>>On 1/1/07, Amit Choudhary <[EMAIL PROTECTED]> wrote:
>> +#define KFREE(x) \
>> + do { \
>> + kfree(x); \
>> + x = NULL; \
>> + } while(0)
>>NAK until you have actual callers for it. CONFIG_SLAB_DEBUG already
>>catches use after free and double-free so I don't see the point of
>>this.
Well, I am not proposing this as a debugging aid. The idea is about correct
programming, atleast
from my view. Ideally, if you kfree(x), then you should set x to NULL. So,
either programmers do
it themselves or a ready made macro do it for them.
In my opinion, the programmers may welcome the macro that does it for them.
There is another good part about it that results in better programming and
shorter code.
Consider an array x[10]. I allocate memory for all of them and then kfree them
- kfree(x[0]),
kfree(x[1]), etc. But I do not set these elements to NULL. So,
x[0] = _location_0_already_freed
x[1] = _location_1_already_freed
... and so on...
Now, consider that when I try to allocate memory again, memory allocation fails
at x[2]. So, we
have now,
x[0] = _valid_location_0
x[1] = _valid_location_1
x[2] = NULL
x[3] = _location_3_already_freed
So, now to avoid error path leak I have to free all the allocated memory. For
this I have to
remember where the allocation failed because I cannot do kfree(x[3]) because it
will crash.
You can easily visualize that how KFREE would have helped. Since, I have
already KFREE'D them
earlier, x[3] is guaranteed to be NULL and I can free the entire array 'x'
wihtout worrying.
So, the code becomes simpler.
Now, consider that there are two more arrays like 'x' being used in the same
function. So, now we
have 'x', 'y', 'z' all allocating memory in the same function. Memory
allocation can fail at
anytime. So, not only do I have to remember the element location where the
allocation failed but
also the array that contains that element.
So, to avoid error path leak, we will have something like this (assume same
number of elements in
all arrays):
case z_nomem:
free_from_0_to_i
free_all_'y'
free_all_'x'
return;
case y_nomem:
free_from_0_to_i
free_all_'x'
return;
case x_nomem:
free_from_0_to_i
return;
However, if the programmer would have used KFREE, then the error path leak code
have been shorter
and easier.
case nomem:
free_all_'z'
free_all_'y'
free_all_'x'
return;
I hope that I have made my point but please let me know if I have missed out
something or made
some assumption that is not correct.
Regards,
Amit
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
