NetBSD/sys/dev/pci/pci.c
dyoung 0d1ba3e899 During shutdown, detach devices in an orderly fashion.
Call the detach routine for every device in the device tree, starting
with the leaves and moving toward the root, expecting that each
(pseudo-)device driver will use the opportunity to gracefully commit
outstandings transactions to the underlying (pseudo-)device and to
relinquish control of the hardware to the system BIOS.

Detaching devices is not suitable for every shutdown: in an emergency,
or if the system state is inconsistent, we should resort to a fast,
simple shutdown that uses only the pmf(9) shutdown hooks and the
(deprecated) shutdownhooks.  For now, if the flag RB_NOSYNC is set in
boothowto, opt for the fast, simple shutdown.

Add a device flag, DVF_DETACH_SHUTDOWN, that indicates by its presence
that it is safe to detach a device during shutdown.  Introduce macros
CFATTACH_DECL3() and CFATTACH_DECL3_NEW() for creating autoconf
attachments with default device flags.  Add DVF_DETACH_SHUTDOWN
to configuration attachments for atabus(4), atw(4) at cardbus(4),
cardbus(4), cardslot(4), com(4) at isa(4), elanpar(4), elanpex(4),
elansc(4), gpio(4), npx(4) at isa(4), nsphyter(4), pci(4), pcib(4),
pcmcia(4), ppb(4), sip(4), wd(4), and wdc(4) at isa(4).

Add a device-detachment "reason" flag, DETACH_SHUTDOWN, that tells the
autoconf code and a device driver that the reason for detachment is
system shutdown.

Add a sysctl, kern.detachall, that tells the system to try to detach
every device at shutdown, regardless of any device's DVF_DETACH_SHUTDOWN
flag.  The default for kern.detachall is 0.  SET IT TO 1, PLEASE, TO
HELP TEST AND DEBUG DEVICE DETACHMENT AT SHUTDOWN.

This is a work in progress.  In future work, I aim to treat
pseudo-devices more thoroughly, and to gracefully tear down a stack of
(pseudo-)disk drivers and filesystems, including cgd(4), vnd(4), and
raid(4) instances at shutdown.

Also commit some changes that are not easily untangled from the rest:

(1) begin to simplify device_t locking: rename struct pmf_private to
device_lock, and incorporate device_lock into struct device.

(2) #include <sys/device.h> in sys/pmf.h in order to get some
definitions that it needs.  Stop unnecessarily #including <sys/device.h>
in sys/arch/x86/include/pic.h to keep the amd64, xen, and i386 releases
building.
2009-04-02 00:09:32 +00:00

974 lines
25 KiB
C

/* $NetBSD: pci.c,v 1.122 2009/04/02 00:09:33 dyoung Exp $ */
/*
* Copyright (c) 1995, 1996, 1997, 1998
* Christopher G. Demetriou. All rights reserved.
* Copyright (c) 1994 Charles M. Hannum. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Charles M. Hannum.
* 4. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/*
* PCI bus autoconfiguration.
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: pci.c,v 1.122 2009/04/02 00:09:33 dyoung Exp $");
#include "opt_pci.h"
#include <sys/param.h>
#include <sys/malloc.h>
#include <sys/systm.h>
#include <sys/device.h>
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
#include <dev/pci/pcidevs.h>
#include <uvm/uvm_extern.h>
#include <net/if.h>
#include "locators.h"
static bool pci_child_register(device_t);
#ifdef PCI_CONFIG_DUMP
int pci_config_dump = 1;
#else
int pci_config_dump = 0;
#endif
int pciprint(void *, const char *);
#ifdef PCI_MACHDEP_ENUMERATE_BUS
#define pci_enumerate_bus PCI_MACHDEP_ENUMERATE_BUS
#else
int pci_enumerate_bus(struct pci_softc *, const int *,
int (*)(struct pci_attach_args *), struct pci_attach_args *);
#endif
/*
* Important note about PCI-ISA bridges:
*
* Callbacks are used to configure these devices so that ISA/EISA bridges
* can attach their child busses after PCI configuration is done.
*
* This works because:
* (1) there can be at most one ISA/EISA bridge per PCI bus, and
* (2) any ISA/EISA bridges must be attached to primary PCI
* busses (i.e. bus zero).
*
* That boils down to: there can only be one of these outstanding
* at a time, it is cleared when configuring PCI bus 0 before any
* subdevices have been found, and it is run after all subdevices
* of PCI bus 0 have been found.
*
* This is needed because there are some (legacy) PCI devices which
* can show up as ISA/EISA devices as well (the prime example of which
* are VGA controllers). If you attach ISA from a PCI-ISA/EISA bridge,
* and the bridge is seen before the video board is, the board can show
* up as an ISA device, and that can (bogusly) complicate the PCI device's
* attach code, or make the PCI device not be properly attached at all.
*
* We use the generic config_defer() facility to achieve this.
*/
int
pcirescan(device_t self, const char *ifattr, const int *locators)
{
struct pci_softc *sc = device_private(self);
KASSERT(ifattr && !strcmp(ifattr, "pci"));
KASSERT(locators);
pci_enumerate_bus(sc, locators, NULL, NULL);
return 0;
}
int
pcimatch(device_t parent, cfdata_t cf, void *aux)
{
struct pcibus_attach_args *pba = aux;
/* Check the locators */
if (cf->cf_loc[PCIBUSCF_BUS] != PCIBUSCF_BUS_DEFAULT &&
cf->cf_loc[PCIBUSCF_BUS] != pba->pba_bus)
return (0);
/* sanity */
if (pba->pba_bus < 0 || pba->pba_bus > 255)
return (0);
/*
* XXX check other (hardware?) indicators
*/
return (1);
}
void
pciattach(device_t parent, device_t self, void *aux)
{
struct pcibus_attach_args *pba = aux;
struct pci_softc *sc = device_private(self);
int io_enabled, mem_enabled, mrl_enabled, mrm_enabled, mwi_enabled;
const char *sep = "";
static const int wildcard[PCICF_NLOCS] = {
PCICF_DEV_DEFAULT, PCICF_FUNCTION_DEFAULT
};
sc->sc_dev = self;
pci_attach_hook(parent, self, pba);
aprint_naive("\n");
aprint_normal("\n");
io_enabled = (pba->pba_flags & PCI_FLAGS_IO_ENABLED);
mem_enabled = (pba->pba_flags & PCI_FLAGS_MEM_ENABLED);
mrl_enabled = (pba->pba_flags & PCI_FLAGS_MRL_OKAY);
mrm_enabled = (pba->pba_flags & PCI_FLAGS_MRM_OKAY);
mwi_enabled = (pba->pba_flags & PCI_FLAGS_MWI_OKAY);
if (io_enabled == 0 && mem_enabled == 0) {
aprint_error_dev(self, "no spaces enabled!\n");
goto fail;
}
#define PRINT(str) \
do { \
aprint_verbose("%s%s", sep, str); \
sep = ", "; \
} while (/*CONSTCOND*/0)
aprint_verbose_dev(self, "");
if (io_enabled)
PRINT("i/o space");
if (mem_enabled)
PRINT("memory space");
aprint_verbose(" enabled");
if (mrl_enabled || mrm_enabled || mwi_enabled) {
if (mrl_enabled)
PRINT("rd/line");
if (mrm_enabled)
PRINT("rd/mult");
if (mwi_enabled)
PRINT("wr/inv");
aprint_verbose(" ok");
}
aprint_verbose("\n");
#undef PRINT
sc->sc_iot = pba->pba_iot;
sc->sc_memt = pba->pba_memt;
sc->sc_dmat = pba->pba_dmat;
sc->sc_dmat64 = pba->pba_dmat64;
sc->sc_pc = pba->pba_pc;
sc->sc_bus = pba->pba_bus;
sc->sc_bridgetag = pba->pba_bridgetag;
sc->sc_maxndevs = pci_bus_maxdevs(pba->pba_pc, pba->pba_bus);
sc->sc_intrswiz = pba->pba_intrswiz;
sc->sc_intrtag = pba->pba_intrtag;
sc->sc_flags = pba->pba_flags;
device_pmf_driver_set_child_register(sc->sc_dev, pci_child_register);
pcirescan(sc->sc_dev, "pci", wildcard);
fail:
if (!pmf_device_register(self, NULL, NULL))
aprint_error_dev(self, "couldn't establish power handler\n");
}
int
pcidetach(device_t self, int flags)
{
int rc;
if ((rc = config_detach_children(self, flags)) != 0)
return rc;
pmf_device_deregister(self);
return 0;
}
int
pciprint(void *aux, const char *pnp)
{
struct pci_attach_args *pa = aux;
char devinfo[256];
const struct pci_quirkdata *qd;
if (pnp) {
pci_devinfo(pa->pa_id, pa->pa_class, 1, devinfo, sizeof(devinfo));
aprint_normal("%s at %s", devinfo, pnp);
}
aprint_normal(" dev %d function %d", pa->pa_device, pa->pa_function);
if (pci_config_dump) {
printf(": ");
pci_conf_print(pa->pa_pc, pa->pa_tag, NULL);
if (!pnp)
pci_devinfo(pa->pa_id, pa->pa_class, 1, devinfo, sizeof(devinfo));
printf("%s at %s", devinfo, pnp ? pnp : "?");
printf(" dev %d function %d (", pa->pa_device, pa->pa_function);
#ifdef __i386__
printf("tag %#lx, intrtag %#lx, intrswiz %#lx, intrpin %#lx",
*(long *)&pa->pa_tag, *(long *)&pa->pa_intrtag,
(long)pa->pa_intrswiz, (long)pa->pa_intrpin);
#else
printf("intrswiz %#lx, intrpin %#lx",
(long)pa->pa_intrswiz, (long)pa->pa_intrpin);
#endif
printf(", i/o %s, mem %s,",
pa->pa_flags & PCI_FLAGS_IO_ENABLED ? "on" : "off",
pa->pa_flags & PCI_FLAGS_MEM_ENABLED ? "on" : "off");
qd = pci_lookup_quirkdata(PCI_VENDOR(pa->pa_id),
PCI_PRODUCT(pa->pa_id));
if (qd == NULL) {
printf(" no quirks");
} else {
snprintb(devinfo, sizeof (devinfo),
"\002\001multifn\002singlefn\003skipfunc0"
"\004skipfunc1\005skipfunc2\006skipfunc3"
"\007skipfunc4\010skipfunc5\011skipfunc6"
"\012skipfunc7", qd->quirks);
printf(" quirks %s", devinfo);
}
printf(")");
}
return (UNCONF);
}
int
pci_probe_device(struct pci_softc *sc, pcitag_t tag,
int (*match)(struct pci_attach_args *), struct pci_attach_args *pap)
{
pci_chipset_tag_t pc = sc->sc_pc;
struct pci_attach_args pa;
pcireg_t id, csr, class, intr, bhlcr;
int ret, pin, bus, device, function;
int locs[PCICF_NLOCS];
device_t subdev;
pci_decompose_tag(pc, tag, &bus, &device, &function);
/* a driver already attached? */
if (sc->PCI_SC_DEVICESC(device, function).c_dev != NULL && !match)
return (0);
bhlcr = pci_conf_read(pc, tag, PCI_BHLC_REG);
if (PCI_HDRTYPE_TYPE(bhlcr) > 2)
return (0);
id = pci_conf_read(pc, tag, PCI_ID_REG);
csr = pci_conf_read(pc, tag, PCI_COMMAND_STATUS_REG);
class = pci_conf_read(pc, tag, PCI_CLASS_REG);
/* Invalid vendor ID value? */
if (PCI_VENDOR(id) == PCI_VENDOR_INVALID)
return (0);
/* XXX Not invalid, but we've done this ~forever. */
if (PCI_VENDOR(id) == 0)
return (0);
pa.pa_iot = sc->sc_iot;
pa.pa_memt = sc->sc_memt;
pa.pa_dmat = sc->sc_dmat;
pa.pa_dmat64 = sc->sc_dmat64;
pa.pa_pc = pc;
pa.pa_bus = bus;
pa.pa_device = device;
pa.pa_function = function;
pa.pa_tag = tag;
pa.pa_id = id;
pa.pa_class = class;
/*
* Set up memory, I/O enable, and PCI command flags
* as appropriate.
*/
pa.pa_flags = sc->sc_flags;
if ((csr & PCI_COMMAND_IO_ENABLE) == 0)
pa.pa_flags &= ~PCI_FLAGS_IO_ENABLED;
if ((csr & PCI_COMMAND_MEM_ENABLE) == 0)
pa.pa_flags &= ~PCI_FLAGS_MEM_ENABLED;
/*
* If the cache line size is not configured, then
* clear the MRL/MRM/MWI command-ok flags.
*/
if (PCI_CACHELINE(bhlcr) == 0)
pa.pa_flags &= ~(PCI_FLAGS_MRL_OKAY|
PCI_FLAGS_MRM_OKAY|PCI_FLAGS_MWI_OKAY);
if (sc->sc_bridgetag == NULL) {
pa.pa_intrswiz = 0;
pa.pa_intrtag = tag;
} else {
pa.pa_intrswiz = sc->sc_intrswiz + device;
pa.pa_intrtag = sc->sc_intrtag;
}
intr = pci_conf_read(pc, tag, PCI_INTERRUPT_REG);
pin = PCI_INTERRUPT_PIN(intr);
pa.pa_rawintrpin = pin;
if (pin == PCI_INTERRUPT_PIN_NONE) {
/* no interrupt */
pa.pa_intrpin = 0;
} else {
/*
* swizzle it based on the number of busses we're
* behind and our device number.
*/
pa.pa_intrpin = /* XXX */
((pin + pa.pa_intrswiz - 1) % 4) + 1;
}
pa.pa_intrline = PCI_INTERRUPT_LINE(intr);
if (match != NULL) {
ret = (*match)(&pa);
if (ret != 0 && pap != NULL)
*pap = pa;
} else {
struct pci_child *c;
locs[PCICF_DEV] = device;
locs[PCICF_FUNCTION] = function;
subdev = config_found_sm_loc(sc->sc_dev, "pci", locs, &pa,
pciprint, config_stdsubmatch);
c = &sc->PCI_SC_DEVICESC(device, function);
c->c_dev = subdev;
pci_conf_capture(pc, tag, &c->c_conf);
if (pci_get_powerstate(pc, tag, &c->c_powerstate) == 0)
c->c_psok = true;
else
c->c_psok = false;
ret = (subdev != NULL);
}
return (ret);
}
void
pcidevdetached(device_t self, device_t child)
{
struct pci_softc *sc = device_private(self);
int d, f;
pcitag_t tag;
struct pci_child *c;
d = device_locator(child, PCICF_DEV);
f = device_locator(child, PCICF_FUNCTION);
c = &sc->PCI_SC_DEVICESC(d, f);
KASSERT(c->c_dev == child);
tag = pci_make_tag(sc->sc_pc, sc->sc_bus, d, f);
if (c->c_psok)
pci_set_powerstate(sc->sc_pc, tag, c->c_powerstate);
pci_conf_restore(sc->sc_pc, tag, &c->c_conf);
c->c_dev = NULL;
}
CFATTACH_DECL3_NEW(pci, sizeof(struct pci_softc),
pcimatch, pciattach, pcidetach, NULL, pcirescan, pcidevdetached,
DVF_DETACH_SHUTDOWN);
int
pci_get_capability(pci_chipset_tag_t pc, pcitag_t tag, int capid,
int *offset, pcireg_t *value)
{
pcireg_t reg;
unsigned int ofs;
reg = pci_conf_read(pc, tag, PCI_COMMAND_STATUS_REG);
if (!(reg & PCI_STATUS_CAPLIST_SUPPORT))
return (0);
/* Determine the Capability List Pointer register to start with. */
reg = pci_conf_read(pc, tag, PCI_BHLC_REG);
switch (PCI_HDRTYPE_TYPE(reg)) {
case 0: /* standard device header */
case 1: /* PCI-PCI bridge header */
ofs = PCI_CAPLISTPTR_REG;
break;
case 2: /* PCI-CardBus Bridge header */
ofs = PCI_CARDBUS_CAPLISTPTR_REG;
break;
default:
return (0);
}
ofs = PCI_CAPLIST_PTR(pci_conf_read(pc, tag, ofs));
while (ofs != 0) {
if ((ofs & 3) || (ofs < 0x40)) {
int bus, device, function;
pci_decompose_tag(pc, tag, &bus, &device, &function);
printf("Skipping broken PCI header on %d:%d:%d\n",
bus, device, function);
break;
}
reg = pci_conf_read(pc, tag, ofs);
if (PCI_CAPLIST_CAP(reg) == capid) {
if (offset)
*offset = ofs;
if (value)
*value = reg;
return (1);
}
ofs = PCI_CAPLIST_NEXT(reg);
}
return (0);
}
int
pci_find_device(struct pci_attach_args *pa,
int (*match)(struct pci_attach_args *))
{
extern struct cfdriver pci_cd;
device_t pcidev;
int i;
static const int wildcard[2] = {
PCICF_DEV_DEFAULT,
PCICF_FUNCTION_DEFAULT
};
for (i = 0; i < pci_cd.cd_ndevs; i++) {
pcidev = device_lookup(&pci_cd, i);
if (pcidev != NULL &&
pci_enumerate_bus(device_private(pcidev), wildcard,
match, pa) != 0)
return (1);
}
return (0);
}
#ifndef PCI_MACHDEP_ENUMERATE_BUS
/*
* Generic PCI bus enumeration routine. Used unless machine-dependent
* code needs to provide something else.
*/
int
pci_enumerate_bus(struct pci_softc *sc, const int *locators,
int (*match)(struct pci_attach_args *), struct pci_attach_args *pap)
{
pci_chipset_tag_t pc = sc->sc_pc;
int device, function, nfunctions, ret;
const struct pci_quirkdata *qd;
pcireg_t id, bhlcr;
pcitag_t tag;
#ifdef __PCI_BUS_DEVORDER
char devs[32];
int i;
#endif
#ifdef __PCI_BUS_DEVORDER
pci_bus_devorder(sc->sc_pc, sc->sc_bus, devs);
for (i = 0; (device = devs[i]) < 32 && device >= 0; i++)
#else
for (device = 0; device < sc->sc_maxndevs; device++)
#endif
{
if ((locators[PCICF_DEV] != PCICF_DEV_DEFAULT) &&
(locators[PCICF_DEV] != device))
continue;
tag = pci_make_tag(pc, sc->sc_bus, device, 0);
bhlcr = pci_conf_read(pc, tag, PCI_BHLC_REG);
if (PCI_HDRTYPE_TYPE(bhlcr) > 2)
continue;
id = pci_conf_read(pc, tag, PCI_ID_REG);
/* Invalid vendor ID value? */
if (PCI_VENDOR(id) == PCI_VENDOR_INVALID)
continue;
/* XXX Not invalid, but we've done this ~forever. */
if (PCI_VENDOR(id) == 0)
continue;
qd = pci_lookup_quirkdata(PCI_VENDOR(id), PCI_PRODUCT(id));
if (qd != NULL &&
(qd->quirks & PCI_QUIRK_MULTIFUNCTION) != 0)
nfunctions = 8;
else if (qd != NULL &&
(qd->quirks & PCI_QUIRK_MONOFUNCTION) != 0)
nfunctions = 1;
else
nfunctions = PCI_HDRTYPE_MULTIFN(bhlcr) ? 8 : 1;
for (function = 0; function < nfunctions; function++) {
if ((locators[PCICF_FUNCTION] != PCICF_FUNCTION_DEFAULT)
&& (locators[PCICF_FUNCTION] != function))
continue;
if (qd != NULL &&
(qd->quirks & PCI_QUIRK_SKIP_FUNC(function)) != 0)
continue;
tag = pci_make_tag(pc, sc->sc_bus, device, function);
ret = pci_probe_device(sc, tag, match, pap);
if (match != NULL && ret != 0)
return (ret);
}
}
return (0);
}
#endif /* PCI_MACHDEP_ENUMERATE_BUS */
/*
* Vital Product Data (PCI 2.2)
*/
int
pci_vpd_read(pci_chipset_tag_t pc, pcitag_t tag, int offset, int count,
pcireg_t *data)
{
uint32_t reg;
int ofs, i, j;
KASSERT(data != NULL);
KASSERT((offset + count) < 0x7fff);
if (pci_get_capability(pc, tag, PCI_CAP_VPD, &ofs, &reg) == 0)
return (1);
for (i = 0; i < count; offset += sizeof(*data), i++) {
reg &= 0x0000ffff;
reg &= ~PCI_VPD_OPFLAG;
reg |= PCI_VPD_ADDRESS(offset);
pci_conf_write(pc, tag, ofs, reg);
/*
* PCI 2.2 does not specify how long we should poll
* for completion nor whether the operation can fail.
*/
j = 0;
do {
if (j++ == 20)
return (1);
delay(4);
reg = pci_conf_read(pc, tag, ofs);
} while ((reg & PCI_VPD_OPFLAG) == 0);
data[i] = pci_conf_read(pc, tag, PCI_VPD_DATAREG(ofs));
}
return (0);
}
int
pci_vpd_write(pci_chipset_tag_t pc, pcitag_t tag, int offset, int count,
pcireg_t *data)
{
pcireg_t reg;
int ofs, i, j;
KASSERT(data != NULL);
KASSERT((offset + count) < 0x7fff);
if (pci_get_capability(pc, tag, PCI_CAP_VPD, &ofs, &reg) == 0)
return (1);
for (i = 0; i < count; offset += sizeof(*data), i++) {
pci_conf_write(pc, tag, PCI_VPD_DATAREG(ofs), data[i]);
reg &= 0x0000ffff;
reg |= PCI_VPD_OPFLAG;
reg |= PCI_VPD_ADDRESS(offset);
pci_conf_write(pc, tag, ofs, reg);
/*
* PCI 2.2 does not specify how long we should poll
* for completion nor whether the operation can fail.
*/
j = 0;
do {
if (j++ == 20)
return (1);
delay(1);
reg = pci_conf_read(pc, tag, ofs);
} while (reg & PCI_VPD_OPFLAG);
}
return (0);
}
int
pci_dma64_available(struct pci_attach_args *pa)
{
#ifdef _PCI_HAVE_DMA64
if (BUS_DMA_TAG_VALID(pa->pa_dmat64))
return 1;
#endif
return 0;
}
void
pci_conf_capture(pci_chipset_tag_t pc, pcitag_t tag,
struct pci_conf_state *pcs)
{
int off;
for (off = 0; off < 16; off++)
pcs->reg[off] = pci_conf_read(pc, tag, (off * 4));
return;
}
void
pci_conf_restore(pci_chipset_tag_t pc, pcitag_t tag,
struct pci_conf_state *pcs)
{
int off;
pcireg_t val;
for (off = 15; off >= 0; off--) {
val = pci_conf_read(pc, tag, (off * 4));
if (val != pcs->reg[off])
pci_conf_write(pc, tag, (off * 4), pcs->reg[off]);
}
return;
}
/*
* Power Management Capability (Rev 2.2)
*/
static int
pci_get_powerstate_int(pci_chipset_tag_t pc, pcitag_t tag , pcireg_t *state,
int offset)
{
pcireg_t value, now;
value = pci_conf_read(pc, tag, offset + PCI_PMCSR);
now = value & PCI_PMCSR_STATE_MASK;
switch (now) {
case PCI_PMCSR_STATE_D0:
case PCI_PMCSR_STATE_D1:
case PCI_PMCSR_STATE_D2:
case PCI_PMCSR_STATE_D3:
*state = now;
return 0;
default:
return EINVAL;
}
}
int
pci_get_powerstate(pci_chipset_tag_t pc, pcitag_t tag , pcireg_t *state)
{
int offset;
pcireg_t value;
if (!pci_get_capability(pc, tag, PCI_CAP_PWRMGMT, &offset, &value))
return EOPNOTSUPP;
return pci_get_powerstate_int(pc, tag, state, offset);
}
static int
pci_set_powerstate_int(pci_chipset_tag_t pc, pcitag_t tag, pcireg_t state,
int offset, pcireg_t cap_reg)
{
pcireg_t value, cap, now;
cap = cap_reg >> PCI_PMCR_SHIFT;
value = pci_conf_read(pc, tag, offset + PCI_PMCSR);
now = value & PCI_PMCSR_STATE_MASK;
value &= ~PCI_PMCSR_STATE_MASK;
if (now == state)
return 0;
switch (state) {
case PCI_PMCSR_STATE_D0:
break;
case PCI_PMCSR_STATE_D1:
if (now == PCI_PMCSR_STATE_D2 || now == PCI_PMCSR_STATE_D3) {
printf("invalid transition from %d to D1\n", (int)now);
return EINVAL;
}
if (!(cap & PCI_PMCR_D1SUPP)) {
printf("D1 not supported\n");
return EOPNOTSUPP;
}
break;
case PCI_PMCSR_STATE_D2:
if (now == PCI_PMCSR_STATE_D3) {
printf("invalid transition from %d to D2\n", (int)now);
return EINVAL;
}
if (!(cap & PCI_PMCR_D2SUPP)) {
printf("D2 not supported\n");
return EOPNOTSUPP;
}
break;
case PCI_PMCSR_STATE_D3:
break;
default:
return EINVAL;
}
value |= state;
pci_conf_write(pc, tag, offset + PCI_PMCSR, value);
/* delay according to pcipm1.2, ch. 5.6.1 */
if (state == PCI_PMCSR_STATE_D3 || now == PCI_PMCSR_STATE_D3)
DELAY(10000);
else if (state == PCI_PMCSR_STATE_D2 || now == PCI_PMCSR_STATE_D2)
DELAY(200);
return 0;
}
int
pci_set_powerstate(pci_chipset_tag_t pc, pcitag_t tag, pcireg_t state)
{
int offset;
pcireg_t value;
if (!pci_get_capability(pc, tag, PCI_CAP_PWRMGMT, &offset, &value)) {
printf("pci_set_powerstate not supported\n");
return EOPNOTSUPP;
}
return pci_set_powerstate_int(pc, tag, state, offset, value);
}
int
pci_activate(pci_chipset_tag_t pc, pcitag_t tag, device_t dev,
int (*wakefun)(pci_chipset_tag_t, pcitag_t, device_t, pcireg_t))
{
pcireg_t pmode;
int error;
if ((error = pci_get_powerstate(pc, tag, &pmode)))
return error;
switch (pmode) {
case PCI_PMCSR_STATE_D0:
break;
case PCI_PMCSR_STATE_D3:
if (wakefun == NULL) {
/*
* The card has lost all configuration data in
* this state, so punt.
*/
aprint_error_dev(dev,
"unable to wake up from power state D3\n");
return EOPNOTSUPP;
}
/*FALLTHROUGH*/
default:
if (wakefun) {
error = (*wakefun)(pc, tag, dev, pmode);
if (error)
return error;
}
aprint_normal_dev(dev, "waking up from power state D%d\n",
pmode);
if ((error = pci_set_powerstate(pc, tag, PCI_PMCSR_STATE_D0)))
return error;
}
return 0;
}
int
pci_activate_null(pci_chipset_tag_t pc, pcitag_t tag,
device_t dev, pcireg_t state)
{
return 0;
}
/* I have disabled this code for now. --dyoung
*
* Insofar as I understand what the PCI retry timeout is [1],
* I see no justification for any driver to disable when it
* attaches/resumes a device.
*
* A PCI bus bridge may tell a bus master to retry its transaction
* at a later time if the resources to complete the transaction
* are not immediately available. Taking a guess, PCI bus masters
* that implement a PCI retry timeout register count down from the
* retry timeout to 0 while it retries a delayed PCI transaction.
* When it reaches 0, it stops retrying. A PCI master is *never*
* supposed to stop retrying a delayed transaction, though.
*
* Incidentally, I initially suspected that writing 0 to the register
* would not disable *retries*, but would disable the timeout.
* That is, any device whose retry timeout was set to 0 would
* *never* timeout. However, I found out, by using PCI debug
* facilities on the AMD Elan SC520, that if I write 0 to the retry
* timeout register on an ath(4) MiniPCI card, the card really does
* not retry transactions.
*
* Some uses of this register have mentioned "interference" with
* a CPU's "C3 sleep state." It seems to me that if a bus master
* is properly put to sleep, it will neither initiate new transactions,
* nor retry delayed transactions, so disabling retries should not
* be necessary.
*
* [1] The timeout does not appear to be documented in any PCI
* standard, and we have no documentation of it for the devices by
* Atheros, and others, that supposedly implement it.
*/
void
pci_disable_retry(pci_chipset_tag_t pc, pcitag_t tag)
{
#if 0
pcireg_t retry;
/*
* Disable retry timeout to keep PCI Tx retries from
* interfering with ACPI C3 CPU state.
*/
retry = pci_conf_read(pc, tag, PCI_RETRY_TIMEOUT_REG);
retry &= ~PCI_RETRY_TIMEOUT_REG_MASK;
pci_conf_write(pc, tag, PCI_RETRY_TIMEOUT_REG, retry);
#endif
}
struct pci_child_power {
struct pci_conf_state p_pciconf;
pci_chipset_tag_t p_pc;
pcitag_t p_tag;
bool p_has_pm;
int p_pm_offset;
pcireg_t p_pm_cap;
pcireg_t p_class;
};
static bool
pci_child_suspend(device_t dv PMF_FN_ARGS)
{
struct pci_child_power *priv = device_pmf_bus_private(dv);
pcireg_t ocsr, csr;
pci_conf_capture(priv->p_pc, priv->p_tag, &priv->p_pciconf);
if (!priv->p_has_pm)
return true; /* ??? hopefully handled by ACPI */
if (PCI_CLASS(priv->p_class) == PCI_CLASS_DISPLAY)
return true; /* XXX */
/* disable decoding and busmastering, see pcipm1.2 ch. 8.2.1 */
ocsr = pci_conf_read(priv->p_pc, priv->p_tag, PCI_COMMAND_STATUS_REG);
csr = ocsr & ~(PCI_COMMAND_IO_ENABLE | PCI_COMMAND_MEM_ENABLE
| PCI_COMMAND_MASTER_ENABLE);
pci_conf_write(priv->p_pc, priv->p_tag, PCI_COMMAND_STATUS_REG, csr);
if (pci_set_powerstate_int(priv->p_pc, priv->p_tag,
PCI_PMCSR_STATE_D3, priv->p_pm_offset, priv->p_pm_cap)) {
pci_conf_write(priv->p_pc, priv->p_tag,
PCI_COMMAND_STATUS_REG, ocsr);
aprint_error_dev(dv, "unsupported state, continuing.\n");
return false;
}
return true;
}
static bool
pci_child_resume(device_t dv PMF_FN_ARGS)
{
struct pci_child_power *priv = device_pmf_bus_private(dv);
if (priv->p_has_pm &&
pci_set_powerstate_int(priv->p_pc, priv->p_tag,
PCI_PMCSR_STATE_D0, priv->p_pm_offset, priv->p_pm_cap)) {
aprint_error_dev(dv, "unsupported state, continuing.\n");
return false;
}
pci_conf_restore(priv->p_pc, priv->p_tag, &priv->p_pciconf);
return true;
}
static bool
pci_child_shutdown(device_t dv, int how)
{
struct pci_child_power *priv = device_pmf_bus_private(dv);
pcireg_t csr;
/* disable busmastering */
csr = pci_conf_read(priv->p_pc, priv->p_tag, PCI_COMMAND_STATUS_REG);
csr &= ~PCI_COMMAND_MASTER_ENABLE;
pci_conf_write(priv->p_pc, priv->p_tag, PCI_COMMAND_STATUS_REG, csr);
return true;
}
static void
pci_child_deregister(device_t dv)
{
struct pci_child_power *priv = device_pmf_bus_private(dv);
free(priv, M_DEVBUF);
}
static bool
pci_child_register(device_t child)
{
device_t self = device_parent(child);
struct pci_softc *sc = device_private(self);
struct pci_child_power *priv;
int device, function, off;
pcireg_t reg;
priv = malloc(sizeof(*priv), M_DEVBUF, M_WAITOK);
device = device_locator(child, PCICF_DEV);
function = device_locator(child, PCICF_FUNCTION);
priv->p_pc = sc->sc_pc;
priv->p_tag = pci_make_tag(priv->p_pc, sc->sc_bus, device,
function);
priv->p_class = pci_conf_read(priv->p_pc, priv->p_tag, PCI_CLASS_REG);
if (pci_get_capability(priv->p_pc, priv->p_tag,
PCI_CAP_PWRMGMT, &off, &reg)) {
priv->p_has_pm = true;
priv->p_pm_offset = off;
priv->p_pm_cap = reg;
} else {
priv->p_has_pm = false;
priv->p_pm_offset = -1;
}
device_pmf_bus_register(child, priv, pci_child_suspend,
pci_child_resume, pci_child_shutdown, pci_child_deregister);
return true;
}