fix ROM enable/disable in r128 and radeon fb drivers

PATCH In review Jesse Barnes 2 participants 4 messages Dec 14, 2004 – Dec 23, 2004 Framebuffer layer

Share & tools

Jesse Barnes

author

2004-12-14 20:56 UTC #0 lore raw

PATCH 1/1 fix ROM enable/disable in r128 and radeon fb drivers

My mail to the maintainer bounced, but these should be safe for everybody. Both the r128 and radeon fb drivers do bad things with the PCI BAR corresponding to their option ROM. They incorrectly use the host address instead of the BAR address to enable the ROM, and then incorrectly lose the original value on unmap. This patch fixes both problems. Tested on Altix.

Signed-off-by Jesse Barnes <[email protected]>

Thanks,
Jesse

===== drivers/video/aty/aty128fb.c 1.52 vs edited =====

drivers/video/aty/aty128fb.c +7 -3

@@ -791,6 +791,7 @@
 static void __init aty128_unmap_ROM(struct pci_dev *dev, void __iomem * rom)
 {
 	struct resource *r = &dev->resource[PCI_ROM_RESOURCE];
+	u32 val;
 	
 	iounmap(rom);

@@ -801,8 +802,9 @@
 		r->end -= r->start;
 		r->start = 0;
 	}
-	/* This will disable and set address to unassigned */
-	pci_write_config_dword(dev, dev->rom_base_reg, 0);
+	/* This will disable it again */
+	pci_read_config_dword(dev, dev->rom_base_reg, &val);
+	pci_write_config_dword(dev, dev->rom_base_reg, val & ~PCI_ROM_ADDRESS_ENABLE);
 }

@@ -830,8 +832,10 @@
 	
 	/* enable if needed */
 	if (!(r->flags & PCI_ROM_ADDRESS_ENABLE)) {
+		u32 val;
+		pci_read_config_dword(dev, dev->rom_base_reg, &val);
 		pci_write_config_dword(dev, dev->rom_base_reg,
-				       r->start | PCI_ROM_ADDRESS_ENABLE);
+				       val | PCI_ROM_ADDRESS_ENABLE);
 		r->flags |= PCI_ROM_ADDRESS_ENABLE;
 	}

===== drivers/video/aty/radeon_base.c 1.35 vs edited =====

drivers/video/aty/radeon_base.c +7 -3

@@ -265,6 +265,7 @@
 {
 	// leave it disabled and unassigned
 	struct resource *r = &dev->resource[PCI_ROM_RESOURCE];
+	u32 val;
 	
 	if (!rinfo->bios_seg)
 		return;

@@ -277,8 +278,9 @@
 		r->end -= r->start;
 		r->start = 0;
 	}
-	/* This will disable and set address to unassigned */
-	pci_write_config_dword(dev, dev->rom_base_reg, 0);
+	/* This will disable it again */
+	pci_read_config_dword(dev, dev->rom_base_reg, &val);
+	pci_write_config_dword(dev, dev->rom_base_reg, val & ~PCI_ROM_ADDRESS_ENABLE);
 }
 
 static int __devinit radeon_map_ROM(struct radeonfb_info *rinfo, struct pci_dev *dev)

@@ -310,8 +312,10 @@
 	
 	/* enable if needed */
 	if (!(r->flags & PCI_ROM_ADDRESS_ENABLE)) {
+		u32 val;
+		pci_read_config_dword(dev, dev->rom_base_reg, &val);
 		pci_write_config_dword(dev, dev->rom_base_reg,
-				       r->start | PCI_ROM_ADDRESS_ENABLE);
+				       val | PCI_ROM_ADDRESS_ENABLE);
 		r->flags |= PCI_ROM_ADDRESS_ENABLE;
 	}

Jon Smirl

2004-12-14 21:46 UTC #1 lore raw

These drivers should be using the new ROM API in the PCI driver instead of manipulating the ROMs directly. Now that the ROM API is in the kernel all direct use of PCI_ENABLE_ROM should be removed. There are about thirty places in the kernel doing direct access. Kernel janitors would probably be a good place to track removing PCI_ENABLE_ROM.

I'm tied up taking care of premature baby twins so it is going to be quite a while until I can work on things like this.

Signature · 2 lines

Jon Smirl
[email protected]

Jesse Barnes

author

2004-12-14 22:23 UTC #2 lore raw

Quoted 6 lines · Jon Smirl

These drivers should be using the new ROM API in the PCI driver instead of manipulating the ROMs directly. Now that the ROM API is in the kernel all direct use of PCI_ENABLE_ROM should be removed. There are about thirty places in the kernel doing direct access. Kernel janitors would probably be a good place to track removing PCI_ENABLE_ROM.

Sure... but in the meantime the drivers should probably be trivially fixed like this so things don't break.

Thanks,
Jesse

Jesse Barnes

author

2004-12-23 18:16 UTC #3 lore raw

Quoted 10 lines · Jesse Barnes

Quoted 6 lines · Jon Smirl
These drivers should be using the new ROM API in the PCI driver instead of manipulating the ROMs directly. Now that the ROM API is in the kernel all direct use of PCI_ENABLE_ROM should be removed. There are about thirty places in the kernel doing direct access. Kernel janitors would probably be a good place to track removing PCI_ENABLE_ROM.
Sure... but in the meantime the drivers should probably be trivially fixed like this so things don't break.

Oops, didn't realize that changing them to use pci_rom_enable/disable would be just as trivial as my fix. Tested on sn2.

This patch fixes a few errors in the aty fb drivers by making them use pci_rom_enable/disable instead of manipulating BARs directly. They were both
a) incorrectly programming resource values into the BARs instead of actual BAR values (resources are cookies and shouldn't be used that way), and b) they were both clobbering the original value of the BAR corresponding to their ROM, also a no-no. Switching to pci_rom_enable/disable fixes both at once and kills some code.

Signed-off-by Jesse Barnes <[email protected]>

Thanks,
Jesse

===== drivers/video/aty/aty128fb.c 1.53 vs edited =====

drivers/video/aty/aty128fb.c +5 ~~-37~~

@@ -452,7 +452,6 @@
 static void __init aty128_get_pllinfo(struct aty128fb_par *par,
 				      void __iomem *bios);
 static void __init __iomem *aty128_map_ROM(struct pci_dev *pdev, const struct aty128fb_par *par);
-static void __init aty128_unmap_ROM(struct pci_dev *dev, void __iomem * rom);
 #endif
 static void aty128_timings(struct aty128fb_par *par);
 static void aty128_init_engine(struct aty128fb_par *par);

@@ -788,30 +787,12 @@
 
 
 #ifndef __sparc__
-static void __init aty128_unmap_ROM(struct pci_dev *dev, void __iomem * rom)
-{
-	struct resource *r = &dev->resource[PCI_ROM_RESOURCE];
-	
-	iounmap(rom);
-	
-	/* Release the ROM resource if we used it in the first place */
-	if (r->parent && r->flags & PCI_ROM_ADDRESS_ENABLE) {
-		release_resource(r);
-		r->flags &= ~PCI_ROM_ADDRESS_ENABLE;
-		r->end -= r->start;
-		r->start = 0;
-	}
-	/* This will disable and set address to unassigned */
-	pci_write_config_dword(dev, dev->rom_base_reg, 0);
-}
-
-
 static void __iomem * __init aty128_map_ROM(const struct aty128fb_par *par, struct pci_dev *dev)
 {
-	struct resource *r;
 	u16 dptr;
 	u8 rom_type;
 	void __iomem *bios;
+	size_t rom_size;
 
     	/* Fix from ATI for problem with Rage128 hardware not leaving ROM enabled */
     	unsigned int temp;

@@ -821,26 +802,13 @@
 	aty_st_le32(RAGE128_MPP_TB_CONFIG, temp);
 	temp = aty_ld_le32(RAGE128_MPP_TB_CONFIG);
 
-	/* no need to search for the ROM, just ask the card where it is. */
-	r = &dev->resource[PCI_ROM_RESOURCE];
+	bios = pci_map_rom(dev, &rom_size);
 
-	/* assign the ROM an address if it doesn't have one */
-	if (r->parent == NULL)
-		pci_assign_resource(dev, PCI_ROM_RESOURCE);
-	
-	/* enable if needed */
-	if (!(r->flags & PCI_ROM_ADDRESS_ENABLE)) {
-		pci_write_config_dword(dev, dev->rom_base_reg,
-				       r->start | PCI_ROM_ADDRESS_ENABLE);
-		r->flags |= PCI_ROM_ADDRESS_ENABLE;
-	}
-	
-	bios = ioremap(r->start, r->end - r->start + 1);
 	if (!bios) {
 		printk(KERN_ERR "aty128fb: ROM failed to map\n");
 		return NULL;
 	}
-	
+
 	/* Very simple test to make sure it appeared */
 	if (BIOS_IN16(0) != 0xaa55) {
 		printk(KERN_ERR "aty128fb: Invalid ROM signature %x should be 0xaa55\n",

@@ -899,7 +867,7 @@
 	return bios;
 
  failed:
-	aty128_unmap_ROM(dev, bios);
+	pci_unmap_rom(dev, bios);
 	return NULL;
 }

@@ -1959,7 +1927,7 @@
 	else {
 		printk(KERN_INFO "aty128fb: Rage128 BIOS located\n");
 		aty128_get_pllinfo(par, bios);
-		aty128_unmap_ROM(pdev, bios);
+		pci_unmap_rom(pdev, bios);
 	}
 #endif /* __sparc__ */

===== drivers/video/aty/radeon_base.c 1.35 vs edited =====

drivers/video/aty/radeon_base.c +3 ~~-30~~

@@ -263,30 +263,17 @@
 
 static void __devexit radeon_unmap_ROM(struct radeonfb_info *rinfo, struct pci_dev *dev)
 {
-	// leave it disabled and unassigned
-	struct resource *r = &dev->resource[PCI_ROM_RESOURCE];
-	
 	if (!rinfo->bios_seg)
 		return;
-	iounmap(rinfo->bios_seg);
-	
-	/* Release the ROM resource if we used it in the first place */
-	if (r->parent && r->flags & PCI_ROM_ADDRESS_ENABLE) {
-		release_resource(r);
-		r->flags &= ~PCI_ROM_ADDRESS_ENABLE;
-		r->end -= r->start;
-		r->start = 0;
-	}
-	/* This will disable and set address to unassigned */
-	pci_write_config_dword(dev, dev->rom_base_reg, 0);
+	pci_unmap_rom(dev, rinfo->bios_seg);
 }
 
 static int __devinit radeon_map_ROM(struct radeonfb_info *rinfo, struct pci_dev *dev)
 {
 	void __iomem *rom;
-	struct resource *r;
 	u16 dptr;
 	u8 rom_type;
+	size_t rom_size;
 
 	/* If this is a primary card, there is a shadow copy of the
 	 * ROM somewhere in the first meg. We will just ignore the copy

@@ -301,21 +288,7 @@
 	OUTREG(MPP_TB_CONFIG, temp);
 	temp = INREG(MPP_TB_CONFIG);
                                                                                                           
-	/* no need to search for the ROM, just ask the card where it is. */
-	r = &dev->resource[PCI_ROM_RESOURCE];
-	
-	/* assign the ROM an address if it doesn't have one */
-	if (r->parent == NULL)
-		pci_assign_resource(dev, PCI_ROM_RESOURCE);
-	
-	/* enable if needed */
-	if (!(r->flags & PCI_ROM_ADDRESS_ENABLE)) {
-		pci_write_config_dword(dev, dev->rom_base_reg,
-				       r->start | PCI_ROM_ADDRESS_ENABLE);
-		r->flags |= PCI_ROM_ADDRESS_ENABLE;
-	}
-	
-	rom = ioremap(r->start, r->end - r->start + 1);
+	rom = pci_map_rom(dev, &rom_size);
 	if (!rom) {
 		printk(KERN_ERR "radeonfb: ROM failed to map\n");
 		return -ENOMEM;