ath10k_pci: QCA9377 hard lock on resume from S3/deep suspend

Qualcomm Atheros QCA9377 using ath10k_pci can cause a complete system
hard lock when resuming from deep/S3 suspend. During the failed
resume, the kernel log shows repeated PCIe AER Physical Layer and Data
Link Layer errors involving the QCA9377.

Unloading ath10k_pci before suspend prevents the problem.

System

Fedora Linux 44

Kernel:
7.1.10-200.fc44.x86_64

Architecture:
x86_64

Kernel taint:
0

BIOS:
1.32.0
09/03/2024

Suspend modes:
s2idle [deep]

Wi-Fi device:
Qualcomm Atheros QCA9377 802.11ac Wireless Network Adapter [168c:0042] (rev 31)

Subsystem:
Dell Device [1028:1810]

Driver:
ath10k_pci

Problem

When the system resumes from deep/S3 suspend with ath10k_pci loaded,
the entire system can hard-freeze.

When this happens:

The system becomes completely unresponsive.
Caps Lock does not toggle.
Magic SysRq does not respond.
A forced power-off is required.

The kernel log around resume contains repeated PCIe AER errors such as:

ath10k_pci 0000:02:00.0: PCIe Bus Error: severity=Correctable,
type=Physical Layer, (Receiver ID)
ath10k_pci 0000:02:00.0: device [168c:0042] error status/mask=000000c1/00006000
ath10k_pci 0000:02:00.0: [ 0] RxErr (First)

pcieport 0000:00:1c.7: AER: Multiple Correctable error message
received from 0000:02:00.0

ath10k_pci 0000:02:00.0: PCIe Bus Error: severity=Correctable,
type=Data Link Layer, (Receiver ID)

There are also ACPI errors during resume:

ACPI BIOS Error (bug): Could not resolve symbol [_SB.PCI0.GLAN.GPEH],
AE_NOT_FOUND
ACPI Error: Aborting method _GPE._L6D due to previous error (AE_NOT_FOUND)
ACPI Error: AE_NOT_FOUND, while evaluating GPE method [_L6D]

Steps to reproduce

Boot normally with ath10k_pci loaded.
Confirm the QCA9377 is using ath10k_pci.
Suspend using deep/S3 suspend.
Wake the system.
The system may completely hard-lock during resume.

Isolation test / workaround

Before suspend:

sudo modprobe -r ath10k_pci
systemctl suspend

With ath10k_pci unloaded, the system resumes normally without the hard lock.

After resume:

sudo modprobe ath10k_pci

The driver reloads and Wi-Fi reconnects normally.

Simply unloading and reloading ath10k_pci without suspending also
works normally.

PCIe information

LnkCap: Speed 2.5GT/s, Width x1, ASPM L0s L1
LnkCtl: ASPM L0s L1 Enabled
LnkSta: Speed 2.5GT/s, Width x1

L1 PM substates are enabled:

PCI-PM_L1.2+
PCI-PM_L1.1+
ASPM_L1.2+
ASPM_L1.1+

This appears to involve the QCA9377/ath10k PCIe resume or PCIe
power-management path, although I cannot determine whether the root
cause is ath10k, PCIe power management, device firmware, or system
ACPI firmware.

I also filed this with Fedora as Bugzilla #2523243.

I can provide the complete previous-boot kernel log and perform
additional tests if needed.