From: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
To: Luiz Augusto von Dentz <luiz.dentz@gmail.com>
Cc: linux-bluetooth@vger.kernel.org, LinMa <linma@zju.edu.cn>
Subject: RE: [RFC] Bluetooth: hci_sock: Fix calling lock_sock when handling HCI_DEV_UNREG
Date: Sun, 18 Jul 2021 11:22:09 +0900 [thread overview]
Message-ID: <7a0405c7-63ba-3e5c-7b22-503d7b326a11@i-love.sakura.ne.jp> (raw)
In-Reply-To: <20210717000731.3836303-1-luiz.dentz@gmail.com>
Luiz Augusto von Dentz wrote:
> This removes the reference of hci_dev to hci_pinfo since the reference
> cannot prevent hdev to be freed hci_pinfo only keeps the index so in
> case the device is unregistered and freed hci_dev_get will return NULL
> thus prevent UAF.
I'm not convinced that this change is safe.
vhci_release() (which will call hci_unregister_dev(hdev)) is called when
refcount to /dev/vchi dropped to 0. That is, vhci_release() might be called
while e.g. hci_sock_bound_ioctl() is in progress.
Since hci_unregister_dev(hdev) calls list_del(&hdev->list) with hci_dev_list_lock
held for write, hci_dev_get(hci_pi(sk)->dev) which scans hci_dev_list with
hci_dev_list_lock held for read will start returning NULL. But I think that
this change introduces two race windows.
Since hci_unregister_dev(hdev) then calls hci_sock_dev_event(hdev, HCI_DEV_UNREG)
and finally calls ida_simple_remove(&hci_index_ida, id), subsequent
hci_register_dev(struct hci_dev *hdev) call can re-assign the id which
hci_pi(sk)->dev is tracking, by calling ida_simple_get() and finally calling
list_add(&hdev->list, &hci_dev_list) with hci_dev_list_lock held for write.
Therefore, I think that first race window is that
+ /* Commands may use copy_from_user which is unsafe while holding hdev as
+ * it could be unregistered in the meantime.
+ */
+ hci_dev_put(hdev);
+ hdev = NULL;
causes hci_sock_bound_ioctl() to check flags on an intended hdev and
e.g. hci_sock_reject_list_add() to operate on an unintended hdev.
Also, even if hci_sock_bound_ioctl() and hci_sock_reject_list_add() reached
the same hdev, I think that there still is second race window that
hci_unregister_dev() { hci_sock_reject_list_add() {
hdev = hci_dev_get(hci_pi(sk)->dev);
write_lock(&hci_dev_list_lock);
list_del(&hdev->list);
write_unlock(&hci_dev_list_lock);
hci_sock_dev_event(hdev, HCI_DEV_UNREG);
hci_dev_lock(hdev);
hci_bdaddr_list_clear(&hdev->reject_list);
hci_dev_unlock(hdev);
hci_dev_lock(hdev);
err = hci_bdaddr_list_add(&hdev->reject_list, &bdaddr, BDADDR_BREDR); // <= Adding after clear all; at least memory leak.
hci_dev_unlock(hdev);
hci_dev_put(hdev);
}
. That is, an attempt to replace pointer reference with index number is racy.
After all, hci_sock_dev_event(hdev, HCI_DEV_UNREG) is responsible for
waiting for already started e.g. hci_sock_reject_list_add().
next prev parent reply other threads:[~2021-07-18 2:22 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-17 0:07 [RFC] Bluetooth: hci_sock: Fix calling lock_sock when handling HCI_DEV_UNREG Luiz Augusto von Dentz
2021-07-17 2:04 ` bluez.test.bot
2021-07-18 2:22 ` Tetsuo Handa [this message]
2021-07-18 5:16 ` Luiz Augusto von Dentz
2021-07-18 6:22 ` Tetsuo Handa
2021-07-18 6:51 ` Luiz Augusto von Dentz
2021-07-18 14:56 ` Tetsuo Handa
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=7a0405c7-63ba-3e5c-7b22-503d7b326a11@i-love.sakura.ne.jp \
--to=penguin-kernel@i-love.sakura.ne.jp \
--cc=linma@zju.edu.cn \
--cc=linux-bluetooth@vger.kernel.org \
--cc=luiz.dentz@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).