Re: [PATCH] move file->f_op->open out of mutex (misc_mtx) in misc_open

Re: [PATCH] move file->f_op->open out of mutex (misc_mtx) in misc_open

[Hillf Danton]

> 
> misc_mtx was used to mutex misc_list.
> But file->f_op->open may be blocked by downloading
> firmware in some devices' open.
> So move file->f_op->open out of mutex (misc_mtx)
> in misc_open to avoid this block.
> It also make code more efficiency.
> 
> Signed-off-by: Zhang Dongxing <dongxing.zhang@intel.com>
> Signed-off-by: xiaoming wang <xiaoming.wang@intel.com>
> ---
>  drivers/char/misc.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/char/misc.c b/drivers/char/misc.c
> index ffa97d2..289c59e 100644
> --- a/drivers/char/misc.c
> +++ b/drivers/char/misc.c
> @@ -144,7 +144,8 @@ static int misc_open(struct inode * inode, struct file * file)
>  	replace_fops(file, new_fops);
>  	if (file->f_op->open) {
>  		file->private_data = c;
> -		err = file->f_op->open(inode,file);
> +		mutex_unlock(&misc_mtx);
> +		return file->f_op->open(inode,file);
>  	}
What if unregister breaks in while opening?

>  fail:
>  	mutex_unlock(&misc_mtx);
> --
> 1.7.9.5

Re: [PATCH] move file->f_op->open out of mutex (misc_mtx) in misc_open

[Greg KH]

On Tue, Jan 06, 2015 at 08:34:07AM +0800, xiaomin1 wrote:
> misc_mtx was used to mutex misc_list.
> But file->f_op->open may be blocked by downloading
> firmware in some devices' open.
> So move file->f_op->open out of mutex (misc_mtx)
> in misc_open to avoid this block.
> It also make code more efficiency.
> 
> Signed-off-by: Zhang Dongxing <dongxing.zhang@intel.com>
> Signed-off-by: xiaoming wang <xiaoming.wang@intel.com>
> ---
>  drivers/char/misc.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)


Please fix up your "From:" line to match your signed-off-by line, as I
can't take it the way you sent this.

greg k-h

[PATCH] move file->f_op->open out of mutex (misc_mtx) in misc_open

[xiaomin1]

misc_mtx was used to mutex misc_list.
But file->f_op->open may be blocked by downloading
firmware in some devices' open.
So move file->f_op->open out of mutex (misc_mtx)
in misc_open to avoid this block.
It also make code more efficiency.

Signed-off-by: Zhang Dongxing <dongxing.zhang@intel.com>
Signed-off-by: xiaoming wang <xiaoming.wang@intel.com>
---
 drivers/char/misc.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/char/misc.c b/drivers/char/misc.c
index ffa97d2..289c59e 100644
--- a/drivers/char/misc.c
+++ b/drivers/char/misc.c
@@ -144,7 +144,8 @@ static int misc_open(struct inode * inode, struct file * file)
 	replace_fops(file, new_fops);
 	if (file->f_op->open) {
 		file->private_data = c;
-		err = file->f_op->open(inode,file);
+		mutex_unlock(&misc_mtx);
+		return file->f_op->open(inode,file);
 	}
 fail:
 	mutex_unlock(&misc_mtx);
-- 
1.7.9.5

Re: [PATCH] move file->f_op->open out of mutex (misc_mtx) in misc_open

[One Thousand Gnomes]

On Tue,  6 Jan 2015 08:34:07 +0800
xiaomin1 <xiaoming.wang@intel.com> wrote:

> misc_mtx was used to mutex misc_list.
> But file->f_op->open may be blocked by downloading
> firmware in some devices' open.
> So move file->f_op->open out of mutex (misc_mtx)
> in misc_open to avoid this block.
> It also make code more efficiency.
> 
> Signed-off-by: Zhang Dongxing <dongxing.zhang@intel.com>
> Signed-off-by: xiaoming wang <xiaoming.wang@intel.com>
> ---
>  drivers/char/misc.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/char/misc.c b/drivers/char/misc.c
> index ffa97d2..289c59e 100644
> --- a/drivers/char/misc.c
> +++ b/drivers/char/misc.c
> @@ -144,7 +144,8 @@ static int misc_open(struct inode * inode, struct file * file)
>  	replace_fops(file, new_fops);
>  	if (file->f_op->open) {
>  		file->private_data = c;
> -		err = file->f_op->open(inode,file);
> +		mutex_unlock(&misc_mtx);
> +		return file->f_op->open(inode,file);
>  	}

This changes all the users of misc device open from singlethreaded to
multi-threaded. Having read through the users I can find this seems safe
and they all seem to have the right checks in place and to assume they
are multi-threaded.

Reviewed-by: Alan Cox <alan@linux.intel.com>