[PATCH] vt: detect and ignore OSC codes.

[PATCH] vt: detect and ignore OSC codes.

[Adam Borowski]

These can be used to send commands consisting of an arbitrary string to the
terminal, most often used to set a terminal's window title or to redefine
the colour palette.  Our console doesn't use OSC, unlike everything else,
which can lead to junk being displayed if a process sends such a code
unconditionally.

Not following Ecma-48, this commit recognizes 7-bit forms (ESC ] ... 0x07,
ESC ] .. ESC \) but not 8-bit (0x9D ... 0x9C).

Signed-off-by: Adam Borowski <kilobyte@angband.pl>
---
 drivers/tty/vt/vt.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/tty/vt/vt.c b/drivers/tty/vt/vt.c
index 61b1137..0377c52 100644
--- a/drivers/tty/vt/vt.c
+++ b/drivers/tty/vt/vt.c
@@ -1590,7 +1590,7 @@ static void restore_cur(struct vc_data *vc)
 
 enum { ESnormal, ESesc, ESsquare, ESgetpars, ESgotpars, ESfunckey,
 	EShash, ESsetG0, ESsetG1, ESpercent, ESignore, ESnonstd,
-	ESpalette };
+	ESpalette, ESosc };
 
 /* console_lock is held (except via vc_init()) */
 static void reset_terminal(struct vc_data *vc, int do_clear)
@@ -1650,11 +1650,15 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
 	 *  Control characters can be used in the _middle_
 	 *  of an escape sequence.
 	 */
+	if (vc->vc_state == ESosc && c>=8 && c<=13) /* ... except for OSC */
+		return;
 	switch (c) {
 	case 0:
 		return;
 	case 7:
-		if (vc->vc_bell_duration)
+		if (vc->vc_state == ESosc)
+			vc->vc_state = ESnormal;
+		else if (vc->vc_bell_duration)
 			kd_mksound(vc->vc_bell_pitch, vc->vc_bell_duration);
 		return;
 	case 8:
@@ -1765,7 +1769,9 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
 		} else if (c=='R') {   /* reset palette */
 			reset_palette(vc);
 			vc->vc_state = ESnormal;
-		} else
+		} else if (c>='0' && c<='9')
+			vc->vc_state = ESosc;
+		else
 			vc->vc_state = ESnormal;
 		return;
 	case ESpalette:
@@ -2023,6 +2029,8 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
 		return;
 	default:
 		vc->vc_state = ESnormal;
+	case ESosc:
+		return;
 	}
 }
 
-- 
1.8.5.2

Re: [PATCH] vt: detect and ignore OSC codes.

[Greg Kroah-Hartman]

On Wed, Jan 15, 2014 at 07:21:04AM +0100, Adam Borowski wrote:
> These can be used to send commands consisting of an arbitrary string to the
> terminal, most often used to set a terminal's window title or to redefine
> the colour palette.  Our console doesn't use OSC, unlike everything else,
> which can lead to junk being displayed if a process sends such a code
> unconditionally.
> 
> Not following Ecma-48, this commit recognizes 7-bit forms (ESC ] ... 0x07,
> ESC ] .. ESC \) but not 8-bit (0x9D ... 0x9C).
> 
> Signed-off-by: Adam Borowski <kilobyte@angband.pl>

Where is this documented?


> ---
>  drivers/tty/vt/vt.c | 14 +++++++++++---
>  1 file changed, 11 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/tty/vt/vt.c b/drivers/tty/vt/vt.c
> index 61b1137..0377c52 100644
> --- a/drivers/tty/vt/vt.c
> +++ b/drivers/tty/vt/vt.c
> @@ -1590,7 +1590,7 @@ static void restore_cur(struct vc_data *vc)
>  
>  enum { ESnormal, ESesc, ESsquare, ESgetpars, ESgotpars, ESfunckey,
>  	EShash, ESsetG0, ESsetG1, ESpercent, ESignore, ESnonstd,
> -	ESpalette };
> +	ESpalette, ESosc };
>  
>  /* console_lock is held (except via vc_init()) */
>  static void reset_terminal(struct vc_data *vc, int do_clear)
> @@ -1650,11 +1650,15 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
>  	 *  Control characters can be used in the _middle_
>  	 *  of an escape sequence.
>  	 */
> +	if (vc->vc_state == ESosc && c>=8 && c<=13) /* ... except for OSC */
> +		return;
>  	switch (c) {
>  	case 0:
>  		return;
>  	case 7:
> -		if (vc->vc_bell_duration)
> +		if (vc->vc_state == ESosc)
> +			vc->vc_state = ESnormal;
> +		else if (vc->vc_bell_duration)
>  			kd_mksound(vc->vc_bell_pitch, vc->vc_bell_duration);
>  		return;
>  	case 8:
> @@ -1765,7 +1769,9 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
>  		} else if (c=='R') {   /* reset palette */
>  			reset_palette(vc);
>  			vc->vc_state = ESnormal;
> -		} else
> +		} else if (c>='0' && c<='9')
> +			vc->vc_state = ESosc;
> +		else
>  			vc->vc_state = ESnormal;
>  		return;
>  	case ESpalette:
> @@ -2023,6 +2029,8 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
>  		return;
>  	default:
>  		vc->vc_state = ESnormal;
> +	case ESosc:
> +		return;

Why below the default: case?

thanks,

greg k-h

Re: [PATCH] vt: detect and ignore OSC codes.

[Adam Borowski]

On Thu, Feb 13, 2014 at 10:39:12AM -0800, Greg Kroah-Hartman wrote:
> On Wed, Jan 15, 2014 at 07:21:04AM +0100, Adam Borowski wrote:
> > These can be used to send commands consisting of an arbitrary string to the
> > terminal, most often used to set a terminal's window title or to redefine
> > the colour palette.  Our console doesn't use OSC, unlike everything else,
> > which can lead to junk being displayed if a process sends such a code
> > unconditionally.
> > 
> > Not following Ecma-48, this commit recognizes 7-bit forms (ESC ] ... 0x07,
> > ESC ] .. ESC \) but not 8-bit (0x9D ... 0x9C).
> > 
> > Signed-off-by: Adam Borowski <kilobyte@angband.pl>
> 
> Where is this documented?

It's a mix of Ecma-48 and undocumented practice.  Ecma-48 says:

# 8.3.89
#   OSC - OPERATING SYSTEM COMMAND
#   Notation: (C1)
#   Representation: 0x9D or ESC 0x5D (])
#
#   OSC is used as the opening delimiter of a control string for operating
#   system use.  The command string following may consist of a sequence of
#   bit combinations in the range 0x08 to 0x0D and 0x20 to 0x7E.  The
#   control string is closed by the terminating delimiter STRING TERMINATOR
#   (ST).  The interpretation of the command string depends on the relevant
#   operating system.

# 8.3.143
#   ST - STRING TERMINATOR
#   Notation: (C1)
#   Representation: 0x9C or ESC 0x5C (\)
#
#   ST is used as the closing delimiter of a control string opened by
#   APPLICATION PROGRAM COMMAND (APC), DEVICE CONTROL STRING (DCS),
#   OPERATING SYSTEM COMMAND (OSC), PRIVACY MESSAGE (PM), or START OF STRING
#   (SOS).

... which doesn't define the behaviour for characters 0x00..0x07, 0x0E..0x1F
or 0x7F..0xFF.  Somehow, using 0x07 for termination became a widespread
idiom, used more often than proper ESC \.  For this reason, implementations
I know all recognize 0x07 as a terminator.  The behaviour for other
characters differs, ie, is truly undefined.

As, unlike what Ecma-48 says, using 8-bit characters in a window title is a
reasonable thing to do, I'd allow 0x80..0xFF as non-terminators.  I have no
idea what to do with remaining control characters: the current patch allows
them to be interpreted, as it's usually the case for control characters
inside terminal codes.  I did not research other implementation here.

I did not recognize 0x9C as ST for two reasons: 1. it'd break non-ASCII
characters that happen to include this byte, and 2. Linux already fails to
recognize 8-bit control codes (with one exception: 0x9B stands for ESC [).


Should I put the above explanation somewhere?  As a comment?  In the commit
message?  Or does it need to be elaborated even further?

> > ---
> >  drivers/tty/vt/vt.c | 14 +++++++++++---
> >  1 file changed, 11 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/tty/vt/vt.c b/drivers/tty/vt/vt.c
> > index 61b1137..0377c52 100644
> > --- a/drivers/tty/vt/vt.c
> > +++ b/drivers/tty/vt/vt.c
> > @@ -1590,7 +1590,7 @@ static void restore_cur(struct vc_data *vc)
> >  
> >  enum { ESnormal, ESesc, ESsquare, ESgetpars, ESgotpars, ESfunckey,
> >  	EShash, ESsetG0, ESsetG1, ESpercent, ESignore, ESnonstd,
> > -	ESpalette };
> > +	ESpalette, ESosc };
> >  
> >  /* console_lock is held (except via vc_init()) */
> >  static void reset_terminal(struct vc_data *vc, int do_clear)
> > @@ -1650,11 +1650,15 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
> >  	 *  Control characters can be used in the _middle_
> >  	 *  of an escape sequence.
> >  	 */
> > +	if (vc->vc_state == ESosc && c>=8 && c<=13) /* ... except for OSC */
> > +		return;
> >  	switch (c) {
> >  	case 0:
> >  		return;
> >  	case 7:
> > -		if (vc->vc_bell_duration)
> > +		if (vc->vc_state == ESosc)
> > +			vc->vc_state = ESnormal;
> > +		else if (vc->vc_bell_duration)
> >  			kd_mksound(vc->vc_bell_pitch, vc->vc_bell_duration);
> >  		return;
> >  	case 8:
> > @@ -1765,7 +1769,9 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
> >  		} else if (c=='R') {   /* reset palette */
> >  			reset_palette(vc);
> >  			vc->vc_state = ESnormal;
> > -		} else
> > +		} else if (c>='0' && c<='9')
> > +			vc->vc_state = ESosc;
> > +		else
> >  			vc->vc_state = ESnormal;
> >  		return;
> >  	case ESpalette:
> > @@ -2023,6 +2029,8 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
> >  		return;
> >  	default:
> >  		vc->vc_state = ESnormal;
> > +	case ESosc:
> > +		return;
> 
> Why below the default: case?

Just to shave a line and a return statement.  From your objection, I guess
this goes against the coding standards, right?


Meow!
-- 
A tit a day keeps the vet away.

Re: [PATCH] vt: detect and ignore OSC codes.

[Greg Kroah-Hartman]

On Sun, Feb 16, 2014 at 03:37:59AM +0100, Adam Borowski wrote:
> On Thu, Feb 13, 2014 at 10:39:12AM -0800, Greg Kroah-Hartman wrote:
> > On Wed, Jan 15, 2014 at 07:21:04AM +0100, Adam Borowski wrote:
> > > These can be used to send commands consisting of an arbitrary string to the
> > > terminal, most often used to set a terminal's window title or to redefine
> > > the colour palette.  Our console doesn't use OSC, unlike everything else,
> > > which can lead to junk being displayed if a process sends such a code
> > > unconditionally.
> > > 
> > > Not following Ecma-48, this commit recognizes 7-bit forms (ESC ] ... 0x07,
> > > ESC ] .. ESC \) but not 8-bit (0x9D ... 0x9C).
> > > 
> > > Signed-off-by: Adam Borowski <kilobyte@angband.pl>
> > 
> > Where is this documented?
> 
> It's a mix of Ecma-48 and undocumented practice.  Ecma-48 says:
> 
> # 8.3.89
> #   OSC - OPERATING SYSTEM COMMAND
> #   Notation: (C1)
> #   Representation: 0x9D or ESC 0x5D (])
> #
> #   OSC is used as the opening delimiter of a control string for operating
> #   system use.  The command string following may consist of a sequence of
> #   bit combinations in the range 0x08 to 0x0D and 0x20 to 0x7E.  The
> #   control string is closed by the terminating delimiter STRING TERMINATOR
> #   (ST).  The interpretation of the command string depends on the relevant
> #   operating system.
> 
> # 8.3.143
> #   ST - STRING TERMINATOR
> #   Notation: (C1)
> #   Representation: 0x9C or ESC 0x5C (\)
> #
> #   ST is used as the closing delimiter of a control string opened by
> #   APPLICATION PROGRAM COMMAND (APC), DEVICE CONTROL STRING (DCS),
> #   OPERATING SYSTEM COMMAND (OSC), PRIVACY MESSAGE (PM), or START OF STRING
> #   (SOS).
> 
> ... which doesn't define the behaviour for characters 0x00..0x07, 0x0E..0x1F
> or 0x7F..0xFF.  Somehow, using 0x07 for termination became a widespread
> idiom, used more often than proper ESC \.  For this reason, implementations
> I know all recognize 0x07 as a terminator.  The behaviour for other
> characters differs, ie, is truly undefined.
> 
> As, unlike what Ecma-48 says, using 8-bit characters in a window title is a
> reasonable thing to do, I'd allow 0x80..0xFF as non-terminators.  I have no
> idea what to do with remaining control characters: the current patch allows
> them to be interpreted, as it's usually the case for control characters
> inside terminal codes.  I did not research other implementation here.
> 
> I did not recognize 0x9C as ST for two reasons: 1. it'd break non-ASCII
> characters that happen to include this byte, and 2. Linux already fails to
> recognize 8-bit control codes (with one exception: 0x9B stands for ESC [).
> 
> 
> Should I put the above explanation somewhere?  As a comment?  In the commit
> message?  Or does it need to be elaborated even further?

In the commit message would be best.

> > > @@ -2023,6 +2029,8 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
> > >  		return;
> > >  	default:
> > >  		vc->vc_state = ESnormal;
> > > +	case ESosc:
> > > +		return;
> > 
> > Why below the default: case?
> 
> Just to shave a line and a return statement.  From your objection, I guess
> this goes against the coding standards, right?

Yes it does.

[PATCH] vt: detect and ignore OSC codes.

[Adam Borowski]

These can be used to send commands consisting of an arbitrary string to the
terminal, most often used to set a terminal's window title or to redefine
the colour palette.  Our console doesn't use OSC, unlike everything else,
which can lead to junk being displayed if a process sends such a code
unconditionally.

The rules for termination follow established practice rather than Ecma-48.
Ecma-48 requires the string to use only byte values 0x08..0x0D and
0x20..0x7E, terminated with either ESC \ or 0x9C.  This would disallow using
8-bit characters, which are reasonable for example in window titles.
A widespread idiom is to terminate with 0x07.  The behaviour for other
control characters differs between terminal emulators, I followed libvte and
xterm:
* 0x07 and ESC anything terminate
* nothing else terminates, all 8-bit values including 0x9C are considered a
  part of the string

Signed-off-by: Adam Borowski <kilobyte@angband.pl>
---
 drivers/tty/vt/vt.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/tty/vt/vt.c b/drivers/tty/vt/vt.c
index b7bde54..3ad0b61 100644
--- a/drivers/tty/vt/vt.c
+++ b/drivers/tty/vt/vt.c
@@ -1592,7 +1592,7 @@ static void restore_cur(struct vc_data *vc)
 
 enum { ESnormal, ESesc, ESsquare, ESgetpars, ESfunckey,
 	EShash, ESsetG0, ESsetG1, ESpercent, ESignore, ESnonstd,
-	ESpalette };
+	ESpalette, ESosc };
 
 /* console_lock is held (except via vc_init()) */
 static void reset_terminal(struct vc_data *vc, int do_clear)
@@ -1652,11 +1652,15 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
 	 *  Control characters can be used in the _middle_
 	 *  of an escape sequence.
 	 */
+	if (vc->vc_state == ESosc && c>=8 && c<=13) /* ... except for OSC */
+		return;
 	switch (c) {
 	case 0:
 		return;
 	case 7:
-		if (vc->vc_bell_duration)
+		if (vc->vc_state == ESosc)
+			vc->vc_state = ESnormal;
+		else if (vc->vc_bell_duration)
 			kd_mksound(vc->vc_bell_pitch, vc->vc_bell_duration);
 		return;
 	case 8:
@@ -1767,7 +1771,9 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
 		} else if (c=='R') {   /* reset palette */
 			reset_palette(vc);
 			vc->vc_state = ESnormal;
-		} else
+		} else if (c>='0' && c<='9')
+			vc->vc_state = ESosc;
+		else
 			vc->vc_state = ESnormal;
 		return;
 	case ESpalette:
@@ -2021,6 +2027,8 @@ static void do_con_trol(struct tty_struct *tty, struct vc_data *vc, int c)
 			vc->vc_translate = set_translate(vc->vc_G1_charset, vc);
 		vc->vc_state = ESnormal;
 		return;
+	case ESosc:
+		return;
 	default:
 		vc->vc_state = ESnormal;
 	}
-- 
1.9.0