coreboot-kgpe-d16/util/sconfig/main.c

775 lines
17 KiB
C
Raw Normal View History

/*
* sconfig, coreboot device tree compiler
*
* Copyright (C) 2010 coresystems GmbH
* written by Patrick Georgi <patrick@georgi-clan.de>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; version 2 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*/
#include <ctype.h>
#include "sconfig.h"
#include "sconfig.tab.h"
extern int linenum;
struct device *head, *lastnode;
static struct chip chip_header;
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
/*
* This is intentionally shared between chip and device structure ids because it
* is easier to track the order of parsing for chip and device.
*/
static int count = 0;
typedef enum {
UNSLASH,
SPLIT_1ST,
TO_LOWER,
TO_UPPER,
} translate_t;
static struct device root;
static struct chip_instance mainboard_instance;
static struct chip mainboard_chip = {
.name = "mainboard",
.name_underscore = "mainboard",
.instance = &mainboard_instance,
};
static struct chip_instance mainboard_instance = {
.id = 0,
.chip = &mainboard_chip,
};
static struct device root = {
.name = "dev_root",
.id = 0,
.chip_instance = &mainboard_instance,
.path = " .type = DEVICE_PATH_ROOT ",
.ops = "&default_dev_ops_root",
.parent = &root,
.enabled = 1
};
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
static struct queue {
void *data;
struct queue *next;
struct queue *prev;
} *q;
#define S_ALLOC(_s) s_alloc(__func__, _s)
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
static void *s_alloc(const char *f, size_t s)
{
void *data = calloc(1, s);
if (!data) {
fprintf(stderr, "%s: Failed to alloc mem!\n", f, s);
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
exit(1);
}
return data;
}
struct queue *new_queue_entry(void *data)
{
struct queue *e = S_ALLOC(sizeof(*e));
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
e->data = data;
e->next = e->prev = e;
return e;
}
void chip_enqueue_tail(void *data)
{
struct queue *tmp = new_queue_entry(data);
if (!q) {
q = tmp;
return;
}
q->prev->next = tmp;
tmp->prev = q->prev;
q->prev = tmp;
tmp->next = q;
}
void *chip_dequeue_tail(void)
{
struct queue *tmp = q->prev;
void *data;
if (tmp == q)
q = NULL;
else {
tmp->prev->next = q;
q->prev = tmp->prev;
}
data = tmp->data;
free(tmp);
return data;
}
static struct device *new_dev(struct device *parent)
{
struct device *dev = S_ALLOC(sizeof(struct device));
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
dev->id = ++count;
dev->parent = parent;
dev->subsystem_vendor = -1;
dev->subsystem_device = -1;
head->next = dev;
head = dev;
return dev;
}
static int device_match(struct device *a, struct device *b)
{
if ((a->bustype == b->bustype) && (a->parent == b->parent)
&& (a->path_a == b->path_a) && (a->path_b == b->path_b))
return 1;
return 0;
}
void fold_in(struct device *parent)
{
struct device *child = parent->children;
struct device *latest = 0;
while (child != latest) {
if (child->children) {
if (!latest)
latest = child->children;
parent->latestchild->next_sibling = child->children;
parent->latestchild = child->latestchild;
}
child = child->next_sibling;
}
}
int yywrap(void)
{
return 1;
}
void yyerror(char const *str)
{
extern char *yytext;
fprintf(stderr, "line %d: %s: %s\n", linenum + 1, yytext, str);
exit(1);
}
void postprocess_devtree(void)
{
root.next_sibling = root.children;
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
/*
* Since root is statically created we need to call fold_in explicitly
* here to have the next_sibling and latestchild setup correctly.
*/
fold_in(&root);
struct device *dev = &root;
while (dev) {
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
/* skip functions of the same device in sibling chain */
while (dev->sibling && dev->sibling->used)
dev->sibling = dev->sibling->sibling;
/* skip duplicate function elements in next chain */
while (dev->next && dev->next->used)
dev->next = dev->next->next;
dev = dev->next_sibling;
}
}
char *translate_name(const char *str, translate_t mode)
{
char *b, *c;
b = c = strdup(str);
while (c && *c) {
if ((mode == SPLIT_1ST) && (*c == '/')) {
*c = 0;
break;
}
if (*c == '/')
*c = '_';
if (*c == '-')
*c = '_';
if (mode == TO_UPPER)
*c = toupper(*c);
if (mode == TO_LOWER)
*c = tolower(*c);
c++;
}
return b;
}
static struct chip *get_chip(char *path)
{
struct chip *h = &chip_header;
while (h->next) {
int result = strcmp(path, h->next->name);
if (result == 0)
return h->next;
if (result < 0)
break;
h = h->next;
}
struct chip *new_chip = S_ALLOC(sizeof(struct chip));
new_chip->next = h->next;
h->next = new_chip;
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
new_chip->chiph_exists = 1;
new_chip->name = path;
new_chip->name_underscore = translate_name(path, UNSLASH);
struct stat st;
char *chip_h = S_ALLOC(strlen(path) + 18);
sprintf(chip_h, "src/%s", path);
if ((stat(chip_h, &st) == -1) && (errno == ENOENT)) {
/* root_complex gets away without a separate directory, but
* exists on on pretty much all AMD chipsets.
*/
if (!strstr(path, "/root_complex")) {
fprintf(stderr, "ERROR: Chip component %s does not exist.\n",
path);
exit(1);
}
}
sprintf(chip_h, "src/%s/chip.h", path);
if ((stat(chip_h, &st) == -1) && (errno == ENOENT))
new_chip->chiph_exists = 0;
free(chip_h);
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
return new_chip;
}
struct chip_instance *new_chip_instance(char *path)
{
struct chip *chip = get_chip(path);
struct chip_instance *instance = S_ALLOC(sizeof(*instance));
instance->id = ++count;
instance->chip = chip;
instance->next = chip->instance;
chip->instance = instance;
return instance;
}
struct device *new_device(struct device *parent,
struct chip_instance *chip_instance,
const int bustype, const char *devnum,
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
int enabled)
{
struct device *new_d = new_dev(parent);
new_d->bustype = bustype;
char *tmp;
new_d->path_a = strtol(devnum, &tmp, 16);
if (*tmp == '.') {
tmp++;
new_d->path_b = strtol(tmp, NULL, 16);
}
char *name = S_ALLOC(10);
sprintf(name, "_dev%d", new_d->id);
new_d->name = name;
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
new_d->enabled = enabled;
new_d->chip_instance = chip_instance;
if (parent->latestchild) {
parent->latestchild->next_sibling = new_d;
parent->latestchild->sibling = new_d;
}
parent->latestchild = new_d;
if (!parent->children)
parent->children = new_d;
lastnode->next = new_d;
lastnode = new_d;
switch (bustype) {
case PCI:
new_d->path = ".type=DEVICE_PATH_PCI,{.pci={ .devfn = PCI_DEVFN(0x%x,%d)}}";
break;
case PNP:
new_d->path = ".type=DEVICE_PATH_PNP,{.pnp={ .port = 0x%x, .device = 0x%x }}";
break;
case I2C:
new_d->path = ".type=DEVICE_PATH_I2C,{.i2c={ .device = 0x%x, .mode_10bit = %d }}";
break;
case APIC:
new_d->path = ".type=DEVICE_PATH_APIC,{.apic={ .apic_id = 0x%x }}";
break;
case CPU_CLUSTER:
new_d->path = ".type=DEVICE_PATH_CPU_CLUSTER,{.cpu_cluster={ .cluster = 0x%x }}";
break;
case CPU:
new_d->path = ".type=DEVICE_PATH_CPU,{.cpu={ .id = 0x%x }}";
break;
case DOMAIN:
new_d->path = ".type=DEVICE_PATH_DOMAIN,{.domain={ .domain = 0x%x }}";
break;
case IOAPIC:
new_d->path = ".type=DEVICE_PATH_IOAPIC,{.ioapic={ .ioapic_id = 0x%x }}";
break;
case GENERIC:
new_d->path = ".type=DEVICE_PATH_GENERIC,{.generic={ .id = 0x%x, .subid = 0x%x }}";
break;
case SPI:
new_d->path = ".type=DEVICE_PATH_SPI,{.spi={ .cs = 0x%x }}";
break;
case USB:
new_d->path = ".type=DEVICE_PATH_USB,{.usb={ .port_type = %d, .port_id = %d }}";
break;
case MMIO:
new_d->path = ".type=DEVICE_PATH_MMIO,{.mmio={ .addr = 0x%x }}";
break;
}
return new_d;
}
void alias_siblings(struct device *d)
{
while (d) {
int link = 0;
struct device *cmp = d->next_sibling;
while (cmp && (cmp->parent == d->parent) && (cmp->path_a == d->path_a)
&& (cmp->path_b == d->path_b)) {
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
if (!cmp->used) {
if (device_match(d, cmp)) {
d->multidev = 1;
cmp->id = d->id;
cmp->name = d->name;
cmp->used = 1;
cmp->link = ++link;
}
}
cmp = cmp->next_sibling;
}
d = d->next_sibling;
}
}
void add_resource(struct device *dev, int type, int index, int base)
{
struct resource *r = S_ALLOC(sizeof(struct resource));
r->type = type;
r->index = index;
r->base = base;
if (dev->res) {
struct resource *head = dev->res;
while (head->next)
head = head->next;
head->next = r;
} else {
dev->res = r;
}
dev->rescnt++;
}
void add_register(struct chip_instance *chip_instance, char *name, char *val)
{
struct reg *r = S_ALLOC(sizeof(struct reg));
r->key = name;
r->value = val;
if (chip_instance->reg) {
struct reg *head = chip_instance->reg;
// sorting to be equal to sconfig's behaviour
int sort = strcmp(r->key, head->key);
if (sort == 0) {
printf("ERROR: duplicate 'register' key.\n");
exit(1);
}
if (sort < 0) {
r->next = head;
chip_instance->reg = r;
} else {
while ((head->next)
&& (strcmp(head->next->key, r->key) < 0))
head = head->next;
r->next = head->next;
head->next = r;
}
} else {
chip_instance->reg = r;
}
}
void add_pci_subsystem_ids(struct device *dev, int vendor, int device,
int inherit)
{
if (dev->bustype != PCI && dev->bustype != DOMAIN) {
printf("ERROR: 'subsystem' only allowed for PCI devices\n");
exit(1);
}
dev->subsystem_vendor = vendor;
dev->subsystem_device = device;
dev->inherit_subsystem = inherit;
}
void add_ioapic_info(struct device *dev, int apicid, const char *_srcpin,
int irqpin)
{
int srcpin;
if (!_srcpin || strlen(_srcpin) < 4 || strncasecmp(_srcpin, "INT", 3) ||
_srcpin[3] < 'A' || _srcpin[3] > 'D') {
printf("ERROR: malformed ioapic_irq args: %s\n", _srcpin);
exit(1);
}
srcpin = _srcpin[3] - 'A';
if (dev->bustype != PCI && dev->bustype != DOMAIN) {
printf("ERROR: ioapic config only allowed for PCI devices\n");
exit(1);
}
if (srcpin > 3) {
printf("ERROR: srcpin '%d' invalid\n", srcpin);
exit(1);
}
dev->pci_irq_info[srcpin].ioapic_irq_pin = irqpin;
dev->pci_irq_info[srcpin].ioapic_dst_id = apicid;
}
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
static void pass0(FILE *fil, struct device *ptr)
{
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
if (ptr->id == 0)
fprintf(fil, "DEVTREE_CONST struct bus %s_links[];\n",
ptr->name);
Make the device tree available in the rom stage We thought about two ways to do this change. The way we decided to try was to 1. drop all ops from devices in romstage 2. constify all devices in romstage (make them read-only) so we can compile static.c into romstage 3. the device tree "devices" can be used to read configuration from the device tree (and nothing else, really) 4. the device tree devices are accessed through struct device * in romstage only. device_t stays the typedef to int in romstage 5. Use the same static.c file in ramstage and romstage We declare structs as follows: ROMSTAGE_CONST struct bus dev_root_links[]; ROMSTAGE_CONST is const in romstage and empty in ramstage; This forces all of the device tree into the text area. So a struct looks like this: static ROMSTAGE_CONST struct device _dev21 = { #ifndef __PRE_RAM__ .ops = 0, #endif .bus = &_dev7_links[0], .path = {.type=DEVICE_PATH_PCI,{.pci={ .devfn = PCI_DEVFN(0x1c,3)}}}, .enabled = 0, .on_mainboard = 1, .subsystem_vendor = 0x1ae0, .subsystem_device = 0xc000, .link_list = NULL, .sibling = &_dev22, #ifndef __PRE_RAM__ .chip_ops = &southbridge_intel_bd82x6x_ops, #endif .chip_info = &southbridge_intel_bd82x6x_info_10, .next=&_dev22 }; Change-Id: I722454d8d3c40baf7df989f5a6891f6ba7db5727 Signed-off-by: Ronald G. Minnich <rminnich@chromium.org> Signed-off-by: Stefan Reinauer <reinauer@google.com> Reviewed-on: http://review.coreboot.org/1398 Tested-by: build bot (Jenkins) Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
2012-08-01 01:47:25 +02:00
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
if ((ptr->id != 0) && (!ptr->used)) {
fprintf(fil, "DEVTREE_CONST static struct device %s;\n",
ptr->name);
if (ptr->rescnt > 0)
fprintf(fil,
"DEVTREE_CONST struct resource %s_res[];\n",
ptr->name);
if (ptr->children || ptr->multidev)
fprintf(fil, "DEVTREE_CONST struct bus %s_links[];\n",
ptr->name);
}
}
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
static void pass1(FILE *fil, struct device *ptr)
{
int pin;
struct chip_instance *chip_ins = ptr->chip_instance;
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
if (!ptr->used) {
if (ptr->id != 0)
fprintf(fil, "static ");
fprintf(fil, "DEVTREE_CONST struct device %s = {\n",
ptr->name);
fprintf(fil, "#if !DEVTREE_EARLY\n");
fprintf(fil, "\t.ops = %s,\n", (ptr->ops) ? (ptr->ops) : "0");
Make the device tree available in the rom stage We thought about two ways to do this change. The way we decided to try was to 1. drop all ops from devices in romstage 2. constify all devices in romstage (make them read-only) so we can compile static.c into romstage 3. the device tree "devices" can be used to read configuration from the device tree (and nothing else, really) 4. the device tree devices are accessed through struct device * in romstage only. device_t stays the typedef to int in romstage 5. Use the same static.c file in ramstage and romstage We declare structs as follows: ROMSTAGE_CONST struct bus dev_root_links[]; ROMSTAGE_CONST is const in romstage and empty in ramstage; This forces all of the device tree into the text area. So a struct looks like this: static ROMSTAGE_CONST struct device _dev21 = { #ifndef __PRE_RAM__ .ops = 0, #endif .bus = &_dev7_links[0], .path = {.type=DEVICE_PATH_PCI,{.pci={ .devfn = PCI_DEVFN(0x1c,3)}}}, .enabled = 0, .on_mainboard = 1, .subsystem_vendor = 0x1ae0, .subsystem_device = 0xc000, .link_list = NULL, .sibling = &_dev22, #ifndef __PRE_RAM__ .chip_ops = &southbridge_intel_bd82x6x_ops, #endif .chip_info = &southbridge_intel_bd82x6x_info_10, .next=&_dev22 }; Change-Id: I722454d8d3c40baf7df989f5a6891f6ba7db5727 Signed-off-by: Ronald G. Minnich <rminnich@chromium.org> Signed-off-by: Stefan Reinauer <reinauer@google.com> Reviewed-on: http://review.coreboot.org/1398 Tested-by: build bot (Jenkins) Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
2012-08-01 01:47:25 +02:00
fprintf(fil, "#endif\n");
fprintf(fil, "\t.bus = &%s_links[%d],\n", ptr->parent->name,
ptr->parent->link);
fprintf(fil, "\t.path = {");
fprintf(fil, ptr->path, ptr->path_a, ptr->path_b);
fprintf(fil, "},\n");
fprintf(fil, "\t.enabled = %d,\n", ptr->enabled);
fprintf(fil, "\t.on_mainboard = 1,\n");
if (ptr->subsystem_vendor > 0)
fprintf(fil, "\t.subsystem_vendor = 0x%04x,\n",
ptr->subsystem_vendor);
for (pin = 0; pin < 4; pin++) {
if (ptr->pci_irq_info[pin].ioapic_irq_pin > 0)
fprintf(fil, "\t.pci_irq_info[%d].ioapic_irq_pin = %d,\n",
pin, ptr->pci_irq_info[pin].ioapic_irq_pin);
if (ptr->pci_irq_info[pin].ioapic_dst_id > 0)
fprintf(fil, "\t.pci_irq_info[%d].ioapic_dst_id = %d,\n",
pin, ptr->pci_irq_info[pin].ioapic_dst_id);
}
if (ptr->subsystem_device > 0)
fprintf(fil, "\t.subsystem_device = 0x%04x,\n",
ptr->subsystem_device);
if (ptr->rescnt > 0) {
fprintf(fil, "\t.resource_list = &%s_res[0],\n",
ptr->name);
}
if (ptr->children || ptr->multidev)
fprintf(fil, "\t.link_list = &%s_links[0],\n",
ptr->name);
else
fprintf(fil, "\t.link_list = NULL,\n");
if (ptr->sibling)
fprintf(fil, "\t.sibling = &%s,\n", ptr->sibling->name);
fprintf(fil, "#if !DEVTREE_EARLY\n");
fprintf(fil, "\t.chip_ops = &%s_ops,\n",
chip_ins->chip->name_underscore);
if (chip_ins == &mainboard_instance)
fprintf(fil, "\t.name = mainboard_name,\n");
fprintf(fil, "#endif\n");
if (chip_ins->chip->chiph_exists)
fprintf(fil, "\t.chip_info = &%s_info_%d,\n",
chip_ins->chip->name_underscore, chip_ins->id);
if (ptr->next)
fprintf(fil, "\t.next=&%s\n", ptr->next->name);
fprintf(fil, "};\n");
}
if (ptr->rescnt > 0) {
int i = 1;
fprintf(fil, "DEVTREE_CONST struct resource %s_res[] = {\n",
ptr->name);
struct resource *r = ptr->res;
while (r) {
fprintf(fil,
"\t\t{ .flags=IORESOURCE_FIXED | IORESOURCE_ASSIGNED | IORESOURCE_");
if (r->type == IRQ)
fprintf(fil, "IRQ");
if (r->type == DRQ)
fprintf(fil, "DRQ");
if (r->type == IO)
fprintf(fil, "IO");
fprintf(fil, ", .index=0x%x, .base=0x%x,", r->index,
r->base);
if (r->next)
fprintf(fil, ".next=&%s_res[%d]},\n", ptr->name,
i++);
else
fprintf(fil, ".next=NULL },\n");
r = r->next;
}
fprintf(fil, "\t };\n");
}
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
if (!ptr->used && (ptr->children || ptr->multidev)) {
fprintf(fil, "DEVTREE_CONST struct bus %s_links[] = {\n",
ptr->name);
if (ptr->multidev) {
struct device *d = ptr;
while (d) {
if (device_match(d, ptr)) {
fprintf(fil, "\t\t[%d] = {\n", d->link);
fprintf(fil, "\t\t\t.link_num = %d,\n",
d->link);
fprintf(fil, "\t\t\t.dev = &%s,\n",
d->name);
if (d->children)
fprintf(fil,
"\t\t\t.children = &%s,\n",
d->children->name);
if (d->next_sibling
&& device_match(d->next_sibling,
ptr))
fprintf(fil,
"\t\t\t.next=&%s_links[%d],\n",
d->name, d->link + 1);
else
fprintf(fil,
"\t\t\t.next = NULL,\n");
fprintf(fil, "\t\t},\n");
}
d = d->next_sibling;
}
} else {
if (ptr->children) {
fprintf(fil, "\t\t[0] = {\n");
fprintf(fil, "\t\t\t.link_num = 0,\n");
fprintf(fil, "\t\t\t.dev = &%s,\n", ptr->name);
fprintf(fil, "\t\t\t.children = &%s,\n",
ptr->children->name);
fprintf(fil, "\t\t\t.next = NULL,\n");
fprintf(fil, "\t\t},\n");
}
}
fprintf(fil, "\t};\n");
}
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
}
static void walk_device_tree(FILE *fil, struct device *ptr,
void (*func)(FILE *, struct device *),
struct device *chips)
{
do {
func(fil, ptr);
ptr = ptr->next_sibling;
} while (ptr);
}
static void emit_chip_headers(FILE *fil, struct chip *chip)
{
struct chip *tmp = chip;
fprintf(fil, "#include <device/device.h>\n");
fprintf(fil, "#include <device/pci.h>\n");
while (chip) {
if (chip->chiph_exists)
fprintf(fil, "#include \"%s/chip.h\"\n", chip->name);
chip = chip->next;
}
fprintf(fil, "\n#if !DEVTREE_EARLY\n");
fprintf(fil,
"__attribute__((weak)) struct chip_operations mainboard_ops = {};\n");
chip = tmp;
while (chip) {
fprintf(fil,
"__attribute__((weak)) struct chip_operations %s_ops = {};\n",
chip->name_underscore);
chip = chip->next;
}
fprintf(fil, "#endif\n");
}
static void emit_chip_instance(FILE *fil, struct chip_instance *instance)
{
fprintf(fil, "DEVTREE_CONST struct %s_config %s_info_%d = {",
instance->chip->name_underscore,
instance->chip->name_underscore,
instance->id);
if (instance->reg) {
fprintf(fil, "\n");
struct reg *r = instance->reg;
while (r) {
fprintf(fil, "\t.%s = %s,\n", r->key, r->value);
r = r->next;
}
}
fprintf(fil, "};\n\n");
}
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
static void emit_chips(FILE *fil)
{
struct chip *chip = chip_header.next;
struct chip_instance *instance;
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
emit_chip_headers(fil, chip);
for (; chip; chip = chip->next) {
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
if (!chip->chiph_exists)
continue;
instance = chip->instance;
while (instance) {
emit_chip_instance(fil, instance);
instance = instance->next;
}
}
}
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
static void inherit_subsystem_ids(FILE *file, struct device *dev)
{
struct device *p;
if (dev->subsystem_vendor != -1 && dev->subsystem_device != -1) {
/* user already gave us a subsystem vendor/device */
return;
}
for (p = dev; p && p != p->parent; p = p->parent) {
if (p->bustype != PCI && p->bustype != DOMAIN)
continue;
if (p->inherit_subsystem) {
dev->subsystem_vendor = p->subsystem_vendor;
dev->subsystem_device = p->subsystem_device;
break;
}
}
}
static void usage(void)
{
printf("usage: sconfig devicetree_file output_file\n");
exit(1);
}
enum {
DEVICEFILE_ARG = 1,
OUTPUTFILE_ARG
};
#define ARG_COUNT 3
int main(int argc, char **argv)
{
if (argc != ARG_COUNT)
usage();
char *devtree = argv[DEVICEFILE_ARG];
char *outputc = argv[OUTPUTFILE_ARG];
FILE *filec = fopen(devtree, "r");
if (!filec) {
perror(NULL);
exit(1);
}
yyrestart(filec);
lastnode = head = &root;
yyparse();
fclose(filec);
FILE *autogen = fopen(outputc, "w");
if (!autogen) {
fprintf(stderr, "Could not open file '%s' for writing: ",
outputc);
perror(NULL);
exit(1);
}
util/sconfig: Re-factor sconfig to not assume chip as device This change adds a new structure "struct chip" to identify elements of type chip rather than re-using the structure for device. Until now chip was treated as a device while generating the parse tree and then device tree postprocessing skipped over all the chip entries in children and sibling pointers of device nodes. With this change, the device tree will only contain struct device in the parsed tree. It helps by avoiding unnecessary pointers to chip structure as children or next_sibling and then skipping those elements in post processing. Every device can then hold a pointer to its chip. When generating static.c, chip structure is emitted before device structure to ensure that the device structure has chip within its scope. Externally, the only visible change in static.c should be the order in which chip/device elements are emitted i.e. previously all chips under a particular device were emitted to static.c and then the devices using those chips. Now, all chips are emitted before all the devices in static.c BUG=b:80081934 TEST=Verified that abuild is successful for all boards. Also, verified that static.c generated for eve, kahlee, scarlet, asrock imb_a180 is unchanged from before in node definitions. Change-Id: I255092f527c8eecb144385eb681df20e54caf8f5 Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/26720 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2018-05-31 00:09:09 +02:00
emit_chips(autogen);
walk_device_tree(autogen, &root, inherit_subsystem_ids, NULL);
fprintf(autogen, "\n/* pass 0 */\n");
walk_device_tree(autogen, &root, pass0, NULL);
fprintf(autogen, "\n/* pass 1 */\n"
"DEVTREE_CONST struct device * DEVTREE_CONST last_dev = &%s;\n",
lastnode->name);
walk_device_tree(autogen, &root, pass1, NULL);
fclose(autogen);
return 0;
}