minix/minix/drivers/bus/pci/pci.c

2657 lines
59 KiB
C
Raw Normal View History

2005-04-21 16:53:53 +02:00
/*
pci.c
Configure devices on the PCI bus
Created: Jan 2000 by Philip Homburg <philip@cs.vu.nl>
*/
#include <minix/driver.h>
#include <minix/acpi.h>
#include <minix/param.h>
#include <minix/ds.h>
#include <minix/rs.h>
2005-04-21 16:53:53 +02:00
#include <machine/pci.h>
#include <machine/pci_amd.h>
#include <machine/pci_intel.h>
#include <machine/pci_sis.h>
#include <machine/pci_via.h>
#include <machine/vmparam.h>
2005-04-21 16:53:53 +02:00
#include <stdlib.h>
#include <stdio.h>
#include "pci.h"
#define irq_mode_pci(irq) ((void)0)
2005-04-21 16:53:53 +02:00
#define PBT_INTEL_HOST 1
2005-04-21 16:53:53 +02:00
#define PBT_PCIBRIDGE 2
#define PBT_CARDBUS 3
#define BAM_NR 6 /* Number of base-address registers */
2005-04-21 16:53:53 +02:00
struct pci_acl pci_acl[NR_DRIVERS];
2005-08-09 13:23:41 +02:00
2012-03-25 20:25:53 +02:00
static struct pcibus
2005-04-21 16:53:53 +02:00
{
int pb_type;
int pb_needinit;
2005-04-21 16:53:53 +02:00
int pb_isabridge_dev;
int pb_isabridge_type;
int pb_devind;
int pb_busnr;
2005-04-21 16:53:53 +02:00
u8_t (*pb_rreg8)(int busind, int devind, int port);
u16_t (*pb_rreg16)(int busind, int devind, int port);
u32_t (*pb_rreg32)(int busind, int devind, int port);
void (*pb_wreg8)(int busind, int devind, int port, u8_t value);
void (*pb_wreg16)(int busind, int devind, int port, u16_t value);
2005-04-21 16:53:53 +02:00
void (*pb_wreg32)(int busind, int devind, int port, u32_t value);
u16_t (*pb_rsts)(int busind);
void (*pb_wsts)(int busind, u16_t value);
2005-04-21 16:53:53 +02:00
} pcibus[NR_PCIBUS];
2012-03-25 20:25:53 +02:00
static int nr_pcibus= 0;
2005-04-21 16:53:53 +02:00
2012-03-25 20:25:53 +02:00
static struct pcidev
2005-04-21 16:53:53 +02:00
{
u8_t pd_busnr;
2005-04-21 16:53:53 +02:00
u8_t pd_dev;
u8_t pd_func;
u8_t pd_baseclass;
u8_t pd_subclass;
u8_t pd_infclass;
u16_t pd_vid;
u16_t pd_did;
u16_t pd_sub_vid;
u16_t pd_sub_did;
u8_t pd_ilr;
2005-04-21 16:53:53 +02:00
u8_t pd_inuse;
endpoint_t pd_proc;
struct bar
{
int pb_flags;
int pb_nr;
u32_t pb_base;
u32_t pb_size;
} pd_bar[BAM_NR];
int pd_bar_nr;
2005-04-21 16:53:53 +02:00
} pcidev[NR_PCIDEV];
/* pb_flags */
#define PBF_IO 1 /* I/O else memory */
#define PBF_INCOMPLETE 2 /* not allocated */
2012-03-25 20:25:53 +02:00
static int nr_pcidev= 0;
static struct machine machine;
static endpoint_t acpi_ep;
/*===========================================================================*
* helper functions for I/O *
*===========================================================================*/
static unsigned
pci_inb(u16_t port) {
2005-10-24 16:02:45 +02:00
u32_t value;
int s;
if ((s=sys_inb(port, &value)) !=OK)
printf("PCI: warning, sys_inb failed: %d\n", s);
return value;
}
static unsigned
pci_inw(u16_t port) {
2005-12-02 15:45:10 +01:00
u32_t value;
int s;
if ((s=sys_inw(port, &value)) !=OK)
printf("PCI: warning, sys_inw failed: %d\n", s);
return value;
}
static unsigned
pci_inl(u16_t port) {
u32_t value;
int s;
if ((s=sys_inl(port, &value)) !=OK)
printf("PCI: warning, sys_inl failed: %d\n", s);
return value;
}
static void
pci_outb(u16_t port, u8_t value) {
int s;
if ((s=sys_outb(port, value)) !=OK)
printf("PCI: warning, sys_outb failed: %d\n", s);
}
static void
pci_outw(u16_t port, u16_t value) {
int s;
if ((s=sys_outw(port, value)) !=OK)
printf("PCI: warning, sys_outw failed: %d\n", s);
}
static void
pci_outl(u16_t port, u32_t value) {
int s;
if ((s=sys_outl(port, value)) !=OK)
printf("PCI: warning, sys_outl failed: %d\n", s);
}
static u8_t
pcii_rreg8(int busind, int devind, int port)
2005-04-21 16:53:53 +02:00
{
u8_t v;
int s;
2005-04-21 16:53:53 +02:00
v= PCII_RREG8_(pcibus[busind].pb_busnr,
pcidev[devind].pd_dev, pcidev[devind].pd_func,
port);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
#if 0
printf("pcii_rreg8(%d, %d, 0x%X): %d.%d.%d= 0x%X\n",
busind, devind, port,
pcibus[busind].pb_bus, pcidev[devind].pd_dev,
pcidev[devind].pd_func, v);
#endif
return v;
2005-04-21 16:53:53 +02:00
}
static u16_t
pcii_rreg16(int busind, int devind, int port)
2005-04-21 16:53:53 +02:00
{
u16_t v;
int s;
2005-04-21 16:53:53 +02:00
v= PCII_RREG16_(pcibus[busind].pb_busnr,
pcidev[devind].pd_dev, pcidev[devind].pd_func,
port);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
#if 0
printf("pcii_rreg16(%d, %d, 0x%X): %d.%d.%d= 0x%X\n",
busind, devind, port,
pcibus[busind].pb_bus, pcidev[devind].pd_dev,
pcidev[devind].pd_func, v);
#endif
return v;
2005-04-21 16:53:53 +02:00
}
static u32_t
pcii_rreg32(int busind, int devind, int port)
2005-04-21 16:53:53 +02:00
{
u32_t v;
int s;
2005-04-21 16:53:53 +02:00
v= PCII_RREG32_(pcibus[busind].pb_busnr,
pcidev[devind].pd_dev, pcidev[devind].pd_func,
port);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
#if 0
printf("pcii_rreg32(%d, %d, 0x%X): %d.%d.%d= 0x%X\n",
busind, devind, port,
pcibus[busind].pb_bus, pcidev[devind].pd_dev,
pcidev[devind].pd_func, v);
#endif
return v;
}
static void
pcii_wreg8(int busind, int devind, int port, u8_t value)
{
int s;
#if 0
printf("pcii_wreg8(%d, %d, 0x%X, 0x%X): %d.%d.%d\n",
busind, devind, port, value,
pcibus[busind].pb_bus, pcidev[devind].pd_dev,
pcidev[devind].pd_func);
#endif
PCII_WREG8_(pcibus[busind].pb_busnr,
pcidev[devind].pd_dev, pcidev[devind].pd_func,
port, value);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
}
static void
pcii_wreg16(int busind, int devind, int port, u16_t value)
{
int s;
#if 0
printf("pcii_wreg16(%d, %d, 0x%X, 0x%X): %d.%d.%d\n",
busind, devind, port, value,
pcibus[busind].pb_bus, pcidev[devind].pd_dev,
pcidev[devind].pd_func);
#endif
PCII_WREG16_(pcibus[busind].pb_busnr,
pcidev[devind].pd_dev, pcidev[devind].pd_func,
port, value);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
}
static void
pcii_wreg32(int busind, int devind, int port, u32_t value)
{
int s;
#if 0
printf("pcii_wreg32(%d, %d, 0x%X, 0x%X): %d.%d.%d\n",
busind, devind, port, value,
pcibus[busind].pb_busnr, pcidev[devind].pd_dev,
pcidev[devind].pd_func);
#endif
PCII_WREG32_(pcibus[busind].pb_busnr,
pcidev[devind].pd_dev, pcidev[devind].pd_func,
port, value);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n",s);
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* ntostr *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
static void
ntostr(unsigned int n, char **str, const char *end)
2005-04-21 16:53:53 +02:00
{
char tmpstr[20];
int i;
2006-01-27 14:18:07 +01:00
if (n == 0)
{
tmpstr[0]= '0';
i= 1;
}
else
2012-03-07 23:45:55 +01:00
{
for (i= 0; n; i++)
{
tmpstr[i]= '0' + (n%10);
n /= 10;
}
2012-03-07 23:45:55 +01:00
}
for (; i>0; i--)
2006-01-27 14:18:07 +01:00
{
if (*str == end)
2006-01-27 14:18:07 +01:00
{
break;
2006-01-27 14:18:07 +01:00
}
**str= tmpstr[i-1];
(*str)++;
2006-01-27 14:18:07 +01:00
}
if (*str == end)
(*str)[-1]= '\0';
else
**str= '\0';
}
/*===========================================================================*
* get_busind *
*===========================================================================*/
static int
get_busind(int busnr)
{
int i;
for (i= 0; i<nr_pcibus; i++)
{
if (pcibus[i].pb_busnr == busnr)
return i;
}
panic("get_busind: can't find bus: %d", busnr);
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* Unprotected helper functions *
*===========================================================================*/
static u8_t
__pci_attr_r8(int devind, int port)
{
int busnr, busind;
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
return pcibus[busind].pb_rreg8(busind, devind, port);
}
static u16_t
__pci_attr_r16(int devind, int port)
2005-04-21 16:53:53 +02:00
{
int busnr, busind;
2005-04-21 16:53:53 +02:00
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
return pcibus[busind].pb_rreg16(busind, devind, port);
2005-04-21 16:53:53 +02:00
}
static u32_t
__pci_attr_r32(int devind, int port)
2005-04-21 16:53:53 +02:00
{
int busnr, busind;
2005-04-21 16:53:53 +02:00
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
return pcibus[busind].pb_rreg32(busind, devind, port);
2005-04-21 16:53:53 +02:00
}
static void
__pci_attr_w8(int devind, int port, u8_t value)
{
int busnr, busind;
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
pcibus[busind].pb_wreg8(busind, devind, port, value);
}
static void
__pci_attr_w16(int devind, int port, u16_t value)
2005-04-21 16:53:53 +02:00
{
int busnr, busind;
2005-04-21 16:53:53 +02:00
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
pcibus[busind].pb_wreg16(busind, devind, port, value);
2005-04-21 16:53:53 +02:00
}
static void
__pci_attr_w32(int devind, int port, u32_t value)
{
int busnr, busind;
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
pcibus[busind].pb_wreg32(busind, devind, port, value);
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* helpers *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
static u16_t
pci_attr_rsts(int devind)
2005-04-21 16:53:53 +02:00
{
int busnr, busind;
2005-04-21 16:53:53 +02:00
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
return pcibus[busind].pb_rsts(busind);
2005-04-21 16:53:53 +02:00
}
static void
pci_attr_wsts(int devind, u16_t value)
2005-04-21 16:53:53 +02:00
{
int busnr, busind;
2005-04-21 16:53:53 +02:00
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
pcibus[busind].pb_wsts(busind, value);
2005-04-21 16:53:53 +02:00
}
static u16_t
pcii_rsts(int busind)
2005-04-21 16:53:53 +02:00
{
u16_t v;
int s;
2005-04-21 16:53:53 +02:00
v= PCII_RREG16_(pcibus[busind].pb_busnr, 0, 0, PCI_SR);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
return v;
}
2005-04-21 16:53:53 +02:00
static void
pcii_wsts(int busind, u16_t value)
{
int s;
PCII_WREG16_(pcibus[busind].pb_busnr, 0, 0, PCI_SR, value);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
}
2005-04-21 16:53:53 +02:00
static int
is_duplicate(u8_t busnr, u8_t dev, u8_t func)
{
int i;
2005-04-21 16:53:53 +02:00
for (i= 0; i<nr_pcidev; i++)
2005-04-21 16:53:53 +02:00
{
if (pcidev[i].pd_busnr == busnr &&
pcidev[i].pd_dev == dev &&
pcidev[i].pd_func == func)
2005-04-21 16:53:53 +02:00
{
return 1;
2005-04-21 16:53:53 +02:00
}
}
return 0;
}
2005-04-21 16:53:53 +02:00
static int
get_freebus(void)
{
int i, freebus;
2005-04-21 16:53:53 +02:00
freebus= 1;
for (i= 0; i<nr_pcibus; i++)
2005-04-21 16:53:53 +02:00
{
if (pcibus[i].pb_needinit)
continue;
if (pcibus[i].pb_type == PBT_INTEL_HOST)
continue;
if (pcibus[i].pb_busnr <= freebus)
freebus= pcibus[i].pb_busnr+1;
printf("get_freebus: should check suboridinate bus number\n");
2005-04-21 16:53:53 +02:00
}
return freebus;
}
2005-04-21 16:53:53 +02:00
static const char *
pci_vid_name(u16_t vid)
{
int i;
2005-04-21 16:53:53 +02:00
for (i= 0; pci_vendor_table[i].name; i++)
2005-04-21 16:53:53 +02:00
{
if (pci_vendor_table[i].vid == vid)
return pci_vendor_table[i].name;
2005-04-21 16:53:53 +02:00
}
return "unknown";
}
2005-04-21 16:53:53 +02:00
static const char *
pci_baseclass_name(u8_t baseclass)
{
int i;
for (i= 0; pci_baseclass_table[i].name; i++)
{
if (pci_baseclass_table[i].baseclass == baseclass)
return pci_baseclass_table[i].name;
}
return NULL;
2005-04-21 16:53:53 +02:00
}
static const char *
pci_subclass_name(u8_t baseclass, u8_t subclass, u8_t infclass)
2005-04-21 16:53:53 +02:00
{
int i;
2005-04-21 16:53:53 +02:00
for (i= 0; pci_subclass_table[i].name; i++)
2005-04-21 16:53:53 +02:00
{
if (pci_subclass_table[i].baseclass != baseclass)
continue;
if (pci_subclass_table[i].subclass != subclass)
continue;
if (pci_subclass_table[i].infclass != infclass &&
pci_subclass_table[i].infclass != (u16_t)-1)
2005-04-21 16:53:53 +02:00
{
continue;
}
return pci_subclass_table[i].name;
}
return NULL;
}
2005-04-21 16:53:53 +02:00
static void
print_hyper_cap(int devind, u8_t capptr)
{
u32_t v;
u16_t cmd;
int type0, type1;
2005-08-09 13:23:41 +02:00
printf("\n");
v= __pci_attr_r32(devind, capptr);
printf("print_hyper_cap: @0x%x, off 0 (cap):", capptr);
cmd= (v >> 16) & 0xffff;
#if 0
if (v & 0x10000)
{
printf(" WarmReset");
v &= ~0x10000;
}
if (v & 0x20000)
{
printf(" DblEnded");
v &= ~0x20000;
}
printf(" DevNum %d", (v & 0x7C0000) >> 18);
v &= ~0x7C0000;
#endif
type0= (cmd & 0xE000) >> 13;
type1= (cmd & 0xF800) >> 11;
if (type0 == 0 || type0 == 1)
{
printf("Capability Type: %s\n",
type0 == 0 ? "Slave or Primary Interface" :
"Host or Secondary Interface");
cmd &= ~0xE000;
}
else
{
printf(" Capability Type 0x%x", type1);
cmd &= ~0xF800;
}
if (cmd)
printf(" undecoded 0x%x\n", cmd);
#if 0
printf("print_hyper_cap: off 4 (ctl): 0x%x\n",
__pci_attr_r32(devind, capptr+4));
printf("print_hyper_cap: off 8 (freq/rev): 0x%x\n",
__pci_attr_r32(devind, capptr+8));
printf("print_hyper_cap: off 12 (cap): 0x%x\n",
__pci_attr_r32(devind, capptr+12));
printf("print_hyper_cap: off 16 (buf count): 0x%x\n",
__pci_attr_r32(devind, capptr+16));
v= __pci_attr_r32(devind, capptr+20);
printf("print_hyper_cap: @0x%x, off 20 (bus nr): ",
capptr+20);
printf("prim %d", v & 0xff);
printf(", sec %d", (v >> 8) & 0xff);
printf(", sub %d", (v >> 16) & 0xff);
if (v >> 24)
printf(", reserved %d", (v >> 24) & 0xff);
printf("\n");
printf("print_hyper_cap: off 24 (type): 0x%x\n",
__pci_attr_r32(devind, capptr+24));
#endif
}
static void
print_capabilities(int devind)
{
u8_t status, capptr, type, next, subtype;
const char *str;
2005-04-21 16:53:53 +02:00
/* Check capabilities bit in the device status register */
status= __pci_attr_r16(devind, PCI_SR);
if (!(status & PSR_CAPPTR))
return;
2005-04-21 16:53:53 +02:00
capptr= (__pci_attr_r8(devind, PCI_CAPPTR) & PCI_CP_MASK);
while (capptr != 0)
{
type = __pci_attr_r8(devind, capptr+CAP_TYPE);
next= (__pci_attr_r8(devind, capptr+CAP_NEXT) & PCI_CP_MASK);
switch(type)
{
case 1: str= "PCI Power Management"; break;
case 2: str= "AGP"; break;
case 3: str= "Vital Product Data"; break;
case 4: str= "Slot Identification"; break;
case 5: str= "Message Signaled Interrupts"; break;
case 6: str= "CompactPCI Hot Swap"; break;
case 8: str= "AMD HyperTransport"; break;
case 0xf: str= "Secure Device"; break;
default: str= "(unknown type)"; break;
}
printf(" @0x%x (0x%08x): capability type 0x%x: %s",
capptr, __pci_attr_r32(devind, capptr), type, str);
if (type == 0x08)
print_hyper_cap(devind, capptr);
else if (type == 0x0f)
{
subtype= (__pci_attr_r8(devind, capptr+2) & 0x07);
switch(subtype)
{
case 0: str= "Device Exclusion Vector"; break;
case 3: str= "IOMMU"; break;
default: str= "(unknown type)"; break;
}
printf(", sub type 0%o: %s", subtype, str);
2005-04-21 16:53:53 +02:00
}
printf("\n");
capptr= next;
2005-04-21 16:53:53 +02:00
}
}
/*===========================================================================*
* ISA Bridge Helpers *
*===========================================================================*/
static void
update_bridge4dev_io(int devind, u32_t io_base, u32_t io_size)
{
int busnr, busind, type, br_devind;
u16_t v16;
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
type= pcibus[busind].pb_type;
if (type == PBT_INTEL_HOST)
return; /* Nothing to do for host controller */
if (type == PBT_PCIBRIDGE)
{
printf(
"update_bridge4dev_io: not implemented for PCI bridges\n");
return;
}
if (type != PBT_CARDBUS)
panic("update_bridge4dev_io: strange bus type: %d", type);
if (debug)
{
printf("update_bridge4dev_io: adding 0x%x at 0x%x\n",
io_size, io_base);
}
br_devind= pcibus[busind].pb_devind;
__pci_attr_w32(br_devind, CBB_IOLIMIT_0, io_base+io_size-1);
__pci_attr_w32(br_devind, CBB_IOBASE_0, io_base);
/* Enable I/O access. Enable busmaster access as well. */
v16= __pci_attr_r16(devind, PCI_CR);
__pci_attr_w16(devind, PCI_CR, v16 | PCI_CR_IO_EN | PCI_CR_MAST_EN);
}
static int
do_piix(int devind)
{
int i, s, irqrc, irq;
u32_t elcr1, elcr2, elcr;
#if DEBUG
printf("in piix\n");
#endif
if (OK != (s=sys_inb(PIIX_ELCR1, &elcr1)))
printf("Warning, sys_inb failed: %d\n", s);
if (OK != (s=sys_inb(PIIX_ELCR2, &elcr2)))
printf("Warning, sys_inb failed: %d\n", s);
elcr= elcr1 | (elcr2 << 8);
for (i= 0; i<4; i++)
{
irqrc= __pci_attr_r8(devind, PIIX_PIRQRCA+i);
if (irqrc & PIIX_IRQ_DI)
{
if (debug)
printf("INT%c: disabled\n", 'A'+i);
}
else
{
irq= irqrc & PIIX_IRQ_MASK;
if (debug)
printf("INT%c: %d\n", 'A'+i, irq);
if (!(elcr & (1 << irq)))
{
if (debug)
{
printf(
"(warning) IRQ %d is not level triggered\n",
irq);
}
}
irq_mode_pci(irq);
}
}
return 0;
}
static int
do_amd_isabr(int devind)
{
int i, busnr, dev, func, xdevind, irq, edge;
u8_t levmask;
u16_t pciirq;
/* Find required function */
func= AMD_ISABR_FUNC;
busnr= pcidev[devind].pd_busnr;
dev= pcidev[devind].pd_dev;
/* Fake a device with the required function */
if (nr_pcidev >= NR_PCIDEV)
panic("too many PCI devices: %d", nr_pcidev);
xdevind= nr_pcidev;
pcidev[xdevind].pd_busnr= busnr;
pcidev[xdevind].pd_dev= dev;
pcidev[xdevind].pd_func= func;
pcidev[xdevind].pd_inuse= 1;
nr_pcidev++;
levmask= __pci_attr_r8(xdevind, AMD_ISABR_PCIIRQ_LEV);
pciirq= __pci_attr_r16(xdevind, AMD_ISABR_PCIIRQ_ROUTE);
for (i= 0; i<4; i++)
{
edge= (levmask >> i) & 1;
irq= (pciirq >> (4*i)) & 0xf;
if (!irq)
{
if (debug)
printf("INT%c: disabled\n", 'A'+i);
}
else
{
if (debug)
printf("INT%c: %d\n", 'A'+i, irq);
if (edge && debug)
{
printf(
"(warning) IRQ %d is not level triggered\n",
irq);
}
irq_mode_pci(irq);
}
}
nr_pcidev--;
return 0;
}
static int
do_sis_isabr(int devind)
{
int i, irq;
irq= 0; /* lint */
for (i= 0; i<4; i++)
{
irq= __pci_attr_r8(devind, SIS_ISABR_IRQ_A+i);
if (irq & SIS_IRQ_DISABLED)
{
if (debug)
printf("INT%c: disabled\n", 'A'+i);
}
else
{
irq &= SIS_IRQ_MASK;
if (debug)
printf("INT%c: %d\n", 'A'+i, irq);
irq_mode_pci(irq);
}
}
return 0;
}
static int
do_via_isabr(int devind)
{
int i, irq, edge;
u8_t levmask;
levmask= __pci_attr_r8(devind, VIA_ISABR_EL);
irq= 0; /* lint */
edge= 0; /* lint */
for (i= 0; i<4; i++)
{
switch(i)
{
case 0:
edge= (levmask & VIA_ISABR_EL_INTA);
irq= __pci_attr_r8(devind, VIA_ISABR_IRQ_R2) >> 4;
break;
case 1:
edge= (levmask & VIA_ISABR_EL_INTB);
irq= __pci_attr_r8(devind, VIA_ISABR_IRQ_R2);
break;
case 2:
edge= (levmask & VIA_ISABR_EL_INTC);
irq= __pci_attr_r8(devind, VIA_ISABR_IRQ_R3) >> 4;
break;
case 3:
edge= (levmask & VIA_ISABR_EL_INTD);
irq= __pci_attr_r8(devind, VIA_ISABR_IRQ_R1) >> 4;
break;
default:
panic("PCI: VIA ISA Bridge IRQ Detection Failed");
}
irq &= 0xf;
if (!irq)
{
if (debug)
printf("INT%c: disabled\n", 'A'+i);
}
else
{
if (debug)
printf("INT%c: %d\n", 'A'+i, irq);
if (edge && debug)
2006-11-01 15:55:00 +01:00
{
printf(
"(warning) IRQ %d is not level triggered\n",
irq);
2006-11-01 15:55:00 +01:00
}
irq_mode_pci(irq);
}
}
return 0;
}
static int
do_isabridge(int busind)
{
int i, j, r, type, busnr, unknown_bridge, bridge_dev;
u16_t vid, did;
u32_t t3;
const char *dstr;
unknown_bridge= -1;
bridge_dev= -1;
j= 0; /* lint */
vid= did= 0; /* lint */
busnr= pcibus[busind].pb_busnr;
for (i= 0; i< nr_pcidev; i++)
{
if (pcidev[i].pd_busnr != busnr)
continue;
t3= ((pcidev[i].pd_baseclass << 16) |
(pcidev[i].pd_subclass << 8) | pcidev[i].pd_infclass);
if (t3 == PCI_T3_ISA)
{
/* ISA bridge. Report if no supported bridge is
* found.
*/
unknown_bridge= i;
}
vid= pcidev[i].pd_vid;
did= pcidev[i].pd_did;
for (j= 0; pci_isabridge[j].vid != 0; j++)
{
if (pci_isabridge[j].vid != vid)
continue;
if (pci_isabridge[j].did != did)
continue;
if (pci_isabridge[j].checkclass &&
unknown_bridge != i)
{
/* This part of multifunction device is
* not the bridge.
*/
continue;
}
break;
}
if (pci_isabridge[j].vid)
{
bridge_dev= i;
break;
}
}
if (bridge_dev != -1)
{
dstr= _pci_dev_name(vid, did);
if (!dstr)
dstr= "unknown device";
if (debug)
{
printf("found ISA bridge (%04X:%04X) %s\n",
vid, did, dstr);
}
pcibus[busind].pb_isabridge_dev= bridge_dev;
type= pci_isabridge[j].type;
pcibus[busind].pb_isabridge_type= type;
switch(type)
{
case PCI_IB_PIIX:
r= do_piix(bridge_dev);
break;
case PCI_IB_VIA:
r= do_via_isabr(bridge_dev);
break;
case PCI_IB_AMD:
r= do_amd_isabr(bridge_dev);
break;
case PCI_IB_SIS:
r= do_sis_isabr(bridge_dev);
break;
default:
panic("unknown ISA bridge type: %d", type);
}
return r;
}
if (unknown_bridge == -1)
{
if (debug)
{
printf("(warning) no ISA bridge found on bus %d\n",
busind);
}
return 0;
}
if (debug)
{
printf(
"(warning) unsupported ISA bridge %04X:%04X for bus %d\n",
pcidev[unknown_bridge].pd_vid,
pcidev[unknown_bridge].pd_did, busind);
}
return 0;
}
/*===========================================================================*
* IRQ handling *
*===========================================================================*/
static int
acpi_get_irq(unsigned bus, unsigned dev, unsigned pin)
{
int err;
message m;
((struct acpi_get_irq_req *)&m)->hdr.request = ACPI_REQ_GET_IRQ;
((struct acpi_get_irq_req *)&m)->bus = bus;
((struct acpi_get_irq_req *)&m)->dev = dev;
((struct acpi_get_irq_req *)&m)->pin = pin;
if ((err = ipc_sendrec(acpi_ep, &m)) != OK)
panic("PCI: error %d while receiveing from ACPI\n", err);
return ((struct acpi_get_irq_resp *)&m)->irq;
}
static int
derive_irq(struct pcidev * dev, int pin)
{
struct pcidev * parent_bridge;
int slot;
parent_bridge = &pcidev[pcibus[get_busind(dev->pd_busnr)].pb_devind];
/*
* We don't support PCI-Express, no ARI, decode the slot of the device
* and mangle the pin as the device is behind a bridge
*/
slot = ((dev->pd_func) >> 3) & 0x1f;
return acpi_get_irq(parent_bridge->pd_busnr,
parent_bridge->pd_dev, (pin + slot) % 4);
}
static void
record_irq(int devind)
{
int ilr, ipr, busnr, busind, cb_devind;
ilr= __pci_attr_r8(devind, PCI_ILR);
ipr= __pci_attr_r8(devind, PCI_IPR);
if (ipr && machine.apic_enabled) {
int irq;
irq = acpi_get_irq(pcidev[devind].pd_busnr,
pcidev[devind].pd_dev, ipr - 1);
if (irq < 0)
irq = derive_irq(&pcidev[devind], ipr - 1);
if (irq >= 0) {
ilr = irq;
__pci_attr_w8(devind, PCI_ILR, ilr);
if (debug)
printf("PCI: ACPI IRQ %d for "
"device %d.%d.%d INT%c\n",
irq,
pcidev[devind].pd_busnr,
pcidev[devind].pd_dev,
pcidev[devind].pd_func,
'A' + ipr-1);
}
else if (debug) {
printf("PCI: no ACPI IRQ routing for "
"device %d.%d.%d INT%c\n",
pcidev[devind].pd_busnr,
pcidev[devind].pd_dev,
pcidev[devind].pd_func,
'A' + ipr-1);
}
}
if (ilr == 0)
{
static int first= 1;
if (ipr && first && debug)
{
first= 0;
printf("PCI: strange, BIOS assigned IRQ0\n");
}
ilr= PCI_ILR_UNKNOWN;
}
pcidev[devind].pd_ilr= ilr;
if (ilr == PCI_ILR_UNKNOWN && !ipr)
{
}
else if (ilr != PCI_ILR_UNKNOWN && ipr)
{
if (debug)
printf("\tIRQ %d for INT%c\n", ilr, 'A' + ipr-1);
}
else if (ilr != PCI_ILR_UNKNOWN)
2010-01-13 11:52:47 +01:00
{
printf(
"PCI: IRQ %d is assigned, but device %d.%d.%d does not need it\n",
ilr, pcidev[devind].pd_busnr, pcidev[devind].pd_dev,
pcidev[devind].pd_func);
}
else
{
/* Check for cardbus devices */
busnr= pcidev[devind].pd_busnr;
busind= get_busind(busnr);
if (pcibus[busind].pb_type == PBT_CARDBUS)
{
cb_devind= pcibus[busind].pb_devind;
ilr= pcidev[cb_devind].pd_ilr;
if (ilr != PCI_ILR_UNKNOWN)
{
if (debug)
{
printf(
"assigning IRQ %d to Cardbus device\n",
ilr);
}
__pci_attr_w8(devind, PCI_ILR, ilr);
pcidev[devind].pd_ilr= ilr;
return;
}
}
if(debug) {
printf(
"PCI: device %d.%d.%d uses INT%c but is not assigned any IRQ\n",
pcidev[devind].pd_busnr, pcidev[devind].pd_dev,
pcidev[devind].pd_func, 'A' + ipr-1);
}
2010-01-13 11:52:47 +01:00
}
}
/*===========================================================================*
* BAR helpers *
*===========================================================================*/
static int
record_bar(int devind, int bar_nr, int last)
{
2010-01-13 11:52:47 +01:00
int reg, prefetch, type, dev_bar_nr, width;
u32_t bar, bar2;
u16_t cmd;
2010-01-13 11:52:47 +01:00
/* Start by assuming that this is a 32-bit bar, taking up one DWORD. */
width = 1;
reg= PCI_BAR+4*bar_nr;
bar= __pci_attr_r32(devind, reg);
if (bar & PCI_BAR_IO)
{
/* Disable I/O access before probing for BAR's size */
cmd = __pci_attr_r16(devind, PCI_CR);
__pci_attr_w16(devind, PCI_CR, cmd & ~PCI_CR_IO_EN);
/* Probe BAR's size */
__pci_attr_w32(devind, reg, 0xffffffff);
bar2= __pci_attr_r32(devind, reg);
/* Restore original state */
__pci_attr_w32(devind, reg, bar);
__pci_attr_w16(devind, PCI_CR, cmd);
bar &= PCI_BAR_IO_MASK; /* Clear non-address bits */
bar2 &= PCI_BAR_IO_MASK;
bar2= (~bar2 & 0xffff)+1;
if (debug)
{
printf("\tbar_%d: %d bytes at 0x%x I/O\n",
bar_nr, bar2, bar);
}
dev_bar_nr= pcidev[devind].pd_bar_nr++;
pcidev[devind].pd_bar[dev_bar_nr].pb_flags= PBF_IO;
pcidev[devind].pd_bar[dev_bar_nr].pb_base= bar;
pcidev[devind].pd_bar[dev_bar_nr].pb_size= bar2;
pcidev[devind].pd_bar[dev_bar_nr].pb_nr= bar_nr;
if (bar == 0)
{
pcidev[devind].pd_bar[dev_bar_nr].pb_flags |=
PBF_INCOMPLETE;
}
}
else
{
2010-01-13 11:52:47 +01:00
type= (bar & PCI_BAR_TYPE);
switch(type) {
case PCI_TYPE_32:
case PCI_TYPE_32_1M:
break;
case PCI_TYPE_64:
/* A 64-bit BAR takes up two consecutive DWORDs. */
if (last)
{
printf("PCI: device %d.%d.%d BAR %d extends"
" beyond designated area\n",
pcidev[devind].pd_busnr,
pcidev[devind].pd_dev,
pcidev[devind].pd_func, bar_nr);
return width;
}
width++;
bar2= __pci_attr_r32(devind, reg+4);
2010-01-13 11:52:47 +01:00
/* If the upper 32 bits of the BAR are not zero, the
* memory is inaccessible to us; ignore the BAR.
*/
if (bar2 != 0)
{
if (debug)
{
printf("\tbar_%d: (64-bit BAR with"
" high bits set)\n", bar_nr);
}
return width;
}
break;
default:
/* Ignore the BAR. */
if (debug)
{
printf("\tbar_%d: (unknown type %x)\n",
bar_nr, type);
}
return width;
}
/* Disable mem access before probing for BAR's size */
cmd = __pci_attr_r16(devind, PCI_CR);
__pci_attr_w16(devind, PCI_CR, cmd & ~PCI_CR_MEM_EN);
/* Probe BAR's size */
__pci_attr_w32(devind, reg, 0xffffffff);
bar2= __pci_attr_r32(devind, reg);
/* Restore original values */
__pci_attr_w32(devind, reg, bar);
__pci_attr_w16(devind, PCI_CR, cmd);
if (bar2 == 0)
2010-01-13 11:52:47 +01:00
return width; /* Reg. is not implemented */
prefetch= !!(bar & PCI_BAR_PREFETCH);
bar &= PCI_BAR_MEM_MASK; /* Clear non-address bits */
bar2 &= PCI_BAR_MEM_MASK;
bar2= (~bar2)+1;
if (debug)
{
2010-01-13 11:52:47 +01:00
printf("\tbar_%d: 0x%x bytes at 0x%x%s memory%s\n",
bar_nr, bar2, bar,
2010-01-13 11:52:47 +01:00
prefetch ? " prefetchable" : "",
type == PCI_TYPE_64 ? ", 64-bit" : "");
}
dev_bar_nr= pcidev[devind].pd_bar_nr++;
pcidev[devind].pd_bar[dev_bar_nr].pb_flags= 0;
pcidev[devind].pd_bar[dev_bar_nr].pb_base= bar;
pcidev[devind].pd_bar[dev_bar_nr].pb_size= bar2;
pcidev[devind].pd_bar[dev_bar_nr].pb_nr= bar_nr;
if (bar == 0)
{
pcidev[devind].pd_bar[dev_bar_nr].pb_flags |=
PBF_INCOMPLETE;
}
}
return width;
}
static void
record_bars(int devind, int last_reg)
{
int i, reg, width;
for (i= 0, reg= PCI_BAR; reg <= last_reg; i += width, reg += 4 * width)
{
width = record_bar(devind, i, reg == last_reg);
}
}
static void
record_bars_normal(int devind)
{
int i, j, clear_01, clear_23, pb_nr;
/* The BAR area of normal devices is six DWORDs in size. */
record_bars(devind, PCI_BAR_6);
/* Special case code for IDE controllers in compatibility mode */
if (pcidev[devind].pd_baseclass == PCI_BCR_MASS_STORAGE &&
pcidev[devind].pd_subclass == PCI_MS_IDE)
{
/* IDE device */
clear_01= 0;
clear_23= 0;
if (!(pcidev[devind].pd_infclass & PCI_IDE_PRI_NATIVE))
{
if (debug)
{
printf(
"primary channel is not in native mode, clearing BARs 0 and 1\n");
}
clear_01= 1;
}
if (!(pcidev[devind].pd_infclass & PCI_IDE_SEC_NATIVE))
{
if (debug)
{
printf(
"secondary channel is not in native mode, clearing BARs 2 and 3\n");
}
clear_23= 1;
}
j= 0;
for (i= 0; i<pcidev[devind].pd_bar_nr; i++)
{
pb_nr= pcidev[devind].pd_bar[i].pb_nr;
if ((pb_nr == 0 || pb_nr == 1) && clear_01)
{
if (debug) printf("skipping bar %d\n", pb_nr);
continue; /* Skip */
}
if ((pb_nr == 2 || pb_nr == 3) && clear_23)
{
if (debug) printf("skipping bar %d\n", pb_nr);
continue; /* Skip */
}
if (i == j)
{
j++;
continue; /* No need to copy */
}
pcidev[devind].pd_bar[j]=
pcidev[devind].pd_bar[i];
j++;
}
pcidev[devind].pd_bar_nr= j;
}
}
static void
record_bars_bridge(int devind)
{
u32_t base, limit, size;
/* The generic BAR area of PCI-to-PCI bridges is two DWORDs in size.
* It may contain up to two 32-bit BARs, or one 64-bit BAR.
*/
record_bars(devind, PCI_BAR_2);
base= ((__pci_attr_r8(devind, PPB_IOBASE) & PPB_IOB_MASK) << 8) |
(__pci_attr_r16(devind, PPB_IOBASEU16) << 16);
limit= 0xff |
((__pci_attr_r8(devind, PPB_IOLIMIT) & PPB_IOL_MASK) << 8) |
((~PPB_IOL_MASK & 0xff) << 8) |
(__pci_attr_r16(devind, PPB_IOLIMITU16) << 16);
size= limit-base + 1;
if (debug)
{
printf("\tI/O window: base 0x%x, limit 0x%x, size %d\n",
base, limit, size);
}
base= ((__pci_attr_r16(devind, PPB_MEMBASE) & PPB_MEMB_MASK) << 16);
limit= 0xffff |
((__pci_attr_r16(devind, PPB_MEMLIMIT) & PPB_MEML_MASK) << 16) |
((~PPB_MEML_MASK & 0xffff) << 16);
size= limit-base + 1;
if (debug)
{
printf("\tMemory window: base 0x%x, limit 0x%x, size 0x%x\n",
base, limit, size);
}
/* Ignore the upper 32 bits */
base= ((__pci_attr_r16(devind, PPB_PFMEMBASE) & PPB_PFMEMB_MASK) << 16);
limit= 0xffff |
((__pci_attr_r16(devind, PPB_PFMEMLIMIT) &
PPB_PFMEML_MASK) << 16) |
((~PPB_PFMEML_MASK & 0xffff) << 16);
size= limit-base + 1;
if (debug)
{
printf(
"\tPrefetchable memory window: base 0x%x, limit 0x%x, size 0x%x\n",
base, limit, size);
}
}
static void
record_bars_cardbus(int devind)
{
u32_t base, limit, size;
/* The generic BAR area of CardBus devices is one DWORD in size. */
record_bars(devind, PCI_BAR);
base= __pci_attr_r32(devind, CBB_MEMBASE_0);
limit= __pci_attr_r32(devind, CBB_MEMLIMIT_0) |
(~CBB_MEML_MASK & 0xffffffff);
size= limit-base + 1;
if (debug)
{
printf("\tMemory window 0: base 0x%x, limit 0x%x, size %d\n",
base, limit, size);
}
base= __pci_attr_r32(devind, CBB_MEMBASE_1);
limit= __pci_attr_r32(devind, CBB_MEMLIMIT_1) |
(~CBB_MEML_MASK & 0xffffffff);
size= limit-base + 1;
if (debug)
{
printf("\tMemory window 1: base 0x%x, limit 0x%x, size %d\n",
base, limit, size);
}
base= __pci_attr_r32(devind, CBB_IOBASE_0);
limit= __pci_attr_r32(devind, CBB_IOLIMIT_0) |
(~CBB_IOL_MASK & 0xffffffff);
size= limit-base + 1;
if (debug)
{
printf("\tI/O window 0: base 0x%x, limit 0x%x, size %d\n",
base, limit, size);
}
base= __pci_attr_r32(devind, CBB_IOBASE_1);
limit= __pci_attr_r32(devind, CBB_IOLIMIT_1) |
(~CBB_IOL_MASK & 0xffffffff);
size= limit-base + 1;
if (debug)
{
printf("\tI/O window 1: base 0x%x, limit 0x%x, size %d\n",
base, limit, size);
}
}
static void
complete_bars(void)
{
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
int i, j, bar_nr, reg;
u32_t memgap_low, memgap_high, iogap_low, iogap_high, io_high,
base, size, v32, diff1, diff2;
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
kinfo_t kinfo;
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
if(OK != sys_getkinfo(&kinfo))
panic("can't get kinfo");
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
/* Set memgap_low to just above physical memory */
memgap_low= kinfo.mem_high_phys;
memgap_high= 0xfe000000; /* Leave space for the CPU (APIC) */
if (debug)
{
printf("complete_bars: initial gap: [0x%x .. 0x%x>\n",
memgap_low, memgap_high);
}
/* Find the lowest memory base */
for (i= 0; i<nr_pcidev; i++)
{
for (j= 0; j<pcidev[i].pd_bar_nr; j++)
{
if (pcidev[i].pd_bar[j].pb_flags & PBF_IO)
continue;
if (pcidev[i].pd_bar[j].pb_flags & PBF_INCOMPLETE)
continue;
base= pcidev[i].pd_bar[j].pb_base;
size= pcidev[i].pd_bar[j].pb_size;
if (base >= memgap_high)
continue; /* Not in the gap */
if (base+size <= memgap_low)
continue; /* Not in the gap */
/* Reduce the gap by the smallest amount */
diff1= base+size-memgap_low;
diff2= memgap_high-base;
if (diff1 < diff2)
memgap_low= base+size;
else
memgap_high= base;
}
}
if (debug)
{
printf("complete_bars: intermediate gap: [0x%x .. 0x%x>\n",
memgap_low, memgap_high);
}
/* Should check main memory size */
if (memgap_high < memgap_low)
{
printf("PCI: bad memory gap: [0x%x .. 0x%x>\n",
memgap_low, memgap_high);
panic(NULL);
}
iogap_high= 0x10000;
iogap_low= 0x400;
/* Find the free I/O space */
for (i= 0; i<nr_pcidev; i++)
{
for (j= 0; j<pcidev[i].pd_bar_nr; j++)
{
if (!(pcidev[i].pd_bar[j].pb_flags & PBF_IO))
continue;
if (pcidev[i].pd_bar[j].pb_flags & PBF_INCOMPLETE)
continue;
base= pcidev[i].pd_bar[j].pb_base;
size= pcidev[i].pd_bar[j].pb_size;
if (base >= iogap_high)
continue;
if (base+size <= iogap_low)
continue;
#if 0
2006-04-11 13:50:29 +02:00
if (debug)
{
printf(
"pci device %d (%04x:%04x), bar %d: base 0x%x, size 0x%x\n",
2006-04-11 13:50:29 +02:00
i, pcidev[i].pd_vid, pcidev[i].pd_did,
j, base, size);
}
#endif
if (base+size-iogap_low < iogap_high-base)
iogap_low= base+size;
else
iogap_high= base;
}
}
if (iogap_high < iogap_low)
2006-04-11 13:50:29 +02:00
{
if (debug)
{
printf("iogap_high too low, should panic\n");
}
else
panic("iogap_high too low: %d", iogap_high);
2006-04-11 13:50:29 +02:00
}
if (debug)
printf("I/O range = [0x%x..0x%x>\n", iogap_low, iogap_high);
for (i= 0; i<nr_pcidev; i++)
{
for (j= 0; j<pcidev[i].pd_bar_nr; j++)
{
if (pcidev[i].pd_bar[j].pb_flags & PBF_IO)
continue;
if (!(pcidev[i].pd_bar[j].pb_flags & PBF_INCOMPLETE))
continue;
size= pcidev[i].pd_bar[j].pb_size;
if (size < PAGE_SIZE)
size= PAGE_SIZE;
base= memgap_high-size;
base &= ~(u32_t)(size-1);
if (base < memgap_low)
panic("memory base too low: %d", base);
memgap_high= base;
bar_nr= pcidev[i].pd_bar[j].pb_nr;
reg= PCI_BAR + 4*bar_nr;
v32= __pci_attr_r32(i, reg);
__pci_attr_w32(i, reg, v32 | base);
if (debug)
{
printf(
"complete_bars: allocated 0x%x size %d to %d.%d.%d, bar_%d\n",
base, size, pcidev[i].pd_busnr,
pcidev[i].pd_dev, pcidev[i].pd_func,
bar_nr);
}
pcidev[i].pd_bar[j].pb_base= base;
pcidev[i].pd_bar[j].pb_flags &= ~PBF_INCOMPLETE;
}
io_high= iogap_high;
for (j= 0; j<pcidev[i].pd_bar_nr; j++)
{
if (!(pcidev[i].pd_bar[j].pb_flags & PBF_IO))
continue;
if (!(pcidev[i].pd_bar[j].pb_flags & PBF_INCOMPLETE))
continue;
size= pcidev[i].pd_bar[j].pb_size;
base= iogap_high-size;
base &= ~(u32_t)(size-1);
/* Assume that ISA compatibility is required. Only
* use the lowest 256 bytes out of every 1024 bytes.
*/
base &= 0xfcff;
if (base < iogap_low)
panic("I/O base too low: %d", base);
iogap_high= base;
bar_nr= pcidev[i].pd_bar[j].pb_nr;
reg= PCI_BAR + 4*bar_nr;
v32= __pci_attr_r32(i, reg);
__pci_attr_w32(i, reg, v32 | base);
if (debug)
{
printf(
"complete_bars: allocated 0x%x size %d to %d.%d.%d, bar_%d\n",
base, size, pcidev[i].pd_busnr,
pcidev[i].pd_dev, pcidev[i].pd_func,
bar_nr);
}
pcidev[i].pd_bar[j].pb_base= base;
pcidev[i].pd_bar[j].pb_flags &= ~PBF_INCOMPLETE;
}
if (iogap_high != io_high)
{
update_bridge4dev_io(i, iogap_high,
io_high-iogap_high);
}
}
for (i= 0; i<nr_pcidev; i++)
{
for (j= 0; j<pcidev[i].pd_bar_nr; j++)
{
if (!(pcidev[i].pd_bar[j].pb_flags & PBF_INCOMPLETE))
continue;
printf("should allocate resources for device %d\n", i);
}
}
return;
}
/*===========================================================================*
* PCI Bridge Helpers *
*===========================================================================*/
static void
probe_bus(int busind)
{
u32_t dev, func, t3;
u16_t vid, did, sts, sub_vid, sub_did;
u8_t headt;
u8_t baseclass, subclass, infclass;
int devind, busnr;
const char *s, *dstr;
if (debug)
printf("probe_bus(%d)\n", busind);
if (nr_pcidev >= NR_PCIDEV)
panic("too many PCI devices: %d", nr_pcidev);
devind= nr_pcidev;
busnr= pcibus[busind].pb_busnr;
for (dev= 0; dev<32; dev++)
{
for (func= 0; func < 8; func++)
{
pcidev[devind].pd_busnr= busnr;
pcidev[devind].pd_dev= dev;
pcidev[devind].pd_func= func;
pci_attr_wsts(devind,
PSR_SSE|PSR_RMAS|PSR_RTAS);
vid= __pci_attr_r16(devind, PCI_VID);
did= __pci_attr_r16(devind, PCI_DID);
headt= __pci_attr_r8(devind, PCI_HEADT);
sts= pci_attr_rsts(devind);
#if 0
printf("vid 0x%x, did 0x%x, headt 0x%x, sts 0x%x\n",
vid, did, headt, sts);
#endif
if (vid == NO_VID && did == NO_VID)
{
if (func == 0)
break; /* Nothing here */
/* Scan all functions of a multifunction
* device.
*/
continue;
}
if (sts & (PSR_SSE|PSR_RMAS|PSR_RTAS))
{
static int warned = 0;
if(!warned) {
printf(
"PCI: ignoring bad value 0x%x in sts for QEMU\n",
sts & (PSR_SSE|PSR_RMAS|PSR_RTAS));
warned = 1;
}
}
sub_vid= __pci_attr_r16(devind, PCI_SUBVID);
sub_did= __pci_attr_r16(devind, PCI_SUBDID);
dstr= _pci_dev_name(vid, did);
if (debug)
{
if (dstr)
{
printf("%d.%lu.%lu: %s (%04X:%04X)\n",
busnr, (unsigned long)dev,
(unsigned long)func, dstr,
vid, did);
}
else
{
printf(
"%d.%lu.%lu: Unknown device, vendor %04X (%s), device %04X\n",
busnr, (unsigned long)dev,
(unsigned long)func, vid,
pci_vid_name(vid), did);
}
printf("Device index: %d\n", devind);
printf("Subsystem: Vid 0x%x, did 0x%x\n",
sub_vid, sub_did);
}
baseclass= __pci_attr_r8(devind, PCI_BCR);
subclass= __pci_attr_r8(devind, PCI_SCR);
infclass= __pci_attr_r8(devind, PCI_PIFR);
s= pci_subclass_name(baseclass, subclass, infclass);
if (!s)
s= pci_baseclass_name(baseclass);
{
if (!s)
s= "(unknown class)";
}
if (debug)
{
printf("\tclass %s (%X/%X/%X)\n", s,
baseclass, subclass, infclass);
}
if (is_duplicate(busnr, dev, func))
{
printf("\tduplicate!\n");
if (func == 0 && !(headt & PHT_MULTIFUNC))
break;
continue;
}
devind= nr_pcidev;
nr_pcidev++;
pcidev[devind].pd_baseclass= baseclass;
pcidev[devind].pd_subclass= subclass;
pcidev[devind].pd_infclass= infclass;
pcidev[devind].pd_vid= vid;
pcidev[devind].pd_did= did;
pcidev[devind].pd_sub_vid= sub_vid;
pcidev[devind].pd_sub_did= sub_did;
pcidev[devind].pd_inuse= 0;
pcidev[devind].pd_bar_nr= 0;
record_irq(devind);
switch(headt & PHT_MASK)
{
case PHT_NORMAL:
record_bars_normal(devind);
break;
case PHT_BRIDGE:
record_bars_bridge(devind);
break;
case PHT_CARDBUS:
record_bars_cardbus(devind);
break;
default:
printf("\t%d.%d.%d: unknown header type %d\n",
busind, dev, func,
headt & PHT_MASK);
break;
}
if (debug)
print_capabilities(devind);
t3= ((baseclass << 16) | (subclass << 8) | infclass);
#if 0
if (t3 == PCI_T3_VGA || t3 == PCI_T3_VGA_OLD)
report_vga(devind);
#endif
if (nr_pcidev >= NR_PCIDEV)
panic("too many PCI devices: %d", nr_pcidev);
devind= nr_pcidev;
if (func == 0 && !(headt & PHT_MULTIFUNC))
break;
}
}
}
static u16_t
pcibr_std_rsts(int busind)
{
int devind;
devind= pcibus[busind].pb_devind;
return __pci_attr_r16(devind, PPB_SSTS);
}
static void
pcibr_std_wsts(int busind, u16_t value)
{
int devind;
devind= pcibus[busind].pb_devind;
#if 0
printf("pcibr_std_wsts(%d, 0x%X), devind= %d\n",
busind, value, devind);
#endif
__pci_attr_w16(devind, PPB_SSTS, value);
}
static u16_t
pcibr_cb_rsts(int busind)
{
int devind;
devind= pcibus[busind].pb_devind;
return __pci_attr_r16(devind, CBB_SSTS);
}
static void
pcibr_cb_wsts(int busind, u16_t value)
{
int devind;
devind= pcibus[busind].pb_devind;
#if 0
printf("pcibr_cb_wsts(%d, 0x%X), devind= %d\n",
busind, value, devind);
#endif
__pci_attr_w16(devind, CBB_SSTS, value);
}
static u16_t
pcibr_via_rsts(int busind)
{
return 0;
}
static void
pcibr_via_wsts(int busind, u16_t value)
2005-04-21 16:53:53 +02:00
{
int devind;
devind= pcibus[busind].pb_devind;
2005-04-21 16:53:53 +02:00
#if 0
printf("pcibr_via_wsts(%d, 0x%X), devind= %d (not implemented)\n",
busind, value, devind);
#endif
}
static void
complete_bridges(void)
{
int i, freebus, devind, prim_busnr;
for (i= 0; i<nr_pcibus; i++)
2005-04-21 16:53:53 +02:00
{
if (!pcibus[i].pb_needinit)
2005-04-21 16:53:53 +02:00
continue;
printf("should allocate bus number for bus %d\n", i);
freebus= get_freebus();
printf("got bus number %d\n", freebus);
2005-04-21 16:53:53 +02:00
devind= pcibus[i].pb_devind;
2005-04-21 16:53:53 +02:00
prim_busnr= pcidev[devind].pd_busnr;
if (prim_busnr != 0)
2005-04-21 16:53:53 +02:00
{
printf(
"complete_bridge: updating subordinate bus number not implemented\n");
2005-04-21 16:53:53 +02:00
}
pcibus[i].pb_needinit= 0;
pcibus[i].pb_busnr= freebus;
printf("devind = %d\n", devind);
printf("prim_busnr= %d\n", prim_busnr);
__pci_attr_w8(devind, PPB_PRIMBN, prim_busnr);
__pci_attr_w8(devind, PPB_SECBN, freebus);
__pci_attr_w8(devind, PPB_SUBORDBN, freebus);
printf("CR = 0x%x\n", __pci_attr_r16(devind, PCI_CR));
printf("SECBLT = 0x%x\n", __pci_attr_r8(devind, PPB_SECBLT));
printf("BRIDGECTRL = 0x%x\n",
__pci_attr_r16(devind, PPB_BRIDGECTRL));
2006-04-11 13:50:29 +02:00
}
2005-04-21 16:53:53 +02:00
}
/*
* tells acpi which two busses are connected by this bridge. The primary bus
* (pbnr) must be already known to acpi and it must map dev as the connection to
* the secondary (sbnr) bus
*/
static void
acpi_map_bridge(unsigned int pbnr, unsigned int dev, unsigned int sbnr)
{
int err;
message m;
((struct acpi_map_bridge_req *)&m)->hdr.request = ACPI_REQ_MAP_BRIDGE;
((struct acpi_map_bridge_req *)&m)->primary_bus = pbnr;
((struct acpi_map_bridge_req *)&m)->secondary_bus = sbnr;
((struct acpi_map_bridge_req *)&m)->device = dev;
if ((err = ipc_sendrec(acpi_ep, &m)) != OK)
panic("PCI: error %d while receiveing from ACPI\n", err);
if (((struct acpi_map_bridge_resp *)&m)->err != OK)
printf("PCI: acpi failed to map pci (%d) to pci (%d) bridge\n",
pbnr, sbnr);
}
static void
do_pcibridge(int busind)
2005-04-21 16:53:53 +02:00
{
int i, devind, busnr;
2005-04-21 16:53:53 +02:00
int ind, type;
u16_t vid, did;
u8_t sbusn, baseclass, subclass, infclass, headt;
u32_t t3;
2005-04-21 16:53:53 +02:00
vid= did= 0; /* lint */
busnr= pcibus[busind].pb_busnr;
2005-04-21 16:53:53 +02:00
for (devind= 0; devind< nr_pcidev; devind++)
{
#if 0
printf("do_pcibridge: trying %u.%u.%u\n",
pcidev[devind].pd_busnr, pcidev[devind].pd_dev,
pcidev[devind].pd_func);
#endif
if (pcidev[devind].pd_busnr != busnr)
{
#if 0
printf("wrong bus\n");
#endif
2005-04-21 16:53:53 +02:00
continue;
}
2005-04-21 16:53:53 +02:00
vid= pcidev[devind].pd_vid;
did= pcidev[devind].pd_did;
for (i= 0; pci_pcibridge[i].vid != 0; i++)
{
if (pci_pcibridge[i].vid != vid)
continue;
if (pci_pcibridge[i].did != did)
continue;
break;
}
type= pci_pcibridge[i].type;
2005-04-21 16:53:53 +02:00
if (pci_pcibridge[i].vid == 0)
{
headt= __pci_attr_r8(devind, PCI_HEADT);
type= 0;
if ((headt & PHT_MASK) == PHT_BRIDGE)
type= PCI_PPB_STD;
else if ((headt & PHT_MASK) == PHT_CARDBUS)
type= PCI_PPB_CB;
else
{
#if 0
printf("not a bridge\n");
#endif
continue; /* Not a bridge */
}
baseclass= __pci_attr_r8(devind, PCI_BCR);
subclass= __pci_attr_r8(devind, PCI_SCR);
infclass= __pci_attr_r8(devind, PCI_PIFR);
t3= ((baseclass << 16) | (subclass << 8) | infclass);
if (type == PCI_PPB_STD &&
t3 != PCI_T3_PCI2PCI &&
t3 != PCI_T3_PCI2PCI_SUBTR)
{
printf(
"Unknown PCI class %02x/%02x/%02x for PCI-to-PCI bridge, device %04X:%04X\n",
baseclass, subclass, infclass,
vid, did);
continue;
}
if (type == PCI_PPB_CB &&
t3 != PCI_T3_CARDBUS)
{
printf(
"Unknown PCI class %02x/%02x/%02x for Cardbus bridge, device %04X:%04X\n",
baseclass, subclass, infclass,
vid, did);
continue;
}
}
2005-04-21 16:53:53 +02:00
if (debug)
{
printf("%u.%u.%u: PCI-to-PCI bridge: %04X:%04X\n",
pcidev[devind].pd_busnr,
pcidev[devind].pd_dev,
pcidev[devind].pd_func, vid, did);
}
2005-04-21 16:53:53 +02:00
/* Assume that the BIOS initialized the secondary bus
* number.
*/
sbusn= __pci_attr_r8(devind, PPB_SECBN);
2005-04-21 16:53:53 +02:00
if (nr_pcibus >= NR_PCIBUS)
panic("too many PCI busses: %d", nr_pcibus);
2005-04-21 16:53:53 +02:00
ind= nr_pcibus;
nr_pcibus++;
pcibus[ind].pb_type= PBT_PCIBRIDGE;
pcibus[ind].pb_needinit= 1;
2005-04-21 16:53:53 +02:00
pcibus[ind].pb_isabridge_dev= -1;
pcibus[ind].pb_isabridge_type= 0;
pcibus[ind].pb_devind= devind;
pcibus[ind].pb_busnr= sbusn;
2005-04-21 16:53:53 +02:00
pcibus[ind].pb_rreg8= pcibus[busind].pb_rreg8;
pcibus[ind].pb_rreg16= pcibus[busind].pb_rreg16;
pcibus[ind].pb_rreg32= pcibus[busind].pb_rreg32;
pcibus[ind].pb_wreg8= pcibus[busind].pb_wreg8;
2005-04-21 16:53:53 +02:00
pcibus[ind].pb_wreg16= pcibus[busind].pb_wreg16;
pcibus[ind].pb_wreg32= pcibus[busind].pb_wreg32;
switch(type)
{
case PCI_PPB_STD:
pcibus[ind].pb_rsts= pcibr_std_rsts;
pcibus[ind].pb_wsts= pcibr_std_wsts;
break;
case PCI_PPB_CB:
pcibus[ind].pb_type= PBT_CARDBUS;
pcibus[ind].pb_rsts= pcibr_cb_rsts;
pcibus[ind].pb_wsts= pcibr_cb_wsts;
2005-04-21 16:53:53 +02:00
break;
case PCI_AGPB_VIA:
pcibus[ind].pb_rsts= pcibr_via_rsts;
pcibus[ind].pb_wsts= pcibr_via_wsts;
break;
default:
panic("unknown PCI-PCI bridge type: %d", type);
2005-04-21 16:53:53 +02:00
}
if (machine.apic_enabled)
acpi_map_bridge(pcidev[devind].pd_busnr,
pcidev[devind].pd_dev, sbusn);
if (debug)
2005-04-21 16:53:53 +02:00
{
printf(
"bus(table) = %d, bus(sec) = %d, bus(subord) = %d\n",
ind, sbusn, __pci_attr_r8(devind, PPB_SUBORDBN));
2005-04-21 16:53:53 +02:00
}
if (sbusn == 0)
2005-04-21 16:53:53 +02:00
{
printf("Secondary bus number not initialized\n");
continue;
2005-04-21 16:53:53 +02:00
}
pcibus[ind].pb_needinit= 0;
probe_bus(ind);
/* Look for PCI bridges */
do_pcibridge(ind);
2005-04-21 16:53:53 +02:00
}
}
/*===========================================================================*
* pci_intel_init *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
static void
pci_intel_init()
2005-04-21 16:53:53 +02:00
{
/* Try to detect a know PCI controller. Read the Vendor ID and
* the Device ID for function 0 of device 0.
* Two times the value 0xffff suggests a system without a (compatible)
* PCI controller.
*/
u32_t bus, dev, func;
u16_t vid, did;
int s, i, r, busind, busnr;
const char *dstr;
2005-04-21 16:53:53 +02:00
bus= 0;
dev= 0;
func= 0;
vid= PCII_RREG16_(bus, dev, func, PCI_VID);
did= PCII_RREG16_(bus, dev, func, PCI_DID);
if (OK != (s=sys_outl(PCII_CONFADD, PCII_UNSEL)))
printf("PCI: warning, sys_outl failed: %d\n", s);
#if 0
if (vid == 0xffff && did == 0xffff)
return; /* Nothing here */
for (i= 0; pci_intel_ctrl[i].vid; i++)
2005-04-21 16:53:53 +02:00
{
if (pci_intel_ctrl[i].vid == vid &&
pci_intel_ctrl[i].did == did)
2005-04-21 16:53:53 +02:00
{
break;
}
}
if (!pci_intel_ctrl[i].vid)
{
printf("pci_intel_init (warning): unknown PCI-controller:\n"
"\tvendor %04X (%s), device %04X\n",
vid, pci_vid_name(vid), did);
}
#endif
if (nr_pcibus >= NR_PCIBUS)
panic("too many PCI busses: %d", nr_pcibus);
busind= nr_pcibus;
nr_pcibus++;
pcibus[busind].pb_type= PBT_INTEL_HOST;
pcibus[busind].pb_needinit= 0;
pcibus[busind].pb_isabridge_dev= -1;
pcibus[busind].pb_isabridge_type= 0;
pcibus[busind].pb_devind= -1;
pcibus[busind].pb_busnr= 0;
pcibus[busind].pb_rreg8= pcii_rreg8;
pcibus[busind].pb_rreg16= pcii_rreg16;
pcibus[busind].pb_rreg32= pcii_rreg32;
pcibus[busind].pb_wreg8= pcii_wreg8;
pcibus[busind].pb_wreg16= pcii_wreg16;
pcibus[busind].pb_wreg32= pcii_wreg32;
pcibus[busind].pb_rsts= pcii_rsts;
pcibus[busind].pb_wsts= pcii_wsts;
dstr= _pci_dev_name(vid, did);
if (!dstr)
dstr= "unknown device";
if (debug)
{
printf("pci_intel_init: %s (%04X:%04X)\n",
dstr, vid, did);
}
probe_bus(busind);
r= do_isabridge(busind);
if (r != OK)
{
busnr= pcibus[busind].pb_busnr;
/* Disable all devices for this bus */
for (i= 0; i<nr_pcidev; i++)
2005-04-21 16:53:53 +02:00
{
if (pcidev[i].pd_busnr != busnr)
continue;
pcidev[i].pd_inuse= 1;
2005-04-21 16:53:53 +02:00
}
return;
2005-04-21 16:53:53 +02:00
}
/* Look for PCI bridges */
do_pcibridge(busind);
/* Allocate bus numbers for uninitialized bridges */
complete_bridges();
/* Allocate I/O and memory resources for uninitialized devices */
complete_bars();
}
#if 0
/*===========================================================================*
* report_vga *
*===========================================================================*/
static void
report_vga(int devind)
{
/* Report the amount of video memory. This is needed by the X11R6
* postinstall script to chmem the X server. Hopefully this can be
* removed when we get virtual memory.
*/
size_t amount, size;
int i;
amount= 0;
for (i= 0; i<pcidev[devind].pd_bar_nr; i++)
{
if (pcidev[devind].pd_bar[i].pb_flags & PBF_IO)
continue;
size= pcidev[devind].pd_bar[i].pb_size;
if (size < amount)
continue;
amount= size;
}
if (size != 0)
{
printf("PCI: video memory for device at %d.%d.%d: %d bytes\n",
pcidev[devind].pd_busnr,
pcidev[devind].pd_dev,
pcidev[devind].pd_func,
amount);
}
}
#endif
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* visible *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
static int
visible(struct rs_pci *aclp, int devind)
2005-04-21 16:53:53 +02:00
{
u16_t acl_sub_vid, acl_sub_did;
2005-04-21 16:53:53 +02:00
int i;
u32_t class_id;
2005-04-21 16:53:53 +02:00
if (!aclp)
return TRUE; /* Should be changed when ACLs become
* mandatory. Do note that procfs relies
* on being able to see all devices.
*/
/* Check whether the caller is allowed to get this device. */
for (i= 0; i<aclp->rsp_nr_device; i++)
2005-04-21 16:53:53 +02:00
{
acl_sub_vid = aclp->rsp_device[i].sub_vid;
acl_sub_did = aclp->rsp_device[i].sub_did;
if (aclp->rsp_device[i].vid == pcidev[devind].pd_vid &&
aclp->rsp_device[i].did == pcidev[devind].pd_did &&
(acl_sub_vid == NO_SUB_VID ||
acl_sub_vid == pcidev[devind].pd_sub_vid) &&
(acl_sub_did == NO_SUB_DID ||
acl_sub_did == pcidev[devind].pd_sub_did))
{
return TRUE;
}
2005-04-21 16:53:53 +02:00
}
if (!aclp->rsp_nr_class)
return FALSE;
2005-04-21 16:53:53 +02:00
class_id= (pcidev[devind].pd_baseclass << 16) |
(pcidev[devind].pd_subclass << 8) |
pcidev[devind].pd_infclass;
for (i= 0; i<aclp->rsp_nr_class; i++)
2005-04-21 16:53:53 +02:00
{
if (aclp->rsp_class[i].pciclass ==
(class_id & aclp->rsp_class[i].mask))
2005-04-21 16:53:53 +02:00
{
return TRUE;
2005-04-21 16:53:53 +02:00
}
}
return FALSE;
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* sef_cb_init_fresh *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
int
sef_cb_init_fresh(int type, sef_init_info_t *info)
2005-04-21 16:53:53 +02:00
{
/* Initialize the pci driver. */
long v;
int i, r;
struct rprocpub rprocpub[NR_BOOT_PROCS];
2005-04-21 16:53:53 +02:00
v= 0;
env_parse("pci_debug", "d", 0, &v, 0, 1);
debug= v;
if (sys_getmachine(&machine)) {
printf("PCI: no machine\n");
return ENODEV;
2005-04-21 16:53:53 +02:00
}
if (machine.apic_enabled &&
ds_retrieve_label_endpt("acpi", &acpi_ep) != OK) {
panic("PCI: Cannot use APIC mode without ACPI!\n");
2005-04-21 16:53:53 +02:00
}
/* Only Intel (compatible) PCI controllers are supported at the
* moment.
*/
pci_intel_init();
/* Map all the services in the boot image. */
if((r = sys_safecopyfrom(RS_PROC_NR, info->rproctab_gid, 0,
(vir_bytes) rprocpub, sizeof(rprocpub))) != OK) {
panic("sys_safecopyfrom failed: %d", r);
}
for(i=0;i < NR_BOOT_PROCS;i++) {
if(rprocpub[i].in_use) {
if((r = map_service(&rprocpub[i])) != OK) {
panic("unable to map service: %d", r);
}
2005-04-21 16:53:53 +02:00
}
}
return(OK);
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* map_service *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
int
map_service(struct rprocpub *rpub)
2005-04-21 16:53:53 +02:00
{
/* Map a new service by registering a new acl entry if required. */
int i;
2005-04-21 16:53:53 +02:00
/* Stop right now if no pci device or class is found. */
if(rpub->pci_acl.rsp_nr_device == 0
&& rpub->pci_acl.rsp_nr_class == 0) {
return(OK);
}
2005-04-21 16:53:53 +02:00
/* Find a free acl slot. */
for (i= 0; i<NR_DRIVERS; i++)
{
if (!pci_acl[i].inuse)
break;
}
if (i >= NR_DRIVERS)
{
printf("PCI: map_service: table is full\n");
return ENOMEM;
}
2005-04-21 16:53:53 +02:00
/* Initialize acl slot. */
pci_acl[i].inuse = 1;
pci_acl[i].acl = rpub->pci_acl;
return(OK);
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* _pci_find_dev *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
int
_pci_find_dev(u8_t bus, u8_t dev, u8_t func, int *devindp)
2005-04-21 16:53:53 +02:00
{
int devind;
for (devind= 0; devind < nr_pcidev; devind++)
{
if (pcidev[devind].pd_busnr == bus &&
pcidev[devind].pd_dev == dev &&
pcidev[devind].pd_func == func)
{
break;
}
}
if (devind >= nr_pcidev)
return 0;
2005-04-21 16:53:53 +02:00
#if 0
if (pcidev[devind].pd_inuse)
return 0;
2005-04-21 16:53:53 +02:00
#endif
*devindp= devind;
return 1;
}
/*===========================================================================*
* _pci_first_dev *
*===========================================================================*/
int
_pci_first_dev(struct rs_pci *aclp, int *devindp, u16_t *vidp,
u16_t *didp)
{
int devind;
for (devind= 0; devind < nr_pcidev; devind++)
{
#if 0
if (pcidev[devind].pd_inuse)
continue;
#endif
if (!visible(aclp, devind))
continue;
break;
}
if (devind >= nr_pcidev)
return 0;
*devindp= devind;
*vidp= pcidev[devind].pd_vid;
*didp= pcidev[devind].pd_did;
return 1;
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* _pci_next_dev *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
int
_pci_next_dev(struct rs_pci *aclp, int *devindp, u16_t *vidp, u16_t *didp)
2005-04-21 16:53:53 +02:00
{
int devind;
for (devind= *devindp+1; devind < nr_pcidev; devind++)
{
2005-04-21 16:53:53 +02:00
#if 0
if (pcidev[devind].pd_inuse)
continue;
2005-04-21 16:53:53 +02:00
#endif
if (!visible(aclp, devind))
continue;
break;
}
if (devind >= nr_pcidev)
return 0;
*devindp= devind;
*vidp= pcidev[devind].pd_vid;
*didp= pcidev[devind].pd_did;
return 1;
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* _pci_reserve *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
int
_pci_reserve(int devind, endpoint_t proc, struct rs_pci *aclp)
2005-04-21 16:53:53 +02:00
{
int i, r;
int ilr;
struct io_range ior;
struct minix_mem_range mr;
2005-04-21 16:53:53 +02:00
if (devind < 0 || devind >= nr_pcidev)
{
printf("pci_reserve_a: bad devind: %d\n", devind);
return EINVAL;
}
if (!visible(aclp, devind))
{
printf("pci_reserve_a: %u is not allowed to reserve %d\n",
proc, devind);
return EPERM;
}
2005-04-21 16:53:53 +02:00
if(pcidev[devind].pd_inuse && pcidev[devind].pd_proc != proc)
return EBUSY;
pcidev[devind].pd_inuse= 1;
pcidev[devind].pd_proc= proc;
2005-04-21 16:53:53 +02:00
for (i= 0; i<pcidev[devind].pd_bar_nr; i++)
{
if (pcidev[devind].pd_bar[i].pb_flags & PBF_INCOMPLETE)
{
printf("pci_reserve_a: BAR %d is incomplete\n", i);
continue;
}
if (pcidev[devind].pd_bar[i].pb_flags & PBF_IO)
{
ior.ior_base= pcidev[devind].pd_bar[i].pb_base;
ior.ior_limit= ior.ior_base +
pcidev[devind].pd_bar[i].pb_size-1;
2005-04-21 16:53:53 +02:00
if(debug) {
printf(
"pci_reserve_a: for proc %d, adding I/O range [0x%x..0x%x]\n",
proc, ior.ior_base, ior.ior_limit);
}
r= sys_privctl(proc, SYS_PRIV_ADD_IO, &ior);
if (r != OK)
{
printf("sys_privctl failed for proc %d: %d\n",
proc, r);
}
}
else
{
mr.mr_base= pcidev[devind].pd_bar[i].pb_base;
mr.mr_limit= mr.mr_base +
pcidev[devind].pd_bar[i].pb_size-1;
2005-04-21 16:53:53 +02:00
r= sys_privctl(proc, SYS_PRIV_ADD_MEM, &mr);
if (r != OK)
{
printf("sys_privctl failed for proc %d: %d\n",
proc, r);
}
}
}
ilr= pcidev[devind].pd_ilr;
if (ilr != PCI_ILR_UNKNOWN)
{
if(debug) printf("pci_reserve_a: adding IRQ %d\n", ilr);
r= sys_privctl(proc, SYS_PRIV_ADD_IRQ, &ilr);
if (r != OK)
{
printf("sys_privctl failed for proc %d: %d\n",
proc, r);
}
}
2005-04-21 16:53:53 +02:00
return OK;
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* _pci_release *
*===========================================================================*/
void
_pci_release(endpoint_t proc)
{
int i;
for (i= 0; i<nr_pcidev; i++)
{
if (!pcidev[i].pd_inuse)
continue;
if (pcidev[i].pd_proc != proc)
continue;
pcidev[i].pd_inuse= 0;
}
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* _pci_ids *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
int
_pci_ids(int devind, u16_t *vidp, u16_t *didp)
2005-04-21 16:53:53 +02:00
{
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
*vidp= pcidev[devind].pd_vid;
*didp= pcidev[devind].pd_did;
return OK;
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* _pci_rescan_bus *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
void
_pci_rescan_bus(u8_t busnr)
2005-04-21 16:53:53 +02:00
{
int busind;
busind= get_busind(busnr);
probe_bus(busind);
/* Allocate bus numbers for uninitialized bridges */
complete_bridges();
/* Allocate I/O and memory resources for uninitialized devices */
complete_bars();
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* _pci_slot_name *
2005-04-21 16:53:53 +02:00
*===========================================================================*/
int
_pci_slot_name(int devind, char **cpp)
2005-04-21 16:53:53 +02:00
{
static char label[]= "ddd.ddd.ddd.ddd";
char *end;
char *p;
2005-04-21 16:53:53 +02:00
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
2006-03-06 16:19:51 +01:00
p= label;
end= label+sizeof(label);
2006-03-06 16:19:51 +01:00
/* FIXME: domain nb is always 0 on 32bit system, but we should
* retrieve it properly, somehow. */
ntostr(0, &p, end);
*p++= '.';
2006-03-06 16:19:51 +01:00
ntostr(pcidev[devind].pd_busnr, &p, end);
*p++= '.';
2006-03-06 16:19:51 +01:00
ntostr(pcidev[devind].pd_dev, &p, end);
*p++= '.';
ntostr(pcidev[devind].pd_func, &p, end);
2006-03-06 16:19:51 +01:00
*cpp= label;
return OK;
}
/*===========================================================================*
* _pci_dev_name *
*===========================================================================*/
const char *
_pci_dev_name(u16_t vid, u16_t did)
{
int i;
for (i= 0; pci_device_table[i].name; i++)
{
if (pci_device_table[i].vid == vid &&
pci_device_table[i].did == did)
{
return pci_device_table[i].name;
}
}
return NULL;
}
/*===========================================================================*
* _pci_get_bar *
*===========================================================================*/
int
_pci_get_bar(int devind, int port, u32_t *base, u32_t *size,
int *ioflag)
{
int i, reg;
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
for (i= 0; i < pcidev[devind].pd_bar_nr; i++)
{
reg= PCI_BAR+4*pcidev[devind].pd_bar[i].pb_nr;
if (reg == port)
{
if (pcidev[devind].pd_bar[i].pb_flags & PBF_INCOMPLETE)
return EINVAL;
*base= pcidev[devind].pd_bar[i].pb_base;
*size= pcidev[devind].pd_bar[i].pb_size;
*ioflag=
!!(pcidev[devind].pd_bar[i].pb_flags & PBF_IO);
return OK;
}
}
return EINVAL;
}
/*===========================================================================*
* _pci_attr_r8 *
*===========================================================================*/
int
_pci_attr_r8(int devind, int port, u8_t *vp)
{
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
if (port < 0 || port > 256-1)
return EINVAL;
*vp= __pci_attr_r8(devind, port);
return OK;
}
/*===========================================================================*
* _pci_attr_r16 *
*===========================================================================*/
int
_pci_attr_r16(int devind, int port, u16_t *vp)
{
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
if (port < 0 || port > 256-2)
return EINVAL;
*vp= __pci_attr_r16(devind, port);
return OK;
}
/*===========================================================================*
* _pci_attr_r32 *
*===========================================================================*/
int
_pci_attr_r32(int devind, int port, u32_t *vp)
{
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
if (port < 0 || port > 256-4)
return EINVAL;
*vp= __pci_attr_r32(devind, port);
return OK;
}
/*===========================================================================*
* _pci_attr_w8 *
*===========================================================================*/
int
_pci_attr_w8(int devind, int port, u8_t value)
{
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
if (port < 0 || port > 256-1)
return EINVAL;
__pci_attr_w8(devind, port, value);
return OK;
}
/*===========================================================================*
* _pci_attr_w16 *
*===========================================================================*/
int
_pci_attr_w16(int devind, int port, u16_t value)
{
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
if (port < 0 || port > 256-2)
return EINVAL;
__pci_attr_w16(devind, port, value);
return OK;
}
/*===========================================================================*
* _pci_attr_w32 *
*===========================================================================*/
int
_pci_attr_w32(int devind, int port, u32_t value)
{
if (devind < 0 || devind >= nr_pcidev)
return EINVAL;
if (port < 0 || port > 256-4)
return EINVAL;
__pci_attr_w32(devind, port, value);
return OK;
}