Re: [PATCH v2 5/5] winebus.sys: Handle IRP_MN_SURPRISE_REMOVAL and set removed flag.

[Zebediah Figura (she/her)]

On 6/29/21 2:21 AM, Rémi Bernon wrote:
> Signed-off-by: Rémi Bernon
>
> dlls/winebus.sys/main.c | 14 ++++++++++----
>   1 file changed, 10 insertions(+), 4 deletions(-)
>
> diff --git a/dlls/winebus.sys/main.c b/dlls/winebus.sys/main.c
> index 897be25ed51..f3e63e87aa2 100644
> --- a/dlls/winebus.sys/main.c
> +++ b/dlls/winebus.sys/main.c
> @@ -374,10 +374,6 @@ void bus_unlink_hid_device(DEVICE_OBJECT *device)
>       EnterCriticalSection(&device_list_cs);
>       list_remove(&pnp_device->entry);
>       LeaveCriticalSection(&device_list_cs);
> -
> -
>
> EnterCriticalSection(&ext->cs);
> -
>
> ext->removed = TRUE;
> -
>
> LeaveCriticalSection(&ext->cs);
> }
>
> void bus_remove_hid_device(DEVICE_OBJECT *device)
>
> @@ -700,11 +696,21 @@ static NTSTATUS pdo_pnp_dispatch(DEVICE_OBJECT *device, IRP *irp)
>               status = STATUS_SUCCESS;
>               break;
> -
>    case IRP_MN_SURPRISE_REMOVAL:
> -
>        EnterCriticalSection(&ext->cs);
> -
>        remove_pending_irps(device);
> -
>        ext->removed = TRUE;
> -
>        LeaveCriticalSection(&ext->cs);
> -
>        break;
> -

This is fine, I guess, but it makes the lock in remove_pending_irps() 
redundant.
>       case IRP_MN_REMOVE_DEVICE:
>       {
>           struct pnp_device *pnp_device = ext->pnp_device;
>
> -
>        EnterCriticalSection(&ext->cs);
>         remove_pending_irps(device);
> -
>        ext->removed = TRUE;
> -
>        LeaveCriticalSection(&ext->cs);
>
>         ext->vtbl->free_device(device);

I believe that ntoskrnl should guarantee we won't receive any IRPs after 
IRP_MN_REMOVE_DEVICE, which makes this hunk redundant.