linux-kernel - Re: [PATCH] usb: gadget: composite: split spinlock to avoid recursion

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: <Pine.LNX.4.44L0.1911131021460.1558-100000@iolanthe.rowland.org>
Date:   Wed, 13 Nov 2019 10:36:25 -0500 (EST)
From:   Alan Stern <stern@...land.harvard.edu>
To:     Peter Chen <peter.chen@....com>
cc:     Michael Olbrich <m.olbrich@...gutronix.de>,
        "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
        Felipe Balbi <balbi@...nel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] usb: gadget: composite: split spinlock to avoid recursion

On Wed, 13 Nov 2019, Peter Chen wrote:

> On 19-11-12 10:33:18, Michael Olbrich wrote:
> > 'delayed_status' and 'deactivations' are used completely independent but
> > they share the same spinlock. This can result in spinlock recursion:
> > 
> > BUG: spinlock recursion on CPU#1, uvc-gadget/322
> >  lock: 0xffffffc0570364e0, .magic: dead4ead, .owner: uvc-gadget/322, .owner_cpu: 1
> > CPU: 1 PID: 322 Comm: uvc-gadget Tainted: G         C O      5.3.0-20190916-1+ #55
> > Hardware name: XXXXX (DT)
> > Call trace:
> >  dump_backtrace+0x0/0x178
> >  show_stack+0x24/0x30
> >  dump_stack+0xc0/0x104
> >  spin_dump+0x90/0xa0
> >  do_raw_spin_lock+0xd8/0x108
> >  _raw_spin_lock_irqsave+0x40/0x50
> >  composite_disconnect+0x2c/0x80
> >  usb_gadget_disconnect+0x84/0x150
> >  usb_gadget_deactivate+0x64/0x120
> >  usb_function_deactivate+0x70/0x80
> >  uvc_function_disconnect+0x20/0x58
> >  uvc_v4l2_release+0x34/0x90
> >  v4l2_release+0xbc/0xf0
> >  __fput+0xb0/0x218
> >  ____fput+0x20/0x30
> >  task_work_run+0xa0/0xd0
> >  do_notify_resume+0x2f4/0x340
> >  work_pending+0x8/0x14
> > 
> > Fix this by using separate spinlocks.
> 
> This issue may be introduced by 0a55187a1ec8c ("USB: gadget core: Issue
> ->disconnect() callback from usb_gadget_disconnect()"), which adds
> gadget's disconnect at usb_gadget_disconnect. Add Alan, if he is Ok
> with your patch, you may cc to stable tree.

I wasn't aware of the dual usage of that lock in the composite core 
(and 0a55187a1ec8c touches only the gadget core, not composite.c).

In any case, I don't have a good feel for how the locking is supposed 
to work in the composite core.  This is really something Felipe should 
look at.

Would a better fix be to change usb_function_deactivate() so that it
doesn't hold the lock while calling usb_gadget_deactivate()?  Maybe
increment cdev->deactivations unconditionally before dropping the lock
(for mutual exclusion) and then decrement it again if the call fails?

Alan Stern