HID: intel-ish-hid: pci-ish: Fix use after free bug in ish_remove due to race condition

  In ish_probe, it calls ish_dev_init to init the device. In this function,
ishtp_device_init is called and &dev->bh_hbm_work is bound with 
bh_hbm_work_fn. recv_hbm may be called to start the timer work.

If we remove the module which will call ish_remove to make cleanup,
there may be an unfinished work. The possible sequence is as follows:

Fix it by canceling the work before cleanup in ishtp_bus_remove_all_clients

CPU0                  CPUc1

                    |bh_hbm_work_fn
ish_remove      |
ishtp_bus_remove_all_clients  |
kfree(ishtp_dev->fw_clients); |
                    |
                    |&dev->fw_clients[...]
                    |   //use

Fixes: 3703f53b99e4 ("HID: intel_ish-hid: ISH Transport layer")
Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
---
 drivers/hid/intel-ish-hid/ishtp/bus.c | 1 +
 1 file changed, 1 insertion(+)
  

Hi,

I missed some detail in the patch. bh_hbm_work_fn will call
ishtp_hbm_dispatch and access &dev->fw_clients in it.

Best regards,
Zheng

Zheng Wang <zyytlz.wz@163.com> 于2023年4月13日周四 01:17写道：
>
> In ish_probe, it calls ish_dev_init to init the device. In this function,
> ishtp_device_init is called and &dev->bh_hbm_work is bound with
> bh_hbm_work_fn. recv_hbm may be called to start the timer work.
>
> If we remove the module which will call ish_remove to make cleanup,
> there may be an unfinished work. The possible sequence is as follows:
>
> Fix it by canceling the work before cleanup in ishtp_bus_remove_all_clients
>
> CPU0                  CPUc1
>
>                     |bh_hbm_work_fn
> ish_remove      |
> ishtp_bus_remove_all_clients  |
> kfree(ishtp_dev->fw_clients); |
>                     |
>                     |&dev->fw_clients[...]
>                     |   //use
>
> Fixes: 3703f53b99e4 ("HID: intel_ish-hid: ISH Transport layer")
> Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
> ---
>  drivers/hid/intel-ish-hid/ishtp/bus.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/hid/intel-ish-hid/ishtp/bus.c b/drivers/hid/intel-ish-hid/ishtp/bus.c
> index 81385ab37fa9..ada7cd08dbeb 100644
> --- a/drivers/hid/intel-ish-hid/ishtp/bus.c
> +++ b/drivers/hid/intel-ish-hid/ishtp/bus.c
> @@ -744,6 +744,7 @@ void ishtp_bus_remove_all_clients(struct ishtp_device *ishtp_dev,
>                  */
>         }
>         spin_unlock_irqrestore(&ishtp_dev->cl_list_lock, flags);
> +       cancel_work_sync(&ishtp_dev->bh_hbm_work);
>
>         /* Release DMA buffers for client messages */
>         ishtp_cl_free_dma_buf(ishtp_dev);
> --
> 2.25.1
>
  

On Thu, 2023-04-13 at 01:14 +0800, Zheng Wang wrote:
> In ish_probe, it calls ish_dev_init to init the device. In this
> function,
> ishtp_device_init is called and &dev->bh_hbm_work is bound with 
> bh_hbm_work_fn. recv_hbm may be called to start the timer work.
> 
> If we remove the module which will call ish_remove to make cleanup,
> there may be an unfinished work. The possible sequence is as follows:
> 
> Fix it by canceling the work before cleanup in
> ishtp_bus_remove_all_clients
> 
> CPU0                  CPUc1
> 
>                     |bh_hbm_work_fn
> ish_remove      |
> ishtp_bus_remove_all_clients  |
> kfree(ishtp_dev->fw_clients); |
>                     |
>                     |&dev->fw_clients[...]
>                     |   //use
> 
> Fixes: 3703f53b99e4 ("HID: intel_ish-hid: ISH Transport layer")
> Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

It is a extremely rare case, so it is OK to not mark for stable.

> ---
>  drivers/hid/intel-ish-hid/ishtp/bus.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/hid/intel-ish-hid/ishtp/bus.c
> b/drivers/hid/intel-ish-hid/ishtp/bus.c
> index 81385ab37fa9..ada7cd08dbeb 100644
> --- a/drivers/hid/intel-ish-hid/ishtp/bus.c
> +++ b/drivers/hid/intel-ish-hid/ishtp/bus.c
> @@ -744,6 +744,7 @@ void ishtp_bus_remove_all_clients(struct
> ishtp_device *ishtp_dev,
>                  */
>         }
>         spin_unlock_irqrestore(&ishtp_dev->cl_list_lock, flags);
> +       cancel_work_sync(&ishtp_dev->bh_hbm_work);
>  
>         /* Release DMA buffers for client messages */
>         ishtp_cl_free_dma_buf(ishtp_dev);
  

srinivas pandruvada <srinivas.pandruvada@linux.intel.com> 于2023年4月25日周二 23:55写道：
>
> On Thu, 2023-04-13 at 01:14 +0800, Zheng Wang wrote:
> > In ish_probe, it calls ish_dev_init to init the device. In this
> > function,
> > ishtp_device_init is called and &dev->bh_hbm_work is bound with
> > bh_hbm_work_fn. recv_hbm may be called to start the timer work.
> >
> > If we remove the module which will call ish_remove to make cleanup,
> > there may be an unfinished work. The possible sequence is as follows:
> >
> > Fix it by canceling the work before cleanup in
> > ishtp_bus_remove_all_clients
> >
> > CPU0                  CPUc1
> >
> >                     |bh_hbm_work_fn
> > ish_remove      |
> > ishtp_bus_remove_all_clients  |
> > kfree(ishtp_dev->fw_clients); |
> >                     |
> >                     |&dev->fw_clients[...]
> >                     |   //use
> >
> > Fixes: 3703f53b99e4 ("HID: intel_ish-hid: ISH Transport layer")
> > Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
> Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
>
> It is a extremely rare case, so it is OK to not mark for stable.

Got it, thanks!

Best regards,
Zheng

>
> > ---
> >  drivers/hid/intel-ish-hid/ishtp/bus.c | 1 +
> >  1 file changed, 1 insertion(+)
> >
> > diff --git a/drivers/hid/intel-ish-hid/ishtp/bus.c
> > b/drivers/hid/intel-ish-hid/ishtp/bus.c
> > index 81385ab37fa9..ada7cd08dbeb 100644
> > --- a/drivers/hid/intel-ish-hid/ishtp/bus.c
> > +++ b/drivers/hid/intel-ish-hid/ishtp/bus.c
> > @@ -744,6 +744,7 @@ void ishtp_bus_remove_all_clients(struct
> > ishtp_device *ishtp_dev,
> >                  */
> >         }
> >         spin_unlock_irqrestore(&ishtp_dev->cl_list_lock, flags);
> > +       cancel_work_sync(&ishtp_dev->bh_hbm_work);
> >
> >         /* Release DMA buffers for client messages */
> >         ishtp_cl_free_dma_buf(ishtp_dev);
>