RE: [PATCH v4] PCI: endpoint: Fix NULL pointer dereference for ->get_features()

["Shradha Todi" <shradha.t@samsung.com>]

> -----Original Message-----
> From: Leon Romanovsky <leon@kernel.org>
> Subject: Re: [PATCH v4] PCI: endpoint: Fix NULL pointer dereference for -
> >get_features()
> 
> On Tue, Jan 19, 2021 at 03:25:10PM +0530, Shradha Todi wrote:
> > > -----Original Message-----
> > > From: Leon Romanovsky <leon@kernel.org>
> > > Subject: Re: [PATCH v4] PCI: endpoint: Fix NULL pointer dereference
> > > for -
> > > >get_features()
> > >
> > > On Tue, Jan 12, 2021 at 07:32:25PM +0530, Shradha Todi wrote:
> > > > get_features ops of pci_epc_ops may return NULL, causing NULL
> > > > pointer dereference in pci_epf_test_bind function. Let us add a
> > > > check for pci_epc_feature pointer in pci_epf_test_bind before we
> > > > access it to avoid any such NULL pointer dereference and return
> > > > -ENOTSUPP in case pci_epc_feature is not found.
> > > >
> > > > When the patch is not applied and EPC features is not implemented
> > > > in the platform driver, we see the following dump due to kernel
> > > > NULL pointer dereference.
> > > >
> > > > [  105.135936] Call trace:
> > > > [  105.138363]  pci_epf_test_bind+0xf4/0x388 [  105.142354]
> > > > pci_epf_bind+0x3c/0x80 [  105.145817]  pci_epc_epf_link+0xa8/0xcc
> > > > [ 105.149632]  configfs_symlink+0x1a4/0x48c [  105.153616]
> > > > vfs_symlink+0x104/0x184 [  105.157169]  do_symlinkat+0x80/0xd4 [
> > > > 105.160636]  __arm64_sys_symlinkat+0x1c/0x24 [  105.164885]
> > > > el0_svc_common.constprop.3+0xb8/0x170
> > > > [  105.169649]  el0_svc_handler+0x70/0x88 [  105.173377]
> > > > el0_svc+0x8/0x640 [  105.176411] Code: d2800581 b9403ab9 f9404ebb
> > > > 8b394f60 (f9400400) [  105.182478] ---[ end trace a438e3c5a24f9df0
> > > > ]---
> > >
> > >
> > > Description and call trace don't correlate with the proposed code
change.
> > >
> > > The code in pci_epf_test_bind() doesn't dereference epc_features, at
> > > least
> > in
> > > direct manner.
> > >
> > > Thanks
> >
> > Thanks for the review. Yes, you're right. The dereference does not
> > happen in the pci_epf_test_bind() itself, but in
> > pci_epf_test_alloc_space() being called within. We will update the
> > line "causing NULL pointer dereference in pci_epf_test_bind function.
> > " in the commit message to "causing NULL pointer dereference in
> > pci_epf_test_alloc_space function. " Would that be good enough?
> 
> I have no idea, just saw that the rest of the code at least partially
prepared to
> deal with epc_features == NULL, maybe
> pci_epf_test_alloc_space() should be update to handle epc_features ==
NULL.
> 
> Thanks
> 

In current implementation, epc_features has been used in quite a few places
in the file without proper checks. Anyway, there is no point in proceeding
to bind the driver with no EPC features implemented, so we feel it's best to
return as soon as possible instead of going ahead and returning error at a
later point. (eg: during allocation). I checked that alloc function is not
being called anywhere else except for the bind function.

Thanks

> >