Re: Re: [PATCH V2] drivers: nfc: nfcmrvl: fix double free bug in nfcmrvl_nci_unregister_dev()

[duoming@zju.edu.cn]

Hello,
On Tue, 12 Apr 2022 00:28:16 -0700 Paolo Abeni wrote:

> > There is a potential double bug in nfcmrvl usb driver between
> > unregister and resume operation.
> > 
> > The race that cause that double free bug can be shown as below:
> > 
> >    (FREE)                   |      (USE)
> >                             | nfcmrvl_resume
> >                             |  nfcmrvl_submit_bulk_urb
> >                             |   nfcmrvl_bulk_complete
> >                             |    nfcmrvl_nci_recv_frame
> >                             |     nfcmrvl_fw_dnld_recv_frame
> >                             |      queue_work
> >                             |       fw_dnld_rx_work
> >                             |        fw_dnld_over
> >                             |         release_firmware
> >                             |          kfree(fw); //(1)
> > nfcmrvl_disconnect          |
> >  nfcmrvl_nci_unregister_dev |
> >   nfcmrvl_fw_dnld_abort     |
> >    fw_dnld_over             |         ...
> >     if (priv->fw_dnld.fw)   |
> >     release_firmware        |
> >      kfree(fw); //(2)       |
> >      ...                    |         priv->fw_dnld.fw = NULL;
> > 
> > When nfcmrvl usb driver is resuming, we detach the device.
> > The release_firmware() will deallocate firmware in position (1),
> > but firmware will be deallocated again in position (2), which
> > leads to double free.
> > 
> > This patch reorders nfcmrvl_fw_dnld_deinit() before nfcmrvl_fw_dnld_abort()
> > in order to prevent double free bug. Because destroy_workqueue() will
> > not return until all work items are finished. The priv->fw_dnld.fw will
> > be set to NULL after work items are finished and fw_dnld_over() called by
> > nfcmrvl_nci_unregister_dev() will check whether priv->fw_dnld.fw is NULL.
> > So the double free bug could be prevented.
> > 
> > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> 
> This looks like a -net candidates, could you please add a suitable
> fixes tag?

I found my patch is not a comprehensive fix, because the fw_dnld_timeout()
could also reach fw_dnld_over(), which leads to double free bugs among 
fw_dnld_rx_work(), fw_dnld_timeout() and nfcmrvl_nci_unregister_dev().

I sent a patch set "Fix double free bugs in nfcmrvl module" just now, and add
a fixes tag in it.

Best regards,
Duoming Zhou