netdev - Re: [PATCH 1/2 v5] usbnet: allow status interrupt URB to always be active

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: <1436202.ggQbKKzqIM@linux-5eaq.site>
Date:	Fri, 19 Apr 2013 10:25:04 +0200
From:	Oliver Neukum <oliver@...kum.org>
To:	Ming Lei <tom.leiming@...il.com>
Cc:	Dan Williams <dcbw@...hat.com>,
	Elina Pasheva <epasheva@...rrawireless.com>,
	Network Development <netdev@...r.kernel.org>,
	linux-usb <linux-usb@...r.kernel.org>,
	Rory Filer <rfiler@...rrawireless.com>,
	Phil Sutter <phil@....cc>,
	Alan Stern <stern@...land.harvard.edu>
Subject: Re: [PATCH 1/2 v5] usbnet: allow status interrupt URB to always be active

On Thursday 18 April 2013 11:51:25 Ming Lei wrote:
> On Wed, Apr 17, 2013 at 2:55 PM, Oliver Neukum <oliver@...kum.org> wrote:
>
> > If we have a function for starting a status URB we want to use it whenever
> > it applies, that is also when we need to poll status for internal reason while
> > an interface is up.
> 
> For other non-sierra usbnet devices, when an interface is up, the status URB
> is scheduled in open() and needn't the API.

But that is the very point. This API is used from _within_ open.
We cannot make every open() use GFP_NOIO

> > You don't need to understand it any more than you need to understand
> > the rule for usb_submit_urb(). The rules are the very same. There is no
> > special effort here.
> 
> No, there isn't one rule for the corner case here, and the corner case should
> have existed in probe() or cancel queue with reset of all USB drivers, instead
> of usbnet only.

The same rule applies to usb_submit_urb(), too.
 
> Also the rule 3 is a bit obscure, maybe not correct, at least there are much
> GFP_KERNEL usages in kthread of usbnet. I am wondering if there are
> other usbnet specific memflags rules except for 1 & 6.
> 
> Strictly speaking, the rule 5 isn't correct, since it might trigger the corner
> case you mentioned, right?
> 
> I think we need to review the memflags part of usb_submit_urb() doc now.

Yes.

> +int usbnet_status_start(struct usbnet *dev, gfp_t mem_flags)
> +{
> +       int ret = 0;
> +
> +       WARN_ON_ONCE(dev->interrupt == NULL);
> +       if (dev->interrupt) {
> +               mutex_lock(&dev->interrupt_mutex);
> 
> ....
> 
> +}
> 
> Obviously, the API can't be called in atomic context, and putting runtime PM
> inside is reasonable and correct.

Yes, but how is it relevant. What allows us to conclude that a driver does not want
runtime PM while a status URB is running?

> > I meant block suspend in the sense of disallowing it. Which is very problematic.
> > The CDC protocols generally support remote wakeup for status information,
> > so we need to be able to sleep while status is running to accomodate devices
> > which are intended to be always online.
> 
> At least now, for non-sierra drivers, it needn't the API to schedule status URB
> which will be started in normal open() path, so won't power on device runtime
> unnecessarily. That is what I say the API shouldn't be for a general usage, :-)

But many drivers, e.g. cdc-ether, cdc-ncm and cdc-mbim want to be able
to runtime suspend while the device is open.

	Regards
		Oliver

--
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