linux-kernel - Re: [PATCH v2] serial: qcom_geni: Fix pinctrl deadlock on runtime resume

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aMvZ10EsMif/DOP4@trex>
Date: Thu, 18 Sep 2025 12:07:19 +0200
From: Jorge Ramirez <jorge.ramirez@....qualcomm.com>
To: Jorge Ramirez <jorge.ramirez@....qualcomm.com>
Cc: Praveen Talari <praveen.talari@....qualcomm.com>,
        Krzysztof Kozlowski <krzk@...nel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Jiri Slaby <jirislaby@...nel.org>,
        Bryan O'Donoghue <bryan.odonoghue@...aro.org>,
        Praveen Talari <quic_ptalari@...cinc.com>,
        linux-arm-msm@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-serial@...r.kernel.org, alexey.klimov@...aro.org,
        dmitry.baryshkov@....qualcomm.com, andersson@...nel.org,
        psodagud@...cinc.com, djaggi@...cinc.com, quic_msavaliy@...cinc.com,
        quic_vtanuku@...cinc.com, quic_arandive@...cinc.com,
        quic_shazhuss@...cinc.com, quic_cchiluve@...cinc.com
Subject: Re: [PATCH v2] serial: qcom_geni: Fix pinctrl deadlock on runtime
 resume

On 18/09/25 09:25:48, Jorge Ramirez wrote:
> On 18/09/25 09:25:53, Praveen Talari wrote:
> > Hi Krzysztof,
> > 
> > On 9/18/2025 5:28 AM, Krzysztof Kozlowski wrote:
> > > On 18/09/2025 03:51, Praveen Talari wrote:
> > > > A stall was observed in disable_irq() during
> > > > pinctrl_pm_select_default_state(), triggered by wakeup IRQ being active
> > > > while the UART port was not yet active. This led to a hang in
> > > > __synchronize_irq(), as shown in the following trace:
> > > > 
> > > > Call trace:
> > > >      __switch_to+0xe0/0x120
> > > >      __schedule+0x39c/0x978
> > > >      schedule+0x5c/0xf8
> > > >      __synchronize_irq+0x88/0xb4
> > > >      disable_irq+0x3c/0x4c
> > > >      msm_pinmux_set_mux+0x508/0x644
> > > >      pinmux_enable_setting+0x190/0x2dc
> > > >      pinctrl_commit_state+0x13c/0x208
> > > >      pinctrl_pm_select_default_state+0x4c/0xa4
> > > >      geni_se_resources_on+0xe8/0x154
> > > >      qcom_geni_serial_runtime_resume+0x4c/0x88
> > > >      pm_generic_runtime_resume+0x2c/0x44
> > > >      __genpd_runtime_resume+0x30/0x80
> > > >      genpd_runtime_resume+0x114/0x29c
> > > >      __rpm_callback+0x48/0x1d8
> > > >      rpm_callback+0x6c/0x78
> > > >      rpm_resume+0x530/0x750
> > > >      __pm_runtime_resume+0x50/0x94
> > > >      handle_threaded_wake_irq+0x30/0x94
> > > >      irq_thread_fn+0x2c/0xa8
> > > >      irq_thread+0x160/0x248
> > > >      kthread+0x110/0x114
> > > >      ret_from_fork+0x10/0x20
> > > > 
> > > > To fix this, wakeup IRQ setup is moved from probe to UART startup,
> > > > ensuring it is only configured when the port is active. Correspondingly,
> > > > the wakeup IRQ is cleared during shutdown. This avoids premature IRQ
> > > > disable during pinctrl setup and prevents the observed stall. The probe
> > > > and remove pathsare simplified by removing redundant wakeup IRQ handling.
> > > > 
> > > > Fixes: 1afa70632c39 ("serial: qcom-geni: Enable PM runtime for serial driver")
> > > > Reported-by: Alexey Klimov <alexey.klimov@...aro.org>
> > > > Closes: https://lore.kernel.org/all/DC0D53ZTNOBU.E8LSD5E5Z8TX@linaro.org/
> > > > Tested-by: Jorge Ramirez <jorge.ramirez@....qualcomm.com>
> > > 
> > > Where did you receive this tag for this patch exactly?
> > 
> > Since Jorge was involved in validating the change, I’ve added him under the
> > Tested-by tag.
> > 
> > Please correct me if I’m not supposed to add this tag myself.
> 
> let's test a bit further Praveen - we need to validate/trace the wake
> path on a real scenairo to make sure it is not cpu intensive (although I
> suspect the 2% was due to the storm you described more than to the code
> path itself)
> 
> I can then provide the tested-by on the list.
> 

um bluetooh comms are broken - reverting the runtime_pm patch fixes it.
and the proposed fix (V2) does not address this scenario.

I agree with the common sentiment, I think the patch should be reverted
in linux-next and better test definition shared.

[    1.451715] Bluetooth: Core ver 2.22
[    1.460668] Bluetooth: HCI device and connection manager initialized
[    1.467034] Bluetooth: HCI socket layer initialized
[    1.471922] Bluetooth: L2CAP socket layer initialized
[    1.476988] Bluetooth: SCO socket layer initialized
[    2.504958] Bluetooth: HCI UART driver ver 2.3
[    2.509427] Bluetooth: HCI UART protocol H4 registered
[    2.514600] Bluetooth: HCI UART protocol LL registered
[    2.519978] Bluetooth: HCI UART protocol Broadcom registered
[    2.525662] Bluetooth: HCI UART protocol QCA registered
[    2.530915] Bluetooth: HCI UART protocol Marvell registered
[    2.764571] Bluetooth: HIDP (Human Interface Emulation) ver 1.2
[    2.770503] Bluetooth: HIDP socket layer initialized
[    3.901958] Bluetooth: hci0: setting up wcn399x
[    6.202761] Bluetooth: hci0: command 0xfc00 tx timeout
[    6.212294] Bluetooth: hci0: Reading QCA version information failed (-110)
[    6.219261] Bluetooth: hci0: Retry BT power ON:0
[    8.538729] Bluetooth: hci0: command 0xfc00 tx timeout
[    8.543988] Bluetooth: hci0: Reading QCA version information failed (-110)
[    8.550989] Bluetooth: hci0: Retry BT power ON:1
[   10.810736] Bluetooth: hci0: command 0xfc00 tx timeout
[   10.816095] Bluetooth: hci0: Reading QCA version information failed (-110)
[   10.816110] Bluetooth: hci0: Retry BT power ON:2
[   13.082946] Bluetooth: hci0: command 0xfc00 tx timeout
[   13.088490] Bluetooth: hci0: Reading QCA version information failed (-110):