From 5cbc30737398b49f62ae8603129ce43ac7db1a41 Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Fri, 25 May 2007 15:49:59 -0700 Subject: [SPARC64]: Use machine description and OBP properly for cpu probing. Signed-off-by: David S. Miller --- arch/sparc64/kernel/mdesc.c | 619 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 619 insertions(+) create mode 100644 arch/sparc64/kernel/mdesc.c (limited to 'arch/sparc64/kernel/mdesc.c') diff --git a/arch/sparc64/kernel/mdesc.c b/arch/sparc64/kernel/mdesc.c new file mode 100644 index 000000000000..9246c2cf9574 --- /dev/null +++ b/arch/sparc64/kernel/mdesc.c @@ -0,0 +1,619 @@ +/* mdesc.c: Sun4V machine description handling. + * + * Copyright (C) 2007 David S. Miller + */ +#include +#include +#include +#include + +#include +#include +#include +#include +#include + +/* Unlike the OBP device tree, the machine description is a full-on + * DAG. An arbitrary number of ARCs are possible from one + * node to other nodes and thus we can't use the OBP device_node + * data structure to represent these nodes inside of the kernel. + * + * Actually, it isn't even a DAG, because there are back pointers + * which create cycles in the graph. + * + * mdesc_hdr and mdesc_elem describe the layout of the data structure + * we get from the Hypervisor. + */ +struct mdesc_hdr { + u32 version; /* Transport version */ + u32 node_sz; /* node block size */ + u32 name_sz; /* name block size */ + u32 data_sz; /* data block size */ +}; + +struct mdesc_elem { + u8 tag; +#define MD_LIST_END 0x00 +#define MD_NODE 0x4e +#define MD_NODE_END 0x45 +#define MD_NOOP 0x20 +#define MD_PROP_ARC 0x61 +#define MD_PROP_VAL 0x76 +#define MD_PROP_STR 0x73 +#define MD_PROP_DATA 0x64 + u8 name_len; + u16 resv; + u32 name_offset; + union { + struct { + u32 data_len; + u32 data_offset; + } data; + u64 val; + } d; +}; + +static struct mdesc_hdr *main_mdesc; +static struct mdesc_node *allnodes; + +static struct mdesc_node *allnodes_tail; +static unsigned int unique_id; + +static struct mdesc_node **mdesc_hash; +static unsigned int mdesc_hash_size; + +static inline unsigned int node_hashfn(u64 node) +{ + return ((unsigned int) (node ^ (node >> 8) ^ (node >> 16))) + & (mdesc_hash_size - 1); +} + +static inline void hash_node(struct mdesc_node *mp) +{ + struct mdesc_node **head = &mdesc_hash[node_hashfn(mp->node)]; + + mp->hash_next = *head; + *head = mp; + + if (allnodes_tail) { + allnodes_tail->allnodes_next = mp; + allnodes_tail = mp; + } else { + allnodes = allnodes_tail = mp; + } +} + +static struct mdesc_node *find_node(u64 node) +{ + struct mdesc_node *mp = mdesc_hash[node_hashfn(node)]; + + while (mp) { + if (mp->node == node) + return mp; + + mp = mp->hash_next; + } + return NULL; +} + +struct property *md_find_property(const struct mdesc_node *mp, + const char *name, + int *lenp) +{ + struct property *pp; + + for (pp = mp->properties; pp != 0; pp = pp->next) { + if (strcasecmp(pp->name, name) == 0) { + if (lenp) + *lenp = pp->length; + break; + } + } + return pp; +} +EXPORT_SYMBOL(md_find_property); + +/* + * Find a property with a given name for a given node + * and return the value. + */ +const void *md_get_property(const struct mdesc_node *mp, const char *name, + int *lenp) +{ + struct property *pp = md_find_property(mp, name, lenp); + return pp ? pp->value : NULL; +} +EXPORT_SYMBOL(md_get_property); + +struct mdesc_node *md_find_node_by_name(struct mdesc_node *from, + const char *name) +{ + struct mdesc_node *mp; + + mp = from ? from->allnodes_next : allnodes; + for (; mp != NULL; mp = mp->allnodes_next) { + if (strcmp(mp->name, name) == 0) + break; + } + return mp; +} +EXPORT_SYMBOL(md_find_node_by_name); + +static unsigned int mdesc_early_allocated; + +static void * __init mdesc_early_alloc(unsigned long size) +{ + void *ret; + + ret = __alloc_bootmem(size, SMP_CACHE_BYTES, 0UL); + if (ret == NULL) { + prom_printf("MDESC: alloc of %lu bytes failed.\n", size); + prom_halt(); + } + + memset(ret, 0, size); + + mdesc_early_allocated += size; + + return ret; +} + +static unsigned int __init count_arcs(struct mdesc_elem *ep) +{ + unsigned int ret = 0; + + ep++; + while (ep->tag != MD_NODE_END) { + if (ep->tag == MD_PROP_ARC) + ret++; + ep++; + } + return ret; +} + +static void __init mdesc_node_alloc(u64 node, struct mdesc_elem *ep, const char *names) +{ + unsigned int num_arcs = count_arcs(ep); + struct mdesc_node *mp; + + mp = mdesc_early_alloc(sizeof(*mp) + + (num_arcs * sizeof(struct mdesc_arc))); + mp->name = names + ep->name_offset; + mp->node = node; + mp->unique_id = unique_id++; + mp->num_arcs = num_arcs; + + hash_node(mp); +} + +static inline struct mdesc_elem *node_block(struct mdesc_hdr *mdesc) +{ + return (struct mdesc_elem *) (mdesc + 1); +} + +static inline void *name_block(struct mdesc_hdr *mdesc) +{ + return ((void *) node_block(mdesc)) + mdesc->node_sz; +} + +static inline void *data_block(struct mdesc_hdr *mdesc) +{ + return ((void *) name_block(mdesc)) + mdesc->name_sz; +} + +/* In order to avoid recursion (the graph can be very deep) we use a + * two pass algorithm. First we allocate all the nodes and hash them. + * Then we iterate over each node, filling in the arcs and properties. + */ +static void __init build_all_nodes(struct mdesc_hdr *mdesc) +{ + struct mdesc_elem *start, *ep; + struct mdesc_node *mp; + const char *names; + void *data; + u64 last_node; + + start = ep = node_block(mdesc); + last_node = mdesc->node_sz / 16; + + names = name_block(mdesc); + + while (1) { + u64 node = ep - start; + + if (ep->tag == MD_LIST_END) + break; + + if (ep->tag != MD_NODE) { + prom_printf("MDESC: Inconsistent element list.\n"); + prom_halt(); + } + + mdesc_node_alloc(node, ep, names); + + if (ep->d.val >= last_node) { + printk("MDESC: Warning, early break out of node scan.\n"); + printk("MDESC: Next node [%lu] last_node [%lu].\n", + node, last_node); + break; + } + + ep = start + ep->d.val; + } + + data = data_block(mdesc); + for (mp = allnodes; mp; mp = mp->allnodes_next) { + struct mdesc_elem *ep = start + mp->node; + struct property **link = &mp->properties; + unsigned int this_arc = 0; + + ep++; + while (ep->tag != MD_NODE_END) { + switch (ep->tag) { + case MD_PROP_ARC: { + struct mdesc_node *target; + + if (this_arc >= mp->num_arcs) { + prom_printf("MDESC: ARC overrun [%u:%u]\n", + this_arc, mp->num_arcs); + prom_halt(); + } + target = find_node(ep->d.val); + if (!target) { + printk("MDESC: Warning, arc points to " + "missing node, ignoring.\n"); + break; + } + mp->arcs[this_arc].name = + (names + ep->name_offset); + mp->arcs[this_arc].arc = target; + this_arc++; + break; + } + + case MD_PROP_VAL: + case MD_PROP_STR: + case MD_PROP_DATA: { + struct property *p = mdesc_early_alloc(sizeof(*p)); + + p->unique_id = unique_id++; + p->name = (char *) names + ep->name_offset; + if (ep->tag == MD_PROP_VAL) { + p->value = &ep->d.val; + p->length = 8; + } else { + p->value = data + ep->d.data.data_offset; + p->length = ep->d.data.data_len; + } + *link = p; + link = &p->next; + break; + } + + case MD_NOOP: + break; + + default: + printk("MDESC: Warning, ignoring unknown tag type %02x\n", + ep->tag); + } + ep++; + } + } +} + +static unsigned int __init count_nodes(struct mdesc_hdr *mdesc) +{ + struct mdesc_elem *ep = node_block(mdesc); + struct mdesc_elem *end; + unsigned int cnt = 0; + + end = ((void *)ep) + mdesc->node_sz; + while (ep < end) { + if (ep->tag == MD_NODE) + cnt++; + ep++; + } + return cnt; +} + +static void __init report_platform_properties(void) +{ + struct mdesc_node *pn = md_find_node_by_name(NULL, "platform"); + const char *s; + const u64 *v; + + if (!pn) { + prom_printf("No platform node in machine-description.\n"); + prom_halt(); + } + + s = md_get_property(pn, "banner-name", NULL); + printk("PLATFORM: banner-name [%s]\n", s); + s = md_get_property(pn, "name", NULL); + printk("PLATFORM: name [%s]\n", s); + + v = md_get_property(pn, "hostid", NULL); + if (v) + printk("PLATFORM: hostid [%08lx]\n", *v); + v = md_get_property(pn, "serial#", NULL); + if (v) + printk("PLATFORM: serial# [%08lx]\n", *v); + v = md_get_property(pn, "stick-frequency", NULL); + printk("PLATFORM: stick-frequency [%08lx]\n", *v); + v = md_get_property(pn, "mac-address", NULL); + if (v) + printk("PLATFORM: mac-address [%lx]\n", *v); + v = md_get_property(pn, "watchdog-resolution", NULL); + if (v) + printk("PLATFORM: watchdog-resolution [%lu ms]\n", *v); + v = md_get_property(pn, "watchdog-max-timeout", NULL); + if (v) + printk("PLATFORM: watchdog-max-timeout [%lu ms]\n", *v); + v = md_get_property(pn, "max-cpus", NULL); + if (v) + printk("PLATFORM: max-cpus [%lu]\n", *v); +} + +static int inline find_in_proplist(const char *list, const char *match, int len) +{ + while (len > 0) { + int l; + + if (!strcmp(list, match)) + return 1; + l = strlen(list) + 1; + list += l; + len -= l; + } + return 0; +} + +static void __init fill_in_one_cache(cpuinfo_sparc *c, struct mdesc_node *mp) +{ + const u64 *level = md_get_property(mp, "level", NULL); + const u64 *size = md_get_property(mp, "size", NULL); + const u64 *line_size = md_get_property(mp, "line-size", NULL); + const char *type; + int type_len; + + type = md_get_property(mp, "type", &type_len); + + switch (*level) { + case 1: + if (find_in_proplist(type, "instn", type_len)) { + c->icache_size = *size; + c->icache_line_size = *line_size; + } else if (find_in_proplist(type, "data", type_len)) { + c->dcache_size = *size; + c->dcache_line_size = *line_size; + } + break; + + case 2: + c->ecache_size = *size; + c->ecache_line_size = *line_size; + break; + + default: + break; + } + + if (*level == 1) { + unsigned int i; + + for (i = 0; i < mp->num_arcs; i++) { + struct mdesc_node *t = mp->arcs[i].arc; + + if (strcmp(mp->arcs[i].name, "fwd")) + continue; + + if (!strcmp(t->name, "cache")) + fill_in_one_cache(c, t); + } + } +} + +static void __init mark_core_ids(struct mdesc_node *mp, int core_id) +{ + unsigned int i; + + for (i = 0; i < mp->num_arcs; i++) { + struct mdesc_node *t = mp->arcs[i].arc; + const u64 *id; + + if (strcmp(mp->arcs[i].name, "back")) + continue; + + if (!strcmp(t->name, "cpu")) { + id = md_get_property(t, "id", NULL); + if (*id < NR_CPUS) + cpu_data(*id).core_id = core_id; + } else { + unsigned int j; + + for (j = 0; j < t->num_arcs; j++) { + struct mdesc_node *n = t->arcs[j].arc; + + if (strcmp(t->arcs[j].name, "back")) + continue; + + if (strcmp(n->name, "cpu")) + continue; + + id = md_get_property(n, "id", NULL); + if (*id < NR_CPUS) + cpu_data(*id).core_id = core_id; + } + } + } +} + +static void __init set_core_ids(void) +{ + struct mdesc_node *mp; + int idx; + + idx = 1; + md_for_each_node_by_name(mp, "cache") { + const u64 *level = md_get_property(mp, "level", NULL); + const char *type; + int len; + + if (*level != 1) + continue; + + type = md_get_property(mp, "type", &len); + if (!find_in_proplist(type, "instn", len)) + continue; + + mark_core_ids(mp, idx); + + idx++; + } +} + +static void __init get_one_mondo_bits(const u64 *p, unsigned int *mask, unsigned char def) +{ + u64 val; + + if (!p) + goto use_default; + val = *p; + + if (!val || val >= 64) + goto use_default; + + *mask = ((1U << val) * 64U) - 1U; + return; + +use_default: + *mask = ((1U << def) * 64U) - 1U; +} + +static void __init get_mondo_data(struct mdesc_node *mp, struct trap_per_cpu *tb) +{ + const u64 *val; + + val = md_get_property(mp, "q-cpu-mondo-#bits", NULL); + get_one_mondo_bits(val, &tb->cpu_mondo_qmask, 7); + + val = md_get_property(mp, "q-dev-mondo-#bits", NULL); + get_one_mondo_bits(val, &tb->dev_mondo_qmask, 7); + + val = md_get_property(mp, "q-resumable-#bits", NULL); + get_one_mondo_bits(val, &tb->resum_qmask, 6); + + val = md_get_property(mp, "q-nonresumable-#bits", NULL); + get_one_mondo_bits(val, &tb->nonresum_qmask, 2); +} + +static void __init mdesc_fill_in_cpu_data(void) +{ + struct mdesc_node *mp; + + ncpus_probed = 0; + md_for_each_node_by_name(mp, "cpu") { + const u64 *id = md_get_property(mp, "id", NULL); + const u64 *cfreq = md_get_property(mp, "clock-frequency", NULL); + struct trap_per_cpu *tb; + cpuinfo_sparc *c; + unsigned int i; + int cpuid; + + ncpus_probed++; + + cpuid = *id; + +#ifdef CONFIG_SMP + if (cpuid >= NR_CPUS) + continue; +#else + /* On uniprocessor we only want the values for the + * real physical cpu the kernel booted onto, however + * cpu_data() only has one entry at index 0. + */ + if (cpuid != real_hard_smp_processor_id()) + continue; + cpuid = 0; +#endif + + c = &cpu_data(cpuid); + c->clock_tick = *cfreq; + + tb = &trap_block[cpuid]; + get_mondo_data(mp, tb); + + for (i = 0; i < mp->num_arcs; i++) { + struct mdesc_node *t = mp->arcs[i].arc; + unsigned int j; + + if (strcmp(mp->arcs[i].name, "fwd")) + continue; + + if (!strcmp(t->name, "cache")) { + fill_in_one_cache(c, t); + continue; + } + + for (j = 0; j < t->num_arcs; j++) { + struct mdesc_node *n; + + n = t->arcs[j].arc; + if (strcmp(t->arcs[j].name, "fwd")) + continue; + + if (!strcmp(n->name, "cache")) + fill_in_one_cache(c, n); + } + } + +#ifdef CONFIG_SMP + cpu_set(cpuid, cpu_present_map); + cpu_set(cpuid, phys_cpu_present_map); +#endif + + c->core_id = 0; + } + + set_core_ids(); + + smp_fill_in_sib_core_maps(); +} + +void __init sun4v_mdesc_init(void) +{ + unsigned long len, real_len, status; + + (void) sun4v_mach_desc(0UL, 0UL, &len); + + printk("MDESC: Size is %lu bytes.\n", len); + + main_mdesc = mdesc_early_alloc(len); + + status = sun4v_mach_desc(__pa(main_mdesc), len, &real_len); + if (status != HV_EOK || real_len > len) { + prom_printf("sun4v_mach_desc fails, err(%lu), " + "len(%lu), real_len(%lu)\n", + status, len, real_len); + prom_halt(); + } + + len = count_nodes(main_mdesc); + printk("MDESC: %lu nodes.\n", len); + + len = roundup_pow_of_two(len); + + mdesc_hash = mdesc_early_alloc(len * sizeof(struct mdesc_node *)); + mdesc_hash_size = len; + + printk("MDESC: Hash size %lu entries.\n", len); + + build_all_nodes(main_mdesc); + + printk("MDESC: Built graph with %u bytes of memory.\n", + mdesc_early_allocated); + + report_platform_properties(); + mdesc_fill_in_cpu_data(); +} -- cgit v1.2.2