/*
* Copyright (c) 2006, 2018, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code 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
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
#include "logging/log.hpp"
#include "precompiled.hpp"
#include "runtime/os.hpp"
#include "vm_version_sparc.hpp"
#define CPUINFO_LINE_SIZE 1024
class CPUinfo {
public:
CPUinfo(const char* field) : _string(NULL) {
char line[CPUINFO_LINE_SIZE];
FILE* fp = fopen("/proc/cpuinfo", "r");
if (fp != NULL) {
while (fgets(line, sizeof(line), fp) != NULL) {
assert(strlen(line) < sizeof(line) - 1,
"buffer too small (%d)", CPUINFO_LINE_SIZE);
const char* vstr = match_field(line, field);
if (vstr != NULL) {
// We have a matching line and a valid starting point to the value of
// the field, copy the string for keeps.
_string = strdup(vstr);
break;
}
}
fclose(fp);
}
}
~CPUinfo() { free((void*)_string); }
const char* value() const { return _string; }
bool valid() const { return _string != NULL; }
bool match(const char* s) const {
return valid() ? strcmp(_string, s) == 0 : false;
}
private:
const char* _string;
const char* match_field(char line[CPUINFO_LINE_SIZE], const char* field);
const char* match_alo(const char* text, const char* exp);
const char* match_seq(const char* text, const char* seq);
};
/* Given a line of text read from /proc/cpuinfo, determine if the property header
* matches the field specified, according to the following regexp: "<field>"\W+:\W+
*
* If we have a matching expression, return a pointer to the first character after
* the matching pattern, i.e. the "value", otherwise return NULL.
*/
const char* CPUinfo::match_field(char line[CPUINFO_LINE_SIZE], const char* field) {
return match_alo(match_seq(match_alo(match_seq(line, field), "\t "), ":"), "\t ");
}
/* Match a sequence of at-least-one character in the string expression (exp) to
* the text input.
*/
const char* CPUinfo::match_alo(const char* text, const char* exp) {
if (text == NULL) return NULL;
const char* chp;
for (chp = &text[0]; *chp != '\0'; chp++) {
if (strchr(exp, *chp) == NULL) break;
}
return text < chp ? chp : NULL;
}
/* Match an exact sequence of characters as specified by the string expression
* (seq) to the text input.
*/
const char* CPUinfo::match_seq(const char* text, const char* seq) {
if (text == NULL) return NULL;
while (*seq != '\0') {
if (*seq != *text++) break; else seq++;
}
return *seq == '\0' ? text : NULL;
}
typedef struct {
const uint32_t hash;
bool seen;
const char* const name;
const uint64_t mask;
} FeatureEntry;
static uint64_t parse_features(FeatureEntry feature_tbl[], const char input[]);
void VM_Version::platform_features() {
// Some of the features reported via "cpucaps", such as; 'flush', 'stbar',
// 'swap', 'muldiv', 'ultra3', 'blkinit', 'n2', 'mul32', 'div32', 'fsmuld'
// and 'v8plus', are either SPARC V8, supported by all HW or simply nonsense
// (the 'ultra3' "property").
//
// Entries marked as 'NYI' are not yet supported via "cpucaps" but are
// expected to have the names used in the table below (these are SPARC M7
// features or more recent).
//
// NOTE: Table sorted on lookup/hash ID.
static FeatureEntry s_feature_tbl[] = {
{ 0x006f, false, "v9", ISA_v9_msk }, // Mandatory
{ 0x00a6, false, "md5", ISA_md5_msk },
{ 0x00ce, false, "adi", ISA_adi_msk }, // NYI
{ 0x00d7, false, "ima", ISA_ima_msk },
{ 0x00d9, false, "aes", ISA_aes_msk },
{ 0x00db, false, "hpc", ISA_hpc_msk },
{ 0x00dc, false, "des", ISA_des_msk },
{ 0x00ed, false, "sha1", ISA_sha1_msk },
{ 0x00f2, false, "vis", ISA_vis1_msk },
{ 0x0104, false, "vis2", ISA_vis2_msk },
{ 0x0105, false, "vis3", ISA_vis3_msk },
{ 0x0114, false, "sha512", ISA_sha512_msk },
{ 0x0119, false, "sha256", ISA_sha256_msk },
{ 0x011a, false, "fmaf", ISA_fmaf_msk },
{ 0x0132, false, "popc", ISA_popc_msk },
{ 0x0140, false, "crc32c", ISA_crc32c_msk },
{ 0x0147, false, "vis3b", ISA_vis3b_msk }, // NYI
{ 0x017e, false, "pause", ISA_pause_msk },
{ 0x0182, false, "mwait", ISA_mwait_msk }, // NYI
{ 0x018b, false, "mpmul", ISA_mpmul_msk },
{ 0x018e, false, "sparc5", ISA_sparc5_msk }, // NYI
{ 0x01a9, false, "cbcond", ISA_cbcond_msk },
{ 0x01c3, false, "vamask", ISA_vamask_msk }, // NYI
{ 0x01ca, false, "kasumi", ISA_kasumi_msk },
{ 0x01e3, false, "xmpmul", ISA_xmpmul_msk }, // NYI
{ 0x022c, false, "montmul", ISA_mont_msk },
{ 0x0234, false, "montsqr", ISA_mont_msk },
{ 0x0238, false, "camellia", ISA_camellia_msk },
{ 0x024a, false, "ASIBlkInit", ISA_blk_init_msk },
{ 0x0284, false, "xmontmul", ISA_xmont_msk }, // NYI
{ 0x02e6, false, "pause_nsec", ISA_pause_nsec_msk }, // NYI
{ 0x0000, false, NULL, 0 }
};
CPUinfo caps("cpucaps"); // Read "cpucaps" from /proc/cpuinfo.
assert(caps.valid(), "must be");
_features = parse_features(s_feature_tbl, caps.value());
assert(has_v9(), "must be"); // Basic SPARC-V9 required (V8 not supported).
CPUinfo type("type");
bool is_sun4v = type.match("sun4v"); // All Oracle SPARC + Fujitsu Athena+
bool is_sun4u = type.match("sun4u"); // All other Fujitsu
uint64_t synthetic = 0;
if (is_sun4v) {
// Indirect and direct branches are equally fast.
synthetic = CPU_fast_ind_br_msk;
// Fast IDIV, BIS and LD available on Niagara Plus.
if (has_vis2()) {
synthetic |= (CPU_fast_idiv_msk | CPU_fast_ld_msk);
// ...on Core C4 however, we prefer not to use BIS.
if (!has_sparc5()) {
synthetic |= CPU_fast_bis_msk;
}
}
// Niagara Core C3 supports fast RDPC and block zeroing.
if (has_ima()) {
synthetic |= (CPU_fast_rdpc_msk | CPU_blk_zeroing_msk);
}
// Niagara Core C3 and C4 have slow CMOVE.
if (!has_ima()) {
synthetic |= CPU_fast_cmove_msk;
}
} else if (is_sun4u) {
// SPARC64 only have fast IDIV and RDPC.
synthetic |= (CPU_fast_idiv_msk | CPU_fast_rdpc_msk);
} else {
log_info(os, cpu)("Unable to derive CPU features: %s", type.value());
}
_features += synthetic; // Including CPU derived/synthetic features.
}
////////////////////////////////////////////////////////////////////////////////
static uint32_t uhash32(const char name[]);
static void update_table(FeatureEntry feature_tbl[], uint32_t hv,
const char* ch1p,
const char* endp);
/* Given a feature table, parse the input text holding the string value of
* 'cpucaps' as reported by '/proc/cpuinfo', in order to complete the table
* with information on each admissible feature (whether present or not).
*
* Return the composite bit-mask representing the features found.
*/
static uint64_t parse_features(FeatureEntry feature_tbl[], const char input[]) {
log_info(os, cpu)("Parse CPU features: %s\n", input);
#ifdef ASSERT
// Verify that hash value entries in the table are unique and ordered.
uint32_t prev = 0;
for (uint k = 0; feature_tbl[k].name != NULL; k++) {
feature_tbl[k].seen = false;
assert(feature_tbl[k].hash == uhash32(feature_tbl[k].name),
"feature '%s' has mismatching hash 0x%08x (expected 0x%08x).\n",
feature_tbl[k].name,
feature_tbl[k].hash,
uhash32(feature_tbl[k].name));
assert(prev < feature_tbl[k].hash,
"feature '%s' has invalid hash 0x%08x (previous is 0x%08x).\n",
feature_tbl[k].name,
feature_tbl[k].hash,
prev);
prev = feature_tbl[k].hash;
}
#endif
// Identify features from the input, consisting of a string with features
// separated by commas (or whitespace), e.g. "flush,muldiv,v9,mul32,div32,
// v8plus,popc,vis".
uint32_t hv = 0;
const char* ch1p = &input[0];
uint i = 0;
do {
char ch = input[i];
if (isalnum(ch) || ch == '_') {
hv += (ch - 32u);
}
else if (isspace(ch) || ch == ',' || ch == '\0') { // end-of-token
if (ch1p < &input[i]) {
update_table(feature_tbl, hv, ch1p, &input[i]);
}
ch1p = &input[i + 1]; hv = 0;
} else {
// Handle non-accepted input robustly.
log_info(os, cpu)("Bad token in feature string: '%c' (0x%02x).\n", ch, ch);
ch1p = &input[i + 1]; hv = 0;
}
}
while (input[i++] != '\0');
// Compute actual bit-mask representation.
uint64_t mask = 0;
for (uint k = 0; feature_tbl[k].name != NULL; k++) {
mask |= feature_tbl[k].seen ? feature_tbl[k].mask : 0;
}
return mask;
}
static uint32_t uhash32(const char name[]) {
uint32_t hv = 0;
for (uint i = 0; name[i] != '\0'; i++) {
hv += (name[i] - 32u);
}
return hv;
}
static bool verify_match(const char name[], const char* ch1p, const char* endp);
static void update_table(FeatureEntry feature_tbl[], uint32_t hv, const char* ch1p, const char* endp) {
assert(ch1p < endp, "at least one character");
// Look for a hash value in the table. Since this table is a small one (and
// is expected to stay small), we use a simple linear search (iff the table
// grows large, we may consider to adopt a binary ditto, or a perfect hash).
for (uint k = 0; feature_tbl[k].name != NULL; k++) {
uint32_t hash = feature_tbl[k].hash;
if (hash < hv) continue;
if (hash == hv) {
const char* name = feature_tbl[k].name;
if (verify_match(name, ch1p, endp)) {
feature_tbl[k].seen = true;
break;
}
}
// Either a non-matching feature (when hash == hv) or hash > hv. In either
// case we break out of the loop and terminate the search (note that the
// table is assumed to be uniquely sorted on the hash).
break;
}
}
static bool verify_match(const char name[], const char* ch1p, const char* endp) {
size_t len = strlen(name);
if (len != static_cast<size_t>(endp - ch1p)) {
return false;
}
for (uint i = 0; ch1p + i < endp; i++) {
if (name[i] != ch1p[i]) return false;
}
return true;
}