diff options
Diffstat (limited to 'arch/powerpc/sysdev')
| -rw-r--r-- | arch/powerpc/sysdev/Makefile | 7 | ||||
| -rw-r--r-- | arch/powerpc/sysdev/dcr.S | 41 | ||||
| -rw-r--r-- | arch/powerpc/sysdev/grackle.c | 64 | ||||
| -rw-r--r-- | arch/powerpc/sysdev/i8259.c | 221 | ||||
| -rw-r--r-- | arch/powerpc/sysdev/indirect_pci.c | 134 | ||||
| -rw-r--r-- | arch/powerpc/sysdev/mpic.c | 927 | ||||
| -rw-r--r-- | arch/powerpc/sysdev/u3_iommu.c | 327 |
7 files changed, 1721 insertions, 0 deletions
diff --git a/arch/powerpc/sysdev/Makefile b/arch/powerpc/sysdev/Makefile new file mode 100644 index 000000000000..8acd21dee05d --- /dev/null +++ b/arch/powerpc/sysdev/Makefile @@ -0,0 +1,7 @@ +obj-$(CONFIG_MPIC) += mpic.o +obj-$(CONFIG_PPC_INDIRECT_PCI) += indirect_pci.o +obj-$(CONFIG_PPC_I8259) += i8259.o +obj-$(CONFIG_PPC_MPC106) += grackle.o +obj-$(CONFIG_BOOKE) += dcr.o +obj-$(CONFIG_40x) += dcr.o +obj-$(CONFIG_U3_DART) += u3_iommu.o diff --git a/arch/powerpc/sysdev/dcr.S b/arch/powerpc/sysdev/dcr.S new file mode 100644 index 000000000000..895f10243a43 --- /dev/null +++ b/arch/powerpc/sysdev/dcr.S @@ -0,0 +1,41 @@ +/* + * arch/ppc/syslib/dcr.S + * + * "Indirect" DCR access + * + * Copyright (c) 2004 Eugene Surovegin <ebs@ebshome.net> + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License as published by the + * Free Software Foundation; either version 2 of the License, or (at your + * option) any later version. + */ + +#include <asm/ppc_asm.h> +#include <asm/processor.h> + +#define DCR_ACCESS_PROLOG(table) \ + rlwinm r3,r3,4,18,27; \ + lis r5,table@h; \ + ori r5,r5,table@l; \ + add r3,r3,r5; \ + mtctr r3; \ + bctr + +_GLOBAL(__mfdcr) + DCR_ACCESS_PROLOG(__mfdcr_table) + +_GLOBAL(__mtdcr) + DCR_ACCESS_PROLOG(__mtdcr_table) + +__mfdcr_table: + mfdcr r3,0; blr +__mtdcr_table: + mtdcr 0,r4; blr + +dcr = 1 + .rept 1023 + mfdcr r3,dcr; blr + mtdcr dcr,r4; blr + dcr = dcr + 1 + .endr diff --git a/arch/powerpc/sysdev/grackle.c b/arch/powerpc/sysdev/grackle.c new file mode 100644 index 000000000000..b6ec793a23be --- /dev/null +++ b/arch/powerpc/sysdev/grackle.c @@ -0,0 +1,64 @@ +/* + * Functions for setting up and using a MPC106 northbridge + * Extracted from arch/powerpc/platforms/powermac/pci.c. + * + * Copyright (C) 2003 Benjamin Herrenschmuidt (benh@kernel.crashing.org) + * Copyright (C) 1997 Paul Mackerras (paulus@samba.org) + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include <linux/kernel.h> +#include <linux/pci.h> +#include <linux/init.h> + +#include <asm/io.h> +#include <asm/prom.h> +#include <asm/pci-bridge.h> +#include <asm/grackle.h> + +#define GRACKLE_CFA(b, d, o) (0x80 | ((b) << 8) | ((d) << 16) \ + | (((o) & ~3) << 24)) + +#define GRACKLE_PICR1_STG 0x00000040 +#define GRACKLE_PICR1_LOOPSNOOP 0x00000010 + +/* N.B. this is called before bridges is initialized, so we can't + use grackle_pcibios_{read,write}_config_dword. */ +static inline void grackle_set_stg(struct pci_controller* bp, int enable) +{ + unsigned int val; + + out_be32(bp->cfg_addr, GRACKLE_CFA(0, 0, 0xa8)); + val = in_le32(bp->cfg_data); + val = enable? (val | GRACKLE_PICR1_STG) : + (val & ~GRACKLE_PICR1_STG); + out_be32(bp->cfg_addr, GRACKLE_CFA(0, 0, 0xa8)); + out_le32(bp->cfg_data, val); + (void)in_le32(bp->cfg_data); +} + +static inline void grackle_set_loop_snoop(struct pci_controller *bp, int enable) +{ + unsigned int val; + + out_be32(bp->cfg_addr, GRACKLE_CFA(0, 0, 0xa8)); + val = in_le32(bp->cfg_data); + val = enable? (val | GRACKLE_PICR1_LOOPSNOOP) : + (val & ~GRACKLE_PICR1_LOOPSNOOP); + out_be32(bp->cfg_addr, GRACKLE_CFA(0, 0, 0xa8)); + out_le32(bp->cfg_data, val); + (void)in_le32(bp->cfg_data); +} + +void __init setup_grackle(struct pci_controller *hose) +{ + setup_indirect_pci(hose, 0xfec00000, 0xfee00000); + if (machine_is_compatible("AAPL,PowerBook1998")) + grackle_set_loop_snoop(hose, 1); +#if 0 /* Disabled for now, HW problems ??? */ + grackle_set_stg(hose, 1); +#endif +} diff --git a/arch/powerpc/sysdev/i8259.c b/arch/powerpc/sysdev/i8259.c new file mode 100644 index 000000000000..90bce6e0c191 --- /dev/null +++ b/arch/powerpc/sysdev/i8259.c @@ -0,0 +1,221 @@ +/* + * i8259 interrupt controller driver. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include <linux/init.h> +#include <linux/ioport.h> +#include <linux/interrupt.h> +#include <asm/io.h> +#include <asm/i8259.h> + +static volatile void __iomem *pci_intack; /* RO, gives us the irq vector */ + +static unsigned char cached_8259[2] = { 0xff, 0xff }; +#define cached_A1 (cached_8259[0]) +#define cached_21 (cached_8259[1]) + +static DEFINE_SPINLOCK(i8259_lock); + +static int i8259_pic_irq_offset; + +/* + * Acknowledge the IRQ using either the PCI host bridge's interrupt + * acknowledge feature or poll. How i8259_init() is called determines + * which is called. It should be noted that polling is broken on some + * IBM and Motorola PReP boxes so we must use the int-ack feature on them. + */ +int i8259_irq(struct pt_regs *regs) +{ + int irq; + + spin_lock(&i8259_lock); + + /* Either int-ack or poll for the IRQ */ + if (pci_intack) + irq = readb(pci_intack); + else { + /* Perform an interrupt acknowledge cycle on controller 1. */ + outb(0x0C, 0x20); /* prepare for poll */ + irq = inb(0x20) & 7; + if (irq == 2 ) { + /* + * Interrupt is cascaded so perform interrupt + * acknowledge on controller 2. + */ + outb(0x0C, 0xA0); /* prepare for poll */ + irq = (inb(0xA0) & 7) + 8; + } + } + + if (irq == 7) { + /* + * This may be a spurious interrupt. + * + * Read the interrupt status register (ISR). If the most + * significant bit is not set then there is no valid + * interrupt. + */ + if (!pci_intack) + outb(0x0B, 0x20); /* ISR register */ + if(~inb(0x20) & 0x80) + irq = -1; + } + + spin_unlock(&i8259_lock); + return irq + i8259_pic_irq_offset; +} + +int i8259_irq_cascade(struct pt_regs *regs, void *unused) +{ + return i8259_irq(regs); +} + +static void i8259_mask_and_ack_irq(unsigned int irq_nr) +{ + unsigned long flags; + + spin_lock_irqsave(&i8259_lock, flags); + irq_nr -= i8259_pic_irq_offset; + if (irq_nr > 7) { + cached_A1 |= 1 << (irq_nr-8); + inb(0xA1); /* DUMMY */ + outb(cached_A1, 0xA1); + outb(0x20, 0xA0); /* Non-specific EOI */ + outb(0x20, 0x20); /* Non-specific EOI to cascade */ + } else { + cached_21 |= 1 << irq_nr; + inb(0x21); /* DUMMY */ + outb(cached_21, 0x21); + outb(0x20, 0x20); /* Non-specific EOI */ + } + spin_unlock_irqrestore(&i8259_lock, flags); +} + +static void i8259_set_irq_mask(int irq_nr) +{ + outb(cached_A1,0xA1); + outb(cached_21,0x21); +} + +static void i8259_mask_irq(unsigned int irq_nr) +{ + unsigned long flags; + + spin_lock_irqsave(&i8259_lock, flags); + irq_nr -= i8259_pic_irq_offset; + if (irq_nr < 8) + cached_21 |= 1 << irq_nr; + else + cached_A1 |= 1 << (irq_nr-8); + i8259_set_irq_mask(irq_nr); + spin_unlock_irqrestore(&i8259_lock, flags); +} + +static void i8259_unmask_irq(unsigned int irq_nr) +{ + unsigned long flags; + + spin_lock_irqsave(&i8259_lock, flags); + irq_nr -= i8259_pic_irq_offset; + if (irq_nr < 8) + cached_21 &= ~(1 << irq_nr); + else + cached_A1 &= ~(1 << (irq_nr-8)); + i8259_set_irq_mask(irq_nr); + spin_unlock_irqrestore(&i8259_lock, flags); +} + +static void i8259_end_irq(unsigned int irq) +{ + if (!(irq_desc[irq].status & (IRQ_DISABLED|IRQ_INPROGRESS)) + && irq_desc[irq].action) + i8259_unmask_irq(irq); +} + +struct hw_interrupt_type i8259_pic = { + .typename = " i8259 ", + .enable = i8259_unmask_irq, + .disable = i8259_mask_irq, + .ack = i8259_mask_and_ack_irq, + .end = i8259_end_irq, +}; + +static struct resource pic1_iores = { + .name = "8259 (master)", + .start = 0x20, + .end = 0x21, + .flags = IORESOURCE_BUSY, +}; + +static struct resource pic2_iores = { + .name = "8259 (slave)", + .start = 0xa0, + .end = 0xa1, + .flags = IORESOURCE_BUSY, +}; + +static struct resource pic_edgectrl_iores = { + .name = "8259 edge control", + .start = 0x4d0, + .end = 0x4d1, + .flags = IORESOURCE_BUSY, +}; + +static struct irqaction i8259_irqaction = { + .handler = no_action, + .flags = SA_INTERRUPT, + .mask = CPU_MASK_NONE, + .name = "82c59 secondary cascade", +}; + +/* + * i8259_init() + * intack_addr - PCI interrupt acknowledge (real) address which will return + * the active irq from the 8259 + */ +void __init i8259_init(unsigned long intack_addr, int offset) +{ + unsigned long flags; + int i; + + spin_lock_irqsave(&i8259_lock, flags); + i8259_pic_irq_offset = offset; + + /* init master interrupt controller */ + outb(0x11, 0x20); /* Start init sequence */ + outb(0x00, 0x21); /* Vector base */ + outb(0x04, 0x21); /* edge tiggered, Cascade (slave) on IRQ2 */ + outb(0x01, 0x21); /* Select 8086 mode */ + + /* init slave interrupt controller */ + outb(0x11, 0xA0); /* Start init sequence */ + outb(0x08, 0xA1); /* Vector base */ + outb(0x02, 0xA1); /* edge triggered, Cascade (slave) on IRQ2 */ + outb(0x01, 0xA1); /* Select 8086 mode */ + + /* always read ISR */ + outb(0x0B, 0x20); + outb(0x0B, 0xA0); + + /* Mask all interrupts */ + outb(cached_A1, 0xA1); + outb(cached_21, 0x21); + + spin_unlock_irqrestore(&i8259_lock, flags); + + /* reserve our resources */ + setup_irq(offset + 2, &i8259_irqaction); + request_resource(&ioport_resource, &pic1_iores); + request_resource(&ioport_resource, &pic2_iores); + request_resource(&ioport_resource, &pic_edgectrl_iores); + + if (intack_addr != 0) + pci_intack = ioremap(intack_addr, 1); + + for (i = 0; i < NUM_ISA_INTERRUPTS; ++i) + irq_desc[offset + i].handler = &i8259_pic; +} diff --git a/arch/powerpc/sysdev/indirect_pci.c b/arch/powerpc/sysdev/indirect_pci.c new file mode 100644 index 000000000000..e71488469704 --- /dev/null +++ b/arch/powerpc/sysdev/indirect_pci.c @@ -0,0 +1,134 @@ +/* + * Support for indirect PCI bridges. + * + * Copyright (C) 1998 Gabriel Paubert. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#include <linux/kernel.h> +#include <linux/pci.h> +#include <linux/delay.h> +#include <linux/string.h> +#include <linux/init.h> + +#include <asm/io.h> +#include <asm/prom.h> +#include <asm/pci-bridge.h> +#include <asm/machdep.h> + +#ifdef CONFIG_PPC_INDIRECT_PCI_BE +#define PCI_CFG_OUT out_be32 +#else +#define PCI_CFG_OUT out_le32 +#endif + +static int +indirect_read_config(struct pci_bus *bus, unsigned int devfn, int offset, + int len, u32 *val) +{ + struct pci_controller *hose = bus->sysdata; + volatile void __iomem *cfg_data; + u8 cfg_type = 0; + + if (ppc_md.pci_exclude_device) + if (ppc_md.pci_exclude_device(bus->number, devfn)) + return PCIBIOS_DEVICE_NOT_FOUND; + + if (hose->set_cfg_type) + if (bus->number != hose->first_busno) + cfg_type = 1; + + PCI_CFG_OUT(hose->cfg_addr, + (0x80000000 | ((bus->number - hose->bus_offset) << 16) + | (devfn << 8) | ((offset & 0xfc) | cfg_type))); + + /* + * Note: the caller has already checked that offset is + * suitably aligned and that len is 1, 2 or 4. + */ + cfg_data = hose->cfg_data + (offset & 3); + switch (len) { + case 1: + *val = in_8(cfg_data); + break; + case 2: + *val = in_le16(cfg_data); + break; + default: + *val = in_le32(cfg_data); + break; + } + return PCIBIOS_SUCCESSFUL; +} + +static int +indirect_write_config(struct pci_bus *bus, unsigned int devfn, int offset, + int len, u32 val) +{ + struct pci_controller *hose = bus->sysdata; + volatile void __iomem *cfg_data; + u8 cfg_type = 0; + + if (ppc_md.pci_exclude_device) + if (ppc_md.pci_exclude_device(bus->number, devfn)) + return PCIBIOS_DEVICE_NOT_FOUND; + + if (hose->set_cfg_type) + if (bus->number != hose->first_busno) + cfg_type = 1; + + PCI_CFG_OUT(hose->cfg_addr, + (0x80000000 | ((bus->number - hose->bus_offset) << 16) + | (devfn << 8) | ((offset & 0xfc) | cfg_type))); + + /* + * Note: the caller has already checked that offset is + * suitably aligned and that len is 1, 2 or 4. + */ + cfg_data = hose->cfg_data + (offset & 3); + switch (len) { + case 1: + out_8(cfg_data, val); + break; + case 2: + out_le16(cfg_data, val); + break; + default: + out_le32(cfg_data, val); + break; + } + return PCIBIOS_SUCCESSFUL; +} + +static struct pci_ops indirect_pci_ops = +{ + indirect_read_config, + indirect_write_config +}; + +void __init +setup_indirect_pci_nomap(struct pci_controller* hose, void __iomem * cfg_addr, + void __iomem * cfg_data) +{ + hose->cfg_addr = cfg_addr; + hose->cfg_data = cfg_data; + hose->ops = &indirect_pci_ops; +} + +void __init +setup_indirect_pci(struct pci_controller* hose, u32 cfg_addr, u32 cfg_data) +{ + unsigned long base = cfg_addr & PAGE_MASK; + void __iomem *mbase, *addr, *data; + + mbase = ioremap(base, PAGE_SIZE); + addr = mbase + (cfg_addr & ~PAGE_MASK); + if ((cfg_data & PAGE_MASK) != base) + mbase = ioremap(cfg_data & PAGE_MASK, PAGE_SIZE); + data = mbase + (cfg_data & ~PAGE_MASK); + setup_indirect_pci_nomap(hose, addr, data); +} diff --git a/arch/powerpc/sysdev/mpic.c b/arch/powerpc/sysdev/mpic.c new file mode 100644 index 000000000000..105f05341a41 --- /dev/null +++ b/arch/powerpc/sysdev/mpic.c @@ -0,0 +1,927 @@ +/* + * arch/powerpc/kernel/mpic.c + * + * Driver for interrupt controllers following the OpenPIC standard, the + * common implementation beeing IBM's MPIC. This driver also can deal + * with various broken implementations of this HW. + * + * Copyright (C) 2004 Benjamin Herrenschmidt, IBM Corp. + * + * 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. + */ + +#undef DEBUG + +#include <linux/config.h> +#include <linux/types.h> +#include <linux/kernel.h> +#include <linux/init.h> +#include <linux/irq.h> +#include <linux/smp.h> +#include <linux/interrupt.h> +#include <linux/bootmem.h> +#include <linux/spinlock.h> +#include <linux/pci.h> + +#include <asm/ptrace.h> +#include <asm/signal.h> +#include <asm/io.h> +#include <asm/pgtable.h> +#include <asm/irq.h> +#include <asm/machdep.h> +#include <asm/mpic.h> +#include <asm/smp.h> + +#ifdef DEBUG +#define DBG(fmt...) printk(fmt) +#else +#define DBG(fmt...) +#endif + +static struct mpic *mpics; +static struct mpic *mpic_primary; +static DEFINE_SPINLOCK(mpic_lock); + +#ifdef CONFIG_PPC32 /* XXX for now */ +#define distribute_irqs CONFIG_IRQ_ALL_CPUS +#endif + +/* + * Register accessor functions + */ + + +static inline u32 _mpic_read(unsigned int be, volatile u32 __iomem *base, + unsigned int reg) +{ + if (be) + return in_be32(base + (reg >> 2)); + else + return in_le32(base + (reg >> 2)); +} + +static inline void _mpic_write(unsigned int be, volatile u32 __iomem *base, + unsigned int reg, u32 value) +{ + if (be) + out_be32(base + (reg >> 2), value); + else + out_le32(base + (reg >> 2), value); +} + +static inline u32 _mpic_ipi_read(struct mpic *mpic, unsigned int ipi) +{ + unsigned int be = (mpic->flags & MPIC_BIG_ENDIAN) != 0; + unsigned int offset = MPIC_GREG_IPI_VECTOR_PRI_0 + (ipi * 0x10); + + if (mpic->flags & MPIC_BROKEN_IPI) + be = !be; + return _mpic_read(be, mpic->gregs, offset); +} + +static inline void _mpic_ipi_write(struct mpic *mpic, unsigned int ipi, u32 value) +{ + unsigned int offset = MPIC_GREG_IPI_VECTOR_PRI_0 + (ipi * 0x10); + + _mpic_write(mpic->flags & MPIC_BIG_ENDIAN, mpic->gregs, offset, value); +} + +static inline u32 _mpic_cpu_read(struct mpic *mpic, unsigned int reg) +{ + unsigned int cpu = 0; + + if (mpic->flags & MPIC_PRIMARY) + cpu = hard_smp_processor_id(); + + return _mpic_read(mpic->flags & MPIC_BIG_ENDIAN, mpic->cpuregs[cpu], reg); +} + +static inline void _mpic_cpu_write(struct mpic *mpic, unsigned int reg, u32 value) +{ + unsigned int cpu = 0; + + if (mpic->flags & MPIC_PRIMARY) + cpu = hard_smp_processor_id(); + + _mpic_write(mpic->flags & MPIC_BIG_ENDIAN, mpic->cpuregs[cpu], reg, value); +} + +static inline u32 _mpic_irq_read(struct mpic *mpic, unsigned int src_no, unsigned int reg) +{ + unsigned int isu = src_no >> mpic->isu_shift; + unsigned int idx = src_no & mpic->isu_mask; + + return _mpic_read(mpic->flags & MPIC_BIG_ENDIAN, mpic->isus[isu], + reg + (idx * MPIC_IRQ_STRIDE)); +} + +static inline void _mpic_irq_write(struct mpic *mpic, unsigned int src_no, + unsigned int reg, u32 value) +{ + unsigned int isu = src_no >> mpic->isu_shift; + unsigned int idx = src_no & mpic->isu_mask; + + _mpic_write(mpic->flags & MPIC_BIG_ENDIAN, mpic->isus[isu], + reg + (idx * MPIC_IRQ_STRIDE), value); +} + +#define mpic_read(b,r) _mpic_read(mpic->flags & MPIC_BIG_ENDIAN,(b),(r)) +#define mpic_write(b,r,v) _mpic_write(mpic->flags & MPIC_BIG_ENDIAN,(b),(r),(v)) +#define mpic_ipi_read(i) _mpic_ipi_read(mpic,(i)) +#define mpic_ipi_write(i,v) _mpic_ipi_write(mpic,(i),(v)) +#define mpic_cpu_read(i) _mpic_cpu_read(mpic,(i)) +#define mpic_cpu_write(i,v) _mpic_cpu_write(mpic,(i),(v)) +#define mpic_irq_read(s,r) _mpic_irq_read(mpic,(s),(r)) +#define mpic_irq_write(s,r,v) _mpic_irq_write(mpic,(s),(r),(v)) + + +/* + * Low level utility functions + */ + + + +/* Check if we have one of those nice broken MPICs with a flipped endian on + * reads from IPI registers + */ +static void __init mpic_test_broken_ipi(struct mpic *mpic) +{ + u32 r; + + mpic_write(mpic->gregs, MPIC_GREG_IPI_VECTOR_PRI_0, MPIC_VECPRI_MASK); + r = mpic_read(mpic->gregs, MPIC_GREG_IPI_VECTOR_PRI_0); + + if (r == le32_to_cpu(MPIC_VECPRI_MASK)) { + printk(KERN_INFO "mpic: Detected reversed IPI registers\n"); + mpic->flags |= MPIC_BROKEN_IPI; + } +} + +#ifdef CONFIG_MPIC_BROKEN_U3 + +/* Test if an interrupt is sourced from HyperTransport (used on broken U3s) + * to force the edge setting on the MPIC and do the ack workaround. + */ +static inline int mpic_is_ht_interrupt(struct mpic *mpic, unsigned int source_no) +{ + if (source_no >= 128 || !mpic->fixups) + return 0; + return mpic->fixups[source_no].base != NULL; +} + +static inline void mpic_apic_end_irq(struct mpic *mpic, unsigned int source_no) +{ + struct mpic_irq_fixup *fixup = &mpic->fixups[source_no]; + u32 tmp; + + spin_lock(&mpic->fixup_lock); + writeb(0x11 + 2 * fixup->irq, fixup->base); + tmp = readl(fixup->base + 2); + writel(tmp | 0x80000000ul, fixup->base + 2); + /* config writes shouldn't be posted but let's be safe ... */ + (void)readl(fixup->base + 2); + spin_unlock(&mpic->fixup_lock); +} + + +static void __init mpic_amd8111_read_irq(struct mpic *mpic, u8 __iomem *devbase) +{ + int i, irq; + u32 tmp; + + printk(KERN_INFO "mpic: - Workarounds on AMD 8111 @ %p\n", devbase); + + for (i=0; i < 24; i++) { + writeb(0x10 + 2*i, devbase + 0xf2); + tmp = readl(devbase + 0xf4); + if ((tmp & 0x1) || !(tmp & 0x20)) + continue; + irq = (tmp >> 16) & 0xff; + mpic->fixups[irq].irq = i; + mpic->fixups[irq].base = devbase + 0xf2; + } +} + +static void __init mpic_amd8131_read_irq(struct mpic *mpic, u8 __iomem *devbase) +{ + int i, irq; + u32 tmp; + + printk(KERN_INFO "mpic: - Workarounds on AMD 8131 @ %p\n", devbase); + + for (i=0; i < 4; i++) { + writeb(0x10 + 2*i, devbase + 0xba); + tmp = readl(devbase + 0xbc); + if ((tmp & 0x1) || !(tmp & 0x20)) + continue; + irq = (tmp >> 16) & 0xff; + mpic->fixups[irq].irq = i; + mpic->fixups[irq].base = devbase + 0xba; + } +} + +static void __init mpic_scan_ioapics(struct mpic *mpic) +{ + unsigned int devfn; + u8 __iomem *cfgspace; + + printk(KERN_INFO "mpic: Setting up IO-APICs workarounds for U3\n"); + + /* Allocate fixups array */ + mpic->fixups = alloc_bootmem(128 * sizeof(struct mpic_irq_fixup)); + BUG_ON(mpic->fixups == NULL); + memset(mpic->fixups, 0, 128 * sizeof(struct mpic_irq_fixup)); + + /* Init spinlock */ + spin_lock_init(&mpic->fixup_lock); + + /* Map u3 config space. We assume all IO-APICs are on the primary bus + * and slot will never be above "0xf" so we only need to map 32k + */ + cfgspace = (unsigned char __iomem *)ioremap(0xf2000000, 0x8000); + BUG_ON(cfgspace == NULL); + + /* Now we scan all slots. We do a very quick scan, we read the header type, + * vendor ID and device ID only, that's plenty enough + */ + for (devfn = 0; devfn < PCI_DEVFN(0x10,0); devfn ++) { + u8 __iomem *devbase = cfgspace + (devfn << 8); + u8 hdr_type = readb(devbase + PCI_HEADER_TYPE); + u32 l = readl(devbase + PCI_VENDOR_ID); + u16 vendor_id, device_id; + int multifunc = 0; + + DBG("devfn %x, l: %x\n", devfn, l); + + /* If no device, skip */ + if (l == 0xffffffff || l == 0x00000000 || + l == 0x0000ffff || l == 0xffff0000) + goto next; + + /* Check if it's a multifunction device (only really used + * to function 0 though + */ + multifunc = !!(hdr_type & 0x80); + vendor_id = l & 0xffff; + device_id = (l >> 16) & 0xffff; + + /* If a known device, go to fixup setup code */ + if (vendor_id == PCI_VENDOR_ID_AMD && device_id == 0x7460) + mpic_amd8111_read_irq(mpic, devbase); + if (vendor_id == PCI_VENDOR_ID_AMD && device_id == 0x7450) + mpic_amd8131_read_irq(mpic, devbase); + next: + /* next device, if function 0 */ + if ((PCI_FUNC(devfn) == 0) && !multifunc) + devfn += 7; + } +} + +#endif /* CONFIG_MPIC_BROKEN_U3 */ + + +/* Find an mpic associated with a given linux interrupt */ +static struct mpic *mpic_find(unsigned int irq, unsigned int *is_ipi) +{ + struct mpic *mpic = mpics; + + while(mpic) { + /* search IPIs first since they may override the main interrupts */ + if (irq >= mpic->ipi_offset && irq < (mpic->ipi_offset + 4)) { + if (is_ipi) + *is_ipi = 1; + return mpic; + } + if (irq >= mpic->irq_offset && + irq < (mpic->irq_offset + mpic->irq_count)) { + if (is_ipi) + *is_ipi = 0; + return mpic; + } + mpic = mpic -> next; + } + return NULL; +} + +/* Convert a cpu mask from logical to physical cpu numbers. */ +static inline u32 mpic_physmask(u32 cpumask) +{ + int i; + u32 mask = 0; + + for (i = 0; i < NR_CPUS; ++i, cpumask >>= 1) + mask |= (cpumask & 1) << get_hard_smp_processor_id(i); + return mask; +} + +#ifdef CONFIG_SMP +/* Get the mpic structure from the IPI number */ +static inline struct mpic * mpic_from_ipi(unsigned int ipi) +{ + return container_of(irq_desc[ipi].handler, struct mpic, hc_ipi); +} +#endif + +/* Get the mpic structure from the irq number */ +static inline struct mpic * mpic_from_irq(unsigned int irq) +{ + return container_of(irq_desc[irq].handler, struct mpic, hc_irq); +} + +/* Send an EOI */ +static inline void mpic_eoi(struct mpic *mpic) +{ + mpic_cpu_write(MPIC_CPU_EOI, 0); + (void)mpic_cpu_read(MPIC_CPU_WHOAMI); +} + +#ifdef CONFIG_SMP +static irqreturn_t mpic_ipi_action(int irq, void *dev_id, struct pt_regs *regs) +{ + struct mpic *mpic = dev_id; + + smp_message_recv(irq - mpic->ipi_offset, regs); + return IRQ_HANDLED; +} +#endif /* CONFIG_SMP */ + +/* + * Linux descriptor level callbacks + */ + + +static void mpic_enable_irq(unsigned int irq) +{ + unsigned int loops = 100000; + struct mpic *mpic = mpic_from_irq(irq); + unsigned int src = irq - mpic->irq_offset; + + DBG("%p: %s: enable_irq: %d (src %d)\n", mpic, mpic->name, irq, src); + + mpic_irq_write(src, MPIC_IRQ_VECTOR_PRI, + mpic_irq_read(src, MPIC_IRQ_VECTOR_PRI) & ~MPIC_VECPRI_MASK); + + /* make sure mask gets to controller before we return to user */ + do { + if (!loops--) { + printk(KERN_ERR "mpic_enable_irq timeout\n"); + break; + } + } while(mpic_irq_read(src, MPIC_IRQ_VECTOR_PRI) & MPIC_VECPRI_MASK); +} + +static void mpic_disable_irq(unsigned int irq) +{ + unsigned int loops = 100000; + struct mpic *mpic = mpic_from_irq(irq); + unsigned int src = irq - mpic->irq_offset; + + DBG("%s: disable_irq: %d (src %d)\n", mpic->name, irq, src); + + mpic_irq_write(src, MPIC_IRQ_VECTOR_PRI, + mpic_irq_read(src, MPIC_IRQ_VECTOR_PRI) | MPIC_VECPRI_MASK); + + /* make sure mask gets to controller before we return to user */ + do { + if (!loops--) { + printk(KERN_ERR "mpic_enable_irq timeout\n"); + break; + } + } while(!(mpic_irq_read(src, MPIC_IRQ_VECTOR_PRI) & MPIC_VECPRI_MASK)); +} + +static void mpic_end_irq(unsigned int irq) +{ + struct mpic *mpic = mpic_from_irq(irq); + + DBG("%s: end_irq: %d\n", mpic->name, irq); + + /* We always EOI on end_irq() even for edge interrupts since that + * should only lower the priority, the MPIC should have properly + * latched another edge interrupt coming in anyway + */ + +#ifdef CONFIG_MPIC_BROKEN_U3 + if (mpic->flags & MPIC_BROKEN_U3) { + unsigned int src = irq - mpic->irq_offset; + if (mpic_is_ht_interrupt(mpic, src)) + mpic_apic_end_irq(mpic, src); + } +#endif /* CONFIG_MPIC_BROKEN_U3 */ + + mpic_eoi(mpic); +} + +#ifdef CONFIG_SMP + +static void mpic_enable_ipi(unsigned int irq) +{ + struct mpic *mpic = mpic_from_ipi(irq); + unsigned int src = irq - mpic->ipi_offset; + + DBG("%s: enable_ipi: %d (ipi %d)\n", mpic->name, irq, src); + mpic_ipi_write(src, mpic_ipi_read(src) & ~MPIC_VECPRI_MASK); +} + +static void mpic_disable_ipi(unsigned int irq) +{ + /* NEVER disable an IPI... that's just plain wrong! */ +} + +static void mpic_end_ipi(unsigned int irq) +{ + struct mpic *mpic = mpic_from_ipi(irq); + + /* + * IPIs are marked IRQ_PER_CPU. This has the side effect of + * preventing the IRQ_PENDING/IRQ_INPROGRESS logic from + * applying to them. We EOI them late to avoid re-entering. + * We mark IPI's with SA_INTERRUPT as they must run with + * irqs disabled. + */ + mpic_eoi(mpic); +} + +#endif /* CONFIG_SMP */ + +static void mpic_set_affinity(unsigned int irq, cpumask_t cpumask) +{ + struct mpic *mpic = mpic_from_irq(irq); + + cpumask_t tmp; + + cpus_and(tmp, cpumask, cpu_online_map); + + mpic_irq_write(irq - mpic->irq_offset, MPIC_IRQ_DESTINATION, + mpic_physmask(cpus_addr(tmp)[0])); +} + + +/* + * Exported functions + */ + + +struct mpic * __init mpic_alloc(unsigned long phys_addr, + unsigned int flags, + unsigned int isu_size, + unsigned int irq_offset, + unsigned int irq_count, + unsigned int ipi_offset, + unsigned char *senses, + unsigned int senses_count, + const char *name) +{ + struct mpic *mpic; + u32 reg; + const char *vers; + int i; + + mpic = alloc_bootmem(sizeof(struct mpic)); + if (mpic == NULL) + return NULL; + + + memset(mpic, 0, sizeof(struct mpic)); + mpic->name = name; + + mpic->hc_irq.typename = name; + mpic->hc_irq.enable = mpic_enable_irq; + mpic->hc_irq.disable = mpic_disable_irq; + mpic->hc_irq.end = mpic_end_irq; + if (flags & MPIC_PRIMARY) + mpic->hc_irq.set_affinity = mpic_set_affinity; +#ifdef CONFIG_SMP + mpic->hc_ipi.typename = name; + mpic->hc_ipi.enable = mpic_enable_ipi; + mpic->hc_ipi.disable = mpic_disable_ipi; + mpic->hc_ipi.end = mpic_end_ipi; +#endif /* CONFIG_SMP */ + + mpic->flags = flags; + mpic->isu_size = isu_size; + mpic->irq_offset = irq_offset; + mpic->irq_count = irq_count; + mpic->ipi_offset = ipi_offset; + mpic->num_sources = 0; /* so far */ + mpic->senses = senses; + mpic->senses_count = senses_count; + + /* Map the global registers */ + mpic->gregs = ioremap(phys_addr + MPIC_GREG_BASE, 0x1000); + mpic->tmregs = mpic->gregs + ((MPIC_TIMER_BASE - MPIC_GREG_BASE) >> 2); + BUG_ON(mpic->gregs == NULL); + + /* Reset */ + if (flags & MPIC_WANTS_RESET) { + mpic_write(mpic->gregs, MPIC_GREG_GLOBAL_CONF_0, + mpic_read(mpic->gregs, MPIC_GREG_GLOBAL_CONF_0) + | MPIC_GREG_GCONF_RESET); + while( mpic_read(mpic->gregs, MPIC_GREG_GLOBAL_CONF_0) + & MPIC_GREG_GCONF_RESET) + mb(); + } + + /* Read feature register, calculate num CPUs and, for non-ISU + * MPICs, num sources as well. On ISU MPICs, sources are counted + * as ISUs are added + */ + reg = mpic_read(mpic->gregs, MPIC_GREG_FEATURE_0); + mpic->num_cpus = ((reg & MPIC_GREG_FEATURE_LAST_CPU_MASK) + >> MPIC_GREG_FEATURE_LAST_CPU_SHIFT) + 1; + if (isu_size == 0) + mpic->num_sources = ((reg & MPIC_GREG_FEATURE_LAST_SRC_MASK) + >> MPIC_GREG_FEATURE_LAST_SRC_SHIFT) + 1; + + /* Map the per-CPU registers */ + for (i = 0; i < mpic->num_cpus; i++) { + mpic->cpuregs[i] = ioremap(phys_addr + MPIC_CPU_BASE + + i * MPIC_CPU_STRIDE, 0x1000); + BUG_ON(mpic->cpuregs[i] == NULL); + } + + /* Initialize main ISU if none provided */ + if (mpic->isu_size == 0) { + mpic->isu_size = mpic->num_sources; + mpic->isus[0] = ioremap(phys_addr + MPIC_IRQ_BASE, + MPIC_IRQ_STRIDE * mpic->isu_size); + BUG_ON(mpic->isus[0] == NULL); + } + mpic->isu_shift = 1 + __ilog2(mpic->isu_size - 1); + mpic->isu_mask = (1 << mpic->isu_shift) - 1; + + /* Display version */ + switch (reg & MPIC_GREG_FEATURE_VERSION_MASK) { + case 1: + vers = "1.0"; + break; + case 2: + vers = "1.2"; + break; + case 3: + vers = "1.3"; + break; + default: + vers = "<unknown>"; + break; + } + printk(KERN_INFO "mpic: Setting up MPIC \"%s\" version %s at %lx, max %d CPUs\n", + name, vers, phys_addr, mpic->num_cpus); + printk(KERN_INFO "mpic: ISU size: %d, shift: %d, mask: %x\n", mpic->isu_size, + mpic->isu_shift, mpic->isu_mask); + + mpic->next = mpics; + mpics = mpic; + + if (flags & MPIC_PRIMARY) + mpic_primary = mpic; + + return mpic; +} + +void __init mpic_assign_isu(struct mpic *mpic, unsigned int isu_num, + unsigned long phys_addr) +{ + unsigned int isu_first = isu_num * mpic->isu_size; + + BUG_ON(isu_num >= MPIC_MAX_ISU); + + mpic->isus[isu_num] = ioremap(phys_addr, MPIC_IRQ_STRIDE * mpic->isu_size); + if ((isu_first + mpic->isu_size) > mpic->num_sources) + mpic->num_sources = isu_first + mpic->isu_size; +} + +void __init mpic_setup_cascade(unsigned int irq, mpic_cascade_t handler, + void *data) +{ + struct mpic *mpic = mpic_find(irq, NULL); + unsigned long flags; + + /* Synchronization here is a bit dodgy, so don't try to replace cascade + * interrupts on the fly too often ... but normally it's set up at boot. + */ + spin_lock_irqsave(&mpic_lock, flags); + if (mpic->cascade) + mpic_disable_irq(mpic->cascade_vec + mpic->irq_offset); + mpic->cascade = NULL; + wmb(); + mpic->cascade_vec = irq - mpic->irq_offset; + mpic->cascade_data = data; + wmb(); + mpic->cascade = handler; + mpic_enable_irq(irq); + spin_unlock_irqrestore(&mpic_lock, flags); +} + +void __init mpic_init(struct mpic *mpic) +{ + int i; + + BUG_ON(mpic->num_sources == 0); + + printk(KERN_INFO "mpic: Initializing for %d sources\n", mpic->num_sources); + + /* Set current processor priority to max */ + mpic_cpu_write(MPIC_CPU_CURRENT_TASK_PRI, 0xf); + + /* Initialize timers: just disable them all */ + for (i = 0; i < 4; i++) { + mpic_write(mpic->tmregs, + i * MPIC_TIMER_STRIDE + MPIC_TIMER_DESTINATION, 0); + mpic_write(mpic->tmregs, + i * MPIC_TIMER_STRIDE + MPIC_TIMER_VECTOR_PRI, + MPIC_VECPRI_MASK | + (MPIC_VEC_TIMER_0 + i)); + } + + /* Initialize IPIs to our reserved vectors and mark them disabled for now */ + mpic_test_broken_ipi(mpic); + for (i = 0; i < 4; i++) { + mpic_ipi_write(i, + MPIC_VECPRI_MASK | + (10 << MPIC_VECPRI_PRIORITY_SHIFT) | + (MPIC_VEC_IPI_0 + i)); +#ifdef CONFIG_SMP + if (!(mpic->flags & MPIC_PRIMARY)) + continue; + irq_desc[mpic->ipi_offset+i].status |= IRQ_PER_CPU; + irq_desc[mpic->ipi_offset+i].handler = &mpic->hc_ipi; +#endif /* CONFIG_SMP */ + } + + /* Initialize interrupt sources */ + if (mpic->irq_count == 0) + mpic->irq_count = mpic->num_sources; + +#ifdef CONFIG_MPIC_BROKEN_U3 + /* Do the ioapic fixups on U3 broken mpic */ + DBG("MPIC flags: %x\n", mpic->flags); + if ((mpic->flags & MPIC_BROKEN_U3) && (mpic->flags & MPIC_PRIMARY)) + mpic_scan_ioapics(mpic); +#endif /* CONFIG_MPIC_BROKEN_U3 */ + + for (i = 0; i < mpic->num_sources; i++) { + /* start with vector = source number, and masked */ + u32 vecpri = MPIC_VECPRI_MASK | i | (8 << MPIC_VECPRI_PRIORITY_SHIFT); + int level = 0; + + /* if it's an IPI, we skip it */ + if ((mpic->irq_offset + i) >= (mpic->ipi_offset + i) && + (mpic->irq_offset + i) < (mpic->ipi_offset + i + 4)) + continue; + + /* do senses munging */ + if (mpic->senses && i < mpic->senses_count) { + if (mpic->senses[i] & IRQ_SENSE_LEVEL) + vecpri |= MPIC_VECPRI_SENSE_LEVEL; + if (mpic->senses[i] & IRQ_POLARITY_POSITIVE) + vecpri |= MPIC_VECPRI_POLARITY_POSITIVE; + } else + vecpri |= MPIC_VECPRI_SENSE_LEVEL; + + /* remember if it was a level interrupts */ + level = (vecpri & MPIC_VECPRI_SENSE_LEVEL); + + /* deal with broken U3 */ + if (mpic->flags & MPIC_BROKEN_U3) { +#ifdef CONFIG_MPIC_BROKEN_U3 + if (mpic_is_ht_interrupt(mpic, i)) { + vecpri &= ~(MPIC_VECPRI_SENSE_MASK | + MPIC_VECPRI_POLARITY_MASK); + vecpri |= MPIC_VECPRI_POLARITY_POSITIVE; + } +#else + printk(KERN_ERR "mpic: BROKEN_U3 set, but CONFIG doesn't match\n"); +#endif + } + + DBG("setup source %d, vecpri: %08x, level: %d\n", i, vecpri, + (level != 0)); + + /* init hw */ + mpic_irq_write(i, MPIC_IRQ_VECTOR_PRI, vecpri); + mpic_irq_write(i, MPIC_IRQ_DESTINATION, + 1 << hard_smp_processor_id()); + + /* init linux descriptors */ + if (i < mpic->irq_count) { + irq_desc[mpic->irq_offset+i].status = level ? IRQ_LEVEL : 0; + irq_desc[mpic->irq_offset+i].handler = &mpic->hc_irq; + } + } + + /* Init spurrious vector */ + mpic_write(mpic->gregs, MPIC_GREG_SPURIOUS, MPIC_VEC_SPURRIOUS); + + /* Disable 8259 passthrough */ + mpic_write(mpic->gregs, MPIC_GREG_GLOBAL_CONF_0, + mpic_read(mpic->gregs, MPIC_GREG_GLOBAL_CONF_0) + | MPIC_GREG_GCONF_8259_PTHROU_DIS); + + /* Set current processor priority to 0 */ + mpic_cpu_write(MPIC_CPU_CURRENT_TASK_PRI, 0); +} + + + +void mpic_irq_set_priority(unsigned int irq, unsigned int pri) +{ + int is_ipi; + struct mpic *mpic = mpic_find(irq, &is_ipi); + unsigned long flags; + u32 reg; + + spin_lock_irqsave(&mpic_lock, flags); + if (is_ipi) { + reg = mpic_ipi_read(irq - mpic->ipi_offset) & MPIC_VECPRI_PRIORITY_MASK; + mpic_ipi_write(irq - mpic->ipi_offset, + reg | (pri << MPIC_VECPRI_PRIORITY_SHIFT)); + } else { + reg = mpic_irq_read(irq - mpic->irq_offset, MPIC_IRQ_VECTOR_PRI) + & MPIC_VECPRI_PRIORITY_MASK; + mpic_irq_write(irq - mpic->irq_offset, MPIC_IRQ_VECTOR_PRI, + reg | (pri << MPIC_VECPRI_PRIORITY_SHIFT)); + } + spin_unlock_irqrestore(&mpic_lock, flags); +} + +unsigned int mpic_irq_get_priority(unsigned int irq) +{ + int is_ipi; + struct mpic *mpic = mpic_find(irq, &is_ipi); + unsigned long flags; + u32 reg; + + spin_lock_irqsave(&mpic_lock, flags); + if (is_ipi) + reg = mpic_ipi_read(irq - mpic->ipi_offset); + else + reg = mpic_irq_read(irq - mpic->irq_offset, MPIC_IRQ_VECTOR_PRI); + spin_unlock_irqrestore(&mpic_lock, flags); + return (reg & MPIC_VECPRI_PRIORITY_MASK) >> MPIC_VECPRI_PRIORITY_SHIFT; +} + +void mpic_setup_this_cpu(void) +{ +#ifdef CONFIG_SMP + struct mpic *mpic = mpic_primary; + unsigned long flags; + u32 msk = 1 << hard_smp_processor_id(); + unsigned int i; + + BUG_ON(mpic == NULL); + + DBG("%s: setup_this_cpu(%d)\n", mpic->name, hard_smp_processor_id()); + + spin_lock_irqsave(&mpic_lock, flags); + + /* let the mpic know we want intrs. default affinity is 0xffffffff + * until changed via /proc. That's how it's done on x86. If we want + * it differently, then we should make sure we also change the default + * values of irq_affinity in irq.c. + */ + if (distribute_irqs) { + for (i = 0; i < mpic->num_sources ; i++) + mpic_irq_write(i, MPIC_IRQ_DESTINATION, + mpic_irq_read(i, MPIC_IRQ_DESTINATION) | msk); + } + + /* Set current processor priority to 0 */ + mpic_cpu_write(MPIC_CPU_CURRENT_TASK_PRI, 0); + + spin_unlock_irqrestore(&mpic_lock, flags); +#endif /* CONFIG_SMP */ +} + +int mpic_cpu_get_priority(void) +{ + struct mpic *mpic = mpic_primary; + + return mpic_cpu_read(MPIC_CPU_CURRENT_TASK_PRI); +} + +void mpic_cpu_set_priority(int prio) +{ + struct mpic *mpic = mpic_primary; + + prio &= MPIC_CPU_TASKPRI_MASK; + mpic_cpu_write(MPIC_CPU_CURRENT_TASK_PRI, prio); +} + +/* + * XXX: someone who knows mpic should check this. + * do we need to eoi the ipi including for kexec cpu here (see xics comments)? + * or can we reset the mpic in the new kernel? + */ +void mpic_teardown_this_cpu(int secondary) +{ + struct mpic *mpic = mpic_primary; + unsigned long flags; + u32 msk = 1 << hard_smp_processor_id(); + unsigned int i; + + BUG_ON(mpic == NULL); + + DBG("%s: teardown_this_cpu(%d)\n", mpic->name, hard_smp_processor_id()); + spin_lock_irqsave(&mpic_lock, flags); + + /* let the mpic know we don't want intrs. */ + for (i = 0; i < mpic->num_sources ; i++) + mpic_irq_write(i, MPIC_IRQ_DESTINATION, + mpic_irq_read(i, MPIC_IRQ_DESTINATION) & ~msk); + + /* Set current processor priority to max */ + mpic_cpu_write(MPIC_CPU_CURRENT_TASK_PRI, 0xf); + + spin_unlock_irqrestore(&mpic_lock, flags); +} + + +void mpic_send_ipi(unsigned int ipi_no, unsigned int cpu_mask) +{ + struct mpic *mpic = mpic_primary; + + BUG_ON(mpic == NULL); + + DBG("%s: send_ipi(ipi_no: %d)\n", mpic->name, ipi_no); + + mpic_cpu_write(MPIC_CPU_IPI_DISPATCH_0 + ipi_no * 0x10, + mpic_physmask(cpu_mask & cpus_addr(cpu_online_map)[0])); +} + +int mpic_get_one_irq(struct mpic *mpic, struct pt_regs *regs) +{ + u32 irq; + + irq = mpic_cpu_read(MPIC_CPU_INTACK) & MPIC_VECPRI_VECTOR_MASK; + DBG("%s: get_one_irq(): %d\n", mpic->name, irq); + + if (mpic->cascade && irq == mpic->cascade_vec) { + DBG("%s: cascading ...\n", mpic->name); + irq = mpic->cascade(regs, mpic->cascade_data); + mpic_eoi(mpic); + return irq; + } + if (unlikely(irq == MPIC_VEC_SPURRIOUS)) + return -1; + if (irq < MPIC_VEC_IPI_0) + return irq + mpic->irq_offset; + DBG("%s: ipi %d !\n", mpic->name, irq - MPIC_VEC_IPI_0); + return irq - MPIC_VEC_IPI_0 + mpic->ipi_offset; +} + +int mpic_get_irq(struct pt_regs *regs) +{ + struct mpic *mpic = mpic_primary; + + BUG_ON(mpic == NULL); + + return mpic_get_one_irq(mpic, regs); +} + + +#ifdef CONFIG_SMP +void mpic_request_ipis(void) +{ + struct mpic *mpic = mpic_primary; + + BUG_ON(mpic == NULL); + + printk("requesting IPIs ... \n"); + + /* IPIs are marked SA_INTERRUPT as they must run with irqs disabled */ + request_irq(mpic->ipi_offset+0, mpic_ipi_action, SA_INTERRUPT, + "IPI0 (call function)", mpic); + request_irq(mpic->ipi_offset+1, mpic_ipi_action, SA_INTERRUPT, + "IPI1 (reschedule)", mpic); + request_irq(mpic->ipi_offset+2, mpic_ipi_action, SA_INTERRUPT, + "IPI2 (unused)", mpic); + request_irq(mpic->ipi_offset+3, mpic_ipi_action, SA_INTERRUPT, + "IPI3 (debugger break)", mpic); + + printk("IPIs requested... \n"); +} + +void smp_mpic_message_pass(int target, int msg) +{ + /* make sure we're sending something that translates to an IPI */ + if ((unsigned int)msg > 3) { + printk("SMP %d: smp_message_pass: unknown msg %d\n", + smp_processor_id(), msg); + return; + } + switch (target) { + case MSG_ALL: + mpic_send_ipi(msg, 0xffffffff); + break; + case MSG_ALL_BUT_SELF: + mpic_send_ipi(msg, 0xffffffff & ~(1 << smp_processor_id())); + break; + default: + mpic_send_ipi(msg, 1 << target); + break; + } +} +#endif /* CONFIG_SMP */ diff --git a/arch/powerpc/sysdev/u3_iommu.c b/arch/powerpc/sysdev/u3_iommu.c new file mode 100644 index 000000000000..fba871a1bda5 --- /dev/null +++ b/arch/powerpc/sysdev/u3_iommu.c @@ -0,0 +1,327 @@ +/* + * arch/ppc64/kernel/u3_iommu.c + * + * Copyright (C) 2004 Olof Johansson <olof@austin.ibm.com>, IBM Corporation + * + * Based on pSeries_iommu.c: + * Copyright (C) 2001 Mike Corrigan & Dave Engebretsen, IBM Corporation + * Copyright (C) 2004 Olof Johansson <olof@austin.ibm.com>, IBM Corporation + * + * Dynamic DMA mapping support, Apple U3 & IBM CPC925 "DART" iommu. + * + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#include <linux/config.h> +#include <linux/init.h> +#include <linux/types.h> +#include <linux/slab.h> +#include <linux/mm.h> +#include <linux/spinlock.h> +#include <linux/string.h> +#include <linux/pci.h> +#include <linux/dma-mapping.h> +#include <linux/vmalloc.h> +#include <asm/io.h> +#include <asm/prom.h> +#include <asm/ppcdebug.h> +#include <asm/iommu.h> +#include <asm/pci-bridge.h> +#include <asm/machdep.h> +#include <asm/abs_addr.h> +#include <asm/cacheflush.h> +#include <asm/lmb.h> +#include <asm/dart.h> +#include <asm/ppc-pci.h> + +extern int iommu_force_on; + +/* Physical base address and size of the DART table */ +unsigned long dart_tablebase; /* exported to htab_initialize */ +static unsigned long dart_tablesize; + +/* Virtual base address of the DART table */ +static u32 *dart_vbase; + +/* Mapped base address for the dart */ +static unsigned int *dart; + +/* Dummy val that entries are set to when unused */ +static unsigned int dart_emptyval; + +static struct iommu_table iommu_table_u3; +static int iommu_table_u3_inited; +static int dart_dirty; + +#define DBG(...) + +static inline void dart_tlb_invalidate_all(void) +{ + unsigned long l = 0; + unsigned int reg; + unsigned long limit; + + DBG("dart: flush\n"); + + /* To invalidate the DART, set the DARTCNTL_FLUSHTLB bit in the + * control register and wait for it to clear. + * + * Gotcha: Sometimes, the DART won't detect that the bit gets + * set. If so, clear it and set it again. + */ + + limit = 0; + +retry: + reg = in_be32((unsigned int *)dart+DARTCNTL); + reg |= DARTCNTL_FLUSHTLB; + out_be32((unsigned int *)dart+DARTCNTL, reg); + + l = 0; + while ((in_be32((unsigned int *)dart+DARTCNTL) & DARTCNTL_FLUSHTLB) && + l < (1L<<limit)) { + l++; + } + if (l == (1L<<limit)) { + if (limit < 4) { + limit++; + reg = in_be32((unsigned int *)dart+DARTCNTL); + reg &= ~DARTCNTL_FLUSHTLB; + out_be32((unsigned int *)dart+DARTCNTL, reg); + goto retry; + } else + panic("U3-DART: TLB did not flush after waiting a long " + "time. Buggy U3 ?"); + } +} + +static void dart_flush(struct iommu_table *tbl) +{ + if (dart_dirty) + dart_tlb_invalidate_all(); + dart_dirty = 0; +} + +static void dart_build(struct iommu_table *tbl, long index, + long npages, unsigned long uaddr, + enum dma_data_direction direction) +{ + unsigned int *dp; + unsigned int rpn; + + DBG("dart: build at: %lx, %lx, addr: %x\n", index, npages, uaddr); + + index <<= DART_PAGE_FACTOR; + npages <<= DART_PAGE_FACTOR; + + dp = ((unsigned int*)tbl->it_base) + index; + + /* On U3, all memory is contigous, so we can move this + * out of the loop. + */ + while (npages--) { + rpn = virt_to_abs(uaddr) >> DART_PAGE_SHIFT; + + *(dp++) = DARTMAP_VALID | (rpn & DARTMAP_RPNMASK); + + rpn++; + uaddr += DART_PAGE_SIZE; + } + + dart_dirty = 1; +} + + +static void dart_free(struct iommu_table *tbl, long index, long npages) +{ + unsigned int *dp; + + /* We don't worry about flushing the TLB cache. The only drawback of + * not doing it is that we won't catch buggy device drivers doing + * bad DMAs, but then no 32-bit architecture ever does either. + */ + + DBG("dart: free at: %lx, %lx\n", index, npages); + + index <<= DART_PAGE_FACTOR; + npages <<= DART_PAGE_FACTOR; + + dp = ((unsigned int *)tbl->it_base) + index; + + while (npages--) + *(dp++) = dart_emptyval; +} + + +static int dart_init(struct device_node *dart_node) +{ + unsigned int regword; + unsigned int i; + unsigned long tmp; + + if (dart_tablebase == 0 || dart_tablesize == 0) { + printk(KERN_INFO "U3-DART: table not allocated, using direct DMA\n"); + return -ENODEV; + } + + /* Make sure nothing from the DART range remains in the CPU cache + * from a previous mapping that existed before the kernel took + * over + */ + flush_dcache_phys_range(dart_tablebase, dart_tablebase + dart_tablesize); + + /* Allocate a spare page to map all invalid DART pages. We need to do + * that to work around what looks like a problem with the HT bridge + * prefetching into invalid pages and corrupting data + */ + tmp = lmb_alloc(DART_PAGE_SIZE, DART_PAGE_SIZE); + if (!tmp) + panic("U3-DART: Cannot allocate spare page!"); + dart_emptyval = DARTMAP_VALID | ((tmp >> DART_PAGE_SHIFT) & DARTMAP_RPNMASK); + + /* Map in DART registers. FIXME: Use device node to get base address */ + dart = ioremap(DART_BASE, 0x7000); + if (dart == NULL) + panic("U3-DART: Cannot map registers!"); + + /* Set initial control register contents: table base, + * table size and enable bit + */ + regword = DARTCNTL_ENABLE | + ((dart_tablebase >> DART_PAGE_SHIFT) << DARTCNTL_BASE_SHIFT) | + (((dart_tablesize >> DART_PAGE_SHIFT) & DARTCNTL_SIZE_MASK) + << DARTCNTL_SIZE_SHIFT); + dart_vbase = ioremap(virt_to_abs(dart_tablebase), dart_tablesize); + + /* Fill initial table */ + for (i = 0; i < dart_tablesize/4; i++) + dart_vbase[i] = dart_emptyval; + + /* Initialize DART with table base and enable it. */ + out_be32((unsigned int *)dart, regword); + + /* Invalidate DART to get rid of possible stale TLBs */ + dart_tlb_invalidate_all(); + + printk(KERN_INFO "U3/CPC925 DART IOMMU initialized\n"); + + return 0; +} + +static void iommu_table_u3_setup(void) +{ + iommu_table_u3.it_busno = 0; + iommu_table_u3.it_offset = 0; + /* it_size is in number of entries */ + iommu_table_u3.it_size = dart_tablesize / sizeof(u32); + + /* Initialize the common IOMMU code */ + iommu_table_u3.it_base = (unsigned long)dart_vbase; + iommu_table_u3.it_index = 0; + iommu_table_u3.it_blocksize = 1; + iommu_init_table(&iommu_table_u3); + + /* Reserve the last page of the DART to avoid possible prefetch + * past the DART mapped area + */ + set_bit(iommu_table_u3.it_size - 1, iommu_table_u3.it_map); +} + +static void iommu_dev_setup_u3(struct pci_dev *dev) +{ + struct device_node *dn; + + /* We only have one iommu table on the mac for now, which makes + * things simple. Setup all PCI devices to point to this table + * + * We must use pci_device_to_OF_node() to make sure that + * we get the real "final" pointer to the device in the + * pci_dev sysdata and not the temporary PHB one + */ + dn = pci_device_to_OF_node(dev); + + if (dn) + PCI_DN(dn)->iommu_table = &iommu_table_u3; +} + +static void iommu_bus_setup_u3(struct pci_bus *bus) +{ + struct device_node *dn; + + if (!iommu_table_u3_inited) { + iommu_table_u3_inited = 1; + iommu_table_u3_setup(); + } + + dn = pci_bus_to_OF_node(bus); + + if (dn) + PCI_DN(dn)->iommu_table = &iommu_table_u3; +} + +static void iommu_dev_setup_null(struct pci_dev *dev) { } +static void iommu_bus_setup_null(struct pci_bus *bus) { } + +void iommu_init_early_u3(void) +{ + struct device_node *dn; + + /* Find the DART in the device-tree */ + dn = of_find_compatible_node(NULL, "dart", "u3-dart"); + if (dn == NULL) + return; + + /* Setup low level TCE operations for the core IOMMU code */ + ppc_md.tce_build = dart_build; + ppc_md.tce_free = dart_free; + ppc_md.tce_flush = dart_flush; + + /* Initialize the DART HW */ + if (dart_init(dn)) { + /* If init failed, use direct iommu and null setup functions */ + ppc_md.iommu_dev_setup = iommu_dev_setup_null; + ppc_md.iommu_bus_setup = iommu_bus_setup_null; + + /* Setup pci_dma ops */ + pci_direct_iommu_init(); + } else { + ppc_md.iommu_dev_setup = iommu_dev_setup_u3; + ppc_md.iommu_bus_setup = iommu_bus_setup_u3; + + /* Setup pci_dma ops */ + pci_iommu_init(); + } +} + + +void __init alloc_u3_dart_table(void) +{ + /* Only reserve DART space if machine has more than 2GB of RAM + * or if requested with iommu=on on cmdline. + */ + if (lmb_end_of_DRAM() <= 0x80000000ull && !iommu_force_on) + return; + + /* 512 pages (2MB) is max DART tablesize. */ + dart_tablesize = 1UL << 21; + /* 16MB (1 << 24) alignment. We allocate a full 16Mb chuck since we + * will blow up an entire large page anyway in the kernel mapping + */ + dart_tablebase = (unsigned long) + abs_to_virt(lmb_alloc_base(1UL<<24, 1UL<<24, 0x80000000L)); + + printk(KERN_INFO "U3-DART allocated at: %lx\n", dart_tablebase); +} |