[RFC/RFT] ssb: Avoid system hang when SPROM read fails

[RFC/RFT] ssb: Avoid system hang when SPROM read fails

[Larry Finger]

In kernel Bugzilla #15825, the OP reports a case of intermittent reading
of the SPROM. If such reads fail, the box hangs. Thanks to careful testing
by bugzillakernelorg@lez.ath.cx has shown that the first read of the
SPROM returns 0xFFFF with the hang happening on the next read.

The source of the read failure is still under investigation; however,
this patch does avoid the system hang.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
---

John,

Does this patch avoid the system hang on your box?

Larry
---

===================================================================
--- wireless-testing.orig/drivers/ssb/pci.c
+++ wireless-testing/drivers/ssb/pci.c
@@ -253,6 +253,11 @@ static int sprom_do_read(struct ssb_bus
 {
 	int i;
 
+	/* Check if SPROM can be read */
+	if (ioread16(bus->mmio + SSB_SPROM_BASE) == 0xFFFF) {
+		ssb_printk(KERN_ERR PFX "Unable to read SPROM\n");
+		return -ENODEV;
+	}
 	for (i = 0; i < bus->sprom_size; i++)
 		sprom[i] = ioread16(bus->mmio + SSB_SPROM_BASE + (i * 2));
 
@@ -625,17 +630,23 @@ static int ssb_pci_sprom_get(struct ssb_
 	if (!buf)
 		goto out;
 	bus->sprom_size = SSB_SPROMSIZE_WORDS_R123;
-	sprom_do_read(bus, buf);
+	err = sprom_do_read(bus, buf);
+	if (err)
+		goto out_free;
 	err = sprom_check_crc(buf, bus->sprom_size);
 	if (err) {
 		/* try for a 440 byte SPROM - revision 4 and higher */
 		kfree(buf);
 		buf = kcalloc(SSB_SPROMSIZE_WORDS_R4, sizeof(u16),
 			      GFP_KERNEL);
-		if (!buf)
+		if (!buf) {
+			err = -ENOMEM;
 			goto out;
+		}
 		bus->sprom_size = SSB_SPROMSIZE_WORDS_R4;
-		sprom_do_read(bus, buf);
+		err = sprom_do_read(bus, buf);
+		if (err)
+			goto out_free;
 		err = sprom_check_crc(buf, bus->sprom_size);
 		if (err) {
 			/* All CRC attempts failed.

Re: [RFC/RFT] ssb: Avoid system hang when SPROM read fails

[John W. Linville]

On Sun, Apr 25, 2010 at 05:30:28PM -0500, Larry Finger wrote:
> In kernel Bugzilla #15825, the OP reports a case of intermittent reading
> of the SPROM. If such reads fail, the box hangs. Thanks to careful testing
> by bugzillakernelorg@lez.ath.cx has shown that the first read of the
> SPROM returns 0xFFFF with the hang happening on the next read.
> 
> The source of the read failure is still under investigation; however,
> this patch does avoid the system hang.
> 
> Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
> ---
> 
> John,
> 
> Does this patch avoid the system hang on your box?
> 
> Larry
> ---
> 
> ===================================================================
> --- wireless-testing.orig/drivers/ssb/pci.c
> +++ wireless-testing/drivers/ssb/pci.c
> @@ -253,6 +253,11 @@ static int sprom_do_read(struct ssb_bus
>  {
>  	int i;
>  
> +	/* Check if SPROM can be read */
> +	if (ioread16(bus->mmio + SSB_SPROM_BASE) == 0xFFFF) {
> +		ssb_printk(KERN_ERR PFX "Unable to read SPROM\n");
> +		return -ENODEV;
> +	}
>  	for (i = 0; i < bus->sprom_size; i++)
>  		sprom[i] = ioread16(bus->mmio + SSB_SPROM_BASE + (i * 2));
>  

Well, the "good" news is that I hit the "Unable to read SPROM" case.
The bad news is that the box still hangs after the -ENODEV.  I have
not yet tracked-down the exact location of the current hang.

John

P.S.  Sorry about the 'radio silence' -- I've been distracted with
some other things.  Also, the box in question originally belonged
to someone else who had configured it in a way that was less than
friendly to kernel development.  I finally reinstalled it...
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: [RFC/RFT] ssb: Avoid system hang when SPROM read fails

[Larry Finger]

On 04/26/2010 12:15 PM, John W. Linville wrote:
> 
> Well, the "good" news is that I hit the "Unable to read SPROM" case.
> The bad news is that the box still hangs after the -ENODEV.  I have
> not yet tracked-down the exact location of the current hang.
> 
> John
> 
> P.S.  Sorry about the 'radio silence' -- I've been distracted with
> some other things.  Also, the box in question originally belonged
> to someone else who had configured it in a way that was less than
> friendly to kernel development.  I finally reinstalled it...

That box is certainly resistant!

I certainly understand your distractions. That has to be a major
understatement.

As the patch makes some improvement in both your system and the one in
Bug #15825, I'll push it as a real patch.

Would your box be available to me as a loaner? I have considered buying
a netbook for debugging purposes, but I cannot quite justify the cost.
In addition, Murphy's law would make whatever one I bought work without
any problems.

Larry

Re: [RFC/RFT] ssb: Avoid system hang when SPROM read fails

[John W. Linville]

On Mon, Apr 26, 2010 at 01:04:33PM -0500, Larry Finger wrote:
> On 04/26/2010 12:15 PM, John W. Linville wrote:
> > 
> > Well, the "good" news is that I hit the "Unable to read SPROM" case.
> > The bad news is that the box still hangs after the -ENODEV.  I have
> > not yet tracked-down the exact location of the current hang.
> > 
> > John
> > 
> > P.S.  Sorry about the 'radio silence' -- I've been distracted with
> > some other things.  Also, the box in question originally belonged
> > to someone else who had configured it in a way that was less than
> > friendly to kernel development.  I finally reinstalled it...
> 
> That box is certainly resistant!
> 
> I certainly understand your distractions. That has to be a major
> understatement.
> 
> As the patch makes some improvement in both your system and the one in
> Bug #15825, I'll push it as a real patch.
> 
> Would your box be available to me as a loaner? I have considered buying
> a netbook for debugging purposes, but I cannot quite justify the cost.
> In addition, Murphy's law would make whatever one I bought work without
> any problems.

I think that can be arranged...

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.