linux-kernel - Re: [PATCH v2 1/2] tty: Convert tty

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Thu, 3 Nov 2022 12:11:26 +0200 (EET)
From:   Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>
To:     Greg KH <gregkh@...uxfoundation.org>
cc:     linux-serial <linux-serial@...r.kernel.org>,
        Jiri Slaby <jirislaby@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 1/2] tty: Convert tty_buffer flags to bool

On Thu, 3 Nov 2022, Greg KH wrote:

> On Wed, Oct 19, 2022 at 01:55:03PM +0300, Ilpo Järvinen wrote:
> > The struct tty_buffer has flags which is only used for storing TTYB_NORMAL.
> > There is also a few quite confusing operations for checking the presense
> > of TTYB_NORMAL. Simplify things by converting flags to bool.
> > 
> > Despite the name remaining the same, the meaning of "flags" is altered
> > slightly by this change. Previously it referred to flags of the buffer
> > (only TTYB_NORMAL being used as a flag). After this change, flags tell
> > whether the buffer contains/should be allocated with flags array along
> > with character data array. It is much more suitable name that
> > TTYB_NORMAL was for this purpose, thus the name remains.
> > 
> > Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>
> > ---
> > 
> > v2:
> > - Make it more obvious why flags is not renamed (both in kerneldoc
> >   comment and commit message).
> > 
> >  drivers/tty/tty_buffer.c   | 28 ++++++++++++++--------------
> >  include/linux/tty_buffer.h |  5 +----
> >  include/linux/tty_flip.h   |  4 ++--
> >  3 files changed, 17 insertions(+), 20 deletions(-)
> > 
> > diff --git a/drivers/tty/tty_buffer.c b/drivers/tty/tty_buffer.c
> > index 5e287dedce01..b408d830fcbc 100644
> > --- a/drivers/tty/tty_buffer.c
> > +++ b/drivers/tty/tty_buffer.c
> > @@ -107,7 +107,7 @@ static void tty_buffer_reset(struct tty_buffer *p, size_t size)
> >  	p->commit = 0;
> >  	p->lookahead = 0;
> >  	p->read = 0;
> > -	p->flags = 0;
> > +	p->flags = true;
> >  }
> >  
> >  /**
> > @@ -249,7 +249,7 @@ void tty_buffer_flush(struct tty_struct *tty, struct tty_ldisc *ld)
> >   * __tty_buffer_request_room	-	grow tty buffer if needed
> >   * @port: tty port
> >   * @size: size desired
> > - * @flags: buffer flags if new buffer allocated (default = 0)
> > + * @flags: buffer has to store flags along character data
> >   *
> >   * Make at least @size bytes of linear space available for the tty buffer.
> >   *
> > @@ -260,19 +260,19 @@ void tty_buffer_flush(struct tty_struct *tty, struct tty_ldisc *ld)
> >   * Returns: the size we managed to find.
> >   */
> >  static int __tty_buffer_request_room(struct tty_port *port, size_t size,
> > -				     int flags)
> > +				     bool flags)
> >  {
> >  	struct tty_bufhead *buf = &port->buf;
> >  	struct tty_buffer *b, *n;
> >  	int left, change;
> >  
> >  	b = buf->tail;
> > -	if (b->flags & TTYB_NORMAL)
> > +	if (!b->flags)
> >  		left = 2 * b->size - b->used;
> >  	else
> >  		left = b->size - b->used;
> >  
> > -	change = (b->flags & TTYB_NORMAL) && (~flags & TTYB_NORMAL);
> > +	change = !b->flags && flags;
> >  	if (change || left < size) {
> >  		/* This is the slow path - looking for new buffers to use */
> >  		n = tty_buffer_alloc(port, size);
> > @@ -300,7 +300,7 @@ static int __tty_buffer_request_room(struct tty_port *port, size_t size,
> >  
> >  int tty_buffer_request_room(struct tty_port *port, size_t size)
> >  {
> > -	return __tty_buffer_request_room(port, size, 0);
> > +	return __tty_buffer_request_room(port, size, true);
> 
> Did this logic just get inverted?
>
> Maybe it's the jet-lag, but this feels like it's not correct anymore.

As you can see, the old way is sooo confusing :-). I'll admit I stumbled 
myself with this same default thing first. It's even more confusing than 
the other places.

This check is true when flag bytes are present / required to be present:
	(~flags & TTYB_NORMAL)
It's very very confusing way to check such condition due to layered 
reverse logic.

With old code, the per character flag bytes won't be there in the buffer 
if TTYB_NORMAL is present. Thus, the old default of 0 means 
__tty_buffer_request_room will allocate room for those flag bytes.

If you think about it carefully, the old code passed 0. Therefore, ~0 & 
TTYB_NORMAL is going to be true. After my change true is passed and true 
matches to the original code.

So the logic was not inverted. I just cleared those layered reverse logic 
traps the original had which makes my patch look it's inverting things.

I really appreciate you took your time to find out this little detail
from it! This is far from a simple change because of how trappy the old
way of doing things is.

> Maybe a commet up above where you calculate "left" would make more sense
> as to what is going on?

Do you mean you want me to add a comment there? I don't see any 
pre-existing comments that you could be pointing me to.


Should I resubmit it since you probably dropped the patch?


While doing this cleanup, I realized there would likely be room for 
some improvements which would avoid allocing a new tty_buffer. I'll 
probably look into those at some point.


-- 
 i.