MIPS: Add Cavium OCTEON PCI support.

This patch adds support for PCI and PCIe to the base Cavium OCTEON
processor support.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

arch/mips/Kconfig

@@ -618,6 +618,8 @@ config CAVIUM_OCTEON_REFERENCE_BOARD
 	select SYS_HAS_EARLY_PRINTK
 	select SYS_HAS_CPU_CAVIUM_OCTEON
 	select SWAP_IO_SPACE
+	select HW_HAS_PCI
+	select ARCH_SUPPORTS_MSI
 	help
 	  This option supports all of the Octeon reference boards from Cavium
 	  Networks. It builds a kernel that dynamically determines the Octeon

arch/mips/cavium-octeon/Makefile

@@ -14,5 +14,9 @@ obj-y += dma-octeon.o flash_setup.o
 obj-y += octeon-memcpy.o
 
 obj-$(CONFIG_SMP)                     += smp.o
+obj-$(CONFIG_PCI)                     += pci-common.o
+obj-$(CONFIG_PCI)                     += pci.o
+obj-$(CONFIG_PCI)                     += pcie.o
+obj-$(CONFIG_PCI_MSI)                 += msi.o
 
 EXTRA_CFLAGS += -Werror

arch/mips/cavium-octeon/dma-octeon.c

@@ -13,20 +13,327 @@
  */
 #include <linux/types.h>
 #include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/dma-mapping.h>
+#include <linux/platform_device.h>
+#include <linux/scatterlist.h>
+
+#include <linux/cache.h>
+#include <linux/io.h>
+
+#include <asm/octeon/octeon.h>
+#include <asm/octeon/cvmx-npi-defs.h>
+#include <asm/octeon/cvmx-pci-defs.h>
 
 #include <dma-coherence.h>
 
+#ifdef CONFIG_PCI
+#include "pci-common.h"
+#endif
+
+#define BAR2_PCI_ADDRESS 0x8000000000ul
+
+struct bar1_index_state {
+	int16_t ref_count;	/* Number of PCI mappings using this index */
+	uint16_t address_bits;	/* Upper bits of physical address. This is
+				   shifted 22 bits */
+};
+
+#ifdef CONFIG_PCI
+static DEFINE_SPINLOCK(bar1_lock);
+static struct bar1_index_state bar1_state[32];
+#endif
+
 dma_addr_t octeon_map_dma_mem(struct device *dev, void *ptr, size_t size)
 {
+#ifndef CONFIG_PCI
 	/* Without PCI/PCIe this function can be called for Octeon internal
 	   devices such as USB. These devices all support 64bit addressing */
 	mb();
 	return virt_to_phys(ptr);
+#else
+	unsigned long flags;
+	uint64_t dma_mask;
+	int64_t start_index;
+	dma_addr_t result = -1;
+	uint64_t physical = virt_to_phys(ptr);
+	int64_t index;
+
+	mb();
+	/*
+	 * Use the DMA masks to determine the allowed memory
+	 * region. For us it doesn't limit the actual memory, just the
+	 * address visible over PCI.  Devices with limits need to use
+	 * lower indexed Bar1 entries.
+	 */
+	if (dev) {
+		dma_mask = dev->coherent_dma_mask;
+		if (dev->dma_mask)
+			dma_mask = *dev->dma_mask;
+	} else {
+		dma_mask = 0xfffffffful;
+	}
+
+	/*
+	 * Platform devices, such as the internal USB, skip all
+	 * translation and use Octeon physical addresses directly.
+	 */
+	if (!dev || dev->bus == &platform_bus_type)
+		return physical;
+
+	switch (octeon_dma_bar_type) {
+	case OCTEON_DMA_BAR_TYPE_PCIE:
+		if (unlikely(physical < (16ul << 10)))
+			panic("dma_map_single: Not allowed to map first 16KB."
+			      " It interferes with BAR0 special area\n");
+		else if ((physical + size >= (256ul << 20)) &&
+			 (physical < (512ul << 20)))
+			panic("dma_map_single: Not allowed to map bootbus\n");
+		else if ((physical + size >= 0x400000000ull) &&
+			 physical < 0x410000000ull)
+			panic("dma_map_single: "
+			      "Attempt to map illegal memory address 0x%llx\n",
+			      physical);
+		else if (physical >= 0x420000000ull)
+			panic("dma_map_single: "
+			      "Attempt to map illegal memory address 0x%llx\n",
+			      physical);
+		else if ((physical + size >=
+			  (4ull<<30) - (OCTEON_PCI_BAR1_HOLE_SIZE<<20))
+			 && physical < (4ull<<30))
+			pr_warning("dma_map_single: Warning: "
+				   "Mapping memory address that might "
+				   "conflict with devices 0x%llx-0x%llx\n",
+				   physical, physical+size-1);
+		/* The 2nd 256MB is mapped at 256<<20 instead of 0x410000000 */
+		if ((physical >= 0x410000000ull) && physical < 0x420000000ull)
+			result = physical - 0x400000000ull;
+		else
+			result = physical;
+		if (((result+size-1) & dma_mask) != result+size-1)
+			panic("dma_map_single: Attempt to map address "
+			      "0x%llx-0x%llx, which can't be accessed "
+			      "according to the dma mask 0x%llx\n",
+			      physical, physical+size-1, dma_mask);
+		goto done;
+
+	case OCTEON_DMA_BAR_TYPE_BIG:
+#ifdef CONFIG_64BIT
+		/* If the device supports 64bit addressing, then use BAR2 */
+		if (dma_mask > BAR2_PCI_ADDRESS) {
+			result = physical + BAR2_PCI_ADDRESS;
+			goto done;
+		}
+#endif
+		if (unlikely(physical < (4ul << 10))) {
+			panic("dma_map_single: Not allowed to map first 4KB. "
+			      "It interferes with BAR0 special area\n");
+		} else if (physical < (256ul << 20)) {
+			if (unlikely(physical + size > (256ul << 20)))
+				panic("dma_map_single: Requested memory spans "
+				      "Bar0 0:256MB and bootbus\n");
+			result = physical;
+			goto done;
+		} else if (unlikely(physical < (512ul << 20))) {
+			panic("dma_map_single: Not allowed to map bootbus\n");
+		} else if (physical < (2ul << 30)) {
+			if (unlikely(physical + size > (2ul << 30)))
+				panic("dma_map_single: Requested memory spans "
+				      "Bar0 512MB:2GB and BAR1\n");
+			result = physical;
+			goto done;
+		} else if (physical < (2ul << 30) + (128 << 20)) {
+			/* Fall through */
+		} else if (physical <
+			   (4ul << 30) - (OCTEON_PCI_BAR1_HOLE_SIZE << 20)) {
+			if (unlikely
+			    (physical + size >
+			     (4ul << 30) - (OCTEON_PCI_BAR1_HOLE_SIZE << 20)))
+				panic("dma_map_single: Requested memory "
+				      "extends past Bar1 (4GB-%luMB)\n",
+				      OCTEON_PCI_BAR1_HOLE_SIZE);
+			result = physical;
+			goto done;
+		} else if ((physical >= 0x410000000ull) &&
+			   (physical < 0x420000000ull)) {
+			if (unlikely(physical + size > 0x420000000ull))
+				panic("dma_map_single: Requested memory spans "
+				      "non existant memory\n");
+			/* BAR0 fixed mapping 256MB:512MB ->
+			 * 16GB+256MB:16GB+512MB */
+			result = physical - 0x400000000ull;
+			goto done;
+		} else {
+			/* Continued below switch statement */
+		}
+		break;
+
+	case OCTEON_DMA_BAR_TYPE_SMALL:
+#ifdef CONFIG_64BIT
+		/* If the device supports 64bit addressing, then use BAR2 */
+		if (dma_mask > BAR2_PCI_ADDRESS) {
+			result = physical + BAR2_PCI_ADDRESS;
+			goto done;
+		}
+#endif
+		/* Continued below switch statement */
+		break;
+
+	default:
+		panic("dma_map_single: Invalid octeon_dma_bar_type\n");
+	}
+
+	/* Don't allow mapping to span multiple Bar entries. The hardware guys
+	   won't guarantee that DMA across boards work */
+	if (unlikely((physical >> 22) != ((physical + size - 1) >> 22)))
+		panic("dma_map_single: "
+		      "Requested memory spans more than one Bar1 entry\n");
+
+	if (octeon_dma_bar_type == OCTEON_DMA_BAR_TYPE_BIG)
+		start_index = 31;
+	else if (unlikely(dma_mask < (1ul << 27)))
+		start_index = (dma_mask >> 22);
+	else
+		start_index = 31;
+
+	/* Only one processor can access the Bar register at once */
+	spin_lock_irqsave(&bar1_lock, flags);
+
+	/* Look through Bar1 for existing mapping that will work */
+	for (index = start_index; index >= 0; index--) {
+		if ((bar1_state[index].address_bits == physical >> 22) &&
+		    (bar1_state[index].ref_count)) {
+			/* An existing mapping will work, use it */
+			bar1_state[index].ref_count++;
+			if (unlikely(bar1_state[index].ref_count < 0))
+				panic("dma_map_single: "
+				      "Bar1[%d] reference count overflowed\n",
+				      (int) index);
+			result = (index << 22) | (physical & ((1 << 22) - 1));
+			/* Large BAR1 is offset at 2GB */
+			if (octeon_dma_bar_type == OCTEON_DMA_BAR_TYPE_BIG)
+				result += 2ul << 30;
+			goto done_unlock;
+		}
+	}
+
+	/* No existing mappings, look for a free entry */
+	for (index = start_index; index >= 0; index--) {
+		if (unlikely(bar1_state[index].ref_count == 0)) {
+			union cvmx_pci_bar1_indexx bar1_index;
+			/* We have a free entry, use it */
+			bar1_state[index].ref_count = 1;
+			bar1_state[index].address_bits = physical >> 22;
+			bar1_index.u32 = 0;
+			/* Address bits[35:22] sent to L2C */
+			bar1_index.s.addr_idx = physical >> 22;
+			/* Don't put PCI accesses in L2. */
+			bar1_index.s.ca = 1;
+			/* Endian Swap Mode */
+			bar1_index.s.end_swp = 1;
+			/* Set '1' when the selected address range is valid. */
+			bar1_index.s.addr_v = 1;
+			octeon_npi_write32(CVMX_NPI_PCI_BAR1_INDEXX(index),
+					   bar1_index.u32);
+			/* An existing mapping will work, use it */
+			result = (index << 22) | (physical & ((1 << 22) - 1));
+			/* Large BAR1 is offset at 2GB */
+			if (octeon_dma_bar_type == OCTEON_DMA_BAR_TYPE_BIG)
+				result += 2ul << 30;
+			goto done_unlock;
+		}
+	}
+
+	pr_err("dma_map_single: "
+	       "Can't find empty BAR1 index for physical mapping 0x%llx\n",
+	       (unsigned long long) physical);
+
+done_unlock:
+	spin_unlock_irqrestore(&bar1_lock, flags);
+done:
+	pr_debug("dma_map_single 0x%llx->0x%llx\n", physical, result);
+	return result;
+#endif
 }
 
 void octeon_unmap_dma_mem(struct device *dev, dma_addr_t dma_addr)
 {
-	/* Without PCI/PCIe this function can be called for Octeon internal
-	 * devices such as USB. These devices all support 64bit addressing */
+#ifndef CONFIG_PCI
+	/*
+	 * Without PCI/PCIe this function can be called for Octeon internal
+	 * devices such as USB. These devices all support 64bit addressing.
+	 */
 	return;
+#else
+	unsigned long flags;
+	uint64_t index;
+
+	/*
+	 * Platform devices, such as the internal USB, skip all
+	 * translation and use Octeon physical addresses directly.
+	 */
+	if (dev->bus == &platform_bus_type)
+		return;
+
+	switch (octeon_dma_bar_type) {
+	case OCTEON_DMA_BAR_TYPE_PCIE:
+		/* Nothing to do, all mappings are static */
+		goto done;
+
+	case OCTEON_DMA_BAR_TYPE_BIG:
+#ifdef CONFIG_64BIT
+		/* Nothing to do for addresses using BAR2 */
+		if (dma_addr >= BAR2_PCI_ADDRESS)
+			goto done;
+#endif
+		if (unlikely(dma_addr < (4ul << 10)))
+			panic("dma_unmap_single: Unexpect DMA address 0x%llx\n",
+			      dma_addr);
+		else if (dma_addr < (2ul << 30))
+			/* Nothing to do for addresses using BAR0 */
+			goto done;
+		else if (dma_addr < (2ul << 30) + (128ul << 20))
+			/* Need to unmap, fall through */
+			index = (dma_addr - (2ul << 30)) >> 22;
+		else if (dma_addr <
+			 (4ul << 30) - (OCTEON_PCI_BAR1_HOLE_SIZE << 20))
+			goto done;	/* Nothing to do for the rest of BAR1 */
+		else
+			panic("dma_unmap_single: Unexpect DMA address 0x%llx\n",
+			      dma_addr);
+		/* Continued below switch statement */
+		break;
+
+	case OCTEON_DMA_BAR_TYPE_SMALL:
+#ifdef CONFIG_64BIT
+		/* Nothing to do for addresses using BAR2 */
+		if (dma_addr >= BAR2_PCI_ADDRESS)
+			goto done;
+#endif
+		index = dma_addr >> 22;
+		/* Continued below switch statement */
+		break;
+
+	default:
+		panic("dma_unmap_single: Invalid octeon_dma_bar_type\n");
+	}
+
+	if (unlikely(index > 31))
+		panic("dma_unmap_single: "
+		      "Attempt to unmap an invalid address (0x%llx)\n",
+		      dma_addr);
+
+	spin_lock_irqsave(&bar1_lock, flags);
+	bar1_state[index].ref_count--;
+	if (bar1_state[index].ref_count == 0)
+		octeon_npi_write32(CVMX_NPI_PCI_BAR1_INDEXX(index), 0);
+	else if (unlikely(bar1_state[index].ref_count < 0))
+		panic("dma_unmap_single: Bar1[%u] reference count < 0\n",
+		      (int) index);
+	spin_unlock_irqrestore(&bar1_lock, flags);
+done:
+	pr_debug("dma_unmap_single 0x%llx\n", dma_addr);
+	return;
+#endif
 }

arch/mips/cavium-octeon/executive/Makefile

@@ -11,3 +11,4 @@
 
 obj-y += cvmx-bootmem.o cvmx-l2c.o cvmx-sysinfo.o octeon-model.o
 
+obj-$(CONFIG_PCI) += cvmx-helper-errata.o cvmx-helper-jtag.o

arch/mips/cavium-octeon/executive/cvmx-helper-errata.c

@@ -0,0 +1,70 @@
+/***********************license start***************
+ * Author: Cavium Networks
+ *
+ * Contact: support@caviumnetworks.com
+ * This file is part of the OCTEON SDK
+ *
+ * Copyright (c) 2003-2008 Cavium Networks
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, Version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful, but
+ * AS-IS and WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, TITLE, or
+ * NONINFRINGEMENT.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this file; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * or visit http://www.gnu.org/licenses/.
+ *
+ * This file may also be available under a different license from Cavium.
+ * Contact Cavium Networks for more information
+ ***********************license end**************************************/
+
+/**
+ *
+ * Fixes and workaround for Octeon chip errata. This file
+ * contains functions called by cvmx-helper to workaround known
+ * chip errata. For the most part, code doesn't need to call
+ * these functions directly.
+ *
+ */
+#include <asm/octeon/octeon.h>
+
+#include <asm/octeon/cvmx-helper-jtag.h>
+
+/**
+ * Due to errata G-720, the 2nd order CDR circuit on CN52XX pass
+ * 1 doesn't work properly. The following code disables 2nd order
+ * CDR for the specified QLM.
+ *
+ * @qlm:    QLM to disable 2nd order CDR for.
+ */
+void __cvmx_helper_errata_qlm_disable_2nd_order_cdr(int qlm)
+{
+	int lane;
+	cvmx_helper_qlm_jtag_init();
+	/* We need to load all four lanes of the QLM, a total of 1072 bits */
+	for (lane = 0; lane < 4; lane++) {
+		/*
+		 * Each lane has 268 bits. We need to set
+		 * cfg_cdr_incx<67:64> = 3 and cfg_cdr_secord<77> =
+		 * 1. All other bits are zero. Bits go in LSB first,
+		 * so start off with the zeros for bits <63:0>.
+		 */
+		cvmx_helper_qlm_jtag_shift_zeros(qlm, 63 - 0 + 1);
+		/* cfg_cdr_incx<67:64>=3 */
+		cvmx_helper_qlm_jtag_shift(qlm, 67 - 64 + 1, 3);
+		/* Zeros for bits <76:68> */
+		cvmx_helper_qlm_jtag_shift_zeros(qlm, 76 - 68 + 1);
+		/* cfg_cdr_secord<77>=1 */
+		cvmx_helper_qlm_jtag_shift(qlm, 77 - 77 + 1, 1);
+		/* Zeros for bits <267:78> */
+		cvmx_helper_qlm_jtag_shift_zeros(qlm, 267 - 78 + 1);
+	}
+	cvmx_helper_qlm_jtag_update(qlm);
+}

arch/mips/cavium-octeon/executive/cvmx-helper-jtag.c

@@ -0,0 +1,144 @@
+
+/***********************license start***************
+ * Author: Cavium Networks
+ *
+ * Contact: support@caviumnetworks.com
+ * This file is part of the OCTEON SDK
+ *
+ * Copyright (c) 2003-2008 Cavium Networks
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, Version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful, but
+ * AS-IS and WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, TITLE, or
+ * NONINFRINGEMENT.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this file; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * or visit http://www.gnu.org/licenses/.
+ *
+ * This file may also be available under a different license from Cavium.
+ * Contact Cavium Networks for more information
+ ***********************license end**************************************/
+
+/**
+ *
+ * Helper utilities for qlm_jtag.
+ *
+ */
+
+#include <asm/octeon/octeon.h>
+#include <asm/octeon/cvmx-helper-jtag.h>
+
+
+/**
+ * Initialize the internal QLM JTAG logic to allow programming
+ * of the JTAG chain by the cvmx_helper_qlm_jtag_*() functions.
+ * These functions should only be used at the direction of Cavium
+ * Networks. Programming incorrect values into the JTAG chain
+ * can cause chip damage.
+ */
+void cvmx_helper_qlm_jtag_init(void)
+{
+	union cvmx_ciu_qlm_jtgc jtgc;
+	uint32_t clock_div = 0;
+	uint32_t divisor = cvmx_sysinfo_get()->cpu_clock_hz / (25 * 1000000);
+	divisor = (divisor - 1) >> 2;
+	/* Convert the divisor into a power of 2 shift */
+	while (divisor) {
+		clock_div++;
+		divisor = divisor >> 1;
+	}
+
+	/*
+	 * Clock divider for QLM JTAG operations.  eclk is divided by
+	 * 2^(CLK_DIV + 2)
+	 */
+	jtgc.u64 = 0;
+	jtgc.s.clk_div = clock_div;
+	jtgc.s.mux_sel = 0;
+	if (OCTEON_IS_MODEL(OCTEON_CN52XX))
+		jtgc.s.bypass = 0x3;
+	else
+		jtgc.s.bypass = 0xf;
+	cvmx_write_csr(CVMX_CIU_QLM_JTGC, jtgc.u64);
+	cvmx_read_csr(CVMX_CIU_QLM_JTGC);
+}
+
+/**
+ * Write up to 32bits into the QLM jtag chain. Bits are shifted
+ * into the MSB and out the LSB, so you should shift in the low
+ * order bits followed by the high order bits. The JTAG chain is
+ * 4 * 268 bits long, or 1072.
+ *
+ * @qlm:    QLM to shift value into
+ * @bits:   Number of bits to shift in (1-32).
+ * @data:   Data to shift in. Bit 0 enters the chain first, followed by
+ *               bit 1, etc.
+ *
+ * Returns The low order bits of the JTAG chain that shifted out of the
+ *         circle.
+ */
+uint32_t cvmx_helper_qlm_jtag_shift(int qlm, int bits, uint32_t data)
+{
+	union cvmx_ciu_qlm_jtgd jtgd;
+	jtgd.u64 = 0;
+	jtgd.s.shift = 1;
+	jtgd.s.shft_cnt = bits - 1;
+	jtgd.s.shft_reg = data;
+	if (!OCTEON_IS_MODEL(OCTEON_CN56XX_PASS1_X))
+		jtgd.s.select = 1 << qlm;
+	cvmx_write_csr(CVMX_CIU_QLM_JTGD, jtgd.u64);
+	do {
+		jtgd.u64 = cvmx_read_csr(CVMX_CIU_QLM_JTGD);
+	} while (jtgd.s.shift);
+	return jtgd.s.shft_reg >> (32 - bits);
+}
+
+/**
+ * Shift long sequences of zeros into the QLM JTAG chain. It is
+ * common to need to shift more than 32 bits of zeros into the
+ * chain. This function is a convience wrapper around
+ * cvmx_helper_qlm_jtag_shift() to shift more than 32 bits of
+ * zeros at a time.
+ *
+ * @qlm:    QLM to shift zeros into
+ * @bits:
+ */
+void cvmx_helper_qlm_jtag_shift_zeros(int qlm, int bits)
+{
+	while (bits > 0) {
+		int n = bits;
+		if (n > 32)
+			n = 32;
+		cvmx_helper_qlm_jtag_shift(qlm, n, 0);
+		bits -= n;
+	}
+}
+
+/**
+ * Program the QLM JTAG chain into all lanes of the QLM. You must
+ * have already shifted in 268*4, or 1072 bits into the JTAG
+ * chain. Updating invalid values can possibly cause chip damage.
+ *
+ * @qlm:    QLM to program
+ */
+void cvmx_helper_qlm_jtag_update(int qlm)
+{
+	union cvmx_ciu_qlm_jtgd jtgd;
+
+	/* Update the new data */
+	jtgd.u64 = 0;
+	jtgd.s.update = 1;
+	if (!OCTEON_IS_MODEL(OCTEON_CN56XX_PASS1_X))
+		jtgd.s.select = 1 << qlm;
+	cvmx_write_csr(CVMX_CIU_QLM_JTGD, jtgd.u64);
+	do {
+		jtgd.u64 = cvmx_read_csr(CVMX_CIU_QLM_JTGD);
+	} while (jtgd.s.update);
+}

arch/mips/cavium-octeon/msi.c

@@ -0,0 +1,288 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005-2007 Cavium Networks
+ */
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/msi.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+
+#include <asm/octeon/octeon.h>
+#include <asm/octeon/cvmx-npi-defs.h>
+#include <asm/octeon/cvmx-pci-defs.h>
+#include <asm/octeon/cvmx-npei-defs.h>
+#include <asm/octeon/cvmx-pexp-defs.h>
+
+#include "pci-common.h"
+
+/*
+ * Each bit in msi_free_irq_bitmask represents a MSI interrupt that is
+ * in use.
+ */
+static uint64_t msi_free_irq_bitmask;
+
+/*
+ * Each bit in msi_multiple_irq_bitmask tells that the device using
+ * this bit in msi_free_irq_bitmask is also using the next bit. This
+ * is used so we can disable all of the MSI interrupts when a device
+ * uses multiple.
+ */
+static uint64_t msi_multiple_irq_bitmask;
+
+/*
+ * This lock controls updates to msi_free_irq_bitmask and
+ * msi_multiple_irq_bitmask.
+ */
+static DEFINE_SPINLOCK(msi_free_irq_bitmask_lock);
+
+
+/**
+ * Called when a driver request MSI interrupts instead of the
+ * legacy INT A-D. This routine will allocate multiple interrupts
+ * for MSI devices that support them. A device can override this by
+ * programming the MSI control bits [6:4] before calling
+ * pci_enable_msi().
+ *
+ * @param dev    Device requesting MSI interrupts
+ * @param desc   MSI descriptor
+ *
+ * Returns 0 on success.
+ */
+int arch_setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc)
+{
+	struct msi_msg msg;
+	uint16_t control;
+	int configured_private_bits;
+	int request_private_bits;
+	int irq;
+	int irq_step;
+	uint64_t search_mask;
+
+	/*
+	 * Read the MSI config to figure out how many IRQs this device
+	 * wants.  Most devices only want 1, which will give
+	 * configured_private_bits and request_private_bits equal 0.
+	 */
+	pci_read_config_word(dev, desc->msi_attrib.pos + PCI_MSI_FLAGS,
+			     &control);
+
+	/*
+	 * If the number of private bits has been configured then use
+	 * that value instead of the requested number. This gives the
+	 * driver the chance to override the number of interrupts
+	 * before calling pci_enable_msi().
+	 */
+	configured_private_bits = (control & PCI_MSI_FLAGS_QSIZE) >> 4;
+	if (configured_private_bits == 0) {
+		/* Nothing is configured, so use the hardware requested size */
+		request_private_bits = (control & PCI_MSI_FLAGS_QMASK) >> 1;
+	} else {
+		/*
+		 * Use the number of configured bits, assuming the
+		 * driver wanted to override the hardware request
+		 * value.
+		 */
+		request_private_bits = configured_private_bits;
+	}
+
+	/*
+	 * The PCI 2.3 spec mandates that there are at most 32
+	 * interrupts. If this device asks for more, only give it one.
+	 */
+	if (request_private_bits > 5)
+		request_private_bits = 0;
+
+try_only_one:
+	/*
+	 * The IRQs have to be aligned on a power of two based on the
+	 * number being requested.
+	 */
+	irq_step = 1 << request_private_bits;
+
+	/* Mask with one bit for each IRQ */
+	search_mask = (1 << irq_step) - 1;
+
+	/*
+	 * We're going to search msi_free_irq_bitmask_lock for zero
+	 * bits. This represents an MSI interrupt number that isn't in
+	 * use.
+	 */
+	spin_lock(&msi_free_irq_bitmask_lock);
+	for (irq = 0; irq < 64; irq += irq_step) {
+		if ((msi_free_irq_bitmask & (search_mask << irq)) == 0) {
+			msi_free_irq_bitmask |= search_mask << irq;
+			msi_multiple_irq_bitmask |= (search_mask >> 1) << irq;
+			break;
+		}
+	}
+	spin_unlock(&msi_free_irq_bitmask_lock);
+
+	/* Make sure the search for available interrupts didn't fail */
+	if (irq >= 64) {
+		if (request_private_bits) {
+			pr_err("arch_setup_msi_irq: Unable to find %d free "
+			       "interrupts, trying just one",
+			       1 << request_private_bits);
+			request_private_bits = 0;
+			goto try_only_one;
+		} else
+			panic("arch_setup_msi_irq: Unable to find a free MSI "
+			      "interrupt");
+	}
+
+	/* MSI interrupts start at logical IRQ OCTEON_IRQ_MSI_BIT0 */
+	irq += OCTEON_IRQ_MSI_BIT0;
+
+	switch (octeon_dma_bar_type) {
+	case OCTEON_DMA_BAR_TYPE_SMALL:
+		/* When not using big bar, Bar 0 is based at 128MB */
+		msg.address_lo =
+			((128ul << 20) + CVMX_PCI_MSI_RCV) & 0xffffffff;
+		msg.address_hi = ((128ul << 20) + CVMX_PCI_MSI_RCV) >> 32;
+	case OCTEON_DMA_BAR_TYPE_BIG:
+		/* When using big bar, Bar 0 is based at 0 */
+		msg.address_lo = (0 + CVMX_PCI_MSI_RCV) & 0xffffffff;
+		msg.address_hi = (0 + CVMX_PCI_MSI_RCV) >> 32;
+		break;
+	case OCTEON_DMA_BAR_TYPE_PCIE:
+		/* When using PCIe, Bar 0 is based at 0 */
+		/* FIXME CVMX_NPEI_MSI_RCV* other than 0? */
+		msg.address_lo = (0 + CVMX_NPEI_PCIE_MSI_RCV) & 0xffffffff;
+		msg.address_hi = (0 + CVMX_NPEI_PCIE_MSI_RCV) >> 32;
+		break;
+	default:
+		panic("arch_setup_msi_irq: Invalid octeon_dma_bar_type\n");
+	}
+	msg.data = irq - OCTEON_IRQ_MSI_BIT0;
+
+	/* Update the number of IRQs the device has available to it */
+	control &= ~PCI_MSI_FLAGS_QSIZE;
+	control |= request_private_bits << 4;
+	pci_write_config_word(dev, desc->msi_attrib.pos + PCI_MSI_FLAGS,
+			      control);
+
+	set_irq_msi(irq, desc);
+	write_msi_msg(irq, &msg);
+	return 0;
+}
+
+
+/**
+ * Called when a device no longer needs its MSI interrupts. All
+ * MSI interrupts for the device are freed.
+ *
+ * @irq:    The devices first irq number. There may be multple in sequence.
+ */
+void arch_teardown_msi_irq(unsigned int irq)
+{
+	int number_irqs;
+	uint64_t bitmask;
+
+	if ((irq < OCTEON_IRQ_MSI_BIT0) || (irq > OCTEON_IRQ_MSI_BIT63))
+		panic("arch_teardown_msi_irq: Attempted to teardown illegal "
+		      "MSI interrupt (%d)", irq);
+	irq -= OCTEON_IRQ_MSI_BIT0;
+
+	/*
+	 * Count the number of IRQs we need to free by looking at the
+	 * msi_multiple_irq_bitmask. Each bit set means that the next
+	 * IRQ is also owned by this device.
+	 */
+	number_irqs = 0;
+	while ((irq+number_irqs < 64) &&
+	       (msi_multiple_irq_bitmask & (1ull << (irq + number_irqs))))
+		number_irqs++;
+	number_irqs++;
+	/* Mask with one bit for each IRQ */
+	bitmask = (1 << number_irqs) - 1;
+	/* Shift the mask to the correct bit location */
+	bitmask <<= irq;
+	if ((msi_free_irq_bitmask & bitmask) != bitmask)
+		panic("arch_teardown_msi_irq: Attempted to teardown MSI "
+		      "interrupt (%d) not in use", irq);
+
+	/* Checks are done, update the in use bitmask */
+	spin_lock(&msi_free_irq_bitmask_lock);
+	msi_free_irq_bitmask &= ~bitmask;
+	msi_multiple_irq_bitmask &= ~bitmask;
+	spin_unlock(&msi_free_irq_bitmask_lock);
+}
+
+
+/**
+ * Called by the interrupt handling code when an MSI interrupt
+ * occurs.
+ *
+ * @param cpl
+ * @param dev_id
+ *
+ * @return
+ */
+static irqreturn_t octeon_msi_interrupt(int cpl, void *dev_id)
+{
+	uint64_t msi_bits;
+	int irq;
+
+	if (octeon_dma_bar_type == OCTEON_DMA_BAR_TYPE_PCIE)
+		msi_bits = cvmx_read_csr(CVMX_PEXP_NPEI_MSI_RCV0);
+	else
+		msi_bits = cvmx_read_csr(CVMX_NPI_NPI_MSI_RCV);
+	irq = fls64(msi_bits);
+	if (irq) {
+		irq += OCTEON_IRQ_MSI_BIT0 - 1;
+		if (irq_desc[irq].action) {
+			do_IRQ(irq);
+			return IRQ_HANDLED;
+		} else {
+			pr_err("Spurious MSI interrupt %d\n", irq);
+			if (octeon_has_feature(OCTEON_FEATURE_PCIE)) {
+				/* These chips have PCIe */
+				cvmx_write_csr(CVMX_PEXP_NPEI_MSI_RCV0,
+					       1ull << (irq -
+							OCTEON_IRQ_MSI_BIT0));
+			} else {
+				/* These chips have PCI */
+				cvmx_write_csr(CVMX_NPI_NPI_MSI_RCV,
+					       1ull << (irq -
+							OCTEON_IRQ_MSI_BIT0));
+			}
+		}
+	}
+	return IRQ_NONE;
+}
+
+
+/**
+ * Initializes the MSI interrupt handling code
+ *
+ * @return
+ */
+int octeon_msi_initialize(void)
+{
+	int r;
+	if (octeon_has_feature(OCTEON_FEATURE_PCIE)) {
+		r = request_irq(OCTEON_IRQ_PCI_MSI0, octeon_msi_interrupt,
+				IRQF_SHARED,
+				"MSI[0:63]", octeon_msi_interrupt);
+	} else if (octeon_is_pci_host()) {
+		r = request_irq(OCTEON_IRQ_PCI_MSI0, octeon_msi_interrupt,
+				IRQF_SHARED,
+				"MSI[0:15]", octeon_msi_interrupt);
+		r += request_irq(OCTEON_IRQ_PCI_MSI1, octeon_msi_interrupt,
+				 IRQF_SHARED,
+				 "MSI[16:31]", octeon_msi_interrupt);
+		r += request_irq(OCTEON_IRQ_PCI_MSI2, octeon_msi_interrupt,
+				 IRQF_SHARED,
+				 "MSI[32:47]", octeon_msi_interrupt);
+		r += request_irq(OCTEON_IRQ_PCI_MSI3, octeon_msi_interrupt,
+				 IRQF_SHARED,
+				 "MSI[48:63]", octeon_msi_interrupt);
+	}
+	return 0;
+}
+
+subsys_initcall(octeon_msi_initialize);

arch/mips/cavium-octeon/octeon-irq.c

@@ -10,6 +10,8 @@
 #include <linux/hardirq.h>
 
 #include <asm/octeon/octeon.h>
+#include <asm/octeon/cvmx-pexp-defs.h>
+#include <asm/octeon/cvmx-npi-defs.h>
 
 DEFINE_RWLOCK(octeon_irq_ciu0_rwlock);
 DEFINE_RWLOCK(octeon_irq_ciu1_rwlock);

arch/mips/cavium-octeon/pci-common.c

@@ -0,0 +1,137 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005-2007 Cavium Networks
+ */
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/interrupt.h>
+#include <linux/time.h>
+#include <linux/delay.h>
+#include "pci-common.h"
+
+typeof(pcibios_map_irq) *octeon_pcibios_map_irq;
+enum octeon_dma_bar_type octeon_dma_bar_type = OCTEON_DMA_BAR_TYPE_INVALID;
+
+/**
+ * Map a PCI device to the appropriate interrupt line
+ *
+ * @param dev    The Linux PCI device structure for the device to map
+ * @param slot   The slot number for this device on __BUS 0__. Linux
+ *               enumerates through all the bridges and figures out the
+ *               slot on Bus 0 where this device eventually hooks to.
+ * @param pin    The PCI interrupt pin read from the device, then swizzled
+ *               as it goes through each bridge.
+ * @return Interrupt number for the device
+ */
+int __init pcibios_map_irq(const struct pci_dev *dev, u8 slot, u8 pin)
+{
+	if (octeon_pcibios_map_irq)
+		return octeon_pcibios_map_irq(dev, slot, pin);
+	else
+		panic("octeon_pcibios_map_irq doesn't point to a "
+		      "pcibios_map_irq() function");
+}
+
+
+/**
+ * Called to perform platform specific PCI setup
+ *
+ * @param dev
+ * @return
+ */
+int pcibios_plat_dev_init(struct pci_dev *dev)
+{
+	uint16_t config;
+	uint32_t dconfig;
+	int pos;
+	/*
+	 * Force the Cache line setting to 64 bytes. The standard
+	 * Linux bus scan doesn't seem to set it. Octeon really has
+	 * 128 byte lines, but Intel bridges get really upset if you
+	 * try and set values above 64 bytes. Value is specified in
+	 * 32bit words.
+	 */
+	pci_write_config_byte(dev, PCI_CACHE_LINE_SIZE, 64 / 4);
+	/* Set latency timers for all devices */
+	pci_write_config_byte(dev, PCI_LATENCY_TIMER, 48);
+
+	/* Enable reporting System errors and parity errors on all devices */
+	/* Enable parity checking and error reporting */
+	pci_read_config_word(dev, PCI_COMMAND, &config);
+	config |= PCI_COMMAND_PARITY | PCI_COMMAND_SERR;
+	pci_write_config_word(dev, PCI_COMMAND, config);
+
+	if (dev->subordinate) {
+		/* Set latency timers on sub bridges */
+		pci_write_config_byte(dev, PCI_SEC_LATENCY_TIMER, 48);
+		/* More bridge error detection */
+		pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &config);
+		config |= PCI_BRIDGE_CTL_PARITY | PCI_BRIDGE_CTL_SERR;
+		pci_write_config_word(dev, PCI_BRIDGE_CONTROL, config);
+	}
+
+	/* Enable the PCIe normal error reporting */
+	pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
+	if (pos) {
+		/* Update Device Control */
+		pci_read_config_word(dev, pos + PCI_EXP_DEVCTL, &config);
+		/* Correctable Error Reporting */
+		config |= PCI_EXP_DEVCTL_CERE;
+		/* Non-Fatal Error Reporting */
+		config |= PCI_EXP_DEVCTL_NFERE;
+		/* Fatal Error Reporting */
+		config |= PCI_EXP_DEVCTL_FERE;
+		/* Unsupported Request */
+		config |= PCI_EXP_DEVCTL_URRE;
+		pci_write_config_word(dev, pos + PCI_EXP_DEVCTL, config);
+	}
+
+	/* Find the Advanced Error Reporting capability */
+	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
+	if (pos) {
+		/* Clear Uncorrectable Error Status */
+		pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
+				      &dconfig);
+		pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
+				       dconfig);
+		/* Enable reporting of all uncorrectable errors */
+		/* Uncorrectable Error Mask - turned on bits disable errors */
+		pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, 0);
+		/*
+		 * Leave severity at HW default. This only controls if
+		 * errors are reported as uncorrectable or
+		 * correctable, not if the error is reported.
+		 */
+		/* PCI_ERR_UNCOR_SEVER - Uncorrectable Error Severity */
+		/* Clear Correctable Error Status */
+		pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &dconfig);
+		pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, dconfig);
+		/* Enable reporting of all correctable errors */
+		/* Correctable Error Mask - turned on bits disable errors */
+		pci_write_config_dword(dev, pos + PCI_ERR_COR_MASK, 0);
+		/* Advanced Error Capabilities */
+		pci_read_config_dword(dev, pos + PCI_ERR_CAP, &dconfig);
+		/* ECRC Generation Enable */
+		if (config & PCI_ERR_CAP_ECRC_GENC)
+			config |= PCI_ERR_CAP_ECRC_GENE;
+		/* ECRC Check Enable */
+		if (config & PCI_ERR_CAP_ECRC_CHKC)
+			config |= PCI_ERR_CAP_ECRC_CHKE;
+		pci_write_config_dword(dev, pos + PCI_ERR_CAP, dconfig);
+		/* PCI_ERR_HEADER_LOG - Header Log Register (16 bytes) */
+		/* Report all errors to the root complex */
+		pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND,
+				       PCI_ERR_ROOT_CMD_COR_EN |
+				       PCI_ERR_ROOT_CMD_NONFATAL_EN |
+				       PCI_ERR_ROOT_CMD_FATAL_EN);
+		/* Clear the Root status register */
+		pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, &dconfig);
+		pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, dconfig);
+	}
+
+	return 0;
+}

arch/mips/cavium-octeon/pci-common.h

@@ -0,0 +1,39 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005-2007 Cavium Networks
+ */
+#ifndef __OCTEON_PCI_COMMON_H__
+#define __OCTEON_PCI_COMMON_H__
+
+#include <linux/pci.h>
+
+/* Some PCI cards require delays when accessing config space. */
+#define PCI_CONFIG_SPACE_DELAY 10000
+
+/* pcibios_map_irq() is defined inside pci-common.c. All it does is call the
+   Octeon specific version pointed to by this variable. This function needs to
+   change for PCI or PCIe based hosts */
+extern typeof(pcibios_map_irq) *octeon_pcibios_map_irq;
+
+/* The following defines are only used when octeon_dma_bar_type =
+   OCTEON_DMA_BAR_TYPE_BIG */
+#define OCTEON_PCI_BAR1_HOLE_BITS 5
+#define OCTEON_PCI_BAR1_HOLE_SIZE (1ul<<(OCTEON_PCI_BAR1_HOLE_BITS+3))
+
+enum octeon_dma_bar_type {
+	OCTEON_DMA_BAR_TYPE_INVALID,
+	OCTEON_DMA_BAR_TYPE_SMALL,
+	OCTEON_DMA_BAR_TYPE_BIG,
+	OCTEON_DMA_BAR_TYPE_PCIE
+};
+
+/**
+ * This is a variable to tell the DMA mapping system in dma-octeon.c
+ * how to map PCI DMA addresses.
+ */
+extern enum octeon_dma_bar_type octeon_dma_bar_type;
+
+#endif

arch/mips/include/asm/octeon/cvmx-helper-errata.h

@@ -0,0 +1,33 @@
+/***********************license start***************
+ * Author: Cavium Networks
+ *
+ * Contact: support@caviumnetworks.com
+ * This file is part of the OCTEON SDK
+ *
+ * Copyright (c) 2003-2008 Cavium Networks
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, Version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful, but
+ * AS-IS and WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, TITLE, or
+ * NONINFRINGEMENT.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this file; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * or visit http://www.gnu.org/licenses/.
+ *
+ * This file may also be available under a different license from Cavium.
+ * Contact Cavium Networks for more information
+ ***********************license end**************************************/
+
+#ifndef __CVMX_HELPER_ERRATA_H__
+#define __CVMX_HELPER_ERRATA_H__
+
+extern void __cvmx_helper_errata_qlm_disable_2nd_order_cdr(int qlm);
+
+#endif

arch/mips/include/asm/octeon/cvmx-helper-jtag.h

@@ -0,0 +1,43 @@
+/***********************license start***************
+ * Author: Cavium Networks
+ *
+ * Contact: support@caviumnetworks.com
+ * This file is part of the OCTEON SDK
+ *
+ * Copyright (c) 2003-2008 Cavium Networks
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, Version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful, but
+ * AS-IS and WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, TITLE, or
+ * NONINFRINGEMENT.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this file; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * or visit http://www.gnu.org/licenses/.
+ *
+ * This file may also be available under a different license from Cavium.
+ * Contact Cavium Networks for more information
+ ***********************license end**************************************/
+
+/**
+ * @file
+ *
+ *  Helper utilities for qlm_jtag.
+ *
+ */
+
+#ifndef __CVMX_HELPER_JTAG_H__
+#define __CVMX_HELPER_JTAG_H__
+
+extern void cvmx_helper_qlm_jtag_init(void);
+extern uint32_t cvmx_helper_qlm_jtag_shift(int qlm, int bits, uint32_t data);
+extern void cvmx_helper_qlm_jtag_shift_zeros(int qlm, int bits);
+extern void cvmx_helper_qlm_jtag_update(int qlm);
+
+#endif /* __CVMX_HELPER_JTAG_H__ */

arch/mips/include/asm/octeon/cvmx.h

@@ -375,6 +375,18 @@ static inline uint64_t cvmx_get_cycle(void)
 	return cycle;
 }
 
+/**
+ * Wait for the specified number of cycle
+ *
+ */
+static inline void cvmx_wait(uint64_t cycles)
+{
+	uint64_t done = cvmx_get_cycle() + cycles;
+
+	while (cvmx_get_cycle() < done)
+		; /* Spin */
+}
+
 /**
  * Reads a chip global cycle counter.  This counts CPU cycles since
  * chip reset.  The counter is 64 bit.

arch/mips/include/asm/octeon/octeon.h

@@ -245,4 +245,6 @@ static inline uint32_t octeon_npi_read32(uint64_t address)
 	return cvmx_read64_uint32(address ^ 4);
 }
 
+extern struct cvmx_bootinfo *octeon_bootinfo;
+
 #endif /* __ASM_OCTEON_OCTEON_H */