Add objsize_max

CHANGES.rst

+..
+	Copyright 2023 UPLEX Nils Goroll Systemoptimierung
+	SPDX-License-Identifier: BSD-2-Clause
+	See LICENSE file for full text of license
+
+===================
+About this document
+===================
+
+.. keep this section at the top!
+
+This document contains a log of noteworthy changes to SLASH/ from new
+to old.
+
+==============
+SLASH/ Changes
+==============
+
+* The ``objsize_max`` tunable has been added, which defaults to 1/4 of
+  *dsksize*.
+
+  The tunable is going to be used to allow SLASH/ to make an educated
+  guess about how much space to allocate in advance when the object
+  size is not known upfront, for example with *chunked encoding*.
+
+Base of the changelog: e3a454096759cafd30d9008f4b11cc754297b250

src/fellow_cache.c

@@ -3553,6 +3553,11 @@ fellow_busy_obj_getspace(struct fellow_busy *fbo, size_t *sz, uint8_t **ptr)
 	max = (size_t)1 << fbo->fc->tune->chunk_exponent;
 	assert(max <= FIO_MAX);
 
+	if (fbo->sz_returned >= fbo->fc->tune->objsize_max)
+		return (FCR_ALLOCFAIL("objsize_max reached"));
+	if (*sz + fbo->sz_returned > fbo->fc->tune->objsize_max)
+		*sz = fbo->fc->tune->objsize_max - fbo->sz_returned;
+
 	/*
 	 * for content-length, varnish-cache will ask for less and less.
 	 *

src/fellow_storage.c

@@ -744,6 +744,10 @@ fellow_stream_f(void *priv, unsigned flush, const void *ptr, ssize_t len)
 	int r;
 
 	CAST_OBJ_NOTNULL(fs, priv, FELLOW_STREAM_MAGIC);
+
+	if (fs->state == BOS_FAILED)
+		return (-1);
+
 	assert((fs->state == BOS_FINISHED && fs->available >= fs->written) ||
 	    (fs->state == BOS_STREAM && fs->available > fs->written));
 	assert(fs->checkpoint == fs->written);

src/fellow_tune.c

@@ -100,6 +100,9 @@ stvfe_tune_check(struct stvfe_tune *tune)
 		tune->dsk_reserve_chunks = l;
 	}
 
+	if (tune->objsize_max == 0)
+		tune->objsize_max = (tune->dsksz >> 2);
+
 //lint --e{685,568} misc const comparisons
 #define TUNE(t, n, d, min, max)						\
 	if (tune->n < (min))						\

src/tbl/fellow_tunables.h

@@ -45,6 +45,7 @@ TUNE(uint8_t, wait_table_exponent, 10, 6, 32);
 TUNE(unsigned, dsk_reserve_chunks, 4, 2, UINT_MAX);
 TUNE(unsigned, mem_reserve_chunks, 1, 0, UINT_MAX);
 TUNE(size_t, objsize_hint, 256 * 1024, 4096, SIZE_MAX);
+TUNE(size_t, objsize_max, 0, 0, SIZE_MAX);
 TUNE(size_t, discard_immediate, 256 * 1024, 4096, SIZE_MAX);
 TUNE(unsigned, readahead, 2, 0, UINT_MAX);
 TUNE(unsigned, io_batch_min, 8, 1, UINT_MAX);

src/vmod_slash.man.rst

@@ -51,7 +51,7 @@ vcl storage objects and methods
 
   new xfellow = slash.fellow(STRING filename, BYTES dsksize, BYTES memsize, BYTES objsize_hint, BOOL delete)
 
-      STRING xfellow.tune([INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
+      STRING xfellow.tune([INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [BYTES objsize_max], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
 
       STEVEDORE xfellow.storage()
 
@@ -68,7 +68,7 @@ vcl functions
 
   STRING tune_buddy(STEVEDORE storage, [INT chunk_exponent], [BYTES chunk_bytes], [INT reserve_chunks], [INT cram])
 
-  STRING tune_fellow(STEVEDORE storage, [INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
+  STRING tune_fellow(STEVEDORE storage, [INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [BYTES objsize_max], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
 
 vcl loadmasters (storage routers)
 ---------------------------------
@@ -479,8 +479,8 @@ will be used (which might fail of insufficient memory is available).
 
 .. _xfellow.tune():
 
-STRING xfellow.tune([INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT wait_table_exponent], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+STRING xfellow.tune([INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT wait_table_exponent], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [BYTES objsize_max], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 
 ::
 
@@ -494,6 +494,7 @@ STRING xfellow.tune([INT logbuffer_size], [DURATION logbuffer_flush_interval], [
             [INT dsk_reserve_chunks],
             [INT mem_reserve_chunks],
             [BYTES objsize_hint],
+            [BYTES objsize_max],
             [INT cram],
             [INT readahead],
             [BYTES discard_immediate],
@@ -630,6 +631,18 @@ fellow storage can be fine tuned:
   used to (re)size the fixed log space. The value should be chosen as
   a safe *lower* bound of average object size.
 
+* *objsize_max*
+
+  - unit: bytes
+  - default: 0
+
+  specifies the maxiumum object size which fellow will accept.
+
+  The default of ``0`` represents 1/4th of *dsksize*. It is strongly
+  recommended to not use a value higher than that.
+
+  The effectively enforced value is rounded up to 4KB.
+
 * *cram* is documented in `xbuddy.tune()`_
 
 * *readahead*
@@ -811,8 +824,8 @@ Can only be called from ``vcl_init {}``.
 
 .. _slash.tune_fellow():
 
-STRING tune_fellow(STEVEDORE storage, [INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT wait_table_exponent], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+STRING tune_fellow(STEVEDORE storage, [INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT wait_table_exponent], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [BYTES objsize_max], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 
 ::
 
@@ -827,6 +840,7 @@ STRING tune_fellow(STEVEDORE storage, [INT logbuffer_size], [DURATION logbuffer_
       [INT dsk_reserve_chunks],
       [INT mem_reserve_chunks],
       [BYTES objsize_hint],
+      [BYTES objsize_max],
       [INT cram],
       [INT readahead],
       [BYTES discard_immediate],

src/vmod_slash.vcc

@@ -36,7 +36,7 @@ vcl storage objects and methods
 
   new xfellow = slash.fellow(STRING filename, BYTES dsksize, BYTES memsize, BYTES objsize_hint, BOOL delete)
 
-      STRING xfellow.tune([INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
+      STRING xfellow.tune([INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [BYTES objsize_max], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
 
       STEVEDORE xfellow.storage()
 
@@ -53,7 +53,7 @@ vcl functions
 
   STRING tune_buddy(STEVEDORE storage, [INT chunk_exponent], [BYTES chunk_bytes], [INT reserve_chunks], [INT cram])
 
-  STRING tune_fellow(STEVEDORE storage, [INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
+  STRING tune_fellow(STEVEDORE storage, [INT logbuffer_size], [DURATION logbuffer_flush_interval], [REAL log_rewrite_ratio], [INT chunk_exponent], [BYTES chunk_bytes], [INT dsk_reserve_chunks], [INT mem_reserve_chunks], [BYTES objsize_hint], [BYTES objsize_max], [INT cram], [INT readahead], [BYTES discard_immediate], [INT io_batch_min], [INT io_batch_max], [ENUM hash_obj], [ENUM hash_log], [ENUM ioerr_obj], [ENUM ioerr_log], [ENUM allocerr_obj], [ENUM allocerr_log])
 
 vcl loadmasters (storage routers)
 ---------------------------------
@@ -433,6 +433,7 @@ $Method STRING .tune(
 	[ INT		dsk_reserve_chunks ],
 	[ INT		mem_reserve_chunks ],
 	[ BYTES		objsize_hint ],
+	[ BYTES		objsize_max ],
 	[ INT		cram ],
 	[ INT		readahead ],
 	[ BYTES		discard_immediate ],
@@ -568,6 +569,18 @@ fellow storage can be fine tuned:
   used to (re)size the fixed log space. The value should be chosen as
   a safe *lower* bound of average object size.
 
+* *objsize_max*
+
+  - unit: bytes
+  - default: 0
+
+  specifies the maxiumum object size which fellow will accept.
+
+  The default of ``0`` represents 1/4th of *dsksize*. It is strongly
+  recommended to not use a value higher than that.
+
+  The effectively enforced value is rounded up to 4KB.
+
 * *cram* is documented in `xbuddy.tune()`_
 
 * *readahead*
@@ -749,6 +762,7 @@ $Function STRING tune_fellow(
 	[ INT		dsk_reserve_chunks ],
 	[ INT		mem_reserve_chunks ],
 	[ BYTES		objsize_hint ],
+	[ BYTES		objsize_max ],
 	[ INT		cram ],
 	[ INT		readahead ],
 	[ BYTES		discard_immediate ],

src/vtc/fellow_coverage.vtc

@@ -7,7 +7,10 @@ shell -err -expect " fellow file can only be used once" \
      -sfellow2=fellow,clashfile,100MB,1MB,64KB \
      }
 
-varnish v1 -vcl {
+varnish v1 \
+    -arg "-E${libvmod_slash}" \
+    -arg "-sfellow1=fellow,fellowfile,1MB,1MB,64KB" \
+    -vcl {
 	backend none none;
 } -start
 
@@ -19,4 +22,67 @@ varnish v1 -errvcl {can only be used on a fellow storage} {
 	sub vcl_init {
 		slash.tune_fellow(storage.Transient, chunk_bytes = 1KB);
 	}
-}
\ No newline at end of file
+}
+
+server s1 {
+	rxreq
+	expect req.url == "/256k"
+	txresp -bodylen 262144
+	rxreq
+	expect req.url == "/256k+1"
+	txresp -bodylen 262145
+} -start
+
+varnish v1 -vcl+backend {
+	sub vcl_backend_response {
+		set beresp.do_stream = false;
+	}
+}
+
+client c1 {
+	txreq -url "/256k"
+	rxresp
+	expect resp.status == 200
+	expect resp.bodylen == 262144
+
+	txreq -url "/256k+1"
+	rxresp
+	expect resp.status == 503
+} -start
+
+client c2 {
+	txreq -url "/256k"
+	rxresp
+	expect resp.status == 200
+	expect resp.bodylen == 262144
+
+	txreq -url "/256k+1"
+	rxresp
+	expect resp.status == 503
+} -start
+
+client c1 -wait
+client c2 -wait
+server s1 -wait
+
+server s1 {
+	rxreq
+	expect req.url == "/16k"
+	txresp -bodylen 16384
+} -start
+
+varnish v1 -vcl+backend {
+	import slash;
+	sub vcl_init {
+		slash.tune_fellow(storage.fellow1, objsize_max = 12K);
+	}
+	sub vcl_backend_response {
+		set beresp.do_stream = false;
+	}
+}
+
+client c3 {
+	txreq -url "/16k"
+	rxresp
+	expect resp.status == 503
+} -run