Received-SPF: pass (google.com: domain of linux-wireless+bounces-8188-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240527173454.459264-1-marcin.slusarz@gmail.com> <0f1aef25-2b91-408d-aca9-9943dc9367b3@gmail.com>
In-Reply-To: <0f1aef25-2b91-408d-aca9-9943dc9367b3@gmail.com>
From: =?UTF-8?Q?Marcin_=C5=9Alusarz?= <marcin.slusarz@gmail.com>
Date: Tue, 28 May 2024 12:42:46 +0200
Message-ID: <CA+GA0_vz7S++ttaug2kGZWyzyc-PMZoVSsiSH-sXn4uctX5V2A@mail.gmail.com>
Subject: Re: wifi: rtw88: 8821CU hangs after some number of power-off/on cycles
To: Bitterblue Smith <rtl8821cerfe2@gmail.com>
Cc: linux-wireless@vger.kernel.org, =?UTF-8?Q?Marcin_=C5=9Alusarz?= <mslusarz@renau.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

pon., 27 maj 2024 o 20:43 Bitterblue Smith <rtl8821cerfe2@gmail.com> napisa=
=C5=82(a):
>
> On 27/05/2024 20:34, Marcin =C5=9Alusarz wrote:
> > From: Marcin =C5=9Alusarz <mslusarz@renau.com>
> >
> > If I don't connect to any Wifi network, after around 10 minutes, the de=
vice
> > hangs with endless spamming of:
> > rtw_8821cu 1-1:1.2: rtw_usb_reg_sec: reg 0x4e0, usb write 1 fail, statu=
s: -71
> > killing both Wifi and Bluetooth part of the device.
> >
> > On arm, just leaving the wifi device unconnected kills it in up to 20 m=
inutes.
> > If I keep restarting wpa_supplicant I can trigger it within a minute.
> > Looping "ifconfig wlan0 down; ifconfig wlan0 up" also triggers it withi=
n a minute.
> >
> > On x86_64 system the only way I could trigger this was via ifconfig loo=
p,
> > but it took 3 hours and 20 minutes to do it.
> >
> > The only thing that can "fix" the device is replugging it.
> >
> > I found out that the reason for those hangs is a power-off+on sequence =
that's
> > triggered by the above steps.
> >
> > Disabling power-off for that chip "fixes" the issue. The patches below
> > implement that, but I'm not seriously proposing them for merging.
> >
> > Marcin =C5=9Alusarz (2):
> >   wifi: rtw88: use RTW_FLAG_RUNNING for deciding whether to enter/leave
> >     IPS
> >   wifi: rtw88: disable power offs for 8821C
> >
> >  drivers/net/wireless/realtek/rtw88/main.c | 14 ++++++++------
> >  drivers/net/wireless/realtek/rtw88/ps.c   |  4 ++--
> >  2 files changed, 10 insertions(+), 8 deletions(-)
> >
>
> The first patch alone doesn't fix it?

The first patch exists only to make the second patch work.
Without the first one, rtw_enter_ips will perform all actions except actual=
ly
doing the power off, which clears the POWERON flag.
After that, rtw_leave_ips will happily return early success without actuall=
y
undoing the stuff that rtw_enter_ips did (like canceling all work_structs).