staging: comedi: amplc_dio200: store the pci_dev in the comedi_device

PATCH In review H Hartley Sweeten 2 participants 6 messages Jul 19, 2012 Staging

Share & tools

H Hartley Sweeten

author

2012-07-19 01:37 UTC #0 lore raw

PATCH 30/90 staging: comedi: amplc_dio200: store the pci_dev in the comedi_device

Use the hw_dev pointer in the comedi_device struct to hold the pci_dev instead of carrying it in the private data.

Since the pci_dev is no longer held in the provate data, we can also cleanup the detach a bit. Remove the IS_ENABLED() tests in the detach. If the pci_dev is non NULL it's a PCI device otherwise it's an ISA device. Using IS_ENABLED() to omit the code paths makes the code a bit confusing and doesn't save much.

Signed-off-by H Hartley Sweeten <[email protected]>

Cc Ian Abbott <[email protected]>

Cc Greg Kroah-Hartman <[email protected]>

---
 drivers/staging/comedi/drivers/amplc_dio200.c | 31 +++++++++++----------------
 1 file changed, 12 insertions(+), 19 deletions(-)

drivers/staging/comedi/drivers/amplc_dio200.c +12 ~~-19~~

index ca9d4f2..6c81e377 100644
@@ -424,7 +424,6 @@ static const struct dio200_layout_struct dio200_layouts[] = {
    feel free to suggest moving the variable to the struct comedi_device struct.
  */
 struct dio200_private {
-	struct pci_dev *pci_dev;	/* PCI device */
 	int intr_sd;
 };

@@ -1225,7 +1224,7 @@ dio200_subdev_8254_cleanup(struct comedi_device *dev,
 static void dio200_report_attach(struct comedi_device *dev, unsigned int irq)
 {
 	const struct dio200_board *thisboard = comedi_board(dev);
-	struct dio200_private *devpriv = dev->private;
+	struct pci_dev *pcidev = comedi_to_pci_dev(dev);
 	char tmpbuf[60];
 	int tmplen;

@@ -1236,7 +1235,7 @@ static void dio200_report_attach(struct comedi_device *dev, unsigned int irq)
 	else if (IS_ENABLED(CONFIG_COMEDI_AMPLC_DIO200_PCI) &&
 		 thisboard->bustype == pci_bustype)
 		tmplen = scnprintf(tmpbuf, sizeof(tmpbuf),
-				   "(pci %s) ", pci_name(devpriv->pci_dev));
+				   "(pci %s) ", pci_name(pcidev));
 	else
 		tmplen = 0;
 	if (irq)

@@ -1327,11 +1326,11 @@ static int dio200_common_attach(struct comedi_device *dev, unsigned long iobase,
 static int dio200_pci_common_attach(struct comedi_device *dev,
 				    struct pci_dev *pci_dev)
 {
-	struct dio200_private *devpriv = dev->private;
 	unsigned long iobase;
 	int ret;
 
-	devpriv->pci_dev = pci_dev;
+	comedi_set_hw_dev(dev, &pci_dev->dev);
+
 	ret = comedi_pci_enable(pci_dev, DIO200_DRIVER_NAME);
 	if (ret < 0) {
 		dev_err(dev->class_dev,

@@ -1419,7 +1418,7 @@ static int __devinit dio200_attach_pci(struct comedi_device *dev,
 static void dio200_detach(struct comedi_device *dev)
 {
 	const struct dio200_board *thisboard = comedi_board(dev);
-	struct dio200_private *devpriv = dev->private;
+	struct pci_dev *pcidev = comedi_to_pci_dev(dev);
 	const struct dio200_layout_struct *layout;
 	unsigned n;

@@ -1444,19 +1443,13 @@ static void dio200_detach(struct comedi_device *dev)
 			}
 		}
 	}
-	if (devpriv) {
-		if (IS_ENABLED(CONFIG_COMEDI_AMPLC_DIO200_PCI) &&
-		    thisboard->bustype == pci_bustype) {
-			if (devpriv->pci_dev) {
-				if (dev->iobase)
-					comedi_pci_disable(devpriv->pci_dev);
-				pci_dev_put(devpriv->pci_dev);
-			}
-		} else if (IS_ENABLED(CONFIG_COMEDI_AMPLC_DIO200_ISA) &&
-			   thisboard->bustype == isa_bustype) {
-			if (dev->iobase)
-				release_region(dev->iobase, DIO200_IO_SIZE);
-		}
+	if (pcidev) {
+		if (dev->iobase)
+			comedi_pci_disable(pcidev);
+		pci_dev_put(pcidev);
+	} else {
+		if (dev->iobase)
+			release_region(dev->iobase, DIO200_IO_SIZE);
 	}
 }

Signature · 1 line

1.7.11

Ian Abbott

2012-07-19 10:10 UTC #1 lore raw

Quoted 8 lines · H Hartley Sweeten

Use the hw_dev pointer in the comedi_device struct to hold the pci_dev instead of carrying it in the private data.
Since the pci_dev is no longer held in the provate data, we can also cleanup the detach a bit. Remove the IS_ENABLED() tests in the detach. If the pci_dev is non NULL it's a PCI device otherwise it's an ISA device. Using IS_ENABLED() to omit the code paths makes the code a bit confusing and doesn't save much.

No, you really need to check thisboard->bustype in the detach() because hw_dev might be NULL for a PCI board if the attach() failed.

Signature · 2 lines

-=( Ian Abbott @ MEV Ltd.    E-mail: <[email protected]>        )=-
-=( Tel: +44 (0)161 477 1898   FAX: +44 (0)161 718 3587         )=-

Ian Abbott

2012-07-19 10:27 UTC #2 lore raw

Quoted 4 lines · Ian Abbott

Quoted 3 lines · H Hartley Sweeten
Use the hw_dev pointer in the comedi_device struct to hold the pci_dev instead of carrying it in the private data.

Quoted 8 lines

Quoted 5 lines
Since the pci_dev is no longer held in the provate data, we can also cleanup the detach a bit. Remove the IS_ENABLED() tests in the detach. If the pci_dev is non NULL it's a PCI device otherwise it's an ISA device. Using IS_ENABLED() to omit the code paths makes the code a bit confusing and doesn't save much.
No, you really need to check thisboard->bustype in the detach() because hw_dev might be NULL for a PCI board if the attach() failed.

Actually, your patch wouldn't do any harm as dev->iobase would be 0. It would just go through the "failed ISA device" path instead of the "failed PCI device" path. It would be more robust to check thisboard->bustype though.

Signature · 2 lines

-=( Ian Abbott @ MEV Ltd.    E-mail: <[email protected]>        )=-
-=( Tel: +44 (0)161 477 1898   FAX: +44 (0)161 718 3587         )=-

Ian Abbott

2012-07-19 10:38 UTC #3 lore raw

Quoted 11 lines · Ian Abbott

Quoted 10 lines · Ian Abbott
Quoted 8 lines · H Hartley Sweeten
Use the hw_dev pointer in the comedi_device struct to hold the pci_dev instead of carrying it in the private data.
Since the pci_dev is no longer held in the provate data, we can also cleanup the detach a bit. Remove the IS_ENABLED() tests in the detach. If the pci_dev is non NULL it's a PCI device otherwise it's an ISA device. Using IS_ENABLED() to omit the code paths makes the code a bit confusing and doesn't save much.

Quoted 7 lines

Quoted 2 lines
No, you really need to check thisboard->bustype in the detach() because hw_dev might be NULL for a PCI board if the attach() failed.
Actually, your patch wouldn't do any harm as dev->iobase would be 0. It would just go through the "failed ISA device" path instead of the "failed PCI device" path. It would be more robust to check thisboard->bustype though.

I don't mind if your original patch is applied as it does no harm. If so, I'll submit a patch afterwards to check the bus type in the detach().

Signature · 2 lines

-=( Ian Abbott @ MEV Ltd.    E-mail: <[email protected]>        )=-
-=( Tel: +44 (0)161 477 1898   FAX: +44 (0)161 718 3587         )=-

H Hartley Sweeten

author

2012-07-19 17:20 UTC #4 lore raw

In reply to Ian Abbott

Quoted 11 lines · Ian Abbott

Quoted 10 lines · Ian Abbott
Quoted 8 lines · H Hartley Sweeten
Use the hw_dev pointer in the comedi_device struct to hold the pci_dev instead of carrying it in the private data.
Since the pci_dev is no longer held in the provate data, we can also cleanup the detach a bit. Remove the IS_ENABLED() tests in the detach. If the pci_dev is non NULL it's a PCI device otherwise it's an ISA device. Using IS_ENABLED() to omit the code paths makes the code a bit confusing and doesn't save much.

Quoted 7 lines

Quoted 2 lines
No, you really need to check thisboard->bustype in the detach() because hw_dev might be NULL for a PCI board if the attach() failed.
Actually, your patch wouldn't do any harm as dev->iobase would be 0. It would just go through the "failed ISA device" path instead of the "failed PCI device" path. It would be more robust to check thisboard->bustype though.

Would this be cleaner:

	If (dev->iobase) {
		If (pcidev)
			comedi_pci_disable(pcidev);
		else
			release_region(dev->iobase, DIO200_IO_SIZE);
	}
	If (pcidev)
		pci_dev_put(pcidev);

Same number of lines but the i/o disable/release and the put of the pcidev would then be distinctly separate.

Regards,
Hartley

��{.n�+��+%��ݶ��w��{.n�+��{��G��{ay�ʇڙ�,j��f��h��z_�(�階�ݢj"��m��G��?��&��~��iO��z��v�^�m��?�I�

H Hartley Sweeten

author

2012-07-19 17:14 UTC #5 lore raw

In reply to Ian Abbott

Quoted 4 lines · Ian Abbott

Quoted 3 lines · H Hartley Sweeten
Use the hw_dev pointer in the comedi_device struct to hold the pci_dev instead of carrying it in the private data.

Quoted 8 lines

Quoted 5 lines
Since the pci_dev is no longer held in the provate data, we can also cleanup the detach a bit. Remove the IS_ENABLED() tests in the detach. If the pci_dev is non NULL it's a PCI device otherwise it's an ISA device. Using IS_ENABLED() to omit the code paths makes the code a bit confusing and doesn't save much.
No, you really need to check thisboard->bustype in the detach() because hw_dev might be NULL for a PCI board if the attach() failed.

Ah, overlooked that.

I still think the IS_ENABLED() stuff should be removed where possible. It just looks confusing in the if().

Regards,
Hartley