On Fri, Nov 18, 2022 at 10:15:50PM -0500, Xin Long wrote:
> On Fri, Nov 18, 2022 at 3:48 AM Zhengchao Shao <shaozhengchao@xxxxxxxxxx> wrote:
> >
> > When sctp_stream_outq_migrate() is called to release stream out resources,
> > the memory pointed to by prio_head in stream out is not released.
> >
> > The memory leak information is as follows:
> > unreferenced object 0xffff88801fe79f80 (size 64):
> > comm "sctp_repo", pid 7957, jiffies 4294951704 (age 36.480s)
> > hex dump (first 32 bytes):
> > 80 9f e7 1f 80 88 ff ff 80 9f e7 1f 80 88 ff ff ................
> > 90 9f e7 1f 80 88 ff ff 90 9f e7 1f 80 88 ff ff ................
> > backtrace:
> > [<ffffffff81b215c6>] kmalloc_trace+0x26/0x60
> > [<ffffffff88ae517c>] sctp_sched_prio_set+0x4cc/0x770
> > [<ffffffff88ad64f2>] sctp_stream_init_ext+0xd2/0x1b0
> > [<ffffffff88aa2604>] sctp_sendmsg_to_asoc+0x1614/0x1a30
> > [<ffffffff88ab7ff1>] sctp_sendmsg+0xda1/0x1ef0
> > [<ffffffff87f765ed>] inet_sendmsg+0x9d/0xe0
> > [<ffffffff8754b5b3>] sock_sendmsg+0xd3/0x120
> > [<ffffffff8755446a>] __sys_sendto+0x23a/0x340
> > [<ffffffff87554651>] __x64_sys_sendto+0xe1/0x1b0
> > [<ffffffff89978b49>] do_syscall_64+0x39/0xb0
> > [<ffffffff89a0008b>] entry_SYSCALL_64_after_hwframe+0x63/0xcd
> >
> > Fixes: 637784ade221 ("sctp: introduce priority based stream scheduler")
> > Reported-by: syzbot+29c402e56c4760763cc0@xxxxxxxxxxxxxxxxxxxxxxxxx
> > Signed-off-by: Zhengchao Shao <shaozhengchao@xxxxxxxxxx>
> > ---
> > net/sctp/stream.c | 6 ++++++
> > 1 file changed, 6 insertions(+)
> >
> > diff --git a/net/sctp/stream.c b/net/sctp/stream.c
> > index ef9fceadef8d..a17dc368876f 100644
> > --- a/net/sctp/stream.c
> > +++ b/net/sctp/stream.c
> > @@ -70,6 +70,9 @@ static void sctp_stream_outq_migrate(struct sctp_stream *stream,
> > * sctp_stream_update will swap ->out pointers.
> > */
> > for (i = 0; i < outcnt; i++) {
> > + if (SCTP_SO(new, i)->ext)
> > + kfree(SCTP_SO(new, i)->ext->prio_head);
> > +
> > kfree(SCTP_SO(new, i)->ext);
> > SCTP_SO(new, i)->ext = SCTP_SO(stream, i)->ext;
> > SCTP_SO(stream, i)->ext = NULL;
> > @@ -77,6 +80,9 @@ static void sctp_stream_outq_migrate(struct sctp_stream *stream,
> > }
> >
> > for (i = outcnt; i < stream->outcnt; i++) {
> > + if (SCTP_SO(stream, i)->ext)
> > + kfree(SCTP_SO(stream, i)->ext->prio_head);
> > +
> > kfree(SCTP_SO(stream, i)->ext);
> > SCTP_SO(stream, i)->ext = NULL;
> > }
> > --
> > 2.17.1
> >
> This is not a proper fix:
> 1. you shouldn't access "prio_head" outside stream_sched_prio.c.
> 2. the prio_head you freed might be used by other out streams, freeing
> it unconditionally would cause either a double free or use after free.
>
> I'm afraid we have to add a ".free_sid" in sctp_sched_ops, and
> implement it for sctp_sched_prio, like:
>
> +static void sctp_sched_prio_free_sid(struct sctp_stream *stream, __u16 sid)
> +{
> + struct sctp_stream_priorities *prio = SCTP_SO(stream,
> sid)->ext->prio_head;
> + int i;
> +
> + if (!prio)
> + return;
> +
> + SCTP_SO(stream, sid)->ext->prio_head = NULL;
> + for (i = 0; i < stream->outcnt; i++) {
Instead of checking all streams, the for() can/should be replaced by
(from sctp_sched_prio_free):
if (!list_empty(&prio->prio_sched))
return;
> + if (SCTP_SO(stream, i)->ext &&
> + SCTP_SO(stream, i)->ext->prio_head == prio)
> + return;
> + }
> + kfree(prio);
> +}
> +
> static void sctp_sched_prio_free(struct sctp_stream *stream)
> {
> struct sctp_stream_priorities *prio, *n;
> @@ -323,6 +340,7 @@ static struct sctp_sched_ops sctp_sched_prio = {
> .get = sctp_sched_prio_get,
> .init = sctp_sched_prio_init,
> .init_sid = sctp_sched_prio_init_sid,
> + .free_sid = sctp_sched_prio_free_sid,
> .free = sctp_sched_prio_free,
> .enqueue = sctp_sched_prio_enqueue,
> .dequeue = sctp_sched_prio_dequeue,
>
> then call it in sctp_stream_outq_migrate(), like:
>
> +static void sctp_stream_free_ext(struct sctp_stream *stream, __u16 sid)
> +{
> + struct sctp_sched_ops *sched = sctp_sched_ops_from_stream(stream);
> +
> + sched->free_sid(stream, sid);
> + kfree(SCTP_SO(stream, sid)->ext);
> + SCTP_SO(stream, sid)->ext = NULL;
> +}
> +
> /* Migrates chunks from stream queues to new stream queues if needed,
> * but not across associations. Also, removes those chunks to streams
> * higher than the new max.
> @@ -70,16 +79,14 @@ static void sctp_stream_outq_migrate(struct
> sctp_stream *stream,
> * sctp_stream_update will swap ->out pointers.
> */
> for (i = 0; i < outcnt; i++) {
> - kfree(SCTP_SO(new, i)->ext);
> + sctp_stream_free_ext(new, i);
> SCTP_SO(new, i)->ext = SCTP_SO(stream, i)->ext;
> SCTP_SO(stream, i)->ext = NULL;
> }
> }
>
> - for (i = outcnt; i < stream->outcnt; i++) {
> - kfree(SCTP_SO(stream, i)->ext);
> - SCTP_SO(stream, i)->ext = NULL;
> - }
> + for (i = outcnt; i < stream->outcnt; i++)
> + sctp_stream_free_ext(new, i);
> }
>
> Marcelo, do you see a better solution?
No. Your suggestion is the best I could think of too.
Another approach would be to expose sched->free and do all the freeing
at once, like sctp_stream_free() does. But the above is looks cleaner
and makes it evident that freeing 'ext' is not trivial.
With the proposal above, sctp_sched_prio_free() becomes an
optimization, if we can call it that. With the for/if replacement
above, not even that, and should be removed. Including sctp_sched_ops
'free' pointer.
sctp_stream_free() then should be updated to use the new
sctp_stream_free_ext() instead, instead of mangling it directly.
Makes sense?
Thanks,
Marcelo
