[SRU][R][S][PATCH 0/1] System freezes on shutdown/reboot after BIOS update on HP ZBook Power 16 G11

[AceLan Kao]

From: "Chia-Lin Kao (AceLan)" <acelan.kao at canonical.com>

BugLink: https://bugs.launchpad.net/bugs/2132119

[Impact]
On HP ZBook Power 16 G11 the system freezes at shutdown or reboot after BIOS 01.07.01 or later. The system stays at the power-off screen and only a long power button press resets it. Plain shutdown also leaves the power LED on and makes the next boot require a long press.

The problem starts when the Realtek RTS525A SD card reader (10ec:525a) fires a Correctable Replay Timer Timeout storm. BIOS 01.07.01 turns on AER correctable error reporting, which the old BIOS had off. The storm floods the kernel AER handler with about 240K events in 11 seconds, for example:

pcieport 0000:00:1c.0: AER: Correctable error message received from 0000:58:00.0
rtsx_pci 0000:58:00.0: PCIe Bus Error: severity=Correctable, type=Data Link Layer, (Transmitter ID)
rtsx_pci 0000:58:00.0: device [10ec:525a] error status/mask=00001000/00006000
rtsx_pci 0000:58:00.0: [12] Timeout

The storm also happens when the rtsx_pci driver is blacklisted, and blocks the shutdown/reboot path. Affects 3 users.

[Fix]
Add a PCI fixup that removes the L0s bits from the RTS525A advertised ASPM capability (drivers/pci/quirks.c). The RTS525A link generates the Replay Timer Timeout storm when ASPM L0s is on. Testing showed that disabling only L0s stops the interrupt storm while keeping L1 and L1 substates working. Disabling L1, L1 substates, or Clock PM alone did not stop it.

Upstream: ec3d987fcaf9 ("PCI/ASPM: Avoid L0s for Realtek RTS525A"), merged in linux-next, queued for v7.3 (not yet in a released tag; based on v7.2-rc1).
Review/pending: https://lore.kernel.org/linux-pci/20260811093944.GA605422@bhelgaas/, first version at https://lore.kernel.org/linux-pci/20260528032347.644365-1-max.lee@canonical.com/

[Test Plan]
On an affected HP ZBook Power 16 G11 with BIOS 01.07.01 or later, boot the test kernel and run:

$ sudo dmesg | grep -i "AER\|rtsx_pci"

Without the patch: the AER correctable Replay Timer Timeout messages above repeat and the system hangs at shutdown/reboot.
With the patch: no new AER messages about 0000:58:00.0 / 10ec:525a, and shutdown and reboot complete.

Then:
$ sudo reboot
$ sudo poweroff

Without the patch: system freezes at power-off (needs long power button press).
With the patch: reboot and poweroff complete normally.

[Where problems could occur]
Could affect the PCI ASPM code and the Realtek rtsx_pci card reader (10ec:525a).

The fix only removes L0s from the RTS525A advertised capability. If the capability handling is wrong, the card reader link could lose ASPM or power management and stay at full power, raising battery drain. The change is a one-line DECLARE_PCI_FIXUP_HEADER quirk scoped to one device ID, so the risk to other devices is small. If the quirk misfires, the symptom would be a card reader that keeps the link in a higher-power state instead of a driver regression; the reader itself would still work.

Max Lee (1):
  PCI/ASPM: Avoid L0s for Realtek RTS525A

 drivers/pci/quirks.c | 3 +++
 1 file changed, 3 insertions(+)

-- 
2.53.0