diff mbox series

USB: usbfs: Suppress emission of uevents for interfaces handled via usbfs

Message ID 20191009123829.07eacc7f@ingpc3.intern.lauterbach.com (mailing list archive)
State New, archived
Headers show
Series USB: usbfs: Suppress emission of uevents for interfaces handled via usbfs | expand

Commit Message

Ingo Rohloff Oct. 9, 2019, 10:38 a.m. UTC
From 17d1e75543e26cfe702e7f5b0d4e07e0e45e5250 Mon Sep 17 00:00:00 2001
From: Ingo Rohloff <ingo.rohloff@lauterbach.com>
Date: Tue, 8 Oct 2019 20:27:57 +0200
Subject: [PATCH] USB: usbfs: Suppress emission of uevents for interfaces
 handled via usbfs.

commit 1455cf8dbfd0
("driver core: emit uevents when device is bound to a driver")
added bind/unbind uevents when a driver is bound/unbound
to a physical device.

For USB devices which are handled via the generic usbfs layer
(via libusb for example), this is problematic:
Each time a user space program calls
   ioctl(usb_fd, USBDEVFS_CLAIMINTERFACE, &usb_intf_nr);
and then later
   ioctl(usb_fd, USBDEVFS_RELEASEINTERFACE, &usb_intf_nr);
The kernel will now produce a bind/unbind event,
which does not really contain any useful information.

This allows a user space program to run a DoS attack against
programs which listen to uevents (in particular systemd/eudev/upowerd):
A malicious user space program just has to call in a tight loop

   ioctl(usb_fd, USBDEVFS_CLAIMINTERFACE, &usb_intf_nr);
   ioctl(usb_fd, USBDEVFS_RELEASEINTERFACE, &usb_intf_nr);

With this loop the malicious user space program floods
the kernel and all programs listening to uevents with
tons of bind/unbind events.

This patch suppresses uevents for interfaces claimed via usbfs.

Signed-off-by: Ingo Rohloff <ingo.rohloff@lauterbach.com>
---
 drivers/usb/core/devio.c  | 7 ++++++-
 drivers/usb/core/driver.c | 2 ++
 2 files changed, 8 insertions(+), 1 deletion(-)

Comments

Greg Kroah-Hartman Oct. 10, 2019, 10:24 a.m. UTC | #1
On Wed, Oct 09, 2019 at 12:38:35PM +0200, Ingo Rohloff wrote:
> >From 17d1e75543e26cfe702e7f5b0d4e07e0e45e5250 Mon Sep 17 00:00:00 2001
> From: Ingo Rohloff <ingo.rohloff@lauterbach.com>
> Date: Tue, 8 Oct 2019 20:27:57 +0200
> Subject: [PATCH] USB: usbfs: Suppress emission of uevents for interfaces
>  handled via usbfs.

No need for this in the changelog body :)

> commit 1455cf8dbfd0
> ("driver core: emit uevents when device is bound to a driver")
> added bind/unbind uevents when a driver is bound/unbound
> to a physical device.

You can wrap the line a bit nicer:
	commit 1455cf8dbfd0 ("driver core: emit uevents when device is bound to
	a driver") added bind/unbind uevents when a driver is bound/unbound to a
	physical device.

> For USB devices which are handled via the generic usbfs layer
> (via libusb for example), this is problematic:
> Each time a user space program calls
>    ioctl(usb_fd, USBDEVFS_CLAIMINTERFACE, &usb_intf_nr);
> and then later
>    ioctl(usb_fd, USBDEVFS_RELEASEINTERFACE, &usb_intf_nr);
> The kernel will now produce a bind/unbind event,
> which does not really contain any useful information.
> 
> This allows a user space program to run a DoS attack against
> programs which listen to uevents (in particular systemd/eudev/upowerd):
> A malicious user space program just has to call in a tight loop
> 
>    ioctl(usb_fd, USBDEVFS_CLAIMINTERFACE, &usb_intf_nr);
>    ioctl(usb_fd, USBDEVFS_RELEASEINTERFACE, &usb_intf_nr);
> 
> With this loop the malicious user space program floods
> the kernel and all programs listening to uevents with
> tons of bind/unbind events.
> 
> This patch suppresses uevents for interfaces claimed via usbfs.
> 
> Signed-off-by: Ingo Rohloff <ingo.rohloff@lauterbach.com>
> ---
>  drivers/usb/core/devio.c  | 7 ++++++-
>  drivers/usb/core/driver.c | 2 ++
>  2 files changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/usb/core/devio.c b/drivers/usb/core/devio.c
> index 3f899552f6e3..a1af1d9b2ae7 100644
> --- a/drivers/usb/core/devio.c
> +++ b/drivers/usb/core/devio.c
> @@ -764,8 +764,13 @@ static int claimintf(struct usb_dev_state *ps, unsigned int ifnum)
>  	intf = usb_ifnum_to_if(dev, ifnum);
>  	if (!intf)
>  		err = -ENOENT;
> -	else
> +	else {
> +		/* suppress uevents for devices handled by usbfs */
> +		dev_set_uevent_suppress(&intf->dev, 1);
>  		err = usb_driver_claim_interface(&usbfs_driver, intf, ps);
> +		if (err != 0)

Did checkpatch let this go through?  Shouldn't that be:
		if (err)

And did you send this patch twice?

Anyway, if you fix those minor things up, it looks good to me.

thanks,

greg k-h
Ingo Rohloff Oct. 10, 2019, 12:53 p.m. UTC | #2
Hello Greg

> > +	else {
> > +		/* suppress uevents for devices handled by usbfs */
> > +		dev_set_uevent_suppress(&intf->dev, 1);
> >  		err = usb_driver_claim_interface(&usbfs_driver, intf, ps);
> > +		if (err != 0)  
> > +			dev_set_uevent_suppress(&intf->dev, 0);

> Did checkpatch let this go through?  Shouldn't that be:
> 		if (err)

I actually wanted it the way it is, but it really might not be the best option.
Let me explain:
The main goal was to suppress bind/unbind uevents produced by libusb
or any other user space program which calls
ioctl USBDEVFS_CLAIMINTERFACE/USBDEVFS_RELEASEINTERFACE .

Now I can suppress uevents produced by usb_driver_claim_interface
with the code above.
But I was not sure how to handle the call to usb_driver_release_interface 
from devio.c/releaseintf()

The strategy I used was: 
1) Set suppression of uevents when user space program tries to claim interface
2) If claiming the interface works, then KEEP uevents suppressed,
   otherwise undo suppression.
   That's why its "if err !=0"; error happened => undo suppression.
3) When interface is released make sure suppression is undone AFTER unbinding the driver.

Thinking about your comment: It might be better + simpler to just use
1) Suppress uevents when calling usb_driver_claim_interface. Undo suppression right after the call.
2) Suppress uevents when calling usb_driver_release_interface. Undo suppression right after the call.

The main semantic problem I do not know about: 
Is it correct to modify uevent suppression of an USB interface device 
even if it CANNOT be claimed by usbfs ?
I grepped the source code for usage of dev_set_uevent_suppress, but it seems not to be 100%
clear how that should be used (sometimes uevents are only suppressed temporarily to implement
a delay, sometimes they are actually kept suppressed).

I will prepare/send an alternative.

with best regards
  Ingo

PS:
> ...
> No need for this in the changelog body :)

I should have read the documentation about how to send correct E-Mails for patches more intensively.
I just found out about "git send-email" and had not set it up (did now...). I am sorry.

> And did you send this patch twice?

Unfortunately yes: I was struggling how to format this correctly.
diff mbox series

Patch

diff --git a/drivers/usb/core/devio.c b/drivers/usb/core/devio.c
index 3f899552f6e3..a1af1d9b2ae7 100644
--- a/drivers/usb/core/devio.c
+++ b/drivers/usb/core/devio.c
@@ -764,8 +764,13 @@  static int claimintf(struct usb_dev_state *ps, unsigned int ifnum)
 	intf = usb_ifnum_to_if(dev, ifnum);
 	if (!intf)
 		err = -ENOENT;
-	else
+	else {
+		/* suppress uevents for devices handled by usbfs */
+		dev_set_uevent_suppress(&intf->dev, 1);
 		err = usb_driver_claim_interface(&usbfs_driver, intf, ps);
+		if (err != 0)
+			dev_set_uevent_suppress(&intf->dev, 0);
+	}
 	if (err == 0)
 		set_bit(ifnum, &ps->ifclaimed);
 	return err;
diff --git a/drivers/usb/core/driver.c b/drivers/usb/core/driver.c
index 2b27d232d7a7..6a15bc5c2869 100644
--- a/drivers/usb/core/driver.c
+++ b/drivers/usb/core/driver.c
@@ -594,6 +594,8 @@  void usb_driver_release_interface(struct usb_driver *driver,
 	 */
 	if (device_is_registered(dev)) {
 		device_release_driver(dev);
+		/* make sure we allow uevents again */
+		dev_set_uevent_suppress(dev, 0);
 	} else {
 		device_lock(dev);
 		usb_unbind_interface(dev);