Skip to contentSkip to conversation

PCI: also read revision ID for sparc64, ppc, read class at the same time

PATCH In review Auke Kok 5 participants 7 messages Jun 9, 2007 – Jun 22, 2007

Subsystems (2)
lore
Share & tools
lore

Auke Kok

author
#0 lore raw

PATCH 1/1 PCI: also read revision ID for sparc64, ppc, read class at the same time

Since the PCI bus class and revision are in the same dword, we can fill in both in the same read.

Other non-x86 arches use different methods to fill in these values, make sure to adjust those too.

---

 arch/powerpc/kernel/pci_64.c |    2 ++
 arch/sparc64/kernel/pci.c    |    1 +
 drivers/pci/probe.c          |    4 +---
 3 files changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kernel/pci_64.c +2 -0
index 249cca2..b2c55ca 100644
@@ -367,8 +367,10 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
 	sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 		dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
 	dev->class = get_int_prop(node, "class-code", 0);
+	dev->revision = get_int_prop(node, "revision-id", 0);
 
 	DBG("    class: 0x%x\n", dev->class);
+	DBG("    revision: 0x%x\n", dev->revision);
 
 	dev->current_state = 4;		/* unknown power state */
 	dev->error_state = pci_channel_io_normal;
arch/sparc64/kernel/pci.c +1 -0
index 38a32bc..759cc20 100644
@@ -448,6 +448,7 @@ struct pci_dev *of_create_pci_dev(struct pci_pbm_info *pbm,
 		 */
 		pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
 		dev->class = class >> 8;
+		dev->revision = class & 0xff;
 
 		sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 			dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
drivers/pci/probe.c +1 -3
index 0fdb71d..a574b7f 100644
@@ -702,6 +702,7 @@ static int pci_setup_device(struct pci_dev * dev)
 		dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn));
 
 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
+	dev->revision = class & 0xff;
 	class >>= 8;				    /* upper 3 bytes */
 	dev->class = class;
 	class >>= 8;
@@ -918,9 +919,6 @@ pci_scan_device(struct pci_bus *bus, int devfn)
 	dev->cfg_size = pci_cfg_space_size(dev);
 	dev->error_state = pci_channel_io_normal;
 
-	/* read the PCI revision: 1 byte */
-	pci_read_config_byte(dev, PCI_REVISION_ID, &dev->revision);
-
 	/* Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
 	   set this higher, assuming the system even supports it.  */
 	dev->dma_mask = 0xffffffff;
Acked-by on v1 1/1
#1 lore raw
Quoted 5 lines · Auke Kok

Since the PCI bus class and revision are in the same dword, we can fill in both in the same read.

Other non-x86 arches use different methods to fill in these values, make sure to adjust those too.

Yeah, this looks good to me.

I didn't know ppc and sparc had their own methods for device discovery.

thanks,
grant

Quoted 58 lines

---

 arch/powerpc/kernel/pci_64.c |    2 ++
 arch/sparc64/kernel/pci.c    |    1 +
 drivers/pci/probe.c          |    4 +---
 3 files changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kernel/pci_64.c +2 -0
index 249cca2..b2c55ca 100644
@@ -367,8 +367,10 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
 	sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 		dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
 	dev->class = get_int_prop(node, "class-code", 0);
+	dev->revision = get_int_prop(node, "revision-id", 0);
 
 	DBG("    class: 0x%x\n", dev->class);
+	DBG("    revision: 0x%x\n", dev->revision);
 
 	dev->current_state = 4;		/* unknown power state */
 	dev->error_state = pci_channel_io_normal;
arch/sparc64/kernel/pci.c +1 -0
index 38a32bc..759cc20 100644
@@ -448,6 +448,7 @@ struct pci_dev *of_create_pci_dev(struct pci_pbm_info *pbm,
 		 */
 		pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
 		dev->class = class >> 8;
+		dev->revision = class & 0xff;
 
 		sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 			dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
drivers/pci/probe.c +1 -3
index 0fdb71d..a574b7f 100644
@@ -702,6 +702,7 @@ static int pci_setup_device(struct pci_dev * dev)
 		dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn));
 
 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
+	dev->revision = class & 0xff;
 	class >>= 8;				    /* upper 3 bytes */
 	dev->class = class;
 	class >>= 8;
@@ -918,9 +919,6 @@ pci_scan_device(struct pci_bus *bus, int devfn)
 	dev->cfg_size = pci_cfg_space_size(dev);
 	dev->error_state = pci_channel_io_normal;
 
-	/* read the PCI revision: 1 byte */
-	pci_read_config_byte(dev, PCI_REVISION_ID, &dev->revision);
-
 	/* Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
 	   set this higher, assuming the system even supports it.  */
 	dev->dma_mask = 0xffffffff;
author
#2 lore raw
Quoted 67 lines · Grant Grundler
Quoted 5 lines · Auke Kok

Since the PCI bus class and revision are in the same dword, we can fill in both in the same read.

Other non-x86 arches use different methods to fill in these values, make sure to adjust those too.

Yeah, this looks good to me.

I didn't know ppc and sparc had their own methods for device discovery.

thanks,
grant

Quoted 51 lines

---

 arch/powerpc/kernel/pci_64.c |    2 ++
 arch/sparc64/kernel/pci.c    |    1 +
 drivers/pci/probe.c          |    4 +---
 3 files changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kernel/pci_64.c +2 -0
index 249cca2..b2c55ca 100644
@@ -367,8 +367,10 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
 	sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 		dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
 	dev->class = get_int_prop(node, "class-code", 0);
+	dev->revision = get_int_prop(node, "revision-id", 0);
 
 	DBG("    class: 0x%x\n", dev->class);
+	DBG("    revision: 0x%x\n", dev->revision);
 
 	dev->current_state = 4;		/* unknown power state */
 	dev->error_state = pci_channel_io_normal;
arch/sparc64/kernel/pci.c +1 -0
index 38a32bc..759cc20 100644
@@ -448,6 +448,7 @@ struct pci_dev *of_create_pci_dev(struct pci_pbm_info *pbm,
 		 */
 		pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
 		dev->class = class >> 8;
+		dev->revision = class & 0xff;
 
 		sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 			dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
drivers/pci/probe.c +1 -0
index 0fdb71d..a574b7f 100644
@@ -702,6 +702,7 @@ static int pci_setup_device(struct pci_dev * dev)
 		dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn));
 
 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
+	dev->revision = class & 0xff;
 	class >>= 8;				    /* upper 3 bytes */
 	dev->class = class;
 	class >>= 8;
@@ -918,9 +919,6 @@ pci_scan_device(struct pci_bus *bus, int devfn)
 	dev->cfg_size = pci_cfg_space_size(dev);
 	dev->error_state = pci_channel_io_normal;
 
Quoted 6 lines

drivers/pci/probe.c pci_scan_device(struct pci_bus *bus, int devfn) · line 924

Quoted 6 lines
drivers/pci/probe.c +0 -3
-	/* read the PCI revision: 1 byte */
-	pci_read_config_byte(dev, PCI_REVISION_ID, &dev->revision);
-
 	/* Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
 	   set this higher, assuming the system even supports it.  */
 	dev->dma_mask = 0xffffffff;

Greg, did you end up pulling this patch? If you pulled my first patch, then you *must* pull this one too. Otherwise a lot of drivers will fail on ppc/sparc64.

Auke

#3 lore raw
Quoted 64 lines · Auke
Quoted 62 lines · Grant Grundler
Quoted 5 lines · Auke Kok

Since the PCI bus class and revision are in the same dword, we can fill in both in the same read.

Other non-x86 arches use different methods to fill in these values, make sure to adjust those too.

Yeah, this looks good to me.
I didn't know ppc and sparc had their own methods for device discovery.

thanks,
grant

Quoted 51 lines

---

 arch/powerpc/kernel/pci_64.c |    2 ++
 arch/sparc64/kernel/pci.c    |    1 +
 drivers/pci/probe.c          |    4 +---
 3 files changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kernel/pci_64.c +2 -0
index 249cca2..b2c55ca 100644
@@ -367,8 +367,10 @@ struct pci_dev *of_create_pci_dev(struct device_node 
 *node,
 	sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 		dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
 	dev->class = get_int_prop(node, "class-code", 0);
+	dev->revision = get_int_prop(node, "revision-id", 0);
  	DBG("    class: 0x%x\n", dev->class);
+	DBG("    revision: 0x%x\n", dev->revision);
  	dev->current_state = 4;		/* unknown power state */
 	dev->error_state = pci_channel_io_normal;
arch/sparc64/kernel/pci.c +1 -0
index 38a32bc..759cc20 100644
@@ -448,6 +448,7 @@ struct pci_dev *of_create_pci_dev(struct pci_pbm_info 
 *pbm,
 		 */
 		pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
 		dev->class = class >> 8;
+		dev->revision = class & 0xff;
  		sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 			dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
drivers/pci/probe.c +1 -2
index 0fdb71d..a574b7f 100644
@@ -702,6 +702,7 @@ static int pci_setup_device(struct pci_dev * dev)
 		dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn));
  	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
+	dev->revision = class & 0xff;
 	class >>= 8;				    /* upper 3 bytes */
 	dev->class = class;
 	class >>= 8;
@@ -918,9 +919,6 @@ pci_scan_device(struct pci_bus *bus, int devfn)
 	dev->cfg_size = pci_cfg_space_size(dev);
 	dev->error_state = pci_channel_io_normal;
 -	/* read the PCI revision: 1 byte */
-	pci_read_config_byte(dev, PCI_REVISION_ID, &dev->revision);
-
Quoted 8 lines
Quoted 3 lines

drivers/pci/probe.c pci_scan_device(struct pci_bus *bus, int devfn) · line 924

Quoted 3 lines
drivers/pci/probe.c +0 -0
 	/* Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
 	   set this higher, assuming the system even supports it.  */
 	dev->dma_mask = 0xffffffff;

Greg, did you end up pulling this patch? If you pulled my first patch, then you *must* pull this one too. Otherwise a lot of drivers will fail on ppc/sparc64.

Care to merge this with your first one and resend it so that no one doing a 'git bisect' will hit a bad point in the tree?

If you do that, a resend of all of the patches in this series would be wise, to make sure that I didn't mess up in applying them.

thanks,

greg k-h

#4 lore raw
In reply to Auke Kok
Quoted 19 lines · Auke Kok

Since the PCI bus class and revision are in the same dword, we can fill in both in the same read.

Other non-x86 arches use different methods to fill in these values, make sure to adjust those too.

---

 arch/powerpc/kernel/pci_64.c |    2 ++
 arch/sparc64/kernel/pci.c    |    1 +
 drivers/pci/probe.c          |    4 +---
 3 files changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kernel/pci_64.c +0 -0
index 249cca2..b2c55ca 100644

arch/powerpc/kernel/pci_64.c struct pci_dev *of_create_pci_dev(struct device_node *node, · line 370

Quoted 4 lines
arch/powerpc/kernel/pci_64.c +1 -0
@@ -367,8 +367,10 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
 	sprintf(pci_name(dev), "%04x:%02x:%02x.%d", pci_domain_nr(bus),
 		dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
 	dev->class = get_int_prop(node, "class-code", 0);
+	dev->revision = get_int_prop(node, "revision-id", 0);

It's not clear to me in the spec if nodes are required to have the "revision-id" property. I guess 0 is a safe default though.

cheers

Signature · 8 lines
Michael Ellerman
OzLabs, IBM Australia Development Lab

wwweb: http://michael.ellerman.id.au
phone: +61 2 6212 1183 (tie line 70 21183)

We do not inherit the earth from our ancestors,
we borrow it from our children. - S.M.A.R.T Person
#5 lore raw
Quoted 4 lines

arch/powerpc/kernel/pci_64.c struct pci_dev *of_create_pci_dev(struct device_node *node, · line 370

Quoted 1 line
arch/powerpc/kernel/pci_64.c +1 -0
+	dev->revision = get_int_prop(node, "revision-id", 0);

It's not clear to me in the spec if nodes are required to have the "revision-id" property.

It is required for every PCI node.

Quoted 1 line

I guess 0 is a safe default though.

As good as any -- that is, no default is save. 0 should be "good enough" here I guess.

Segher

#6 lore raw
Quoted 6 lines · Segher Boessenkool
Quoted 4 lines

arch/powerpc/kernel/pci_64.c struct pci_dev *of_create_pci_dev(struct device_node *node, · line 370

Quoted 1 line
arch/powerpc/kernel/pci_64.c +1 -0
+	dev->revision = get_int_prop(node, "revision-id", 0);

It's not clear to me in the spec if nodes are required to have the "revision-id" property.

It is required for every PCI node.

Yep. I was reading the wrong spec :)

cheers

Signature · 8 lines
Michael Ellerman
OzLabs, IBM Australia Development Lab

wwweb: http://michael.ellerman.id.au
phone: +61 2 6212 1183 (tie line 70 21183)

We do not inherit the earth from our ancestors,
we borrow it from our children. - S.M.A.R.T Person