On 8/15/20 11:16 AM, Jens Axboe wrote: > On 8/15/20 10:00 AM, syzbot wrote: >> Hello, >> >> syzbot found the following issue on: >> >> HEAD commit: 7fca4dee Merge tag 'powerpc-5.9-2' of git://git.kernel.org.. >> git tree: upstream >> console output: https://syzkaller.appspot.com/x/log.txt?x=1264d116900000 >> kernel config: https://syzkaller.appspot.com/x/.config?x=21f0d1d2df6d5fc >> dashboard link: https://syzkaller.appspot.com/bug?extid=7f617d4a9369028b8a2c >> compiler: clang version 10.0.0 (https://github.com/llvm/llvm-project/ c2443155a0fb245c8f17f2c1c72b6ea391e86e81) >> syz repro: https://syzkaller.appspot.com/x/repro.syz?x=10f211d2900000 >> C reproducer: https://syzkaller.appspot.com/x/repro.c?x=1721b0ce900000 >> >> The issue was bisected to: >> >> commit 18bceab101adde8f38de76016bc77f3f25cf22f4 >> Author: Jens Axboe <axboe@xxxxxxxxx> >> Date: Fri May 15 17:56:54 2020 +0000 >> >> io_uring: allow POLL_ADD with double poll_wait() users > > I can reproduce this, I'll fix it up. Thanks! This should fix it: >From 34fc8d0b76572c9fb184ab589d682dccfeb5c039 Mon Sep 17 00:00:00 2001 From: Jens Axboe <axboe@xxxxxxxxx> Date: Sat, 15 Aug 2020 11:44:50 -0700 Subject: [PATCH] io_uring: sanitize double poll handling There's a bit of confusion on the matching pairs of poll vs double poll, depending on if the request is a pure poll (IORING_OP_POLL_ADD) or poll driven retry. Add io_poll_get_double() that returns the double poll waitqueue, if any, and io_poll_get_single() that returns the original poll waitqueue. With that, remove the argument to io_poll_remove_double(). Finally ensure that wait->private is cleared once the double poll handler has run, so that remove knows it's already been seen. Cc: stable@xxxxxxxxxxxxxxx # v5.8 Reported-by: syzbot+7f617d4a9369028b8a2c@xxxxxxxxxxxxxxxxxxxxxxxxx Fixes: 18bceab101ad ("io_uring: allow POLL_ADD with double poll_wait() users") Signed-off-by: Jens Axboe <axboe@xxxxxxxxx> --- fs/io_uring.c | 34 +++++++++++++++++++++++++--------- 1 file changed, 25 insertions(+), 9 deletions(-) diff --git a/fs/io_uring.c b/fs/io_uring.c index 7dd6df15bc49..cb030912bf5e 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -4649,9 +4649,24 @@ static bool io_poll_rewait(struct io_kiocb *req, struct io_poll_iocb *poll) return false; } -static void io_poll_remove_double(struct io_kiocb *req, void *data) +static struct io_poll_iocb *io_poll_get_double(struct io_kiocb *req) { - struct io_poll_iocb *poll = data; + /* pure poll stashes this in ->io, poll driven retry elsewhere */ + if (req->opcode == IORING_OP_POLL_ADD) + return (struct io_poll_iocb *) req->io; + return req->apoll->double_poll; +} + +static struct io_poll_iocb *io_poll_get_single(struct io_kiocb *req) +{ + if (req->opcode == IORING_OP_POLL_ADD) + return &req->poll; + return &req->apoll->poll; +} + +static void io_poll_remove_double(struct io_kiocb *req) +{ + struct io_poll_iocb *poll = io_poll_get_double(req); lockdep_assert_held(&req->ctx->completion_lock); @@ -4671,7 +4686,7 @@ static void io_poll_complete(struct io_kiocb *req, __poll_t mask, int error) { struct io_ring_ctx *ctx = req->ctx; - io_poll_remove_double(req, req->io); + io_poll_remove_double(req); req->poll.done = true; io_cqring_fill_event(req, error ? error : mangle_poll(mask)); io_commit_cqring(ctx); @@ -4711,7 +4726,7 @@ static int io_poll_double_wake(struct wait_queue_entry *wait, unsigned mode, int sync, void *key) { struct io_kiocb *req = wait->private; - struct io_poll_iocb *poll = req->apoll->double_poll; + struct io_poll_iocb *poll = io_poll_get_single(req); __poll_t mask = key_to_poll(key); /* for instances that support it check for an event match first: */ @@ -4725,6 +4740,8 @@ static int io_poll_double_wake(struct wait_queue_entry *wait, unsigned mode, done = list_empty(&poll->wait.entry); if (!done) list_del_init(&poll->wait.entry); + /* make sure double remove sees this as being gone */ + wait->private = NULL; spin_unlock(&poll->head->lock); if (!done) __io_async_wake(req, poll, mask, io_poll_task_func); @@ -4808,7 +4825,7 @@ static void io_async_task_func(struct callback_head *cb) if (hash_hashed(&req->hash_node)) hash_del(&req->hash_node); - io_poll_remove_double(req, apoll->double_poll); + io_poll_remove_double(req); spin_unlock_irq(&ctx->completion_lock); if (!READ_ONCE(apoll->poll.canceled)) @@ -4919,7 +4936,7 @@ static bool io_arm_poll_handler(struct io_kiocb *req) ret = __io_arm_poll_handler(req, &apoll->poll, &ipt, mask, io_async_wake); if (ret || ipt.error) { - io_poll_remove_double(req, apoll->double_poll); + io_poll_remove_double(req); spin_unlock_irq(&ctx->completion_lock); kfree(apoll->double_poll); kfree(apoll); @@ -4951,14 +4968,13 @@ static bool io_poll_remove_one(struct io_kiocb *req) { bool do_complete; + io_poll_remove_double(req); + if (req->opcode == IORING_OP_POLL_ADD) { - io_poll_remove_double(req, req->io); do_complete = __io_poll_remove_one(req, &req->poll); } else { struct async_poll *apoll = req->apoll; - io_poll_remove_double(req, apoll->double_poll); - /* non-poll requests have submit ref still */ do_complete = __io_poll_remove_one(req, &apoll->poll); if (do_complete) { -- 2.28.0 -- Jens Axboe
