Re: [ 1/1] printk: fix buffer overflow when calling log_prefix function from call_console_drivers

[Satoru Takeuchi <satoru.takeuchi@gmail.com>]

Hi Alexandre,

At Mon, 18 Feb 2013 10:24:56 -0800,
Greg Kroah-Hartman wrote:
> 
> 3.4-stable review patch.  If anyone has any objections, please let me know.
> 
> ------------------
> 
> From: Alexandre SIMON <Alexandre.Simon@univ-lorraine.fr>
> 
> This patch corrects a buffer overflow in kernels from 3.0 to 3.4 when calling
> log_prefix() function from call_console_drivers().
> 
> This bug existed in previous releases but has been revealed with commit
> 162a7e7500f9664636e649ba59defe541b7c2c60 (2.6.39 => 3.0) that made changes
> about how to allocate memory for early printk buffer (use of memblock_alloc).
> It disappears with commit 7ff9554bb578ba02166071d2d487b7fc7d860d62 (3.4 => 3.5)
> that does a refactoring of printk buffer management.
> 
> In log_prefix(), the access to "p[0]", "p[1]", "p[2]" or
> "simple_strtoul(&p[1], &endp, 10)" may cause a buffer overflow as this
> function is called from call_console_drivers by passing "&LOG_BUF(cur_index)"
> where the index must be masked to do not exceed the buffer's boundary.

I reviewed this patch and it seems to be good for me. Since I'm not good at
printk code, I want to confirm whether what I think is correct or not.
Is the following my understanding correct?

> -			cur_index += log_prefix(&LOG_BUF(cur_index), &msg_level, NULL);

Here is one example of the problematic case.

  +---- start of log_buf
  |
  |                           +--- end of log_buf
  |                           |
  v                           v
  <-------- log_buf ----------><------- * outside of log_buf. Don't access here !!! * --- ~~~
                              ^
                              |
                              cur_index

In this case, only LOG_BUF(cur_index) is safe to access and

 - "LOG_BUF(cur_index) + 1" as p[1],
 - "LOG_BUF(cur_index) + 2" as p[2], and
 - "LOG_BUF(cur_index) + 1 or more" as simple_strtoul(&p[1], &endp, 10)

in log_prefix() are not to do so. Hence touching them would cause the system hang as you
said as follows.

> 
> The trick is to prepare in call_console_drivers() a buffer with the necessary
> data (PRI field of syslog message) to be safely evaluated in log_prefix().
> 
> This patch can be applied to stable kernel branches 3.0.y, 3.2.y and 3.4.y.
> 
> Without this patch, one can freeze a server running this loop from shell :
>   $ export DUMMY=`cat /dev/urandom | tr -dc '12345AZERTYUIOPQSDFGHJKLMWXCVBNazertyuiopqsdfghjklmwxcvbn' | head -c255`
>   $ while true do ; echo $DUMMY > /dev/kmsg ; done
> 
> The "server freeze" depends on where memblock_alloc does allocate printk buffer :
> if the buffer overflow is inside another kernel allocation the problem may not
> be revealed, else the server may hangs up.
...
> --- a/kernel/printk.c
> +++ b/kernel/printk.c
> @@ -638,8 +638,19 @@ static void call_console_drivers(unsigne
>  	start_print = start;
>  	while (cur_index != end) {
>  		if (msg_level < 0 && ((end - cur_index) > 2)) {
> +			/*
> +			 * prepare buf_prefix, as a contiguous array,
> +			 * to be processed by log_prefix function
> +			 */
> +			char buf_prefix[SYSLOG_PRI_MAX_LENGTH+1];
> +			unsigned i;
> +			for (i = 0; i < ((end - cur_index)) && (i < SYSLOG_PRI_MAX_LENGTH); i++) {

The condition, "i < ((end - cur_index)) && (i < SYSLOG_PRI_MAX_LENGTH)", is to prevent
access over

 - the region to write out, and
 - the max length of log_prefix.

In addition, "min(end - cur_index, SYSLOG_PRI_MAX_LENGTH)" has the same meaning here.

> +				buf_prefix[i] = LOG_BUF(cur_index + i);

You ensure that all characters (not only first character) in the candidate of log_prefix
are inside of log_buf here by copying each character of them one by one.

Thanks,
Satoru