On Fri 15 Jan 03:13 CST 2021, Arnaud POULIQUEN wrote:
> Hi Mathieu,
>
>
> On 1/14/21 8:05 PM, Mathieu Poirier wrote:
> > On Wed, Jan 06, 2021 at 02:37:14PM +0100, Arnaud Pouliquen wrote:
> >> The rpmsg_create_ept function is invoked when the device is opened.
> >> As only one endpoint must be created per device. It is not
> >> possible to open the same device twice.
> >> The fix consists in returning -EBUSY when device is already
> >> opened.
> >>
> >> Fixes: c0cdc19f84a4 ("rpmsg: Driver for user space endpoint interface")
> >> Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@xxxxxxxxxxx>
> >> ---
> >> drivers/rpmsg/rpmsg_char.c | 3 +++
> >> 1 file changed, 3 insertions(+)
> >>
> >> diff --git a/drivers/rpmsg/rpmsg_char.c b/drivers/rpmsg/rpmsg_char.c
> >> index 4bbbacdbf3bb..360a1ab0a9c4 100644
> >> --- a/drivers/rpmsg/rpmsg_char.c
> >> +++ b/drivers/rpmsg/rpmsg_char.c
> >> @@ -127,6 +127,9 @@ static int rpmsg_eptdev_open(struct inode *inode, struct file *filp)
> >> struct rpmsg_device *rpdev = eptdev->rpdev;
> >> struct device *dev = &eptdev->dev;
> >>
> >> + if (eptdev->ept)
> >> + return -EBUSY;
> >> +
> >
> > I rarely had to work so hard to review a 2 line patch...
>
> That means that my commit description was not enough explicit...
>
> >
> > As far as I can tell the actual code is doing the right thing. If user space is
> > trying to open the same eptdev more than once function rpmsg_create_ept() should
> > complain and the operation denied, wich is what the current code is doing.
> >
> > There is currently two customers for this API - SMD and GLINK. The SMD code is
> > quite clear that if the channel is already open, the operation will be
> > denied [1]. The GLINK code isn't as clear but the fact that it returns NULL on
> > error conditions [2] is a good indication that things are working the same way.
> >
> > What kind of use case are you looking to address? Is there any way you can use
> > rpdev->ops->create_ept() as it is currently done?
>
> This patch was part of the IOCTL rpmsg series. I sent it separately at Bjorn's
> request [1].
>
I apparently didn't spend as much effort as Mathieu thinking about the
details. I do believe that he's right, at least both GLINK and SMD
_should_ return -EBUSY if we try to open an already open channel -
either because the kernel has bound a driver to the channel or because
rpmsg_char already has it opened.
> I detect the issue using the RPMSG_ADDR_ANY for the source address when tested
> it with the rpmsf_virtio bus. In this case at each sys open of the device, a new
> endpoint is created because a new source address is allocated.
>
In SMD and GLINK channels are identified solely by their name and hence
it's not possible to have duplicates. As this isn't the case for virtio
I didn't have any objections to it and that's why I asked you to resend
it separately.
But in line with GLINK/SMD, what would the expected behavior be if I
with the virtio backend open a rpmsg_char which is already bound to a
kernel driver?
Do you think we should get another "channel" or do you think the virtio
driver should detect this and return -EBUSY? (I.e. render this patch
unnecessary)
Regards,
Bjorn
> [1]https://patchwork.kernel.org/project/linux-remoteproc/patch/20201222105726.16906-11-arnaud.pouliquen@xxxxxxxxxxx/
>
> Thanks,
> Arnaud
>
> >
> > Thanks,
> > Mathieu
> >
> > [1]. https://elixir.bootlin.com/linux/v5.11-rc3/source/drivers/rpmsg/qcom_smd.c#L920
> > [2]. https://elixir.bootlin.com/linux/v5.11-rc3/source/drivers/rpmsg/qcom_glink_native.c#L1149
> >
> >> get_device(dev);
> >>
> >> ept = rpmsg_create_ept(rpdev, rpmsg_ept_cb, eptdev, eptdev->chinfo);
> >> --
> >> 2.17.1
> >>
> > _______________________________________________
> > Linux-stm32 mailing list
> > Linux-stm32@xxxxxxxxxxxxxxxxxxxxxxxxxxxx
> > https://st-md-mailman.stormreply.com/mailman/listinfo/linux-stm32
> >
