[PATCH] Fix potential memory leak in tipc_named_node_up()

[PATCH] Fix potential memory leak in tipc_named_node_up()

[Jesper Juhl]

There seems to be a memory leak in net/tipc/name_distr.c::tipc_named_node_up()

The function, with comments, is this :

void tipc_named_node_up(unsigned long node)
{
        struct publication *publ;
        struct distr_item *item = NULL;
        struct sk_buff *buf = NULL;
        u32 left = 0;
        u32 rest;
        u32 max_item_buf;

        read_lock_bh(&tipc_nametbl_lock);
        max_item_buf = TIPC_MAX_USER_MSG_SIZE / ITEM_SIZE;
        max_item_buf *= ITEM_SIZE;
        rest = publ_cnt * ITEM_SIZE;

        list_for_each_entry(publ, &publ_root, local_list) {
-----> If we stop processing here after doing 1, 2 & 3 below we end up at (4) (below).
                if (!buf) {
                        left = (rest <= max_item_buf) ? rest : max_item_buf;
                        rest -= left;
                        buf = named_prepare_buf(PUBLICATION, left, node);
-----> (1) here we allocate memory and store a pointer to it in 'buf'.

                        if (!buf) {
-----> (2) This test needs to fail, meaning we did allocate some memory.
                                warn("Bulk publication distribution failure\n");
                                goto exit;
                        }
                        item = (struct distr_item *)msg_data(buf_msg(buf));
                }
                publ_to_item(item, publ);
                item++;
                left -= ITEM_SIZE;
                if (!left) {
-----> (3) If this test fails we loop and do nothing to 'buf'.
                        msg_set_link_selector(buf_msg(buf), node);
                        dbg("tipc_named_node_up: sending publish msg to "
                            "<%u.%u.%u>\n", tipc_zone(node),
                            tipc_cluster(node), tipc_node(node));
                        tipc_link_send(buf, node, node);
                        buf = NULL;
                }
        }
exit:
        read_unlock_bh(&tipc_nametbl_lock);
-----> (4) here we return without freeing 'buf' - memory leak.
}

Luckily this is easy to fix, since we can only leave the function with 'buf'
either set to NULL or (in the leak case) set to a valid address, and since 
kfree() handles being passed NULL gracefully we can simply kfree(buf) just 
before we leave the function.

The patch below frees 'buf' before leaving the function, thus avoiding the 
potential leak.



Fix a potential memory leak in tipc_named_node_up()

Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
---

 net/tipc/name_distr.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/net/tipc/name_distr.c b/net/tipc/name_distr.c
index 39fd161..228ccfe 100644
--- a/net/tipc/name_distr.c
+++ b/net/tipc/name_distr.c
@@ -204,6 +204,7 @@ void tipc_named_node_up(unsigned long node)
 	}
 exit:
 	read_unlock_bh(&tipc_nametbl_lock);
+	kfree(buf);
 }
 
 /**

Re: [PATCH] Fix potential memory leak in tipc_named_node_up()

[Björn Steinbrink]

On 2007.05.28 22:58:08 +0200, Jesper Juhl wrote:
> There seems to be a memory leak in net/tipc/name_distr.c::tipc_named_node_up()
> 
> The function, with comments, is this :
> 
> void tipc_named_node_up(unsigned long node)
> {
>         struct publication *publ;
>         struct distr_item *item = NULL;
>         struct sk_buff *buf = NULL;
>         u32 left = 0;
>         u32 rest;
>         u32 max_item_buf;
> 
>         read_lock_bh(&tipc_nametbl_lock);
>         max_item_buf = TIPC_MAX_USER_MSG_SIZE / ITEM_SIZE;
>         max_item_buf *= ITEM_SIZE;
>         rest = publ_cnt * ITEM_SIZE;
> 
>         list_for_each_entry(publ, &publ_root, local_list) {
> -----> If we stop processing here after doing 1, 2 & 3 below we end up at (4) (below).
>                 if (!buf) {
>                         left = (rest <= max_item_buf) ? rest : max_item_buf;
>                         rest -= left;
>                         buf = named_prepare_buf(PUBLICATION, left, node);
> -----> (1) here we allocate memory and store a pointer to it in 'buf'.
> 
>                         if (!buf) {
> -----> (2) This test needs to fail, meaning we did allocate some memory.
>                                 warn("Bulk publication distribution failure\n");
>                                 goto exit;
>                         }
>                         item = (struct distr_item *)msg_data(buf_msg(buf));
>                 }
>                 publ_to_item(item, publ);
>                 item++;
>                 left -= ITEM_SIZE;
>                 if (!left) {
> -----> (3) If this test fails we loop and do nothing to 'buf'.
>                         msg_set_link_selector(buf_msg(buf), node);
>                         dbg("tipc_named_node_up: sending publish msg to "
>                             "<%u.%u.%u>\n", tipc_zone(node),
>                             tipc_cluster(node), tipc_node(node));
>                         tipc_link_send(buf, node, node);
>                         buf = NULL;
>                 }
>         }
> exit:
>         read_unlock_bh(&tipc_nametbl_lock);
> -----> (4) here we return without freeing 'buf' - memory leak.
> }
> 
> Luckily this is easy to fix, since we can only leave the function with 'buf'
> either set to NULL or (in the leak case) set to a valid address, and since 
> kfree() handles being passed NULL gracefully we can simply kfree(buf) just 
> before we leave the function.

Actually, I don't think that there's a leak.

publ_cnt: Number of items in the list

rest = publ_cnt * ITEM_SIZE
max_item_buf = n * ITEM_SIZE (Buffer can hold n elements at most)

1)
If publ_cnt <= n, "rest" becomes 0 and "left" becomes publ_cnt * ITEM_SIZE,
so for the last iteration "left" becomes 0 and "buf" is freed.

2)
And if publ_cnt > n, "left" becomes 0 in the nth iteration. As "rest"
already got decrement by n * ITEM_SIZE, you now got:
rest = (publ_cnt - n) * ITEM_SIZE
Then after 2*n iterations:
rest = (publ_cnt - 2*n) * ITEM_SIZE
and so on, until publ_cnt - m*n < n

At that point, "left" becomes (publ_cnt - m*n) * ITEM_SIZE, and there are
also (publ_cnt - m*n) iterations left, so "left" again becomes 0 in the
last iteration and "buf" is freed.


Besides that, is it valid to call kfree() on a buffer allocated by
alloc_skb()?

Thanks,
Björn

Re: [PATCH] Fix potential memory leak in tipc_named_node_up()

[Jesper Juhl]

On 29/05/07, Björn Steinbrink <B.Steinbrink@gmx.de> wrote:
> On 2007.05.28 22:58:08 +0200, Jesper Juhl wrote:
> > There seems to be a memory leak in net/tipc/name_distr.c::tipc_named_node_up()
> >
> > The function, with comments, is this :
> >
> > void tipc_named_node_up(unsigned long node)
> > {
> >         struct publication *publ;
> >         struct distr_item *item = NULL;
> >         struct sk_buff *buf = NULL;
> >         u32 left = 0;
> >         u32 rest;
> >         u32 max_item_buf;
> >
> >         read_lock_bh(&tipc_nametbl_lock);
> >         max_item_buf = TIPC_MAX_USER_MSG_SIZE / ITEM_SIZE;
> >         max_item_buf *= ITEM_SIZE;
> >         rest = publ_cnt * ITEM_SIZE;
> >
> >         list_for_each_entry(publ, &publ_root, local_list) {
> > -----> If we stop processing here after doing 1, 2 & 3 below we end up at (4) (below).
> >                 if (!buf) {
> >                         left = (rest <= max_item_buf) ? rest : max_item_buf;
> >                         rest -= left;
> >                         buf = named_prepare_buf(PUBLICATION, left, node);
> > -----> (1) here we allocate memory and store a pointer to it in 'buf'.
> >
> >                         if (!buf) {
> > -----> (2) This test needs to fail, meaning we did allocate some memory.
> >                                 warn("Bulk publication distribution failure\n");
> >                                 goto exit;
> >                         }
> >                         item = (struct distr_item *)msg_data(buf_msg(buf));
> >                 }
> >                 publ_to_item(item, publ);
> >                 item++;
> >                 left -= ITEM_SIZE;
> >                 if (!left) {
> > -----> (3) If this test fails we loop and do nothing to 'buf'.
> >                         msg_set_link_selector(buf_msg(buf), node);
> >                         dbg("tipc_named_node_up: sending publish msg to "
> >                             "<%u.%u.%u>\n", tipc_zone(node),
> >                             tipc_cluster(node), tipc_node(node));
> >                         tipc_link_send(buf, node, node);
> >                         buf = NULL;
> >                 }
> >         }
> > exit:
> >         read_unlock_bh(&tipc_nametbl_lock);
> > -----> (4) here we return without freeing 'buf' - memory leak.
> > }
> >
> > Luckily this is easy to fix, since we can only leave the function with 'buf'
> > either set to NULL or (in the leak case) set to a valid address, and since
> > kfree() handles being passed NULL gracefully we can simply kfree(buf) just
> > before we leave the function.
>
> Actually, I don't think that there's a leak.
>
> publ_cnt: Number of items in the list
>
> rest = publ_cnt * ITEM_SIZE
> max_item_buf = n * ITEM_SIZE (Buffer can hold n elements at most)
>
> 1)
> If publ_cnt <= n, "rest" becomes 0 and "left" becomes publ_cnt * ITEM_SIZE,
> so for the last iteration "left" becomes 0 and "buf" is freed.
>
> 2)
> And if publ_cnt > n, "left" becomes 0 in the nth iteration. As "rest"
> already got decrement by n * ITEM_SIZE, you now got:
> rest = (publ_cnt - n) * ITEM_SIZE
> Then after 2*n iterations:
> rest = (publ_cnt - 2*n) * ITEM_SIZE
> and so on, until publ_cnt - m*n < n
>
> At that point, "left" becomes (publ_cnt - m*n) * ITEM_SIZE, and there are
> also (publ_cnt - m*n) iterations left, so "left" again becomes 0 in the
> last iteration and "buf" is freed.
>
Ok, creating patches late in the evening after several cups of coffee
is clearly a bad idea. I think you are right and there is really no
leak. Thanks.


>
> Besides that, is it valid to call kfree() on a buffer allocated by
> alloc_skb()?
>
Arrgh, no of course not, seems I messed this one up real good :(


-- 
Jesper Juhl <jesper.juhl@gmail.com>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html