[v1] Revert "char: pcmcia: cm4000_cs: Replace mdelay with usleep_range in set_protocol"

[PATCH] Revert "char: pcmcia: cm4000_cs: Replace mdelay with usleep_range in set_protocol"

Posted by Duoming Zhou 2 years, 9 months ago

This reverts commit be826ada52f1fcabed5b5217c94609ebf5967211.

The function monitor_card() is a timer handler that runs in an
atomic context, but it calls usleep_range() that can sleep.
As a result, the sleep-in-atomic-context bugs will happen.
The process is shown below:

    (atomic context)
monitor_card()
  set_protocol()
    usleep_range() //sleep

The origin commit c1986ee9bea3 ("[PATCH] New Omnikey Cardman
4000 driver") works fine.

Fixes: be826ada52f1 ("char: pcmcia: cm4000_cs: Replace mdelay with usleep_range in set_protocol")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
---
 drivers/char/pcmcia/cm4000_cs.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/char/pcmcia/cm4000_cs.c b/drivers/char/pcmcia/cm4000_cs.c
index adaec8fd4b1..7125f89531f 100644
--- a/drivers/char/pcmcia/cm4000_cs.c
+++ b/drivers/char/pcmcia/cm4000_cs.c
@@ -529,7 +529,7 @@ static int set_protocol(struct cm4000_dev *dev, struct ptsreq *ptsreq)
 			DEBUGP(5, dev, "NumRecBytes is valid\n");
 			break;
 		}
-		usleep_range(10000, 11000);
+		mdelay(10);
 	}
 	if (i == 100) {
 		DEBUGP(5, dev, "Timeout waiting for NumRecBytes getting "
@@ -549,7 +549,7 @@ static int set_protocol(struct cm4000_dev *dev, struct ptsreq *ptsreq)
 			}
 			break;
 		}
-		usleep_range(10000, 11000);
+		mdelay(10);
 	}
 
 	/* check whether it is a short PTS reply? */
-- 
2.17.1

Re: [PATCH] Revert "char: pcmcia: cm4000_cs: Replace mdelay with usleep_range in set_protocol"

Posted by Greg KH 2 years, 9 months ago

On Mon, Nov 28, 2022 at 01:39:15PM +0800, Duoming Zhou wrote:
> This reverts commit be826ada52f1fcabed5b5217c94609ebf5967211.
> 
> The function monitor_card() is a timer handler that runs in an
> atomic context, but it calls usleep_range() that can sleep.
> As a result, the sleep-in-atomic-context bugs will happen.
> The process is shown below:
> 
>     (atomic context)
> monitor_card()
>   set_protocol()
>     usleep_range() //sleep
> 
> The origin commit c1986ee9bea3 ("[PATCH] New Omnikey Cardman
> 4000 driver") works fine.
> 
> Fixes: be826ada52f1 ("char: pcmcia: cm4000_cs: Replace mdelay with usleep_range in set_protocol")
> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> ---
>  drivers/char/pcmcia/cm4000_cs.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/char/pcmcia/cm4000_cs.c b/drivers/char/pcmcia/cm4000_cs.c
> index adaec8fd4b1..7125f89531f 100644
> --- a/drivers/char/pcmcia/cm4000_cs.c
> +++ b/drivers/char/pcmcia/cm4000_cs.c
> @@ -529,7 +529,7 @@ static int set_protocol(struct cm4000_dev *dev, struct ptsreq *ptsreq)
>  			DEBUGP(5, dev, "NumRecBytes is valid\n");
>  			break;
>  		}
> -		usleep_range(10000, 11000);
> +		mdelay(10);
>  	}
>  	if (i == 100) {
>  		DEBUGP(5, dev, "Timeout waiting for NumRecBytes getting "
> @@ -549,7 +549,7 @@ static int set_protocol(struct cm4000_dev *dev, struct ptsreq *ptsreq)
>  			}
>  			break;
>  		}
> -		usleep_range(10000, 11000);
> +		mdelay(10);

Can you add a comment that says "can not sleep as this is atomic
context" for these calls so that this doesn't get changed again in the
future?

thanks,

greg k-h

Re: [PATCH] Revert "char: pcmcia: cm4000_cs: Replace mdelay with usleep_range in set_protocol"

Posted by duoming@zju.edu.cn 2 years, 9 months ago

Hello,

On Sun, 18 Dec 2022 12:43:27 +0100 Greg KH wrote:

> On Mon, Nov 28, 2022 at 01:39:15PM +0800, Duoming Zhou wrote:
> > This reverts commit be826ada52f1fcabed5b5217c94609ebf5967211.
> > 
> > The function monitor_card() is a timer handler that runs in an
> > atomic context, but it calls usleep_range() that can sleep.
> > As a result, the sleep-in-atomic-context bugs will happen.
> > The process is shown below:
> > 
> >     (atomic context)
> > monitor_card()
> >   set_protocol()
> >     usleep_range() //sleep
> > 
> > The origin commit c1986ee9bea3 ("[PATCH] New Omnikey Cardman
> > 4000 driver") works fine.
> > 
> > Fixes: be826ada52f1 ("char: pcmcia: cm4000_cs: Replace mdelay with usleep_range in set_protocol")
> > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > ---
> >  drivers/char/pcmcia/cm4000_cs.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/char/pcmcia/cm4000_cs.c b/drivers/char/pcmcia/cm4000_cs.c
> > index adaec8fd4b1..7125f89531f 100644
> > --- a/drivers/char/pcmcia/cm4000_cs.c
> > +++ b/drivers/char/pcmcia/cm4000_cs.c
> > @@ -529,7 +529,7 @@ static int set_protocol(struct cm4000_dev *dev, struct ptsreq *ptsreq)
> >  			DEBUGP(5, dev, "NumRecBytes is valid\n");
> >  			break;
> >  		}
> > -		usleep_range(10000, 11000);
> > +		mdelay(10);
> >  	}
> >  	if (i == 100) {
> >  		DEBUGP(5, dev, "Timeout waiting for NumRecBytes getting "
> > @@ -549,7 +549,7 @@ static int set_protocol(struct cm4000_dev *dev, struct ptsreq *ptsreq)
> >  			}
> >  			break;
> >  		}
> > -		usleep_range(10000, 11000);
> > +		mdelay(10);
> 
> Can you add a comment that says "can not sleep as this is atomic
> context" for these calls so that this doesn't get changed again in the
> future?

Thank you for your reply and suggestions! I will add the comment and re-send this patch.

Best regards,
Duoming Zhou