Skip to contentSkip to conversation

viafb: use proper pci config API

PATCH In review Harald Welte 2 participants 3 messages May 19, 2009 – May 20, 2009 Framebuffer layer

lore
Share & tools
lore
#0 lore raw

PATCH 2/2 viafb: use proper pci config API

This patch alters viafb to use the proper Linux in-kernel API to access PCI configuration space, rather than poking at I/O ports by itself.

It also removes a minor inconsistency with naming of constants for the PCI device ID of the FUNCTION3 devices.

---
 drivers/video/via/hw.c |   64 +++++++++++++++++++++++++++++------------------
 drivers/video/via/hw.h |    4 +-
 2 files changed, 41 insertions(+), 27 deletions(-)
drivers/video/via/hw.c +39 -25
index 6306c88..40b7c0c 100644
@@ -2599,24 +2599,37 @@ static void disable_second_display_channel(void)
 	viafb_write_reg_mask(CR6A, VIACR, BIT6, BIT6);
 }
 
+static u_int16_t via_function3[] = {
+	CLE266_FUNCTION3, KM400_FUNCTION3, CN400_FUNCTION3, CN700_FUNCTION3,
+	CX700_FUNCTION3, KM800_FUNCTION3, KM890_FUNCTION3, P4M890_FUNCTION3,
+	P4M900_FUNCTION3, VX800_FUNCTION3, VX855_FUNCTION3,
+};
+
+/* Get the BIOS-configured framebuffer size from PCI configuration space
+ * of function 3 in the respective chipset */
 int viafb_get_fb_size_from_pci(void)
 {
-	unsigned long configid, deviceid, FBSize = 0;
-	int VideoMemSize;
-	int DeviceFound = false;
-
-	for (configid = 0x80000000; configid < 0x80010800; configid += 0x100) {
-		outl(configid, (unsigned long)0xCF8);
-		deviceid = (inl((unsigned long)0xCFC) >> 16) & 0xffff;
-
-		switch (deviceid) {
-		case CLE266:
-		case KM400:
-			outl(configid + 0xE0, (unsigned long)0xCF8);
-			FBSize = inl((unsigned long)0xCFC);
-			DeviceFound = true;	/* Found device id */
-			break;
+	int i;
+	u_int8_t offset = 0;
+	u_int32_t FBSize;
+	u_int32_t VideoMemSize;
+
+	/* search for the "FUNCTION3" device in this chipset */
+	for (i = 0; i < ARRAY_SIZE(via_function3); i++) {
+		struct pci_dev *pdev;
+
+		pdev = pci_get_device(PCI_VENDOR_ID_VIA, via_function3[i],
+				      NULL);
+		if (!pdev)
+			continue;
+
+		DEBUG_MSG(KERN_INFO "Device ID = %x\n", pdev->device);
 
+		switch (pdev->device) {
+		case CLE266_FUNCTION3:
+		case KM400_FUNCTION3:
+			offset = 0xE0;
+			break;
 		case CN400_FUNCTION3:
 		case CN700_FUNCTION3:
 		case CX700_FUNCTION3:
@@ -2626,21 +2639,22 @@ int viafb_get_fb_size_from_pci(void)
 		case P4M900_FUNCTION3:
 		case VX800_FUNCTION3:
 		case VX855_FUNCTION3:
-			/*case CN750_FUNCTION3: */
-			outl(configid + 0xA0, (unsigned long)0xCF8);
-			FBSize = inl((unsigned long)0xCFC);
-			DeviceFound = true;	/* Found device id */
-			break;
-
-		default:
+		/*case CN750_FUNCTION3: */
+			offset = 0xA0;
 			break;
 		}
-
-		if (DeviceFound)
+		
+		if (!offset)
 			break;
+
+		pci_read_config_dword(pdev, offset, &FBSize);
+		pci_dev_put(pdev);
 	}
 
-	DEBUG_MSG(KERN_INFO "Device ID = %lx\n", deviceid);
+	if (!offset) {
+		printk(KERN_ERR "cannot determine framebuffer size\n");
+		return -EIO;
+	}
 
 	FBSize = FBSize & 0x00007000;
 	DEBUG_MSG(KERN_INFO "FB Size = %x\n", FBSize);
drivers/video/via/hw.h +2 -2
index 1cd0648..a69e907 100644
@@ -850,8 +850,8 @@ struct iga2_crtc_timing {
 };
 
 /* device ID */
-#define CLE266              0x3123
-#define KM400               0x3205
+#define CLE266_FUNCTION3    0x3123
+#define KM400_FUNCTION3     0x3205
 #define CN400_FUNCTION2     0x2259
 #define CN400_FUNCTION3     0x3259
 /* support VT3314 chipset */
Signature · 9 lines
1.6.2.4


------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables 
unlimited royalty-free distribution of the report engine 
for externally facing server and web deployment. 
http://p.sf.net/sfu/businessobjects
#1 lore raw
Quoted 46 lines · On Tue, 19 May 2009 16:35:34 +0800 Harald Welte <[email protected]> wrote:

This patch alters viafb to use the proper Linux in-kernel API to access PCI configuration space, rather than poking at I/O ports by itself.

It also removes a minor inconsistency with naming of constants for the PCI device ID of the FUNCTION3 devices.

---
 drivers/video/via/hw.c |   64 +++++++++++++++++++++++++++++------------------
 drivers/video/via/hw.h |    4 +-
 2 files changed, 41 insertions(+), 27 deletions(-)
drivers/video/via/hw.c +9 -15
index 6306c88..40b7c0c 100644
@@ -2599,24 +2599,37 @@ static void disable_second_display_channel(void)
 	viafb_write_reg_mask(CR6A, VIACR, BIT6, BIT6);
 }
 
+static u_int16_t via_function3[] = {
+	CLE266_FUNCTION3, KM400_FUNCTION3, CN400_FUNCTION3, CN700_FUNCTION3,
+	CX700_FUNCTION3, KM800_FUNCTION3, KM890_FUNCTION3, P4M890_FUNCTION3,
+	P4M900_FUNCTION3, VX800_FUNCTION3, VX855_FUNCTION3,
+};
+
+/* Get the BIOS-configured framebuffer size from PCI configuration space
+ * of function 3 in the respective chipset */
 int viafb_get_fb_size_from_pci(void)
 {
-	unsigned long configid, deviceid, FBSize = 0;
-	int VideoMemSize;
-	int DeviceFound = false;
-
-	for (configid = 0x80000000; configid < 0x80010800; configid += 0x100) {
-		outl(configid, (unsigned long)0xCF8);
-		deviceid = (inl((unsigned long)0xCFC) >> 16) & 0xffff;
-
-		switch (deviceid) {
-		case CLE266:
-		case KM400:
-			outl(configid + 0xE0, (unsigned long)0xCF8);
-			FBSize = inl((unsigned long)0xCFC);
-			DeviceFound = true;	/* Found device id */
-			break;
+	int i;

drivers/video/via/hw.c static void disable_second_display_channel(void) · line 2622

Quoted 10 lines
drivers/video/via/hw.c +10 -0
+	u_int8_t offset = 0;
+	u_int32_t FBSize;
+	u_int32_t VideoMemSize;
+
+	/* search for the "FUNCTION3" device in this chipset */
+	for (i = 0; i < ARRAY_SIZE(via_function3); i++) {
+		struct pci_dev *pdev;
+
+		pdev = pci_get_device(PCI_VENDOR_ID_VIA, via_function3[i],
+				      NULL);

Can you pass the pdev pointer from the probe function here and skip the whole searching of the PCI device?

Regards,
Krzysztof

----------------------------------------------------------------------
Audi kilka tysiecy zlotych taniej? Przebieraj wsrod tysiecy ogloszen!
Sprawdz: http://link.interia.pl/f216f

------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment.
http://p.sf.net/sfu/businessobjects

#2 lore raw
Quoted 6 lines · Krzysztof Helt
Quoted 3 lines · On Tue, 19 May 2009 16:35:34 +0800 Harald Welte <[email protected]> wrote:

This patch alters viafb to use the proper Linux in-kernel API to access PCI configuration space, rather than poking at I/O ports by itself.

Quoted 7 lines
Quoted 4 lines

It also removes a minor inconsistency with naming of constants for the PCI device ID of the FUNCTION3 devices.

Can you pass the pdev pointer from the probe function here and skip the whole searching of the PCI device?

No. The FUNCTION3 device that we're searching is the SDRAM controller function integrated in the north bridge, not the integrated graphics.

Only the PCI config space of the SDRAM controller contains the information about the size of the graphics memory.

I was thinking of a 1:1 mapping between <pci device id for the graphics> : <pci device id for FUNCTION3>, but I'm not entirely sure if this is always correct or if there is some hardware that has the ID of one SDRAM controller and a different integrated graphics chipet, especially in legacy products.

So for that reason I decided not to change the logic for searching the FUNCTION3, merely altering the pci config space access function.

Regards,

Signature · 10 lines
- Harald Welte <[email protected]>	    http://linux.via.com.tw/
============================================================================
VIA Open Source Liaison

------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables 
unlimited royalty-free distribution of the report engine 
for externally facing server and web deployment. 
http://p.sf.net/sfu/businessobjects