| Message ID | BLU436-SMTP119C5643AF40BB3897C39CEBF870@phx.gbl |
|---|---|
| State | Accepted |
| Delegated to: | Simon Glass |
| Headers | show |
Hi Bin, On 18 July 2015 at 10:20, Bin Meng <bmeng.cn@gmail.com> wrote: > In driver model, each pci bridge device has its own hose structure. > hose->first_busno points to the bridge device's device number, so > we should not substract hose->first_busno before programming the > bridge device's primary/secondary/subordinate bus number registers. > > Signed-off-by: Bin Meng <bmeng.cn@gmail.com> > --- > > drivers/pci/pci_auto.c | 9 +++++++++ > 1 file changed, 9 insertions(+) Acked-by: Simon Glass <sjg@chromium.org> Tested-by: Simon Glass <sjg@chromium.org> I'm not happy with how each PCI bridge has its own hose structure. I think that was a mistake. It just creates problems. I'm thinking we should instead have a UCLASS_PCI which is strictly for the PCI controller, and a new UCLASS_PCI_BRIDGE for bridges. What do you think? Regards, Simon
Hi Simon, On Mon, Jul 20, 2015 at 9:59 AM, Simon Glass <sjg@chromium.org> wrote: > Hi Bin, > > On 18 July 2015 at 10:20, Bin Meng <bmeng.cn@gmail.com> wrote: >> In driver model, each pci bridge device has its own hose structure. >> hose->first_busno points to the bridge device's device number, so >> we should not substract hose->first_busno before programming the >> bridge device's primary/secondary/subordinate bus number registers. >> >> Signed-off-by: Bin Meng <bmeng.cn@gmail.com> >> --- >> >> drivers/pci/pci_auto.c | 9 +++++++++ >> 1 file changed, 9 insertions(+) > > Acked-by: Simon Glass <sjg@chromium.org> > Tested-by: Simon Glass <sjg@chromium.org> > > I'm not happy with how each PCI bridge has its own hose structure. I > think that was a mistake. It just creates problems. > > I'm thinking we should instead have a UCLASS_PCI which is strictly for > the PCI controller, and a new UCLASS_PCI_BRIDGE for bridges. > > What do you think? > I guess that may help. During the debug, I feel the dm pci codes are really not that intuitive than the previous non-dm version. It took me sometime to figure out where is the problem. Regards, Bin
On 19 July 2015 at 19:59, Simon Glass <sjg@chromium.org> wrote: > Hi Bin, > > On 18 July 2015 at 10:20, Bin Meng <bmeng.cn@gmail.com> wrote: >> In driver model, each pci bridge device has its own hose structure. >> hose->first_busno points to the bridge device's device number, so >> we should not substract hose->first_busno before programming the >> bridge device's primary/secondary/subordinate bus number registers. >> >> Signed-off-by: Bin Meng <bmeng.cn@gmail.com> >> --- >> >> drivers/pci/pci_auto.c | 9 +++++++++ >> 1 file changed, 9 insertions(+) > > Acked-by: Simon Glass <sjg@chromium.org> > Tested-by: Simon Glass <sjg@chromium.org> > Applied to u-boot-x86, thanks! [snip]
Hi Bin, On 19 July 2015 at 20:38, Bin Meng <bmeng.cn@gmail.com> wrote: > Hi Simon, > > On Mon, Jul 20, 2015 at 9:59 AM, Simon Glass <sjg@chromium.org> wrote: >> Hi Bin, >> >> On 18 July 2015 at 10:20, Bin Meng <bmeng.cn@gmail.com> wrote: >>> In driver model, each pci bridge device has its own hose structure. >>> hose->first_busno points to the bridge device's device number, so >>> we should not substract hose->first_busno before programming the >>> bridge device's primary/secondary/subordinate bus number registers. >>> >>> Signed-off-by: Bin Meng <bmeng.cn@gmail.com> >>> --- >>> >>> drivers/pci/pci_auto.c | 9 +++++++++ >>> 1 file changed, 9 insertions(+) >> >> Acked-by: Simon Glass <sjg@chromium.org> >> Tested-by: Simon Glass <sjg@chromium.org> >> >> I'm not happy with how each PCI bridge has its own hose structure. I >> think that was a mistake. It just creates problems. >> >> I'm thinking we should instead have a UCLASS_PCI which is strictly for >> the PCI controller, and a new UCLASS_PCI_BRIDGE for bridges. >> >> What do you think? >> > > I guess that may help. During the debug, I feel the dm pci codes are > really not that intuitive than the previous non-dm version. It took me > sometime to figure out where is the problem. Agreed. Part of the problem I think is that it is still compatible with the old code. We should be able to do things like pci_read_config16(struct udevice *dev, ...) I guess we could move to separate things more. Regards, Simon
diff --git a/drivers/pci/pci_auto.c b/drivers/pci/pci_auto.c index e034ed1..ef6dc4f 100644 --- a/drivers/pci/pci_auto.c +++ b/drivers/pci/pci_auto.c @@ -224,10 +224,15 @@ void pciauto_prescan_setup_bridge(struct pci_controller *hose, prefechable_64 &= PCI_PREF_RANGE_TYPE_MASK; /* Configure bus number registers */ +#ifdef CONFIG_DM_PCI + pci_hose_write_config_byte(hose, dev, PCI_PRIMARY_BUS, PCI_BUS(dev)); + pci_hose_write_config_byte(hose, dev, PCI_SECONDARY_BUS, sub_bus); +#else pci_hose_write_config_byte(hose, dev, PCI_PRIMARY_BUS, PCI_BUS(dev) - hose->first_busno); pci_hose_write_config_byte(hose, dev, PCI_SECONDARY_BUS, sub_bus - hose->first_busno); +#endif pci_hose_write_config_byte(hose, dev, PCI_SUBORDINATE_BUS, 0xff); if (pci_mem) { @@ -295,8 +300,12 @@ void pciauto_postscan_setup_bridge(struct pci_controller *hose, struct pci_region *pci_io = hose->pci_io; /* Configure bus number registers */ +#ifdef CONFIG_DM_PCI + pci_hose_write_config_byte(hose, dev, PCI_SUBORDINATE_BUS, sub_bus); +#else pci_hose_write_config_byte(hose, dev, PCI_SUBORDINATE_BUS, sub_bus - hose->first_busno); +#endif if (pci_mem) { /* Round memory allocator to 1MB boundary */
In driver model, each pci bridge device has its own hose structure. hose->first_busno points to the bridge device's device number, so we should not substract hose->first_busno before programming the bridge device's primary/secondary/subordinate bus number registers. Signed-off-by: Bin Meng <bmeng.cn@gmail.com> --- drivers/pci/pci_auto.c | 9 +++++++++ 1 file changed, 9 insertions(+)