On Wed, 2010-03-24 at 08:48 +0100, Giampaolo Bellini wrote: > Hello Otavio and Bjorn > > here are ICOP's notes about PCI collision on Vortex86-DX: > > >See the attachment for Memory mapping of Vortex86SX/DX > >E segment will be occuiped by USB decice and our engineer > suggest you > >to :disable" USB device or you can use D segment to slove PCI > address collision > > I don't test this yet... but I don't think that disable USB could be a > good solution :-( Giampaolo, as far as I can tell, everything is working correctly on your machine, and your only complaint is that we print a KERN_ERR message that makes the system switch to text mode. We already do relocate the BARs with the collisions: pci 0000:00:0b.0: address space collision: [mem 0x000e0000-0x000e0fff] already in use pci 0000:00:0b.1: address space collision: [mem 0x000e1000-0x000e10ff] already in use pci 0000:00:0b.0: BAR 0: assigned [mem 0x20000000-0x20000fff] pci 0000:00:0b.1: BAR 0: assigned [mem 0x20001000-0x200010ff] and this relocation works fine, as far as I can tell. I think the collision is probably related to the fact that we get this from the BIOS: #1 [000009fc00 - 0000100000] BIOS reserved and then we find these PCI devices using that same region, but I don't know enough about the BIOS memory map to know whether it's fixable. I cc'd Yinghai, who knows more about that. Otavio's issue is completely different: his disk doesn't work unless he uses "irqpoll". Bjorn > 2010/3/22 Bjorn Helgaas <bjorn.helgaas@xxxxxx> > On Monday 22 March 2010 03:27:18 pm Otavio Salvador wrote: > > Hello Bjorn, > > > > On Mon, Mar 22, 2010 at 11:50 AM, Bjorn Helgaas > <bjorn.helgaas@xxxxxx> wrote: > > > On Monday 22 March 2010 08:41:03 am Otavio Salvador wrote: > > >> I have this same problem in VortexSX machine. > > > Can you post your entire dmesg log and a description of > what > > > problem you're seeing (e.g., device X doesn't work or > whatever)? > > > If this is a regression, i.e., your system used to work > and > > > now doesn't work, please tell us which version used to > work > > > and when it broke. > > > Since you don't mention any working version, I assume that > Linux > has never worked correctly on this system. > > > This has already been done at > http://article.gmane.org/gmane.linux.kernel/916237 > > > That post is about the fact that your disk doesn't work unless > you > use "irqpoll". That just means the interrupts don't appear > where > we expect them, which has nothing to do with the PCI address > space > collisions. > > I'm sorry I don't know enough about interrupts to help more. > If > I were debugging it, I would start by adding printks to the > pirq_enable_irq() path and see if I could figure out why it > works > for some devices, but for IDE. > > Bjorn > -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
