netdev - RE: [PATCH net v3] ravb: Fix races between ravb_tx_timeout

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID:
 <TYBPR01MB5341209D8B6E97F8C0AE9DCFD8B0A@TYBPR01MB5341.jpnprd01.prod.outlook.com>
Date: Thu, 16 Nov 2023 02:43:27 +0000
From: Yoshihiro Shimoda <yoshihiro.shimoda.uh@...esas.com>
To: Sergey Shtylyov <s.shtylyov@....ru>, "davem@...emloft.net"
	<davem@...emloft.net>, "edumazet@...gle.com" <edumazet@...gle.com>,
	"kuba@...nel.org" <kuba@...nel.org>, "pabeni@...hat.com" <pabeni@...hat.com>
CC: "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	"linux-renesas-soc@...r.kernel.org" <linux-renesas-soc@...r.kernel.org>,
	Simon Horman <horms@...nel.org>
Subject: RE: [PATCH net v3] ravb: Fix races between ravb_tx_timeout_work() and
 net related ops

Hello,

> From: Sergey Shtylyov, Sent: Thursday, November 16, 2023 3:37 AM
> 
> On 11/15/23 5:26 AM, Yoshihiro Shimoda wrote:
> 
> > Fix races between ravb_tx_timeout_work() and functions of net_device_ops
> > and ethtool_ops by using rtnl_trylock() and rtnl_unlock(). Note that
> > since ravb_close() is under the rtnl lock and calls cancel_work_sync(),
> > ravb_tx_timeout_work() should calls rtnl_trylock(). Otherwise, a deadlock
> > may happen in ravb_tx_timeout_work() like below:
> >
> > CPU0			CPU1
> > 			ravb_tx_timeout()
> > 			schedule_work()
> > ...
> > __dev_close_many()
> > // Under rtnl lock
> > ravb_close()
> > cancel_work_sync()
> > // Waiting
> > 			ravb_tx_timeout_work()
> > 			rtnl_lock()
> > 			// This is possible to cause a deadlock
> >
> > And, if rtnl_trylock() fails and the netif is still running,
> > rescheduling the work with 1 msec delayed. So, using
> > schedule_delayed_work() instead of schedule_work().
> >
> > Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
> > Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@...esas.com>
> > Reviewed-by: Sergey Shtylyov <s.shtylyov@....ru>
> 
>    Hm, I haven't reviewed this version... :-)

Oops, I should have dropped the tag...

> > Reviewed-by: Simon Horman <horms@...nel.org>
> [...]
> > diff --git a/drivers/net/ethernet/renesas/ravb.h b/drivers/net/ethernet/renesas/ravb.h
> > index e0f8276cffed..e9bb8ee3ba2d 100644
> > --- a/drivers/net/ethernet/renesas/ravb.h
> > +++ b/drivers/net/ethernet/renesas/ravb.h
> > @@ -1081,7 +1081,7 @@ struct ravb_private {
> >  	u32 cur_tx[NUM_TX_QUEUE];
> >  	u32 dirty_tx[NUM_TX_QUEUE];
> >  	struct napi_struct napi[NUM_RX_QUEUE];
> > -	struct work_struct work;
> > +	struct delayed_work work;
> 
>    Not sure this is justified...
>    Then again, what do I know about workqueues? Not much... :-)

I thought that the schedule_work() called the work function immediately.
So, I thought call*ing the schedule_work() from the work function caused
endless loop. However, it is not true. The schedule_work() just inserts
a work queue, and then the kernel calls the work function later.

So, changing from work_struct to delayed_work is not needed for fixing
this issue, I think now. However, I have another concern about rescheduling
this work by schedule_work() here because it's possible to cause high CPU load
while the rtnl_lock() is held. So, I think we should call a sleep function
like usleep_range(1000, 2000) for instance before schedule_work().
But, what do you think?

> [...]
> > diff --git a/drivers/net/ethernet/renesas/ravb_main.c b/drivers/net/ethernet/renesas/ravb_main.c
> > index c70cff80cc99..ca7db8a5b412 100644
> > --- a/drivers/net/ethernet/renesas/ravb_main.c
> > +++ b/drivers/net/ethernet/renesas/ravb_main.c
> > @@ -1863,17 +1863,24 @@ static void ravb_tx_timeout(struct net_device *ndev, unsigned int txqueue)
> >  	/* tx_errors count up */
> >  	ndev->stats.tx_errors++;
> >
> > -	schedule_work(&priv->work);
> > +	schedule_delayed_work(&priv->work, 0);
> >  }
> >
> >  static void ravb_tx_timeout_work(struct work_struct *work)
> >  {
> > -	struct ravb_private *priv = container_of(work, struct ravb_private,
> > +	struct delayed_work *dwork = to_delayed_work(work);
> > +	struct ravb_private *priv = container_of(dwork, struct ravb_private,
> >  						 work);
> >  	const struct ravb_hw_info *info = priv->info;
> >  	struct net_device *ndev = priv->ndev;
> >  	int error;
> >
> > +	if (!rtnl_trylock()) {
> > +		if (netif_running(ndev))
> > +			schedule_delayed_work(&priv->work, msecs_to_jiffies(10));
> 
>    The delay is rather arbitrary. Why not e.g. 1 ms?

I think that 1 ms is enough.

Best regards,
Yoshihiro Shimoda

> [...]
> 
> MBR, Sergey