Skip to contentSkip to conversation

PCI: add pci_try_set_mwi

PATCH Matthew Wilcox 4 participants 5 messages Jul 12, 2007 10 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

#10 lore raw
Quoted 2 lines · Greg Kroah-Hartman

As suggested by Andrew, add pci_try_set_mwi(), which does not require return-value checking.

Seems like a daft suggestion. What's wrong with just removing the __must_check from pci_set_mwi()? Did it find any bugs?

drivers/scsi/lpfc/lpfc_init.c lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid) · line 1581

Quoted 11 lines
drivers/scsi/lpfc/lpfc_init.c +1 -4
@@ -1578,10 +1578,7 @@ lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid)
 	INIT_LIST_HEAD(&phba->fc_nodes);
 
 	pci_set_master(pdev);
-	retval = pci_set_mwi(pdev);
-	if (retval)
-		dev_printk(KERN_WARNING, &pdev->dev,
-			   "Warning: pci_set_mwi returned %d\n", retval);
+	pci_try_set_mwi(pdev);

Why remove the warning? Presumably people want to know if pci_set_mwi failed.

Signature · 3 lines
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."
#11 lore raw
Quoted 12 lines · Matthew Wilcox
Quoted 2 lines · Greg Kroah-Hartman

As suggested by Andrew, add pci_try_set_mwi(), which does not require return-value checking.

Seems like a daft suggestion. What's wrong with just removing the __must_check from pci_set_mwi()? Did it find any bugs?

Quoted 5 lines
drivers/scsi/lpfc/lpfc_init.c +0 -0
@@ -1578,10 +1578,7 @@ lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid)
 	INIT_LIST_HEAD(&phba->fc_nodes);
 
Quoted 9 lines

drivers/scsi/lpfc/lpfc_init.c lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid) · line 1581

Quoted 6 lines
drivers/scsi/lpfc/lpfc_init.c +1 -4
 	pci_set_master(pdev);
-	retval = pci_set_mwi(pdev);
-	if (retval)
-		dev_printk(KERN_WARNING, &pdev->dev,
-			   "Warning: pci_set_mwi returned %d\n", retval);
+	pci_try_set_mwi(pdev);

Why remove the warning? Presumably people want to know if pci_set_mwi failed.

Randy, this was your change, right?

thanks,

greg k-h

#12 lore raw
Quoted 23 lines · On Wed, 11 Jul 2007 17:37:31 -0700 Greg KH wrote:
Quoted 12 lines · Matthew Wilcox
Quoted 2 lines · Greg Kroah-Hartman

As suggested by Andrew, add pci_try_set_mwi(), which does not require return-value checking.

Seems like a daft suggestion. What's wrong with just removing the __must_check from pci_set_mwi()? Did it find any bugs?

Quoted 5 lines
drivers/scsi/lpfc/lpfc_init.c +0 -0
@@ -1578,10 +1578,7 @@ lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid)
 	INIT_LIST_HEAD(&phba->fc_nodes);
 
Quoted 9 lines

drivers/scsi/lpfc/lpfc_init.c lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid) · line 1581

Quoted 6 lines
drivers/scsi/lpfc/lpfc_init.c +1 -4
 	pci_set_master(pdev);
-	retval = pci_set_mwi(pdev);
-	if (retval)
-		dev_printk(KERN_WARNING, &pdev->dev,
-			   "Warning: pci_set_mwi returned %d\n", retval);
+	pci_try_set_mwi(pdev);

Why remove the warning? Presumably people want to know if pci_set_mwi failed.

Quoted 1 line

Randy, this was your change, right?

Yes, I'm trying to track that down...

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#13 lore raw
In reply to Greg KH
Quoted 7 lines · On Wed, 11 Jul 2007 17:37:31 -0700 Greg KH wrote:
Quoted 6 lines · Matthew Wilcox
Quoted 2 lines · Greg Kroah-Hartman

As suggested by Andrew, add pci_try_set_mwi(), which does not require return-value checking.

Seems like a daft suggestion. What's wrong with just removing the __must_check from pci_set_mwi()? Did it find any bugs?

(a) Alan suggested just dropping __must_check IIRC.
And David Brownell even sent a patch to do that (which Alan acked).

(b) not that I know of.

Quoted 15 lines
Quoted 5 lines
Quoted 5 lines
drivers/scsi/lpfc/lpfc_init.c +0 -0
@@ -1578,10 +1578,7 @@ lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid)
 	INIT_LIST_HEAD(&phba->fc_nodes);
 
Quoted 9 lines

drivers/scsi/lpfc/lpfc_init.c lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid) · line 1581

Quoted 6 lines
drivers/scsi/lpfc/lpfc_init.c +1 -4
 	pci_set_master(pdev);
-	retval = pci_set_mwi(pdev);
-	if (retval)
-		dev_printk(KERN_WARNING, &pdev->dev,
-			   "Warning: pci_set_mwi returned %d\n", retval);
+	pci_try_set_mwi(pdev);

Why remove the warning? Presumably people want to know if pci_set_mwi failed.

Quoted 1 line

Randy, this was your change, right?

Uh, I think that my thinking was like this:

pci_try_set_mwi() and pci_set_mwi() are both "try best effort" functions. Neither of them guarantees that pci_set_cacheline_size() will succeed. And in case of serious problems, pci_set_cacheline_size() will print a (KERN_DEBUG) message.

Anyway, I don't mind restoring the former lpfc code if that is what should be done.

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#14 lore raw

I'm agnostic on the change... As long as we get a message somewhere when the failure is meaningful, I'm fine with this change. I didn't like setting mwi by the driver anyway - it should have already been done by the platform.

-- james s

Quoted 30 lines · Randy Dunlap
Quoted 13 lines
Quoted 5 lines
Quoted 5 lines
drivers/scsi/lpfc/lpfc_init.c +0 -0
@@ -1578,10 +1578,7 @@ lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid)
 	INIT_LIST_HEAD(&phba->fc_nodes);
 
Quoted 8 lines

drivers/scsi/lpfc/lpfc_init.c lpfc_pci_probe_one(struct pci_dev *pdev, const struct pci_device_id *pid) · line 1581

Quoted 6 lines
drivers/scsi/lpfc/lpfc_init.c +1 -4
 	pci_set_master(pdev);
-	retval = pci_set_mwi(pdev);
-	if (retval)
-		dev_printk(KERN_WARNING, &pdev->dev,
-			   "Warning: pci_set_mwi returned %d\n", retval);
+	pci_try_set_mwi(pdev);

Why remove the warning? Presumably people want to know if pci_set_mwi failed.

Quoted 1 line

Randy, this was your change, right?

Uh, I think that my thinking was like this:

pci_try_set_mwi() and pci_set_mwi() are both "try best effort" functions. Neither of them guarantees that pci_set_cacheline_size() will succeed. And in case of serious problems, pci_set_cacheline_size() will print a (KERN_DEBUG) message.

Anyway, I don't mind restoring the former lpfc code if that is what should be done.

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***