
> K1 was re-enabled on MTL and later platforms by
> commit 578294b8b60d ("e1000e: Reconfigure PLL clock gate timeout and re-enable K1 on Meteor Lake").
> Since then, the Dell Pro 5 Micro P5M1260 (Panther Lake, I219
> [8086:57b3]) no longer enters S0ix during s2idle. 

I have independently reproduced a similar S0ix issue on a different
machine, an HP ZBook 8 G2i 14 (also Panther Lake, with Intel I219 NIC
[8086:57b3] using the e1000e driver).

Machine tested:
- Model: HP ZBook 8 G2i 14
- DMI system vendor: HP
- DMI product name: HP ZBook 8 G2i 14 inch Mobile Workstation PC
- DMI base board product name: 8ED9
- NIC: Intel I219-LM [8086:57b3], subsystem Hewlett-Packard Company
  Device [103c:8ed9], driver e1000e
- Distribution: Fedora 45
- Kernel: 7.3.4-300.fc45.x86_64
- BIOS: Y70 Ver. 01.01.07

With K1 enabled by default (disable-k1: off), S0ix entry is
intermittent. Some suspend go into S0ix while others do not, even with
CPU reaching Package C10.

> Every suspend ends with:
> 
>    intel_pmc_core INT33A1:00: CPU did not enter SLP_S0!!! (S0ix cnt=0)

Failed suspend produces the same message:
   intel_pmc_core INT33A1:00: CPU did not enter SLP_S0!!!

> The PMC low power mode status dump shows the GbE IP is never power
> gated (GBE_PGD0_PG_STS, GBE_D3_STS and GBE_VNN_REQ_STS all 0), while
> the CPU itself reaches C10.

The PMC dump shows the following:
   CPU_C10_REQ_STS     1
   GBE_PGD0_PG_STS     0
   GBE_D3_STS          1
   GBE_VNN_REQ_STS     0

> Disabling K1 at runtime restores S0ix on the same unit, kernel and
> BIOS:
> 
>    # ethtool --set-priv-flags enp0s31f6 disable-k1 on

Disabling K1 using the same command appears to resolve the problem. I
performed the following A/B/A test within a single boot, using 120
second rtcwake -m mem cycles and measuring slp_s0_residency_usec before
and after each cycle:

disable-k1 on:   6/6 successful (~116-119 s in S0ix)
disable-k1 off:  1/4 successful (118, 0, 0, 0 s in S0ix)
disable-k1 on:   4/4 successful (~115-119 s in S0ix)

Since this issue also reproduces on a different machine with the same
PCI device ID for NIC, would you prefer adding another HP-specific DMI
entry to disable_k1_list, or do you think that disabling K1 for all NICs
with PCI ID [8086:57b3] would be more appropriate?

I'm happy to submit a separate disable_k1_list patch or test any
proposed changes.

Thank you!
Ye

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2026-11-20  1:54 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
     [not found] <20261119001818.513503-1-acelan.kao@canonical.com>
2026-11-20  1:54 ` [PATCH] e1000e: add Dell Pro 5 Micro P5M1260 to disable K1 list Ye Shu

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox