[PATCH] PM / devfreq: Return -ENODEV from try_then_request_governor

[PATCH] PM / devfreq: Return -ENODEV from try_then_request_governor

[Tomeu Vizoso]

Callers don't expect it to return NULL, but an error code.

Fixes Oops such as the one below, when one tries to set a governor that
isn't available:

Unable to handle kernel NULL pointer dereference at virtual address 00000018

[<c0dcf718>] (governor_store) from [<c0539558>] (kernfs_fop_write+0x100/0x1e0)
[<c0539558>] (kernfs_fop_write) from [<c04b09b4>] (__vfs_write+0x2c/0x17c)
[<c04b09b4>] (__vfs_write) from [<c04b33fc>] (vfs_write+0xa4/0x184)
[<c04b33fc>] (vfs_write) from [<c04b361c>] (ksys_write+0x4c/0xac)
[<c04b361c>] (ksys_write) from [<c0301000>] (ret_fast_syscall+0x0/0x28)

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Fixes: 23c7b54ca1cd ("PM / devfreq: Fix devfreq_add_device() when drivers are built as modules.")
Reported-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Cc: Enric Balletbo i Serra <enric.balletbo@collabora.com>
---
 drivers/devfreq/devfreq.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
index 0ae3de76833b..5539e9be718d 100644
--- a/drivers/devfreq/devfreq.c
+++ b/drivers/devfreq/devfreq.c
@@ -254,7 +254,7 @@ static struct devfreq_governor *try_then_request_governor(const char *name)
 		/* Restore previous state before return */
 		mutex_lock(&devfreq_list_lock);
 		if (err)
-			return NULL;
+			return ERR_PTR(-ENODEV);
 
 		governor = find_devfreq_governor(name);
 	}
-- 
2.20.1

Re: [PATCH] PM / devfreq: Return -ENODEV from try_then_request_governor

[Enric Balletbo i Serra]

Hi Tomeu,

On 23/4/19 10:11, Tomeu Vizoso wrote:
> Callers don't expect it to return NULL, but an error code.
> 
> Fixes Oops such as the one below, when one tries to set a governor that
> isn't available:
> 
> Unable to handle kernel NULL pointer dereference at virtual address 00000018
> 
> [<c0dcf718>] (governor_store) from [<c0539558>] (kernfs_fop_write+0x100/0x1e0)
> [<c0539558>] (kernfs_fop_write) from [<c04b09b4>] (__vfs_write+0x2c/0x17c)
> [<c04b09b4>] (__vfs_write) from [<c04b33fc>] (vfs_write+0xa4/0x184)
> [<c04b33fc>] (vfs_write) from [<c04b361c>] (ksys_write+0x4c/0xac)
> [<c04b361c>] (ksys_write) from [<c0301000>] (ret_fast_syscall+0x0/0x28)
> 
> Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
> Fixes: 23c7b54ca1cd ("PM / devfreq: Fix devfreq_add_device() when drivers are built as modules.")
> Reported-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
> Cc: Enric Balletbo i Serra <enric.balletbo@collabora.com>
> ---

There is already a fix for that. The fix was initially sent in October [2] but
unfortunately it got lost. I resend and now is queued [1]. Hopefully the Fixes
tag will help to pick the fix to the proper kernel releases.

Thanks,
 Enric

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/mzx/devfreq.git/commit/?h=for-next&id=b53b0128052ffd687797d5f4deeb76327e7b5711

[2] https://lkml.org/lkml/2018/10/16/744

>  drivers/devfreq/devfreq.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
> index 0ae3de76833b..5539e9be718d 100644
> --- a/drivers/devfreq/devfreq.c
> +++ b/drivers/devfreq/devfreq.c
> @@ -254,7 +254,7 @@ static struct devfreq_governor *try_then_request_governor(const char *name)
>  		/* Restore previous state before return */
>  		mutex_lock(&devfreq_list_lock);
>  		if (err)
> -			return NULL;
> +			return ERR_PTR(-ENODEV);
>  
>  		governor = find_devfreq_governor(name);
>  	}
> 

Re: [PATCH] PM / devfreq: Return -ENODEV from try_then_request_governor

[Tomeu Vizoso]

On Tue, 23 Apr 2019 at 11:56, Enric Balletbo i Serra
<enric.balletbo@collabora.com> wrote:
>
> Hi Tomeu,
>
> On 23/4/19 10:11, Tomeu Vizoso wrote:
> > Callers don't expect it to return NULL, but an error code.
> >
> > Fixes Oops such as the one below, when one tries to set a governor that
> > isn't available:
> >
> > Unable to handle kernel NULL pointer dereference at virtual address 00000018
> >
> > [<c0dcf718>] (governor_store) from [<c0539558>] (kernfs_fop_write+0x100/0x1e0)
> > [<c0539558>] (kernfs_fop_write) from [<c04b09b4>] (__vfs_write+0x2c/0x17c)
> > [<c04b09b4>] (__vfs_write) from [<c04b33fc>] (vfs_write+0xa4/0x184)
> > [<c04b33fc>] (vfs_write) from [<c04b361c>] (ksys_write+0x4c/0xac)
> > [<c04b361c>] (ksys_write) from [<c0301000>] (ret_fast_syscall+0x0/0x28)
> >
> > Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
> > Fixes: 23c7b54ca1cd ("PM / devfreq: Fix devfreq_add_device() when drivers are built as modules.")
> > Reported-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
> > Cc: Enric Balletbo i Serra <enric.balletbo@collabora.com>
> > ---
>
> There is already a fix for that. The fix was initially sent in October [2] but
> unfortunately it got lost. I resend and now is queued [1]. Hopefully the Fixes
> tag will help to pick the fix to the proper kernel releases.

Actually, Steve Price sent a third patch for this same issue.

Glad to read that it's being merged.

Thanks,

Tomeu



> Thanks,
>  Enric
>
> [1]
> https://git.kernel.org/pub/scm/linux/kernel/git/mzx/devfreq.git/commit/?h=for-next&id=b53b0128052ffd687797d5f4deeb76327e7b5711
>
> [2] https://lkml.org/lkml/2018/10/16/744
>
> >  drivers/devfreq/devfreq.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
> > index 0ae3de76833b..5539e9be718d 100644
> > --- a/drivers/devfreq/devfreq.c
> > +++ b/drivers/devfreq/devfreq.c
> > @@ -254,7 +254,7 @@ static struct devfreq_governor *try_then_request_governor(const char *name)
> >               /* Restore previous state before return */
> >               mutex_lock(&devfreq_list_lock);
> >               if (err)
> > -                     return NULL;
> > +                     return ERR_PTR(-ENODEV);
> >
> >               governor = find_devfreq_governor(name);
> >       }
> >