Fw: [Bugme-new] [Bug 7189] New: Inconsistent /proc/fb behavior

Fw: [Bugme-new] [Bug 7189] New: Inconsistent /proc/fb behavior

[Andrew Morton]

Begin forwarded message:

Date: Fri, 22 Sep 2006 19:27:02 -0700
From: bugme-daemon@bugzilla.kernel.org
To: bugme-new@lists.osdl.org
Subject: [Bugme-new] [Bug 7189] New: Inconsistent /proc/fb behavior


http://bugzilla.kernel.org/show_bug.cgi?id=7189

           Summary: Inconsistent /proc/fb behavior
    Kernel Version: 2.6.17.13
            Status: NEW
          Severity: normal
             Owner: jsimmons@infradead.org
         Submitter: jurij@wooyd.org


Distribution: Debian unstable
Hardware environment: Sparc Ultra60 workstation

Problem Description:
It appears that the function fbmem_read_proc, which serves as a backend for the
/proc/fb file, has a problem. When constructing the list of available frame
buffers to return to the user, it uses the following for-cycle;

for (fi = registered_fb; fi < &registered_fb[FB_MAX] && len < 4000; fi++)

Here len is the parameter passed to the function, that it the amount of data the
user is requesting from the file. So if the user requests a chunk larger than
4000 bytes, nothing is returned, leading to the peculiar behaviour described below.

Steps to reproduce:

$ cat /proc/fb
0 Creator 3D

This works fine, because strace shows that cat is reading data in 1024-byte chunks:
[..]
read(3, "0 Creator 3D\n", 1024)         = 13

OTOH, grep is reading data in 32kB chunks:
[..]
read(3, "", 32768)                      = 0

so the command 'grep Creator /proc/fb' returns nothing (quite unexpectingly). I
suspect that for-loop should have 'clen' rather than 'len', the local variable
which tracks the size of the buffer (even though I am not sure why one would
want to impose a 4000 byte limit here).

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: Fw: [Bugme-new] [Bug 7189] New: Inconsistent /proc/fb behavior

[Geert Uytterhoeven]

On Fri, 22 Sep 2006, Andrew Morton wrote:
> Begin forwarded message:
> 
> http://bugzilla.kernel.org/show_bug.cgi?id=7189
> 
>            Summary: Inconsistent /proc/fb behavior
>     Kernel Version: 2.6.17.13
>             Status: NEW
>           Severity: normal
>              Owner: jsimmons@infradead.org
>          Submitter: jurij@wooyd.org
> 
> 
> Distribution: Debian unstable
> Hardware environment: Sparc Ultra60 workstation
> 
> Problem Description:
> It appears that the function fbmem_read_proc, which serves as a backend for the
> /proc/fb file, has a problem. When constructing the list of available frame
> buffers to return to the user, it uses the following for-cycle;
> 
> for (fi = registered_fb; fi < &registered_fb[FB_MAX] && len < 4000; fi++)
> 
> Here len is the parameter passed to the function, that it the amount of data the
> user is requesting from the file. So if the user requests a chunk larger than
> 4000 bytes, nothing is returned, leading to the peculiar behaviour described below.
> 
> Steps to reproduce:
> 
> $ cat /proc/fb
> 0 Creator 3D
> 
> This works fine, because strace shows that cat is reading data in 1024-byte chunks:
> [..]
> read(3, "0 Creator 3D\n", 1024)         = 13
> 
> OTOH, grep is reading data in 32kB chunks:
> [..]
> read(3, "", 32768)                      = 0
> 
> so the command 'grep Creator /proc/fb' returns nothing (quite unexpectingly). I
> suspect that for-loop should have 'clen' rather than 'len', the local variable
> which tracks the size of the buffer (even though I am not sure why one would
> want to impose a 4000 byte limit here).

I guess it should check `clen', not `len'. Can you please try this
patch? It should apply to 2.4.34-pre3, too.

[PATCH] Correct buffer size limit in fbmem_read_proc()

Signed-Off-By: Geert Uytterhoeven <geert@linux-m68k.org>

--- linux-2.6.18/drivers/video/fbmem.c.orig	2006-09-04 11:02:45.000000000 +0200
+++ linux-2.6.18/drivers/video/fbmem.c	2006-09-24 12:17:07.000000000 +0200
@@ -554,7 +554,8 @@ static int fbmem_read_proc(char *buf, ch
 	int clen;
 
 	clen = 0;
-	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && len < 4000; fi++)
+	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && clen < 4000;
+	     fi++)
 		if (*fi)
 			clen += sprintf(buf + clen, "%d %s\n",
 				        (*fi)->node,

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: Fw: [Bugme-new] [Bug 7189] New: Inconsistent /proc/fb behavior

[Andrew Morton]

On Sun, 24 Sep 2006 12:31:30 +0200 (CEST)
Geert Uytterhoeven <geert@linux-m68k.org> wrote:

> On Fri, 22 Sep 2006, Andrew Morton wrote:
> > Begin forwarded message:
> > 
> > http://bugzilla.kernel.org/show_bug.cgi?id=7189
> > 
> >            Summary: Inconsistent /proc/fb behavior
> >     Kernel Version: 2.6.17.13
> >             Status: NEW
> >           Severity: normal
> >              Owner: jsimmons@infradead.org
> >          Submitter: jurij@wooyd.org
> > 
> > 
> > Distribution: Debian unstable
> > Hardware environment: Sparc Ultra60 workstation
> > 
> > Problem Description:
> > It appears that the function fbmem_read_proc, which serves as a backend for the
> > /proc/fb file, has a problem. When constructing the list of available frame
> > buffers to return to the user, it uses the following for-cycle;
> > 
> > for (fi = registered_fb; fi < &registered_fb[FB_MAX] && len < 4000; fi++)
> > 
> > Here len is the parameter passed to the function, that it the amount of data the
> > user is requesting from the file. So if the user requests a chunk larger than
> > 4000 bytes, nothing is returned, leading to the peculiar behaviour described below.
> > 
> > Steps to reproduce:
> > 
> > $ cat /proc/fb
> > 0 Creator 3D
> > 
> > This works fine, because strace shows that cat is reading data in 1024-byte chunks:
> > [..]
> > read(3, "0 Creator 3D\n", 1024)         = 13
> > 
> > OTOH, grep is reading data in 32kB chunks:
> > [..]
> > read(3, "", 32768)                      = 0
> > 
> > so the command 'grep Creator /proc/fb' returns nothing (quite unexpectingly). I
> > suspect that for-loop should have 'clen' rather than 'len', the local variable
> > which tracks the size of the buffer (even though I am not sure why one would
> > want to impose a 4000 byte limit here).
> 
> I guess it should check `clen', not `len'. Can you please try this
> patch? It should apply to 2.4.34-pre3, too.
> 
> [PATCH] Correct buffer size limit in fbmem_read_proc()
> 
> Signed-Off-By: Geert Uytterhoeven <geert@linux-m68k.org>
> 
> --- linux-2.6.18/drivers/video/fbmem.c.orig	2006-09-04 11:02:45.000000000 +0200
> +++ linux-2.6.18/drivers/video/fbmem.c	2006-09-24 12:17:07.000000000 +0200
> @@ -554,7 +554,8 @@ static int fbmem_read_proc(char *buf, ch
>  	int clen;
>  
>  	clen = 0;
> -	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && len < 4000; fi++)
> +	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && clen < 4000;
> +	     fi++)
>  		if (*fi)
>  			clen += sprintf(buf + clen, "%d %s\n",
>  				        (*fi)->node,
> 

Thanks.   Do we have confirmation that this patch fixes the bug?

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: Fw: [Bugme-new] [Bug 7189] New: Inconsistent /proc/fb behavior

[Willy Tarreau]

On Wed, Sep 27, 2006 at 12:39:30AM -0700, Jurij Smakov wrote:
> On Tue, Sep 26, 2006 at 01:58:09PM -0700, Andrew Morton wrote:
> > On Sun, 24 Sep 2006 12:31:30 +0200 (CEST)
> > Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> [...]
> > > [PATCH] Correct buffer size limit in fbmem_read_proc()
> > > 
> > > Signed-Off-By: Geert Uytterhoeven <geert@linux-m68k.org>
> > > 
> > > --- linux-2.6.18/drivers/video/fbmem.c.orig	2006-09-04 11:02:45.000000000 +0200
> > > +++ linux-2.6.18/drivers/video/fbmem.c	2006-09-24 12:17:07.000000000 +0200
> > > @@ -554,7 +554,8 @@ static int fbmem_read_proc(char *buf, ch
> > >  	int clen;
> > >  
> > >  	clen = 0;
> > > -	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && len < 4000; fi++)
> > > +	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && clen < 4000;
> > > +	     fi++)
> > >  		if (*fi)
> > >  			clen += sprintf(buf + clen, "%d %s\n",
> > >  				        (*fi)->node,
> > > 
> > 
> > Thanks.   Do we have confirmation that this patch fixes the bug?
> 
> Sorry for not replying sooner. Today I've built a 2.6.18 kernel with this patch 
> and it did fix the problem. Grepping through /proc/fb now works as expected on my 
> Sparc.

Thanks for the feedback, then I'm queueing it for 2.4.34-pre4 too.

Cheers,
Willy


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: Fw: [Bugme-new] [Bug 7189] New: Inconsistent /proc/fb behavior

[Jurij Smakov]

On Tue, Sep 26, 2006 at 01:58:09PM -0700, Andrew Morton wrote:
> On Sun, 24 Sep 2006 12:31:30 +0200 (CEST)
> Geert Uytterhoeven <geert@linux-m68k.org> wrote:
[...]
> > [PATCH] Correct buffer size limit in fbmem_read_proc()
> > 
> > Signed-Off-By: Geert Uytterhoeven <geert@linux-m68k.org>
> > 
> > --- linux-2.6.18/drivers/video/fbmem.c.orig	2006-09-04 11:02:45.000000000 +0200
> > +++ linux-2.6.18/drivers/video/fbmem.c	2006-09-24 12:17:07.000000000 +0200
> > @@ -554,7 +554,8 @@ static int fbmem_read_proc(char *buf, ch
> >  	int clen;
> >  
> >  	clen = 0;
> > -	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && len < 4000; fi++)
> > +	for (fi = registered_fb; fi < &registered_fb[FB_MAX] && clen < 4000;
> > +	     fi++)
> >  		if (*fi)
> >  			clen += sprintf(buf + clen, "%d %s\n",
> >  				        (*fi)->node,
> > 
> 
> Thanks.   Do we have confirmation that this patch fixes the bug?

Sorry for not replying sooner. Today I've built a 2.6.18 kernel with this patch 
and it did fix the problem. Grepping through /proc/fb now works as expected on my 
Sparc.

Best regards,
-- 
Jurij Smakov                                        jurij@wooyd.org
Key: http://www.wooyd.org/pgpkey/                   KeyID: C99E03CC

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV