[PATCH 01/17] ckpt_write_err: use single format with %(T) style tokens

Oren Laadan orenl at librato.com
Fri Oct 30 08:45:07 PDT 2009



Serge E. Hallyn wrote:
> Quoting Oren Laadan (orenl at librato.com):
>>
>> serue at us.ibm.com wrote:
>>> From: Serge E. Hallyn <serue at us.ibm.com>

[...]
>>> +	 */
>>> +	alloclen = 37 + 8 * strlen(fmt);
>> This calculation assumed that @fmt had only format string...
>> At the very minimum you could take strlen(fmt)/3 (+1 to round up)
> 
> Yeah, I didn't want to think about that in detail yet :)
> 
>> I thought you were going to use a @ctx->buffer or something ?
> 
> And I am, for my string.  We need one for the expanded fmt here,
> and then one to snprintf the final string into so we can write it
> out.
> 
> Shall I just add a @ctx->fmtbuf?

Sure.

> 
>>> +	format = kzalloc(alloclen, GFP_KERNEL);
>>> +	if (!format)
>>> +		return NULL;
>>> +
>>> +	for (; *fmt; fmt++) {
>>> +		BUG_ON(len > alloclen);
>>> +		if (*fmt != '%' || fmt[1] != '(' || fmt[3] != ')') {
>> This is still a bit risky .. how about adding
>> 	|| fmt[2] == '\0'
>> between the 2nd and 3rd test ?
> 
> Well I can do that, but since we provide the fmt strings and there is
> no risk for an information leak I didn't think it was worth making
> the line even longer.  But ok, I'll add it...

Well, if a developer gives a string like "hello %(", then you
will potentially go past the end of the string and eventually
crash (or worse).

Oren.



More information about the Containers mailing list