[linux-yocto] [PATCH 26/87] arm/dts: updating PCIe related files for TPAGE and BAR0
Paul Butler
butler.paul at gmail.com
Mon May 27 09:55:57 PDT 2013
From: SangeethaRao <sangeetha.rao at lsi.com>
Signed-off-by: SangeethaRao <sangeetha.rao at lsi.com>
---
arch/arm/boot/dts/axm55xxsim.dts | 15 +-
arch/arm/mach-axxia/pci.c | 505 ++++++++++++++++++++++-----------------
2 files changed, 294 insertions(+), 226 deletions(-)
diff --git a/arch/arm/boot/dts/axm55xxsim.dts b/arch/arm/boot/dts/axm55xxsim.dts
index 2147ceb..169aa82 100644
--- a/arch/arm/boot/dts/axm55xxsim.dts
+++ b/arch/arm/boot/dts/axm55xxsim.dts
@@ -300,16 +300,15 @@
0x20 0x20120000 0x0 0x00008000 >;
bus-range = <0 0x0f>;
/* Outbound ranges */
- /* < <3-cell PCI addr> <2-cell CPU (PLB) addr> <2-cell size> > *
-/
+ /* < <3-cell PCI addr> <2-cell CPU (PLB) addr> <2-cell size> >*/
ranges = <0x03000000 0x00000000 0xa0000000
0x30 0x00000000
0x00 0x10000000>;
/* Inbound ranges */
/* < <3-cell PCI addr> <2-cell CPU addr> <2-cell size> > */
- dma-ranges = <0x03000000 0x00000000 0xa0000000
+ dma-ranges = <0x03000000 0x00000000 0x00000000
0x00 0x00000000
- 0x00 0x10000000>;
+ 0x00 0x40000000>;
interrupt-parent = <&gic>;
interrupts = <29 2>;
interrupt-map-mask = <0000 0 0 7>;
@@ -343,9 +342,9 @@
0x00 0x10000000>;
/* Inbound ranges */
/* < <3-cell PCI addr> <2-cell CPU addr> <2-cell size> > */
- dma-ranges = <0x03000000 0x00000000 0xb0000000
+ dma-ranges = <0x03000000 0x00000000 0x00000000
0x00 0x00000000
- 0x00 0x10000000>;
+ 0x00 0x40000000>;
interrupt-parent = <&gic>;
interrupts = <72 2>;
interrupt-map-mask = <0000 0 0 7>;
@@ -379,9 +378,9 @@
0x00 0x10000000>;
/* Inbound ranges */
/* < <3-cell PCI addr> <2-cell CPU addr> <2-cell size> > */
- dma-ranges = <0x03000000 0x00000000 0xc0000000
+ dma-ranges = <0x03000000 0x00000000 0x00000000
0x00 0x00000000
- 0x00 0x10000000>;
+ 0x00 0x40000000>;
interrupt-parent = <&gic>;
interrupts = <73 2>;
diff --git a/arch/arm/mach-axxia/pci.c b/arch/arm/mach-axxia/pci.c
index b0f1a9a..ea9a7a5 100644
--- a/arch/arm/mach-axxia/pci.c
+++ b/arch/arm/mach-axxia/pci.c
@@ -17,7 +17,7 @@
#include <linux/of_irq.h>
#include <linux/interrupt.h>
-#include <asm/io.h>
+#include <linux/io.h>
#include <asm/sizes.h>
#include <asm/mach/pci.h>
@@ -54,7 +54,8 @@ struct axxia_pciex_port {
struct device_node *node;
struct resource utl_regs;
struct resource cfg_space;
- struct resource res;
+ struct resource outbound, inbound;
+ dma_addr_t msi_phys;
};
static struct axxia_pciex_port *axxia_pciex_ports;
@@ -75,7 +76,6 @@ fixup_axxia_pci_bridge(struct pci_dev *dev)
printk(KERN_INFO
"PCI: Setting PCI Class to PCI_CLASS_BRIDGE_HOST for %04x:%04x\n",
dev->vendor, dev->device);
-
dev->class = PCI_CLASS_BRIDGE_HOST << 8;
/*
* Make the bridge transparent
@@ -87,6 +87,7 @@ fixup_axxia_pci_bridge(struct pci_dev *dev)
DECLARE_PCI_FIXUP_HEADER(0x1000, 0x5101, fixup_axxia_pci_bridge);
DECLARE_PCI_FIXUP_HEADER(0x1000, 0x5108, fixup_axxia_pci_bridge);
+DECLARE_PCI_FIXUP_HEADER(0x1000, 0x5120, fixup_axxia_pci_bridge);
/* Convert to Bus# to PCIe port# */
static struct axxia_pciex_port *bus_to_port(struct pci_bus *bus)
@@ -95,8 +96,7 @@ static struct axxia_pciex_port *bus_to_port(struct pci_bus *bus)
}
/* Validate the Bus#/Device#/Function# */
-static int axxia_pciex_validate_bdf(struct pci_bus *bus,
- unsigned int devfn)
+static int axxia_pciex_validate_bdf(struct pci_bus *bus, unsigned int devfn)
{
struct axxia_pciex_port *port;
@@ -106,10 +106,10 @@ static int axxia_pciex_validate_bdf(struct pci_bus *bus,
if (port->endpoint)
return PCIBIOS_DEVICE_NOT_FOUND;
- if (((!((PCI_FUNC(devfn) == 0) && (PCI_SLOT(devfn) == 0))) &&
- (bus->number == port->root_bus_nr)) ||
- (!(PCI_SLOT(devfn) == 0) &&
- (bus->number == port->root_bus_nr+1))) {
+ if (((!((PCI_FUNC(devfn) == 0) && (PCI_SLOT(devfn) == 0)))
+ && (bus->number == port->root_bus_nr))
+ || (!(PCI_SLOT(devfn) == 0)
+ && (bus->number == port->root_bus_nr+1))) {
return PCIBIOS_DEVICE_NOT_FOUND;
}
return 0;
@@ -147,12 +147,15 @@ static void __iomem *axxia_pciex_get_config_base(struct axxia_pciex_port *port,
mpage |= (fn << 19);
if ((mpage != last_mpage) || (port->index != last_port)) {
- addr = ((u32)port->cfg_addr) +
- ACPX1_PCIE_MPAGE_UPPER(7);
+ addr = ((u32)port->cfg_addr)
+ + ACPX1_PCIE_MPAGE_UPPER(7);
writel(0x0, (u32 *) addr);
- addr = ((u32)port->cfg_addr) +
- ACPX1_PCIE_MPAGE_LOWER(7);
+ addr = ((u32)port->cfg_addr)
+ + ACPX1_PCIE_MPAGE_LOWER(7);
writel(mpage, (u32 *) addr);
+ /* printk("pcie_get_base: %02x:%02x:%02x setting
+ mpage = 0x%08x in addr = 0x%08x\n", bus->number,
+ dev, fn, mpage, addr);*/
last_mpage = mpage;
last_port = port->index;
}
@@ -198,10 +201,10 @@ arm_pciex_axxia_read_config(struct pci_bus *bus, unsigned int devfn,
* AXI address [3:0] is not used at all.
* AXI address[9:4] becomes register number.
* AXI address[13:10] becomes Ext. register number
- * AXI address[17:14] becomes 1st DWBE
- * for configuration read only.
- * AXI address[29:27] is used to select
- * one of 8 Mpage registers.
+ * AXI address[17:14] becomes 1st DWBE for configuration
+ * read only.
+ * AXI address[29:27] is used to select one of 8 Mpage
+ * registers.
*/
bus_addr = (u32) addr + (offset << 2);
bus_addr1 = bus_addr;
@@ -237,9 +240,9 @@ arm_pciex_axxia_read_config(struct pci_bus *bus, unsigned int devfn,
#ifdef PRINT_CONFIG_ACCESSES
printk(KERN_INFO
- "acp_read_config for PEI%d: %3d fn=0x%04x o=0x%04x l=%d a=0x%08x v=0x%08x, dev=%d\n",
- port->index, bus->number, devfn, offset,
- len, bus_addr, *val, PCI_SLOT(devfn));
+ "acp_read_config for PEI%d: %3d fn=0x%04x o=0x%04x l=%d a=0x%08x v=0x%08x, dev=%d\n",
+ port->index, bus->number, devfn, offset, len,
+ bus_addr, *val, PCI_SLOT(devfn));
#endif
return rc;
}
@@ -280,12 +283,13 @@ static int arm_pciex_axxia_write_config(struct pci_bus *bus, unsigned int devfn,
val32 = readl((u32 *)bus_addr);
if (len == 2) {
- val32 = (val32 & ~(0xffff << bs)) |
- ((val & 0xffff) << bs);
+ val32 = (val32 & ~(0xffff << bs))
+ | ((val & 0xffff) << bs);
} else {
- val32 = (val32 & ~(0xff << bs)) |
- ((val & 0xff) << bs);
+ val32 = (val32 & ~(0xff << bs))
+ | ((val & 0xff) << bs);
}
+
val = val32;
len = 4;
}
@@ -295,8 +299,8 @@ static int arm_pciex_axxia_write_config(struct pci_bus *bus, unsigned int devfn,
#ifdef PRINT_CONFIG_ACCESSES
printk(KERN_INFO
- "acp_write_config: bus=%3d devfn=0x%04x offset=0x%04x len=%d, addr=0x%08x val=0x%08x\n",
- bus->number, devfn, offset, len, bus_addr, val);
+ "acp_write_config: bus=%3d devfn=0x%04x offset=0x%04x len=%d addr=0x%08x val=0x%08x\n",
+ bus->number, devfn, offset, len, bus_addr, val);
#endif
switch (len) {
@@ -335,6 +339,7 @@ acp_pcie_isr(int irq, void *arg)
/* check if this is a PCIe message from an external device */
if (intr_status & 0x00000010) {
externalPciIntr = 1;
+
msg_fifo_stat = readl(mbase+0x10b4);
/* loop until the message fifo is empty */
@@ -342,59 +347,51 @@ acp_pcie_isr(int irq, void *arg)
u8 bus, dev, fn;
u8 msg_type;
msg_fifo_info = readl(mbase+0x10b0);
+
bus = (msg_fifo_info >> 16) & 0xff;
dev = (msg_fifo_info >> 11) & 0x1f;
fn = (msg_fifo_info >> 8) & 0x07;
msg_type = msg_fifo_info & 0xff;
/* print out the BDF and message type.
- * We ignore the common message types.
- */
+ * We ignore the common message types.
+ */
switch (msg_type) {
case 0x20: /* assert_INTa */
- printk(KERN_INFO
- "PEI%d --> INTa asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTa asserted\n",
+ port->index);
break;
case 0x21: /* assert_INTb */
- printk(KERN_INFO
- "PEI%d --> INTb asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTb asserted\n",
+ port->index);
break;
case 0x22: /* assert_INTc */
- printk(KERN_INFO
- "PEI%d --> INTc asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTc asserted\n",
+ port->index);
break;
case 0x23: /* assert_INTd */
- printk(KERN_INFO
- "PEI%d --> INTd asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTd asserted\n",
+ port->index);
break;
case 0x24: /* de-assert_INTa */
- printk(KERN_INFO
- "PEI%d --> INTa de-asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTa de-asserted\n",
+ port->index);
break;
case 0x25: /* de-assert_INTb */
- printk(KERN_INFO
- "PEI%d --> INTb de-asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTb de-asserted\n",
+ port->index);
break;
case 0x26: /* de-assert_INTc */
- printk(KERN_INFO
- "PEI%d --> INTc de-asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTc de-asserted\n",
+ port->index);
break;
case 0x27: /* de-assert_INTd */
- printk(KERN_INFO
- "PEI%d --> INTd de-asserted\n",
- port->index);
+ printk(KERN_INFO "PEI%d --> INTd de-asserted\n",
+ port->index);
break;
default:
- printk(KERN_INFO
- "BDF %02x:%02x.%x sent msgtype 0x%02x\n",
- bus, dev, fn, msg_type);
+ printk(KERN_INFO "BDF %02x:%02x.%x sent msgtype 0x%02x\n",
+ bus, dev, fn, msg_type);
break;
}
@@ -402,10 +399,8 @@ acp_pcie_isr(int irq, void *arg)
msg_fifo_stat = readl(mbase+0x10b4);
}
} else {
- /*
- Ignore the common interrupts, still need to figure
- out what they all mean.
- */
+ /* Ignore the common interrupts, still need to figure out what
+ * they all mean.*/
if (intr_status & 0xf3ffffab) {
u32 t2a_err_stat;
u32 t2a_other_err_stat;
@@ -414,19 +409,18 @@ acp_pcie_isr(int irq, void *arg)
u32 offset;
printk(KERN_ERR
- "ACP_PCIE_ISR: got PEI%d error interrupt 0x%08x\n",
- intr_status, port->index);
+ "ACP_PCIE_ISR: got PEI%d error interrupt 0x%08x\n",
+ intr_status, port->index);
linkStatus = readl(mbase+0x117c);
- printk(KERN_ERR
- "link_status (0x117c) = 0x%08x\n",
- linkStatus);
+ printk(KERN_ERR "link_status (0x117c) = 0x%08x\n",
+ linkStatus);
if (intr_status & 0x00020000) {
t2a_err_stat = readl(mbase+0x1170);
- printk(KERN_ERR
- "t2a_fn_indp_err_stat = 0x%08x\n",
- t2a_err_stat);
+ printk(KERN_ERR "t2a_fn_indp_err_stat = 0x%08x\n",
+ t2a_err_stat);
+
int_enb = readl(mbase+0x10c4);
int_enb &= 0xfffdffff;
writel(int_enb, mbase + 0x10c4);
@@ -435,51 +429,44 @@ acp_pcie_isr(int irq, void *arg)
if (intr_status & 0x00040000) {
t2a_other_err_stat = readl(mbase+0x1174);
printk(KERN_ERR "t2a_fn_indp_other_err_stat = 0x%08x\n",
- t2a_other_err_stat);
+ t2a_other_err_stat);
int_enb = readl(mbase+0x10c4);
int_enb &= 0xfffbffff;
writel(int_enb, mbase + 0x10c4);
}
if (intr_status & 0x00000800) {
- printk(KERN_INFO
- "pci_config = 0x%08x\n",
- readl(mbase + 0x1000));
- printk(KERN_INFO
- "pci_status = 0x%08x\n",
- readl(mbase + 0x1004));
-
+ printk(KERN_INFO "pci_config = 0x%08x\n",
+ readl(mbase + 0x1000));
+ printk(KERN_INFO "pci_status = 0x%08x\n",
+ readl(mbase + 0x1004));
int_enb = readl(mbase + 0x10c4);
int_enb &= 0xfffff7ff;
writel(int_enb, mbase + 0x10c4);
}
-
/*
- * dump all the potentially interesting PEI registers
- */
+ * dump all the potentially interesting PEI registers
+ */
for (offset = 0x114c; offset <= 0x1180; offset += 4) {
- printk(KERN_INFO
- " 0x%04x : 0x%08x\n",
- offset, readl(mbase + offset));
+ printk(KERN_INFO " 0x%04x : 0x%08x\n", offset,
+ readl(mbase + offset));
}
}
}
/*
- * We clear all the interrupts in the PEI status, even though
- * interrupts from external devices have not yet been handled.
- * That should be okay, since the PCI IRQ in the MPIC won't be
- * re-enabled until all external handlers have been called.
- */
+ * We clear all the interrupts in the PEI status, even though
+ * interrupts from external devices have not yet been handled.
+ * That should be okay, since the PCI IRQ in the MPIC won't be
+ * re-enabled until all external handlers have been called.
+ */
writel(intr_status, mbase + 0x10c0);
-
return externalPciIntr ? IRQ_NONE : IRQ_HANDLED;
}
/* ACP PCIe ISR to handle MSI Interrupts */
static irqreturn_t
-acp_pcie_MSI_isr(int irq, void *arg)
-{
+acp_pcie_MSI_isr(int irq, void *arg) {
u32 intr_status;
u8 msiIntr = 0, bit = 0;
struct axxia_pciex_port *port = (struct axxia_pciex_port *)arg;
@@ -494,26 +481,23 @@ acp_pcie_MSI_isr(int irq, void *arg)
msiIntr = 1;
for (bit = 0; bit < 16; bit++) {
if (intr_status & (0x1 << bit)) {
- printk(KERN_INFO
- "PEI%d --> MSI%d-%d interrupt asserted\n",
- port->index, bit*16, ((bit+1)*16)-1);
+ printk(KERN_INFO "PEI%d --> MSI%d-%d interrupt asserted\n",
+ port->index, bit*16, ((bit+1)*16)-1);
/* MSI Level 1 interrupt status */
statusReg = (0x123c + (0xc * bit));
statusVal = readl(mbase+statusReg);
- printk(KERN_INFO
- "MSI status Reg 0x%x val = 0x%x\n",
- statusReg, statusVal);
+ printk(KERN_INFO "MSI status Reg 0x%x val = 0x%x\n",
+ statusReg, statusVal);
/* clear statusReg */
writel(statusVal, mbase+statusReg);
}
}
}
/*
- * We clear all the interrupts in the PEI status, even though
- * interrupts from MSI devices have not yet been handled.
- */
+ * We clear all the interrupts in the PEI status, even though
+ * interrupts from MSI devices have not yet been handled.
+ */
writel(intr_status, mbase + 0x1230);
-
return msiIntr ? IRQ_NONE : IRQ_HANDLED;
}
@@ -523,21 +507,18 @@ int axxia_pcie_setup(int portno, struct pci_sys_data *sys)
{
struct axxia_pciex_port *port;
u32 pci_config, pci_status, link_state;
- int i, num_pages;
+ int i, num_pages, err;
u32 mpage_lower, pciah, pcial;
- u32 msiLower, msiHigher, msiFinal;
- u64 size, msiAddr;
- void __iomem *cfg_addr, *cfg_data;
+ u64 size, bar0_size;
+ void __iomem *cfg_addr, *cfg_data, *tpage_base;
int mappedIrq;
- int err;
+ u32 inbound_size;
port = &axxia_pciex_ports[sys->domain];
- printk(KERN_INFO
- "cfg_space start = 0x%012llx, end = 0x%012llx\n",
- port->cfg_space.start, port->cfg_space.end);
- printk(KERN_INFO
- "utl_regs start = 0x%012llx, end = 0x%012llx\n",
- port->utl_regs.start, port->utl_regs.end);
+ printk(KERN_INFO "cfg_space start = 0x%012llx, end = 0x%012llx\n",
+ port->cfg_space.start, port->cfg_space.end);
+ printk(KERN_INFO "utl_regs start = 0x%012llx, end = 0x%012llx\n",
+ port->utl_regs.start, port->utl_regs.end);
port->root_bus_nr = sys->busnr;
/* 1M external config */
@@ -550,21 +531,22 @@ int axxia_pcie_setup(int portno, struct pci_sys_data *sys)
port->cfg_data = cfg_data;
/* IORESOURCE_MEM */
- port->res.name = "PCIe MEM";
- port->res.start = port->cfg_space.start - 0x38000000;
+ port->outbound.name = "PCIe MEM";
+ port->outbound.start = port->cfg_space.start - 0x38000000;
/* allocate 256 M -- 2 MPAGEs worth */
- port->res.end = port->res.start + 0x10000000 - 1;
- port->res.flags = IORESOURCE_MEM;
+ port->outbound.end = port->outbound.start + 0x10000000 - 1;
+ port->outbound.flags = IORESOURCE_MEM;
- if (request_resource(&iomem_resource, &port->res))
+ if (request_resource(&iomem_resource, &port->outbound))
panic("Request PCIe Memory resource failed for port %d\n",
portno);
- pci_add_resource_offset(&sys->resources, &port->res, sys->mem_offset);
+ pci_add_resource_offset(&sys->resources, &port->outbound,
+ sys->mem_offset);
printk(KERN_INFO "port res start = 0x%012llx, end = 0x%012llx\n",
- port->res.start, port->res.end);
+ port->outbound.start, port->outbound.end);
printk(KERN_INFO "port system mem_offset start = 0x%012llx\n",
- sys->mem_offset);
+ sys->mem_offset);
/* 4K internal config */
cfg_addr = ioremap(port->utl_regs.start, 0x10000);
@@ -574,8 +556,7 @@ int axxia_pcie_setup(int portno, struct pci_sys_data *sys)
goto fail;
}
port->cfg_addr = cfg_addr;
- printk(KERN_INFO "cfg_addr for port %d = 0x%8x\n",
- port->index, port->cfg_addr);
+ printk("cfg_addr for port %d = 0x%8x\n", port->index, port->cfg_addr);
pci_config = readl(cfg_addr);
#ifdef PRINT_CONFIG_ACCESSES
printk(KERN_INFO "pci_vendor = 0x%08x\n", pci_config);
@@ -592,63 +573,70 @@ int axxia_pcie_setup(int portno, struct pci_sys_data *sys)
/* IRQ# 70 for PEI2 */
mappedIrq = 102;
}
- printk(KERN_INFO
- "Requesting irq#%d for PEI%d Legacy INTs\n",
- mappedIrq, port->index);
- err = request_irq(mappedIrq, acp_pcie_isr, IRQF_SHARED,
- "acp_pcie", port);
+ printk(KERN_INFO "Requesting irq#%d for PEI%d Legacy INTs\n",
+ mappedIrq, port->index);
+ err = request_irq(mappedIrq, acp_pcie_isr,
+ IRQF_SHARED, "acp_pcie", port);
if (err) {
- printk(KERN_ERR
- "request_irq failed!!!! for IRQ# %d err = %d\n",
- mappedIrq, err);
+ printk(KERN_ERR "request_irq failed!!!! for IRQ# %d err = %d\n",
+ mappedIrq, err);
goto fail;
}
+ /* MSI INTS for PEI0 */
+ if (sys->domain == 0) {
+ /* IRQ# 73-88 for PEI0 MSI INTs */
+ for (mappedIrq = 73; mappedIrq <= 88; mappedIrq++) {
+ printk(KERN_INFO
+ "Requesting irq#%d for PEI0 MSI INTs\n",
+ mappedIrq+32);
+ err = request_irq(mappedIrq+32, acp_pcie_MSI_isr,
+ IRQF_SHARED, "acp_pcie_MSI", port);
+ if (err) {
+ printk(KERN_ERR
+ "request_irq failed!!!! for IRQ# %d err = %d\n",
+ mappedIrq+32, err);
+ goto fail;
+ }
+ }
+ }
+
/* Setup as root complex */
pci_config = readl(cfg_addr + 0x1000);
#ifdef PRINT_CONFIG_ACCESSES
- printk(KERN_INFO "pci_config = 0x%08x\n", pci_config);
+ printk("pci_config = 0x%08x\n", pci_config);
#endif
pci_status = readl(cfg_addr + 0x1004);
link_state = (pci_status & 0x3f00) >> 8;
printk(KERN_INFO
- "PCIE%d status = 0x%08x : PCI link state = 0x%x\n",
+ "PCIE%d status = 0x%08x : PCI link state = 0x%x\n",
port->index, pci_status, link_state);
/* make sure the ACP device is configured as PCI Root Complex */
if ((pci_status & 0x18) != 0x18) {
- printk(KERN_ERR
- "ACP device is not PCI Root Complex! status = 0x%08x\n",
- pci_status);
+ printk(KERN_INFO
+ "ACP device is not PCI Root Complex! status = 0x%08x\n",
+ pci_status);
goto fail;
}
/* make sure the link is up */
if (link_state != 0xb) {
/* reset */
- printk(KERN_ERR
- "PCI link in bad state - resetting\n");
+ printk("PCI link in bad state - resetting\n");
pci_config |= 1;
writel(pci_config, cfg_addr + 0x1000);
msleep(1000);
pci_status = readl(cfg_addr + 0x1004);
link_state = (pci_status & 0x3f00) >> 8;
- printk(KERN_ERR
- "PCI link state now = 0x%x\n", link_state);
+ printk(KERN_INFO "PCI link state now = 0x%x\n", link_state);
if (link_state != 0xb) {
- printk(KERN_ERR
- "PCI link still in bad state - giving up!\n");
+ printk(KERN_INFO "PCI link still in bad state - giving up!\n");
goto fail;
}
}
- /* setup ACP for 4GB 1=Prefetchable, 10=Locate anywhere in
- * 64 bit address space */
- writel(0x1000000c, cfg_addr + 0x10);
- writel(0x0, cfg_addr + 0x14);
- writel(((0xff << 16) | (0x1 << 8) | 0x0), cfg_addr + PCI_PRIMARY_BUS);
-
/* ACP X1 setup MPAGE registers */
/*
* MPAGE7 is dedicated to config access, so we only
@@ -657,9 +645,9 @@ int axxia_pcie_setup(int portno, struct pci_sys_data *sys)
*/
size = 7 * 1024*128*1024;
num_pages = ((size - 1) >> 27) + 1;
+ pciah = U64_TO_U32_HIGH(port->pci_addr);
+ pcial = U64_TO_U32_LOW(port->pci_addr);
for (i = 0; i < num_pages; i++) {
- pciah = U64_TO_U32_HIGH(port->pci_addr);
- pcial = U64_TO_U32_LOW(port->pci_addr);
mpage_lower = (pcial & 0xf8000000);
mpage_lower |= 0x0;
writel(pciah, cfg_addr + ACPX1_PCIE_MPAGE_UPPER(i));
@@ -667,22 +655,34 @@ int axxia_pcie_setup(int portno, struct pci_sys_data *sys)
pcial += 0x08000000;
}
- if (sys->domain == 0) {
- /* Setup MSI address for PEI0 only since it support MSI */
- msiAddr = port->cfg_space.end & 0xfffffffffffffC00;
- msiLower = U64_TO_U32_LOW(msiAddr);
- msiLower = msiLower >> 10;
- msiHigher = U64_TO_U32_HIGH(msiAddr);
- msiHigher = msiHigher & 0x3F;
- msiFinal = ((msiHigher << 22) | (msiLower));
- printk(KERN_ERR
- "PEI%d msiHigher = 0x%x, msiLower = 0x%x\n",
- sys->domain, msiHigher, msiLower);
- printk(KERN_ERR
- "PEI%d msiAddr = 0x%012llx, msiFinal = 0x%x\n",
- sys->domain, msiAddr, msiFinal);
- writel(msiFinal, cfg_addr + 0x1190);
+ inbound_size = (u32) (port->inbound.end - port->inbound.start + 1);
+
+ /* configures the RC Memory Space Configuration Register */
+ writel(inbound_size, cfg_addr + 0x11f4);
+
+ /* write all 1s to BAR0 register */
+ writel(0xffffffff, cfg_addr + 0x10);
+
+ /* read back BAR0 */
+ bar0_size = readl(cfg_addr + 0x10);
+ if ((bar0_size & inbound_size) != inbound_size)
+ printk(KERN_INFO "Writing/Reading BAR0 reg failed\n");
+
+ /* set the BASE0 address to start of PCIe base */
+ writel(port->inbound.start, cfg_addr + 0x10);
+
+ /* setup TPAGE registers for inbound mapping */
+ /* We set the MSB of each TPAGE to select 128-bit AXI access.
+ * For the address field we simply program an incrementing value
+ * to map consecutive pages
+ */
+ tpage_base = cfg_addr + 0x1050;
+ for (i = 0; i < 8; i++) {
+ writel((0x80000000 + i), tpage_base);
+ tpage_base += 4;
}
+
+
return 1;
fail:
if (cfg_data)
@@ -717,8 +717,7 @@ axxia_pcie_scan_bus(int nr, struct pci_sys_data *sys)
if (nr < axxia_pciex_port_count) {
bus = pci_scan_root_bus(NULL, sys->busnr,
- &axxia_pciex_pci_ops,
- sys, &sys->resources);
+ &axxia_pciex_pci_ops, sys, &sys->resources);
} else {
bus = NULL;
BUG();
@@ -727,47 +726,75 @@ axxia_pcie_scan_bus(int nr, struct pci_sys_data *sys)
return bus;
}
-/* Tegra PCIE requires relaxed ordering */
+/* MSI setup */
static void __devinit axxia_pcie_msi_enable(struct pci_dev *dev)
{
- u32 msi_lower, msi_higher = 0;
+ u32 pci_higher = 0, msi_lower = 0;
+ u16 flag_val;
int pos = pci_find_capability(dev, PCI_CAP_ID_MSI);
struct axxia_pciex_port *port;
- int mappedIrq;
- int err;
+ void *msi_virt;
+ int device, fn, bus_num;
+ static u32 pci_lower, msi_count;
port = bus_to_port(dev->bus);
if (pos <= 0) {
dev_err(&dev->dev, "skipping MSI enable\n");
- printk(KERN_ERR "skipping MSI enable\n");
+ printk(KERN_INFO "skipping MSI enable\n");
return;
}
+ /* MSI applicable only to PEI0 */
if (port->index == 0) {
- /* MSI support only in PEI0 */
- msi_lower = port->pci_addr | (port->cfg_space.end & 0xfffc00);
- printk(KERN_ERR
- "PEI%d axxia_pcie_msi_enable Found MSI, msi_lower = 0x%x\n",
- port->index, msi_lower);
- pci_write_config_dword(dev, pos + PCI_MSI_ADDRESS_LO,
- msi_lower);
- pci_write_config_dword(dev, pos + PCI_MSI_ADDRESS_HI,
- msi_higher);
- /* IRQ# 73-88 for PEI0 MSI INTs */
- for (mappedIrq = 73; mappedIrq <= 88; mappedIrq++) {
- printk(KERN_ERR
- "Requesting irq#%d for PEI0 MSI INTs\n",
- mappedIrq+32);
- err = request_irq(mappedIrq+32, acp_pcie_MSI_isr,
- IRQF_SHARED, "acp_pcie_MSI", port);
- if (err) {
- printk(KERN_ERR
- "request_irq failed!!!! for IRQ# %d err = %d\n",
- mappedIrq+32, err);
+ /* MSI is generated when EP writes to address mapped */
+ if (msi_lower == 0) {
+ /* MSI support only in PEI0 */
+ /* allocate 1K to manage 256 MSIs
+ * one for each endpoint */
+
+ if ((!dma_set_coherent_mask(&dev->dev,
+ DMA_BIT_MASK(64))) ||
+ (!dma_set_coherent_mask(dev,
+ DMA_BIT_MASK(32)))) {
+ msi_virt = dma_alloc_coherent(&dev->dev,
+ 1024, &(port->msi_phys), GFP_KERNEL);
+ } else {
+ printk(KERN_INFO
+ "No suitable DMA available. MSI cannot be supported\n");
return;
}
+ msi_lower = (u32)port->msi_phys;
+ /* Please note we have 1:1 mapping for inbound */
+ pci_lower = port->inbound.start + msi_lower;
+ printk("PEI%d dma_alloc_coherent msiAddr = 0x%x\n",
+ port->index, msi_lower);
+ writel(msi_lower>>10, port->cfg_addr + 0x1190);
}
+ device = ((PCI_SLOT(dev->devfn) & 0xf8) >> 3);
+ fn = PCI_FUNC(dev->devfn) & 0x7;
+ bus_num = dev->bus->number;
+
+ printk(KERN_INFO
+ "PEI%d axxia_pcie_msi_enable Found MSI%d, msi_lower = 0x%x for bus_num = %d, dev = %d, fn = %d\n",
+ port->index, msi_count, msi_lower,
+ bus_num, device, fn);
+ pci_write_config_dword(dev, pos + PCI_MSI_ADDRESS_LO,
+ pci_lower);
+ pci_write_config_dword(dev, pos + PCI_MSI_ADDRESS_HI,
+ pci_higher);
+
+ /* Enable MSI */
+ dev->msi_enabled = 1;
+
+ pci_read_config_word(dev, pos + PCI_MSI_FLAGS, &flag_val);
+ flag_val = flag_val | 0x1;
+ pci_write_config_word(dev, pos + PCI_MSI_FLAGS,
+ (flag_val | 0x1));
+
+ /* for next EP MSI */
+ pci_lower = pci_lower + 0x4;
+ msi_count++;
}
return;
}
@@ -775,24 +802,24 @@ DECLARE_PCI_FIXUP_HEADER(PCI_ANY_ID, PCI_ANY_ID, axxia_pcie_msi_enable);
/* Port definition struct
* Please note: PEI core#1 is not used in AXM5500 */
-static struct hw_pci __refdata axxia_pcie_hw[] = {
+static struct hw_pci axxia_pcie_hw[] = {
[0] = {
- .nr_controllers = 1,
- .domain = 0,
- .swizzle = pci_std_swizzle,
- .setup = axxia_pcie_setup,
- .scan = axxia_pcie_scan_bus
+ .nr_controllers = 1,
+ .domain = 0,
+ .swizzle = pci_std_swizzle,
+ .setup = axxia_pcie_setup,
+ .scan = axxia_pcie_scan_bus
},
[1] = {
- .nr_controllers = 0,
- .domain = 1
+ .nr_controllers = 0,
+ .domain = 1
},
[2] = {
- .nr_controllers = 1,
- .domain = 2,
- .swizzle = pci_std_swizzle,
- .setup = axxia_pcie_setup,
- .scan = axxia_pcie_scan_bus
+ .nr_controllers = 1,
+ .domain = 2,
+ .swizzle = pci_std_swizzle,
+ .setup = axxia_pcie_setup,
+ .scan = axxia_pcie_scan_bus
}
};
@@ -801,9 +828,8 @@ void __init axxia_pcie_init(void)
{
struct device_node *np;
/* allocate memory */
- axxia_pciex_ports = kzalloc(axxia_pciex_port_count *
- sizeof(struct axxia_pciex_port),
- GFP_KERNEL);
+ axxia_pciex_ports = kzalloc(axxia_pciex_port_count
+ * sizeof(struct axxia_pciex_port), GFP_KERNEL);
if (!axxia_pciex_ports) {
printk(KERN_WARNING "PCIE: failed to allocate ports array\n");
@@ -829,6 +855,7 @@ static void axxia_probe_pciex_bridge(struct device_node *np)
int rlen;
int pna = of_n_addr_cells(np);
int num = pna + 5;
+ u64 size;
/* Get the port number from the device-tree */
if (!of_property_read_u32(np, "port", &pval)) {
@@ -852,6 +879,7 @@ static void axxia_probe_pciex_bridge(struct device_node *np)
port = &axxia_pciex_ports[portno];
port->index = portno;
+
port->node = of_node_get(np);
/* Check if device_type property is set to "pci" or "pci-endpoint".
@@ -864,8 +892,7 @@ static void axxia_probe_pciex_bridge(struct device_node *np)
} else if (!strcmp(val, "pci")) {
port->endpoint = 0;
} else {
- printk(KERN_ERR
- "PCIE%d: missing or incorrect device_type for %s\n",
+ printk(KERN_ERR "PCIE%d: missing or incorrect device_type for %s\n",
portno, np->full_name);
return;
}
@@ -877,8 +904,9 @@ static void axxia_probe_pciex_bridge(struct device_node *np)
np->full_name);
return;
}
- printk(KERN_ERR "cfg_space start = 0x%012llx, end = 0x%012llx\n",
- port->cfg_space.start, port->cfg_space.end);
+ printk(KERN_INFO
+ "cfg_space start = 0x%012llx, end = 0x%012llx\n",
+ port->cfg_space.start, port->cfg_space.end);
/* Fetch host bridge internal registers address */
if (of_address_to_resource(np, 1, &port->utl_regs)) {
@@ -886,21 +914,62 @@ static void axxia_probe_pciex_bridge(struct device_node *np)
np->full_name);
return;
}
- printk(KERN_ERR "utl_regs start = 0x%012llx, end = 0x%012llx\n",
- port->utl_regs.start, port->utl_regs.end);
+ printk(KERN_INFO
+ "utl_regs start = 0x%012llx, end = 0x%012llx\n",
+ port->utl_regs.start, port->utl_regs.end);
- field = of_get_property(np, "dma-ranges", &rlen);
+ field = of_get_property(np, "ranges", &rlen);
if (field == NULL)
- printk(KERN_ERR "not able to get dma-ranges\n");
+ printk("not able to get ranges\n");
+
+ u64 pci_addr = of_read_number(field + 1, 2);
+ printk(KERN_INFO "pci_addr = 0x%012llx\n", pci_addr);
+ port->pci_addr = pci_addr;
+
+ printk(KERN_INFO "%s PCIE%d config base = 0x%012llx\n",
+ np->full_name, port->index, port->utl_regs.start);
+
+ /* Default 256 MB */
+ port->inbound.start = 0;
+ size = 0x10000000;
+ port->inbound.end = size - 1;
+ port->inbound.flags = IORESOURCE_MEM | IORESOURCE_PREFETCH;
+
+ /* Get dma-ranges property */
+ field = of_get_property(np, "dma-ranges", &rlen);
+ if (field == NULL) {
+ printk(KERN_INFO "not able to get dma-ranges\n");
+ /* use default */
+ return;
+ }
/* Walk it */
while ((rlen -= num * 4) >= 0) {
u64 pci_addr = of_read_number(field + 1, 2);
- printk(KERN_ERR "pci_addr = 0x%012llx\n", pci_addr);
- port->pci_addr = pci_addr;
+ u64 cpu_addr = of_read_number(field + 3, 1);
+ size = of_read_number(field + pna + 3, 2);
+ field += num;
+
+ /* We currently only support memory at 0, and pci_addr
+ * within 32 bits space and 1:1 mapping
+ */
+ if (cpu_addr != 0 || pci_addr > 0xffffffff) {
+ printk(KERN_WARNING
+ "%s: Ignored unsupported dma range 0x%016llx...0x%016llx -> 0x%016llx\n",
+ np->full_name, pci_addr,
+ pci_addr + size - 1, cpu_addr);
+ continue;
+ }
+ /* Use that */
+ port->inbound.start = pci_addr;
+ /* Beware of 32 bits resources */
+ if (sizeof(resource_size_t) == sizeof(u32) &&
+ (pci_addr + size) > 0x100000000ull) {
+ port->inbound.end = 0xffffffff;
+ } else
+ port->inbound.end = port->inbound.start + size - 1;
break;
}
- printk(KERN_ERR "%s PCIE%d config base = 0x%012llx\n", np->full_name,
- port->index, port->utl_regs.start);
+ printk(KERN_INFO "inbound start = 0x%016llx, end = 0x%016llx\n",\
+ port->inbound.start, port->inbound.end);
}
-
--
1.8.3
More information about the linux-yocto
mailing list