On 06.05.2008 07:54, ron minnich wrote:
I know some people will get upset by the find_pci_device, but:
Indeed. Since we don't use the struct device passed in to ide_init, why don't we give Geode IDE its own struct device_operations and store the IDE config there?
This patch fixes a long-standing problem in the cs5536 driver, that is probably also in v2.
The ide_init is called with the sb device but needs the IDE device, which is different. It also did a write_config8,not write_config32; this bug has been in there since the code was written for v2. I don't know how or why it ever worked, since PWB is 1<<14.
Works fine with 2.6.24. 2.6.15 hangs on boot, but I don't think it was
^^^^^^ 2.6.25?
ever IDE. POST on the 2.6.25 boot hang is 00.
I have problems parsing the sentence above: "... I don't think it was ever IDE."
Signed-off-by: Ronald G. Minnich rminnich@gmail.com
Index: southbridge/amd/cs5536/cs5536.c
--- southbridge/amd/cs5536/cs5536.c (revision 676) +++ southbridge/amd/cs5536/cs5536.c (working copy) @@ -589,15 +589,18 @@ static void ide_init(struct device *dev) { u32 ide_cfg;
struct device *ide_dev;
printk(BIOS_DEBUG, "cs5536_ide: %s\n", __func__); /* GPIO and IRQ setup are handled in the main chipset code. */
ide_dev = dev_find_pci_device(PCI_VENDOR_ID_AMD,
PCI_DEVICE_ID_AMD_CS5536_B0_IDE, 0);// Enable the channel and Post Write Buffer // NOTE: Only 32-bit writes to the data buffer are allowed when PWB is set
- ide_cfg = pci_read_config32(dev, IDE_CFG);
- ide_cfg = pci_read_config32(ide_dev, IDE_CFG); ide_cfg |= CHANEN | PWB;
- pci_write_config8(dev, IDE_CFG, ide_cfg);
- pci_write_config32(ide_dev, IDE_CFG, ide_cfg);
}
You know my preference for killing dev_find_pci_device, but that can come later if it is too difficult.
Acked-by: Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net
Regards, Carl-Daniel