From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
To: Sean Young <sean@mess.org>
Cc: Alan Stern <stern@rowland.harvard.edu>,
Mauro Carvalho Chehab <mchehab@kernel.org>,
Jarod Wilson <jarod@redhat.com>,
syzbot <syzbot+c558267ad910fc494497@syzkaller.appspotmail.com>,
andreyknvl@google.com, linux-media@vger.kernel.org,
linux-usb@vger.kernel.org, syzkaller-bugs@googlegroups.com
Subject: Re: [PATCH v2 (resend)] media: imon: reorganize serialization
Date: Mon, 2 May 2022 19:26:30 +0900 [thread overview]
Message-ID: <e144d930-7b82-3310-1163-dae9f7914fa9@I-love.SAKURA.ne.jp> (raw)
In-Reply-To: <Ym+myt+bqap21r0O@gofer.mess.org>
On 2022/05/02 18:39, Sean Young wrote:
> So this part of the patch address the issue of driver_lock being
> unnecessary. This should be in its own patch, so I am going to merge:
>
> https://patchwork.linuxtv.org/project/linux-media/patch/349f3e34-41ed-f832-3b22-ae10c50e3868@I-love.SAKURA.ne.jp/
driver_lock is not unnecessary, unless combined with kfree_rcu() approach.
>> + /*
>> + * We need to wait for RCU grace period in order to allow
>> + * display_open() to safely check ->disconnected and increment ->users.
>> + */
>> + struct rcu_head rcu;
>
> Is it possible to modify the users/disconnected fields while holding the
> lock mutex in imon_context? This would make it unnecessary to use rcu and
> simplify the code.
I don't think it is possible, and I don't think it simplifies the code.
Unless we revive global driver_lock lock (untested delta diff is shown below),
nothing prevents from calling kfree(ictx) before holding ictx->lock or
checking ->disconnected or incrementing ->users.
As a whole, I think kfree_rcu() is simpler while removing global lock.
diff --git a/drivers/media/rc/imon.c b/drivers/media/rc/imon.c
index 9a4f24e294bc..469a2f869572 100644
--- a/drivers/media/rc/imon.c
+++ b/drivers/media/rc/imon.c
@@ -166,11 +166,6 @@ struct imon_context {
* imon_disconnect() was already called.
*/
bool disconnected;
- /*
- * We need to wait for RCU grace period in order to allow
- * display_open() to safely check ->disconnected and increment ->users.
- */
- struct rcu_head rcu;
};
#define TOUCH_TIMEOUT (HZ/30)
@@ -457,6 +452,9 @@ static struct usb_driver imon_driver = {
.id_table = imon_usb_id_table,
};
+/* to prevent races between open() and kfree() */
+static DEFINE_MUTEX(driver_lock);
+
/* Module bookkeeping bits */
MODULE_AUTHOR(MOD_AUTHOR);
MODULE_DESCRIPTION(MOD_DESC);
@@ -516,6 +514,8 @@ static int display_open(struct inode *inode, struct file *file)
int subminor;
int retval = 0;
+ mutex_lock(&driver_lock);
+
subminor = iminor(inode);
interface = usb_find_interface(&imon_driver, subminor);
if (!interface) {
@@ -524,15 +524,13 @@ static int display_open(struct inode *inode, struct file *file)
goto exit;
}
- rcu_read_lock();
ictx = usb_get_intfdata(interface);
if (!ictx || ictx->disconnected || !refcount_inc_not_zero(&ictx->users)) {
- rcu_read_unlock();
pr_err("no context found for minor %d\n", subminor);
retval = -ENODEV;
goto exit;
}
- rcu_read_unlock();
+ mutex_unlock(&driver_lock);
mutex_lock(&ictx->lock);
@@ -550,10 +548,15 @@ static int display_open(struct inode *inode, struct file *file)
mutex_unlock(&ictx->lock);
- if (retval && refcount_dec_and_test(&ictx->users))
- free_imon_context(ictx);
-
+ if (retval) {
+ mutex_unlock(&driver_lock);
+ if (refcount_dec_and_test(&ictx->users))
+ free_imon_context(ictx);
+ mutex_unlock(&driver_lock);
+ }
+ return retval;
exit:
+ mutex_unlock(&driver_lock);
return retval;
}
@@ -2497,6 +2500,8 @@ static void imon_disconnect(struct usb_interface *interface)
struct device *dev;
int ifnum;
+ mutex_lock(&driver_lock);
+
ictx = usb_get_intfdata(interface);
ictx->disconnected = true;
dev = ictx->dev;
@@ -2542,6 +2547,8 @@ static void imon_disconnect(struct usb_interface *interface)
if (refcount_dec_and_test(&ictx->users))
free_imon_context(ictx);
+ mutex_unlock(&driver_lock);
+
dev_dbg(dev, "%s: iMON device (intf%d) disconnected\n",
__func__, ifnum);
}
next prev parent reply other threads:[~2022-05-02 10:28 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-09 13:18 possible deadlock in display_open syzbot
2022-04-25 5:29 ` Tetsuo Handa
2022-04-25 9:20 ` Sean Young
2022-04-25 11:14 ` Tetsuo Handa
2022-04-25 11:56 ` Sean Young
2022-04-25 16:01 ` Alan Stern
2022-04-25 16:21 ` Sean Young
2022-04-26 0:54 ` [PATCH] media: imon: remove redundant serialization Tetsuo Handa
2022-04-26 7:57 ` Sean Young
2022-04-26 10:32 ` [PATCH v2] media: imon: reorganize serialization Tetsuo Handa
2022-05-02 3:49 ` [PATCH v2 (resend)] " Tetsuo Handa
2022-05-02 9:39 ` Sean Young
2022-05-02 10:26 ` Tetsuo Handa [this message]
2022-05-02 10:40 ` Tetsuo Handa
2022-05-02 10:46 ` Oliver Neukum
2022-05-02 11:05 ` Tetsuo Handa
2022-05-03 7:41 ` Oliver Neukum
2022-05-02 12:53 ` Greg KH
2022-05-02 13:04 ` Tetsuo Handa
2022-05-02 13:06 ` Greg KH
2022-05-02 13:40 ` 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=e144d930-7b82-3310-1163-dae9f7914fa9@I-love.SAKURA.ne.jp \
--to=penguin-kernel@i-love.sakura.ne.jp \
--cc=andreyknvl@google.com \
--cc=jarod@redhat.com \
--cc=linux-media@vger.kernel.org \
--cc=linux-usb@vger.kernel.org \
--cc=mchehab@kernel.org \
--cc=sean@mess.org \
--cc=stern@rowland.harvard.edu \
--cc=syzbot+c558267ad910fc494497@syzkaller.appspotmail.com \
--cc=syzkaller-bugs@googlegroups.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).