[master] fe2ad14 Finally split cache_center.c into a HTTP/1 and a Request Finite State Machine.

[Poul-Henning Kamp]

commit fe2ad14cdc379db7c5de9454ac1d8135849e1af3
Author: Poul-Henning Kamp <phk at FreeBSD.org>
Date:   Mon Aug 6 09:58:49 2012 +0000

    Finally split cache_center.c into a HTTP/1 and a Request Finite State Machine.

diff --git a/bin/varnishd/Makefile.am b/bin/varnishd/Makefile.am
index bd91dd7..967d3bb 100644
--- a/bin/varnishd/Makefile.am
+++ b/bin/varnishd/Makefile.am
@@ -17,7 +17,6 @@ varnishd_SOURCES = \
 	cache/cache_backend_poll.c \
 	cache/cache_ban.c \
 	cache/cache_busyobj.c \
-	cache/cache_center.c \
 	cache/cache_cli.c \
 	cache/cache_dir.c \
 	cache/cache_dir_dns.c \
@@ -31,6 +30,7 @@ varnishd_SOURCES = \
 	cache/cache_gzip.c \
 	cache/cache_hash.c \
 	cache/cache_http.c \
+	cache/cache_http1_fsm.c \
 	cache/cache_httpconn.c \
 	cache/cache_lck.c \
 	cache/cache_main.c \
@@ -38,6 +38,7 @@ varnishd_SOURCES = \
 	cache/cache_panic.c \
 	cache/cache_pipe.c \
 	cache/cache_pool.c \
+	cache/cache_req_fsm.c \
 	cache/cache_response.c \
 	cache/cache_rfc2616.c \
 	cache/cache_session.c \
diff --git a/bin/varnishd/cache/cache_center.c b/bin/varnishd/cache/cache_center.c
deleted file mode 100644
index dc7b4cc..0000000
--- a/bin/varnishd/cache/cache_center.c
+++ /dev/null
@@ -1,1662 +0,0 @@
-/*-
- * Copyright (c) 2006 Verdens Gang AS
- * Copyright (c) 2006-2011 Varnish Software AS
- * All rights reserved.
- *
- * Author: Poul-Henning Kamp <phk at phk.freebsd.dk>
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- *
- * This file contains the two central state machine for pushing
- * sessions and requests.
- *
- * The first part of the file, entrypoint CNT_Session() and down to
- * the ==== separator, is concerned with sessions.  When a session has
- * a request to deal with, it calls into the second half of the file.
- * This part is for all practical purposes HTTP/1.x specific.
- *
- * The second part of the file, entrypoint CNT_Request() and below the
- * ==== separator, is intended to (over time) be(ome) protocol agnostic.
- * We already use this now with ESI:includes, which are for all relevant
- * purposes a different "protocol"
- *
- * A special complication is the fact that we can suspend processing of
- * a request when hash-lookup finds a busy objhdr.
- *
- * Since the states are rather nasty in detail, I have decided to embedd
- * a dot(1) graph in the source code comments.  So to see the big picture,
- * extract the DOT lines and run though dot(1), for instance with the
- * command:
- *	sed -n '/^DOT/s///p' cache/cache_center.c | dot -Tps > /tmp/_.ps
- */
-
-/*
-DOT digraph vcl_center {
-xDOT	page="8.2,11.5"
-DOT	size="7.2,10.5"
-DOT	margin="0.5"
-DOT	center="1"
-DOT acceptor [
-DOT	shape=hexagon
-DOT	label="Request received"
-DOT ]
-DOT ERROR [shape=plaintext]
-DOT RESTART [shape=plaintext]
-DOT acceptor -> first [style=bold,color=green]
- */
-
-#include "config.h"
-
-#include <math.h>
-#include <poll.h>
-#include <stdio.h>
-#include <stdlib.h>
-
-#include "cache.h"
-
-#include "hash/hash_slinger.h"
-#include "vcl.h"
-#include "vcli_priv.h"
-#include "vsha256.h"
-#include "vtcp.h"
-#include "vtim.h"
-
-#ifndef HAVE_SRANDOMDEV
-#include "compat/srandomdev.h"
-#endif
-
-static unsigned xids;
-
-/*--------------------------------------------------------------------
- * WAIT
- * Collect the request from the client.
- *
-DOT subgraph xcluster_wait {
-DOT	wait [
-DOT		shape=box
-DOT		label="cnt_sess_wait:\nwait for\ncomplete\nrequest"
-DOT	]
-DOT	herding [shape=hexagon]
-DOT	wait -> start [label="got req",style=bold,color=green]
-DOT	wait -> "SES_Delete()" [label="errors"]
-DOT	wait -> herding [label="timeout_linger"]
-DOT	herding -> wait [label="fd read_ready"]
-DOT }
- */
-
-static int
-cnt_sess_wait(struct sess *sp, struct worker *wrk, struct req *req)
-{
-	int j, tmo;
-	struct pollfd pfd[1];
-	double now, when;
-	enum sess_close why = SC_NULL;
-	enum htc_status_e hs;
-
-	CHECK_OBJ_NOTNULL(sp, SESS_MAGIC);
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-
-	assert(req->sp == sp);
-
-
-	AZ(req->vcl);
-	AZ(req->obj);
-	AZ(req->esi_level);
-	assert(req->xid == 0);
-	assert(isnan(req->t_req));
-	assert(isnan(req->t_resp));
-
-	tmo = (int)(1e3 * cache_param->timeout_linger);
-	while (1) {
-		pfd[0].fd = sp->fd;
-		pfd[0].events = POLLIN;
-		pfd[0].revents = 0;
-		j = poll(pfd, 1, tmo);
-		assert(j >= 0);
-		now = VTIM_real();
-		if (j != 0)
-			hs = HTC_Rx(req->htc);
-		else
-			hs = HTC_Complete(req->htc);
-		if (hs == HTC_COMPLETE) {
-			/* Got it, run with it */
-			req->t_req = now;
-			return (0);
-		} else if (hs == HTC_ERROR_EOF) {
-			why = SC_REM_CLOSE;
-			break;
-		} else if (hs == HTC_OVERFLOW) {
-			why = SC_RX_OVERFLOW;
-			break;
-		} else if (hs == HTC_ALL_WHITESPACE) {
-			/* Nothing but whitespace */
-			when = sp->t_idle + cache_param->timeout_idle;
-			if (when < now) {
-				why = SC_RX_TIMEOUT;
-				break;
-			}
-			when = sp->t_idle + cache_param->timeout_linger;
-			tmo = (int)(1e3 * (when - now));
-			if (when < now || tmo == 0) {
-				req->t_req = NAN;
-				wrk->stats.sess_herd++;
-				SES_ReleaseReq(req);
-				WAIT_Enter(sp);
-				return (1);
-			}
-		} else {
-			/* Working on it */
-			if (isnan(req->t_req))
-				req->t_req = now;
-			when = req->t_req + cache_param->timeout_req;
-			tmo = (int)(1e3 * (when - now));
-			if (when < now || tmo == 0) {
-				why = SC_RX_TIMEOUT;
-				break;
-			}
-		}
-	}
-	SES_ReleaseReq(req);
-	assert(why != SC_NULL);
-	SES_Delete(sp, why, now);
-	return (1);
-}
-
-/*--------------------------------------------------------------------
- * This is the final state, figure out if we should close or recycle
- * the client connection
- *
-DOT	DONE [
-DOT		shape=record
-DOT		label="{cnt_done:|Request completed}"
-DOT	]
-DOT	ESI_RESP [ shape=hexagon ]
-DOT	DONE -> start [label="full pipeline"]
-DOT	DONE -> wait
-DOT	DONE -> ESI_RESP
- */
-
-enum cnt_sess_done_ret {
-	SESS_DONE_RET_GONE,
-	SESS_DONE_RET_WAIT,
-	SESS_DONE_RET_START,
-};
-
-static enum cnt_sess_done_ret
-cnt_sess_done(struct sess *sp, struct worker *wrk, struct req *req)
-{
-
-	CHECK_OBJ_NOTNULL(sp, SESS_MAGIC);
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	CHECK_OBJ_ORNULL(req->vcl, VCL_CONF_MAGIC);
-
-	AZ(req->obj);
-	AZ(req->busyobj);
-	req->director = NULL;
-	req->restarts = 0;
-
-	AZ(req->esi_level);
-
-	if (req->vcl != NULL) {
-		if (wrk->vcl != NULL)
-			VCL_Rel(&wrk->vcl);
-		wrk->vcl = req->vcl;
-		req->vcl = NULL;
-	}
-
-	sp->t_idle = W_TIM_real(wrk);
-	if (req->xid == 0) 
-		req->t_resp = sp->t_idle;
-	req->xid = 0;
-	VSL_Flush(req->vsl, 0);
-
-	req->t_req = NAN;
-	req->t_resp = NAN;
-
-	req->req_bodybytes = 0;
-
-	req->hash_always_miss = 0;
-	req->hash_ignore_busy = 0;
-
-	if (sp->fd >= 0 && req->doclose != SC_NULL)
-		SES_Close(sp, req->doclose);
-
-	if (sp->fd < 0) {
-		wrk->stats.sess_closed++;
-		AZ(req->vcl);
-		SES_ReleaseReq(req);
-		SES_Delete(sp, SC_NULL, NAN);
-		return (SESS_DONE_RET_GONE);
-	}
-
-	if (wrk->stats.client_req >= cache_param->wthread_stats_rate)
-		WRK_SumStat(wrk);
-
-	WS_Reset(req->ws, NULL);
-	WS_Reset(wrk->aws, NULL);
-	req->vxid = VXID_Get(&wrk->vxid_pool);
-
-	if (HTC_Reinit(req->htc) == HTC_COMPLETE) {
-		req->t_req = sp->t_idle;
-		wrk->stats.sess_pipeline++;
-		return (SESS_DONE_RET_START);
-	} else {
-		if (Tlen(req->htc->rxbuf))
-			wrk->stats.sess_readahead++;
-		return (SESS_DONE_RET_WAIT);
-	}
-}
-
-/*--------------------------------------------------------------------
- */
-
-void
-CNT_Session(struct worker *wrk, struct req *req)
-{
-	int done;
-	struct sess *sp;
-	enum cnt_sess_done_ret sdr;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	sp = req->sp;
-	CHECK_OBJ_NOTNULL(sp, SESS_MAGIC);
-
-	/*
-	 * Whenever we come in from the acceptor or waiter, we need to set
-	 * blocking mode, but there is no point in setting it when we come from
-	 * ESI or when a parked sessions returns.
-	 * It would be simpler to do this in the acceptor or waiter, but we'd
-	 * rather do the syscall in the worker thread.
-	 * On systems which return errors for ioctl, we close early
-	 */
-	if (sp->sess_step == S_STP_NEWREQ && VTCP_blocking(sp->fd)) {
-		if (errno == ECONNRESET)
-			SES_Close(sp, SC_REM_CLOSE);
-		else
-			SES_Close(sp, SC_TX_ERROR);
-		sdr = cnt_sess_done(sp, wrk, req);
-		assert(sdr == SESS_DONE_RET_GONE);
-		return;
-	}
-
-	if (sp->sess_step == S_STP_NEWREQ) {
-		HTC_Init(req->htc, req->ws, sp->fd, req->vsl,
-		    cache_param->http_req_size,
-		    cache_param->http_req_hdr_len);
-	}
-
-	while (1) {
-		/*
-		 * Possible entrance states
-		 */
-
-		assert(
-		    sp->sess_step == S_STP_NEWREQ ||
-		    req->req_step == R_STP_LOOKUP ||
-		    req->req_step == R_STP_START);
-
-		if (sp->sess_step == S_STP_WORKING) {
-			done = CNT_Request(wrk, req);
-			if (done == 2)
-				return;
-			assert(done == 1);
-			sdr = cnt_sess_done(sp, wrk, req);
-			switch (sdr) {
-			case SESS_DONE_RET_GONE:
-				return;
-			case SESS_DONE_RET_WAIT:
-				sp->sess_step = S_STP_NEWREQ;
-				break;
-			case SESS_DONE_RET_START:
-				sp->sess_step = S_STP_WORKING;
-				req->req_step = R_STP_START;
-				break;
-			default:
-				WRONG("Illegal enum cnt_sess_done_ret");
-			}
-		}
-
-		if (sp->sess_step == S_STP_NEWREQ) {
-			done = cnt_sess_wait(sp, wrk, req);
-			if (done)
-				return;
-			sp->sess_step = S_STP_WORKING;
-			req->req_step = R_STP_START;
-		}
-	}
-}
-
-/*====================================================================*/
-
-
-/*--------------------------------------------------------------------
- * We have a refcounted object on the session, and possibly the busyobj
- * which is fetching it, prepare a response.
- *
-DOT subgraph xcluster_prepresp {
-DOT	prepresp [
-DOT		shape=record
-DOT		label="{cnt_prepresp:|Filter obj.-\>resp.|{vcl_deliver\{\}|{req.|resp.}}|{error?|restart?}|stream ?}"
-DOT	]
-DOT	prepresp -> deliver [style=bold,color=green,label=deliver]
-DOT	prepresp -> deliver [style=bold,color=red]
-DOT	prepresp -> deliver [style=bold,color=blue]
-DOT }
- *
- */
-
-static int
-cnt_prepresp(struct worker *wrk, struct req *req)
-{
-	struct busyobj *bo;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	bo = req->busyobj;
-	CHECK_OBJ_ORNULL(bo, BUSYOBJ_MAGIC);
-
-	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-
-	req->res_mode = 0;
-
-	if (bo == NULL) {
-		if (!req->disable_esi && req->obj->esidata != NULL) {
-			/* In ESI mode, we can't know the aggregate length */
-			req->res_mode &= ~RES_LEN;
-			req->res_mode |= RES_ESI;
-		} else {
-			req->res_mode |= RES_LEN;
-		}
-	} else {
-		AZ(bo->do_esi);
-	}
-
-	if (req->esi_level > 0) {
-		/* Included ESI object, always CHUNKED or EOF */
-		req->res_mode &= ~RES_LEN;
-		req->res_mode |= RES_ESI_CHILD;
-	}
-
-	if (cache_param->http_gzip_support && req->obj->gziped &&
-	    !RFC2616_Req_Gzip(req->http)) {
-		/*
-		 * We don't know what it uncompresses to
-		 * XXX: we could cache that
-		 */
-		req->res_mode &= ~RES_LEN;
-		req->res_mode |= RES_GUNZIP;
-	}
-
-	if (!(req->res_mode & (RES_LEN|RES_CHUNKED|RES_EOF))) {
-		/* We havn't chosen yet, do so */
-		if (!req->wantbody) {
-			/* Nothing */
-		} else if (req->http->protover >= 11) {
-			req->res_mode |= RES_CHUNKED;
-		} else {
-			req->res_mode |= RES_EOF;
-			req->doclose = SC_TX_EOF;
-		}
-	}
-
-	req->t_resp = W_TIM_real(wrk);
-	if (req->obj->objcore->objhead != NULL) {
-		if ((req->t_resp - req->obj->last_lru) >
-		    cache_param->lru_timeout &&
-		    EXP_Touch(req->obj->objcore))
-			req->obj->last_lru = req->t_resp;
-		if (!cache_param->obj_readonly)
-			req->obj->last_use = req->t_resp; /* XXX: locking ? */
-	}
-	HTTP_Setup(req->resp, req->ws, req->vsl, HTTP_Resp);
-	RES_BuildHttp(req);
-
-	VCL_deliver_method(req);
-	switch (req->handling) {
-	case VCL_RET_DELIVER:
-		break;
-	case VCL_RET_RESTART:
-		if (req->restarts >= cache_param->max_restarts)
-			break;
-		if (bo != NULL) {
-			AN(bo->do_stream);
-			(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
-			VBO_DerefBusyObj(wrk, &req->busyobj);
-		} else {
-			(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
-		}
-		AZ(req->obj);
-		http_Teardown(req->resp);
-		req->req_step = R_STP_RESTART;
-		return (0);
-	default:
-		WRONG("Illegal action in vcl_deliver{}");
-	}
-	req->req_step = R_STP_DELIVER;
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * Deliver an already stored object
- *
-DOT subgraph xcluster_deliver {
-DOT	deliver [
-DOT		shape=record
-DOT		label="{cnt_deliver:|Send body}"
-DOT	]
-DOT }
-DOT deliver -> DONE [style=bold,color=green]
-DOT deliver -> DONE [style=bold,color=red]
-DOT deliver -> DONE [style=bold,color=blue]
- *
- */
-
-static int
-cnt_deliver(struct worker *wrk, struct req *req)
-{
-	struct busyobj *bo;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
-	bo = req->busyobj;
-	CHECK_OBJ_ORNULL(bo, BUSYOBJ_MAGIC);
-
-	if (bo != NULL) {
-		while (bo->state < BOS_FAILED)
-			(void)usleep(10000);
-		assert(bo->state >= BOS_FAILED);
-
-		if (bo->state == BOS_FAILED) {
-			HSH_Deref(&wrk->stats, NULL, &req->obj);
-			VBO_DerefBusyObj(wrk, &req->busyobj);
-			req->err_code = 503;
-			req->req_step = R_STP_ERROR;
-			return (0);
-		}
-		VBO_DerefBusyObj(wrk, &req->busyobj);
-	}
-
-	AZ(req->busyobj);
-	req->director = NULL;
-	req->restarts = 0;
-
-	RES_WriteObj(req);
-
-	/* No point in saving the body if it is hit-for-pass */
-	if (req->obj->objcore->flags & OC_F_PASS)
-		STV_Freestore(req->obj);
-
-	assert(WRW_IsReleased(wrk));
-	(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
-	http_Teardown(req->resp);
-	return (1);
-}
-/*--------------------------------------------------------------------
- * Emit an error
- *
-DOT subgraph xcluster_error {
-DOT	vcl_error [
-DOT		shape=record
-DOT		label="vcl_error()|resp."
-DOT	]
-DOT	ERROR -> vcl_error
-DOT	vcl_error-> prepresp [label=deliver]
-DOT }
-DOT vcl_error-> rsterr [label="restart",color=purple]
-DOT rsterr [label="RESTART",shape=plaintext]
- */
-
-static int
-cnt_error(struct worker *wrk, struct req *req)
-{
-	struct http *h;
-	struct busyobj *bo;
-	char date[40];
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	AZ(req->objcore);
-	AZ(req->obj);
-	AZ(req->busyobj);
-
-	bo = VBO_GetBusyObj(wrk);
-	req->busyobj = bo;
-	bo->vsl->wid = req->sp->vsl_id;
-	AZ(bo->stats);
-	bo->stats = &wrk->stats;
-	req->objcore = HSH_NewObjCore(wrk);
-	req->obj = STV_NewObject(bo, &req->objcore,
-	    TRANSIENT_STORAGE, cache_param->http_resp_size,
-	    (uint16_t)cache_param->http_max_hdr);
-	bo->stats = NULL;
-	if (req->obj == NULL) {
-		req->doclose = SC_OVERLOAD;
-		req->director = NULL;
-		http_Teardown(bo->beresp);
-		http_Teardown(bo->bereq);
-		return(1);
-	}
-	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
-	req->obj->xid = req->xid;
-	req->obj->exp.entered = req->t_req;
-
-	h = req->obj->http;
-
-	if (req->err_code < 100 || req->err_code > 999)
-		req->err_code = 501;
-
-	http_PutProtocol(h, "HTTP/1.1");
-	http_PutStatus(h, req->err_code);
-	VTIM_format(W_TIM_real(wrk), date);
-	http_PrintfHeader(h, "Date: %s", date);
-	http_SetHeader(h, "Server: Varnish");
-
-	if (req->err_reason != NULL)
-		http_PutResponse(h, req->err_reason);
-	else
-		http_PutResponse(h, http_StatusMessage(req->err_code));
-	VCL_error_method(req);
-
-	if (req->handling == VCL_RET_RESTART &&
-	    req->restarts <  cache_param->max_restarts) {
-		HSH_Drop(wrk, &req->obj);
-		VBO_DerefBusyObj(wrk, &req->busyobj);
-		req->req_step = R_STP_RESTART;
-		return (0);
-	} else if (req->handling == VCL_RET_RESTART)
-		req->handling = VCL_RET_DELIVER;
-
-
-	/* We always close when we take this path */
-	req->doclose = SC_TX_ERROR;
-	req->wantbody = 1;
-
-	assert(req->handling == VCL_RET_DELIVER);
-	req->err_code = 0;
-	req->err_reason = NULL;
-	http_Teardown(bo->bereq);
-	VBO_DerefBusyObj(wrk, &req->busyobj);
-	req->req_step = R_STP_PREPRESP;
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * Fetch response headers from the backend
- *
-DOT subgraph xcluster_fetch {
-DOT	fetch [
-DOT		shape=record
-DOT		label="{cnt_fetch:|fetch hdr\nfrom backend|(find obj.ttl)|{vcl_fetch\{\}|{req.|bereq.|beresp.}}|{<err>error?|<rst>restart?}}"
-DOT	]
-DOT }
-DOT fetch -> fetchbody [style=bold,color=red]
-DOT fetch -> fetchbody [style=bold,color=blue]
- */
-
-static int
-cnt_fetch(struct worker *wrk, struct req *req)
-{
-	int i, need_host_hdr;
-	struct busyobj *bo;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-	bo = req->busyobj;
-	CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
-
-	AN(req->director);
-	AZ(bo->vbc);
-	AZ(bo->should_close);
-	AZ(req->storage_hint);
-
-	HTTP_Setup(bo->beresp, bo->ws, bo->vsl, HTTP_Beresp);
-
-	need_host_hdr = !http_GetHdr(bo->bereq, H_Host, NULL);
-
-	req->acct_req.fetch++;
-
-	i = FetchHdr(req, need_host_hdr, req->objcore->objhead == NULL);
-	/*
-	 * If we recycle a backend connection, there is a finite chance
-	 * that the backend closed it before we get a request to it.
-	 * Do a single retry in that case.
-	 */
-	if (i == 1) {
-		VSC_C_main->backend_retry++;
-		i = FetchHdr(req, need_host_hdr, req->objcore->objhead == NULL);
-	}
-
-	if (i) {
-		req->handling = VCL_RET_ERROR;
-		req->err_code = 503;
-	} else {
-		/*
-		 * These two headers can be spread over multiple actual headers
-		 * and we rely on their content outside of VCL, so collect them
-		 * into one line here.
-		 */
-		http_CollectHdr(bo->beresp, H_Cache_Control);
-		http_CollectHdr(bo->beresp, H_Vary);
-
-		/*
-		 * Figure out how the fetch is supposed to happen, before the
-		 * headers are adultered by VCL
-		 * NB: Also sets other wrk variables
-		 */
-		bo->body_status = RFC2616_Body(bo, &wrk->stats);
-
-		req->err_code = http_GetStatus(bo->beresp);
-
-		/*
-		 * What does RFC2616 think about TTL ?
-		 */
-		EXP_Clr(&bo->exp);
-		bo->exp.entered = W_TIM_real(wrk);
-		RFC2616_Ttl(bo, req->xid);
-
-		/* pass from vclrecv{} has negative TTL */
-		if (req->objcore->objhead == NULL)
-			bo->exp.ttl = -1.;
-
-		AZ(bo->do_esi);
-		AZ(bo->do_pass);
-
-		VCL_fetch_method(req);
-
-		if (bo->do_pass)
-			req->objcore->flags |= OC_F_PASS;
-
-		switch (req->handling) {
-		case VCL_RET_DELIVER:
-			req->req_step = R_STP_FETCHBODY;
-			return (0);
-		default:
-			break;
-		}
-
-		/* We are not going to fetch the body, Close the connection */
-		VDI_CloseFd(&bo->vbc);
-	}
-
-	/* Clean up partial fetch */
-	AZ(bo->vbc);
-
-	if (req->objcore->objhead != NULL || req->handling == VCL_RET_ERROR) {
-		CHECK_OBJ_NOTNULL(req->objcore, OBJCORE_MAGIC);
-		AZ(HSH_Deref(&wrk->stats, req->objcore, NULL));
-		req->objcore = NULL;
-	}
-	assert(bo->refcount == 2);
-	VBO_DerefBusyObj(wrk, &bo);
-	VBO_DerefBusyObj(wrk, &req->busyobj);
-	req->director = NULL;
-	req->storage_hint = NULL;
-
-	switch (req->handling) {
-	case VCL_RET_RESTART:
-		req->req_step = R_STP_RESTART;
-		return (0);
-	case VCL_RET_ERROR:
-		req->req_step = R_STP_ERROR;
-		return (0);
-	default:
-		WRONG("Illegal action in vcl_fetch{}");
-	}
-}
-
-/*--------------------------------------------------------------------
- * Prepare to fetch body from backend
- *
-DOT subgraph xcluster_body {
-DOT	fetchbody [
-DOT		shape=record
-DOT		label="{cnt_fetchbody:|start fetch_thread}"
-DOT	]
-DOT }
-DOT fetchbody:out -> prepresp [style=bold,color=red]
-DOT fetchbody:out -> prepresp [style=bold,color=blue]
- */
-
-static int
-cnt_fetchbody(struct worker *wrk, struct req *req)
-{
-	struct http *hp, *hp2;
-	char *b;
-	uint16_t nhttp;
-	unsigned l;
-	struct vsb *vary = NULL;
-	int varyl = 0, pass;
-	struct busyobj *bo;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	bo = req->busyobj;
-	CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
-
-	assert(req->handling == VCL_RET_DELIVER);
-
-	if (req->objcore->objhead == NULL) {
-		/* This is a pass from vcl_recv */
-		pass = 1;
-		/* VCL may have fiddled this, but that doesn't help */
-		bo->exp.ttl = -1.;
-	} else if (bo->do_pass) {
-		pass = 1;
-	} else {
-		/* regular object */
-		pass = 0;
-	}
-
-	/*
-	 * The VCL variables beresp.do_g[un]zip tells us how we want the
-	 * object processed before it is stored.
-	 *
-	 * The backend Content-Encoding header tells us what we are going
-	 * to receive, which we classify in the following three classes:
-	 *
-	 *	"Content-Encoding: gzip"	--> object is gzip'ed.
-	 *	no Content-Encoding		--> object is not gzip'ed.
-	 *	anything else			--> do nothing wrt gzip
-	 *
-	 */
-
-	/* We do nothing unless the param is set */
-	if (!cache_param->http_gzip_support)
-		bo->do_gzip = bo->do_gunzip = 0;
-
-	bo->is_gzip = http_HdrIs(bo->beresp, H_Content_Encoding, "gzip");
-
-	bo->is_gunzip = !http_GetHdr(bo->beresp, H_Content_Encoding, NULL);
-
-	/* It can't be both */
-	assert(bo->is_gzip == 0 || bo->is_gunzip == 0);
-
-	/* We won't gunzip unless it is gzip'ed */
-	if (bo->do_gunzip && !bo->is_gzip)
-		bo->do_gunzip = 0;
-
-	/* If we do gunzip, remove the C-E header */
-	if (bo->do_gunzip)
-		http_Unset(bo->beresp, H_Content_Encoding);
-
-	/* We wont gzip unless it is ungziped */
-	if (bo->do_gzip && !bo->is_gunzip)
-		bo->do_gzip = 0;
-
-	/* If we do gzip, add the C-E header */
-	if (bo->do_gzip)
-		http_SetHeader(bo->beresp, "Content-Encoding: gzip");
-
-	/* But we can't do both at the same time */
-	assert(bo->do_gzip == 0 || bo->do_gunzip == 0);
-
-	/* ESI takes precedence and handles gzip/gunzip itself */
-	if (bo->do_esi)
-		bo->vfp = &vfp_esi;
-	else if (bo->do_gunzip)
-		bo->vfp = &vfp_gunzip;
-	else if (bo->do_gzip)
-		bo->vfp = &vfp_gzip;
-	else if (bo->is_gzip)
-		bo->vfp = &vfp_testgzip;
-
-	if (bo->do_esi || req->esi_level > 0)
-		bo->do_stream = 0;
-	if (!req->wantbody)
-		bo->do_stream = 0;
-
-	/* No reason to try streaming a non-existing body */
-	if (bo->body_status == BS_NONE)
-		bo->do_stream = 0;
-
-	l = http_EstimateWS(bo->beresp,
-	    pass ? HTTPH_R_PASS : HTTPH_A_INS, &nhttp);
-
-	/* Create Vary instructions */
-	if (req->objcore->objhead != NULL) {
-		CHECK_OBJ_NOTNULL(req->objcore, OBJCORE_MAGIC);
-		vary = VRY_Create(req, bo->beresp);
-		if (vary != NULL) {
-			varyl = VSB_len(vary);
-			assert(varyl > 0);
-			l += varyl;
-		}
-	}
-
-	/*
-	 * Space for producing a Content-Length: header including padding
-	 * A billion gigabytes is enough for anybody.
-	 */
-	l += strlen("Content-Length: XxxXxxXxxXxxXxxXxx") + sizeof(void *);
-
-	if (bo->exp.ttl < cache_param->shortlived ||
-	    req->objcore == NULL)
-		req->storage_hint = TRANSIENT_STORAGE;
-
-	AZ(bo->stats);
-	bo->stats = &wrk->stats;
-	req->obj = STV_NewObject(bo, &req->objcore, req->storage_hint, l,
-	    nhttp);
-	if (req->obj == NULL) {
-		/*
-		 * Try to salvage the transaction by allocating a
-		 * shortlived object on Transient storage.
-		 */
-		if (bo->exp.ttl > cache_param->shortlived)
-			bo->exp.ttl = cache_param->shortlived;
-		bo->exp.grace = 0.0;
-		bo->exp.keep = 0.0;
-		req->obj = STV_NewObject(bo, &req->objcore, TRANSIENT_STORAGE,
-		    l, nhttp);
-	}
-	bo->stats = NULL;
-	if (req->obj == NULL) {
-		req->err_code = 503;
-		req->req_step = R_STP_ERROR;
-		VDI_CloseFd(&bo->vbc);
-		VBO_DerefBusyObj(wrk, &req->busyobj);
-		return (0);
-	}
-	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
-
-	req->storage_hint = NULL;
-
-	AZ(bo->fetch_obj);
-	bo->fetch_obj = req->obj;
-
-	if (bo->do_gzip || (bo->is_gzip && !bo->do_gunzip))
-		req->obj->gziped = 1;
-
-	if (vary != NULL) {
-		req->obj->vary = (void *)WS_Alloc(req->obj->http->ws, varyl);
-		AN(req->obj->vary);
-		memcpy(req->obj->vary, VSB_data(vary), varyl);
-		VRY_Validate(req->obj->vary);
-		VSB_delete(vary);
-	}
-
-	req->obj->xid = req->xid;
-	req->obj->response = req->err_code;
-	WS_Assert(req->obj->ws_o);
-
-	/* Filter into object */
-	hp = bo->beresp;
-	hp2 = req->obj->http;
-
-	hp2->logtag = HTTP_Obj;
-	http_FilterResp(hp, hp2, pass ? HTTPH_R_PASS : HTTPH_A_INS);
-	http_CopyHome(hp2);
-
-	if (http_GetHdr(hp, H_Last_Modified, &b))
-		req->obj->last_modified = VTIM_parse(b);
-	else
-		req->obj->last_modified = floor(bo->exp.entered);
-
-	assert(WRW_IsReleased(wrk));
-
-	/*
-	 * If we can deliver a 304 reply, we don't bother streaming.
-	 * Notice that vcl_deliver{} could still nuke the headers
-	 * that allow the 304, in which case we return 200 non-stream.
-	 */
-	if (req->obj->response == 200 &&
-	    req->http->conds &&
-	    RFC2616_Do_Cond(req))
-		bo->do_stream = 0;
-
-	/*
-	 * Ready to fetch the body
-	 */
-	bo->fetch_task.func = FetchBody;
-	bo->fetch_task.priv = bo;
-
-	assert(bo->refcount == 2);	/* one for each thread */
-
-	if (req->obj->objcore->objhead != NULL) {
-		EXP_Insert(req->obj);
-		AN(req->obj->objcore->ban);
-		AZ(req->obj->ws_o->overflow);
-		HSH_Unbusy(&wrk->stats, req->obj->objcore);
-	}
-
-	if (!bo->do_stream ||
-	    Pool_Task(wrk->pool, &bo->fetch_task, POOL_NO_QUEUE))
-		FetchBody(wrk, bo);
-
-	if (req->obj->objcore->objhead != NULL)
-		HSH_Ref(req->obj->objcore);
-
-	if (bo->state == BOS_FINISHED) {
-		VBO_DerefBusyObj(wrk, &req->busyobj);
-	} else if (bo->state == BOS_FAILED) {
-		/* handle early failures */
-		HSH_Deref(&wrk->stats, NULL, &req->obj);
-		VBO_DerefBusyObj(wrk, &req->busyobj);
-		req->err_code = 503;
-		req->req_step = R_STP_ERROR;
-		return (0);
-	}
-
-	assert(WRW_IsReleased(wrk));
-	req->req_step = R_STP_PREPRESP;
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * HIT
- * We had a cache hit.  Ask VCL, then march off as instructed.
- *
-DOT subgraph xcluster_hit {
-DOT	hit [
-DOT		shape=record
-DOT		label="{cnt_hit:|{vcl_hit()|{req.|obj.}}|{<err>error?|<rst>restart?}|{<del>deliver?|<pass>pass?}}"
-DOT	]
-DOT }
-XDOT hit:err -> err_hit [label="error"]
-XDOT err_hit [label="ERROR",shape=plaintext]
-XDOT hit:rst -> rst_hit [label="restart",color=purple]
-XDOT rst_hit [label="RESTART",shape=plaintext]
-DOT hit:pass -> pass [label=pass,style=bold,color=red]
-DOT hit:del -> prepresp [label="deliver",style=bold,color=green]
- */
-
-static int
-cnt_hit(struct worker *wrk, struct req *req)
-{
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-
-	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-	AZ(req->objcore);
-	AZ(req->busyobj);
-
-	assert(!(req->obj->objcore->flags & OC_F_PASS));
-
-	VCL_hit_method(req);
-
-	if (req->handling == VCL_RET_DELIVER) {
-		//AZ(req->busyobj->bereq->ws);
-		//AZ(req->busyobj->beresp->ws);
-		(void)FetchReqBody(req, 0);
-		req->req_step = R_STP_PREPRESP;
-		return (0);
-	}
-
-	/* Drop our object, we won't need it */
-	(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
-	req->objcore = NULL;
-
-	switch(req->handling) {
-	case VCL_RET_PASS:
-		req->req_step = R_STP_PASS;
-		return (0);
-	case VCL_RET_ERROR:
-		req->req_step = R_STP_ERROR;
-		return (0);
-	case VCL_RET_RESTART:
-		req->req_step = R_STP_RESTART;
-		return (0);
-	default:
-		WRONG("Illegal action in vcl_hit{}");
-	}
-}
-
-/*--------------------------------------------------------------------
- * LOOKUP
- * Hash things together and look object up in hash-table.
- *
- * LOOKUP consists of two substates so that we can reenter if we
- * encounter a busy object.
- *
-DOT subgraph xcluster_lookup {
-DOT	lookup [
-DOT		shape=record
-DOT		label="{<top>cnt_lookup:|hash lookup|{<busy>busy ?|<miss>miss ?}|{<no>no|obj.f.pass?|<yes>yes}}"
-DOT	]
-DOT }
-DOT lookup:busy -> lookup:top [label="(waitinglist)"]
-DOT lookup:miss -> miss [style=bold,color=blue]
-DOT lookup:no -> hit [style=bold,color=green]
-DOT lookup:yes -> pass [style=bold,color=red]
- */
-
-static int
-cnt_lookup(struct worker *wrk, struct req *req)
-{
-	struct objcore *oc;
-	struct object *o;
-	struct objhead *oh;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	AZ(req->objcore);
-
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-	AZ(req->busyobj);
-
-	VRY_Prep(req);
-
-	AZ(req->objcore);
-	oc = HSH_Lookup(req);
-	if (oc == NULL) {
-		/*
-		 * We lost the session to a busy object, disembark the
-		 * worker thread.   We return to STP_LOOKUP when the busy
-		 * object has been unbusied, and still have the hash digest
-		 * around to do the lookup with.
-		 */
-		return (2);
-	}
-	AZ(req->objcore);
-
-	CHECK_OBJ_NOTNULL(oc, OBJCORE_MAGIC);
-	oh = oc->objhead;
-	CHECK_OBJ_NOTNULL(oh, OBJHEAD_MAGIC);
-
-	/* If we inserted a new object it's a miss */
-	if (oc->flags & OC_F_BUSY) {
-		CHECK_OBJ_NOTNULL(oc->busyobj, BUSYOBJ_MAGIC);
-		assert(oc->busyobj == req->busyobj);
-		wrk->stats.cache_miss++;
-
-		if (req->vary_l != NULL) {
-			assert(oc->busyobj->vary == req->vary_b);
-			VRY_Validate(oc->busyobj->vary);
-			WS_ReleaseP(req->ws, (void*)req->vary_l);
-		} else {
-			AZ(oc->busyobj->vary);
-			WS_Release(req->ws, 0);
-		}
-		req->vary_b = NULL;
-		req->vary_l = NULL;
-		req->vary_e = NULL;
-
-		req->objcore = oc;
-		req->req_step = R_STP_MISS;
-		return (0);
-	}
-
-	/* We are not prepared to do streaming yet */
-	XXXAZ(req->busyobj);
-
-	o = oc_getobj(&wrk->stats, oc);
-	CHECK_OBJ_NOTNULL(o, OBJECT_MAGIC);
-	req->obj = o;
-
-	WS_Release(req->ws, 0);
-	req->vary_b = NULL;
-	req->vary_l = NULL;
-	req->vary_e = NULL;
-
-	if (oc->flags & OC_F_PASS) {
-		wrk->stats.cache_hitpass++;
-		VSLb(req->vsl, SLT_HitPass, "%u", req->obj->xid);
-		(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
-		AZ(req->objcore);
-		req->req_step = R_STP_PASS;
-		return (0);
-	}
-
-	wrk->stats.cache_hit++;
-	VSLb(req->vsl, SLT_Hit, "%u", req->obj->xid);
-	req->req_step = R_STP_HIT;
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * We had a miss, ask VCL, proceed as instructed
- *
-DOT subgraph xcluster_miss {
-DOT	miss [
-DOT		shape=record
-DOT		label="{cnt_miss:|filter req.-\>bereq.|{vcl_miss\{\}|{req.*|bereq.*}}|{<err>error?|<rst>restart?}|{<pass>pass?|<fetch>fetch?}}"
-DOT	]
-DOT }
-DOT miss:fetch -> fetch [label="fetch",style=bold,color=blue]
-DOT miss:pass -> pass [label="pass",style=bold,color=red]
-DOT
- */
-
-static int
-cnt_miss(struct worker *wrk, struct req *req)
-{
-	struct busyobj *bo;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-	CHECK_OBJ_NOTNULL(req->objcore, OBJCORE_MAGIC);
-	bo = req->busyobj;
-	CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
-	AZ(req->obj);
-
-	HTTP_Setup(bo->bereq, bo->ws, bo->vsl, HTTP_Bereq);
-	http_FilterReq(req, HTTPH_R_FETCH);
-	http_ForceGet(bo->bereq);
-	if (cache_param->http_gzip_support) {
-		/*
-		 * We always ask the backend for gzip, even if the
-		 * client doesn't grok it.  We will uncompress for
-		 * the minority of clients which don't.
-		 */
-		http_Unset(bo->bereq, H_Accept_Encoding);
-		http_SetHeader(bo->bereq, "Accept-Encoding: gzip");
-	}
-
-	VCL_miss_method(req);
-
-	if (req->handling == VCL_RET_FETCH) {
-		CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
-		req->req_step = R_STP_FETCH;
-		return (0);
-	}
-
-	AZ(HSH_Deref(&wrk->stats, req->objcore, NULL));
-	req->objcore = NULL;
-	http_Teardown(bo->bereq);
-	VBO_DerefBusyObj(wrk, &req->busyobj);
-
-	switch(req->handling) {
-	case VCL_RET_ERROR:
-		req->req_step = R_STP_ERROR;
-		break;
-	case VCL_RET_PASS:
-		req->req_step = R_STP_PASS;
-		break;
-	case VCL_RET_RESTART:
-		req->req_step = R_STP_RESTART;
-		break;
-	default:
-		WRONG("Illegal action in vcl_miss{}");
-	}
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * Start pass processing by getting headers from backend, then
- * continue in passbody.
- *
-DOT subgraph xcluster_pass {
-DOT	pass [
-DOT		shape=record
-DOT		label="{cnt_pass:|(XXX: deref obj.)|filter req.*-\>bereq.|{vcl_pass\{\}|{req.*|bereq.*}}|{<err>error?|<rst>restart?}|<pass>create anon obj}"
-DOT	]
-DOT }
-DOT pass:pass -> fetch [style=bold, color=red]
-XDOT pass:rst -> rst_pass [label="restart",color=purple]
-XDOT rst_pass [label="RESTART",shape=plaintext]
-XDOT pass:err -> err_pass [label="error"]
-XDOT err_pass [label="ERROR",shape=plaintext]
- */
-
-static int
-cnt_pass(struct worker *wrk, struct req *req)
-{
-	struct busyobj *bo;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-	AZ(req->objcore);
-	AZ(req->obj);
-	AZ(req->busyobj);
-
-	req->busyobj = VBO_GetBusyObj(wrk);
-	bo = req->busyobj;
-	bo->vsl->wid = req->sp->vsl_id;
-	bo->refcount = 2;
-	HTTP_Setup(bo->bereq, bo->ws, bo->vsl, HTTP_Bereq);
-	http_FilterReq(req, HTTPH_R_PASS);
-
-	VCL_pass_method(req);
-
-	if (req->handling == VCL_RET_ERROR) {
-		http_Teardown(bo->bereq);
-		VBO_DerefBusyObj(wrk, &req->busyobj);
-		req->req_step = R_STP_ERROR;
-		return (0);
-	}
-	assert(req->handling == VCL_RET_PASS);
-	req->acct_req.pass++;
-	req->req_step = R_STP_FETCH;
-
-	req->objcore = HSH_NewObjCore(wrk);
-	req->objcore->busyobj = bo;
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * Ship the request header to the backend unchanged, then pipe
- * until one of the ends close the connection.
- *
-DOT subgraph xcluster_pipe {
-DOT	pipe [
-DOT		shape=ellipse
-DOT		label="Filter req.->bereq."
-DOT	]
-DOT	vcl_pipe [
-DOT		shape=record
-DOT		label="vcl_pipe()|req.\nbereq\."
-DOT	]
-DOT	pipe_do [
-DOT		shape=ellipse
-DOT		label="send bereq.\npipe until close"
-DOT	]
-DOT	vcl_pipe -> pipe_do [label="pipe",style=bold,color=orange]
-DOT	pipe -> vcl_pipe [style=bold,color=orange]
-DOT }
-DOT pipe_do -> DONE [style=bold,color=orange]
-DOT vcl_pipe -> err_pipe [label="error"]
-DOT err_pipe [label="ERROR",shape=plaintext]
- */
-
-static int
-cnt_pipe(struct worker *wrk, struct req *req)
-{
-	struct busyobj *bo;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-	AZ(req->busyobj);
-
-	req->acct_req.pipe++;
-	req->busyobj = VBO_GetBusyObj(wrk);
-	bo = req->busyobj;
-	bo->vsl->wid = req->sp->vsl_id;
-	HTTP_Setup(bo->bereq, bo->ws, bo->vsl, HTTP_Bereq);
-	http_FilterReq(req, 0);
-
-	VCL_pipe_method(req);
-
-	if (req->handling == VCL_RET_ERROR)
-		INCOMPL();
-	assert(req->handling == VCL_RET_PIPE);
-
-	PipeRequest(req);
-	assert(WRW_IsReleased(wrk));
-	http_Teardown(bo->bereq);
-	VBO_DerefBusyObj(wrk, &req->busyobj);
-	return (1);
-}
-
-/*--------------------------------------------------------------------
- *
-DOT subgraph xcluster_restart {
-DOT	restart [
-DOT		shape=record
-DOT		label="{cnt_restart}"
-DOT	]
-DOT }
-DOT RESTART -> restart [color=purple]
-DOT restart -> recv [color=purple]
- */
-
-static int
-cnt_restart(const struct worker *wrk, struct req *req)
-{
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-
-	req->director = NULL;
-	if (++req->restarts >= cache_param->max_restarts) {
-		req->err_code = 503;
-		req->req_step = R_STP_ERROR;
-	} else {
-		req->err_code = 0;
-		req->req_step = R_STP_RECV;
-	}
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * RECV
- * We have a complete request, set everything up and start it.
- * We can come here both with a request from the client and with
- * a interior request during ESI delivery.
- *
-DOT subgraph xcluster_recv {
-DOT	recv [
-DOT		shape=record
-DOT		label="{cnt_recv:|{vcl_recv\{\}|req.*}|{<pipe>pipe?|<pass>pass?|<error>error?|<lookup>lookup?}}"
-DOT	]
-DOT }
-DOT subgraph xcluster_hash {
-DOT	hash [
-DOT		shape=record
-DOT		label="{cnt_recv:|{vcl_hash\{\}|req.*}}"
-DOT	]
-DOT }
-DOT ESI_REQ [ shape=hexagon ]
-DOT ESI_REQ -> recv
-DOT recv:pipe -> pipe [style=bold,color=orange]
-DOT recv:pass -> pass [style=bold,color=red]
-#DOT recv:error -> err_recv
-#DOT err_recv [label="ERROR",shape=plaintext]
-DOT recv:lookup -> hash [style=bold,color=green]
-DOT hash -> lookup [label="hash",style=bold,color=green]
- */
-
-static int
-cnt_recv(const struct worker *wrk, struct req *req)
-{
-	unsigned recv_handling;
-	struct SHA256Context sha256ctx;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
-	AZ(req->obj);
-	AZ(req->busyobj);
-
-	/* By default we use the first backend */
-	AZ(req->director);
-	req->director = req->vcl->director[0];
-	AN(req->director);
-
-	req->disable_esi = 0;
-	req->hash_always_miss = 0;
-	req->hash_ignore_busy = 0;
-	req->client_identity = NULL;
-
-	http_CollectHdr(req->http, H_Cache_Control);
-
-	VCL_recv_method(req);
-	recv_handling = req->handling;
-
-	if (cache_param->http_gzip_support &&
-	     (recv_handling != VCL_RET_PIPE) &&
-	     (recv_handling != VCL_RET_PASS)) {
-		if (RFC2616_Req_Gzip(req->http)) {
-			http_Unset(req->http, H_Accept_Encoding);
-			http_SetHeader(req->http, "Accept-Encoding: gzip");
-		} else {
-			http_Unset(req->http, H_Accept_Encoding);
-		}
-	}
-
-	req->sha256ctx = &sha256ctx;	/* so HSH_AddString() can find it */
-	SHA256_Init(req->sha256ctx);
-	VCL_hash_method(req);
-	assert(req->handling == VCL_RET_HASH);
-	SHA256_Final(req->digest, req->sha256ctx);
-	req->sha256ctx = NULL;
-
-	if (!strcmp(req->http->hd[HTTP_HDR_REQ].b, "HEAD"))
-		req->wantbody = 0;
-	else
-		req->wantbody = 1;
-
-	switch(recv_handling) {
-	case VCL_RET_LOOKUP:
-		req->req_step = R_STP_LOOKUP;
-		return (0);
-	case VCL_RET_PIPE:
-		if (req->esi_level > 0) {
-			/* XXX: VSL something */
-			INCOMPL();
-			return (1);
-		}
-		req->req_step = R_STP_PIPE;
-		return (0);
-	case VCL_RET_PASS:
-		req->req_step = R_STP_PASS;
-		return (0);
-	case VCL_RET_ERROR:
-		req->req_step = R_STP_ERROR;
-		return (0);
-	default:
-		WRONG("Illegal action in vcl_recv{}");
-	}
-}
-
-/*--------------------------------------------------------------------
- * START
- * First time we see a request
- *
-DOT start [
-DOT	shape=box
-DOT	label="cnt_start:\nDissect request\nHandle expect"
-DOT ]
-DOT start -> recv [style=bold,color=green]
-DOT start -> DONE [label=errors]
- */
-
-static int
-cnt_start(struct worker *wrk, struct req *req)
-{
-	char *p;
-	const char *r = "HTTP/1.1 100 Continue\r\n\r\n";
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-	AZ(req->restarts);
-	AZ(req->obj);
-	AZ(req->vcl);
-	AZ(req->esi_level);
-	assert(!isnan(req->t_req));
-
-	/* Update stats of various sorts */
-	wrk->stats.client_req++;
-	req->acct_req.req++;
-
-	/* Assign XID and log */
-	req->xid = ++xids;				/* XXX not locked */
-	VSLb(req->vsl, SLT_ReqStart, "%s %s %u",
-	    req->sp->addr, req->sp->port, req->xid);
-
-	/* Borrow VCL reference from worker thread */
-	VCL_Refresh(&wrk->vcl);
-	req->vcl = wrk->vcl;
-	wrk->vcl = NULL;
-
-	EXP_Clr(&req->exp);
-
-	HTTP_Setup(req->http, req->ws, req->vsl, HTTP_Req);
-	req->err_code = http_DissectRequest(req);
-
-	/* If we could not even parse the request, just close */
-	if (req->err_code == 400) {
-		SES_Close(req->sp, SC_RX_JUNK);
-		return (1);
-	}
-
-	req->ws_req = WS_Snapshot(req->ws);
-
-	req->doclose = http_DoConnection(req->http);
-
-	/*
-	 * We want to deal with Expect: headers the first time we
-	 * attempt the request, and remove them before we move on.
-	 */
-	if (req->err_code == 0 && http_GetHdr(req->http, H_Expect, &p)) {
-		if (strcasecmp(p, "100-continue")) {
-			req->err_code = 417;
-		} else if (strlen(r) != write(req->sp->fd, r, strlen(r))) {
-			SES_Close(req->sp, SC_REM_CLOSE);
-			return (1);
-		}
-	}
-	http_Unset(req->http, H_Expect);
-
-	/* XXX: pull in req-body and make it available instead. */
-	req->reqbodydone = 0;
-
-	HTTP_Copy(req->http0, req->http);	/* Copy for restart/ESI use */
-
-	if (req->err_code)
-		req->req_step = R_STP_ERROR;
-	else
-		req->req_step = R_STP_RECV;
-	return (0);
-}
-
-/*--------------------------------------------------------------------
- * Central state engine dispatcher.
- *
- * Kick the session around until it has had enough.
- *
- */
-
-static void
-cnt_diag(struct req *req, const char *state)
-{
-
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-
-	VSLb(req->vsl,  SLT_Debug, "vsl_id %u STP_%s sp %p obj %p vcl %p",
-	    req->sp->vsl_id, state, req->sp, req->obj, req->vcl);
-	VSL_Flush(req->vsl, 0);
-}
-
-int
-CNT_Request(struct worker *wrk, struct req *req)
-{
-	int done;
-
-	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-
-	/*
-	 * Possible entrance states
-	 */
-	assert(
-	    req->req_step == R_STP_LOOKUP ||
-	    req->req_step == R_STP_START ||
-	    req->req_step == R_STP_RECV);
-
-	req->wrk = wrk;
-
-	for (done = 0; !done; ) {
-		/*
-		 * This is a good place to be paranoid about the various
-		 * pointers still pointing to the things we expect.
-		 */
-		CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
-		CHECK_OBJ_ORNULL(wrk->nobjhead, OBJHEAD_MAGIC);
-		WS_Assert(wrk->aws);
-		CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
-
-		switch (req->req_step) {
-#define REQ_STEP(l,u,arg) \
-		    case R_STP_##u: \
-			if (cache_param->diag_bitmap & 0x01) \
-				cnt_diag(req, #u); \
-			done = cnt_##l arg; \
-		        break;
-#include "tbl/steps.h"
-#undef REQ_STEP
-		default:
-			WRONG("State engine misfire");
-		}
-		WS_Assert(wrk->aws);
-		CHECK_OBJ_ORNULL(wrk->nobjhead, OBJHEAD_MAGIC);
-	}
-	if (done == 1) {
-		/* XXX: Workaround for pipe */
-		if (req->sp->fd >= 0) {
-			VSLb(req->vsl, SLT_Length, "%ju",
-			    (uintmax_t)req->req_bodybytes);
-		}
-		VSLb(req->vsl, SLT_ReqEnd, "%u %.9f %.9f %.9f %.9f %.9f",
-		    req->xid,
-		    req->t_req,
-		    req->sp->t_idle,
-		    req->sp->t_idle - req->t_resp,
-		    req->t_resp - req->t_req,
-		    req->sp->t_idle - req->t_resp);
-
-		/* done == 2 was charged by cache_hash.c */
-		SES_Charge(wrk, req);
-	}
-
-	req->wrk = NULL;
-
-	assert(WRW_IsReleased(wrk));
-	return (done);
-}
-
-/*
-DOT }
-*/
-
-/*--------------------------------------------------------------------
- * Debugging aids
- */
-
-static void
-cli_debug_xid(struct cli *cli, const char * const *av, void *priv)
-{
-	(void)priv;
-	if (av[2] != NULL)
-		xids = strtoul(av[2], NULL, 0);
-	VCLI_Out(cli, "XID is %u", xids);
-}
-
-/*
- * Default to seed=1, this is the only seed value POSIXl guarantees will
- * result in a reproducible random number sequence.
- */
-static void
-cli_debug_srandom(struct cli *cli, const char * const *av, void *priv)
-{
-	(void)priv;
-	unsigned seed = 1;
-
-	if (av[2] != NULL)
-		seed = strtoul(av[2], NULL, 0);
-	srandom(seed);
-	srand48(random());
-	VCLI_Out(cli, "Random(3) seeded with %u", seed);
-}
-
-static struct cli_proto debug_cmds[] = {
-	{ "debug.xid", "debug.xid",
-		"\tExamine or set XID\n", 0, 1, "d", cli_debug_xid },
-	{ "debug.srandom", "debug.srandom",
-		"\tSeed the random(3) function\n", 0, 1, "d",
-		cli_debug_srandom },
-	{ NULL }
-};
-
-/*--------------------------------------------------------------------
- *
- */
-
-void
-CNT_Init(void)
-{
-
-	srandomdev();
-	srand48(random());
-	xids = random();
-	CLI_AddFuncs(debug_cmds);
-}
diff --git a/bin/varnishd/cache/cache_http1_fsm.c b/bin/varnishd/cache/cache_http1_fsm.c
new file mode 100644
index 0000000..586fd50
--- /dev/null
+++ b/bin/varnishd/cache/cache_http1_fsm.c
@@ -0,0 +1,348 @@
+/*-
+ * Copyright (c) 2006 Verdens Gang AS
+ * Copyright (c) 2006-2011 Varnish Software AS
+ * All rights reserved.
+ *
+ * Author: Poul-Henning Kamp <phk at phk.freebsd.dk>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * This file contains the two central state machine for pushing
+ * sessions and requests.
+ *
+ * The first part of the file, entrypoint CNT_Session() and down to
+ * the ==== separator, is concerned with sessions.  When a session has
+ * a request to deal with, it calls into the second half of the file.
+ * This part is for all practical purposes HTTP/1.x specific.
+ *
+ * The second part of the file, entrypoint CNT_Request() and below the
+ * ==== separator, is intended to (over time) be(ome) protocol agnostic.
+ * We already use this now with ESI:includes, which are for all relevant
+ * purposes a different "protocol"
+ *
+ * A special complication is the fact that we can suspend processing of
+ * a request when hash-lookup finds a busy objhdr.
+ *
+ * Since the states are rather nasty in detail, I have decided to embedd
+ * a dot(1) graph in the source code comments.  So to see the big picture,
+ * extract the DOT lines and run though dot(1), for instance with the
+ * command:
+ *	sed -n '/^DOT/s///p' cache/cache_center.c | dot -Tps > /tmp/_.ps
+ */
+
+/*
+DOT digraph vcl_center {
+xDOT	page="8.2,11.5"
+DOT	size="7.2,10.5"
+DOT	margin="0.5"
+DOT	center="1"
+DOT acceptor [
+DOT	shape=hexagon
+DOT	label="Request received"
+DOT ]
+DOT ERROR [shape=plaintext]
+DOT RESTART [shape=plaintext]
+DOT acceptor -> first [style=bold,color=green]
+ */
+
+#include "config.h"
+
+#include <math.h>
+#include <poll.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "cache.h"
+
+#include "hash/hash_slinger.h"
+#include "vcl.h"
+#include "vcli_priv.h"
+#include "vsha256.h"
+#include "vtcp.h"
+#include "vtim.h"
+
+#ifndef HAVE_SRANDOMDEV
+#include "compat/srandomdev.h"
+#endif
+
+
+/*--------------------------------------------------------------------
+ * WAIT
+ * Collect the request from the client.
+ *
+DOT subgraph xcluster_wait {
+DOT	wait [
+DOT		shape=box
+DOT		label="cnt_sess_wait:\nwait for\ncomplete\nrequest"
+DOT	]
+DOT	herding [shape=hexagon]
+DOT	wait -> start [label="got req",style=bold,color=green]
+DOT	wait -> "SES_Delete()" [label="errors"]
+DOT	wait -> herding [label="timeout_linger"]
+DOT	herding -> wait [label="fd read_ready"]
+DOT }
+ */
+
+static int
+cnt_sess_wait(struct sess *sp, struct worker *wrk, struct req *req)
+{
+	int j, tmo;
+	struct pollfd pfd[1];
+	double now, when;
+	enum sess_close why = SC_NULL;
+	enum htc_status_e hs;
+
+	CHECK_OBJ_NOTNULL(sp, SESS_MAGIC);
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+
+	assert(req->sp == sp);
+
+
+	AZ(req->vcl);
+	AZ(req->obj);
+	AZ(req->esi_level);
+	assert(req->xid == 0);
+	assert(isnan(req->t_req));
+	assert(isnan(req->t_resp));
+
+	tmo = (int)(1e3 * cache_param->timeout_linger);
+	while (1) {
+		pfd[0].fd = sp->fd;
+		pfd[0].events = POLLIN;
+		pfd[0].revents = 0;
+		j = poll(pfd, 1, tmo);
+		assert(j >= 0);
+		now = VTIM_real();
+		if (j != 0)
+			hs = HTC_Rx(req->htc);
+		else
+			hs = HTC_Complete(req->htc);
+		if (hs == HTC_COMPLETE) {
+			/* Got it, run with it */
+			req->t_req = now;
+			return (0);
+		} else if (hs == HTC_ERROR_EOF) {
+			why = SC_REM_CLOSE;
+			break;
+		} else if (hs == HTC_OVERFLOW) {
+			why = SC_RX_OVERFLOW;
+			break;
+		} else if (hs == HTC_ALL_WHITESPACE) {
+			/* Nothing but whitespace */
+			when = sp->t_idle + cache_param->timeout_idle;
+			if (when < now) {
+				why = SC_RX_TIMEOUT;
+				break;
+			}
+			when = sp->t_idle + cache_param->timeout_linger;
+			tmo = (int)(1e3 * (when - now));
+			if (when < now || tmo == 0) {
+				req->t_req = NAN;
+				wrk->stats.sess_herd++;
+				SES_ReleaseReq(req);
+				WAIT_Enter(sp);
+				return (1);
+			}
+		} else {
+			/* Working on it */
+			if (isnan(req->t_req))
+				req->t_req = now;
+			when = req->t_req + cache_param->timeout_req;
+			tmo = (int)(1e3 * (when - now));
+			if (when < now || tmo == 0) {
+				why = SC_RX_TIMEOUT;
+				break;
+			}
+		}
+	}
+	SES_ReleaseReq(req);
+	assert(why != SC_NULL);
+	SES_Delete(sp, why, now);
+	return (1);
+}
+
+/*--------------------------------------------------------------------
+ * This is the final state, figure out if we should close or recycle
+ * the client connection
+ *
+DOT	DONE [
+DOT		shape=record
+DOT		label="{cnt_done:|Request completed}"
+DOT	]
+DOT	ESI_RESP [ shape=hexagon ]
+DOT	DONE -> start [label="full pipeline"]
+DOT	DONE -> wait
+DOT	DONE -> ESI_RESP
+ */
+
+enum cnt_sess_done_ret {
+	SESS_DONE_RET_GONE,
+	SESS_DONE_RET_WAIT,
+	SESS_DONE_RET_START,
+};
+
+static enum cnt_sess_done_ret
+cnt_sess_done(struct sess *sp, struct worker *wrk, struct req *req)
+{
+
+	CHECK_OBJ_NOTNULL(sp, SESS_MAGIC);
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	CHECK_OBJ_ORNULL(req->vcl, VCL_CONF_MAGIC);
+
+	AZ(req->obj);
+	AZ(req->busyobj);
+	req->director = NULL;
+	req->restarts = 0;
+
+	AZ(req->esi_level);
+
+	if (req->vcl != NULL) {
+		if (wrk->vcl != NULL)
+			VCL_Rel(&wrk->vcl);
+		wrk->vcl = req->vcl;
+		req->vcl = NULL;
+	}
+
+	sp->t_idle = W_TIM_real(wrk);
+	if (req->xid == 0) 
+		req->t_resp = sp->t_idle;
+	req->xid = 0;
+	VSL_Flush(req->vsl, 0);
+
+	req->t_req = NAN;
+	req->t_resp = NAN;
+
+	req->req_bodybytes = 0;
+
+	req->hash_always_miss = 0;
+	req->hash_ignore_busy = 0;
+
+	if (sp->fd >= 0 && req->doclose != SC_NULL)
+		SES_Close(sp, req->doclose);
+
+	if (sp->fd < 0) {
+		wrk->stats.sess_closed++;
+		AZ(req->vcl);
+		SES_ReleaseReq(req);
+		SES_Delete(sp, SC_NULL, NAN);
+		return (SESS_DONE_RET_GONE);
+	}
+
+	if (wrk->stats.client_req >= cache_param->wthread_stats_rate)
+		WRK_SumStat(wrk);
+
+	WS_Reset(req->ws, NULL);
+	WS_Reset(wrk->aws, NULL);
+	req->vxid = VXID_Get(&wrk->vxid_pool);
+
+	if (HTC_Reinit(req->htc) == HTC_COMPLETE) {
+		req->t_req = sp->t_idle;
+		wrk->stats.sess_pipeline++;
+		return (SESS_DONE_RET_START);
+	} else {
+		if (Tlen(req->htc->rxbuf))
+			wrk->stats.sess_readahead++;
+		return (SESS_DONE_RET_WAIT);
+	}
+}
+
+/*--------------------------------------------------------------------
+ */
+
+void
+CNT_Session(struct worker *wrk, struct req *req)
+{
+	int done;
+	struct sess *sp;
+	enum cnt_sess_done_ret sdr;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	sp = req->sp;
+	CHECK_OBJ_NOTNULL(sp, SESS_MAGIC);
+
+	/*
+	 * Whenever we come in from the acceptor or waiter, we need to set
+	 * blocking mode, but there is no point in setting it when we come from
+	 * ESI or when a parked sessions returns.
+	 * It would be simpler to do this in the acceptor or waiter, but we'd
+	 * rather do the syscall in the worker thread.
+	 * On systems which return errors for ioctl, we close early
+	 */
+	if (sp->sess_step == S_STP_NEWREQ && VTCP_blocking(sp->fd)) {
+		if (errno == ECONNRESET)
+			SES_Close(sp, SC_REM_CLOSE);
+		else
+			SES_Close(sp, SC_TX_ERROR);
+		sdr = cnt_sess_done(sp, wrk, req);
+		assert(sdr == SESS_DONE_RET_GONE);
+		return;
+	}
+
+	if (sp->sess_step == S_STP_NEWREQ) {
+		HTC_Init(req->htc, req->ws, sp->fd, req->vsl,
+		    cache_param->http_req_size,
+		    cache_param->http_req_hdr_len);
+	}
+
+	while (1) {
+		/*
+		 * Possible entrance states
+		 */
+
+		assert(
+		    sp->sess_step == S_STP_NEWREQ ||
+		    req->req_step == R_STP_LOOKUP ||
+		    req->req_step == R_STP_START);
+
+		if (sp->sess_step == S_STP_WORKING) {
+			done = CNT_Request(wrk, req);
+			if (done == 2)
+				return;
+			assert(done == 1);
+			sdr = cnt_sess_done(sp, wrk, req);
+			switch (sdr) {
+			case SESS_DONE_RET_GONE:
+				return;
+			case SESS_DONE_RET_WAIT:
+				sp->sess_step = S_STP_NEWREQ;
+				break;
+			case SESS_DONE_RET_START:
+				sp->sess_step = S_STP_WORKING;
+				req->req_step = R_STP_START;
+				break;
+			default:
+				WRONG("Illegal enum cnt_sess_done_ret");
+			}
+		}
+
+		if (sp->sess_step == S_STP_NEWREQ) {
+			done = cnt_sess_wait(sp, wrk, req);
+			if (done)
+				return;
+			sp->sess_step = S_STP_WORKING;
+			req->req_step = R_STP_START;
+		}
+	}
+}
diff --git a/bin/varnishd/cache/cache_req_fsm.c b/bin/varnishd/cache/cache_req_fsm.c
new file mode 100644
index 0000000..34cfbfa
--- /dev/null
+++ b/bin/varnishd/cache/cache_req_fsm.c
@@ -0,0 +1,1395 @@
+/*-
+ * Copyright (c) 2006 Verdens Gang AS
+ * Copyright (c) 2006-2011 Varnish Software AS
+ * All rights reserved.
+ *
+ * Author: Poul-Henning Kamp <phk at phk.freebsd.dk>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * This file contains the two central state machine for pushing
+ * sessions and requests.
+ *
+ * The first part of the file, entrypoint CNT_Session() and down to
+ * the ==== separator, is concerned with sessions.  When a session has
+ * a request to deal with, it calls into the second half of the file.
+ * This part is for all practical purposes HTTP/1.x specific.
+ *
+ * The second part of the file, entrypoint CNT_Request() and below the
+ * ==== separator, is intended to (over time) be(ome) protocol agnostic.
+ * We already use this now with ESI:includes, which are for all relevant
+ * purposes a different "protocol"
+ *
+ * A special complication is the fact that we can suspend processing of
+ * a request when hash-lookup finds a busy objhdr.
+ *
+ * Since the states are rather nasty in detail, I have decided to embedd
+ * a dot(1) graph in the source code comments.  So to see the big picture,
+ * extract the DOT lines and run though dot(1), for instance with the
+ * command:
+ *	sed -n '/^DOT/s///p' cache/cache_center.c | dot -Tps > /tmp/_.ps
+ */
+
+/*
+DOT digraph vcl_center {
+xDOT	page="8.2,11.5"
+DOT	size="7.2,10.5"
+DOT	margin="0.5"
+DOT	center="1"
+DOT acceptor [
+DOT	shape=hexagon
+DOT	label="Request received"
+DOT ]
+DOT ERROR [shape=plaintext]
+DOT RESTART [shape=plaintext]
+DOT acceptor -> first [style=bold,color=green]
+ */
+
+#include "config.h"
+
+#include <math.h>
+#include <poll.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "cache.h"
+
+#include "hash/hash_slinger.h"
+#include "vcl.h"
+#include "vcli_priv.h"
+#include "vsha256.h"
+#include "vtim.h"
+
+#ifndef HAVE_SRANDOMDEV
+#include "compat/srandomdev.h"
+#endif
+
+static unsigned xids;
+/*--------------------------------------------------------------------
+ * We have a refcounted object on the session, and possibly the busyobj
+ * which is fetching it, prepare a response.
+ *
+DOT subgraph xcluster_prepresp {
+DOT	prepresp [
+DOT		shape=record
+DOT		label="{cnt_prepresp:|Filter obj.-\>resp.|{vcl_deliver\{\}|{req.|resp.}}|{error?|restart?}|stream ?}"
+DOT	]
+DOT	prepresp -> deliver [style=bold,color=green,label=deliver]
+DOT	prepresp -> deliver [style=bold,color=red]
+DOT	prepresp -> deliver [style=bold,color=blue]
+DOT }
+ *
+ */
+
+static int
+cnt_prepresp(struct worker *wrk, struct req *req)
+{
+	struct busyobj *bo;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	bo = req->busyobj;
+	CHECK_OBJ_ORNULL(bo, BUSYOBJ_MAGIC);
+
+	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+
+	req->res_mode = 0;
+
+	if (bo == NULL) {
+		if (!req->disable_esi && req->obj->esidata != NULL) {
+			/* In ESI mode, we can't know the aggregate length */
+			req->res_mode &= ~RES_LEN;
+			req->res_mode |= RES_ESI;
+		} else {
+			req->res_mode |= RES_LEN;
+		}
+	} else {
+		AZ(bo->do_esi);
+	}
+
+	if (req->esi_level > 0) {
+		/* Included ESI object, always CHUNKED or EOF */
+		req->res_mode &= ~RES_LEN;
+		req->res_mode |= RES_ESI_CHILD;
+	}
+
+	if (cache_param->http_gzip_support && req->obj->gziped &&
+	    !RFC2616_Req_Gzip(req->http)) {
+		/*
+		 * We don't know what it uncompresses to
+		 * XXX: we could cache that
+		 */
+		req->res_mode &= ~RES_LEN;
+		req->res_mode |= RES_GUNZIP;
+	}
+
+	if (!(req->res_mode & (RES_LEN|RES_CHUNKED|RES_EOF))) {
+		/* We havn't chosen yet, do so */
+		if (!req->wantbody) {
+			/* Nothing */
+		} else if (req->http->protover >= 11) {
+			req->res_mode |= RES_CHUNKED;
+		} else {
+			req->res_mode |= RES_EOF;
+			req->doclose = SC_TX_EOF;
+		}
+	}
+
+	req->t_resp = W_TIM_real(wrk);
+	if (req->obj->objcore->objhead != NULL) {
+		if ((req->t_resp - req->obj->last_lru) >
+		    cache_param->lru_timeout &&
+		    EXP_Touch(req->obj->objcore))
+			req->obj->last_lru = req->t_resp;
+		if (!cache_param->obj_readonly)
+			req->obj->last_use = req->t_resp; /* XXX: locking ? */
+	}
+	HTTP_Setup(req->resp, req->ws, req->vsl, HTTP_Resp);
+	RES_BuildHttp(req);
+
+	VCL_deliver_method(req);
+	switch (req->handling) {
+	case VCL_RET_DELIVER:
+		break;
+	case VCL_RET_RESTART:
+		if (req->restarts >= cache_param->max_restarts)
+			break;
+		if (bo != NULL) {
+			AN(bo->do_stream);
+			(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
+			VBO_DerefBusyObj(wrk, &req->busyobj);
+		} else {
+			(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
+		}
+		AZ(req->obj);
+		http_Teardown(req->resp);
+		req->req_step = R_STP_RESTART;
+		return (0);
+	default:
+		WRONG("Illegal action in vcl_deliver{}");
+	}
+	req->req_step = R_STP_DELIVER;
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * Deliver an already stored object
+ *
+DOT subgraph xcluster_deliver {
+DOT	deliver [
+DOT		shape=record
+DOT		label="{cnt_deliver:|Send body}"
+DOT	]
+DOT }
+DOT deliver -> DONE [style=bold,color=green]
+DOT deliver -> DONE [style=bold,color=red]
+DOT deliver -> DONE [style=bold,color=blue]
+ *
+ */
+
+static int
+cnt_deliver(struct worker *wrk, struct req *req)
+{
+	struct busyobj *bo;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
+	bo = req->busyobj;
+	CHECK_OBJ_ORNULL(bo, BUSYOBJ_MAGIC);
+
+	if (bo != NULL) {
+		while (bo->state < BOS_FAILED)
+			(void)usleep(10000);
+		assert(bo->state >= BOS_FAILED);
+
+		if (bo->state == BOS_FAILED) {
+			HSH_Deref(&wrk->stats, NULL, &req->obj);
+			VBO_DerefBusyObj(wrk, &req->busyobj);
+			req->err_code = 503;
+			req->req_step = R_STP_ERROR;
+			return (0);
+		}
+		VBO_DerefBusyObj(wrk, &req->busyobj);
+	}
+
+	AZ(req->busyobj);
+	req->director = NULL;
+	req->restarts = 0;
+
+	RES_WriteObj(req);
+
+	/* No point in saving the body if it is hit-for-pass */
+	if (req->obj->objcore->flags & OC_F_PASS)
+		STV_Freestore(req->obj);
+
+	assert(WRW_IsReleased(wrk));
+	(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
+	http_Teardown(req->resp);
+	return (1);
+}
+/*--------------------------------------------------------------------
+ * Emit an error
+ *
+DOT subgraph xcluster_error {
+DOT	vcl_error [
+DOT		shape=record
+DOT		label="vcl_error()|resp."
+DOT	]
+DOT	ERROR -> vcl_error
+DOT	vcl_error-> prepresp [label=deliver]
+DOT }
+DOT vcl_error-> rsterr [label="restart",color=purple]
+DOT rsterr [label="RESTART",shape=plaintext]
+ */
+
+static int
+cnt_error(struct worker *wrk, struct req *req)
+{
+	struct http *h;
+	struct busyobj *bo;
+	char date[40];
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	AZ(req->objcore);
+	AZ(req->obj);
+	AZ(req->busyobj);
+
+	bo = VBO_GetBusyObj(wrk);
+	req->busyobj = bo;
+	bo->vsl->wid = req->sp->vsl_id;
+	AZ(bo->stats);
+	bo->stats = &wrk->stats;
+	req->objcore = HSH_NewObjCore(wrk);
+	req->obj = STV_NewObject(bo, &req->objcore,
+	    TRANSIENT_STORAGE, cache_param->http_resp_size,
+	    (uint16_t)cache_param->http_max_hdr);
+	bo->stats = NULL;
+	if (req->obj == NULL) {
+		req->doclose = SC_OVERLOAD;
+		req->director = NULL;
+		http_Teardown(bo->beresp);
+		http_Teardown(bo->bereq);
+		return(1);
+	}
+	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
+	req->obj->xid = req->xid;
+	req->obj->exp.entered = req->t_req;
+
+	h = req->obj->http;
+
+	if (req->err_code < 100 || req->err_code > 999)
+		req->err_code = 501;
+
+	http_PutProtocol(h, "HTTP/1.1");
+	http_PutStatus(h, req->err_code);
+	VTIM_format(W_TIM_real(wrk), date);
+	http_PrintfHeader(h, "Date: %s", date);
+	http_SetHeader(h, "Server: Varnish");
+
+	if (req->err_reason != NULL)
+		http_PutResponse(h, req->err_reason);
+	else
+		http_PutResponse(h, http_StatusMessage(req->err_code));
+	VCL_error_method(req);
+
+	if (req->handling == VCL_RET_RESTART &&
+	    req->restarts <  cache_param->max_restarts) {
+		HSH_Drop(wrk, &req->obj);
+		VBO_DerefBusyObj(wrk, &req->busyobj);
+		req->req_step = R_STP_RESTART;
+		return (0);
+	} else if (req->handling == VCL_RET_RESTART)
+		req->handling = VCL_RET_DELIVER;
+
+
+	/* We always close when we take this path */
+	req->doclose = SC_TX_ERROR;
+	req->wantbody = 1;
+
+	assert(req->handling == VCL_RET_DELIVER);
+	req->err_code = 0;
+	req->err_reason = NULL;
+	http_Teardown(bo->bereq);
+	VBO_DerefBusyObj(wrk, &req->busyobj);
+	req->req_step = R_STP_PREPRESP;
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * Fetch response headers from the backend
+ *
+DOT subgraph xcluster_fetch {
+DOT	fetch [
+DOT		shape=record
+DOT		label="{cnt_fetch:|fetch hdr\nfrom backend|(find obj.ttl)|{vcl_fetch\{\}|{req.|bereq.|beresp.}}|{<err>error?|<rst>restart?}}"
+DOT	]
+DOT }
+DOT fetch -> fetchbody [style=bold,color=red]
+DOT fetch -> fetchbody [style=bold,color=blue]
+ */
+
+static int
+cnt_fetch(struct worker *wrk, struct req *req)
+{
+	int i, need_host_hdr;
+	struct busyobj *bo;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+	bo = req->busyobj;
+	CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
+
+	AN(req->director);
+	AZ(bo->vbc);
+	AZ(bo->should_close);
+	AZ(req->storage_hint);
+
+	HTTP_Setup(bo->beresp, bo->ws, bo->vsl, HTTP_Beresp);
+
+	need_host_hdr = !http_GetHdr(bo->bereq, H_Host, NULL);
+
+	req->acct_req.fetch++;
+
+	i = FetchHdr(req, need_host_hdr, req->objcore->objhead == NULL);
+	/*
+	 * If we recycle a backend connection, there is a finite chance
+	 * that the backend closed it before we get a request to it.
+	 * Do a single retry in that case.
+	 */
+	if (i == 1) {
+		VSC_C_main->backend_retry++;
+		i = FetchHdr(req, need_host_hdr, req->objcore->objhead == NULL);
+	}
+
+	if (i) {
+		req->handling = VCL_RET_ERROR;
+		req->err_code = 503;
+	} else {
+		/*
+		 * These two headers can be spread over multiple actual headers
+		 * and we rely on their content outside of VCL, so collect them
+		 * into one line here.
+		 */
+		http_CollectHdr(bo->beresp, H_Cache_Control);
+		http_CollectHdr(bo->beresp, H_Vary);
+
+		/*
+		 * Figure out how the fetch is supposed to happen, before the
+		 * headers are adultered by VCL
+		 * NB: Also sets other wrk variables
+		 */
+		bo->body_status = RFC2616_Body(bo, &wrk->stats);
+
+		req->err_code = http_GetStatus(bo->beresp);
+
+		/*
+		 * What does RFC2616 think about TTL ?
+		 */
+		EXP_Clr(&bo->exp);
+		bo->exp.entered = W_TIM_real(wrk);
+		RFC2616_Ttl(bo, req->xid);
+
+		/* pass from vclrecv{} has negative TTL */
+		if (req->objcore->objhead == NULL)
+			bo->exp.ttl = -1.;
+
+		AZ(bo->do_esi);
+		AZ(bo->do_pass);
+
+		VCL_fetch_method(req);
+
+		if (bo->do_pass)
+			req->objcore->flags |= OC_F_PASS;
+
+		switch (req->handling) {
+		case VCL_RET_DELIVER:
+			req->req_step = R_STP_FETCHBODY;
+			return (0);
+		default:
+			break;
+		}
+
+		/* We are not going to fetch the body, Close the connection */
+		VDI_CloseFd(&bo->vbc);
+	}
+
+	/* Clean up partial fetch */
+	AZ(bo->vbc);
+
+	if (req->objcore->objhead != NULL || req->handling == VCL_RET_ERROR) {
+		CHECK_OBJ_NOTNULL(req->objcore, OBJCORE_MAGIC);
+		AZ(HSH_Deref(&wrk->stats, req->objcore, NULL));
+		req->objcore = NULL;
+	}
+	assert(bo->refcount == 2);
+	VBO_DerefBusyObj(wrk, &bo);
+	VBO_DerefBusyObj(wrk, &req->busyobj);
+	req->director = NULL;
+	req->storage_hint = NULL;
+
+	switch (req->handling) {
+	case VCL_RET_RESTART:
+		req->req_step = R_STP_RESTART;
+		return (0);
+	case VCL_RET_ERROR:
+		req->req_step = R_STP_ERROR;
+		return (0);
+	default:
+		WRONG("Illegal action in vcl_fetch{}");
+	}
+}
+
+/*--------------------------------------------------------------------
+ * Prepare to fetch body from backend
+ *
+DOT subgraph xcluster_body {
+DOT	fetchbody [
+DOT		shape=record
+DOT		label="{cnt_fetchbody:|start fetch_thread}"
+DOT	]
+DOT }
+DOT fetchbody:out -> prepresp [style=bold,color=red]
+DOT fetchbody:out -> prepresp [style=bold,color=blue]
+ */
+
+static int
+cnt_fetchbody(struct worker *wrk, struct req *req)
+{
+	struct http *hp, *hp2;
+	char *b;
+	uint16_t nhttp;
+	unsigned l;
+	struct vsb *vary = NULL;
+	int varyl = 0, pass;
+	struct busyobj *bo;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	bo = req->busyobj;
+	CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
+
+	assert(req->handling == VCL_RET_DELIVER);
+
+	if (req->objcore->objhead == NULL) {
+		/* This is a pass from vcl_recv */
+		pass = 1;
+		/* VCL may have fiddled this, but that doesn't help */
+		bo->exp.ttl = -1.;
+	} else if (bo->do_pass) {
+		pass = 1;
+	} else {
+		/* regular object */
+		pass = 0;
+	}
+
+	/*
+	 * The VCL variables beresp.do_g[un]zip tells us how we want the
+	 * object processed before it is stored.
+	 *
+	 * The backend Content-Encoding header tells us what we are going
+	 * to receive, which we classify in the following three classes:
+	 *
+	 *	"Content-Encoding: gzip"	--> object is gzip'ed.
+	 *	no Content-Encoding		--> object is not gzip'ed.
+	 *	anything else			--> do nothing wrt gzip
+	 *
+	 */
+
+	/* We do nothing unless the param is set */
+	if (!cache_param->http_gzip_support)
+		bo->do_gzip = bo->do_gunzip = 0;
+
+	bo->is_gzip = http_HdrIs(bo->beresp, H_Content_Encoding, "gzip");
+
+	bo->is_gunzip = !http_GetHdr(bo->beresp, H_Content_Encoding, NULL);
+
+	/* It can't be both */
+	assert(bo->is_gzip == 0 || bo->is_gunzip == 0);
+
+	/* We won't gunzip unless it is gzip'ed */
+	if (bo->do_gunzip && !bo->is_gzip)
+		bo->do_gunzip = 0;
+
+	/* If we do gunzip, remove the C-E header */
+	if (bo->do_gunzip)
+		http_Unset(bo->beresp, H_Content_Encoding);
+
+	/* We wont gzip unless it is ungziped */
+	if (bo->do_gzip && !bo->is_gunzip)
+		bo->do_gzip = 0;
+
+	/* If we do gzip, add the C-E header */
+	if (bo->do_gzip)
+		http_SetHeader(bo->beresp, "Content-Encoding: gzip");
+
+	/* But we can't do both at the same time */
+	assert(bo->do_gzip == 0 || bo->do_gunzip == 0);
+
+	/* ESI takes precedence and handles gzip/gunzip itself */
+	if (bo->do_esi)
+		bo->vfp = &vfp_esi;
+	else if (bo->do_gunzip)
+		bo->vfp = &vfp_gunzip;
+	else if (bo->do_gzip)
+		bo->vfp = &vfp_gzip;
+	else if (bo->is_gzip)
+		bo->vfp = &vfp_testgzip;
+
+	if (bo->do_esi || req->esi_level > 0)
+		bo->do_stream = 0;
+	if (!req->wantbody)
+		bo->do_stream = 0;
+
+	/* No reason to try streaming a non-existing body */
+	if (bo->body_status == BS_NONE)
+		bo->do_stream = 0;
+
+	l = http_EstimateWS(bo->beresp,
+	    pass ? HTTPH_R_PASS : HTTPH_A_INS, &nhttp);
+
+	/* Create Vary instructions */
+	if (req->objcore->objhead != NULL) {
+		CHECK_OBJ_NOTNULL(req->objcore, OBJCORE_MAGIC);
+		vary = VRY_Create(req, bo->beresp);
+		if (vary != NULL) {
+			varyl = VSB_len(vary);
+			assert(varyl > 0);
+			l += varyl;
+		}
+	}
+
+	/*
+	 * Space for producing a Content-Length: header including padding
+	 * A billion gigabytes is enough for anybody.
+	 */
+	l += strlen("Content-Length: XxxXxxXxxXxxXxxXxx") + sizeof(void *);
+
+	if (bo->exp.ttl < cache_param->shortlived ||
+	    req->objcore == NULL)
+		req->storage_hint = TRANSIENT_STORAGE;
+
+	AZ(bo->stats);
+	bo->stats = &wrk->stats;
+	req->obj = STV_NewObject(bo, &req->objcore, req->storage_hint, l,
+	    nhttp);
+	if (req->obj == NULL) {
+		/*
+		 * Try to salvage the transaction by allocating a
+		 * shortlived object on Transient storage.
+		 */
+		if (bo->exp.ttl > cache_param->shortlived)
+			bo->exp.ttl = cache_param->shortlived;
+		bo->exp.grace = 0.0;
+		bo->exp.keep = 0.0;
+		req->obj = STV_NewObject(bo, &req->objcore, TRANSIENT_STORAGE,
+		    l, nhttp);
+	}
+	bo->stats = NULL;
+	if (req->obj == NULL) {
+		req->err_code = 503;
+		req->req_step = R_STP_ERROR;
+		VDI_CloseFd(&bo->vbc);
+		VBO_DerefBusyObj(wrk, &req->busyobj);
+		return (0);
+	}
+	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
+
+	req->storage_hint = NULL;
+
+	AZ(bo->fetch_obj);
+	bo->fetch_obj = req->obj;
+
+	if (bo->do_gzip || (bo->is_gzip && !bo->do_gunzip))
+		req->obj->gziped = 1;
+
+	if (vary != NULL) {
+		req->obj->vary = (void *)WS_Alloc(req->obj->http->ws, varyl);
+		AN(req->obj->vary);
+		memcpy(req->obj->vary, VSB_data(vary), varyl);
+		VRY_Validate(req->obj->vary);
+		VSB_delete(vary);
+	}
+
+	req->obj->xid = req->xid;
+	req->obj->response = req->err_code;
+	WS_Assert(req->obj->ws_o);
+
+	/* Filter into object */
+	hp = bo->beresp;
+	hp2 = req->obj->http;
+
+	hp2->logtag = HTTP_Obj;
+	http_FilterResp(hp, hp2, pass ? HTTPH_R_PASS : HTTPH_A_INS);
+	http_CopyHome(hp2);
+
+	if (http_GetHdr(hp, H_Last_Modified, &b))
+		req->obj->last_modified = VTIM_parse(b);
+	else
+		req->obj->last_modified = floor(bo->exp.entered);
+
+	assert(WRW_IsReleased(wrk));
+
+	/*
+	 * If we can deliver a 304 reply, we don't bother streaming.
+	 * Notice that vcl_deliver{} could still nuke the headers
+	 * that allow the 304, in which case we return 200 non-stream.
+	 */
+	if (req->obj->response == 200 &&
+	    req->http->conds &&
+	    RFC2616_Do_Cond(req))
+		bo->do_stream = 0;
+
+	/*
+	 * Ready to fetch the body
+	 */
+	bo->fetch_task.func = FetchBody;
+	bo->fetch_task.priv = bo;
+
+	assert(bo->refcount == 2);	/* one for each thread */
+
+	if (req->obj->objcore->objhead != NULL) {
+		EXP_Insert(req->obj);
+		AN(req->obj->objcore->ban);
+		AZ(req->obj->ws_o->overflow);
+		HSH_Unbusy(&wrk->stats, req->obj->objcore);
+	}
+
+	if (!bo->do_stream ||
+	    Pool_Task(wrk->pool, &bo->fetch_task, POOL_NO_QUEUE))
+		FetchBody(wrk, bo);
+
+	if (req->obj->objcore->objhead != NULL)
+		HSH_Ref(req->obj->objcore);
+
+	if (bo->state == BOS_FINISHED) {
+		VBO_DerefBusyObj(wrk, &req->busyobj);
+	} else if (bo->state == BOS_FAILED) {
+		/* handle early failures */
+		HSH_Deref(&wrk->stats, NULL, &req->obj);
+		VBO_DerefBusyObj(wrk, &req->busyobj);
+		req->err_code = 503;
+		req->req_step = R_STP_ERROR;
+		return (0);
+	}
+
+	assert(WRW_IsReleased(wrk));
+	req->req_step = R_STP_PREPRESP;
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * HIT
+ * We had a cache hit.  Ask VCL, then march off as instructed.
+ *
+DOT subgraph xcluster_hit {
+DOT	hit [
+DOT		shape=record
+DOT		label="{cnt_hit:|{vcl_hit()|{req.|obj.}}|{<err>error?|<rst>restart?}|{<del>deliver?|<pass>pass?}}"
+DOT	]
+DOT }
+XDOT hit:err -> err_hit [label="error"]
+XDOT err_hit [label="ERROR",shape=plaintext]
+XDOT hit:rst -> rst_hit [label="restart",color=purple]
+XDOT rst_hit [label="RESTART",shape=plaintext]
+DOT hit:pass -> pass [label=pass,style=bold,color=red]
+DOT hit:del -> prepresp [label="deliver",style=bold,color=green]
+ */
+
+static int
+cnt_hit(struct worker *wrk, struct req *req)
+{
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+
+	CHECK_OBJ_NOTNULL(req->obj, OBJECT_MAGIC);
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+	AZ(req->objcore);
+	AZ(req->busyobj);
+
+	assert(!(req->obj->objcore->flags & OC_F_PASS));
+
+	VCL_hit_method(req);
+
+	if (req->handling == VCL_RET_DELIVER) {
+		//AZ(req->busyobj->bereq->ws);
+		//AZ(req->busyobj->beresp->ws);
+		(void)FetchReqBody(req, 0);
+		req->req_step = R_STP_PREPRESP;
+		return (0);
+	}
+
+	/* Drop our object, we won't need it */
+	(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
+	req->objcore = NULL;
+
+	switch(req->handling) {
+	case VCL_RET_PASS:
+		req->req_step = R_STP_PASS;
+		return (0);
+	case VCL_RET_ERROR:
+		req->req_step = R_STP_ERROR;
+		return (0);
+	case VCL_RET_RESTART:
+		req->req_step = R_STP_RESTART;
+		return (0);
+	default:
+		WRONG("Illegal action in vcl_hit{}");
+	}
+}
+
+/*--------------------------------------------------------------------
+ * LOOKUP
+ * Hash things together and look object up in hash-table.
+ *
+ * LOOKUP consists of two substates so that we can reenter if we
+ * encounter a busy object.
+ *
+DOT subgraph xcluster_lookup {
+DOT	lookup [
+DOT		shape=record
+DOT		label="{<top>cnt_lookup:|hash lookup|{<busy>busy ?|<miss>miss ?}|{<no>no|obj.f.pass?|<yes>yes}}"
+DOT	]
+DOT }
+DOT lookup:busy -> lookup:top [label="(waitinglist)"]
+DOT lookup:miss -> miss [style=bold,color=blue]
+DOT lookup:no -> hit [style=bold,color=green]
+DOT lookup:yes -> pass [style=bold,color=red]
+ */
+
+static int
+cnt_lookup(struct worker *wrk, struct req *req)
+{
+	struct objcore *oc;
+	struct object *o;
+	struct objhead *oh;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	AZ(req->objcore);
+
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+	AZ(req->busyobj);
+
+	VRY_Prep(req);
+
+	AZ(req->objcore);
+	oc = HSH_Lookup(req);
+	if (oc == NULL) {
+		/*
+		 * We lost the session to a busy object, disembark the
+		 * worker thread.   We return to STP_LOOKUP when the busy
+		 * object has been unbusied, and still have the hash digest
+		 * around to do the lookup with.
+		 */
+		return (2);
+	}
+	AZ(req->objcore);
+
+	CHECK_OBJ_NOTNULL(oc, OBJCORE_MAGIC);
+	oh = oc->objhead;
+	CHECK_OBJ_NOTNULL(oh, OBJHEAD_MAGIC);
+
+	/* If we inserted a new object it's a miss */
+	if (oc->flags & OC_F_BUSY) {
+		CHECK_OBJ_NOTNULL(oc->busyobj, BUSYOBJ_MAGIC);
+		assert(oc->busyobj == req->busyobj);
+		wrk->stats.cache_miss++;
+
+		if (req->vary_l != NULL) {
+			assert(oc->busyobj->vary == req->vary_b);
+			VRY_Validate(oc->busyobj->vary);
+			WS_ReleaseP(req->ws, (void*)req->vary_l);
+		} else {
+			AZ(oc->busyobj->vary);
+			WS_Release(req->ws, 0);
+		}
+		req->vary_b = NULL;
+		req->vary_l = NULL;
+		req->vary_e = NULL;
+
+		req->objcore = oc;
+		req->req_step = R_STP_MISS;
+		return (0);
+	}
+
+	/* We are not prepared to do streaming yet */
+	XXXAZ(req->busyobj);
+
+	o = oc_getobj(&wrk->stats, oc);
+	CHECK_OBJ_NOTNULL(o, OBJECT_MAGIC);
+	req->obj = o;
+
+	WS_Release(req->ws, 0);
+	req->vary_b = NULL;
+	req->vary_l = NULL;
+	req->vary_e = NULL;
+
+	if (oc->flags & OC_F_PASS) {
+		wrk->stats.cache_hitpass++;
+		VSLb(req->vsl, SLT_HitPass, "%u", req->obj->xid);
+		(void)HSH_Deref(&wrk->stats, NULL, &req->obj);
+		AZ(req->objcore);
+		req->req_step = R_STP_PASS;
+		return (0);
+	}
+
+	wrk->stats.cache_hit++;
+	VSLb(req->vsl, SLT_Hit, "%u", req->obj->xid);
+	req->req_step = R_STP_HIT;
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * We had a miss, ask VCL, proceed as instructed
+ *
+DOT subgraph xcluster_miss {
+DOT	miss [
+DOT		shape=record
+DOT		label="{cnt_miss:|filter req.-\>bereq.|{vcl_miss\{\}|{req.*|bereq.*}}|{<err>error?|<rst>restart?}|{<pass>pass?|<fetch>fetch?}}"
+DOT	]
+DOT }
+DOT miss:fetch -> fetch [label="fetch",style=bold,color=blue]
+DOT miss:pass -> pass [label="pass",style=bold,color=red]
+DOT
+ */
+
+static int
+cnt_miss(struct worker *wrk, struct req *req)
+{
+	struct busyobj *bo;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+	CHECK_OBJ_NOTNULL(req->objcore, OBJCORE_MAGIC);
+	bo = req->busyobj;
+	CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
+	AZ(req->obj);
+
+	HTTP_Setup(bo->bereq, bo->ws, bo->vsl, HTTP_Bereq);
+	http_FilterReq(req, HTTPH_R_FETCH);
+	http_ForceGet(bo->bereq);
+	if (cache_param->http_gzip_support) {
+		/*
+		 * We always ask the backend for gzip, even if the
+		 * client doesn't grok it.  We will uncompress for
+		 * the minority of clients which don't.
+		 */
+		http_Unset(bo->bereq, H_Accept_Encoding);
+		http_SetHeader(bo->bereq, "Accept-Encoding: gzip");
+	}
+
+	VCL_miss_method(req);
+
+	if (req->handling == VCL_RET_FETCH) {
+		CHECK_OBJ_NOTNULL(bo, BUSYOBJ_MAGIC);
+		req->req_step = R_STP_FETCH;
+		return (0);
+	}
+
+	AZ(HSH_Deref(&wrk->stats, req->objcore, NULL));
+	req->objcore = NULL;
+	http_Teardown(bo->bereq);
+	VBO_DerefBusyObj(wrk, &req->busyobj);
+
+	switch(req->handling) {
+	case VCL_RET_ERROR:
+		req->req_step = R_STP_ERROR;
+		break;
+	case VCL_RET_PASS:
+		req->req_step = R_STP_PASS;
+		break;
+	case VCL_RET_RESTART:
+		req->req_step = R_STP_RESTART;
+		break;
+	default:
+		WRONG("Illegal action in vcl_miss{}");
+	}
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * Start pass processing by getting headers from backend, then
+ * continue in passbody.
+ *
+DOT subgraph xcluster_pass {
+DOT	pass [
+DOT		shape=record
+DOT		label="{cnt_pass:|(XXX: deref obj.)|filter req.*-\>bereq.|{vcl_pass\{\}|{req.*|bereq.*}}|{<err>error?|<rst>restart?}|<pass>create anon obj}"
+DOT	]
+DOT }
+DOT pass:pass -> fetch [style=bold, color=red]
+XDOT pass:rst -> rst_pass [label="restart",color=purple]
+XDOT rst_pass [label="RESTART",shape=plaintext]
+XDOT pass:err -> err_pass [label="error"]
+XDOT err_pass [label="ERROR",shape=plaintext]
+ */
+
+static int
+cnt_pass(struct worker *wrk, struct req *req)
+{
+	struct busyobj *bo;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+	AZ(req->objcore);
+	AZ(req->obj);
+	AZ(req->busyobj);
+
+	req->busyobj = VBO_GetBusyObj(wrk);
+	bo = req->busyobj;
+	bo->vsl->wid = req->sp->vsl_id;
+	bo->refcount = 2;
+	HTTP_Setup(bo->bereq, bo->ws, bo->vsl, HTTP_Bereq);
+	http_FilterReq(req, HTTPH_R_PASS);
+
+	VCL_pass_method(req);
+
+	if (req->handling == VCL_RET_ERROR) {
+		http_Teardown(bo->bereq);
+		VBO_DerefBusyObj(wrk, &req->busyobj);
+		req->req_step = R_STP_ERROR;
+		return (0);
+	}
+	assert(req->handling == VCL_RET_PASS);
+	req->acct_req.pass++;
+	req->req_step = R_STP_FETCH;
+
+	req->objcore = HSH_NewObjCore(wrk);
+	req->objcore->busyobj = bo;
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * Ship the request header to the backend unchanged, then pipe
+ * until one of the ends close the connection.
+ *
+DOT subgraph xcluster_pipe {
+DOT	pipe [
+DOT		shape=ellipse
+DOT		label="Filter req.->bereq."
+DOT	]
+DOT	vcl_pipe [
+DOT		shape=record
+DOT		label="vcl_pipe()|req.\nbereq\."
+DOT	]
+DOT	pipe_do [
+DOT		shape=ellipse
+DOT		label="send bereq.\npipe until close"
+DOT	]
+DOT	vcl_pipe -> pipe_do [label="pipe",style=bold,color=orange]
+DOT	pipe -> vcl_pipe [style=bold,color=orange]
+DOT }
+DOT pipe_do -> DONE [style=bold,color=orange]
+DOT vcl_pipe -> err_pipe [label="error"]
+DOT err_pipe [label="ERROR",shape=plaintext]
+ */
+
+static int
+cnt_pipe(struct worker *wrk, struct req *req)
+{
+	struct busyobj *bo;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+	AZ(req->busyobj);
+
+	req->acct_req.pipe++;
+	req->busyobj = VBO_GetBusyObj(wrk);
+	bo = req->busyobj;
+	bo->vsl->wid = req->sp->vsl_id;
+	HTTP_Setup(bo->bereq, bo->ws, bo->vsl, HTTP_Bereq);
+	http_FilterReq(req, 0);
+
+	VCL_pipe_method(req);
+
+	if (req->handling == VCL_RET_ERROR)
+		INCOMPL();
+	assert(req->handling == VCL_RET_PIPE);
+
+	PipeRequest(req);
+	assert(WRW_IsReleased(wrk));
+	http_Teardown(bo->bereq);
+	VBO_DerefBusyObj(wrk, &req->busyobj);
+	return (1);
+}
+
+/*--------------------------------------------------------------------
+ *
+DOT subgraph xcluster_restart {
+DOT	restart [
+DOT		shape=record
+DOT		label="{cnt_restart}"
+DOT	]
+DOT }
+DOT RESTART -> restart [color=purple]
+DOT restart -> recv [color=purple]
+ */
+
+static int
+cnt_restart(const struct worker *wrk, struct req *req)
+{
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+
+	req->director = NULL;
+	if (++req->restarts >= cache_param->max_restarts) {
+		req->err_code = 503;
+		req->req_step = R_STP_ERROR;
+	} else {
+		req->err_code = 0;
+		req->req_step = R_STP_RECV;
+	}
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * RECV
+ * We have a complete request, set everything up and start it.
+ * We can come here both with a request from the client and with
+ * a interior request during ESI delivery.
+ *
+DOT subgraph xcluster_recv {
+DOT	recv [
+DOT		shape=record
+DOT		label="{cnt_recv:|{vcl_recv\{\}|req.*}|{<pipe>pipe?|<pass>pass?|<error>error?|<lookup>lookup?}}"
+DOT	]
+DOT }
+DOT subgraph xcluster_hash {
+DOT	hash [
+DOT		shape=record
+DOT		label="{cnt_recv:|{vcl_hash\{\}|req.*}}"
+DOT	]
+DOT }
+DOT ESI_REQ [ shape=hexagon ]
+DOT ESI_REQ -> recv
+DOT recv:pipe -> pipe [style=bold,color=orange]
+DOT recv:pass -> pass [style=bold,color=red]
+#DOT recv:error -> err_recv
+#DOT err_recv [label="ERROR",shape=plaintext]
+DOT recv:lookup -> hash [style=bold,color=green]
+DOT hash -> lookup [label="hash",style=bold,color=green]
+ */
+
+static int
+cnt_recv(const struct worker *wrk, struct req *req)
+{
+	unsigned recv_handling;
+	struct SHA256Context sha256ctx;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	CHECK_OBJ_NOTNULL(req->vcl, VCL_CONF_MAGIC);
+	AZ(req->obj);
+	AZ(req->busyobj);
+
+	/* By default we use the first backend */
+	AZ(req->director);
+	req->director = req->vcl->director[0];
+	AN(req->director);
+
+	req->disable_esi = 0;
+	req->hash_always_miss = 0;
+	req->hash_ignore_busy = 0;
+	req->client_identity = NULL;
+
+	http_CollectHdr(req->http, H_Cache_Control);
+
+	VCL_recv_method(req);
+	recv_handling = req->handling;
+
+	if (cache_param->http_gzip_support &&
+	     (recv_handling != VCL_RET_PIPE) &&
+	     (recv_handling != VCL_RET_PASS)) {
+		if (RFC2616_Req_Gzip(req->http)) {
+			http_Unset(req->http, H_Accept_Encoding);
+			http_SetHeader(req->http, "Accept-Encoding: gzip");
+		} else {
+			http_Unset(req->http, H_Accept_Encoding);
+		}
+	}
+
+	req->sha256ctx = &sha256ctx;	/* so HSH_AddString() can find it */
+	SHA256_Init(req->sha256ctx);
+	VCL_hash_method(req);
+	assert(req->handling == VCL_RET_HASH);
+	SHA256_Final(req->digest, req->sha256ctx);
+	req->sha256ctx = NULL;
+
+	if (!strcmp(req->http->hd[HTTP_HDR_REQ].b, "HEAD"))
+		req->wantbody = 0;
+	else
+		req->wantbody = 1;
+
+	switch(recv_handling) {
+	case VCL_RET_LOOKUP:
+		req->req_step = R_STP_LOOKUP;
+		return (0);
+	case VCL_RET_PIPE:
+		if (req->esi_level > 0) {
+			/* XXX: VSL something */
+			INCOMPL();
+			return (1);
+		}
+		req->req_step = R_STP_PIPE;
+		return (0);
+	case VCL_RET_PASS:
+		req->req_step = R_STP_PASS;
+		return (0);
+	case VCL_RET_ERROR:
+		req->req_step = R_STP_ERROR;
+		return (0);
+	default:
+		WRONG("Illegal action in vcl_recv{}");
+	}
+}
+
+/*--------------------------------------------------------------------
+ * START
+ * First time we see a request
+ *
+DOT start [
+DOT	shape=box
+DOT	label="cnt_start:\nDissect request\nHandle expect"
+DOT ]
+DOT start -> recv [style=bold,color=green]
+DOT start -> DONE [label=errors]
+ */
+
+static int
+cnt_start(struct worker *wrk, struct req *req)
+{
+	char *p;
+	const char *r = "HTTP/1.1 100 Continue\r\n\r\n";
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+	AZ(req->restarts);
+	AZ(req->obj);
+	AZ(req->vcl);
+	AZ(req->esi_level);
+	assert(!isnan(req->t_req));
+
+	/* Update stats of various sorts */
+	wrk->stats.client_req++;
+	req->acct_req.req++;
+
+	/* Assign XID and log */
+	req->xid = ++xids;				/* XXX not locked */
+	VSLb(req->vsl, SLT_ReqStart, "%s %s %u",
+	    req->sp->addr, req->sp->port, req->xid);
+
+	/* Borrow VCL reference from worker thread */
+	VCL_Refresh(&wrk->vcl);
+	req->vcl = wrk->vcl;
+	wrk->vcl = NULL;
+
+	EXP_Clr(&req->exp);
+
+	HTTP_Setup(req->http, req->ws, req->vsl, HTTP_Req);
+	req->err_code = http_DissectRequest(req);
+
+	/* If we could not even parse the request, just close */
+	if (req->err_code == 400) {
+		SES_Close(req->sp, SC_RX_JUNK);
+		return (1);
+	}
+
+	req->ws_req = WS_Snapshot(req->ws);
+
+	req->doclose = http_DoConnection(req->http);
+
+	/*
+	 * We want to deal with Expect: headers the first time we
+	 * attempt the request, and remove them before we move on.
+	 */
+	if (req->err_code == 0 && http_GetHdr(req->http, H_Expect, &p)) {
+		if (strcasecmp(p, "100-continue")) {
+			req->err_code = 417;
+		} else if (strlen(r) != write(req->sp->fd, r, strlen(r))) {
+			SES_Close(req->sp, SC_REM_CLOSE);
+			return (1);
+		}
+	}
+	http_Unset(req->http, H_Expect);
+
+	/* XXX: pull in req-body and make it available instead. */
+	req->reqbodydone = 0;
+
+	HTTP_Copy(req->http0, req->http);	/* Copy for restart/ESI use */
+
+	if (req->err_code)
+		req->req_step = R_STP_ERROR;
+	else
+		req->req_step = R_STP_RECV;
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * Central state engine dispatcher.
+ *
+ * Kick the session around until it has had enough.
+ *
+ */
+
+static void
+cnt_diag(struct req *req, const char *state)
+{
+
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+
+	VSLb(req->vsl,  SLT_Debug, "vsl_id %u STP_%s sp %p obj %p vcl %p",
+	    req->sp->vsl_id, state, req->sp, req->obj, req->vcl);
+	VSL_Flush(req->vsl, 0);
+}
+
+int
+CNT_Request(struct worker *wrk, struct req *req)
+{
+	int done;
+
+	CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+	CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+
+	/*
+	 * Possible entrance states
+	 */
+	assert(
+	    req->req_step == R_STP_LOOKUP ||
+	    req->req_step == R_STP_START ||
+	    req->req_step == R_STP_RECV);
+
+	req->wrk = wrk;
+
+	for (done = 0; !done; ) {
+		/*
+		 * This is a good place to be paranoid about the various
+		 * pointers still pointing to the things we expect.
+		 */
+		CHECK_OBJ_NOTNULL(wrk, WORKER_MAGIC);
+		CHECK_OBJ_ORNULL(wrk->nobjhead, OBJHEAD_MAGIC);
+		WS_Assert(wrk->aws);
+		CHECK_OBJ_NOTNULL(req, REQ_MAGIC);
+
+		switch (req->req_step) {
+#define REQ_STEP(l,u,arg) \
+		    case R_STP_##u: \
+			if (cache_param->diag_bitmap & 0x01) \
+				cnt_diag(req, #u); \
+			done = cnt_##l arg; \
+		        break;
+#include "tbl/steps.h"
+#undef REQ_STEP
+		default:
+			WRONG("State engine misfire");
+		}
+		WS_Assert(wrk->aws);
+		CHECK_OBJ_ORNULL(wrk->nobjhead, OBJHEAD_MAGIC);
+	}
+	if (done == 1) {
+		/* XXX: Workaround for pipe */
+		if (req->sp->fd >= 0) {
+			VSLb(req->vsl, SLT_Length, "%ju",
+			    (uintmax_t)req->req_bodybytes);
+		}
+		VSLb(req->vsl, SLT_ReqEnd, "%u %.9f %.9f %.9f %.9f %.9f",
+		    req->xid,
+		    req->t_req,
+		    req->sp->t_idle,
+		    req->sp->t_idle - req->t_resp,
+		    req->t_resp - req->t_req,
+		    req->sp->t_idle - req->t_resp);
+
+		/* done == 2 was charged by cache_hash.c */
+		SES_Charge(wrk, req);
+	}
+
+	req->wrk = NULL;
+
+	assert(WRW_IsReleased(wrk));
+	return (done);
+}
+
+/*
+DOT }
+*/
+
+/*--------------------------------------------------------------------
+ * Debugging aids
+ */
+
+static void
+cli_debug_xid(struct cli *cli, const char * const *av, void *priv)
+{
+	(void)priv;
+	if (av[2] != NULL)
+		xids = strtoul(av[2], NULL, 0);
+	VCLI_Out(cli, "XID is %u", xids);
+}
+
+/*
+ * Default to seed=1, this is the only seed value POSIXl guarantees will
+ * result in a reproducible random number sequence.
+ */
+static void
+cli_debug_srandom(struct cli *cli, const char * const *av, void *priv)
+{
+	(void)priv;
+	unsigned seed = 1;
+
+	if (av[2] != NULL)
+		seed = strtoul(av[2], NULL, 0);
+	srandom(seed);
+	srand48(random());
+	VCLI_Out(cli, "Random(3) seeded with %u", seed);
+}
+
+static struct cli_proto debug_cmds[] = {
+	{ "debug.xid", "debug.xid",
+		"\tExamine or set XID\n", 0, 1, "d", cli_debug_xid },
+	{ "debug.srandom", "debug.srandom",
+		"\tSeed the random(3) function\n", 0, 1, "d",
+		cli_debug_srandom },
+	{ NULL }
+};
+
+/*--------------------------------------------------------------------
+ *
+ */
+
+void
+CNT_Init(void)
+{
+
+	srandomdev();
+	srand48(random());
+	xids = random();
+	CLI_AddFuncs(debug_cmds);
+}