All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jean Delvare <jdelvare@suse.de>
To: "Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>
Cc: Linux I2C <linux-i2c@vger.kernel.org>
Subject: Re: i2c-tools: i2cbusses: Avoid buffer overflows in sysfs paths
Date: Wed, 8 Nov 2017 09:57:52 +0100	[thread overview]
Message-ID: <20171108095752.300e95c6@endymion> (raw)
In-Reply-To: <20171106120426.lyo6azbrxxok4gan@pengutronix.de>

Hi Uwe,

Thanks for the review, very appreciated.

On Mon, 6 Nov 2017 13:04:26 +0100, Uwe Kleine-König wrote:
> Hello Jean,
> 
> On Tue, Oct 31, 2017 at 08:16:04AM +0100, Jean Delvare wrote:
> > sprintf isn't safe, use snprintf instead.
> > ---
> >  tools/i2cbusses.c |   10 +++++-----
> >  1 file changed, 5 insertions(+), 5 deletions(-)
> > 
> > --- a/tools/i2cbusses.c
> > +++ b/tools/i2cbusses.c
> > @@ -220,18 +220,18 @@ struct i2c_adap *gather_i2c_busses(void)
> >  
> >  		/* this should work for kernels 2.6.5 or higher and */
> >  		/* is preferred because is unambiguous */
> > -		sprintf(n, "%s/%s/name", sysfs, de->d_name);
> > +		snprintf(n, NAME_MAX, "%s/%s/name", sysfs, de->d_name);  
> 
> OK, now instead of running in a buffer overflow in sprintf you might
> call fopen with a partial (maybe unterminated?) filename. While this is
> definitively better, you should check the return value of snprintf to be
> completely safe here.

To be honest, I never thought the buffer overflows could ever happen,
my motivation to fix them was to allow the code to build in OBS, where
FORTIFY_SOURCE is enabled. So I went for the most simple change that
made gcc happy.

That being said, I have no problem additionally checking the value
returned by snprintf. Something like this?

From: Jean Delvare
Subject: i2cbusses: Check the return value of snprintf

It's very unlikely that these paths will ever be truncated, but
better safe than sorry.
---
 tools/i2cbusses.c |   34 ++++++++++++++++++++++++++++------
 1 file changed, 28 insertions(+), 6 deletions(-)

--- i2c-tools.orig/tools/i2cbusses.c	2017-11-02 16:17:50.698383029 +0100
+++ i2c-tools/tools/i2cbusses.c	2017-11-08 09:49:40.365339644 +0100
@@ -137,7 +137,7 @@ struct i2c_adap *gather_i2c_busses(void)
 	FILE *f;
 	char fstype[NAME_MAX], sysfs[NAME_MAX], n[NAME_MAX];
 	int foundsysfs = 0;
-	int count=0;
+	int len, count = 0;
 	struct i2c_adap *adapters;
 
 	adapters = calloc(BUNCH, sizeof(struct i2c_adap));
@@ -220,18 +220,32 @@ struct i2c_adap *gather_i2c_busses(void)
 
 		/* this should work for kernels 2.6.5 or higher and */
 		/* is preferred because is unambiguous */
-		snprintf(n, NAME_MAX, "%s/%s/name", sysfs, de->d_name);
+		len = snprintf(n, NAME_MAX, "%s/%s/name", sysfs, de->d_name);
+		if (len >= NAME_MAX) {
+			fprintf(stderr, "%s: path truncated\n", n);
+			continue;
+		}
 		f = fopen(n, "r");
 		/* this seems to work for ISA */
 		if(f == NULL) {
-			snprintf(n, NAME_MAX, "%s/%s/device/name", sysfs, de->d_name);
+			len = snprintf(n, NAME_MAX, "%s/%s/device/name", sysfs,
+				       de->d_name);
+			if (len >= NAME_MAX) {
+				fprintf(stderr, "%s: path truncated\n", n);
+				continue;
+			}
 			f = fopen(n, "r");
 		}
 		/* non-ISA is much harder */
 		/* and this won't find the correct bus name if a driver
 		   has more than one bus */
 		if(f == NULL) {
-			snprintf(n, NAME_MAX, "%s/%s/device", sysfs, de->d_name);
+			len = snprintf(n, NAME_MAX, "%s/%s/device", sysfs,
+				       de->d_name);
+			if (len >= NAME_MAX) {
+				fprintf(stderr, "%s: path truncated\n", n);
+				continue;
+			}
 			if(!(ddir = opendir(n)))
 				continue;
 			while ((dde = readdir(ddir)) != NULL) {
@@ -240,8 +254,16 @@ struct i2c_adap *gather_i2c_busses(void)
 				if (!strcmp(dde->d_name, ".."))
 					continue;
 				if ((!strncmp(dde->d_name, "i2c-", 4))) {
-					snprintf(n, NAME_MAX, "%s/%s/device/%s/name",
-						 sysfs, de->d_name, dde->d_name);
+					len = snprintf(n, NAME_MAX,
+						       "%s/%s/device/%s/name",
+						       sysfs, de->d_name,
+						       dde->d_name);
+					if (len >= NAME_MAX) {
+						fprintf(stderr,
+							"%s: path truncated\n",
+							n);
+						continue;
+					}
 					if((f = fopen(n, "r")))
 						goto found;
 				}


-- 
Jean Delvare
SUSE L3 Support

  reply	other threads:[~2017-11-08  8:57 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-10-31  7:16 i2c-tools: i2cbusses: Avoid buffer overflows in sysfs paths Jean Delvare
2017-11-06 12:04 ` Uwe Kleine-König
2017-11-08  8:57   ` Jean Delvare [this message]
2017-11-08  9:29     ` Uwe Kleine-König
2017-11-08 21:14       ` Jean Delvare

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20171108095752.300e95c6@endymion \
    --to=jdelvare@suse.de \
    --cc=linux-i2c@vger.kernel.org \
    --cc=u.kleine-koenig@pengutronix.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.