QUIC: avoid excessive buffer allocations in stream output.

Previously, when a few bytes were send to a QUIC stream by the application, a
4K buffer was allocated for these bytes.  Then a STREAM frame was created and
that entire buffer was used as data for that frame.  The frame with the buffer
were in use up until the frame was acked by client.  Meanwhile, when more
bytes were send to the stream, more buffers were allocated and assigned as
data to newer STREAM frames.  In this scenario most buffer memory is unused.

Now the unused part of the stream output buffer is available for further
stream output while earlier parts of the buffer are waiting to be acked.
This is achieved by splitting the output buffer.

5 files changed, 33 insertions, 15 deletions

diff --git a/src/event/quic/ngx_event_quic.h b/src/event/quic/ngx_event_quic.h
index 8ae7bf643..9481fef62 100644
--- a/src/event/quic/ngx_event_quic.h
+++ b/src/event/quic/ngx_event_quic.h
@@ -63,6 +63,7 @@ struct ngx_quic_stream_s {
     uint64_t                   recv_last;
     uint64_t                   final_size;
     ngx_chain_t               *in;
+    ngx_chain_t               *out;
     ngx_uint_t                 cancelable;  /* unsigned  cancelable:1; */
 };
 
diff --git a/src/event/quic/ngx_event_quic_frames.c b/src/event/quic/ngx_event_quic_frames.c
index 4a3902f7f..71ed981e6 100644
--- a/src/event/quic/ngx_event_quic_frames.c
+++ b/src/event/quic/ngx_event_quic_frames.c
@@ -482,14 +482,14 @@ done:
 
 ngx_int_t
 ngx_quic_order_bufs(ngx_connection_t *c, ngx_chain_t **out, ngx_chain_t *in,
-    size_t offset)
+    off_t limit, off_t offset)
 {
+    off_t         n;
     u_char       *p;
-    size_t        n;
     ngx_buf_t    *b;
     ngx_chain_t  *cl, *sl;
 
-    while (in) {
+    while (in && limit) {
         cl = *out;
 
         if (cl == NULL) {
@@ -523,8 +523,9 @@ ngx_quic_order_bufs(ngx_connection_t *c, ngx_chain_t **out, ngx_chain_t *in,
             continue;
         }
 
-        for (p = b->pos + offset; p != b->last && in; /* void */ ) {
+        for (p = b->pos + offset; p != b->last && in && limit; /* void */ ) {
             n = ngx_min(b->last - p, in->buf->last - in->buf->pos);
+            n = ngx_min(n, limit);
 
             if (b->sync) {
                 ngx_memcpy(p, in->buf->pos, n);
@@ -533,6 +534,7 @@ ngx_quic_order_bufs(ngx_connection_t *c, ngx_chain_t **out, ngx_chain_t *in,
             p += n;
             in->buf->pos += n;
             offset += n;
+            limit -= n;
 
             if (in->buf->pos == in->buf->last) {
                 in = in->next;
diff --git a/src/event/quic/ngx_event_quic_frames.h b/src/event/quic/ngx_event_quic_frames.h
index f4c147682..1310f6dfd 100644
--- a/src/event/quic/ngx_event_quic_frames.h
+++ b/src/event/quic/ngx_event_quic_frames.h
@@ -31,7 +31,7 @@ ngx_chain_t *ngx_quic_copy_chain(ngx_connection_t *c, ngx_chain_t *in,
 void ngx_quic_trim_bufs(ngx_chain_t *in, size_t size);
 void ngx_quic_free_bufs(ngx_connection_t *c, ngx_chain_t *in);
 ngx_int_t ngx_quic_order_bufs(ngx_connection_t *c, ngx_chain_t **out,
-    ngx_chain_t *in, size_t offset);
+    ngx_chain_t *in, off_t limit, off_t offset);
 
 #if (NGX_DEBUG)
 void ngx_quic_log_frame(ngx_log_t *log, ngx_quic_frame_t *f, ngx_uint_t tx);
diff --git a/src/event/quic/ngx_event_quic_ssl.c b/src/event/quic/ngx_event_quic_ssl.c
index fb4b1af85..d2f1237a5 100644
--- a/src/event/quic/ngx_event_quic_ssl.c
+++ b/src/event/quic/ngx_event_quic_ssl.c
@@ -369,7 +369,7 @@ ngx_quic_handle_crypto_frame(ngx_connection_t *c, ngx_quic_header_t *pkt,
     }
 
     if (f->offset > ctx->crypto_received) {
-        return ngx_quic_order_bufs(c, &ctx->crypto, frame->data,
+        return ngx_quic_order_bufs(c, &ctx->crypto, frame->data, f->length,
                                    f->offset - ctx->crypto_received);
     }
 
diff --git a/src/event/quic/ngx_event_quic_streams.c b/src/event/quic/ngx_event_quic_streams.c
index 295aa54aa..d5facc270 100644
--- a/src/event/quic/ngx_event_quic_streams.c
+++ b/src/event/quic/ngx_event_quic_streams.c
@@ -838,8 +838,9 @@ static ngx_chain_t *
 ngx_quic_stream_send_chain(ngx_connection_t *c, ngx_chain_t *in, off_t limit)
 {
     size_t                  n, flow;
+    ngx_buf_t              *b;
     ngx_event_t            *wev;
-    ngx_chain_t            *cl;
+    ngx_chain_t            *out, **ll;
     ngx_connection_t       *pc;
     ngx_quic_frame_t       *frame;
     ngx_quic_stream_t      *qs;
@@ -862,18 +863,30 @@ ngx_quic_stream_send_chain(ngx_connection_t *c, ngx_chain_t *in, off_t limit)
 
     n = (limit && (size_t) limit < flow) ? (size_t) limit : flow;
 
-    frame = ngx_quic_alloc_frame(pc);
-    if (frame == NULL) {
+    if (ngx_quic_order_bufs(pc, &qs->out, in, n, 0) != NGX_OK) {
         return NGX_CHAIN_ERROR;
     }
 
-    frame->data = ngx_quic_copy_chain(pc, in, n);
-    if (frame->data == NGX_CHAIN_ERROR) {
-        return NGX_CHAIN_ERROR;
+    n = 0;
+    out = qs->out;
+
+    for (ll = &out; *ll; ll = &(*ll)->next) {
+        b = (*ll)->buf;
+
+        if (b->sync) {
+            /* hole */
+            break;
+        }
+
+        n += b->last - b->pos;
     }
 
-    for (n = 0, cl = frame->data; cl; cl = cl->next) {
-        n += ngx_buf_size(cl->buf);
+    qs->out = *ll;
+    *ll = NULL;
+
+    frame = ngx_quic_alloc_frame(pc);
+    if (frame == NULL) {
+        return NGX_CHAIN_ERROR;
     }
 
     while (in && ngx_buf_size(in->buf) == 0) {
@@ -882,6 +895,7 @@ ngx_quic_stream_send_chain(ngx_connection_t *c, ngx_chain_t *in, off_t limit)
 
     frame->level = ssl_encryption_application;
     frame->type = NGX_QUIC_FT_STREAM;
+    frame->data = out;
     frame->u.stream.off = 1;
     frame->u.stream.len = 1;
     frame->u.stream.fin = 0;
@@ -977,6 +991,7 @@ ngx_quic_stream_cleanup_handler(void *data)
 
     ngx_rbtree_delete(&qc->streams.tree, &qs->node);
     ngx_quic_free_bufs(pc, qs->in);
+    ngx_quic_free_bufs(pc, qs->out);
 
     if (qc->closing) {
         /* schedule handler call to continue ngx_quic_close_connection() */
@@ -1098,7 +1113,7 @@ ngx_quic_handle_stream_frame(ngx_connection_t *c, ngx_quic_header_t *pkt,
         qs->final_size = last;
     }
 
-    if (ngx_quic_order_bufs(c, &qs->in, frame->data,
+    if (ngx_quic_order_bufs(c, &qs->in, frame->data, f->length,
                             f->offset - qs->recv_offset)
         != NGX_OK)
     {