mirror of
https://github.com/torvalds/linux.git
synced 2024-10-25 14:40:58 +00:00
42fddcc7c6
Prior the state variable was introduced by Ilya, the dev member was
used to determine whether the socket was bound or not. However, when
dev was read, proper SMP barriers and READ_ONCE were missing. In order
to address the missing barriers and READ_ONCE, we start using the
state variable as a point of synchronization. The state member
read/write is paired with proper SMP barriers, and from this follows
that the members described above does not need READ_ONCE if used in
conjunction with state check.
In all syscalls and the xsk_rcv path we check if state is
XSK_BOUND. If that is the case we do a SMP read barrier, and this
implies that the dev, umem and all rings are correctly setup. Note
that no READ_ONCE are needed for these variable if used when state is
XSK_BOUND (plus the read barrier).
To summarize: The members struct xdp_sock members dev, queue_id, umem,
fq, cq, tx, rx, and state were read lock-less, with incorrect barriers
and missing {READ, WRITE}_ONCE. Now, umem, fq, cq, tx, rx, and state
are read lock-less. When these members are updated, WRITE_ONCE is
used. When read, READ_ONCE are only used when read outside the control
mutex (e.g. mmap) or, not synchronized with the state member
(XSK_BOUND plus smp_rmb())
Note that dev and queue_id do not need a WRITE_ONCE or READ_ONCE, due
to the introduce state synchronization (XSK_BOUND plus smp_rmb()).
Introducing the state check also fixes a race, found by syzcaller, in
xsk_poll() where umem could be accessed when stale.
Suggested-by: Hillf Danton <hdanton@sina.com>
Reported-by: syzbot+c82697e3043781e08802@syzkaller.appspotmail.com
Fixes:
|
||
|---|---|---|
| .. | ||
| Kconfig | ||
| Makefile | ||
| xdp_umem.c | ||
| xdp_umem.h | ||
| xsk_diag.c | ||
| xsk_queue.c | ||
| xsk_queue.h | ||
| xsk.c | ||
| xsk.h | ||