Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Subject: Re: [PATCH RFC] block: fix use after free for bd_holder_dir/slave_dir
To:     Christoph Hellwig <hch@infradead.org>,
        Yu Kuai <yukuai1@huaweicloud.com>
Cc:     axboe@kernel.dk, gregkh@linuxfoundation.org, willy@infradead.org,
        kch@nvidia.com, martin.petersen@oracle.com,
        johannes.thumshirn@wdc.com, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org, yi.zhang@huawei.com,
        "yukuai3@huawei.com >> yukuai (C)" <yukuai3@huawei.com>
References: <20221012125838.1608619-1-yukuai1@huaweicloud.com>
 <Y00LTH0yk3obS22m@infradead.org>
 <dadef634-9ec3-aa7b-f1c0-65a98a0fafb1@huaweicloud.com>
 <Y00+SsxI7IJ56im8@infradead.org>
From:   Yu Kuai <yukuai1@huaweicloud.com>
Message-ID: <21375b39-234a-cffd-40db-f78fae3aee60@huaweicloud.com>
Date:   Mon, 17 Oct 2022 21:17:03 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:60.0) Gecko/20100101
 Thunderbird/60.8.0
MIME-Version: 1.0
In-Reply-To: <Y00+SsxI7IJ56im8@infradead.org>
Content-Type: text/plain; charset=gbk; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

Hi,

?? 2022/10/17 19:36, Christoph Hellwig д??:
> On Mon, Oct 17, 2022 at 05:24:20PM +0800, Yu Kuai wrote:
>> I agree that this patch do make sense, however, it seems to me this
>> patch should fix the problem that kobject is leaked, not uaf... And I
>> verified that the problem can still be reporduced with this patch.
> 
> Can you share your reproducer?
> .
> 

Of course, I add some delay in kernel to make sure the problem is
reproduced 100%:

diff --git a/block/holder.c b/block/holder.c
index 5283bc804cc1..be8f4b4245f3 100644
--- a/block/holder.c
+++ b/block/holder.c
@@ -1,6 +1,7 @@
  // SPDX-License-Identifier: GPL-2.0-only
  #include <linux/blkdev.h>
  #include <linux/slab.h>
+#include <linux/delay.h>

  struct bd_holder_disk {
         struct list_head        list;
@@ -33,6 +34,8 @@ static int __link_disk_holder(struct block_device 
*bdev, struct gendisk *disk)
  {
         int ret;

+       printk("%s: delay 5s\n", __func__);
+       msleep(5000);
         ret = add_symlink(disk->slave_dir, bdev_kobj(bdev));
         if (ret)
                 return ret;

test cmd is very simple:

dmsetup create test1 --table "0 100000 linear /dev/sda 0" &
sleep 1
echo 1 > /sys/block/sda/device/delete