Re: [PATCH] usb: xhci: check Null pointer in segment alloc

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Mon, Dec 22, 2025 at 06:42:52AM +0000, Lee Jones wrote:
> On Sat, 20 Dec 2025, Michal Pecio wrote:
> 
> > Hi,
> > 
> > On Fri, 19 Dec 2025 15:53:08 +0000, 胡连勤 wrote:
> > > [ 4021.987665][  T332] Call trace:
> > > [ 4021.987668][  T332]  dma_pool_alloc+0x3c/0x248
> > > [ 4021.987676][  T332]  xhci_segment_alloc+0x9c/0x184
> > > [ 4021.987682][  T332]  xhci_alloc_segments_for_ring+0xcc/0x1cc
> > > [ 4021.987688][  T332]  xhci_ring_alloc+0xc4/0x1a8
> > > [ 4021.987693][  T332]  xhci_endpoint_init+0x36c/0x4ac
> > > [ 4021.987698][  T332]  xhci_add_endpoint+0x18c/0x2a4
> > > [ 4021.987702][  T332]  usb_hcd_alloc_bandwidth+0x384/0x3e4
> > > [ 4021.987711][  T332]  usb_set_interface+0x144/0x510
> > > [ 4021.987716][  T332]  usb_reset_and_verify_device+0x248/0x5fc
> > > [ 4021.987723][  T332]  usb_port_resume+0x580/0x700
> > > [ 4021.987730][  T332]  usb_generic_driver_resume+0x24/0x5c
> > > [ 4021.987735][  T332]  usb_resume_both+0x104/0x32c
> > > [ 4021.987740][  T332]  usb_runtime_resume+0x18/0x28
> > > [ 4021.987746][  T332]  __rpm_callback+0x94/0x3d4
> > > [ 4021.987754][  T332]  rpm_resume+0x3f8/0x5fc
> > > [ 4021.987762][  T332]  rpm_resume+0x1fc/0x5fc
> > > [ 4021.987769][  T332]  __pm_runtime_resume+0x4c/0x90
> > > [ 4021.987777][  T332]  usb_autopm_get_interface+0x20/0x4c
> > > [ 4021.987783][  T332]  snd_usb_autoresume+0x68/0x124
> > > [ 4021.987792][  T332]  suspend_resume_store+0x2a0/0x2b4 [dwc3_msm a4b7997a2e35cfe1a4a429762003b34dd4e85076]
> > 
> > This looks like some out of tree driver tries to resume a sound device,
> > and apparently it's doing it while xhci_hcd isn't ready, perhaps during
> > the power_lost branch in xhci_resume() after full system suspend.
> > 
> > I suppose dynamic debug could show better what's going on:
> > echo 'module usbcore +p' >/proc/dynamic_debug/control
> > echo 'module xhci_hcd +p' >/proc/dynamic_debug/control
> > 
> > If my guess is right then USB core is failing to prevent device resume
> > during HC resume, but IDK whether it's supposed to prevent that or if
> > the out of tree driver simply shouldn't be trying such things.
> 
> Lower-level functionality shouldn't be able to attack / fuzz core-code
> in this way.  Shouldn't the core be resistant to any possible mistakes
> or a lack of education exhibited by it's consumers?

Not always, we rely on drivers "doing the right thing" in almost all of
our in-kernel apis because we have access to the source of those drivers
to fix them to do the right thing.

> An API that insists on its users exercising care, knowledge and
> cognisance sounds fragile and vulnerable.

Fragile yes, vulnerable no.  Let's fix the fragility then, but as has
been pointed out in this thread, we don't know the root cause, and I
don't even think this "fix" would do the right thing anyway.

thanks,

greg k-h