Re: ali15x3 on alpha dies in 2.4.21-BK

Ivan Kokshaysky (ink@jurassic.park.msu.ru)
Fri, 6 Dec 2002 13:44:34 +0300


On Thu, Dec 05, 2002 at 09:20:01PM -0500, Jeff Garzik wrote:
> ali15x3: not 100% native mode, will probe irqs later
> ide0: bm-dma at 0x8420-0x8427, bios settings: hda:pio, hdb:pio
> ide1: bm-dma at 0x8428-0x842f, bios settings: hdc:pio, hdd:pio
> hdc:hdc: status timeout: status=0xff {busy}

Fixed in 2.5-ac two months ago... ;-)

Well, in addition to true legacy and native modes, the ALi chip supports
"intermediate" one which is utterly broken wrt PCI spec.
In this mode
BAR0 = 0x1f1
BAR1 = 0x3f5
BAR2 = 0x171
BAR3 = 0x375

and these BARs are *read-only*. This causes a lot of confusion:
a) sizes calculated by pci_read_bases() for these IO ranges are
incorrect;
b) generic resource allocation code assumes that these BARs
can be reassigned, and this actually happens on Alpha.

As there are other PCI devices with the same problem, I prefer
generic solution instead of individual quirk routines.

Here's the patch re-diffed vs. 2.4.

Ivan.

--- 2.4bk/drivers/pci/pci.c Fri Dec 6 11:07:52 2002
+++ linux/drivers/pci/pci.c Fri Dec 6 11:09:23 2002
@@ -1039,13 +1039,20 @@ static inline unsigned int pci_calc_reso
}

/*
- * Find the extent of a PCI decode..
+ * Find the extent of a PCI decode, do sanity checks.
*/
-static u32 pci_size(u32 base, unsigned long mask)
+static u32 pci_size(u32 base, u32 maxbase, unsigned long mask)
{
- u32 size = mask & base; /* Find the significant bits */
+ u32 size = mask & maxbase; /* Find the significant bits */
+ if (!size)
+ return 0;
size = size & ~(size-1); /* Get the lowest of them to find the decode size */
- return size-1; /* extent = size - 1 */
+ size -= 1; /* extent = size - 1 */
+ if (base == maxbase && ((base | size) & mask) != mask)
+ return 0; /* base == maxbase can be valid only
+ if the BAR has been already
+ programmed with all 1s */
+ return size;
}

static void pci_read_bases(struct pci_dev *dev, unsigned int howmany, int rom)
@@ -1068,13 +1075,17 @@ static void pci_read_bases(struct pci_de
if (l == 0xffffffff)
l = 0;
if ((l & PCI_BASE_ADDRESS_SPACE) == PCI_BASE_ADDRESS_SPACE_MEMORY) {
+ sz = pci_size(l, sz, PCI_BASE_ADDRESS_MEM_MASK);
+ if (!sz)
+ continue;
res->start = l & PCI_BASE_ADDRESS_MEM_MASK;
res->flags |= l & ~PCI_BASE_ADDRESS_MEM_MASK;
- sz = pci_size(sz, PCI_BASE_ADDRESS_MEM_MASK);
} else {
+ sz = pci_size(l, sz, PCI_BASE_ADDRESS_IO_MASK & 0xffff);
+ if (!sz)
+ continue;
res->start = l & PCI_BASE_ADDRESS_IO_MASK;
res->flags |= l & ~PCI_BASE_ADDRESS_IO_MASK;
- sz = pci_size(sz, PCI_BASE_ADDRESS_IO_MASK & 0xffff);
}
res->end = res->start + (unsigned long) sz;
res->flags |= pci_calc_resource_flags(l);
@@ -1104,6 +1115,7 @@ static void pci_read_bases(struct pci_de
if (rom) {
dev->rom_base_reg = rom;
res = &dev->resource[PCI_ROM_RESOURCE];
+ res->name = dev->name;
pci_read_config_dword(dev, rom, &l);
pci_write_config_dword(dev, rom, ~PCI_ROM_ADDRESS_ENABLE);
pci_read_config_dword(dev, rom, &sz);
@@ -1111,13 +1123,14 @@ static void pci_read_bases(struct pci_de
if (l == 0xffffffff)
l = 0;
if (sz && sz != 0xffffffff) {
+ sz = pci_size(l, sz, PCI_ROM_ADDRESS_MASK);
+ if (!sz)
+ return;
res->flags = (l & PCI_ROM_ADDRESS_ENABLE) |
IORESOURCE_MEM | IORESOURCE_PREFETCH | IORESOURCE_READONLY | IORESOURCE_CACHEABLE;
res->start = l & PCI_ROM_ADDRESS_MASK;
- sz = pci_size(sz, PCI_ROM_ADDRESS_MASK);
res->end = res->start + (unsigned long) sz;
}
- res->name = dev->name;
}
}

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/