MEDIUM: h1-htx: Add a function to parse contiguous small chunks
Add h1_parse_full_contig_chunks() function to parse full contiguous chunks.
This function neither handles incomplete chunks nor wrapping buffers. It is
designed to efficiently parse a buffer with several small chunks. Of course,
there is no zero copy here because it is not possible. This function is a
bit tricky and all changes may a have a impact. This one may probably be
optimized, but it is good enough for now and not too complex.
The main function (h1_parse_msg_chunks) always tries to use this function
when the HTTP parser is waiting for a chunk size. In this case, there is no
zero-copy, so there is no reason to call the generic version to parse the
chunk. However, if some unparsed data remain after this step, the generic
function is called. This way, wrapping data and incomplete chunks may be
parsed.
Quick tests show it is now slightly faster in all cases than the legacy
mode.
diff --git a/src/h1_htx.c b/src/h1_htx.c
index 7ac2708..adb96e1 100644
--- a/src/h1_htx.c
+++ b/src/h1_htx.c
@@ -534,7 +534,204 @@
return total;
}
-/* Parse HTTP chunks */
+/* Parses full contiguous HTTP chunks. This version is optimized for small
+ * chunks and does not performed zero-copy. It must be called in
+ * H1_MSG_CHUNK_SIZE state. Be carefull if you change something in this
+ * function. It is really sensitive, any change may have an impact on
+ * performance.
+ */
+static size_t h1_parse_full_contig_chunks(struct h1m *h1m, struct htx **dsthtx,
+ struct buffer *srcbuf, size_t ofs, size_t *max,
+ struct buffer *htxbuf)
+{
+ char *start, *end, *dptr;
+ ssize_t dpos, ridx, save;
+ size_t lmax, total = 0;
+ uint64_t chksz;
+ struct htx_ret htxret;
+
+ /* source info :
+ * start : pointer at <ofs> position
+ * end : pointer marking the end of data to parse
+ * ridx : the reverse index (negative) marking the parser position (end[ridx])
+ */
+ ridx = -b_contig_data(srcbuf, ofs);
+ if (!ridx)
+ goto out;
+ start = b_peek(srcbuf, ofs);
+ end = start - ridx;
+
+ /* Reserve the maximum possible size for the data */
+ htxret = htx_reserve_max_data(*dsthtx);
+ if (!htxret.blk)
+ goto out;
+
+ /* destination info :
+ * dptr : pointer on the beginning of the data
+ * dpos : current position where to copy data
+ */
+ dptr = htx_get_blk_ptr(*dsthtx, htxret.blk);
+ dpos = htxret.ret;
+
+ /* Empty DATA block is not possible, thus if <dpos> is the beginning of
+ * the block, it means it is a new block. We can remove the block size
+ * from <max>. Then we must adjust it if it exceeds the free size in the
+ * block.
+ */
+ lmax = *max;
+ if (!dpos)
+ lmax -= sizeof(struct htx_blk);
+ if (lmax > htx_get_blksz(htxret.blk) - dpos)
+ lmax = htx_get_blksz(htxret.blk) - dpos;
+
+ while (1) {
+ /* The chunk size is in the following form, though we are only
+ * interested in the size and CRLF :
+ * 1*HEXDIGIT *WSP *[ ';' extensions ] CRLF
+ */
+ chksz = 0;
+ save = ridx; /* Save the parser position to rewind if necessary */
+ while (1) {
+ int c;
+
+ if (!ridx)
+ goto end_parsing;
+
+ /* Convert current character */
+ c = hex2i(end[ridx]);
+
+ /* not a hex digit anymore */
+ if (c < 0)
+ break;
+
+ /* Update current chunk size */
+ chksz = (chksz << 4) + c;
+
+ if (unlikely(chksz & 0xF0000000000000)) {
+ /* Don't get more than 13 hexa-digit (2^52 - 1)
+ * to never fed possibly bogus values from
+ * languages that use floats for their integers
+ */
+ goto parsing_error;
+ }
+ ++ridx;
+ }
+
+ if (unlikely(chksz > lmax))
+ goto end_parsing;
+
+ if (unlikely(ridx == save)) {
+ /* empty size not allowed */
+ goto parsing_error;
+ }
+
+ /* Skip spaces */
+ while (HTTP_IS_SPHT(end[ridx])) {
+ if (!++ridx)
+ goto end_parsing;
+ }
+
+ /* Up to there, we know that at least one byte is present. Check
+ * for the end of chunk size.
+ */
+ while (1) {
+ if (likely(end[ridx] == '\r')) {
+ /* Parse CRLF */
+ if (!++ridx)
+ goto end_parsing;
+ if (unlikely(end[ridx] != '\n')) {
+ /* CR must be followed by LF */
+ goto parsing_error;
+ }
+
+ /* done */
+ ++ridx;
+ break;
+ }
+ else if (end[ridx] == '\n') {
+ /* Parse LF only, nothing more to do */
+ ++ridx;
+ break;
+ }
+ else if (likely(end[ridx] == ';')) {
+ /* chunk extension, ends at next CRLF */
+ if (!++ridx)
+ goto end_parsing;
+ while (!HTTP_IS_CRLF(end[ridx])) {
+ if (!++ridx)
+ goto end_parsing;
+ }
+ /* we have a CRLF now, loop above */
+ continue;
+ }
+ else {
+ /* all other characters are unexpected */
+ goto parsing_error;
+ }
+ }
+
+ /* Exit if it is the last chunk */
+ if (unlikely(!chksz)) {
+ h1m->state = H1_MSG_TRAILERS;
+ save = ridx;
+ goto end_parsing;
+ }
+
+ /* Now check if the whole chunk is here (including the CRLF at
+ * the end), otherise we switch in H1_MSG_DATA stae.
+ */
+ if (chksz + 2 > -ridx) {
+ h1m->curr_len = chksz;
+ h1m->body_len += chksz;
+ h1m->state = H1_MSG_DATA;
+ (*dsthtx)->extra = h1m->curr_len;
+ save = ridx;
+ goto end_parsing;
+ }
+
+ memcpy(dptr + dpos, end + ridx, chksz);
+ h1m->body_len += chksz;
+ lmax -= chksz;
+ dpos += chksz;
+ ridx += chksz;
+
+ /* Parse CRLF or LF (always present) */
+ if (likely(end[ridx] == '\r'))
+ ++ridx;
+ if (end[ridx] != '\n') {
+ h1m->state = H1_MSG_CHUNK_CRLF;
+ goto parsing_error;
+ }
+ ++ridx;
+ }
+
+ end_parsing:
+ ridx = save;
+
+ /* Adjust the HTX block size or remove the block if nothing was copied
+ * (Empty HTX data block are not supported).
+ */
+ if (!dpos)
+ htx_remove_blk(*dsthtx, htxret.blk);
+ else
+ htx_change_blk_value_len(*dsthtx, htxret.blk, dpos);
+ total = end + ridx - start;
+ *max = lmax;
+
+ out:
+ return total;
+
+ parsing_error:
+ (*dsthtx)->flags |= HTX_FL_PARSING_ERROR;
+ h1m->err_state = h1m->state;
+ h1m->err_pos = ofs + end + ridx - start;
+ return 0;
+}
+
+/* Parse HTTP chunks. This function relies on an optimized function to parse
+ * contiguous chunks if possible. Otherwise, when a chunk is incomplete or when
+ * the underlying buffer is wrapping, a generic function is used.
+ */
static size_t h1_parse_msg_chunks(struct h1m *h1m, struct htx **dsthtx,
struct buffer *srcbuf, size_t ofs, size_t max,
struct buffer *htxbuf)
@@ -544,16 +741,36 @@
while (ofs < b_data(srcbuf)) {
ret = 0;
+ /* First parse full contiguous chunks. It is only possible if we
+ * are waiting for the next chunk size.
+ */
+ if (h1m->state == H1_MSG_CHUNK_SIZE) {
+ ret = h1_parse_full_contig_chunks(h1m, dsthtx, srcbuf, ofs, &max, htxbuf);
+ /* exit on error */
+ if (!ret && (*dsthtx)->flags & HTX_FL_PARSING_ERROR) {
+ total = 0;
+ break;
+ }
+ /* or let a chance to parse remaining data */
+ total += ret;
+ ofs += ret;
+ ret = 0;
+ }
+
- if (h1m->state < H1_MSG_TRAILERS) {
+ /* If some data remains, try to parse it using the generic
+ * function handling incomplete chunks and splitted chunks
+ * because of a wrapping buffer.
+ */
+ if (h1m->state < H1_MSG_TRAILERS && ofs < b_data(srcbuf)) {
ret = h1_parse_chunk(h1m, dsthtx, srcbuf, ofs, &max, htxbuf);
total += ret;
ofs += ret;
}
- /* nothing more was parsed, we can exit, handling parsing error
- * if necessary.
+ /* nothing more was parsed or parsing was stopped on incomplete
+ * chunk, we can exit, handling parsing error if necessary.
*/
- if (!ret) {
+ if (!ret || h1m->state != H1_MSG_CHUNK_SIZE) {
if ((*dsthtx)->flags & HTX_FL_PARSING_ERROR)
total = 0;
break;