mirror of
https://github.com/jart/cosmopolitan.git
synced 2025-01-31 03:27:39 +00:00
9f9b240f3d
ParseHttpMessage() now does a better job avoiding malloc() calls and headers commonly sent by CloudFlare can now be consulted in constant time using our hard-coded perfect hash table. This increases /claim performance from 321k to 337k qps which is 5% faster.
277 lines
9.9 KiB
C
277 lines
9.9 KiB
C
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
|
|
│vi: set net ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi│
|
|
╞══════════════════════════════════════════════════════════════════════════════╡
|
|
│ Copyright 2020 Justine Alexandra Roberts Tunney │
|
|
│ │
|
|
│ Permission to use, copy, modify, and/or distribute this software for │
|
|
│ any purpose with or without fee is hereby granted, provided that the │
|
|
│ above copyright notice and this permission notice appear in all copies. │
|
|
│ │
|
|
│ THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL │
|
|
│ WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED │
|
|
│ WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE │
|
|
│ AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL │
|
|
│ DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR │
|
|
│ PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER │
|
|
│ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR │
|
|
│ PERFORMANCE OF THIS SOFTWARE. │
|
|
╚─────────────────────────────────────────────────────────────────────────────*/
|
|
#include "libc/assert.h"
|
|
#include "libc/intrin/bits.h"
|
|
#include "libc/limits.h"
|
|
#include "libc/macros.internal.h"
|
|
#include "libc/mem/alg.h"
|
|
#include "libc/mem/arraylist.internal.h"
|
|
#include "libc/mem/mem.h"
|
|
#include "libc/stdio/stdio.h"
|
|
#include "libc/str/str.h"
|
|
#include "libc/sysv/errfuns.h"
|
|
#include "libc/x/x.h"
|
|
#include "net/http/http.h"
|
|
|
|
#define LIMIT (SHRT_MAX - 2)
|
|
|
|
/**
|
|
* Initializes HTTP message parser.
|
|
*/
|
|
void InitHttpMessage(struct HttpMessage *r, int type) {
|
|
assert(type == kHttpRequest || type == kHttpResponse);
|
|
bzero(r, sizeof(*r));
|
|
r->type = type;
|
|
}
|
|
|
|
/**
|
|
* Destroys HTTP message parser.
|
|
*/
|
|
void DestroyHttpMessage(struct HttpMessage *r) {
|
|
if (r->xheaders.p) {
|
|
free(r->xheaders.p);
|
|
r->xheaders.p = NULL;
|
|
r->xheaders.n = 0;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Parses HTTP request or response.
|
|
*
|
|
* This parser is responsible for determining the length of a message
|
|
* and slicing the strings inside it. Performance is attained using
|
|
* perfect hash tables. No memory allocation is performed for normal
|
|
* messages. Line folding is forbidden. State persists across calls so
|
|
* that fragmented messages can be handled efficiently. A limitation on
|
|
* message size is imposed to make the header data structures smaller.
|
|
*
|
|
* This parser assumes ISO-8859-1 and guarantees no C0 or C1 control
|
|
* codes are present in message fields, with the exception of tab.
|
|
* Please note that fields like kHttpStateUri may use UTF-8 percent encoding.
|
|
* This parser doesn't care if you choose ASA X3.4-1963 or MULTICS newlines.
|
|
*
|
|
* kHttpRepeatable defines which standard header fields are O(1) and
|
|
* which ones may have comma entries spilled over into xheaders. For
|
|
* most headers it's sufficient to simply check the static slice. If
|
|
* r->headers[kHttpFoo].a is zero then the header is totally absent.
|
|
*
|
|
* This parser has linear complexity. Each character only needs to be
|
|
* considered a single time. That's the case even if messages are
|
|
* fragmented. If a message is valid but incomplete, this function will
|
|
* return zero so that it can be resumed as soon as more data arrives.
|
|
*
|
|
* This parser takes about 400 nanoseconds to parse a 403 byte Chrome
|
|
* HTTP request under MODE=rel on a Core i9 which is about three cycles
|
|
* per byte or a gigabyte per second of throughput per core.
|
|
*
|
|
* @note we assume p points to a buffer that has >=SHRT_MAX bytes
|
|
* @see HTTP/1.1 RFC2616 RFC2068
|
|
* @see HTTP/1.0 RFC1945
|
|
*/
|
|
int ParseHttpMessage(struct HttpMessage *r, const char *p, size_t n) {
|
|
int c, h, i;
|
|
for (n = MIN(n, LIMIT); r->i < n; ++r->i) {
|
|
c = p[r->i] & 0xff;
|
|
switch (r->t) {
|
|
case kHttpStateStart:
|
|
if (c == '\r' || c == '\n') break; /* RFC7230 § 3.5 */
|
|
if (!kHttpToken[c]) return ebadmsg();
|
|
r->t = r->type == kHttpRequest ? kHttpStateMethod : kHttpStateVersion;
|
|
r->a = r->i;
|
|
break;
|
|
case kHttpStateMethod:
|
|
for (;;) {
|
|
if (c == ' ') {
|
|
r->method = GetHttpMethod(p + r->a, r->i - r->a);
|
|
r->xmethod.a = r->a;
|
|
r->xmethod.b = r->i;
|
|
r->a = r->i + 1;
|
|
r->t = kHttpStateUri;
|
|
break;
|
|
} else if (!kHttpToken[c]) {
|
|
return ebadmsg();
|
|
}
|
|
if (++r->i == n) break;
|
|
c = p[r->i] & 0xff;
|
|
}
|
|
break;
|
|
case kHttpStateUri:
|
|
for (;;) {
|
|
if (c == ' ' || c == '\r' || c == '\n') {
|
|
if (r->i == r->a) return ebadmsg();
|
|
r->uri.a = r->a;
|
|
r->uri.b = r->i;
|
|
if (c == ' ') {
|
|
r->a = r->i + 1;
|
|
r->t = kHttpStateVersion;
|
|
} else {
|
|
r->version = 9;
|
|
r->t = c == '\r' ? kHttpStateCr : kHttpStateLf1;
|
|
}
|
|
break;
|
|
} else if (c < 0x20 || (0x7F <= c && c < 0xA0)) {
|
|
return ebadmsg();
|
|
}
|
|
if (++r->i == n) break;
|
|
c = p[r->i] & 0xff;
|
|
}
|
|
break;
|
|
case kHttpStateVersion:
|
|
if (c == ' ' || c == '\r' || c == '\n') {
|
|
if (r->i - r->a == 8 &&
|
|
(READ64BE(p + r->a) & 0xFFFFFFFFFF00FF00) == 0x485454502F002E00 &&
|
|
isdigit(p[r->a + 5]) && isdigit(p[r->a + 7])) {
|
|
r->version = (p[r->a + 5] - '0') * 10 + (p[r->a + 7] - '0');
|
|
if (r->type == kHttpRequest) {
|
|
r->t = c == '\r' ? kHttpStateCr : kHttpStateLf1;
|
|
} else {
|
|
r->t = kHttpStateStatus;
|
|
}
|
|
} else {
|
|
return ebadmsg();
|
|
}
|
|
}
|
|
break;
|
|
case kHttpStateStatus:
|
|
for (;;) {
|
|
if (c == ' ' || c == '\r' || c == '\n') {
|
|
if (r->status < 100) return ebadmsg();
|
|
if (c == ' ') {
|
|
r->a = r->i + 1;
|
|
r->t = kHttpStateMessage;
|
|
} else {
|
|
r->t = c == '\r' ? kHttpStateCr : kHttpStateLf1;
|
|
}
|
|
break;
|
|
} else if ('0' <= c && c <= '9') {
|
|
r->status *= 10;
|
|
r->status += c - '0';
|
|
if (r->status > 999) return ebadmsg();
|
|
} else {
|
|
return ebadmsg();
|
|
}
|
|
if (++r->i == n) break;
|
|
c = p[r->i] & 0xff;
|
|
}
|
|
break;
|
|
case kHttpStateMessage:
|
|
for (;;) {
|
|
if (c == '\r' || c == '\n') {
|
|
r->message.a = r->a;
|
|
r->message.b = r->i;
|
|
r->t = c == '\r' ? kHttpStateCr : kHttpStateLf1;
|
|
break;
|
|
} else if (c < 0x20 || (0x7F <= c && c < 0xA0)) {
|
|
return ebadmsg();
|
|
}
|
|
if (++r->i == n) break;
|
|
c = p[r->i] & 0xff;
|
|
}
|
|
break;
|
|
case kHttpStateCr:
|
|
if (c != '\n') return ebadmsg();
|
|
r->t = kHttpStateLf1;
|
|
break;
|
|
case kHttpStateLf1:
|
|
if (c == '\r') {
|
|
r->t = kHttpStateLf2;
|
|
break;
|
|
} else if (c == '\n') {
|
|
return ++r->i;
|
|
} else if (!kHttpToken[c]) {
|
|
/*
|
|
* 1. Forbid empty header name (RFC2616 §2.2)
|
|
* 2. Forbid line folding (RFC7230 §3.2.4)
|
|
*/
|
|
return ebadmsg();
|
|
}
|
|
r->k.a = r->i;
|
|
r->t = kHttpStateName;
|
|
break;
|
|
case kHttpStateName:
|
|
for (;;) {
|
|
if (c == ':') {
|
|
r->k.b = r->i;
|
|
r->t = kHttpStateColon;
|
|
break;
|
|
} else if (!kHttpToken[c]) {
|
|
return ebadmsg();
|
|
}
|
|
if (++r->i == n) break;
|
|
c = p[r->i] & 0xff;
|
|
}
|
|
break;
|
|
case kHttpStateColon:
|
|
if (c == ' ' || c == '\t') break;
|
|
r->a = r->i;
|
|
r->t = kHttpStateValue;
|
|
/* fallthrough */
|
|
case kHttpStateValue:
|
|
for (;;) {
|
|
if (c == '\r' || c == '\n') {
|
|
i = r->i;
|
|
while (i > r->a && (p[i - 1] == ' ' || p[i - 1] == '\t')) --i;
|
|
if ((h = GetHttpHeader(p + r->k.a, r->k.b - r->k.a)) != -1 &&
|
|
(!r->headers[h].a || !kHttpRepeatable[h])) {
|
|
r->headers[h].a = r->a;
|
|
r->headers[h].b = i;
|
|
} else {
|
|
if (r->xheaders.n == r->xheaders.c) {
|
|
unsigned c2;
|
|
struct HttpHeader *p1, *p2;
|
|
p1 = r->xheaders.p;
|
|
c2 = r->xheaders.c + 2;
|
|
c2 = c2 >> 1;
|
|
if ((p2 = realloc(p1, c2 * sizeof(*p1)))) {
|
|
r->xheaders.p = p2;
|
|
r->xheaders.c = c2;
|
|
}
|
|
}
|
|
if (r->xheaders.n < r->xheaders.c) {
|
|
r->xheaders.p[r->xheaders.n].k = r->k;
|
|
r->xheaders.p[r->xheaders.n].v.a = r->a;
|
|
r->xheaders.p[r->xheaders.n].v.b = i;
|
|
r->xheaders.p = r->xheaders.p;
|
|
++r->xheaders.n;
|
|
}
|
|
}
|
|
r->t = c == '\r' ? kHttpStateCr : kHttpStateLf1;
|
|
break;
|
|
} else if ((c < 0x20 && c != '\t') || (0x7F <= c && c < 0xA0)) {
|
|
return ebadmsg();
|
|
}
|
|
if (++r->i == n) break;
|
|
c = p[r->i] & 0xff;
|
|
}
|
|
break;
|
|
case kHttpStateLf2:
|
|
if (c == '\n') {
|
|
return ++r->i;
|
|
}
|
|
return ebadmsg();
|
|
default:
|
|
unreachable;
|
|
}
|
|
}
|
|
if (r->i < LIMIT) {
|
|
return 0;
|
|
} else {
|
|
return ebadmsg();
|
|
}
|
|
}
|