[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 07 Jul 2008 15:57:23 -0700 (PDT)
From: David Miller <davem@...emloft.net>
To: rdreier@...co.com
Cc: shemminger@...tta.com, arjan@...radead.org, netdev@...r.kernel.org
Subject: Re: Printing the driver name as part of the netdev watchdog message
From: Roland Dreier <rdreier@...co.com>
Date: Mon, 07 Jul 2008 15:45:38 -0700
> > > For consistency each call through ethtool_ops should
> > > be holding rtnl mutex. And since dev_watchdog is a timer routine,
> > > it is not safe to acquire a mutex there.
> >
> > I think "get driver info", which does nothing but copy strings out of
> > the driver software state, is a safe exception to these strict locking
> > rules.
>
> it seems that's not universally the case, eg get_drvinfo() in
> drivers/net/cxgb3/cxgb3_main.c does
>
> t3_get_fw_version(adapter, &fw_vers);
> t3_get_tp_version(adapter, &tp_vers);
>
> and from reading further in the code it's not obvious that this is safe
> to do without serialization (it calls into t3_read_flash(), which does a
> typical "write address reg, read data reg" sequence to access flash,
> without any locking).
I doubt it uses the RTNL semaphore elsewhere to protect
against this path, which is the only protection these
calls currently have.
Please don't bring up scarecrows, this looks like simply
a bug which already exists.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists