[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <db65c9f0-334f-43fb-bddf-316bf883a848@oss.qualcomm.com>
Date: Wed, 18 Jun 2025 11:18:14 +0200
From: Konrad Dybcio <konrad.dybcio@....qualcomm.com>
To: Johan Hovold <johan@...nel.org>
Cc: Baochen Qiang <quic_bqiang@...cinc.com>,
Johannes Berg <johannes@...solutions.net>,
Jeff Johnson
<jjohnson@...nel.org>, linux-wireless@...r.kernel.org,
ath11k@...ts.infradead.org, linux-kernel@...r.kernel.org,
Takashi Iwai <tiwai@...e.de>,
Bjorn Andersson <bjorn.andersson@....qualcomm.com>,
Johan Hovold <johan+linaro@...nel.org>
Subject: Re: [PATCH ath-next v3 5/6] wifi: ath11k: choose default PM policy
for hibernation
On 6/18/25 10:52 AM, Johan Hovold wrote:
> On Wed, Jun 18, 2025 at 02:15:04AM +0200, Konrad Dybcio wrote:
>> On 3/28/25 6:32 AM, Baochen Qiang wrote:
>>> Now WoWLAN mode is chosen for those machines listed in the quirk table.
>>> This works for suspend (S3) but breaks for hibernation (S4), because
>>> WoWLAN mode requires WLAN power to be sustained, which is not the case
>>> during hibernation. For hibernation, the default mode should be used.
>>>
>>> Register a PM notifier with which kernel can notify us of the actual PM
>>> operation: if system is going to suspend, the original PM policy is
>>> honored; while if it is hibernation, overwrite it with default policy.
>>>
>>> To summarize: for suspend (S3), WoWLAN mode is chosen for machines listed
>>> in the quirk table, non-WoWLAN mode for others; for hibernation (S4),
>>> non-WoWLAN mode is chosen for all.
>>>
>>> Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.30
>>>
>>> Tested-by: Takashi Iwai <tiwai@...e.de>
>>> Signed-off-by: Baochen Qiang <quic_bqiang@...cinc.com>
>>> ---
>>
>> I've bisected the following splat to this patch, still happening on
>> linux-next/master, WCN6855 + SC8280XP CRD, 100% reproducibility
>
> WFIW, I'm not seeing this with 6.16-rc2 (which has this patch) on either
> the X13s or sc8280xp-crd (ath11k now fails to resume on the latter
> because of missing regulatory data, but that appears to be a separate
> regression).
Do you have CONFIG_DEBUG_NOTIFIERS?
>> [root@...280xp-crd ~]# echo mem > /sys/power/state
>> [ 20.267830] fb0: Framebuffer is not in virtual address space.
>> [ 39.863070] PM: suspend entry (s2idle)
>> [ 39.908067] Filesystems sync: 0.035 seconds
>> [ 39.934453] ------------[ cut here ]------------
>> [ 39.939259] Invalid notifier called!
>> [ 39.939268] WARNING: CPU: 5 PID: 513 at kernel/notifier.c:79 notifier_call_chain+0x84/0x1a4
>> [ 39.951566] Modules linked in:
>> [ 39.954732] CPU: 5 UID: 0 PID: 513 Comm: bash Not tainted 6.14.0-rc4longbois-01215-g32d93b51bc7e #12177
>
> 6.14?
g show 32d93b51bc7e2e557771abe4a88da69c609e3d52:Makefile | head
# SPDX-License-Identifier: GPL-2.0
VERSION = 6
PATCHLEVEL = 14
SUBLEVEL = 0
EXTRAVERSION = -rc4
it's been in the tree for a while
Konrad
Powered by blists - more mailing lists