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  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]
Date:	Tue, 22 Jul 2014 09:47:04 -0500
From:	Felipe Balbi <balbi@...com>
To:	Lothar Waßmann <LW@...O-electronics.de>
CC:	<balbi@...com>, Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	<linux-kernel@...r.kernel.org>, <linux-usb@...r.kernel.org>,
	Ezequiel Garcia <ezequiel@...guardiasur.com.ar>,
	George Cherian <george.cherian@...com>,
	<linux-arm-kernel@...ts.infradead.org>,
	Roger Quadros <rogerq@...com>, <devicetree@...r.kernel.org>
Subject: Re: [PATCH 9/9] usb: musb: musb_am335x: reinstate module
 loading/unloading support

On Tue, Jul 22, 2014 at 09:49:30AM +0200, Lothar Waßmann wrote:
> Hi,
> 
> Felipe Balbi wrote:
> > On Fri, Jul 18, 2014 at 11:31:30AM +0200, Lothar Waßmann wrote:
> > > There is no need to throw the baby out with the bath due to a bad
> > > failure analysis. The commit:
> > > 7adb5c876e9c usb: musb: Fix panic upon musb_am335x module removal
> > > came to a wrong conclusion about the cause of the crash it was
> > > "fixing". The real culprit was the phy-am335x module that was removed
> > > from underneath its users that were still referencing data from it.
> > > After having fixed this in a previous patch, module unloading can be
> > > reinstated.
> > > 
> > > Another bug with module loading/unloading was the fact, that after
> > > removing the devices instantiated from DT their 'OF_POPULATED' flag
> > > was still set, so that re-loading the module after an unload had no
> > > effect. This is also fixed in this patch.
> > 
> > now this is a good commit log. I still can't see the need for the other
> > patch adding try_module_get(), though. Another thing, this needs to be
> > reviewed by DT folks too.
> > 
> Without holding a reference to the phy module, the module may be
> unloaded when its resources are still in use which may lead to the
> crash observed in the above stated commit.
> 
> > > Signed-off-by: Lothar Waßmann <LW@...O-electronics.de>
> > > ---
> > >  drivers/usb/musb/musb_am335x.c |   23 ++++++++++++++++++-----
> > >  1 file changed, 18 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/drivers/usb/musb/musb_am335x.c b/drivers/usb/musb/musb_am335x.c
> > > index 164c868..152a6f5 100644
> > > --- a/drivers/usb/musb/musb_am335x.c
> > > +++ b/drivers/usb/musb/musb_am335x.c
> > > @@ -19,6 +19,22 @@ err:
> > >  	return ret;
> > >  }
> > >  
> > > +static int of_remove_populated_child(struct device *dev, void *d)
> > > +{
> > > +	struct platform_device *pdev = to_platform_device(dev);
> > > +
> > > +	of_device_unregister(pdev);
> > > +	of_node_clear_flag(pdev->dev.of_node, OF_POPULATED);
> > 
> > I don't think this should be called by drivers; rather by the bus.
> > 
> Possibly the right thing to do would be to use of_platform_depopulate()
> in the remove function to pair up with op_platform_populate() in the
> probe function, but doing this results in a crash in
> platform_device_del() (called from platform_device_unregister()) when
> release_resource() is being called on resources that were never
> properly registered with the device:
> Unable to handle kernel NULL pointer dereference at virtual address 00000018
> pgd = 8dd64000
> [00000018] *pgd=8ddc2831, *pte=00000000, *ppte=00000000
> Internal error: Oops: 17 [#1] PREEMPT ARM
> Modules linked in: musb_am335x(-) [last unloaded: snd]
> CPU: 0 PID: 1435 Comm: modprobe Not tainted 3.16.0-rc5-next-20140717-dbg+ #13
> task: 8da00880 ti: 8dda0000 task.ti: 8dda0000
> PC is at release_resource+0x14/0x7c
> LR is at release_resource+0x10/0x7c
> pc : [<8003165c>]    lr : [<80031658>]    psr: a0000013
> sp : 8dda1ec0  ip : 8dda0000  fp : 00000000
> r10: 00000000  r9 : 8dda0000  r8 : 8deb7c10
> r7 : 8deb7c00  r6 : 00000200  r5 : 00000001  r4 : 8deed380
> r3 : 00000000  r2 : 00000000  r1 : 00000011  r0 : 806772a0
> Flags: NzCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
> Control: 10c5387d  Table: 8dd64019  DAC: 00000015
> Process modprobe (pid: 1435, stack limit = 0x8dda0238)
> Stack: (0x8dda1ec0 to 0x8dda2000)
> 1ec0: 8da00880 8deed380 00000001 802f850c 00000000 8deed380 44e10620 44e1062f
> 1ee0: 8deb7c00 00000000 80398c5c 00000081 8000e904 802f8848 8deb7c10 80398d0c
> 1f00: 00000000 802f3470 8d8d7200 8ddc84b0 8d92e810 7f0f9014 8d92e844 7f0f7010
> 1f20: 8d92e810 802f8110 802f80f8 802f68f8 7f0f9014 8d92e810 7f0f9014 802f7238
> 1f40: 7f0f9014 00000000 20000013 802f6764 7f0f9058 8007ec94 00000000 6273756d
> 1f60: 336d615f 00783533 8da00880 8000e764 00000001 80053ae4 00000058 76f41000
> 1f80: 7eb299e8 01290838 00000011 00000081 60000010 0000e854 7eb299e8 01290838
> 1fa0: 00000011 8000e740 7eb299e8 01290838 012908a0 00000080 00000000 00000001
> 1fc0: 7eb299e8 01290838 00000011 00000081 012908a0 00000000 01290844 00000000
> 1fe0: 76eb76f0 7eb2995c 0000a534 76eb76fc 60000010 012908a0 aaaaaaaa aaaaaaaa
> [<8003165c>] (release_resource) from [<802f850c>] (platform_device_del+0xb4/0xf4)
> [<802f850c>] (platform_device_del) from [<802f8848>] (platform_device_unregister+0xc/0x18)
> [<802f8848>] (platform_device_unregister) from [<80398d0c>] (of_platform_device_destroy+0xb0/0xc8)
> [<80398d0c>] (of_platform_device_destroy) from [<802f3470>] (device_for_each_child+0x34/0x74)
> [<802f3470>] (device_for_each_child) from [<7f0f7010>] (am335x_child_remove+0x10/0x24 [musb_am335x])
> [<7f0f7010>] (am335x_child_remove [musb_am335x]) from [<802f8110>] (platform_drv_remove+0x18/0x1c)
> [<802f8110>] (platform_drv_remove) from [<802f68f8>] (__device_release_driver+0x70/0xc4)
> [<802f68f8>] (__device_release_driver) from [<802f7238>] (driver_detach+0xb4/0xb8)
> [<802f7238>] (driver_detach) from [<802f6764>] (bus_remove_driver+0x5c/0xa4)
> [<802f6764>] (bus_remove_driver) from [<8007ec94>] (SyS_delete_module+0x120/0x18c)
> [<8007ec94>] (SyS_delete_module) from [<8000e740>] (ret_fast_syscall+0x0/0x48)
> Code: e1a04000 e59f0068 eb10bac4 e5943010 (e5932018) 
> ---[ end trace 24ca43dfc1a677d6 ]---
> 
> The cause for this seems to be calling platform_device_unregister() on
> a device that was created with device_add() (rather than
> platform_device_add()).

good, then you found another bug. Let's fix that and use
of_platform_depopulate().

-- 
balbi

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)

Powered by blists - more mailing lists