Re: [PATCH v4 2/2] mailbox: sprd: Add Spreadtrum mailbox driver

[Baolin Wang]

On Fri, May 22, 2020 at 11:48 AM Jassi Brar <jassisinghbrar@xxxxxxxxx> wrote:
>
> On Thu, May 21, 2020 at 7:24 AM Baolin Wang <baolin.wang7@xxxxxxxxx> wrote:
> >
> > Hi Jassi,
> >
> > On Wed, May 13, 2020 at 2:32 PM Baolin Wang <baolin.wang7@xxxxxxxxx> wrote:
> > >
> > > On Wed, May 13, 2020 at 2:05 PM Jassi Brar <jassisinghbrar@xxxxxxxxx> wrote:
> > > >
> > > > On Tue, May 12, 2020 at 11:14 PM Baolin Wang <baolin.wang7@xxxxxxxxx> wrote:
> > > > >
> > > > > Hi Jassi,
> > > > >
> > > > > On Thu, May 7, 2020 at 11:23 AM Baolin Wang <baolin.wang7@xxxxxxxxx> wrote:
> > > > > >
> > > > > > Hi Jassi,
> > > > > >
> > > > > > On Thu, May 7, 2020 at 7:25 AM Jassi Brar <jassisinghbrar@xxxxxxxxx> wrote:
> > > > > > >
> > > > > > > On Wed, May 6, 2020 at 8:29 AM Baolin Wang <baolin.wang7@xxxxxxxxx> wrote:
> > > > > > > >
> > > > > > > > Hi Jassi,
> > > > > > > >
> > > > > > > > On Tue, Apr 28, 2020 at 11:10 AM Baolin Wang <baolin.wang7@xxxxxxxxx> wrote:
> > > > > > > > >
> > > > > > > > > From: Baolin Wang <baolin.wang@xxxxxxxxxx>
> > > > > > > > >
> > > > > > > > > The Spreadtrum mailbox controller supports 8 channels to communicate
> > > > > > > > > with MCUs, and it contains 2 different parts: inbox and outbox, which
> > > > > > > > > are used to send and receive messages by IRQ mode.
> > > > > > > > >
> > > > > > > > > Signed-off-by: Baolin Wang <baolin.wang@xxxxxxxxxx>
> > > > > > > > > Signed-off-by: Baolin Wang <baolin.wang7@xxxxxxxxx>
> > > > > > > > > ---
> > > > > > > > > Changes from v3:
> > > > > > > > >  - Save the id in mbox_chan.con_priv and remove the 'sprd_mbox_chan'
> > > > > > > > >
> > > > > > > > > Changes from v2:
> > > > > > > > >  - None.
> > > > > > > > >
> > > > > > > > > Changes from v1:
> > > > > > > > >  - None
> > > > > > > >
> > > > > > > > Gentle ping, do you have any other comments? Thanks.
> > > > > > > >
> > > > > > > Yea, I am still not sure about the error returned in send_data().  It
> > > > > > > will either never hit or there will be no easy recovery from it. The
> > > > > > > api expects the driver to tell it the last-tx was done only when it
> > > > > > > can send the next message. (There may be case like sending depend on
> > > > > > > remote, which can't be ensured before hand).
> > > > > >
> > > > > > Actually this is an unusual case, suppose the remote target did not
> > > > > > fetch the message as soon as possile, which will cause the FIFO
> > > > > > overflow, so in this case we  can not send messages to the remote
> > > > > > target any more, otherwise messages will be lost. Thus we can return
> > > > > > errors to users to indicate that something wrong with the remote
> > > > > > target need to be checked.
> > > > > >
> > > > > > So this validation in send_data() is mostly for debugging for this
> > > > > > abnormal case and we will not trigger this issue if the remote target
> > > > > > works well. So I think it is useful to keep this validation in
> > > > > > send_data(). Thanks.
> > > > >
> > > > > Any comments? Thanks.
> > > > >
> > > > Same as my last post.
> > >
> > > I think I've explained the reason why we need add this validation in
> > > my previous email, I am not sure how do you think? You still want to
> > > remove this validation?
> >
> > Gentle ping.
> >
> > As I explained in previous email, this validation is for an unusual
> > case, suppose the remote target did not fetch the message as soon as
> > possile, which will cause the FIFO overflow, so in this case we  can
> > not send messages to the remote
> > target any more, otherwise messages will be lost. Thus we can return
> > errors to users to indicate that something wrong with the remote
> > target need to be checked.
> >
> > So this validation in send_data() is mostly for debugging for this
> > abnormal case and we will not trigger this issue if the remote target
> > works well. So I think it is useful to keep this validation in
> > send_data(). What do you think? Thanks.
> >
> I still think the same as before.
> You should do this check before you call mbox_chan_txdone() and wait
> if busy ... which is exactly the purpose of txdone().
> It seems harmless to be paranoid and place a block of code in
> practically "if 0", but that sets bad precedence for other drivers. So
> please move the check before txdone().

OK. I realized I can implement the flush() to make sure the
transmission has been completed. Thanks.

--
Baolin Wang