diff mbox series

core/pci: Document some stuff

Message ID 20180409040149.9625-1-oohall@gmail.com
State Accepted
Headers show
Series core/pci: Document some stuff | expand

Commit Message

Oliver O'Halloran April 9, 2018, 4:01 a.m. UTC 
Document the bridge class code hack and what ibm,pci-config-space-type
actually means. Also replace some of the pci_cap() calls with a variable
to make it a bit more readable.

Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
---
 core/pci.c | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

Comments

Stewart Smith April 10, 2018, 6:36 a.m. UTC | #1 
Oliver O'Halloran <oohall@gmail.com> writes:
> Document the bridge class code hack and what ibm,pci-config-space-type
> actually means. Also replace some of the pci_cap() calls with a variable
> to make it a bit more readable.
>
> Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
> ---
>  core/pci.c | 22 +++++++++++++---------
>  1 file changed, 13 insertions(+), 9 deletions(-)

Merged as of e5fb3b6d17fb74e11259d6f2287a005278bb1932

If only I had a MAINTAINERS file to troll you with by adding you as "pci
docs maintainer" :)
diff mbox series

Patch

diff --git a/core/pci.c b/core/pci.c
index 862f4a6d6517..dab7586f875a 100644
--- a/core/pci.c
+++ b/core/pci.c
@@ -1552,6 +1552,7 @@  static void pci_add_one_device_node(struct phb *phb,
 	uint32_t rev_class, vdid;
 	uint32_t reg[5];
 	uint8_t intpin;
+	bool is_pcie;
 	const uint32_t ranges_direct[] = {
 				/* 64-bit direct mapping. We know the bridges
 				 * don't cover the entire address space so
@@ -1563,14 +1564,14 @@  static void pci_add_one_device_node(struct phb *phb,
 	pci_cfg_read32(phb, pd->bdfn, 0, &vdid);
 	pci_cfg_read32(phb, pd->bdfn, PCI_CFG_REV_ID, &rev_class);
 	pci_cfg_read8(phb, pd->bdfn, PCI_CFG_INT_PIN, &intpin);
+	is_pcie = pci_has_cap(pd, PCI_CFG_CAP_ID_EXP, false);
 
 	/*
-	 * Quirk for IBM bridge bogus class on PCIe root complex.
-	 * Without it, the PCI DN won't be created for its downstream
-	 * devices in Linux.
+	 * Some IBM PHBs (p7ioc?) have an invalid PCI class code. Linux
+	 * uses prefers to read the class code from the DT rather than
+	 * re-reading config space we can hack around it here.
 	 */
-	if (pci_has_cap(pd, PCI_CFG_CAP_ID_EXP, false) &&
-	    parent_node == phb->dt_node)
+	if (is_pcie && parent_node == phb->dt_node)
 		rev_class = (rev_class & 0xff) | 0x6040000;
 	cname = pci_class_name(rev_class >> 8);
 
@@ -1582,9 +1583,12 @@  static void pci_add_one_device_node(struct phb *phb,
 			 cname, (pd->bdfn >> 3) & 0x1f);
 	pd->dn = np = dt_new(parent_node, name);
 
-	/* XXX FIXME: make proper "compatible" properties */
-	if (pci_has_cap(pd, PCI_CFG_CAP_ID_EXP, false) ||
-		phb->phb_type == phb_type_npu_v2_opencapi) {
+	/*
+	 * NB: ibm,pci-config-space-type is the PAPR way of indicating the
+	 * device has a 4KB config space. It's got nothing to do with the
+	 * standard Type 0/1 config spaces defined by PCI.
+	 */
+	if (is_pcie || phb->phb_type == phb_type_npu_v2_opencapi) {
 		snprintf(compat, MAX_NAME, "pciex%x,%x",
 			 vdid & 0xffff, vdid >> 16);
 		dt_add_property_cells(np, "ibm,pci-config-space-type", 1);
@@ -1637,7 +1641,7 @@  static void pci_add_one_device_node(struct phb *phb,
 	dt_add_property_cells(np, "#interrupt-cells", 1);
 
 	/* We want "device_type" for bridges */
-	if (pci_has_cap(pd, PCI_CFG_CAP_ID_EXP, false))
+	if (is_pcie)
 		dt_add_property_string(np, "device_type", "pciex");
 	else
 		dt_add_property_string(np, "device_type", "pci");