#include "php_http.h"
#include "php_http_api.h"
+#ifdef HTTP_HAVE_ZLIB
+# include "php_http_send_api.h"
+# include "php_http_headers_api.h"
+# include <zlib.h>
+#endif
+
ZEND_EXTERN_MODULE_GLOBALS(http);
/* {{{ char *http_encoding_dechunk(char *, size_t, char **, size_t *) */
/* }}} */
#ifdef HTTP_HAVE_ZLIB
-#include <zlib.h>
-#define HTTP_GZMAXTRY 10
-#define HTTP_GZBUFLEN(l) (l + (l / 1000) + 16 + 1)
-
-static const char http_gzencode_header[] = {
- (const char) 0x1f,
- (const char) 0x8b,
- (const char) Z_DEFLATED,
- 0, 0, 0, 0, 0, 0,
- (const char) 0x03
+static const char http_encoding_gzip_header[] = {
+ (const char) 0x1f, // fixed value
+ (const char) 0x8b, // fixed value
+ (const char) Z_DEFLATED, // compression algorithm
+ (const char) 0, // none of the possible flags defined by the GZIP "RFC"
+ (const char) 0, // no MTIME available (4 bytes)
+ (const char) 0, // =*=
+ (const char) 0, // =*=
+ (const char) 0, // =*=
+ (const char) 0, // two possible flag values for 9 compression levels? o_O
+ (const char) 0x03 // assume *nix OS
};
inline void http_init_gzencode_buffer(z_stream *Z, const char *data, size_t data_len, char **buf_ptr)
Z->next_in = (Bytef *) data;
Z->avail_in = data_len;
- Z->avail_out = HTTP_GZBUFLEN(data_len) - 1;
+ Z->avail_out = HTTP_ENCODING_BUFLEN(data_len) + HTTP_ENCODING_SAFPAD - 1;
- *buf_ptr = emalloc(Z->avail_out + sizeof(http_gzencode_header));
- memcpy(*buf_ptr, http_gzencode_header, sizeof(http_gzencode_header));
+ *buf_ptr = emalloc(HTTP_ENCODING_BUFLEN(data_len) + sizeof(http_encoding_gzip_header) + HTTP_ENCODING_SAFPAD);
+ memcpy(*buf_ptr, http_encoding_gzip_header, sizeof(http_encoding_gzip_header));
- Z->next_out = *buf_ptr + sizeof(http_gzencode_header);
+ Z->next_out = *buf_ptr + sizeof(http_encoding_gzip_header);
}
inline void http_init_deflate_buffer(z_stream *Z, const char *data, size_t data_len, char **buf_ptr)
Z->zfree = Z_NULL;
Z->opaque = Z_NULL;
- Z->data_type = Z_ASCII;
+ Z->data_type = Z_UNKNOWN;
Z->next_in = (Bytef *) data;
Z->avail_in = data_len;
- Z->avail_out = HTTP_GZBUFLEN(data_len) - 1;
- Z->next_out = emalloc(Z->avail_out);
+ Z->avail_out = HTTP_ENCODING_BUFLEN(data_len) - 1;
+ Z->next_out = emalloc(HTTP_ENCODING_BUFLEN(data_len));
*buf_ptr = Z->next_out;
}
-inline void http_init_inflate_buffer(z_stream *Z, const char *data, size_t data_len, char **buf_ptr, size_t *buf_len, int iteration)
+inline void http_init_uncompress_buffer(size_t data_len, char **buf_ptr, size_t *buf_len, int *iteration)
{
- Z->zalloc = Z_NULL;
- Z->zfree = Z_NULL;
-
- if (!iteration) {
+ if (!*iteration) {
*buf_len = data_len * 2;
*buf_ptr = emalloc(*buf_len + 1);
} else {
- *buf_len <<= 2;
- *buf_ptr = erealloc(*buf_ptr, *buf_len + 1);
+ size_t new_len = *buf_len << 2;
+ char *new_ptr = erealloc(*buf_ptr, new_len + 1);
+
+ if (new_ptr) {
+ *buf_ptr = new_ptr;
+ *buf_len = new_len;
+ } else {
+ *iteration = INT_MAX;
+ }
}
+}
+
+inline void http_init_inflate_buffer(z_stream *Z, const char *data, size_t data_len, char **buf_ptr, size_t *buf_len, int *iteration)
+{
+ Z->zalloc = Z_NULL;
+ Z->zfree = Z_NULL;
+
+ http_init_uncompress_buffer(data_len, buf_ptr, buf_len, iteration);
Z->next_in = (Bytef *) data;
Z->avail_in = data_len;
crc = crc32(0L, Z_NULL, 0);
crc = crc32(crc, (const Bytef *) data, data_len);
- trailer = *buf_ptr + sizeof(http_gzencode_header) + Z->total_out;
+ trailer = *buf_ptr + sizeof(http_encoding_gzip_header) + Z->total_out;
/* LSB */
trailer[0] = (char) (crc & 0xFF);
trailer[6] = (char) ((Z->total_in >> 16) & 0xFF);
trailer[7] = (char) ((Z->total_in >> 24) & 0xFF);
- return http_finish_buffer(Z->total_out + sizeof(http_gzencode_header) + 8, buf_ptr);
+ return http_finish_buffer(Z->total_out + sizeof(http_encoding_gzip_header) + 8, buf_ptr);
}
+inline STATUS http_verify_gzencode_buffer(const char *data, size_t data_len, const char **encoded, size_t *encoded_len, int error_level TSRMLS_DC)
+{
+ size_t offset = sizeof(http_encoding_gzip_header);
+
+ if (data_len < offset) {
+ goto really_bad_gzip_header;
+ }
+
+ if (data[0] != (const char) 0x1F || data[1] != (const char) 0x8B) {
+ http_error_ex(error_level TSRMLS_CC, HTTP_E_ENCODING, "Unrecognized GZIP header start: 0x%02X 0x%02X", (int) data[0], (int) (data[1] & 0xFF));
+ return FAILURE;
+ }
+
+ if (data[2] != (const char) Z_DEFLATED) {
+ http_error_ex(error_level TSRMLS_CC, HTTP_E_ENCODING, "Unrecognized compression format (%d)", (int) (data[2] & 0xFF));
+ /* still try to decode */
+ }
+ if ((data[3] & 0x4) == 0x4) {
+ if (data_len < offset + 2) {
+ goto really_bad_gzip_header;
+ }
+ /* there are extra fields, the length follows the common header as 2 bytes LSB */
+ offset += (unsigned) ((data[offset] & 0xFF));
+ offset += 1;
+ offset += (unsigned) ((data[offset] & 0xFF) << 8);
+ offset += 1;
+ }
+ if ((data[3] & 0x8) == 0x8) {
+ if (data_len <= offset) {
+ goto really_bad_gzip_header;
+ }
+ /* there's a file name */
+ offset += strlen(&data[offset]) + 1 /*NUL*/;
+ }
+ if ((data[3] & 0x10) == 0x10) {
+ if (data_len <= offset) {
+ goto really_bad_gzip_header;
+ }
+ /* there's a comment */
+ offset += strlen(&data[offset]) + 1 /* NUL */;
+ }
+ if ((data[3] & 0x2) == 0x2) {
+ /* there's a CRC16 of the header */
+ offset += 2;
+ if (data_len <= offset) {
+ goto really_bad_gzip_header;
+ } else {
+ unsigned long crc, cmp;
+
+ cmp = (unsigned) ((data[offset-2] & 0xFF));
+ cmp += (unsigned) ((data[offset-1] & 0xFF) << 8);
+
+ crc = crc32(0L, Z_NULL, 0);
+ crc = crc32(crc, data, sizeof(http_encoding_gzip_header));
+
+ if (cmp != (crc & 0xFFFF)) {
+ http_error_ex(error_level TSRMLS_CC, HTTP_E_ENCODING, "GZIP headers CRC checksums so not match (%lu, %lu)", cmp, crc & 0xFFFF);
+ return FAILURE;
+ }
+ }
+ }
+
+ if (data_len < offset + 8) {
+ http_error(error_level TSRMLS_CC, HTTP_E_ENCODING, "Missing or truncated GZIP footer");
+ return FAILURE;
+ }
+
+ if (encoded) {
+ *encoded = data + offset;
+ }
+ if (encoded_len) {
+ *encoded_len = data_len - offset - 8 /* size of the assumed GZIP footer */;
+ }
+
+ return SUCCESS;
+
+really_bad_gzip_header:
+ http_error(error_level TSRMLS_CC, HTTP_E_ENCODING, "Missing or truncated GZIP header");
+ return FAILURE;
+}
+
+inline STATUS http_verify_gzdecode_buffer(const char *data, size_t data_len, const char *decoded, size_t decoded_len, int error_level TSRMLS_DC)
+{
+ STATUS status = SUCCESS;
+ unsigned long len, cmp, crc;
+
+ crc = crc32(0L, Z_NULL, 0);
+ crc = crc32(crc, (const Bytef *) decoded, decoded_len);
+
+ cmp = (unsigned) ((data[data_len-8] & 0xFF));
+ cmp += (unsigned) ((data[data_len-7] & 0xFF) << 8);
+ cmp += (unsigned) ((data[data_len-6] & 0xFF) << 16);
+ cmp += (unsigned) ((data[data_len-5] & 0xFF) << 24);
+ len = (unsigned) ((data[data_len-4] & 0xFF));
+ len += (unsigned) ((data[data_len-3] & 0xFF) << 8);
+ len += (unsigned) ((data[data_len-2] & 0xFF) << 16);
+ len += (unsigned) ((data[data_len-1] & 0xFF) << 24);
+
+ if (cmp != crc) {
+ http_error_ex(error_level TSRMLS_CC, HTTP_E_ENCODING, "Could not verify data integrity: CRC checksums do not match (%lu, %lu)", cmp, crc);
+ status = FAILURE;
+ }
+ if (len != decoded_len) {
+ http_error_ex(error_level TSRMLS_CC, HTTP_E_ENCODING, "Could not verify data integrity: data sizes do not match (%lu, %lu)", len, decoded_len);
+ status = FAILURE;
+ }
+ return status;
+}
+
+PHP_HTTP_API STATUS _http_encode(http_encoding_type type, int level, const char *data, size_t data_len, char **encoded, size_t *encoded_len TSRMLS_DC)
+{
+ STATUS status = SUCCESS;
+
+ switch (type)
+ {
+ case HTTP_ENCODING_ANY:
+ case HTTP_ENCODING_GZIP:
+ status = http_encoding_gzencode(level, data, data_len, encoded, encoded_len);
+ break;
+
+ case HTTP_ENCODING_DEFLATE:
+ status = http_encoding_deflate(level, data, data_len, encoded, encoded_len);
+ break;
+
+ case HTTP_ENCODING_COMPRESS:
+ status = http_encoding_compress(level, data, data_len, encoded, encoded_len);
+ break;
+
+ case HTTP_ENCODING_NONE:
+ default:
+ *encoded = estrndup(data, data_len);
+ *encoded_len = data_len;
+ break;
+ }
+
+ return status;
+}
+
+PHP_HTTP_API STATUS _http_decode(http_encoding_type type, const char *data, size_t data_len, char **decoded, size_t *decoded_len TSRMLS_DC)
+{
+ STATUS status = SUCCESS;
+
+ switch (type)
+ {
+ case HTTP_ENCODING_ANY:
+ if ( SUCCESS != http_encoding_gzdecode(data, data_len, decoded, decoded_len) &&
+ SUCCESS != http_encoding_inflate(data, data_len, decoded, decoded_len) &&
+ SUCCESS != http_encoding_uncompress(data, data_len, decoded, decoded_len)) {
+ status = FAILURE;
+ }
+ break;
+
+ case HTTP_ENCODING_GZIP:
+ status = http_encoding_gzdecode(data, data_len, decoded, decoded_len);
+ break;
+
+ case HTTP_ENCODING_DEFLATE:
+ status = http_encoding_inflate(data, data_len, decoded, decoded_len);
+ break;
+
+ case HTTP_ENCODING_COMPRESS:
+ status = http_encoding_uncompress(data, data_len, decoded, decoded_len);
+ break;
+
+ case HTTP_ENCODING_NONE:
+ default:
+ *decoded = estrndup(data, data_len);
+ *decoded_len = data_len;
+ break;
+ }
+
+ return status;
+}
PHP_HTTP_API STATUS _http_encoding_gzencode(int level, const char *data, size_t data_len, char **encoded, size_t *encoded_len TSRMLS_DC)
{
{
STATUS status;
- *encoded = emalloc(*encoded_len = HTTP_GZBUFLEN(data_len));
+ *encoded = emalloc(*encoded_len = HTTP_ENCODING_BUFLEN(data_len));
if (Z_OK == (status = compress2(*encoded, encoded_len, data, data_len, level))) {
http_finish_buffer(*encoded_len, encoded);
PHP_HTTP_API STATUS _http_encoding_gzdecode(const char *data, size_t data_len, char **decoded, size_t *decoded_len TSRMLS_DC)
{
- const char *encoded = data + sizeof(http_gzencode_header);
+ const char *encoded;
size_t encoded_len;
- if (data_len <= sizeof(http_gzencode_header) + 8) {
- http_error(HE_WARNING, HTTP_E_ENCODING, "Could not gzdecode data: too short data length");
- } else {
- encoded_len = data_len - sizeof(http_gzencode_header) - 8;
-
- if (SUCCESS == http_encoding_inflate(encoded, encoded_len, decoded, decoded_len)) {
- unsigned long len = 0, cmp = 0, crc = crc32(0L, Z_NULL, 0);
-
- crc = crc32(crc, (const Bytef *) *decoded, *decoded_len);
-
- cmp = (unsigned) ((data[data_len-8] & 0xFF));
- cmp += (unsigned) ((data[data_len-7] & 0xFF) << 8);
- cmp += (unsigned) ((data[data_len-6] & 0xFF) << 16);
- cmp += (unsigned) ((data[data_len-5] & 0xFF) << 24);
- len = (unsigned) ((data[data_len-4] & 0xFF));
- len += (unsigned) ((data[data_len-3] & 0xFF) << 8);
- len += (unsigned) ((data[data_len-2] & 0xFF) << 16);
- len += (unsigned) ((data[data_len-1] & 0xFF) << 24);
-
- if (cmp != crc) {
- http_error_ex(HE_NOTICE, HTTP_E_ENCODING, "Could not verify data integrity: CRC checksums do not match (%lu, %lu)", cmp, crc);
- }
- if (len != *decoded_len) {
- http_error_ex(HE_NOTICE, HTTP_E_ENCODING, "Could not verify data integrity: data sizes do not match (%lu, %lu)", len, *decoded_len);
- }
-
- return SUCCESS;
- }
+ if ( (SUCCESS == http_verify_gzencode_buffer(data, data_len, &encoded, &encoded_len, HE_NOTICE)) &&
+ (SUCCESS == http_encoding_inflate(encoded, encoded_len, decoded, decoded_len))) {
+ http_verify_gzdecode_buffer(data, data_len, *decoded, *decoded_len, HE_NOTICE);
+ return SUCCESS;
}
+
return FAILURE;
}
z_stream Z;
do {
- http_init_inflate_buffer(&Z, data, data_len, decoded, decoded_len, max++);
+ http_init_inflate_buffer(&Z, data, data_len, decoded, decoded_len, &max);
if (Z_OK == (status = inflateInit2(&Z, -MAX_WBITS))) {
if (Z_STREAM_END == (status = inflate(&Z, Z_FINISH))) {
if (Z_OK == (status = inflateEnd(&Z))) {
}
}
}
- } while (max < HTTP_GZMAXTRY);
+ } while (++max < HTTP_ENCODING_MAXTRY && status == Z_BUF_ERROR);
+ efree(*decoded);
http_error_ex(HE_WARNING, HTTP_E_ENCODING, "Could not inflate data: %s", zError(status));
return FAILURE;
}
{
int max = 0;
STATUS status;
- size_t want = data_len * 2;
-
- *decoded = emalloc(want + 1);
- if (Z_BUF_ERROR == (status = uncompress(*decoded, &want, data, data_len))) do {
- /* this is a lot faster with large data than gzuncompress(),
- but could be a problem with a low memory limit */
- want <<= 2;
- *decoded = erealloc(*decoded, want + 1);
- status = uncompress(*decoded, &want, data, data_len);
- } while (++max < HTTP_GZMAXTRY && status == Z_BUF_ERROR);
-
- if (Z_OK == status) {
- *decoded_len = http_finish_buffer(want, decoded);
- return SUCCESS;
- }
+
+ do {
+ http_init_uncompress_buffer(data_len, decoded, decoded_len, &max);
+ if (Z_OK == (status = uncompress(*decoded, decoded_len, data, data_len))) {
+ http_finish_buffer(*decoded_len, decoded);
+ return SUCCESS;
+ }
+ } while (++max < HTTP_ENCODING_MAXTRY && status == Z_BUF_ERROR);
efree(*decoded);
http_error_ex(HE_WARNING, HTTP_E_ENCODING, "Could not uncompress data: %s", zError(status));
return FAILURE;
}
+#define HTTP_ENCODING_STREAM_ERROR(status, tofree) \
+ { \
+ if (tofree) efree(tofree); \
+ http_error_ex(HE_WARNING, HTTP_E_ENCODING, "GZIP stream error: %s", zError(status)); \
+ return FAILURE; \
+ }
+
+PHP_HTTP_API STATUS _http_encoding_stream_init(http_encoding_stream *s, int gzip, int level, char **encoded, size_t *encoded_len TSRMLS_DC)
+{
+ STATUS status;
+
+ memset(s, 0, sizeof(http_encoding_stream));
+ if (Z_OK != (status = deflateInit2(&s->Z, level, Z_DEFLATED, -MAX_WBITS, MAX_MEM_LEVEL, Z_DEFAULT_STRATEGY))) {
+ HTTP_ENCODING_STREAM_ERROR(status, NULL);
+ }
+
+ if (s->gzip = gzip) {
+ s->crc = crc32(0L, Z_NULL, 0);
+ *encoded_len = sizeof(http_encoding_gzip_header);
+ *encoded = emalloc(*encoded_len);
+ memcpy(*encoded, http_encoding_gzip_header, *encoded_len);
+ } else {
+ *encoded_len = 0;
+ *encoded = NULL;
+ }
+
+ return SUCCESS;
+}
+
+PHP_HTTP_API STATUS _http_encoding_stream_update(http_encoding_stream *s, const char *data, size_t data_len, char **encoded, size_t *encoded_len TSRMLS_DC)
+{
+ STATUS status;
+
+ *encoded_len = HTTP_ENCODING_BUFLEN(data_len);
+ *encoded = emalloc(*encoded_len);
+
+ s->Z.next_in = (Bytef *) data;
+ s->Z.avail_in = data_len;
+ s->Z.next_out = *encoded;
+ s->Z.avail_out = *encoded_len;
+
+ status = deflate(&s->Z, Z_SYNC_FLUSH);
+
+ if (Z_OK != status && Z_STREAM_END != status) {
+ HTTP_ENCODING_STREAM_ERROR(status, *encoded);
+ }
+ *encoded_len -= s->Z.avail_out;
+
+ if (s->gzip) {
+ s->crc = crc32(s->crc, (const Bytef *) data, data_len);
+ }
+
+ return SUCCESS;
+}
+
+PHP_HTTP_API STATUS _http_encoding_stream_finish(http_encoding_stream *s, char **encoded, size_t *encoded_len TSRMLS_DC)
+{
+ STATUS status;
+
+ *encoded_len = 1024;
+ *encoded = emalloc(*encoded_len);
+
+ s->Z.next_out = *encoded;
+ s->Z.avail_out = *encoded_len;
+
+ if (Z_STREAM_END != (status = deflate(&s->Z, Z_FINISH)) || Z_OK != (status = deflateEnd(&s->Z))) {
+ HTTP_ENCODING_STREAM_ERROR(status, *encoded);
+ }
+
+ *encoded_len -= s->Z.avail_out;
+ if (s->gzip) {
+ if (s->Z.avail_out < 8) {
+ *encoded = erealloc(*encoded, *encoded_len + 8);
+ }
+ (*encoded)[(*encoded_len)++] = (char) (s->crc & 0xFF);
+ (*encoded)[(*encoded_len)++] = (char) ((s->crc >> 8) & 0xFF);
+ (*encoded)[(*encoded_len)++] = (char) ((s->crc >> 16) & 0xFF);
+ (*encoded)[(*encoded_len)++] = (char) ((s->crc >> 24) & 0xFF);
+ (*encoded)[(*encoded_len)++] = (char) ((s->Z.total_in) & 0xFF);
+ (*encoded)[(*encoded_len)++] = (char) ((s->Z.total_in >> 8) & 0xFF);
+ (*encoded)[(*encoded_len)++] = (char) ((s->Z.total_in >> 16) & 0xFF);
+ (*encoded)[(*encoded_len)++] = (char) ((s->Z.total_in >> 24) & 0xFF);
+ }
+
+ return SUCCESS;
+}
+
#endif /* HTTP_HAVE_ZLIB */
+PHP_HTTP_API zend_bool _http_encoding_response_start(size_t content_length TSRMLS_DC)
+{
+ if ( php_ob_handler_used("ob_gzhandler" TSRMLS_CC) ||
+ php_ob_handler_used("zlib output compression" TSRMLS_CC)) {
+ HTTP_G(send).gzip_encoding = 0;
+ } else {
+ if (!HTTP_G(send).gzip_encoding) {
+ /* emit a content-length header */
+ if (content_length) {
+ char cl_header_str[128];
+ size_t cl_header_len;
+ cl_header_len = snprintf(cl_header_str, lenof(cl_header_str), "Content-Length: %lu", (unsigned long) content_length);
+ http_send_header_string_ex(cl_header_str, cl_header_len, 1);
+ }
+ } else {
+#ifndef HTTP_HAVE_ZLIB
+ php_start_ob_buffer_named("ob_gzhandler", 0, 0 TSRMLS_CC);
+#else
+ HashTable *selected;
+ zval zsupported;
+
+ INIT_PZVAL(&zsupported);
+ array_init(&zsupported);
+ add_next_index_stringl(&zsupported, "gzip", lenof("gzip"), 1);
+ add_next_index_stringl(&zsupported, "deflate", lenof("deflate"), 1);
+
+ if (selected = http_negotiate_encoding(&zsupported)) {
+ STATUS hs = FAILURE;
+ char *encoding = NULL;
+ ulong idx;
+
+ if (HASH_KEY_IS_STRING == zend_hash_get_current_key(selected, &encoding, &idx, 0) && encoding) {
+ if (!strcmp(encoding, "gzip")) {
+ if (SUCCESS == (hs = http_send_header_string("Content-Encoding: gzip"))) {
+ HTTP_G(send).gzip_encoding = HTTP_ENCODING_GZIP;
+ }
+ } else if (!strcmp(encoding, "deflate")) {
+ if (SUCCESS == (hs = http_send_header_string("Content-Encoding: deflate"))) {
+ HTTP_G(send).gzip_encoding = HTTP_ENCODING_DEFLATE;
+ }
+ }
+ if (SUCCESS == hs) {
+ http_send_header_string("Vary: Accept-Encoding");
+ } else {
+ HTTP_G(send).gzip_encoding = 0;
+ }
+ }
+
+ zend_hash_destroy(selected);
+ FREE_HASHTABLE(selected);
+ }
+
+ zval_dtor(&zsupported);
+ return 1;
+#endif
+ }
+ }
+ return 0;
+}
+
/*
* Local variables:
* tab-width: 4
projects / m6w6 / ext-http / blobdiff