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: <4c64791e-a2f0-4878-8025-4ced94da6bc3@rowland.harvard.edu>
Date: Wed, 28 Aug 2024 10:54:18 -0400
From: Alan Stern <stern@...land.harvard.edu>
To: Greg KH <gregkh@...uxfoundation.org>
Cc: Selvarasu Ganesan <selvarasu.g@...sung.com>, royluo@...gle.com,
	paul@...pouillou.net, elder@...nel.org, yuanlinyu@...onor.com,
	quic_kriskura@...cinc.com, crwulff@...il.com,
	linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org,
	jh0801.jung@...sung.com, dh10.jung@...sung.com, naushad@...sung.com,
	akash.m5@...sung.com, rc93.raju@...sung.com,
	taehyun.cho@...sung.com, hongpooh.kim@...sung.com,
	eomji.oh@...sung.com, shijie.cai@...sung.com,
	stable <stable@...nel.org>
Subject: Re: [PATCH] usb: gadget: udc: Add null pointer check for udc in
 gadget_match_driver

On Wed, Aug 28, 2024 at 11:39:58AM +0200, Greg KH wrote:
> On Wed, Aug 28, 2024 at 12:35:04PM +0530, Selvarasu Ganesan wrote:
> > This commit adds a null pointer check for udc in gadget_match_driver to
> > prevent the below potential dangling pointer access. The issue arises
> > due to continuous USB role switch and simultaneous UDC write operations
> > performed by init.rc from user space through configfs.  In these
> > scenarios, there was a possibility of usb_udc_release being done before
> > gadget_match_driver.
> > 
> > [27635.233849]  BUG: KASAN: invalid-access in gadget_match_driver+0x40/0x94
> > [27635.233871]  Read of size 8 at addr d7ffff8837ead080 by task init/1
> > [27635.233881]  Pointer tag: [d7], memory tag: [fe]
> > [27635.233888]
> > [27635.233917]  Call trace:
> > [27635.233923]   dump_backtrace+0xec/0x10c
> > [27635.233935]   show_stack+0x18/0x24
> > [27635.233944]   dump_stack_lvl+0x50/0x6c
> > [27635.233958]   print_report+0x150/0x6b4
> > [27635.233977]   kasan_report+0xe8/0x148
> > [27635.233985]   __hwasan_load8_noabort+0x88/0x98
> > [27635.233995]   gadget_match_driver+0x40/0x94
> > [27635.234005]   __driver_attach+0x60/0x304
> > [27635.234018]   bus_for_each_dev+0x154/0x1b4
> > [27635.234027]   driver_attach+0x34/0x48
> > [27635.234036]   bus_add_driver+0x1ec/0x310
> > [27635.234045]   driver_register+0xc8/0x1b4
> > [27635.234055]   usb_gadget_register_driver_owner+0x7c/0x140
> > [27635.234066]   gadget_dev_desc_UDC_store+0x148/0x19c
> > [27635.234075]   configfs_write_iter+0x180/0x1e0
> > [27635.234087]   vfs_write+0x298/0x3e4
> > [27635.234105]   ksys_write+0x88/0x100
> > [27635.234115]   __arm64_sys_write+0x44/0x5c
> > [27635.234126]   invoke_syscall+0x6c/0x17c
> > [27635.234143]   el0_svc_common+0xf8/0x138
> > [27635.234154]   do_el0_svc+0x30/0x40
> > [27635.234164]   el0_svc+0x38/0x68
> > [27635.234174]   el0t_64_sync_handler+0x68/0xbc
> > [27635.234184]   el0t_64_sync+0x19c/0x1a0
> > 
> > Fixes: fc274c1e9973 ("USB: gadget: Add a new bus for gadgets")
> > Cc: stable <stable@...nel.org>
> > Signed-off-by: Selvarasu Ganesan <selvarasu.g@...sung.com>
> > ---
> >  drivers/usb/gadget/udc/core.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/usb/gadget/udc/core.c b/drivers/usb/gadget/udc/core.c
> > index cf6478f97f4a..77dc0f28ff01 100644
> > --- a/drivers/usb/gadget/udc/core.c
> > +++ b/drivers/usb/gadget/udc/core.c
> > @@ -1338,6 +1338,7 @@ static void usb_udc_release(struct device *dev)
> >  	udc = container_of(dev, struct usb_udc, dev);
> >  	dev_dbg(dev, "releasing '%s'\n", dev_name(dev));
> >  	kfree(udc);
> > +	udc = NULL;
> 
> That's not ok, as what happens if you race right between freeing it and
> accessing it elsewhere?

In fact, this assignment does nothing at all.  This is at the end of the 
function and udc is a local variable, so it's not going to be used 
again.  The compiler won't even generate any code for this.

> >  }
> >  
> >  static const struct attribute_group *usb_udc_attr_groups[];
> > @@ -1574,7 +1575,7 @@ static int gadget_match_driver(struct device *dev, const struct device_driver *d
> >  			struct usb_gadget_driver, driver);
> >  
> >  	/* If the driver specifies a udc_name, it must match the UDC's name */
> > -	if (driver->udc_name &&
> > +	if (driver->udc_name && udc &&
> 
> I agree this isn't good, but you just made the window smaller, please
> fix this properly.

I don't see how udc can possibly be NULL here.  It gets initialized to a 
non-NULL value when usb_add_gadget() does:

	gadget->udc = udc;

and nothing changes its value thereafter.  It seems much more likely 
that the error shown above is an invalid pointer access because 
gadget->udc points to a location that has been deallocated.  Adding this 
NULL check won't fix the bug.

Apparently the problem is caused by the fact that bus_for_each_dev(), 
iterating over the things on the gadget bus, is still using gadget after

	device_del(&gadget->dev);

in usb_del_gadget() returns and while

	device_unregister(&udc->dev);

runs and the udc structure is deallocated.  The only solution I can 
think of is for the gadget to take a reference to the udc and drop the 
reference when the gadget is released.  Unfortunately, several UDC 
drivers define their own gadget-release routines; they will all need to 
be modified.  And the core will need its own gadget-release routine for 
use when the UDC driver does not specify its own.

Alan Stern

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ