what`s wrong?

what`s wrong?

[il boba]

Is there anybody that can help me understand what`s wrong with this code?

Mostly i`m not sure about the correct use of the functions filp_open, 
generic_file_read and filp_close.

When i compile as a module on the kernel and after, when i do the INSMOD  , 
at first said SEGMENTATION FAULT, right now it just blocks the whole 
machine.

Any help or hint is very welcome!!!
thanks
boba

Here is the code:

/* FILE NAME : pops.c */


#include <linux/kernel.h>
#include <linux/module.h>


#define BUFSIZ 8192

#if CONFIG_MODVERSIONS==1
#define MODVERSIONS
#include <linux/modversions.h>
#endif

#include <linux/fs.h>


#if LINUX_VERSION_CODE > KERNEL_VERSION(2,2,0)
#include <asm/uaccess.h>
#endif



int init_err_frame(int []);

/* prints the vector err_frame */

int init_module()
{
  int err_frame[BUFSIZ];
  int i, k;
  k = init_err_frame(err_frame);
  for (i = 0; i < k; i++){
    printk("%d \n", err_frame[i]);
  }
  return 0;
}

/* inizializza il vettore err_frame */

int init_err_frame(int err_frame[]) {

  int i, k = 0, j = 0;
  char buffer[BUFSIZ];
  int temp[2];
  int retval;


  struct file * f = 
filp_open("/usr/src/kernel-source-2.2.19/pinux/misc/err_file", 0, 0);  /* i 
want it only readable */
  if (f == '\0'){
       printk ("errore nell' APERTURA del file d'errore");
       return -1;
  }
  printk("%d ", f->f_count);
  if ((generic_file_read(f, buffer, BUFSIZ, &f->f_pos)) < 0 ){
    printk ("errore nella LETTURA del file");
    if ((retval = filp_close(f, NULL)) < 0){
       printk ("errore nella CHIUSURA del file");
       return -3;
    }
    return -2;
  }
  for (i=0; i < '\0'; i++){
    temp[j] = buffer[i];
    if (j == 1){
      err_frame[k] = (int) temp[0] + (int) temp[1];
      j = 0;
      k++;
    } else
      j = 1;
  }
  filp_close(f,NULL);
  return k;
}

void cleanup_module()
{
  printk(".........");
}



and i compile it as
cc -D__KERNEL__ -I/usr/src/kernel-source-2.2.19/include -Wall 
-Wstrict-prototypes -O2 -fomit-frame-pointer -fno-strict-aliasing -pipe 
-fno-strength-reduce -mpreferred-stack-boundary=2 -m486 -DCPU=486 -DMODULE 
-DEXPORT_SYMTAB -c pops.c



_________________________________________________________________
MSN Foto è il modo più semplice per condividere e stampare le tue foto: 
http://photos.msn.com/support/worldwide.aspx

Re: what`s wrong?

[Andreas Dilger]

On Apr 24, 2002  18:06 +0200, il boba wrote:
> Is there anybody that can help me understand what`s wrong with this code?

Yes, easily spotted a major problem without even reading the whole
thing.

> #define BUFSIZ 8192
> 
> int init_module()
> {
>  int err_frame[BUFSIZ];

The entire kernel stack is only 8kB in size.  You have already killed
a bunch of random memory by allocating this much memory on the stack.
You allocated 4*8192 = 32kB on the stack here.

> #if CONFIG_MODVERSIONS==1
> #define MODVERSIONS
> #include <linux/modversions.h>
> #endif

You shouldn't do things like this.  Rather have a makefile which sets
the correct defines.

> int init_err_frame(int err_frame[]) {
>  int i, k = 0, j = 0;
>  char buffer[BUFSIZ];

Another 8kB on the stack here - further random corruption.

>  struct file * f = 
> filp_open("/usr/src/kernel-source-2.2.19/pinux/misc/err_file", 0, 0);  /* i 
> want it only readable */

More clear to add "O_RDONLY" to filp_open, even though it is still 0.

>  if (f == '\0'){
>       printk ("errore nell' APERTURA del file d'errore");
>       return -1;
>  }

How about "if (f == NULL)"?

>  if ((generic_file_read(f, buffer, BUFSIZ, &f->f_pos)) < 0 ){

Calling generic_file_read() may work in some cases, but not others.  It
depends on the filesystem.  Use "f->f_op->read" instead.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Re: what`s wrong?

[Jerry Cooperstein]

On Wed, Apr 24, 2002 at 10:40:56AM -0600, Andreas Dilger wrote:
> On Apr 24, 2002  18:06 +0200, il boba wrote:
> > Is there anybody that can help me understand what`s wrong with this code?
......

Even with all this he's going to have trouble since his buffer is in
kernel space and the read is going to expect it to be in user-context...

Reading and writing files from within the kernel is almost never a good
idea and has been made difficult by design.  He should look at what
he is trying to do and perhaps look for another interface, such as
use of ioctls, proc, seq_file, etc..  He probably doesn't need to
do this I/O from within the kernel and should rethink it.

 Jerry Cooperstein  <coop@axian.com>
 Senior Consultant
 Axian, Inc.   <info@axian.com>
 4800 SW Griffith Dr., Ste. 202, Beaverton, OR  97005 USA
 http://www.axian.com/               Software Consulting and Training