On 2015/3/25 7:58, Daniel Axtens wrote:

On Tue, 2015-03-24 at 11:34 +0800, Yijing Wang wrote:

quoted

Now we could use pci_scan_host_bridge() to scan
pci buses, provide powerpc specific pci_host_bridge_ops.

Signed-off-by: Yijing Wang <redacted>
CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: linuxppc-dev@lists.ozlabs.org
---
 arch/powerpc/kernel/pci-common.c |   60 +++++++++++++++++++++++--------------
 1 files changed, 37 insertions(+), 23 deletions(-)

quoted

diff --git a/arch/powerpc/kernel/pci-common.c b/arch/powerpc/kernel/pci-common.c
index 2c58200..e2b50a2 100644
--- a/arch/powerpc/kernel/pci-common.c
+++ b/arch/powerpc/kernel/pci-common.c

@@ -773,6 +773,29 @@ void pcibios_set_root_bus_speed(struct pci_host_bridge *bridge)
 		return ppc_md.pcibios_set_root_bus_speed(bridge);
 }
 
+static int pci_host_scan_bus(struct pci_host_bridge *host)
+{
+	int mode = PCI_PROBE_NORMAL;
+	struct pci_bus *bus = host->bus;
+	struct pci_controller *hose = dev_get_drvdata(&host->dev);

Is there any reason this isn't *hose = pci_bus_to_host(bus)?

Hi Daniel, thanks for your review and comments. We want to make a generic pci_host_bridge,
which would hold the common host information, for example, pci domain is common info for
pci host bridge, this series saved domain in pci_host_bridge, then we no need to
extract out domain by pci_bus->sysdata by platform specific pci_domain_nr().
Also we store the sysdata in pci_host_bridge, and pci_bus_to_host() is the platform
interface, I think use the common interface would be better.

quoted

+
+	/* Get probe mode and perform scan */
+	if (hose->dn && ppc_md.pci_probe_mode)
+		mode = ppc_md.pci_probe_mode(bus);
+
+	pr_debug("    probe mode: %d\n", mode);
+	if (mode == PCI_PROBE_DEVTREE)
+		of_scan_bus(hose->dn, bus);
+
+	if (mode == PCI_PROBE_NORMAL) {
+		pci_bus_update_busn_res_end(bus, 255);
+		hose->last_busno = pci_scan_child_bus(bus);
+		pci_bus_update_busn_res_end(bus, hose->last_busno);
+	}
+
+	return pci_bus_child_max_busnr(bus);
+}
+

I'm having trouble convincing myself that this patch covers every
variation within our PCI implementations. In particular, there's a
stanza in of_scan_pci_bridge in kernel/pci_of_scan.c that's almost
identical to this function. Does that implementation need to be cleaned
up and replaced with this function too?

This is a pci_host_bridge_ops hook function, which would be called in
PCI core, and after applied this series, we only need to call pci_scan_host_bridge()
to scan pci devices, and this function is also extracted from the pcibios_scan_phb(),
it's not the redundant code.

quoted

@@ -1641,9 +1655,9 @@ void pcibios_scan_phb(struct pci_controller *hose)
 		ppc_md.pcibios_fixup_phb(hose);
 
 	/* Configure PCI Express settings */
-	if (bus && !pci_has_flag(PCI_PROBE_ONLY)) {
+	if (host->bus && !pci_has_flag(PCI_PROBE_ONLY)) {
 		struct pci_bus *child;
-		list_for_each_entry(child, &bus->children, node)
+		list_for_each_entry(child, &host->bus->children, node)
 			pcie_bus_configure_settings(child);
 	}
 }

Two things: Firstly, the function uses hose throughout, not host.
Secondly, you're not deleting the bus variable: what's the purpose of
this change?

host is the common pci_host_bridge which is created by PCI core for pci host bridge driver,
the hose is the platform data used in powerpc. The purpose of the patch/series is to simplify
pci enumeration interface, and try to reduce the weak functions which were used to setup pci bus/devices
during PCI enumeration.

Regards,
Daniel

-- 
Thanks!
Yijing