Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1512223 > unrolled thread

[PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init()

Started byOndrej Zary <linux@rainbow-software.org>
First post2016-10-30 23:50 +0100
Last post2016-10-31 04:40 +0100
Articles 5 — 2 participants

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init() Ondrej Zary <linux@rainbow-software.org> - 2016-10-30 23:50 +0100
    Re: [PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init() Finn Thain <fthain@telegraphics.com.au> - 2016-10-31 04:20 +0100
      Re: [PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init() Ondrej Zary <linux@rainbow-software.org> - 2016-10-31 09:10 +0100
        Re: [PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init() Finn Thain <fthain@telegraphics.com.au> - 2016-10-31 10:40 +0100
    Re: [PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init() Finn Thain <fthain@telegraphics.com.au> - 2016-10-31 04:40 +0100

#1512223 — [PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init()

FromOndrej Zary <linux@rainbow-software.org>
Date2016-10-30 23:50 +0100
Subject[PATCH 3/3] NCR5380: Check for chip presence in NCR5380_init()
Message-ID<sy75o-2jv-9@gated-at.bofh.it>
Read back MODE_REG after writing it in NCR5380_init() to check if the
chip is really there.

This prevents hang when incorrect I/O address was specified by user.

Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
---
 drivers/scsi/NCR5380.c |    5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/scsi/NCR5380.c b/drivers/scsi/NCR5380.c
index 01c0027..ce3156d 100644
--- a/drivers/scsi/NCR5380.c
+++ b/drivers/scsi/NCR5380.c
@@ -495,6 +495,11 @@ static int NCR5380_init(struct Scsi_Host *instance, int flags)
 
 	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
 	NCR5380_write(MODE_REG, MR_BASE);
+	/* check if the chip is really there */
+	if (NCR5380_read(MODE_REG) != MR_BASE) {
+		NCR5380_exit(instance);
+		return -ENODEV;
+	}
 	NCR5380_write(TARGET_COMMAND_REG, 0);
 	NCR5380_write(SELECT_ENABLE_REG, 0);
 
-- 
Ondrej Zary

[toc] | [next] | [standalone]

#1512300

FromFinn Thain <fthain@telegraphics.com.au>
Date2016-10-31 04:20 +0100
Message-ID<sybiF-5ce-3@gated-at.bofh.it>
In reply to#1512223
On Sun, 30 Oct 2016, Ondrej Zary wrote:

> Read back MODE_REG after writing it in NCR5380_init() to check if the
> chip is really there.
> 
> This prevents hang when incorrect I/O address was specified by user.
> 
> Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
> ---
>  drivers/scsi/NCR5380.c |    5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/scsi/NCR5380.c b/drivers/scsi/NCR5380.c
> index 01c0027..ce3156d 100644
> --- a/drivers/scsi/NCR5380.c
> +++ b/drivers/scsi/NCR5380.c
> @@ -495,6 +495,11 @@ static int NCR5380_init(struct Scsi_Host *instance, int flags)
>  
>  	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
>  	NCR5380_write(MODE_REG, MR_BASE);
> +	/* check if the chip is really there */
> +	if (NCR5380_read(MODE_REG) != MR_BASE) {
> +		NCR5380_exit(instance);
> +		return -ENODEV;
> +	}

This doesn't belong in the core driver. Only the 5380 ISA drivers have 
configurable base addresses.

Also, MR_BASE == 0, so that test is likely to be ineffectual anyway. This 
patch doesn't really add any value AFAICT.

-- 

>  	NCR5380_write(TARGET_COMMAND_REG, 0);
>  	NCR5380_write(SELECT_ENABLE_REG, 0);
>  
>

[toc] | [prev] | [next] | [standalone]

#1512386

FromOndrej Zary <linux@rainbow-software.org>
Date2016-10-31 09:10 +0100
Message-ID<syfPj-89m-9@gated-at.bofh.it>
In reply to#1512300
On Monday 31 October 2016, Finn Thain wrote:
> On Sun, 30 Oct 2016, Ondrej Zary wrote:
> > Read back MODE_REG after writing it in NCR5380_init() to check if the
> > chip is really there.
> >
> > This prevents hang when incorrect I/O address was specified by user.
> >
> > Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
> > ---
> >  drivers/scsi/NCR5380.c |    5 +++++
> >  1 file changed, 5 insertions(+)
> >
> > diff --git a/drivers/scsi/NCR5380.c b/drivers/scsi/NCR5380.c
> > index 01c0027..ce3156d 100644
> > --- a/drivers/scsi/NCR5380.c
> > +++ b/drivers/scsi/NCR5380.c
> > @@ -495,6 +495,11 @@ static int NCR5380_init(struct Scsi_Host *instance,
> > int flags)
> >
> >  	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
> >  	NCR5380_write(MODE_REG, MR_BASE);
> > +	/* check if the chip is really there */
> > +	if (NCR5380_read(MODE_REG) != MR_BASE) {
> > +		NCR5380_exit(instance);
> > +		return -ENODEV;
> > +	}
>
> This doesn't belong in the core driver. Only the 5380 ISA drivers have
> configurable base addresses.
>
> Also, MR_BASE == 0, so that test is likely to be ineffectual anyway. This
> patch doesn't really add any value AFAICT.

This fixes the most common problem: no device present at the specified I/O 
address, all reads result in 0xff.

-- 
Ondrej Zary

[toc] | [prev] | [next] | [standalone]

#1512424

FromFinn Thain <fthain@telegraphics.com.au>
Date2016-10-31 10:40 +0100
Message-ID<syhep-us-3@gated-at.bofh.it>
In reply to#1512386
On Mon, 31 Oct 2016, Ondrej Zary wrote:

> On Monday 31 October 2016, Finn Thain wrote:
> > On Sun, 30 Oct 2016, Ondrej Zary wrote:
> > > Read back MODE_REG after writing it in NCR5380_init() to check if the
> > > chip is really there.
> > >
> > > This prevents hang when incorrect I/O address was specified by user.
> > >
> > > Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
> > > ---
> > >  drivers/scsi/NCR5380.c |    5 +++++
> > >  1 file changed, 5 insertions(+)
> > >
> > > diff --git a/drivers/scsi/NCR5380.c b/drivers/scsi/NCR5380.c
> > > index 01c0027..ce3156d 100644
> > > --- a/drivers/scsi/NCR5380.c
> > > +++ b/drivers/scsi/NCR5380.c
> > > @@ -495,6 +495,11 @@ static int NCR5380_init(struct Scsi_Host *instance,
> > > int flags)
> > >
> > >  	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
> > >  	NCR5380_write(MODE_REG, MR_BASE);
> > > +	/* check if the chip is really there */
> > > +	if (NCR5380_read(MODE_REG) != MR_BASE) {
> > > +		NCR5380_exit(instance);
> > > +		return -ENODEV;
> > > +	}
> >
> > This doesn't belong in the core driver. Only the 5380 ISA drivers have 
> > configurable base addresses.
> >
> > Also, MR_BASE == 0, so that test is likely to be ineffectual anyway. 
> > This patch doesn't really add any value AFAICT.
> 
> This fixes the most common problem: no device present at the specified 
> I/O address, all reads result in 0xff.
> 

ISTR that's true for ISA in general, so I guess it would be okay in 
g_NCR5380.c.

Better do this before NCR5380_init() so you can probe before allocating a 
scsi host. Just write 0 to MODE_REG or SELECT_ENABLE_REG.

--

[toc] | [prev] | [next] | [standalone]

#1512304

FromFinn Thain <fthain@telegraphics.com.au>
Date2016-10-31 04:40 +0100
Message-ID<sybC6-5ja-11@gated-at.bofh.it>
In reply to#1512223
On Sun, 30 Oct 2016, Ondrej Zary wrote:

> Read back MODE_REG after writing it in NCR5380_init() to check if the
> chip is really there.
> 
> This prevents hang when incorrect I/O address was specified by user.

Do you know whereabouts in the driver the hang happens? Maybe there is a 
robustness issue there.

Card type detection (and vacant slot detection) is a good idea but I'm not 
sure how we can detect this chip reliably.

-- 

> 
> Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
> ---
>  drivers/scsi/NCR5380.c |    5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/scsi/NCR5380.c b/drivers/scsi/NCR5380.c
> index 01c0027..ce3156d 100644
> --- a/drivers/scsi/NCR5380.c
> +++ b/drivers/scsi/NCR5380.c
> @@ -495,6 +495,11 @@ static int NCR5380_init(struct Scsi_Host *instance, int flags)
>  
>  	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
>  	NCR5380_write(MODE_REG, MR_BASE);
> +	/* check if the chip is really there */
> +	if (NCR5380_read(MODE_REG) != MR_BASE) {
> +		NCR5380_exit(instance);
> +		return -ENODEV;
> +	}
>  	NCR5380_write(TARGET_COMMAND_REG, 0);
>  	NCR5380_write(SELECT_ENABLE_REG, 0);
>  
>

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web