Re: [BUG] rtw88_8723de: Wi-Fi and Bluetooth fail after warm reboot

[Snake Pilot]

Hello Lovekesh,

I repeated the test using the updated patch on Debian kernel
7.1.3+deb14-amd64.

I started from a cold boot with both Wi-Fi and Bluetooth working, temporarily
neutralized the existing shutdown workaround for this test, and performed a
normal warm reboot to the same patched kernel.

The warm reboot reproduced both failures.

The previous-boot kernel journal shows that the Bluetooth shutdown path
reached the Realtek callback and completed successfully:

Bluetooth: hci0: diag: hci_dev_shtudown: HCI_UP=1 HCI_UNREGISTER=0
Bluetooth: hci0: diag: calling vendor shutdown callback
Bluetooth: hci0: diag: btrtl_shutdown_realtek: sending HCI_OP_RESET
Bluetooth: hci0: diag: btrtl_shutdown_realtek: HCI_OP_RESET completed OK
Bluetooth: hci0: diag: vendor shutdown returned 0

After the reboot, the Bluetooth USB function failed to enumerate again with
repeated error -71 messages.

The Wi-Fi function also failed with the following diagnostics:

rtw88_8723de 0000:02:00.0: diag: pre-toggle
REG_SYS_PW_CTRL=0x12, poll offset=0x5 want=0x0

rtw88_8723de 0000:02:00.0: diag: post-toggle
REG_SYS_PW_CTRL=0x1a, retry poll read=0x01

rtw88_8723de 0000:02:00.0: mac power on failed
rtw88_8723de 0000:02:00.0: probe with driver rtw88_8723de failed
with error -16

I did not find the btusb shutdown or rtw_pci_shutdown/D3hot diagnostic
messages in the persisted previous-boot journal. The journal appears to end
before the final device shutdown stage, so their absence from the journal
does not necessarily mean that the callbacks were not invoked.

The attached archive contains the requested files:

journal-kernel-previous.txt
journal-kernel-current.txt
acpidump.txt

There were two include directives in the updated patch that prevented it from
compiling against the Debian 7.1.3-1 source:

#include "net/bluetooth/hci.h"
in net/bluetooth/hci_sync.c

#include "hci.h"
before main.h in drivers/net/wireless/realtek/rtw88/mac.c

I removed only those two include lines. The diagnostic logging and the
remaining patch contents were left unchanged.

After collecting the logs, I restored the original packaged modules and
performed a complete power-off. Both Wi-Fi and Bluetooth recovered normally,
and the existing shutdown workaround is enabled again.

Regards,
Snake

El sáb, 1 ago 2026 a la(s) 10:10 a.m., Lovekesh Solanki
(lovekeshsolanki00@xxxxxxxxx) escribió:
>
> Thanks for testing,
>
> On Sat, Aug 01, 2026 at 02:00:12AM -0600, Snake Pilot wrote:
> > This explains the missing btusb shutdown log, although it does not by itself
> > explain or fix the retained hardware state. I can test a corrected diagnostic
> > or a specific USB reset/shutdown sequence if useful.
> I didn't notice this, I'll fix it in updated patch.
>
> I'm adding some logging in pci.c around rtw_pci_shutdown(),
> maybe the chip's power or reset state is just stuck before rtw88*
> reaches it and USB for some reason leaves it that way. Let's see if the
> PCIe function even reaches D3hot transition.
>
> Could you redo a warm reboot with only the updated patch and attach the
> previous and current logging same as before?
> Could you also attach your acpidump? I'd like to see if D3cold state is
> possible which may be a possible fix.
>
> Regards,
> Lovekesh

Attachment: rtl8723de-diag2-logs-acpidump.tar.gz
Description: application/gzip