Merge PR #778 (v2016.03 release) onto master

Makefile

@@ -29,4 +29,17 @@ clean:
 	(cd lib/luajit && $(MAKE) clean)
 	(cd src; $(MAKE) clean; rm -rf syscall.lua syscall)
 
+PACKAGE:=snabbswitch
+DIST_BINARY:=snabb
+BUILDDIR:=$(shell pwd)
+
+dist: DISTDIR:=$(BUILDDIR)/$(PACKAGE)-$(shell git describe --tags)
+dist: all
+	mkdir "$(DISTDIR)"
+	git clone "$(BUILDDIR)" "$(DISTDIR)/snabbswitch"
+	rm -rf "$(DISTDIR)/snabbswitch/.git"
+	cp "$(BUILDDIR)/src/snabb" "$(DISTDIR)/$(DIST_BINARY)"
+	cd "$(DISTDIR)/.." && tar cJvf "`basename '$(DISTDIR)'`.tar.xz" "`basename '$(DISTDIR)'`"
+	rm -rf "$(DISTDIR)"
+
 .SERIAL: all

lib/luajit/src/lj_record.c

@@ -719,9 +719,14 @@ void lj_record_tailcall(jit_State *J, BCReg func, ptrdiff_t nargs)
   /* Move func + args down. */
   memmove(&J->base[-1], &J->base[func], sizeof(TRef)*(J->maxslot+1));
   /* Note: the new TREF_FRAME is now at J->base[-1] (even for slot #0). */
-  /* Tailcalls can form a loop, so count towards the loop unroll limit. */
-  if (++J->tailcalled > J->loopunroll)
-    lj_trace_err(J, LJ_TRERR_LUNROLL);
+
+  J->tailcalled++;
+
+  /* Although it's true that tail calls can form a loop, the Lua programming
+  ** idiom does not encourage this.  Instead of counting tail calls towards the
+  ** unroll limit and potentially causing important traces without loops to
+  ** abort, eventually blacklist, and fall back to the interpreter, just rely on
+  ** JIT_P_maxrecord to catch runaway tail-call loops. */
 }
 
 /* Check unroll limits for down-recursion. */

src/Makefile

@@ -16,7 +16,7 @@ CSRC   = $(shell find . -regex '[^\#]*\.c' -not -regex './arch/.*' -printf '%P '
 CHDR   = $(shell find . -regex '[^\#]*\.h' -printf '%P ')
 ASM    = $(shell find . -regex '[^\#]*\.dasl' -printf '%P ')
 ARCHSRC= $(shell find . -regex '^./arch/[^\#]*\.c' -printf '%P ')
-RMSRC  = $(shell find . -name  README.md.src -printf '%P ')
+RMSRC  = $(shell find . -name  README.src.md -printf '%P ')
 # regexp is to include program/foo but not program/foo/bar
 PROGRAM = $(shell find program -regex '^[^/]+/[^/]+' -type d -printf '%P ')
 # sort to eliminate potential duplicate of programs.inc
@@ -30,7 +30,7 @@ ARCHOBJ:= $(patsubst %.c,obj/%_c.o,    $(ARCHSRC))
 ASMOBJ := $(patsubst %.dasl,obj/%_dasl.o,   $(ASM))
 JITOBJS:= $(patsubst %,obj/jit_%.o,$(JITSRC))
 EXTRAOBJS := obj/jit_tprof.o obj/jit_vmprof.o obj/strict.o
-RMOBJS := $(patsubst %.src,%,$(RMSRC))
+RMOBJS := $(patsubst %.src.md,%.md,$(RMSRC))
 INCOBJ := $(patsubst %.inc,obj/%_inc.o, $(INCSRC))
 EXE    := bin/snabb $(patsubst %,bin/%,$(PROGRAM))
 
@@ -152,7 +152,7 @@ $(JITOBJS): obj/jit_%.o: ../lib/luajit/src/jit/%.lua $(OBJDIR)
 	$(Q) luajit -bg -n $(patsubst obj/jit_%.o, jit.%, $@) $< $@
 
 
-$(RMOBJS): %: %.src
+$(RMOBJS): %.md: %.src.md
 	$(E) "MARKDOWN  $@"
 	$(Q) scripts/process-markdown $< > $@
 

src/README.md

@@ -333,49 +333,24 @@ Allocate packet and fill it with the contents of *string*.
 
 ## Memory (core.memory)
 
-Snabb Switch does two things specially when it comes to memory: It
-runs with a fixed physical memory map and it allocates *huge pages*
-from the operating system.
+Snabb Switch allocates special
+[DMA](https://en.wikipedia.org/wiki/Direct_memory_access) memory that
+can be accessed directly by network cards. The important
+characteristic of DMA memory is being located in contiguous physical
+memory at a stable address.
 
-Running with a fixed memory map means that every virtual address in the
-Snabb Switch process has a fixed *physical address* in the RAM
-chips. This means that we are always able to convert from a virtual
-address in our process to a physical address that other hardware (for
-example, a network card) can use for DMA.
-
-Huge pages (also known as *HugeTLB pages*) are how we allocate large
-amounts of contiguous memory, typically 2MB at a time. Hardware devices
-sometimes require this, for example a network card's *descriptor ring*
-may require a list of pointers to available buffers in physical memory.
+— Function **memory.dma_alloc** *bytes*
 
-— Variable **memory.chunks**
+Returns a pointer to *bytes* of new DMA memory.
 
-List of all allocated huge pages. Read-only. Each huge page is
-represented by a table with the following keys:
+— Function **memory.virtual_to_physical** *pointer*
 
-* `pointer` - Virtual address
-* `physical` - Physical address
-* `size` -  Size in bytes
-* `used` - Bytes used
+Returns the physical address (`uint64_t`) the DMA memory at *pointer*.
 
 — Variable **memory.huge_page_size**
 
 Size of a single huge page in bytes. Read-only.
 
-— Variable **huge_page_bits**
-
-Number of address bits per huge page. Read-only.
-
-— Function **memory.dma_alloc** *bytes*
-
-Allocate *bytes* of DMA-friendly memory. Returns virtual memory pointer,
-physical address, and actual size.
-
-— Function **memory.virtual_to_physical** *virtual_address*
-
-Returns the physical address of memory at *virtual_address*.
-
-
 
 ## Lib (core.lib)
 

src/README.md.src → src/README.src.md

@@ -359,56 +359,24 @@ Allocate packet and fill it with the contents of *string*.
 
 ## Memory (core.memory)
 
-Snabb Switch does two things specially when it comes to memory: It
-runs with a fixed physical memory map and it allocates *huge pages*
-from the operating system.
+Snabb Switch allocates special
+[DMA](https://en.wikipedia.org/wiki/Direct_memory_access) memory that
+can be accessed directly by network cards. The important
+characteristic of DMA memory is being located in contiguous physical
+memory at a stable address.
 
-Running with a fixed memory map means that every virtual address in the
-Snabb Switch process has a fixed *physical address* in the RAM
-chips. This means that we are always able to convert from a virtual
-address in our process to a physical address that other hardware (for
-example, a network card) can use for DMA.
-
-Huge pages (also known as *HugeTLB pages*) are how we allocate large
-amounts of contiguous memory, typically 2MB at a time. Hardware devices
-sometimes require this, for example a network card's *descriptor ring*
-may require a list of pointers to available buffers in physical memory.
-
-— Variable **memory.chunks**
-
-List of all allocated huge pages. Read-only. Each huge page is
-represented by a table with the following keys:
-
-* `pointer` - Virtual address
-* `physical` - Physical address
-* `size` -  Size in bytes
-* `used` - Bytes used
+— Function **memory.dma_alloc** *bytes*
 
-— Variable **memory.dma_min_addr**
+Returns a pointer to *bytes* of new DMA memory.
 
-— Variable **memory.dma_max_addr**
+— Function **memory.virtual_to_physical** *pointer*
 
-Lowest and highest addresses of valid DMA memory. Useful information for
-creating memory maps. Read-only.
+Returns the physical address (`uint64_t`) the DMA memory at *pointer*.
 
 — Variable **memory.huge_page_size**
 
 Size of a single huge page in bytes. Read-only.
 
-— Variable **huge_page_bits**
-
-Number of address bits per huge page. Read-only.
-
-— Function **memory.dma_alloc** *bytes*
-
-Allocate *bytes* of DMA-friendly memory. Returns virtual memory pointer,
-physical address, and actual size.
-
-— Function **memory.virtual_to_physical** *virtual_address*
-
-Returns the physical address of memory at *virtual_address*.
-
-
 
 ## Lib (core.lib)
 

src/apps/basic/basic_apps.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local app = require("core.app")

src/apps/bridge/base.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 -- Base class for an Ethernet bridge with split-horizon semantics.
 --
 -- A bridge conists of any number of ports, each of which is a member

src/apps/bridge/flooding.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 -- This class derives from lib.bridge.base and implements the simplest
 -- possible bridge, which floods a packet arriving on a port to all
 -- destination ports within its scope according to the split-horizon

src/apps/bridge/learning.h

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 /* Type for port and split-horizon group handles */
 typedef uint16_t handle_t;
 

src/apps/bridge/learning.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 -- This class derives from lib.bridge.base and implements a "learning
 -- bridge" using a MAC address table provided by apps.bridge.mac_table
 -- to store the set of source addresses of packets arriving on all

src/apps/bridge/mac_table.c

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 #include <inttypes.h>
 #include "learning.h"
 

src/apps/bridge/mac_table.lua

@@ -1,3 +1,4 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
 --
 -- This module implements a MAC address table as part of the learning
 -- bridge app (apps.bridge.learning).  It associates a MAC address

src/apps/csv.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local app = require("core.app")

src/apps/intel/intel.h

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 // RX descriptor written by software.
 struct rx_desc {
    uint64_t address;    // 64-bit address of receive buffer

src/apps/intel/intel10g.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 --- Device driver for the Intel 82599 10-Gigabit Ethernet controller.
 --- This is one of the most popular production 10G Ethernet
 --- controllers on the market and it is readily available in

src/apps/intel/intel1g.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 -- intel1g: Device driver app for Intel 1G network cards
 -- 
 -- This is a device driver for Intel i210, i350 families of 1G network cards.

src/apps/intel/intel_app.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local zone = require("jit.zone")

src/apps/intel/loadgen.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local ffi = require("ffi")

src/apps/ipv6/nd_light.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 -- This app implements a small subset of IPv6 neighbor discovery
 -- (RFC4861).  It has two ports, north and south.  The south port
 -- attaches to a port on which ND must be performed.  The north port

src/apps/ipv6/ns_responder.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 -- This app acts as a responder for neighbor solicitaions for a
 -- specific target address and as a relay for all other packets.  It
 -- has two ports, north and south.  The south port attaches to a port

src/apps/keyed_ipv6_tunnel/tunnel.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 -- http://tools.ietf.org/html/draft-mkonstan-keyed-ipv6-tunnel-01

src/apps/packet_filter/conntrack.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 -- conntrack.lua -- Connection tracking for IPv4/IPv6 TCP/UDP sessions
 --
 -- This module exposes the following API:

src/apps/packet_filter/pcap_filter.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local app = require("core.app")

src/apps/pcap/pcap.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local ffi = require("ffi")

src/apps/rate_limiter/rate_limiter.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(..., package.seeall)
 
 local app = require("core.app")

src/apps/socket/raw.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(..., package.seeall)
 
 local S = require("syscall")
@@ -14,14 +16,13 @@ RawSocket = {}
 
 function RawSocket:new (ifname)
    assert(ifname)
+   local index, err = S.util.if_nametoindex(ifname)
+   if not index then error(err) end
+
    local tp = h.htons(c.ETH_P["ALL"])
    local sock, err = S.socket(c.AF.PACKET, bit.bor(c.SOCK.RAW, c.SOCK.NONBLOCK), tp)
    if not sock then error(err) end
-   local index, err = S.util.if_nametoindex(ifname, sock)
-   if err then
-      S.close(sock)
-      error(err)
-   end
+
    local addr = t.sockaddr_ll{sll_family = c.AF.PACKET, sll_ifindex = index, sll_protocol = tp}
    local ok, err = S.bind(sock, addr)
    if not ok then

src/apps/solarflare/ef_vi.h

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 #include <time.h>
 
 /* etherfabric/base.h */

src/apps/solarflare/poll.c

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 /* poll.c - Poll multiple ef_vi interfaces in one FFI call to save on FFI overhead */
 
 #include <stdint.h>

src/apps/solarflare/selftest.sh

@@ -1,3 +1,3 @@
-#!/bin/bash
+#!/usr/bin/env bash
 
 ./snabb snabbmark solarflare 10e6 128 10

src/apps/solarflare/solarflare.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local lib      = require("core.lib")

src/apps/tap/selftest.sh

@@ -1,4 +1,4 @@
-#!/bin/bash
+#!/usr/bin/env bash
 sudo ip netns add snabbtest || exit $TEST_SKIPPED
 sudo ip netns exec snabbtest ip link add name snabbtest type bridge
 sudo ip netns exec snabbtest ip link set up dev snabbtest

src/apps/tap/tap.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(..., package.seeall)
 
 local S = require("syscall")

src/apps/test/synth.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 module(...,package.seeall)
 
 local ffi = require("ffi")

src/apps/vhost/vhost.h

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 // vhost_memory structure is used to declare which memory address
 // ranges we want to use for DMA. The kernel uses this to create a
 // shared memory mapping.

src/apps/vhost/vhost_user.c

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 #include <assert.h>
 #include <errno.h>
 #include <fcntl.h>

src/apps/vhost/vhost_user.h

@@ -1,3 +1,5 @@
+/* Use of this source code is governed by the Apache 2.0 license; see COPYING. */
+
 enum {
     VHOST_USER_MEMORY_MAX_NREGIONS = 8
 };

src/apps/vhost/vhost_user.lua

@@ -1,3 +1,5 @@
+-- Use of this source code is governed by the Apache 2.0 license; see COPYING.
+
 --
 -- See http://www.virtualopensystems.com/en/solutions/guides/snabbswitch-qemu/