fcpci.c revision 89b43686db1fe9681d80a7cf5662730cb9378cae
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
* Copyright (c) 2011 Bayard G. Bell. All rights reserved.
*/
/*
* fcpci.c: Framework PCI fcode ops
*/
#include <sys/types.h>
#include <sys/kmem.h>
#include <sys/systm.h>
#include <sys/pci.h>
#include <sys/ddi.h>
#include <sys/sunddi.h>
#include <sys/sunndi.h>
#include <sys/ddidmareq.h>
#include <sys/pci.h>
#include <sys/modctl.h>
#include <sys/ndi_impldefs.h>
#include <sys/fcode.h>
#include <sys/promif.h>
#include <sys/promimpl.h>
#include <sys/ddi_implfuncs.h>
#define PCI_NPT_bits (PCI_RELOCAT_B | PCI_PREFETCH_B | PCI_ALIAS_B)
#define PCI_BDF_bits (PCI_REG_BDFR_M & ~PCI_REG_REG_M)
#define PCICFG_CONF_INDIRECT_MAP 1
static int pfc_map_in(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_map_out(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_dma_map_in(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_dma_map_out(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_dma_sync(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_dma_cleanup(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_register_fetch(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_register_store(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_config_fetch(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_config_store(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_probe_address(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_probe_space(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_config_child(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_get_fcode_size(dev_info_t *, fco_handle_t, fc_ci_t *);
static int pfc_get_fcode(dev_info_t *, fco_handle_t, fc_ci_t *);
int prom_get_fcode_size(char *);
int prom_get_fcode(char *, char *);
int pfc_update_assigned_prop(dev_info_t *, pci_regspec_t *);
int pfc_remove_assigned_prop(dev_info_t *, pci_regspec_t *);
int pci_alloc_resource(dev_info_t *, pci_regspec_t);
int pci_free_resource(dev_info_t *, pci_regspec_t);
int pci_alloc_mem_chunk(dev_info_t *, uint64_t, uint64_t *, uint64_t *);
int pci_alloc_io_chunk(dev_info_t *, uint64_t, uint64_t *, uint64_t *);
static int fcpci_indirect_map(dev_info_t *);
int fcpci_unloadable;
static ddi_dma_attr_t fcpci_dma_attr = {
DMA_ATTR_V0, /* version number */
0x0, /* lowest usable address */
0xFFFFFFFFull, /* high DMA address range */
0xFFFFFFFFull, /* DMA counter register */
1, /* DMA address alignment */
1, /* DMA burstsizes */
1, /* min effective DMA size */
0xFFFFFFFFull, /* max DMA xfer size */
0xFFFFFFFFull, /* segment boundary */
1, /* s/g list length */
1, /* granularity of device */
0 /* DMA transfer flags */
};
#define HIADDR(n) ((uint32_t)(((uint64_t)(n) & 0xFFFFFFFF00000000)>> 32))
#define LOADDR(n)((uint32_t)((uint64_t)(n) & 0x00000000FFFFFFFF))
#define LADDR(lo, hi) (((uint64_t)(hi) << 32) | (uint32_t)(lo))
#define PCI_4GIG_LIMIT 0xFFFFFFFFUL
#define PCI_MEMGRAN 0x100000
#define PCI_IOGRAN 0x1000
/*
* Module linkage information for the kernel.
*/
static struct modlmisc modlmisc = {
&mod_miscops, "FCode pci bus functions"
};
static struct modlinkage modlinkage = {
MODREV_1, (void *)&modlmisc, NULL
};
int
_init(void)
{
return (mod_install(&modlinkage));
}
int
_fini(void)
{
if (fcpci_unloadable)
return (mod_remove(&modlinkage));
return (EBUSY);
}
int
_info(struct modinfo *modinfop)
{
return (mod_info(&modlinkage, modinfop));
}
struct pfc_ops_v {
char *svc_name;
fc_ops_t *f;
};
static struct pfc_ops_v pov[] = {
{ "map-in", pfc_map_in},
{ "map-out", pfc_map_out},
{ "dma-map-in", pfc_dma_map_in},
{ "dma-map-out", pfc_dma_map_out},
{ "dma-sync", pfc_dma_sync},
{ "rx@", pfc_register_fetch},
{ "rl@", pfc_register_fetch},
{ "rw@", pfc_register_fetch},
{ "rb@", pfc_register_fetch},
{ "rx!", pfc_register_store},
{ "rl!", pfc_register_store},
{ "rw!", pfc_register_store},
{ "rb!", pfc_register_store},
{ "config-l@", pfc_config_fetch},
{ "config-w@", pfc_config_fetch},
{ "config-b@", pfc_config_fetch},
{ "config-l!", pfc_config_store},
{ "config-w!", pfc_config_store},
{ "config-b!", pfc_config_store},
{ FC_PROBE_ADDRESS, pfc_probe_address},
{ FC_PROBE_SPACE, pfc_probe_space},
{ FC_SVC_EXIT, pfc_dma_cleanup},
{ FC_CONFIG_CHILD, pfc_config_child},
{ FC_GET_FCODE_SIZE, pfc_get_fcode_size},
{ FC_GET_FCODE, pfc_get_fcode},
{ NULL, NULL}
};
static struct pfc_ops_v shared_pov[] = {
{ FC_SVC_EXIT, pfc_dma_cleanup},
{ NULL, NULL}
};
int pci_map_phys(dev_info_t *, pci_regspec_t *,
caddr_t *, ddi_device_acc_attr_t *, ddi_acc_handle_t *);
void pci_unmap_phys(ddi_acc_handle_t *, pci_regspec_t *);
fco_handle_t
pci_fc_ops_alloc_handle(dev_info_t *ap, dev_info_t *child,
void *fcode, size_t fcode_size, char *unit_address,
struct pci_ops_bus_args *up)
{
fco_handle_t rp;
struct pci_ops_bus_args *bp = NULL;
phandle_t h;
rp = kmem_zalloc(sizeof (struct fc_resource_list), KM_SLEEP);
rp->next_handle = fc_ops_alloc_handle(ap, child, fcode, fcode_size,
unit_address, NULL);
rp->ap = ap;
rp->child = child;
rp->fcode = fcode;
rp->fcode_size = fcode_size;
if (unit_address) {
char *buf;
buf = kmem_zalloc(strlen(unit_address) + 1, KM_SLEEP);
(void) strcpy(buf, unit_address);
rp->unit_address = buf;
}
bp = kmem_zalloc(sizeof (struct pci_ops_bus_args), KM_SLEEP);
*bp = *up;
rp->bus_args = bp;
/*
* Add the child's nodeid to our table...
*/
h = ddi_get_nodeid(rp->child);
fc_add_dip_to_phandle(fc_handle_to_phandle_head(rp), rp->child, h);
return (rp);
}
void
pci_fc_ops_free_handle(fco_handle_t rp)
{
struct pci_ops_bus_args *bp;
struct fc_resource *ip, *np;
ASSERT(rp);
if (rp->next_handle)
fc_ops_free_handle(rp->next_handle);
if (rp->unit_address)
kmem_free(rp->unit_address, strlen(rp->unit_address) + 1);
if ((bp = rp->bus_args) != NULL)
kmem_free(bp, sizeof (struct pci_ops_bus_args));
/*
* Release all the resources from the resource list
* XXX: We don't handle 'unknown' types, but we don't create them.
*/
for (ip = rp->head; ip != NULL; ip = np) {
np = ip->next;
switch (ip->type) {
case RT_MAP:
FC_DEBUG1(1, CE_CONT, "pci_fc_ops_free: "
"pci_unmap_phys(%p)\n", ip->fc_map_handle);
pci_unmap_phys(&ip->fc_map_handle, ip->fc_regspec);
kmem_free(ip->fc_regspec, sizeof (pci_regspec_t));
break;
case RT_DMA:
/* DMA has to be freed up at exit time */
cmn_err(CE_CONT, "pfc_fc_ops_free: DMA seen!\n");
break;
default:
cmn_err(CE_CONT, "pci_fc_ops_free: "
"unknown resource type %d\n", ip->type);
break;
}
fc_rem_resource(rp, ip);
kmem_free(ip, sizeof (struct fc_resource));
}
kmem_free(rp, sizeof (struct fc_resource_list));
}
int
pci_fc_ops(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
struct pfc_ops_v *pv;
char *name = fc_cell2ptr(cp->svc_name);
ASSERT(rp);
/*
* First try the generic fc_ops. If the ops is a shared op,
* also call our local function.
*/
if (fc_ops(ap, rp->next_handle, cp) == 0) {
for (pv = shared_pov; pv->svc_name != NULL; ++pv)
if (strcmp(pv->svc_name, name) == 0)
return (pv->f(ap, rp, cp));
return (0);
}
for (pv = pov; pv->svc_name != NULL; ++pv)
if (strcmp(pv->svc_name, name) == 0)
return (pv->f(ap, rp, cp));
FC_DEBUG1(9, CE_CONT, "pci_fc_ops: <%s> not serviced\n", name);
return (-1);
}
/*
* Create a dma mapping for a given user address.
*/
static int
pfc_dma_map_in(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
ddi_dma_handle_t h;
int error;
caddr_t virt;
size_t len;
uint_t flags = DDI_DMA_RDWR | DDI_DMA_CONSISTENT;
struct fc_resource *ip;
ddi_dma_cookie_t c;
struct buf *bp;
uint_t ccnt;
if (fc_cell2int(cp->nargs) != 3)
return (fc_syntax_error(cp, "nargs must be 3"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
/*
* XXX: It's not clear what we should do with a non-cacheable request
*/
virt = fc_cell2ptr(fc_arg(cp, 2));
len = fc_cell2size(fc_arg(cp, 1));
#ifdef notdef
cacheable = fc_cell2int(fc_arg(cp, 0)); /* XXX: do what? */
#endif
FC_DEBUG2(6, CE_CONT, "pcf_dma_map_in: virt %p, len %d\n", virt, len);
/*
* Set up the address space for physio from userland
*/
error = fc_physio_setup(&bp, virt, len);
if (error) {
FC_DEBUG3(1, CE_CONT, "pfc_dma_map_in: fc_physio_setup failed "
"error: %d virt: %p len %d\n", error, virt, len);
return (fc_priv_error(cp, "fc_physio_setup failed"));
}
FC_DEBUG1(9, CE_CONT, "pfc_dma_map_in: dma_map_in; bp = %p\n", bp);
error = fc_ddi_dma_alloc_handle(ap, &fcpci_dma_attr, DDI_DMA_SLEEP,
NULL, &h);
if (error != DDI_SUCCESS) {
FC_DEBUG3(1, CE_CONT, "pfc_dma_map_in: real dma-map-in failed "
"error: %d virt: %p len %d\n", error, virt, len);
return (fc_priv_error(cp, "real dma-map-in failed"));
}
error = fc_ddi_dma_buf_bind_handle(h, bp, flags, DDI_DMA_SLEEP, NULL,
&c, &ccnt);
if ((error != DDI_DMA_MAPPED) || (ccnt != 1)) {
fc_ddi_dma_free_handle(&h);
FC_DEBUG3(1, CE_CONT, "pfc_dma_map_in: real dma-map-in failed "
"error: %d virt: %p len %d\n", error, virt, len);
return (fc_priv_error(cp, "real dma-map-in failed"));
}
if (c.dmac_size < len) {
error = fc_ddi_dma_unbind_handle(h);
if (error != DDI_SUCCESS) {
return (fc_priv_error(cp, "ddi_dma_unbind error"));
}
fc_ddi_dma_free_handle(&h);
return (fc_priv_error(cp, "ddi_dma_buf_bind size < len"));
}
FC_DEBUG1(9, CE_CONT, "pfc_dma_map_in: returning devaddr %x\n",
c.dmac_address);
cp->nresults = fc_int2cell(1);
fc_result(cp, 0) = fc_uint32_t2cell(c.dmac_address); /* XXX size */
/*
* Now we have to log this resource saving the handle and buf header
*/
ip = kmem_zalloc(sizeof (struct fc_resource), KM_SLEEP);
ip->type = RT_DMA;
ip->fc_dma_virt = virt;
ip->fc_dma_len = len;
ip->fc_dma_handle = h;
ip->fc_dma_devaddr = c.dmac_address;
ip->fc_dma_bp = bp;
fc_add_resource(rp, ip);
return (fc_success_op(ap, rp, cp));
}
static int
pfc_dma_sync(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
void *virt;
size_t len;
uint32_t devaddr;
int error;
struct fc_resource *ip;
if (fc_cell2int(cp->nargs) != 3)
return (fc_syntax_error(cp, "nargs must be 3"));
virt = fc_cell2ptr(fc_arg(cp, 2));
devaddr = fc_cell2uint32_t(fc_arg(cp, 1));
len = fc_cell2size(fc_arg(cp, 0));
/*
* Find if this virt is 'within' a request we know about
*/
fc_lock_resource_list(rp);
for (ip = rp->head; ip != NULL; ip = ip->next) {
if (ip->type != RT_DMA)
continue;
if (ip->fc_dma_devaddr != devaddr)
continue;
if (((char *)virt >= (char *)ip->fc_dma_virt) &&
(((char *)virt + len) <=
((char *)ip->fc_dma_virt + ip->fc_dma_len)))
break;
}
fc_unlock_resource_list(rp);
if (ip == NULL)
return (fc_priv_error(cp, "request not within a "
"known dma mapping"));
/*
* We know about this request, so we trust it enough to sync it.
* Unfortunately, we don't know which direction, so we'll do
* both directions.
*/
error = fc_ddi_dma_sync(ip->fc_dma_handle,
(char *)virt - (char *)ip->fc_dma_virt, len, DDI_DMA_SYNC_FORCPU);
error |= fc_ddi_dma_sync(ip->fc_dma_handle,
(char *)virt - (char *)ip->fc_dma_virt, len, DDI_DMA_SYNC_FORDEV);
if (error)
return (fc_priv_error(cp, "Call to ddi_dma_sync failed"));
cp->nresults = fc_int2cell(0);
return (fc_success_op(ap, rp, cp));
}
static int
pfc_dma_map_out(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
void *virt;
size_t len;
uint32_t devaddr;
struct fc_resource *ip;
int e;
if (fc_cell2int(cp->nargs) != 3)
return (fc_syntax_error(cp, "nargs must be 3"));
virt = fc_cell2ptr(fc_arg(cp, 2));
devaddr = fc_cell2uint32_t(fc_arg(cp, 1));
len = fc_cell2size(fc_arg(cp, 0));
/*
* Find if this virt matches a request we know about
*/
fc_lock_resource_list(rp);
for (ip = rp->head; ip != NULL; ip = ip->next) {
if (ip->type != RT_DMA)
continue;
if (ip->fc_dma_devaddr != devaddr)
continue;
if (ip->fc_dma_virt != virt)
continue;
if (len == ip->fc_dma_len)
break;
}
fc_unlock_resource_list(rp);
if (ip == NULL)
return (fc_priv_error(cp, "request doesn't match a "
"known dma mapping"));
/*
* ddi_dma_unbind_handle does an implied sync ...
*/
e = fc_ddi_dma_unbind_handle(ip->fc_dma_handle);
if (e != DDI_SUCCESS) {
cmn_err(CE_CONT, "pfc_dma_map_out: ddi_dma_unbind failed!\n");
}
fc_ddi_dma_free_handle(&ip->fc_dma_handle);
/*
* Tear down the physio mappings
*/
fc_physio_free(&ip->fc_dma_bp, ip->fc_dma_virt, ip->fc_dma_len);
/*
* remove the resource from the list and release it.
*/
fc_rem_resource(rp, ip);
kmem_free(ip, sizeof (struct fc_resource));
cp->nresults = fc_int2cell(0);
return (fc_success_op(ap, rp, cp));
}
static struct fc_resource *
next_dma_resource(fco_handle_t rp)
{
struct fc_resource *ip;
fc_lock_resource_list(rp);
for (ip = rp->head; ip != NULL; ip = ip->next)
if (ip->type == RT_DMA)
break;
fc_unlock_resource_list(rp);
return (ip);
}
static int
pfc_dma_cleanup(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
struct fc_resource *ip;
int e;
while ((ip = next_dma_resource(rp)) != NULL) {
FC_DEBUG2(9, CE_CONT, "pfc_dma_cleanup: virt %x len %x\n",
ip->fc_dma_virt, ip->fc_dma_len);
/*
* Free the dma handle
*/
e = fc_ddi_dma_unbind_handle(ip->fc_dma_handle);
if (e != DDI_SUCCESS) {
cmn_err(CE_CONT, "pfc_dma_cleanup: "
"ddi_dma_unbind failed!\n");
}
fc_ddi_dma_free_handle(&ip->fc_dma_handle);
/*
* Tear down the userland mapping and free the buf header
*/
fc_physio_free(&ip->fc_dma_bp, ip->fc_dma_virt, ip->fc_dma_len);
fc_rem_resource(rp, ip);
kmem_free(ip, sizeof (struct fc_resource));
}
cp->nresults = fc_int2cell(0);
return (fc_success_op(ap, rp, cp));
}
static int
pfc_map_in(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
size_t len;
int error;
caddr_t virt;
pci_regspec_t p, *ph;
struct fc_resource *ip;
ddi_device_acc_attr_t acc;
ddi_acc_handle_t h;
if (fc_cell2int(cp->nargs) != 4)
return (fc_syntax_error(cp, "nargs must be 4"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
p.pci_size_hi = 0;
p.pci_size_low = len = fc_cell2size(fc_arg(cp, 0));
p.pci_phys_hi = fc_cell2uint(fc_arg(cp, 1));
p.pci_phys_mid = fc_cell2uint(fc_arg(cp, 2));
p.pci_phys_low = fc_cell2uint(fc_arg(cp, 3));
acc.devacc_attr_version = DDI_DEVICE_ATTR_V0;
/*
* Fcode is expecting the bytes are not swapped.
*/
acc.devacc_attr_endian_flags = DDI_STRUCTURE_LE_ACC;
acc.devacc_attr_dataorder = DDI_STRICTORDER_ACC;
/*
* First We need to allocate the PCI Resource.
*/
error = pci_alloc_resource(rp->child, p);
if (error) {
return (fc_priv_error(cp, "pci map-in failed"));
}
error = pci_map_phys(rp->child, &p, &virt, &acc, &h);
if (error) {
return (fc_priv_error(cp, "pci map-in failed"));
}
cp->nresults = fc_int2cell(1);
fc_result(cp, 0) = fc_ptr2cell(virt);
/*
* Log this resource ...
*/
ip = kmem_zalloc(sizeof (struct fc_resource), KM_SLEEP);
ip->type = RT_MAP;
ip->fc_map_virt = virt;
ip->fc_map_len = len;
ip->fc_map_handle = h;
ph = kmem_zalloc(sizeof (pci_regspec_t), KM_SLEEP);
*ph = p;
ip->fc_regspec = ph; /* cache a copy of the reg spec */
fc_add_resource(rp, ip);
return (fc_success_op(ap, rp, cp));
}
static int
pfc_map_out(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
caddr_t virt;
size_t len;
struct fc_resource *ip;
if (fc_cell2int(cp->nargs) != 2)
return (fc_syntax_error(cp, "nargs must be 2"));
virt = fc_cell2ptr(fc_arg(cp, 1));
len = fc_cell2size(fc_arg(cp, 0));
/*
* Find if this request matches a mapping resource we set up.
*/
fc_lock_resource_list(rp);
for (ip = rp->head; ip != NULL; ip = ip->next) {
if (ip->type != RT_MAP)
continue;
if (ip->fc_map_virt != virt)
continue;
if (ip->fc_map_len == len)
break;
}
fc_unlock_resource_list(rp);
if (ip == NULL)
return (fc_priv_error(cp, "request doesn't match a "
"known mapping"));
pci_unmap_phys(&ip->fc_map_handle, ip->fc_regspec);
kmem_free(ip->fc_regspec, sizeof (pci_regspec_t));
/*
* remove the resource from the list and release it.
*/
fc_rem_resource(rp, ip);
kmem_free(ip, sizeof (struct fc_resource));
cp->nresults = fc_int2cell(0);
return (fc_success_op(ap, rp, cp));
}
static int
pfc_register_fetch(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
size_t len;
caddr_t virt;
int error;
uint64_t x;
uint32_t l;
uint16_t w;
uint8_t b;
char *name = fc_cell2ptr(cp->svc_name);
struct fc_resource *ip;
if (fc_cell2int(cp->nargs) != 1)
return (fc_syntax_error(cp, "nargs must be 1"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
virt = fc_cell2ptr(fc_arg(cp, 0));
/*
* Determine the access width .. we can switch on the 2nd
* character of the name which is "rx@", "rl@", "rb@" or "rw@"
*/
switch (*(name + 1)) {
case 'x': len = sizeof (x); break;
case 'l': len = sizeof (l); break;
case 'w': len = sizeof (w); break;
case 'b': len = sizeof (b); break;
}
/*
* Check the alignment ...
*/
if (((intptr_t)virt & (len - 1)) != 0)
return (fc_priv_error(cp, "unaligned access"));
/*
* Find if this virt is 'within' a request we know about
*/
fc_lock_resource_list(rp);
for (ip = rp->head; ip != NULL; ip = ip->next) {
if (ip->type != RT_MAP)
continue;
if ((virt >= (caddr_t)ip->fc_map_virt) && ((virt + len) <=
((caddr_t)ip->fc_map_virt + ip->fc_map_len)))
break;
}
fc_unlock_resource_list(rp);
if (ip == NULL)
return (fc_priv_error(cp, "request not within a "
"known mapping"));
/*
* XXX: We need access handle versions of peek/poke to move
* beyond the prototype ... we assume that we have hardware
* byte swapping enabled for pci register access here which
* is a huge dependency on the current implementation.
*/
switch (len) {
case sizeof (x):
error = ddi_peek64(rp->child, (int64_t *)virt, (int64_t *)&x);
break;
case sizeof (l):
error = ddi_peek32(rp->child, (int32_t *)virt, (int32_t *)&l);
break;
case sizeof (w):
error = ddi_peek16(rp->child, (int16_t *)virt, (int16_t *)&w);
break;
case sizeof (b):
error = ddi_peek8(rp->child, (int8_t *)virt, (int8_t *)&b);
break;
}
if (error) {
return (fc_priv_error(cp, "access error"));
}
cp->nresults = fc_int2cell(1);
switch (len) {
case sizeof (x): fc_result(cp, 0) = x; break;
case sizeof (l): fc_result(cp, 0) = fc_uint32_t2cell(l); break;
case sizeof (w): fc_result(cp, 0) = fc_uint16_t2cell(w); break;
case sizeof (b): fc_result(cp, 0) = fc_uint8_t2cell(b); break;
}
return (fc_success_op(ap, rp, cp));
}
static int
pfc_register_store(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
size_t len;
caddr_t virt;
int error;
uint64_t x;
uint32_t l;
uint16_t w;
uint8_t b;
char *name = fc_cell2ptr(cp->svc_name);
struct fc_resource *ip;
if (fc_cell2int(cp->nargs) != 2)
return (fc_syntax_error(cp, "nargs must be 2"));
virt = fc_cell2ptr(fc_arg(cp, 0));
/*
* Determine the access width .. we can switch on the 2nd
* character of the name which is "rl!", "rb!" or "rw!"
*/
switch (*(name + 1)) {
case 'x': len = sizeof (x); x = fc_arg(cp, 1); break;
case 'l': len = sizeof (l); l = fc_cell2uint32_t(fc_arg(cp, 1)); break;
case 'w': len = sizeof (w); w = fc_cell2uint16_t(fc_arg(cp, 1)); break;
case 'b': len = sizeof (b); b = fc_cell2uint8_t(fc_arg(cp, 1)); break;
}
/*
* Check the alignment ...
*/
if (((intptr_t)virt & (len - 1)) != 0)
return (fc_priv_error(cp, "unaligned access"));
/*
* Find if this virt is 'within' a request we know about
*/
fc_lock_resource_list(rp);
for (ip = rp->head; ip != NULL; ip = ip->next) {
if (ip->type != RT_MAP)
continue;
if ((virt >= (caddr_t)ip->fc_map_virt) && ((virt + len) <=
((caddr_t)ip->fc_map_virt + ip->fc_map_len)))
break;
}
fc_unlock_resource_list(rp);
if (ip == NULL)
return (fc_priv_error(cp, "request not within a "
"known mapping"));
/*
* XXX: We need access handle versions of peek/poke to move
* beyond the prototype ... we assume that we have hardware
* byte swapping enabled for pci register access here which
* is a huge dependency on the current implementation.
*/
switch (len) {
case sizeof (x):
error = ddi_poke64(rp->child, (int64_t *)virt, x);
break;
case sizeof (l):
error = ddi_poke32(rp->child, (int32_t *)virt, l);
break;
case sizeof (w):
error = ddi_poke16(rp->child, (int16_t *)virt, w);
break;
case sizeof (b):
error = ddi_poke8(rp->child, (int8_t *)virt, b);
break;
}
if (error) {
return (fc_priv_error(cp, "access error"));
}
cp->nresults = fc_int2cell(0);
return (fc_success_op(ap, rp, cp));
}
static int
pfc_config_fetch(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
caddr_t virt, v;
int error, reg, flags = 0;
size_t len;
uint32_t l, tmp;
uint16_t w;
uint8_t b;
char *name = fc_cell2ptr(cp->svc_name);
pci_regspec_t p;
ddi_device_acc_attr_t acc;
ddi_acc_handle_t h;
if (fc_cell2int(cp->nargs) != 1)
return (fc_syntax_error(cp, "nargs must be 1"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
/*
* Construct a config address pci reg property from the args.
* arg[0] is the configuration address.
*/
p.pci_phys_hi = fc_cell2uint(fc_arg(cp, 0));
p.pci_phys_mid = p.pci_phys_low = 0;
p.pci_size_hi = p.pci_size_low = 0;
/*
* Verify that the address is a configuration space address
* ss must be zero.
*/
if ((p.pci_phys_hi & PCI_ADDR_MASK) != PCI_ADDR_CONFIG) {
cmn_err(CE_CONT, "pfc_config_fetch: "
"invalid config addr: %x\n", p.pci_phys_hi);
return (fc_priv_error(cp, "non-config addr"));
}
/*
* Extract the register number from the config address and
* remove the register number from the physical address.
*/
reg = (p.pci_phys_hi & PCI_REG_REG_M) |
(((p.pci_phys_hi & PCI_REG_EXTREG_M) >> PCI_REG_EXTREG_SHIFT) << 8);
p.pci_phys_hi &= PCI_BDF_bits;
/*
* Determine the access width .. we can switch on the 9th
* character of the name which is "config-{l,w,b}@"
*/
switch (*(name + 7)) {
case 'l': len = sizeof (l); break;
case 'w': len = sizeof (w); break;
case 'b': len = sizeof (b); break;
}
/*
* Verify that the access is properly aligned
*/
if ((reg & (len - 1)) != 0)
return (fc_priv_error(cp, "unaligned access"));
/*
* Map in configuration space (temporarily)
*/
acc.devacc_attr_version = DDI_DEVICE_ATTR_V0;
acc.devacc_attr_endian_flags = DDI_STRUCTURE_LE_ACC;
acc.devacc_attr_dataorder = DDI_STRICTORDER_ACC;
error = pci_map_phys(rp->child, &p, &virt, &acc, &h);
if (error) {
return (fc_priv_error(cp, "pci config map-in failed"));
}
if (fcpci_indirect_map(rp->child) == DDI_SUCCESS)
flags |= PCICFG_CONF_INDIRECT_MAP;
if (flags & PCICFG_CONF_INDIRECT_MAP) {
tmp = (int32_t)ddi_get32(h, (uint32_t *)virt);
error = DDI_SUCCESS;
} else
error = ddi_peek32(rp->child, (int32_t *)virt, (int32_t *)&tmp);
if (error == DDI_SUCCESS)
if ((tmp == (int32_t)0xffffffff) || (tmp == -1)) {
error = DDI_FAILURE;
cmn_err(CE_CONT, "fcpcii: conf probe failed.l=%x", tmp);
}
if (error != DDI_SUCCESS) {
return (fc_priv_error(cp, "pci config fetch failed"));
}
/*
* XXX: We need access handle versions of peek/poke to move
* beyond the prototype ... we assume that we have hardware
* byte swapping enabled for pci register access here which
* is a huge dependency on the current implementation.
*/
v = virt + reg;
switch (len) {
case sizeof (l):
l = (int32_t)ddi_get32(h, (uint32_t *)v);
break;
case sizeof (w):
w = (int16_t)ddi_get16(h, (uint16_t *)v);
break;
case sizeof (b):
b = (int8_t)ddi_get8(h, (uint8_t *)v);
break;
}
/*
* Remove the temporary config space mapping
*/
pci_unmap_phys(&h, &p);
if (error) {
return (fc_priv_error(cp, "access error"));
}
cp->nresults = fc_int2cell(1);
switch (len) {
case sizeof (l): fc_result(cp, 0) = fc_uint32_t2cell(l); break;
case sizeof (w): fc_result(cp, 0) = fc_uint16_t2cell(w); break;
case sizeof (b): fc_result(cp, 0) = fc_uint8_t2cell(b); break;
}
return (fc_success_op(ap, rp, cp));
}
static int
pfc_config_store(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
caddr_t virt, v;
int error, reg, flags = 0;
size_t len;
uint32_t l, tmp;
uint16_t w;
uint8_t b;
char *name = fc_cell2ptr(cp->svc_name);
pci_regspec_t p;
ddi_device_acc_attr_t acc;
ddi_acc_handle_t h;
if (fc_cell2int(cp->nargs) != 2)
return (fc_syntax_error(cp, "nargs must be 2"));
/*
* Construct a config address pci reg property from the args.
* arg[0] is the configuration address. arg[1] is the data.
*/
p.pci_phys_hi = fc_cell2uint(fc_arg(cp, 0));
p.pci_phys_mid = p.pci_phys_low = 0;
p.pci_size_hi = p.pci_size_low = 0;
/*
* Verify that the address is a configuration space address
* ss must be zero.
*/
if ((p.pci_phys_hi & PCI_ADDR_MASK) != PCI_ADDR_CONFIG) {
cmn_err(CE_CONT, "pfc_config_store: "
"invalid config addr: %x\n", p.pci_phys_hi);
return (fc_priv_error(cp, "non-config addr"));
}
/*
* Extract the register number from the config address and
* remove the register number from the physical address.
*/
reg = (p.pci_phys_hi & PCI_REG_REG_M) |
(((p.pci_phys_hi & PCI_REG_EXTREG_M) >> PCI_REG_EXTREG_SHIFT) << 8);
p.pci_phys_hi &= PCI_BDF_bits;
/*
* Determine the access width .. we can switch on the 8th
* character of the name which is "config-{l,w,b}@"
*/
switch (*(name + 7)) {
case 'l': len = sizeof (l); l = fc_cell2uint32_t(fc_arg(cp, 1)); break;
case 'w': len = sizeof (w); w = fc_cell2uint16_t(fc_arg(cp, 1)); break;
case 'b': len = sizeof (b); b = fc_cell2uint8_t(fc_arg(cp, 1)); break;
}
/*
* Verify that the access is properly aligned
*/
if ((reg & (len - 1)) != 0)
return (fc_priv_error(cp, "unaligned access"));
/*
* Map in configuration space (temporarily)
*/
acc.devacc_attr_version = DDI_DEVICE_ATTR_V0;
acc.devacc_attr_endian_flags = DDI_STRUCTURE_LE_ACC;
acc.devacc_attr_dataorder = DDI_STRICTORDER_ACC;
error = pci_map_phys(rp->child, &p, &virt, &acc, &h);
if (error) {
return (fc_priv_error(cp, "pci config map-in failed"));
}
if (fcpci_indirect_map(rp->child) == DDI_SUCCESS)
flags |= PCICFG_CONF_INDIRECT_MAP;
if (flags & PCICFG_CONF_INDIRECT_MAP) {
tmp = (int32_t)ddi_get32(h, (uint32_t *)virt);
error = DDI_SUCCESS;
} else
error = ddi_peek32(rp->child, (int32_t *)virt, (int32_t *)&tmp);
if (error == DDI_SUCCESS)
if ((tmp == (int32_t)0xffffffff) || (tmp == -1)) {
error = DDI_FAILURE;
cmn_err(CE_CONT, "fcpci: conf probe failed.l=%x", tmp);
}
if (error != DDI_SUCCESS) {
return (fc_priv_error(cp, "pci config store failed"));
}
/*
* XXX: We need access handle versions of peek/poke to move
* beyond the prototype ... we assume that we have hardware
* byte swapping enabled for pci register access here which
* is a huge dependency on the current implementation.
*/
v = virt + reg;
switch (len) {
case sizeof (l):
ddi_put32(h, (uint32_t *)v, (uint32_t)l);
break;
case sizeof (w):
ddi_put16(h, (uint16_t *)v, (uint16_t)w);
break;
case sizeof (b):
ddi_put8(h, (uint8_t *)v, (uint8_t)b);
break;
}
/*
* Remove the temporary config space mapping
*/
pci_unmap_phys(&h, &p);
if (error) {
return (fc_priv_error(cp, "access error"));
}
cp->nresults = fc_int2cell(0);
return (fc_success_op(ap, rp, cp));
}
static int
pfc_get_fcode(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
caddr_t name_virt, fcode_virt;
char *name, *fcode;
int fcode_len, status;
if (fc_cell2int(cp->nargs) != 3)
return (fc_syntax_error(cp, "nargs must be 3"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
name_virt = fc_cell2ptr(fc_arg(cp, 0));
fcode_virt = fc_cell2ptr(fc_arg(cp, 1));
fcode_len = fc_cell2int(fc_arg(cp, 2));
name = kmem_zalloc(FC_SVC_NAME_LEN, KM_SLEEP);
if (copyinstr(fc_cell2ptr(name_virt), name,
FC_SVC_NAME_LEN - 1, NULL)) {
status = 0;
} else {
fcode = kmem_zalloc(fcode_len, KM_SLEEP);
if ((status = prom_get_fcode(name, fcode)) != 0) {
if (copyout((void *)fcode, (void *)fcode_virt,
fcode_len)) {
cmn_err(CE_WARN, " pfc_get_fcode: Unable "
"to copy out fcode image\n");
status = 0;
}
}
kmem_free(fcode, fcode_len);
}
kmem_free(name, FC_SVC_NAME_LEN);
cp->nresults = fc_int2cell(1);
fc_result(cp, 0) = status;
return (fc_success_op(ap, rp, cp));
}
static int
pfc_get_fcode_size(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
caddr_t virt;
char *name;
int len;
if (fc_cell2int(cp->nargs) != 1)
return (fc_syntax_error(cp, "nargs must be 1"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
virt = fc_cell2ptr(fc_arg(cp, 0));
name = kmem_zalloc(FC_SVC_NAME_LEN, KM_SLEEP);
if (copyinstr(fc_cell2ptr(virt), name,
FC_SVC_NAME_LEN - 1, NULL)) {
len = 0;
} else {
len = prom_get_fcode_size(name);
}
kmem_free(name, FC_SVC_NAME_LEN);
cp->nresults = fc_int2cell(1);
fc_result(cp, 0) = len;
return (fc_success_op(ap, rp, cp));
}
/*
* Return the physical probe address: lo=0, mid=0, hi-config-addr
*/
static int
pfc_probe_address(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
if (fc_cell2int(cp->nargs) != 0)
return (fc_syntax_error(cp, "nargs must be 0"));
if (fc_cell2int(cp->nresults) < 2)
return (fc_syntax_error(cp, "nresults must be >= 3"));
cp->nresults = fc_int2cell(2);
fc_result(cp, 1) = fc_int2cell(0); /* phys.lo */
fc_result(cp, 0) = fc_int2cell(0); /* phys.mid */
return (fc_success_op(ap, rp, cp));
}
/*
* Return the phys.hi component of the probe address.
*/
static int
pfc_probe_space(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
struct pci_ops_bus_args *ba = rp->bus_args;
ASSERT(ba);
if (fc_cell2int(cp->nargs) != 0)
return (fc_syntax_error(cp, "nargs must be 0"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
cp->nresults = fc_int2cell(1);
fc_result(cp, 0) = fc_uint32_t2cell(ba->config_address); /* phys.hi */
return (fc_success_op(ap, rp, cp));
}
static int
pfc_config_child(dev_info_t *ap, fco_handle_t rp, fc_ci_t *cp)
{
fc_phandle_t h;
if (fc_cell2int(cp->nargs) != 0)
return (fc_syntax_error(cp, "nargs must be 0"));
if (fc_cell2int(cp->nresults) < 1)
return (fc_syntax_error(cp, "nresults must be >= 1"));
h = fc_dip_to_phandle(fc_handle_to_phandle_head(rp), rp->child);
cp->nresults = fc_int2cell(1);
fc_result(cp, 0) = fc_phandle2cell(h);
return (fc_success_op(ap, rp, cp));
}
int
pci_alloc_mem_chunk(dev_info_t *dip, uint64_t mem_align, uint64_t *mem_size,
uint64_t *mem_answer)
{
ndi_ra_request_t req;
int rval;
bzero((caddr_t)&req, sizeof (ndi_ra_request_t));
req.ra_flags = NDI_RA_ALLOC_BOUNDED;
req.ra_boundbase = 0;
req.ra_boundlen = PCI_4GIG_LIMIT;
req.ra_len = *mem_size;
req.ra_align_mask = mem_align - 1;
rval = ndi_ra_alloc(dip, &req, mem_answer, mem_size,
NDI_RA_TYPE_MEM, NDI_RA_PASS);
return (rval);
}
int
pci_alloc_io_chunk(dev_info_t *dip, uint64_t io_align, uint64_t *io_size,
uint64_t *io_answer)
{
ndi_ra_request_t req;
int rval;
bzero((caddr_t)&req, sizeof (ndi_ra_request_t));
req.ra_flags = (NDI_RA_ALLOC_BOUNDED | NDI_RA_ALLOC_PARTIAL_OK);
req.ra_boundbase = 0;
req.ra_boundlen = PCI_4GIG_LIMIT;
req.ra_len = *io_size;
req.ra_align_mask = io_align - 1;
rval = ndi_ra_alloc(dip, &req, io_answer, io_size,
NDI_RA_TYPE_IO, NDI_RA_PASS);
return (rval);
}
int
pci_alloc_resource(dev_info_t *dip, pci_regspec_t phys_spec)
{
uint64_t answer;
uint64_t alen;
int offset, tmp;
pci_regspec_t config;
caddr_t virt, v;
ddi_device_acc_attr_t acc;
ddi_acc_handle_t h;
ndi_ra_request_t request;
pci_regspec_t *assigned;
int assigned_len, entries, i, l, flags = 0, error;
l = phys_spec.pci_size_low;
if (ddi_getlongprop(DDI_DEV_T_ANY, dip,
DDI_PROP_DONTPASS, "assigned-addresses", (caddr_t)&assigned,
&assigned_len) == DDI_PROP_SUCCESS) {
entries = assigned_len / (sizeof (pci_regspec_t));
/*
* Walk through the assigned-addresses entries. If there is
* a match, there is no need to allocate the resource.
*/
for (i = 0; i < entries; i++) {
if (assigned[i].pci_phys_hi == phys_spec.pci_phys_hi) {
if (assigned[i].pci_size_low >=
phys_spec.pci_size_low) {
kmem_free(assigned, assigned_len);
return (0);
}
/*
* Fcode wants to assign more than what
* probe found.
*/
(void) pci_free_resource(dip, assigned[i]);
/*
* Go on to allocate resources.
*/
break;
}
/*
* Check if Fcode wants to map using different
* NPT bits.
*/
if (PCI_REG_BDFR_G(assigned[i].pci_phys_hi) ==
PCI_REG_BDFR_G(phys_spec.pci_phys_hi)) {
/*
* It is an error to change SS bits
*/
if (PCI_REG_ADDR_G(assigned[i].pci_phys_hi) !=
PCI_REG_ADDR_G(phys_spec.pci_phys_hi)) {
FC_DEBUG2(2, CE_WARN, "Fcode changing "
"ss bits in reg %x -- %x",
assigned[i].pci_phys_hi,
phys_spec.pci_phys_hi);
}
/*
* Allocate enough
*/
l = MAX(assigned[i].pci_size_low,
phys_spec.pci_size_low);
phys_spec.pci_size_low = l;
(void) pci_free_resource(dip, assigned[i]);
/*
* Go on to allocate resources.
*/
break;
}
}
kmem_free(assigned, assigned_len);
}
bzero((caddr_t)&request, sizeof (ndi_ra_request_t));
config.pci_phys_hi = PCI_CONF_ADDR_MASK & phys_spec.pci_phys_hi;
config.pci_phys_hi &= ~PCI_REG_REG_M;
config.pci_phys_mid = config.pci_phys_low = 0;
config.pci_size_hi = config.pci_size_low = 0;
/*
* Map in configuration space (temporarily)
*/
acc.devacc_attr_version = DDI_DEVICE_ATTR_V0;
acc.devacc_attr_endian_flags = DDI_STRUCTURE_LE_ACC;
acc.devacc_attr_dataorder = DDI_STRICTORDER_ACC;
if (error = pci_map_phys(dip, &config, &virt, &acc, &h)) {
return (1);
}
if (fcpci_indirect_map(dip) == DDI_SUCCESS)
flags |= PCICFG_CONF_INDIRECT_MAP;
if (flags & PCICFG_CONF_INDIRECT_MAP) {
tmp = (int32_t)ddi_get32(h, (uint32_t *)virt);
error = DDI_SUCCESS;
} else
error = ddi_peek32(dip, (int32_t *)virt, (int32_t *)&tmp);
if (error == DDI_SUCCESS)
if ((tmp == (int32_t)0xffffffff) || (tmp == -1)) {
error = DDI_FAILURE;
}
if (error != DDI_SUCCESS) {
return (1);
}
request.ra_flags |= NDI_RA_ALIGN_SIZE;
request.ra_boundbase = 0;
request.ra_boundlen = PCI_4GIG_LIMIT;
offset = PCI_REG_REG_G(phys_spec.pci_phys_hi);
v = virt + offset;
if (PCI_REG_REG_G(phys_spec.pci_phys_hi) == PCI_CONF_ROM) {
request.ra_len = l;
request.ra_flags ^= NDI_RA_ALLOC_BOUNDED;
/* allocate memory space from the allocator */
if (ndi_ra_alloc(ddi_get_parent(dip),
&request, &answer, &alen, NDI_RA_TYPE_MEM,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
return (1);
}
FC_DEBUG3(1, CE_CONT, "ROM addr = [0x%x.%x] len [0x%x]\n",
HIADDR(answer), LOADDR(answer), alen);
/* program the low word */
ddi_put32(h, (uint32_t *)v, LOADDR(answer));
phys_spec.pci_phys_low = LOADDR(answer);
phys_spec.pci_phys_mid = HIADDR(answer);
} else {
request.ra_len = l;
switch (PCI_REG_ADDR_G(phys_spec.pci_phys_hi)) {
case PCI_REG_ADDR_G(PCI_ADDR_MEM64):
request.ra_flags ^= NDI_RA_ALLOC_BOUNDED;
if (phys_spec.pci_phys_hi & PCI_REG_REL_M) {
/*
* If it is a non relocatable address,
* then specify the address we want.
*/
request.ra_flags = NDI_RA_ALLOC_SPECIFIED;
request.ra_addr = (uint64_t)LADDR(
phys_spec.pci_phys_low,
phys_spec.pci_phys_mid);
}
/* allocate memory space from the allocator */
if (ndi_ra_alloc(ddi_get_parent(dip),
&request, &answer, &alen, NDI_RA_TYPE_MEM,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
if (request.ra_flags == NDI_RA_ALLOC_SPECIFIED)
cmn_err(CE_WARN, "Unable to allocate "
"non relocatable address 0x%p\n",
(void *) request.ra_addr);
return (1);
}
FC_DEBUG3(1, CE_CONT,
"64 addr = [0x%x.%x] len [0x%x]\n",
HIADDR(answer),
LOADDR(answer),
alen);
/* program the low word */
ddi_put32(h, (uint32_t *)v, LOADDR(answer));
/* program the high word with value zero */
v += 4;
ddi_put32(h, (uint32_t *)v, HIADDR(answer));
phys_spec.pci_phys_low = LOADDR(answer);
phys_spec.pci_phys_mid = HIADDR(answer);
/*
* currently support 32b address space
* assignments only.
*/
phys_spec.pci_phys_hi ^= PCI_ADDR_MEM64 ^
PCI_ADDR_MEM32;
break;
case PCI_REG_ADDR_G(PCI_ADDR_MEM32):
request.ra_flags |= NDI_RA_ALLOC_BOUNDED;
if (phys_spec.pci_phys_hi & PCI_REG_REL_M) {
/*
* If it is a non relocatable address,
* then specify the address we want.
*/
request.ra_flags = NDI_RA_ALLOC_SPECIFIED;
request.ra_addr = (uint64_t)
phys_spec.pci_phys_low;
}
/* allocate memory space from the allocator */
if (ndi_ra_alloc(ddi_get_parent(dip),
&request, &answer, &alen, NDI_RA_TYPE_MEM,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
if (request.ra_flags == NDI_RA_ALLOC_SPECIFIED)
cmn_err(CE_WARN, "Unable to allocate "
"non relocatable address 0x%p\n",
(void *) request.ra_addr);
return (1);
}
FC_DEBUG3(1, CE_CONT,
"32 addr = [0x%x.%x] len [0x%x]\n",
HIADDR(answer),
LOADDR(answer),
alen);
/* program the low word */
ddi_put32(h, (uint32_t *)v, LOADDR(answer));
phys_spec.pci_phys_low = LOADDR(answer);
break;
case PCI_REG_ADDR_G(PCI_ADDR_IO):
request.ra_flags |= NDI_RA_ALLOC_BOUNDED;
if (phys_spec.pci_phys_hi & PCI_REG_REL_M) {
/*
* If it is a non relocatable address,
* then specify the address we want.
*/
request.ra_flags = NDI_RA_ALLOC_SPECIFIED;
request.ra_addr = (uint64_t)
phys_spec.pci_phys_low;
}
/* allocate I/O space from the allocator */
if (ndi_ra_alloc(ddi_get_parent(dip),
&request, &answer, &alen, NDI_RA_TYPE_IO,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
if (request.ra_flags ==
NDI_RA_ALLOC_SPECIFIED)
cmn_err(CE_WARN, "Unable to allocate "
"non relocatable IO Space 0x%p\n",
(void *) request.ra_addr);
return (1);
}
FC_DEBUG3(1, CE_CONT,
"I/O addr = [0x%x.%x] len [0x%x]\n",
HIADDR(answer),
LOADDR(answer),
alen);
ddi_put32(h, (uint32_t *)v, LOADDR(answer));
phys_spec.pci_phys_low = LOADDR(answer);
break;
default:
pci_unmap_phys(&h, &config);
return (1);
} /* switch */
}
/*
* Now that memory locations are assigned,
* update the assigned address property.
*/
if (pfc_update_assigned_prop(dip, &phys_spec)) {
pci_unmap_phys(&h, &config);
return (1);
}
pci_unmap_phys(&h, &config);
return (0);
}
int
pci_free_resource(dev_info_t *dip, pci_regspec_t phys_spec)
{
int offset, tmp;
pci_regspec_t config;
caddr_t virt, v;
ddi_device_acc_attr_t acc;
ddi_acc_handle_t h;
ndi_ra_request_t request;
int l, error, flags = 0;
bzero((caddr_t)&request, sizeof (ndi_ra_request_t));
config.pci_phys_hi = PCI_CONF_ADDR_MASK & phys_spec.pci_phys_hi;
config.pci_phys_hi &= ~PCI_REG_REG_M;
config.pci_phys_mid = config.pci_phys_low = 0;
config.pci_size_hi = config.pci_size_low = 0;
/*
* Map in configuration space (temporarily)
*/
acc.devacc_attr_version = DDI_DEVICE_ATTR_V0;
acc.devacc_attr_endian_flags = DDI_STRUCTURE_LE_ACC;
acc.devacc_attr_dataorder = DDI_STRICTORDER_ACC;
if (error = pci_map_phys(dip, &config, &virt, &acc, &h)) {
return (1);
}
if (fcpci_indirect_map(dip) == DDI_SUCCESS)
flags |= PCICFG_CONF_INDIRECT_MAP;
if (flags & PCICFG_CONF_INDIRECT_MAP) {
tmp = (int32_t)ddi_get32(h, (uint32_t *)virt);
error = DDI_SUCCESS;
} else
error = ddi_peek32(dip, (int32_t *)virt, (int32_t *)&tmp);
if (error == DDI_SUCCESS)
if ((tmp == (int32_t)0xffffffff) || (tmp == -1)) {
error = DDI_FAILURE;
}
if (error != DDI_SUCCESS) {
return (1);
}
offset = PCI_REG_REG_G(phys_spec.pci_phys_hi);
v = virt + offset;
/*
* Pick up the size to be freed. It may be different from
* what probe finds.
*/
l = phys_spec.pci_size_low;
if (PCI_REG_REG_G(phys_spec.pci_phys_hi) == PCI_CONF_ROM) {
/* free memory back to the allocator */
if (ndi_ra_free(ddi_get_parent(dip), phys_spec.pci_phys_low,
l, NDI_RA_TYPE_MEM,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
return (1);
}
/* Unmap the BAR by writing a zero */
ddi_put32(h, (uint32_t *)v, 0);
} else {
switch (PCI_REG_ADDR_G(phys_spec.pci_phys_hi)) {
case PCI_REG_ADDR_G(PCI_ADDR_MEM64):
/* free memory back to the allocator */
if (ndi_ra_free(ddi_get_parent(dip),
LADDR(phys_spec.pci_phys_low,
phys_spec.pci_phys_mid),
l, NDI_RA_TYPE_MEM,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
return (1);
}
break;
case PCI_REG_ADDR_G(PCI_ADDR_MEM32):
/* free memory back to the allocator */
if (ndi_ra_free(ddi_get_parent(dip),
phys_spec.pci_phys_low,
l, NDI_RA_TYPE_MEM,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
return (1);
}
break;
case PCI_REG_ADDR_G(PCI_ADDR_IO):
/* free I/O space back to the allocator */
if (ndi_ra_free(ddi_get_parent(dip),
phys_spec.pci_phys_low,
l, NDI_RA_TYPE_IO,
NDI_RA_PASS) != NDI_SUCCESS) {
pci_unmap_phys(&h, &config);
return (1);
}
break;
default:
pci_unmap_phys(&h, &config);
return (1);
} /* switch */
}
/*
* Now that memory locations are assigned,
* update the assigned address property.
*/
FC_DEBUG1(1, CE_CONT, "updating assigned-addresss for %x\n",
phys_spec.pci_phys_hi);
if (pfc_remove_assigned_prop(dip, &phys_spec)) {
pci_unmap_phys(&h, &config);
return (1);
}
pci_unmap_phys(&h, &config);
return (0);
}
int
pci_map_phys(dev_info_t *dip, pci_regspec_t *phys_spec,
caddr_t *addrp, ddi_device_acc_attr_t *accattrp,
ddi_acc_handle_t *handlep)
{
ddi_map_req_t mr;
ddi_acc_hdl_t *hp;
int result;
*handlep = impl_acc_hdl_alloc(KM_SLEEP, NULL);
hp = impl_acc_hdl_get(*handlep);
hp->ah_vers = VERS_ACCHDL;
hp->ah_dip = dip;
hp->ah_rnumber = 0;
hp->ah_offset = 0;
hp->ah_len = 0;
hp->ah_acc = *accattrp;
mr.map_op = DDI_MO_MAP_LOCKED;
mr.map_type = DDI_MT_REGSPEC;
mr.map_obj.rp = (struct regspec *)phys_spec;
mr.map_prot = PROT_READ | PROT_WRITE;
mr.map_flags = DDI_MF_KERNEL_MAPPING;
mr.map_handlep = hp;
mr.map_vers = DDI_MAP_VERSION;
result = ddi_map(dip, &mr, 0, 0, addrp);
if (result != DDI_SUCCESS) {
impl_acc_hdl_free(*handlep);
*handlep = (ddi_acc_handle_t)NULL;
} else {
hp->ah_addr = *addrp;
}
return (result);
}
void
pci_unmap_phys(ddi_acc_handle_t *handlep, pci_regspec_t *ph)
{
ddi_map_req_t mr;
ddi_acc_hdl_t *hp;
hp = impl_acc_hdl_get(*handlep);
ASSERT(hp);
mr.map_op = DDI_MO_UNMAP;
mr.map_type = DDI_MT_REGSPEC;
mr.map_obj.rp = (struct regspec *)ph;
mr.map_prot = PROT_READ | PROT_WRITE;
mr.map_flags = DDI_MF_KERNEL_MAPPING;
mr.map_handlep = hp;
mr.map_vers = DDI_MAP_VERSION;
(void) ddi_map(hp->ah_dip, &mr, hp->ah_offset,
hp->ah_len, &hp->ah_addr);
impl_acc_hdl_free(*handlep);
*handlep = (ddi_acc_handle_t)NULL;
}
int
pfc_update_assigned_prop(dev_info_t *dip, pci_regspec_t *newone)
{
int alen;
pci_regspec_t *assigned;
caddr_t newreg;
uint_t status;
status = ddi_getlongprop(DDI_DEV_T_ANY, dip, DDI_PROP_DONTPASS,
"assigned-addresses", (caddr_t)&assigned, &alen);
switch (status) {
case DDI_PROP_SUCCESS:
break;
case DDI_PROP_NO_MEMORY:
return (1);
default:
(void) ndi_prop_update_int_array(DDI_DEV_T_NONE, dip,
"assigned-addresses", (int *)newone,
sizeof (*newone)/sizeof (int));
return (0);
}
/*
* Allocate memory for the existing
* assigned-addresses(s) plus one and then
* build it.
*/
newreg = kmem_zalloc(alen+sizeof (*newone), KM_SLEEP);
bcopy(assigned, newreg, alen);
bcopy(newone, newreg + alen, sizeof (*newone));
/*
* Write out the new "assigned-addresses" spec
*/
(void) ndi_prop_update_int_array(DDI_DEV_T_NONE, dip,
"assigned-addresses", (int *)newreg,
(alen + sizeof (*newone))/sizeof (int));
kmem_free((caddr_t)newreg, alen+sizeof (*newone));
kmem_free(assigned, alen);
return (0);
}
int
pfc_remove_assigned_prop(dev_info_t *dip, pci_regspec_t *oldone)
{
int alen, new_len, num_entries, i;
pci_regspec_t *assigned;
uint_t status;
status = ddi_getlongprop(DDI_DEV_T_ANY, dip, DDI_PROP_DONTPASS,
"assigned-addresses", (caddr_t)&assigned, &alen);
switch (status) {
case DDI_PROP_SUCCESS:
break;
case DDI_PROP_NO_MEMORY:
return (1);
default:
return (0);
}
num_entries = alen / sizeof (pci_regspec_t);
new_len = alen - sizeof (pci_regspec_t);
/*
* Search for the memory being removed.
*/
for (i = 0; i < num_entries; i++) {
if (assigned[i].pci_phys_hi == oldone->pci_phys_hi) {
if (new_len == 0) {
(void) ndi_prop_remove(DDI_DEV_T_NONE, dip,
"assigned-addresses");
break;
}
if ((new_len - (i * sizeof (pci_regspec_t)))
== 0) {
FC_DEBUG1(1, CE_CONT, "assigned-address entry "
"%x removed from property (last entry)\n",
oldone->pci_phys_hi);
} else {
bcopy((void *)(assigned + i + 1),
(void *)(assigned + i),
(new_len - (i * sizeof (pci_regspec_t))));
FC_DEBUG1(1, CE_CONT, "assigned-address entry "
"%x removed from property\n",
oldone->pci_phys_hi);
}
(void) ndi_prop_update_int_array(DDI_DEV_T_NONE,
dip, "assigned-addresses", (int *)assigned,
(new_len/sizeof (int)));
break;
}
}
kmem_free(assigned, alen);
return (0);
}
/*
* we recognize the non transparent bridge child nodes with the
* following property. This is specific to this implementation only.
* This property is specific to AP nodes only.
*/
#define PCICFG_DEV_CONF_MAP_PROP "pci-parent-indirect"
/*
* If a non transparent bridge drives a hotplug/hotswap bus, then
* the following property must be defined for the node either by
* the driver or the OBP.
*/
#define PCICFG_BUS_CONF_MAP_PROP "pci-conf-indirect"
/*
* this function is called only for SPARC platforms, where we may have
* a mix n' match of direct vs indirectly mapped configuration space.
*/
/*ARGSUSED*/
static int
fcpci_indirect_map(dev_info_t *dip)
{
int rc = DDI_FAILURE;
if (ddi_getprop(DDI_DEV_T_ANY, ddi_get_parent(dip), 0,
PCICFG_DEV_CONF_MAP_PROP, DDI_FAILURE) != DDI_FAILURE)
rc = DDI_SUCCESS;
else
if (ddi_getprop(DDI_DEV_T_ANY, ddi_get_parent(dip),
0, PCICFG_BUS_CONF_MAP_PROP, DDI_FAILURE) != DDI_FAILURE)
rc = DDI_SUCCESS;
return (rc);
}