From 519fdbe6c8459eef86fcc674918da99cace4382d Mon Sep 17 00:00:00 2001
From: sdomi <ja@sdomi.pl>
Date: Wed, 26 Feb 2025 01:22:23 +0100
Subject: [PATCH 1/2] ws: full hecking rewrite

---
 .resources/primary_config.sh |   1 +
 src/response/101.sh          |  42 ++++++------
 src/server.sh                |  82 +++++++++++++----------
 src/ws.sh                    | 123 ++++++++++++++++++-----------------
 4 files changed, 129 insertions(+), 119 deletions(-)

diff --git a/.resources/primary_config.sh b/.resources/primary_config.sh
index dfc96e3..9f0d1d9 100644
--- a/.resources/primary_config.sh
+++ b/.resources/primary_config.sh
@@ -21,6 +21,7 @@ cfg[ssl_cert]=''
 cfg[ssl_key]=''
 
 cfg[extension]='shs'
+cfg[extension_websocket]='shx'
 #cfg[encoding]='UTF-8' # UTF-8 by default, used by iconv
 cfg[extra_headers]="server: HTTP.sh/$HTTPSH_VERSION (devel)"
 
diff --git a/src/response/101.sh b/src/response/101.sh
index 72cf7e1..6d11ec6 100755
--- a/src/response/101.sh
+++ b/src/response/101.sh
@@ -1,31 +1,29 @@
+#!/usr/bin/env bash
+[[ ! -f src/libsh/bin/bin.sh ]] && return # no websocket for you! clone the repo
+source src/libsh/bin/bin.sh
+
+if [[ "${r[status]}" == 101 && "${r[url_clean]}" != *\.${cfg[extension_websocket]} ]] ||
+   [[ "${r[status]}" == 102 && "${r[view]}" != *\.${cfg[extension_websocket]} ]]; then
+	echo "Rejecting an invalid WebSocket connection." >&2
+	declare -p r >&2
+	return
+fi
+
 echo "HTTP/1.1 101 Web Socket Protocol Handshake
 Connection: Upgrade
 Upgrade: WebSocket
 ${cfg[extra_headers]}"
 if [[ ${r[websocket_key]} != '' ]]; then
-	accept=$(echo -ne $(echo "${r[websocket_key]}""258EAFA5-E914-47DA-95CA-C5AB0DC85B11" | sha1sum | sed 's/ //g;s/-//g;s/.\{2\}/\\x&/g') | base64)
-	echo "Sec-WebSocket-Accept: "$accept
+	accept=$(echo -ne $(echo -n "${r[websocket_key]}""258EAFA5-E914-47DA-95CA-C5AB0DC85B11" | sha1sum | sed 's/ //g;s/-//g;s/.\{2\}/\\x&/g') | base64)
+	echo -n "Sec-WebSocket-Accept: $accept"
 fi
-echo -e "\r\n\r\n"
-
-#echo "Laura is cute <3"
-#WebSocket-Location: ws://localhost:1337/
-#WebSocket-Origin: http://localhost:1337/\r\n\r\n  "
+echo -ne "\r\n\r\n"
 
 source ./src/ws.sh 
+unset IFS
 
-#input=''
-#while read -N 1 chr; do
-#	input=$input$chr
-#	if [[ $chr == "\r" ]]; then
-#		break
-#	fi
-#done
-
-
-exit 0
-#while true; do
-#	read test
-#	echo $test
-#	sleep 1
-#done
+if [[ "${r[status]}" == 101 ]]; then
+	source "${r[uri]}"
+elif [[ "${r[status]}" == 102 ]]; then
+	source "${r[view]}"
+fi
diff --git a/src/server.sh b/src/server.sh
index b5ce2cb..5781bea 100755
--- a/src/server.sh
+++ b/src/server.sh
@@ -167,42 +167,44 @@ echo "$(date) - IP: ${r[ip]}, PROTO: ${r[proto]}, URL: ${r[url]}, GET_data: ${ge
 
 [[ -f "${cfg[namespace]}/routes.sh" ]] && source "${cfg[namespace]}/routes.sh"
 
-if [[ ${r[status]} != 101 ]]; then
-	for (( i=0; i<${#route[@]}; i=i+3 )); do
-		if [[ "$(grep -Poh "^${route[$((i+1))]}$" <<< "${r[url_clean]}")" != "" ]] || [[ "$(grep -Poh "^${route[$((i+1))]}$" <<< "${r[url_clean]}/")" != "" ]]; then
-			r[status]=212
-			r[view]="${route[$((i+2))]}"
-			IFS='/'
-			url=(${route[$i]})
-			url_=(${r[url_clean]})
-			unset IFS
-			for (( j=0; j<${#url[@]}; j++ )); do
-				# TODO: think about the significance of this if really hard when i'm less tired
-				if [[ ${url_[$j]} != '' && ${url[$j]} == ":"* ]]; then
-					params[${url[$j]/:/}]="${url_[$j]}"
-				fi
-			done
-			break
-		fi
-	done
-	unset IFS
-	if [[ ${r[status]} != 212 ]]; then
-		if [[ -a "${r[uri]}" && ! -r "${r[uri]}" ]]; then
-			r[status]=403
-		elif [[ "${r[uri]}" != "$(realpath "${cfg[namespace]}/${cfg[root]}")"* ]]; then
-			r[status]=403
-		elif [[ -f "${r[uri]}" ]]; then
-			r[status]=200
-		elif [[ -d "${r[uri]}" ]]; then
-			for name in ${cfg[index]}; do
-				if [[ -f "${r[uri]}/$name" ]]; then
-					r[uri]="${r[uri]}/$name"
-					r[status]=200
-				fi
-			done
+for (( i=0; i<${#route[@]}; i=i+3 )); do
+	if [[ "$(grep -Poh "^${route[$((i+1))]}$" <<< "${r[url_clean]}")" != "" ]] || [[ "$(grep -Poh "^${route[$((i+1))]}$" <<< "${r[url_clean]}/")" != "" ]]; then
+		if [[ "${r[status]}" == 101 ]]; then
+			r[status]=102 # internal, temp
 		else
-			r[status]=404
+			r[status]=212
 		fi
+		r[view]="${route[$((i+2))]}"
+		IFS='/'
+		url=(${route[$i]})
+		url_=(${r[url_clean]})
+		unset IFS
+		for (( j=0; j<${#url[@]}; j++ )); do
+			# TODO: think about the significance of this if really hard when i'm less tired
+			if [[ ${url_[$j]} != '' && ${url[$j]} == ":"* ]]; then
+				params[${url[$j]/:/}]="${url_[$j]}"
+			fi
+		done
+		break
+	fi
+done
+unset IFS
+if [[ ${r[status]} != 212 && ${r[status]} != 102 ]]; then
+	if [[ -a "${r[uri]}" && ! -r "${r[uri]}" ]]; then
+		r[status]=403
+	elif [[ "${r[uri]}" != "$(realpath "${cfg[namespace]}/${cfg[root]}")"* ]]; then
+		r[status]=403
+	elif [[ -f "${r[uri]}" ]]; then
+		[[ "${r[status]}" != 101 ]] && r[status]=200
+	elif [[ -d "${r[uri]}" ]]; then
+		for name in ${cfg[index]}; do
+			if [[ -f "${r[uri]}/$name" ]]; then
+				r[uri]="${r[uri]}/$name"
+				r[status]=200
+			fi
+		done
+	else
+		r[status]=404
 	fi
 fi
 
@@ -266,12 +268,20 @@ fi
 if [[ ${r[status]} == 210 && ${cfg[autoindex]} == true ]]; then
 	source "src/response/listing.sh"
 elif [[ ${r[status]} == 200 || ${r[status]} == 212 ]]; then
-	source "src/response/200.sh"
+	# ban the WS executable extensions
+	# default defined here for legacy compat only
+	if [[ "${r[view]}" == *".${cfg[extension_websocket]-shx}" ||
+		  "${r[url_clean]}" == *".${cfg[extension_websocket]-shx}" ]]; then
+		source "src/response/403.sh" # TODO: should be 400/500, ig
+	else
+		source "src/response/200.sh"
+	fi
 elif [[ ${r[status]} == 401 ]]; then
 	source "src/response/401.sh"
 elif [[ ${r[status]} == 404 ]]; then
 	source "src/response/404.sh"
-elif [[ ${r[status]} == 101 && ${cfg[websocket_enable]} == true ]]; then
+elif [[ ${r[status]} == 101 || ${r[status]} == 102 ]] &&
+     [[ ${cfg[websocket_enable]} == true ]]; then
 	source "src/response/101.sh"
 else
 	source "src/response/403.sh"
diff --git a/src/ws.sh b/src/ws.sh
index bd08867..8e9020a 100755
--- a/src/ws.sh
+++ b/src/ws.sh
@@ -1,69 +1,70 @@
-#!/bin/zsh
-read bytes
+#!/bin/bash
 
-echo "bytes:" > /dev/stderr
-echo $bytes | hexdump -C > /dev/stderr
+# _ws_to_varlen(num) -> $num
+_ws_to_varlen() {
+	if [[ $1 -lt 126 ]]; then
+		num=$(u8 "$1")
+	elif [[ $1 -ge 126 ]]; then # 2 bytes extended
+		num=7e$(u16 "$1")
+	elif [[ $1 -gt 65535 ]]; then # 8 bytes extended
+		num=7f$(u64 "$1")
+	fi
+}
 
-readarray -t data <<< $(echo -n $bytes | hexdump -e '/1 "%u\n"')
+# _ws_from_varlen(len) -> $len
+_ws_from_varlen() {
+	if [[ $1 -lt 126 ]]; then
+		len="$1"
+		return
+	elif [[ $1 == 126 ]]; then
+		bread 2
+	elif [[ $1 == 127 ]]; then
+		bread 8
+	else # what
+		return 1
+	fi
+	len="$((0x$bres))"
+}
 
-echo "data:" > /dev/stderr
-echo ${data[@]} | hexdump -C > /dev/stderr
+# ws_send(payload)
+ws_send() {
+	local num out len
+	len="${#1}"
+	[[ $len == 0 ]] && return
+	_ws_to_varlen $len
 
+	( echo -n "81$num" | xxd -p -r; echo -n "$1" )
+}
 
-echo "FIN bit:" $(( ${data[0]} >> 7 )) > /dev/stderr
+ws_recv() {
+	local flags fin opcode mask mask_ len bres
+	if [[ "$1" ]]; then
+		local -n ws_res=$1
+	fi
+	bread 2 || return 1
+	flags=$bres
 
-echo "opcode:" $(( ${data[0]} & 0x0f )) > /dev/stderr
+	fin=$(bit_slice $flags 15..16)
+	opcode=$(bit_slice $flags 8..12)
+	mask_=$(bit_slice $flags 7)
+	len=$(bit_slice $flags 0..7)
+	_ws_from_varlen "$len" # check for extended length
 
-echo ${data[0]} > /dev/stderr
+	if [[ $mask_ == 1 ]]; then
+		bread 4
+		# split into 4 separate bytes, separate them with dummy items at odd positions
+		# this is saving us a div/mul in the for loop below! :3
+		mask=( ${bres:0:2} . ${bres:2:2} . ${bres:4:2} . ${bres:6:2} )
+	fi
 
-mask_bit=$(( ${data[1]} >> 7 ))
-echo "Mask (bit):" $mask_bit > /dev/stderr
-
-len=$(( ${data[1]} & 0x7f ))
-offset=2 # 2 for header
-
-if [[ $len == 126 ]]; then
-	len=$(( ${data[2]} << 8 + ${data[3]} ))
-	offset=4 # 2 for header, 2 for extended length
-elif [[ $len == 127 ]]; then
-	len=0
-	for i in {0..8}; do
-		len=$(( $len << 8 + ${data[2+i]} ))
-	done
-	offset=10 # 2 for header, 8 for extended length
-fi
-
-echo "Data length:" $len > /dev/stderr
-
-echo "Offset:" $offset > /dev/stderr
-
-if [[ $mask_bit == 1 ]]; then
-	read -ra mask <<< ${data[@]:$offset:4}
-
-	echo "Mask:" $mask > /dev/stderr
-
-	offset=$(( $offset + 4 ))
-fi
-
-read -ra payload <<< ${data[@]:$offset:$len}
-
-echo "Payload:" ${payload[@]} > /dev/stderr
-
-if [[ $mask_bit == 1 ]]; then
-	for ((i = 0; i < $len; i++)); do
-		byte=$(( ${payload[$i]}  ^ ${mask[$i % 4]} ))
-		echo "Byte $i: $byte" > /dev/stderr
-		payload[$i]=$byte
-	done
-
-	echo "Payload unmasked:" ${payload[@]} > /dev/stderr
-fi
-
-payload_hex=""
-for ((i = 0; i < $len; i++)); do
-	hex=$(printf '\\x%x\n' $(( ${payload[$i]} )))
-	payload_hex+=$hex
-done
-
-echo "Payload hex:" $payload_hex > /dev/stderr
-echo -e "Payload: $payload_hex" > /dev/stderr
+	bread $len
+	if [[ $mask_ == 1 ]]; then
+		val=()
+		for (( i=0; i<(len*2); i=i+2 )); do
+			val+=($(( 0x${bres:i:2} ^ 0x${mask[i%8]} )))
+		done
+		ws_res="$(printf '%x' ${val[@]} | unhex)" # TODO: variant that doesn't unhex
+	else
+		ws_res="$(unhex <<< "$bres")"
+	fi
+}

From ec0c81f9f5aa6df2662515ceaed807dfe83e614c Mon Sep 17 00:00:00 2001
From: sdomi <ja@sdomi.pl>
Date: Wed, 26 Feb 2025 01:22:42 +0100
Subject: [PATCH 2/2] docs: websockets api documentation

---
 docs/websockets.md | 76 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 76 insertions(+)
 create mode 100644 docs/websockets.md

diff --git a/docs/websockets.md b/docs/websockets.md
new file mode 100644
index 0000000..cf751a8
--- /dev/null
+++ b/docs/websockets.md
@@ -0,0 +1,76 @@
+# HTTP.sh and WebSockets
+
+WebSocket support is currently experimental; It has been rewritten from scratch and verified, but
+it's a bit rough around the edges. Additionally, you have to manually obtain a copy of libsh
+before using websockets:
+
+    git clone https://git.sakamoto.pl/domi/libsh src/libsh
+
+Afterwards, you need to set `cfg[websocket_enable]` to `true` and optionally customize
+`cfg[extension_websocket]`. Both of those options can be found in `config/master.sh`.
+
+It is **required** that all of your websocket handling scripts have a different extension than
+your normal scripts. By default that's set to `shx` (opposed to default `shs` for normal scripts).
+The way of interfacing with WebSockets is fundamentally different enough that it warrants wrapper
+functions. While one could create scripts compatible with both, in most cases executing a script
+with the wrong interface is NOT what we want to happen.
+
+## The API
+
+`ws.sh` exposes two main functions: `ws_recv` and `ws_send`. An example use is presented below:
+
+    #!/bin/bash
+    while ws_recv; do
+    	ws_send "hiii! here's your message: $ws_res"
+    done
+
+### ws_recv
+
+`ws_recv` waits for a new message, parses and unmasks it, then puts it into `$ws_res`.
+This variable name can be overriden by doing `ws_recv varname`, which will then use `$varname` for
+the payload.
+
+The return codes are:
+- 0 for successful read
+- 1 if EOF is encountered (analogous to builtin `read`)
+
+### ws_send
+
+`ws_send <payload>` generates a header and sends the message out.
+
+## Note about async I/O
+
+Waiting for a message to arrive before sending out our data is not always useful. Thus, some use of
+asynchronous I/O is in order. Bash doesn't make this very easy, but it's not impossible:
+
+    #!/bin/bash
+    {
+    	while sleep 1; do
+    		ws_send "a"
+    	done
+    } &
+    
+    while ws_recv; do
+    	ws_send "$ws_res"
+    done
+    
+    pkill -P $$
+
+This spawns a subshell which sends the letter "a". Then, we create a receive loop where we echo
+everything back. Finally, and most importantly, when that receive loop quits, we kill all children
+of the current shell process, cleaning up the previous subshell. IF THIS ISN'T DONE, BAD THINGS
+HAPPEN TO YOUR SERVER - especially if your loop *doesn't* include a sleep like the example.
+
+With this style of async I/O, you have to take care of the IPC yourself. That is, no variables are
+shared after the subshell spawns. The best way to handle this is to fully decouple the input and
+output; If that's impossible, you may use a file approach (with or without notORM).
+
+## Quirks
+
+- Currently, anything manually sent to stdout will mangle the bitstream,
+  causing all subsequent I/O to fail.
+- Endless loops are dangerous if not cleaned up correctly at the end of the file!
+  Additionally, you **need** a receive loop (it can be a dummy, with `:`) if you want to continously
+  send and cleanup after the connection closes; Otherwise there's nothing indicating whether the
+  connection is still alive or not.
+- There's no way to transmit binary data w/o mangling it. TODO.