freebsd-skq/contrib/serf/buckets/headers_buckets.c
peter 82dd802622 Update serf-1.3.0 -> 1.3.4 - fixes multiple issues (see the CHANGES file)
including an SSL issue that turned up in the cluster with svn-1.8.8.
2014-02-22 01:19:47 +00:00

457 lines
13 KiB
C

/* Copyright 2004 Justin Erenkrantz and Greg Stein
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdlib.h>
#include <apr_general.h> /* for strcasecmp() */
#include "serf.h"
#include "serf_bucket_util.h"
#include "serf_private.h" /* for serf__bucket_headers_remove */
typedef struct header_list {
const char *header;
const char *value;
apr_size_t header_size;
apr_size_t value_size;
int alloc_flags;
#define ALLOC_HEADER 0x0001 /* header lives in our allocator */
#define ALLOC_VALUE 0x0002 /* value lives in our allocator */
struct header_list *next;
} header_list_t;
typedef struct {
header_list_t *list;
header_list_t *last;
header_list_t *cur_read;
enum {
READ_START, /* haven't started reading yet */
READ_HEADER, /* reading cur_read->header */
READ_SEP, /* reading ": " */
READ_VALUE, /* reading cur_read->value */
READ_CRLF, /* reading "\r\n" */
READ_TERM, /* reading the final "\r\n" */
READ_DONE /* no more data to read */
} state;
apr_size_t amt_read; /* how much of the current state we've read */
} headers_context_t;
serf_bucket_t *serf_bucket_headers_create(
serf_bucket_alloc_t *allocator)
{
headers_context_t *ctx;
ctx = serf_bucket_mem_alloc(allocator, sizeof(*ctx));
ctx->list = NULL;
ctx->last = NULL;
ctx->state = READ_START;
return serf_bucket_create(&serf_bucket_type_headers, allocator, ctx);
}
void serf_bucket_headers_setx(
serf_bucket_t *bkt,
const char *header, apr_size_t header_size, int header_copy,
const char *value, apr_size_t value_size, int value_copy)
{
headers_context_t *ctx = bkt->data;
header_list_t *hdr;
#if 0
/* ### include this? */
if (ctx->cur_read) {
/* we started reading. can't change now. */
abort();
}
#endif
hdr = serf_bucket_mem_alloc(bkt->allocator, sizeof(*hdr));
hdr->header_size = header_size;
hdr->value_size = value_size;
hdr->alloc_flags = 0;
hdr->next = NULL;
if (header_copy) {
hdr->header = serf_bstrmemdup(bkt->allocator, header, header_size);
hdr->alloc_flags |= ALLOC_HEADER;
}
else {
hdr->header = header;
}
if (value_copy) {
hdr->value = serf_bstrmemdup(bkt->allocator, value, value_size);
hdr->alloc_flags |= ALLOC_VALUE;
}
else {
hdr->value = value;
}
/* Add the new header at the end of the list. */
if (ctx->last)
ctx->last->next = hdr;
else
ctx->list = hdr;
ctx->last = hdr;
}
void serf_bucket_headers_set(
serf_bucket_t *headers_bucket,
const char *header,
const char *value)
{
serf_bucket_headers_setx(headers_bucket,
header, strlen(header), 0,
value, strlen(value), 1);
}
void serf_bucket_headers_setc(
serf_bucket_t *headers_bucket,
const char *header,
const char *value)
{
serf_bucket_headers_setx(headers_bucket,
header, strlen(header), 1,
value, strlen(value), 1);
}
void serf_bucket_headers_setn(
serf_bucket_t *headers_bucket,
const char *header,
const char *value)
{
serf_bucket_headers_setx(headers_bucket,
header, strlen(header), 0,
value, strlen(value), 0);
}
const char *serf_bucket_headers_get(
serf_bucket_t *headers_bucket,
const char *header)
{
headers_context_t *ctx = headers_bucket->data;
header_list_t *found = ctx->list;
const char *val = NULL;
int value_size = 0;
int val_alloc = 0;
while (found) {
if (strcasecmp(found->header, header) == 0) {
if (val) {
/* The header is already present. RFC 2616, section 4.2
indicates that we should append the new value, separated by
a comma. Reasoning: for headers whose values are known to
be comma-separated, that is clearly the correct behavior;
for others, the correct behavior is undefined anyway. */
/* The "+1" is for the comma; the +1 in the alloc
call is for the terminating '\0' */
apr_size_t new_size = found->value_size + value_size + 1;
char *new_val = serf_bucket_mem_alloc(headers_bucket->allocator,
new_size + 1);
memcpy(new_val, val, value_size);
new_val[value_size] = ',';
memcpy(new_val + value_size + 1, found->value,
found->value_size);
new_val[new_size] = '\0';
/* Copy the new value over the already existing value. */
if (val_alloc)
serf_bucket_mem_free(headers_bucket->allocator, (void*)val);
val_alloc |= ALLOC_VALUE;
val = new_val;
value_size = new_size;
}
else {
val = found->value;
value_size = found->value_size;
}
}
found = found->next;
}
return val;
}
void serf__bucket_headers_remove(serf_bucket_t *bucket, const char *header)
{
headers_context_t *ctx = bucket->data;
header_list_t *scan = ctx->list, *prev = NULL;
/* Find and delete all items with the same header (case insensitive) */
while (scan) {
if (strcasecmp(scan->header, header) == 0) {
if (prev) {
prev->next = scan->next;
} else {
ctx->list = scan->next;
}
if (ctx->last == scan) {
ctx->last = NULL;
}
} else {
prev = scan;
}
scan = scan->next;
}
}
void serf_bucket_headers_do(
serf_bucket_t *headers_bucket,
serf_bucket_headers_do_callback_fn_t func,
void *baton)
{
headers_context_t *ctx = headers_bucket->data;
header_list_t *scan = ctx->list;
while (scan) {
if (func(baton, scan->header, scan->value) != 0) {
break;
}
scan = scan->next;
}
}
static void serf_headers_destroy_and_data(serf_bucket_t *bucket)
{
headers_context_t *ctx = bucket->data;
header_list_t *scan = ctx->list;
while (scan) {
header_list_t *next_hdr = scan->next;
if (scan->alloc_flags & ALLOC_HEADER)
serf_bucket_mem_free(bucket->allocator, (void *)scan->header);
if (scan->alloc_flags & ALLOC_VALUE)
serf_bucket_mem_free(bucket->allocator, (void *)scan->value);
serf_bucket_mem_free(bucket->allocator, scan);
scan = next_hdr;
}
serf_default_destroy_and_data(bucket);
}
static void select_value(
headers_context_t *ctx,
const char **value,
apr_size_t *len)
{
const char *v;
apr_size_t l;
if (ctx->state == READ_START) {
if (ctx->list == NULL) {
/* No headers. Move straight to the TERM state. */
ctx->state = READ_TERM;
}
else {
ctx->state = READ_HEADER;
ctx->cur_read = ctx->list;
}
ctx->amt_read = 0;
}
switch (ctx->state) {
case READ_HEADER:
v = ctx->cur_read->header;
l = ctx->cur_read->header_size;
break;
case READ_SEP:
v = ": ";
l = 2;
break;
case READ_VALUE:
v = ctx->cur_read->value;
l = ctx->cur_read->value_size;
break;
case READ_CRLF:
case READ_TERM:
v = "\r\n";
l = 2;
break;
case READ_DONE:
*len = 0;
return;
default:
/* Not reachable */
return;
}
*value = v + ctx->amt_read;
*len = l - ctx->amt_read;
}
/* the current data chunk has been read/consumed. move our internal state. */
static apr_status_t consume_chunk(headers_context_t *ctx)
{
/* move to the next state, resetting the amount read. */
++ctx->state;
ctx->amt_read = 0;
/* just sent the terminator and moved to DONE. signal completion. */
if (ctx->state == READ_DONE)
return APR_EOF;
/* end of this header. move to the next one. */
if (ctx->state == READ_TERM) {
ctx->cur_read = ctx->cur_read->next;
if (ctx->cur_read != NULL) {
/* We've got another head to send. Reset the read state. */
ctx->state = READ_HEADER;
}
/* else leave in READ_TERM */
}
/* there is more data which can be read immediately. */
return APR_SUCCESS;
}
static apr_status_t serf_headers_peek(serf_bucket_t *bucket,
const char **data,
apr_size_t *len)
{
headers_context_t *ctx = bucket->data;
select_value(ctx, data, len);
/* already done or returning the CRLF terminator? return EOF */
if (ctx->state == READ_DONE || ctx->state == READ_TERM)
return APR_EOF;
return APR_SUCCESS;
}
static apr_status_t serf_headers_read(serf_bucket_t *bucket,
apr_size_t requested,
const char **data, apr_size_t *len)
{
headers_context_t *ctx = bucket->data;
apr_size_t avail;
select_value(ctx, data, &avail);
if (ctx->state == READ_DONE) {
*len = avail;
return APR_EOF;
}
if (requested >= avail) {
/* return everything from this chunk */
*len = avail;
/* we consumed this chunk. advance the state. */
return consume_chunk(ctx);
}
/* return just the amount requested, and advance our pointer */
*len = requested;
ctx->amt_read += requested;
/* there is more that can be read immediately */
return APR_SUCCESS;
}
static apr_status_t serf_headers_readline(serf_bucket_t *bucket,
int acceptable, int *found,
const char **data, apr_size_t *len)
{
headers_context_t *ctx = bucket->data;
apr_status_t status;
/* ### what behavior should we use here? APR_EGENERAL for now */
if ((acceptable & SERF_NEWLINE_CRLF) == 0)
return APR_EGENERAL;
/* get whatever is in this chunk */
select_value(ctx, data, len);
if (ctx->state == READ_DONE)
return APR_EOF;
/* we consumed this chunk. advance the state. */
status = consume_chunk(ctx);
/* the type of newline found is easy... */
*found = (ctx->state == READ_CRLF || ctx->state == READ_TERM)
? SERF_NEWLINE_CRLF : SERF_NEWLINE_NONE;
return status;
}
static apr_status_t serf_headers_read_iovec(serf_bucket_t *bucket,
apr_size_t requested,
int vecs_size,
struct iovec *vecs,
int *vecs_used)
{
apr_size_t avail = requested;
int i;
*vecs_used = 0;
for (i = 0; i < vecs_size; i++) {
const char *data;
apr_size_t len;
apr_status_t status;
/* Calling read() would not be a safe opt in the general case, but it
* is here for the header bucket as it only frees all of the header
* keys and values when the entire bucket goes away - not on a
* per-read() basis as is normally the case.
*/
status = serf_headers_read(bucket, avail, &data, &len);
if (len) {
vecs[*vecs_used].iov_base = (char*)data;
vecs[*vecs_used].iov_len = len;
(*vecs_used)++;
if (avail != SERF_READ_ALL_AVAIL) {
avail -= len;
/* If we reach 0, then read()'s status will suffice. */
if (avail == 0) {
return status;
}
}
}
if (status) {
return status;
}
}
return APR_SUCCESS;
}
const serf_bucket_type_t serf_bucket_type_headers = {
"HEADERS",
serf_headers_read,
serf_headers_readline,
serf_headers_read_iovec,
serf_default_read_for_sendfile,
serf_default_read_bucket,
serf_headers_peek,
serf_headers_destroy_and_data,
};