[PATCH 0/5] kdb: Cleanup code to read user input and handle escape sequences

[PATCH 0/5] kdb: Cleanup code to read user input and handle escape sequences

[Daniel Thompson]

I spend a little time earlier this month looking at kdb_read() and ended
up doing a little light refactoring a bit to make easier to read.

Most of the patches are simple tidy ups although patches 4 and 5
introduce new behaviours. Patch 4 is probably not controversial
but perhaps patch 5 is ;-).

Tested as shown here:
https://github.com/daniel-thompson/kgdbtest/commit/c65e28d99357c2df6dac2cebe195574e634d04dc

Daniel Thompson (5):
  kdb: Tidy up code to handle escape sequences
  kdb: Simplify code to fetch characters from console
  kdb: Remove special case logic from kdb_read()
  kdb: Improve handling of characters from different input sources
  kdb: Tweak escape handling for vi users

 kernel/debug/kdb/kdb_io.c | 218 +++++++++++++++++---------------------
 1 file changed, 99 insertions(+), 119 deletions(-)

--
2.17.1

[PATCH 1/5] kdb: Tidy up code to handle escape sequences

[Daniel Thompson]

kdb_read_get_key() has extremely complex break/continue control flow
managed by state variables and is very hard to review or modify. In
particular the way the escape sequence handling interacts with the
general control flow is hard to follow. Separate out the escape key
handling, without changing the control flow. This makes the main body of
the code easier to review.

Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
---
 kernel/debug/kdb/kdb_io.c | 127 ++++++++++++++++++++------------------
 1 file changed, 66 insertions(+), 61 deletions(-)

diff --git a/kernel/debug/kdb/kdb_io.c b/kernel/debug/kdb/kdb_io.c
index ed5d34925ad0..e54e1e185c47 100644
--- a/kernel/debug/kdb/kdb_io.c
+++ b/kernel/debug/kdb/kdb_io.c
@@ -49,6 +49,63 @@ static int kgdb_transition_check(char *buffer)
 	return 0;
 }
 
+/*
+ * kdb_read_handle_escape
+ *
+ * Run a validity check on an accumulated escape sequence.
+ *
+ * Returns -1 if the escape sequence is unwanted, 0 if it is incomplete,
+ * otherwise it returns a mapped key value to pass to the upper layers.
+ */
+static int kdb_read_handle_escape(char *buf, size_t sz)
+{
+	char *lastkey = buf + sz - 1;
+
+	switch (sz) {
+	case 1:
+		if (*lastkey == '\e')
+			return 0;
+		break;
+
+	case 2: /* \e<something> */
+		if (*lastkey == '[')
+			return 0;
+		break;
+
+	case 3:
+		switch (*lastkey) {
+		case 'A': /* \e[A, up arrow */
+			return 16;
+		case 'B': /* \e[B, down arrow */
+			return 14;
+		case 'C': /* \e[C, right arrow */
+			return 6;
+		case 'D': /* \e[D, left arrow */
+			return 2;
+		case '1': /* \e[<1,3,4>], may be home, del, end */
+		case '3':
+		case '4':
+			return 0;
+		}
+		break;
+
+	case 4:
+		if (*lastkey == '~') {
+			switch (buf[2]) {
+			case '1': /* \e[1~, home */
+				return 1;
+			case '3': /* \e[3~, del */
+				return 4;
+			case '4': /* \e[4~, end */
+				return 5;
+			}
+		}
+		break;
+	}
+
+	return -1;
+}
+
 static int kdb_read_get_key(char *buffer, size_t bufsize)
 {
 #define ESCAPE_UDELAY 1000
@@ -102,68 +159,16 @@ static int kdb_read_get_key(char *buffer, size_t bufsize)
 				escape_delay = 2;
 				continue;
 			}
-			if (ped - escape_data == 1) {
-				/* \e */
-				continue;
-			} else if (ped - escape_data == 2) {
-				/* \e<something> */
-				if (key != '[')
-					escape_delay = 2;
-				continue;
-			} else if (ped - escape_data == 3) {
-				/* \e[<something> */
-				int mapkey = 0;
-				switch (key) {
-				case 'A': /* \e[A, up arrow */
-					mapkey = 16;
-					break;
-				case 'B': /* \e[B, down arrow */
-					mapkey = 14;
-					break;
-				case 'C': /* \e[C, right arrow */
-					mapkey = 6;
-					break;
-				case 'D': /* \e[D, left arrow */
-					mapkey = 2;
-					break;
-				case '1': /* dropthrough */
-				case '3': /* dropthrough */
-				/* \e[<1,3,4>], may be home, del, end */
-				case '4':
-					mapkey = -1;
-					break;
-				}
-				if (mapkey != -1) {
-					if (mapkey > 0) {
-						escape_data[0] = mapkey;
-						escape_data[1] = '\0';
-					}
-					escape_delay = 2;
-				}
-				continue;
-			} else if (ped - escape_data == 4) {
-				/* \e[<1,3,4><something> */
-				int mapkey = 0;
-				if (key == '~') {
-					switch (escape_data[2]) {
-					case '1': /* \e[1~, home */
-						mapkey = 1;
-						break;
-					case '3': /* \e[3~, del */
-						mapkey = 4;
-						break;
-					case '4': /* \e[4~, end */
-						mapkey = 5;
-						break;
-					}
-				}
-				if (mapkey > 0) {
-					escape_data[0] = mapkey;
-					escape_data[1] = '\0';
-				}
-				escape_delay = 2;
-				continue;
+
+			key = kdb_read_handle_escape(escape_data,
+						     ped - escape_data);
+			if (key > 0) {
+				escape_data[0] = key;
+				escape_data[1] = '\0';
 			}
+			if (key)
+				escape_delay = 2;
+			continue;
 		}
 		break;	/* A key to process */
 	}
-- 
2.17.1

[PATCH 2/5] kdb: Simplify code to fetch characters from console

[Daniel Thompson]

Currently kdb_read_get_key() contains complex control flow that, on
close inspection, turns out to be unnecessary. In particular:

1. It is impossible to enter the branch conditioned on (escape_delay == 1)
   except when the loop enters with (escape_delay == 2) allowing us to
   combine the branches.

2. Most of the code conditioned on (escape_delay == 2) simply modifies
   local data and then breaks out of the loop causing the function to
   return escape_data[0].

3. Based on #2 there is not actually any need to ever explicitly set
   escape_delay to 2 because we it is much simpler to directly return
   escape_data[0] instead.

4. escape_data[0] is, for all but one exit path, known to be '\e'.

Simplify the code based on these observations.

There is a subtle (and harmless) change of behaviour resulting from this
simplification: instead of letting the escape timeout after ~1998
milliseconds we now timeout after ~2000 milliseconds

Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
---
 kernel/debug/kdb/kdb_io.c | 38 ++++++++++++++------------------------
 1 file changed, 14 insertions(+), 24 deletions(-)

diff --git a/kernel/debug/kdb/kdb_io.c b/kernel/debug/kdb/kdb_io.c
index e54e1e185c47..0ba490728459 100644
--- a/kernel/debug/kdb/kdb_io.c
+++ b/kernel/debug/kdb/kdb_io.c
@@ -122,25 +122,18 @@ static int kdb_read_get_key(char *buffer, size_t bufsize)
 			touch_nmi_watchdog();
 			f = &kdb_poll_funcs[0];
 		}
-		if (escape_delay == 2) {
-			*ped = '\0';
-			ped = escape_data;
-			--escape_delay;
-		}
-		if (escape_delay == 1) {
-			key = *ped++;
-			if (!*ped)
-				--escape_delay;
-			break;
-		}
+
 		key = (*f)();
+
 		if (key == -1) {
 			if (escape_delay) {
 				udelay(ESCAPE_UDELAY);
-				--escape_delay;
+				if (--escape_delay == 0)
+					return '\e';
 			}
 			continue;
 		}
+
 		if (bufsize <= 2) {
 			if (key == '\r')
 				key = '\n';
@@ -148,28 +141,25 @@ static int kdb_read_get_key(char *buffer, size_t bufsize)
 			*buffer = '\0';
 			return -1;
 		}
+
 		if (escape_delay == 0 && key == '\e') {
 			escape_delay = ESCAPE_DELAY;
 			ped = escape_data;
 			f_escape = f;
 		}
 		if (escape_delay) {
-			*ped++ = key;
-			if (f_escape != f) {
-				escape_delay = 2;
-				continue;
-			}
+			if (f_escape != f)
+				return '\e';
 
+			*ped++ = key;
 			key = kdb_read_handle_escape(escape_data,
 						     ped - escape_data);
-			if (key > 0) {
-				escape_data[0] = key;
-				escape_data[1] = '\0';
-			}
-			if (key)
-				escape_delay = 2;
-			continue;
+			if (key < 0)
+				return '\e';
+			if (key == 0)
+				continue;
 		}
+
 		break;	/* A key to process */
 	}
 	return key;
-- 
2.17.1

[PATCH 3/5] kdb: Remove special case logic from kdb_read()

[Daniel Thompson]

kdb_read() contains special case logic to force it exit after reading
a single character. We can remove all the special case logic by directly
calling the function to read a single character instead. This also
allows us to tidy up the function prototype which, because it now matches
getchar(), we can also rename in order to make its role clearer.

Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
---
 kernel/debug/kdb/kdb_io.c | 56 ++++++++++++++++-----------------------
 1 file changed, 23 insertions(+), 33 deletions(-)

diff --git a/kernel/debug/kdb/kdb_io.c b/kernel/debug/kdb/kdb_io.c
index 0ba490728459..f735d859f28b 100644
--- a/kernel/debug/kdb/kdb_io.c
+++ b/kernel/debug/kdb/kdb_io.c
@@ -106,7 +106,19 @@ static int kdb_read_handle_escape(char *buf, size_t sz)
 	return -1;
 }
 
-static int kdb_read_get_key(char *buffer, size_t bufsize)
+/*
+ * kdb_getchar
+ *
+ * Read a single character from kdb console (or consoles).
+ *
+ * An escape key could be the start of a vt100 control sequence such as \e[D
+ * (left arrow) or it could be a character in its own right.  The standard
+ * method for detecting the difference is to wait for 2 seconds to see if there
+ * are any other characters.  kdb is complicated by the lack of a timer service
+ * (interrupts are off), by multiple input sources. Escape sequence processing
+ * has to be done as states in the polling loop.
+ */
+static int kdb_getchar(void)
 {
 #define ESCAPE_UDELAY 1000
 #define ESCAPE_DELAY (2*1000000/ESCAPE_UDELAY) /* 2 seconds worth of udelays */
@@ -124,7 +136,6 @@ static int kdb_read_get_key(char *buffer, size_t bufsize)
 		}
 
 		key = (*f)();
-
 		if (key == -1) {
 			if (escape_delay) {
 				udelay(ESCAPE_UDELAY);
@@ -134,14 +145,6 @@ static int kdb_read_get_key(char *buffer, size_t bufsize)
 			continue;
 		}
 
-		if (bufsize <= 2) {
-			if (key == '\r')
-				key = '\n';
-			*buffer++ = key;
-			*buffer = '\0';
-			return -1;
-		}
-
 		if (escape_delay == 0 && key == '\e') {
 			escape_delay = ESCAPE_DELAY;
 			ped = escape_data;
@@ -183,17 +186,7 @@ static int kdb_read_get_key(char *buffer, size_t bufsize)
  *	function.  It is not reentrant - it relies on the fact
  *	that while kdb is running on only one "master debug" cpu.
  * Remarks:
- *
- * The buffer size must be >= 2.  A buffer size of 2 means that the caller only
- * wants a single key.
- *
- * An escape key could be the start of a vt100 control sequence such as \e[D
- * (left arrow) or it could be a character in its own right.  The standard
- * method for detecting the difference is to wait for 2 seconds to see if there
- * are any other characters.  kdb is complicated by the lack of a timer service
- * (interrupts are off), by multiple input sources and by the need to sometimes
- * return after just one key.  Escape sequence processing has to be done as
- * states in the polling loop.
+ *	The buffer size must be >= 2.
  */
 
 static char *kdb_read(char *buffer, size_t bufsize)
@@ -228,9 +221,7 @@ static char *kdb_read(char *buffer, size_t bufsize)
 	*cp = '\0';
 	kdb_printf("%s", buffer);
 poll_again:
-	key = kdb_read_get_key(buffer, bufsize);
-	if (key == -1)
-		return buffer;
+	key = kdb_getchar();
 	if (key != 9)
 		tab = 0;
 	switch (key) {
@@ -738,7 +729,7 @@ int vkdb_printf(enum kdb_msgsrc src, const char *fmt, va_list ap)
 
 	/* check for having reached the LINES number of printed lines */
 	if (kdb_nextline >= linecount) {
-		char buf1[16] = "";
+		char ch;
 
 		/* Watch out for recursion here.  Any routine that calls
 		 * kdb_printf will come back through here.  And kdb_read
@@ -773,39 +764,38 @@ int vkdb_printf(enum kdb_msgsrc src, const char *fmt, va_list ap)
 		if (logging)
 			printk("%s", moreprompt);
 
-		kdb_read(buf1, 2); /* '2' indicates to return
-				    * immediately after getting one key. */
+		ch = kdb_getchar();
 		kdb_nextline = 1;	/* Really set output line 1 */
 
 		/* empty and reset the buffer: */
 		kdb_buffer[0] = '\0';
 		next_avail = kdb_buffer;
 		size_avail = sizeof(kdb_buffer);
-		if ((buf1[0] == 'q') || (buf1[0] == 'Q')) {
+		if ((ch == 'q') || (ch == 'Q')) {
 			/* user hit q or Q */
 			KDB_FLAG_SET(CMD_INTERRUPT); /* command interrupted */
 			KDB_STATE_CLEAR(PAGER);
 			/* end of command output; back to normal mode */
 			kdb_grepping_flag = 0;
 			kdb_printf("\n");
-		} else if (buf1[0] == ' ') {
+		} else if (ch == ' ') {
 			kdb_printf("\r");
 			suspend_grep = 1; /* for this recursion */
-		} else if (buf1[0] == '\n') {
+		} else if (ch == '\n' || ch == '\r') {
 			kdb_nextline = linecount - 1;
 			kdb_printf("\r");
 			suspend_grep = 1; /* for this recursion */
-		} else if (buf1[0] == '/' && !kdb_grepping_flag) {
+		} else if (ch == '/' && !kdb_grepping_flag) {
 			kdb_printf("\r");
 			kdb_getstr(kdb_grep_string, KDB_GREP_STRLEN,
 				   kdbgetenv("SEARCHPROMPT") ?: "search> ");
 			*strchrnul(kdb_grep_string, '\n') = '\0';
 			kdb_grepping_flag += KDB_GREPPING_FLAG_SEARCH;
 			suspend_grep = 1; /* for this recursion */
-		} else if (buf1[0] && buf1[0] != '\n') {
+		} else if (ch && ch != '\n') {
 			/* user hit something other than enter */
 			suspend_grep = 1; /* for this recursion */
-			if (buf1[0] != '/')
+			if (ch != '/')
 				kdb_printf(
 				    "\nOnly 'q', 'Q' or '/' are processed at "
 				    "more prompt, input ignored\n");
-- 
2.17.1

[PATCH 4/5] kdb: Improve handling of characters from different input sources

[Daniel Thompson]

Currently if an escape timer is interrupted by a character from a
different input source then the new character is discarded and the
function returns '\e' (which will be discarded by the level above).
It is hard to see why this would ever be the desired behaviour.
Fix this to return the new character rather then the '\e'.

This is a bigger refactor that might be expected because the new
character needs to go through escape sequence detection.

Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
---
 kernel/debug/kdb/kdb_io.c | 33 ++++++++++++++-------------------
 1 file changed, 14 insertions(+), 19 deletions(-)

diff --git a/kernel/debug/kdb/kdb_io.c b/kernel/debug/kdb/kdb_io.c
index f735d859f28b..8f2470016250 100644
--- a/kernel/debug/kdb/kdb_io.c
+++ b/kernel/debug/kdb/kdb_io.c
@@ -122,8 +122,8 @@ static int kdb_getchar(void)
 {
 #define ESCAPE_UDELAY 1000
 #define ESCAPE_DELAY (2*1000000/ESCAPE_UDELAY) /* 2 seconds worth of udelays */
-	char escape_data[5];	/* longest vt100 escape sequence is 4 bytes */
-	char *ped = escape_data;
+	char buf[4];	/* longest vt100 escape sequence is 4 bytes */
+	char *pbuf = buf;
 	int escape_delay = 0;
 	get_char_func *f, *f_escape = NULL;
 	int key;
@@ -145,27 +145,22 @@ static int kdb_getchar(void)
 			continue;
 		}
 
-		if (escape_delay == 0 && key == '\e') {
-			escape_delay = ESCAPE_DELAY;
-			ped = escape_data;
+		/* Reset state on first character from an input source */
+		if (f_escape != f) {
 			f_escape = f;
-		}
-		if (escape_delay) {
-			if (f_escape != f)
-				return '\e';
-
-			*ped++ = key;
-			key = kdb_read_handle_escape(escape_data,
-						     ped - escape_data);
-			if (key < 0)
-				return '\e';
-			if (key == 0)
-				continue;
+			pbuf = buf;
+			escape_delay = ESCAPE_DELAY;
 		}
 
-		break;	/* A key to process */
+		*pbuf++ = key;
+		key = kdb_read_handle_escape(buf, pbuf - buf);
+		if (key < 0) /* no escape sequence; return first character */
+			return buf[0];
+		if (key > 0)
+			return key;
 	}
-	return key;
+
+	unreachable();
 }
 
 /*
-- 
2.17.1

[PATCH 5/5] kdb: Tweak escape handling for vi users

[Daniel Thompson]

Currently if sequences such as "\ehelp\r" are delivered to the console then
the h gets eaten by the escape handling code. Since pressing escape
becomes something of a nervous twitch for vi users (and that escape doesn't
have much effect at a shell prompt) it is more helpful to emit the 'h' than
the '\e'.

We don't simply choose to emit the final character for all escape sequences
since that will do odd things for unsupported escape sequences (in
other words we retain the existing behaviour once we see '\e[').

Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
---
 kernel/debug/kdb/kdb_io.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/debug/kdb/kdb_io.c b/kernel/debug/kdb/kdb_io.c
index 8f2470016250..6ee8f812d656 100644
--- a/kernel/debug/kdb/kdb_io.c
+++ b/kernel/debug/kdb/kdb_io.c
@@ -154,8 +154,8 @@ static int kdb_getchar(void)
 
 		*pbuf++ = key;
 		key = kdb_read_handle_escape(buf, pbuf - buf);
-		if (key < 0) /* no escape sequence; return first character */
-			return buf[0];
+		if (key < 0) /* no escape sequence; return best character */
+			return buf[pbuf - buf != 2 ? 0 : 1];
 		if (key > 0)
 			return key;
 	}
-- 
2.17.1