2009-03-01 20:00:31

by Frank Seidel

[permalink] [raw]
Subject: [PATCH] lkdtm: reduce stack size

From: Frank Seidel <[email protected]>

Reduce stack memory footprint of lkdtm
(recursive_loop). From 1024 bytes on i386 down
to <10.

Signed-off-by: Frank Seidel <[email protected]>
---
drivers/misc/lkdtm.c | 16 +++++++++++++---
1 file changed, 13 insertions(+), 3 deletions(-)

--- a/drivers/misc/lkdtm.c
+++ b/drivers/misc/lkdtm.c
@@ -225,14 +225,24 @@ static int lkdtm_parse_commandline(void)

static int recursive_loop(int a)
{
- char buf[1024];
+ char *buf;
+ int ret;
+
+ buf = kmalloc(1024, GFP_KERNEL);
+ if (!buf) {
+ printk(KERN_ERR "lkdtm : couldn't allocate buffer\n");
+ return -ENOMEM;
+ }

memset(buf,0xFF,1024);
recur_count--;
if (!recur_count)
- return 0;
+ ret = 0;
else
- return recursive_loop(a);
+ ret = recursive_loop(a);
+
+ kfree(buf);
+ return ret;
}

void lkdtm_handler(void)


2009-03-02 04:12:07

by Ankita Garg

[permalink] [raw]
Subject: Re: [PATCH] lkdtm: reduce stack size

Hi Frank,

On Sun, Mar 01, 2009 at 09:00:14PM +0100, Frank Seidel wrote:
> From: Frank Seidel <[email protected]>
>
> Reduce stack memory footprint of lkdtm
> (recursive_loop). From 1024 bytes on i386 down
> to <10.
>

The intention here is infact to trigger a stack overflow. This module is
used to test kernel dumping mechanism like kdump. So the expectation is
that the kernel will dump as soon as the stack overflows.

> Signed-off-by: Frank Seidel <[email protected]>
> ---
> drivers/misc/lkdtm.c | 16 +++++++++++++---
> 1 file changed, 13 insertions(+), 3 deletions(-)
>
> --- a/drivers/misc/lkdtm.c
> +++ b/drivers/misc/lkdtm.c
> @@ -225,14 +225,24 @@ static int lkdtm_parse_commandline(void)
>
> static int recursive_loop(int a)
> {
> - char buf[1024];
> + char *buf;
> + int ret;
> +
> + buf = kmalloc(1024, GFP_KERNEL);
> + if (!buf) {
> + printk(KERN_ERR "lkdtm : couldn't allocate buffer\n");
> + return -ENOMEM;
> + }
>
> memset(buf,0xFF,1024);
> recur_count--;
> if (!recur_count)
> - return 0;
> + ret = 0;
> else
> - return recursive_loop(a);
> + ret = recursive_loop(a);
> +
> + kfree(buf);
> + return ret;
> }
>
> void lkdtm_handler(void)

--
Regards,
Ankita Garg ([email protected])
Linux Technology Center
IBM India Systems & Technology Labs,
Bangalore, India

2009-03-02 08:11:17

by Frank Seidel

[permalink] [raw]
Subject: Re: [PATCH] lkdtm: reduce stack size

Ankita Garg wrote:
> The intention here is infact to trigger a stack overflow. This module is
> used to test kernel dumping mechanism like kdump. So the expectation is
> that the kernel will dump as soon as the stack overflows.

Oh, i see. Then this patch of course doesn't make sense.

Thanks,
Frank

2009-03-02 10:12:50

by Ankita Garg

[permalink] [raw]
Subject: Re: [PATCH] lkdtm: reduce stack size

Hi Bernd,

On Mon, Mar 02, 2009 at 10:51:30AM +0100, Bernd Petrovitsch wrote:
> Hi!
>
> On Mon, 2009-03-02 at 09:41 +0530, Ankita Garg wrote:
> > Hi Frank,
> >
> > On Sun, Mar 01, 2009 at 09:00:14PM +0100, Frank Seidel wrote:
> > > From: Frank Seidel <[email protected]>
> > >
> > > Reduce stack memory footprint of lkdtm
> > > (recursive_loop). From 1024 bytes on i386 down
> > > to <10.
> > >
> >
> > The intention here is infact to trigger a stack overflow. This module is
> > used to test kernel dumping mechanism like kdump. So the expectation is
> > that the kernel will dump as soon as the stack overflows.
>
> That answers why the variable exists at all. It is only zeroed and never
> else used.
> Hmm, adding the above as a comment makes probably sense.
>

There is some text at the head of the module file that explains what the
module does and the usage, etc. I suppose that would be sufficient ?

--
Regards,
Ankita Garg ([email protected])
Linux Technology Center
IBM India Systems & Technology Labs,
Bangalore, India

2009-03-02 10:42:50

by Bernd Petrovitsch

[permalink] [raw]
Subject: Re: [PATCH] lkdtm: reduce stack size

Hi!

On Mon, 2009-03-02 at 09:41 +0530, Ankita Garg wrote:
> Hi Frank,
>
> On Sun, Mar 01, 2009 at 09:00:14PM +0100, Frank Seidel wrote:
> > From: Frank Seidel <[email protected]>
> >
> > Reduce stack memory footprint of lkdtm
> > (recursive_loop). From 1024 bytes on i386 down
> > to <10.
> >
>
> The intention here is infact to trigger a stack overflow. This module is
> used to test kernel dumping mechanism like kdump. So the expectation is
> that the kernel will dump as soon as the stack overflows.

That answers why the variable exists at all. It is only zeroed and never
else used.
Hmm, adding the above as a comment makes probably sense.

> > Signed-off-by: Frank Seidel <[email protected]>
> > ---
> > drivers/misc/lkdtm.c | 16 +++++++++++++---
> > 1 file changed, 13 insertions(+), 3 deletions(-)
> >
> > --- a/drivers/misc/lkdtm.c
> > +++ b/drivers/misc/lkdtm.c
> > @@ -225,14 +225,24 @@ static int lkdtm_parse_commandline(void)
> >
> > static int recursive_loop(int a)
> > {
> > - char buf[1024];
> > + char *buf;
> > + int ret;
> > +
> > + buf = kmalloc(1024, GFP_KERNEL);
> > + if (!buf) {
> > + printk(KERN_ERR "lkdtm : couldn't allocate buffer\n");
> > + return -ENOMEM;
> > + }
> >
> > memset(buf,0xFF,1024);
> > recur_count--;
> > if (!recur_count)
> > - return 0;
> > + ret = 0;
> > else
> > - return recursive_loop(a);
> > + ret = recursive_loop(a);
> > +
> > + kfree(buf);
> > + return ret;
> > }
> >
> > void lkdtm_handler(void)

Bernd
--
Firmix Software GmbH http://www.firmix.at/
mobil: +43 664 4416156 fax: +43 1 7890849-55
Embedded Linux Development and Services