[PATCH] (2.6.x) toshiba_acpi needs copy_from_user (fixes oops)

[PATCH] (2.6.x) toshiba_acpi needs copy_from_user (fixes oops)

[Barry K. Nathan]

On kernels with the 4G/4G patch (like some of the recent kernels in
Fedora Core 2 development), writing stuff to the /proc/acpi/toshiba/*
files causes an oops. As it turns out, this is because the driver is
accessing userspace data without first doing copy_from_user(). IOW, this
is a bug in toshiba_acpi, not a bug in the 4G/4G patch.

Here's a patch to fix this bug. I've tested it on 2.6.4 + some patches
from the FC kernels (including the 4G/4G patch) and it fixes my oopses.
I have also tested it against vanilla 2.6.4 and I haven't encountered
any regressions.

If there are any problems with this patch, let me know.

-Barry K. Nathan <barryn@pobox.com>


diff -ruN linux-2.6.4/drivers/acpi/toshiba_acpi.c linux-2.6.4-bkn1/drivers/acpi/toshiba_acpi.c
--- linux-2.6.4/drivers/acpi/toshiba_acpi.c	2004-03-12 21:31:59.000000000 -0800
+++ linux-2.6.4-bkn1/drivers/acpi/toshiba_acpi.c	2004-03-12 22:27:07.000000000 -0800
@@ -41,6 +41,7 @@
 #include <linux/init.h>
 #include <linux/types.h>
 #include <linux/proc_fs.h>
+#include <asm/uaccess.h>
 
 #include <acpi/acpi_drivers.h>
 
@@ -269,10 +270,18 @@
 }
 
 static int
-dispatch_write(struct file* file, const char* buffer, unsigned long count,
-	ProcItem* item)
+dispatch_write(struct file* file, const char __user *buffer,
+	unsigned long count, ProcItem* item)
 {
-	return item->write_func(buffer, count);
+	char str[48] = {'\0'};
+
+	if (count > sizeof(str) - 1)
+		return count;
+	
+	if (copy_from_user(str, buffer, count))
+		return -EFAULT;
+
+	return item->write_func(str, count);
 }
 
 static char*

Re: [PATCH] (2.6.x) toshiba_acpi needs copy_from_user (fixes oops)

[John Belmonte]

Thank you for tracking this down.

Please don't apply this patch to the kernel tree.  I will submit a 
variation via Len Brown.  In any case, it appears at least one other 
ACPI driver has a similar bug, so best to go through Len.

-John


Barry K. Nathan wrote:
> On kernels with the 4G/4G patch (like some of the recent kernels in
> Fedora Core 2 development), writing stuff to the /proc/acpi/toshiba/*
> files causes an oops. As it turns out, this is because the driver is
> accessing userspace data without first doing copy_from_user(). IOW, this
> is a bug in toshiba_acpi, not a bug in the 4G/4G patch.
> 
> Here's a patch to fix this bug. I've tested it on 2.6.4 + some patches
> from the FC kernels (including the 4G/4G patch) and it fixes my oopses.
> I have also tested it against vanilla 2.6.4 and I haven't encountered
> any regressions.
> 
> If there are any problems with this patch, let me know.
> 
> -Barry K. Nathan <barryn@pobox.com>
> 
> 
> diff -ruN linux-2.6.4/drivers/acpi/toshiba_acpi.c linux-2.6.4-bkn1/drivers/acpi/toshiba_acpi.c
> --- linux-2.6.4/drivers/acpi/toshiba_acpi.c	2004-03-12 21:31:59.000000000 -0800
> +++ linux-2.6.4-bkn1/drivers/acpi/toshiba_acpi.c	2004-03-12 22:27:07.000000000 -0800
> @@ -41,6 +41,7 @@
>  #include <linux/init.h>
>  #include <linux/types.h>
>  #include <linux/proc_fs.h>
> +#include <asm/uaccess.h>
>  
>  #include <acpi/acpi_drivers.h>
>  
> @@ -269,10 +270,18 @@
>  }
>  
>  static int
> -dispatch_write(struct file* file, const char* buffer, unsigned long count,
> -	ProcItem* item)
> +dispatch_write(struct file* file, const char __user *buffer,
> +	unsigned long count, ProcItem* item)
>  {
> -	return item->write_func(buffer, count);
> +	char str[48] = {'\0'};
> +
> +	if (count > sizeof(str) - 1)
> +		return count;
> +	
> +	if (copy_from_user(str, buffer, count))
> +		return -EFAULT;
> +
> +	return item->write_func(str, count);
>  }
>  
>  static char*
> 
> 


-- 
http:// if  ile.org/

Re: [PATCH] (2.6.x) toshiba_acpi needs copy_from_user (fixes oops)

[Barry K. Nathan]

On Sun, Mar 14, 2004 at 12:52:04AM -0500, John Belmonte wrote:
> Thank you for tracking this down.

That reminds me of something I forgot to mention in my original message.

I filed this in Red Hat Bugzilla (#117682), and Arjan van de Ven
mentioned the likely cause. I then confirmed it and coded up my patch.

So, Arjan deserves credit too. I meant to mention this in my original
e-mail, but I accidentally forgot. Sorry about that.

> Please don't apply this patch to the kernel tree.  I will submit a 
> variation via Len Brown.  In any case, it appears at least one other 
> ACPI driver has a similar bug, so best to go through Len.

Ok, sounds good to me.

-Barry K. Nathan <barryn@pobox.com>