Re: [PATCH] block, bfq: fix possible uaf for 'bfqq->bic'

[Jan Kara]

On Sat 10-12-22 18:25:37, Yu Kuai wrote:
> From: Yu Kuai <yukuai3@xxxxxxxxxx>
> 
> Our test report a uaf for 'bfqq->bic' in 5.10:
> 
> ==================================================================
> BUG: KASAN: use-after-free in bfq_select_queue+0x378/0xa30
> Read of size 8 at addr ffff88810efb42d8 by task fsstress/2318352
> 
> CPU: 6 PID: 2318352 Comm: fsstress Kdump: loaded Not tainted 5.10.0-60.18.0.50.h602.kasan.eulerosv2r11.x86_64 #1
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.1-0-ga5cab58-20220320_160524-szxrtosci10000 04/01/2014
> Call Trace:
...
>  bfq_select_queue+0x378/0xa30
>  __bfq_dispatch_request+0x1c4/0x220
>  bfq_dispatch_request+0xe8/0x130
>  __blk_mq_do_dispatch_sched+0x3f4/0x560
>  blk_mq_do_dispatch_sched+0x62/0xb0
>  __blk_mq_sched_dispatch_requests+0x215/0x2a0
>  blk_mq_sched_dispatch_requests+0x8f/0xd0
>  __blk_mq_run_hw_queue+0x98/0x180
>  __blk_mq_delay_run_hw_queue+0x22b/0x240
>  blk_mq_run_hw_queue+0xe3/0x190
>  blk_mq_sched_insert_requests+0x107/0x200
>  blk_mq_flush_plug_list+0x26e/0x3c0
>  blk_finish_plug+0x63/0x90
>  __iomap_dio_rw+0x7b5/0x910
>  iomap_dio_rw+0x36/0x80
>  ext4_dio_read_iter+0x146/0x190 [ext4]
>  ext4_file_read_iter+0x1e2/0x230 [ext4]
>  new_sync_read+0x29f/0x400
>  vfs_read+0x24e/0x2d0
>  ksys_read+0xd5/0x1b0

Perhaps we can trim this UAF report a bit to what I've left above? That
should be enough to give idea about the problem.

> Commit 3bc5e683c67d ("bfq: Split shared queues on move between cgroups")
> changes that move process to a new cgroup will allocate a new bfqq to
> use, however, the old bfqq and new bfqq can point to the same bic:
> 
> 1) Initial state, two process with io in the same cgroup.
> 
> Process 1       Process 2
>  (BIC1)          (BIC2)
>   |  Λ            |  Λ
>   |  |            |  |
>   V  |            V  |
>   bfqq1           bfqq2
> 
> 2) bfqq1 is merged to bfqq2.
> 
> Process 1       Process 2（cg1)
>  (BIC1)          (BIC2)
>   |               |
>    \-------------\|
>                   V
>   bfqq1           bfqq2(coop)
> 
> 3) Process 1 exit, then issue new io(denoce IOA) from Process 2.
> 
>  (BIC2)
>   |  Λ
>   |  |
>   V  |
>   bfqq2(coop)
> 
> 4) Before IOA is completed, move Process 2 to another cgroup and issue io.
> 
> Process 2
>  (BIC2)
>    Λ
>    |\--------------\
>    |                V
>   bfqq2           bfqq3
> 
> Now that BIC2 points to bfqq3, while bfqq2 and bfqq3 both point to BIC2.
> If all the requests are completed, and Process 2 exit, BIC2 will be
> freed while there is no guarantee that bfqq2 will be freed before BIC2.
> 
> Fix the problem by clearing bfqq->bic if process references is decreased
> to zero, since that they are not related anymore.
> 
> Fixes: 3bc5e683c67d ("bfq: Split shared queues on move between cgroups")
> Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>

Thanks for the analysis and the patch! I agree this is a problem.

> diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
> index a72304c728fc..6eada99d1b34 100644
> --- a/block/bfq-iosched.c
> +++ b/block/bfq-iosched.c
> @@ -3036,6 +3036,14 @@ void bfq_release_process_ref(struct bfq_data *bfqd, struct bfq_queue *bfqq)
>  
>  	bfq_reassign_last_bfqq(bfqq, NULL);
>  
> +	/*
> +	 * __bfq_bic_change_cgroup() just reset bic->bfqq so that a new bfqq
> +	 * will be created to handle new io, while old bfqq will stay around
> +	 * until all the requests are completed. It's unsafe to keep bfqq->bic
> +	 * since they are not related anymore.
> +	 */
> +	if (bfqq_process_refs(bfqq) == 1)
> +		bfqq->bic = NULL;
>  	bfq_put_queue(bfqq);

Rather than changing bfq_release_process_ref() I think it would be more
logical to change bic_set_bfqq() like:

	struct bfq_queue *old_bfqq = bic->bfqq[is_sync];

	/* Clear bic pointer if we are detaching bfqq from its bic */
	if (old_bfqq && old_bfqq->bic == bic)
		old_bfqq->bic = NULL;

And then we can also remove several explicit bfqq->bic = NULL statements
from bfq code.

								Honza
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR