2021-10-06 16:30:31

by Mukesh Ojha

[permalink] [raw]
Subject: [PATCH] PM / suspend: Abort suspend if somebody holds wakelock

There could be a scenario, where request_firmware() call results in
user land process trying to load firmwares into memory and
parallely one miscellaneous process is trying to invoke manual
suspend and due to which user mode helper gets disabled during
freezing of processes and that aborts loading of firmware even
though request_firmware() thread has taken wakelock.

Although, we are checking for any wakeup event inside
try_to_freeze_tasks() but that could be too late for the
above scenario.

Let's add a check before freezing/disable user land process in
suspend path.

Signed-off-by: Mukesh Ojha <[email protected]>
---
kernel/power/process.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/kernel/power/process.c b/kernel/power/process.c
index 37401c9..3e0d3d6 100644
--- a/kernel/power/process.c
+++ b/kernel/power/process.c
@@ -124,6 +124,9 @@ int freeze_processes(void)
{
int error;

+ if (pm_wakeup_pending())
+ return -EBUSY;
+
error = __usermodehelper_disable(UMH_FREEZING);
if (error)
return error;
--
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center,
Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project


2021-10-06 16:36:08

by Rafael J. Wysocki

[permalink] [raw]
Subject: Re: [PATCH] PM / suspend: Abort suspend if somebody holds wakelock

On Wed, Oct 6, 2021 at 6:27 PM Mukesh Ojha <[email protected]> wrote:
>
> There could be a scenario, where request_firmware() call results in
> user land process trying to load firmwares into memory and
> parallely one miscellaneous process is trying to invoke manual
> suspend and due to which user mode helper gets disabled during
> freezing of processes and that aborts loading of firmware even
> though request_firmware() thread has taken wakelock.
>
> Although, we are checking for any wakeup event inside
> try_to_freeze_tasks() but that could be too late for the
> above scenario.
>
> Let's add a check before freezing/disable user land process in
> suspend path.

If a laptop lid is closed and the system is expected to suspend, it
must suspend.

This takes precedence over the loading of firmware. Sorry.

> Signed-off-by: Mukesh Ojha <[email protected]>
> ---
> kernel/power/process.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/kernel/power/process.c b/kernel/power/process.c
> index 37401c9..3e0d3d6 100644
> --- a/kernel/power/process.c
> +++ b/kernel/power/process.c
> @@ -124,6 +124,9 @@ int freeze_processes(void)
> {
> int error;
>
> + if (pm_wakeup_pending())
> + return -EBUSY;
> +
> error = __usermodehelper_disable(UMH_FREEZING);
> if (error)
> return error;
> --
> Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center,
> Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
>