Re: [PATCH 0/9] tty: tty_struct dependency clean-ups
On Fri, Sep 09, 2016 at 05:37:01PM -0500, Rob Herring wrote: > This patch series removes or prepares to remove some of the dependencies > on tty_struct within tty_port drivers. This will allow using tty_ports > directly for so called UART slave devices. > > The first patch fixes a regression from the last round with tty > closing. Patches 2-4 are mechanical conversions generated with > coccinelle. This is my first adventure into coccinelle, so hopefully I > didn't screw things up. :) Only serial_core has some of its tty_struct > dependencies removed as those are the drivers we immediately care about > supporting slave devices with. > > Next up after this are moving some functions to the tty_port ops. I've > got some WIP patches for some of that, but nothing ready to send out > quite yet. I've applied the first patch here, and wanted to apply patch 3, but it depended on #2, so I'll wait for your next round of patches. thanks, greg k-h
Re: [PATCH 0/9] tty: tty_struct dependency clean-ups
On Sun, 11 Sep 2016 22:05:07 -0500 Rob Herring wrote: > On Sun, Sep 11, 2016 at 4:14 PM, One Thousand Gnomes > wrote: > > On Fri, 9 Sep 2016 17:37:01 -0500 > > Rob Herring wrote: > > > >> This patch series removes or prepares to remove some of the dependencies > >> on tty_struct within tty_port drivers. This will allow using tty_ports > >> directly for so called UART slave devices. > > > > You can create a tty_struct kernel side with the two tiny changes I > > posted before. Why do you want to do invasive tree wide changes when you > > can do simple ones ? > > Well, I don't want to do invasive changes, but I thought the idea was > to use tty_port struct without a tty_struct. I posted some tiny patches to break the file/tty requirement in the base tty code for comment a while ago and they were very tiny for most ldiscs (n_tty unsurprisingly wouldn't work this way but does anyone need kernel mode n_tty ?) Moving termios into the tty_port is IMHO a good thing to do whichever approach is taken. > I was planning to keep termios out of tty_port and make clients of > tty_port carry it if for nothing else not quite understanding all the > details around the lifetime, init and locking of it. If there's always > a tty_struct then there's not much point moving it other than which > struct makes more sense. But that would cause some churn. The termios lifetime is the lifetime of the port, although it may get reset at some times. Alan
Re: [PATCH 0/9] tty: tty_struct dependency clean-ups
On Sun, Sep 11, 2016 at 4:14 PM, One Thousand Gnomes wrote: > On Fri, 9 Sep 2016 17:37:01 -0500 > Rob Herring wrote: > >> This patch series removes or prepares to remove some of the dependencies >> on tty_struct within tty_port drivers. This will allow using tty_ports >> directly for so called UART slave devices. > > You can create a tty_struct kernel side with the two tiny changes I > posted before. Why do you want to do invasive tree wide changes when you > can do simple ones ? Well, I don't want to do invasive changes, but I thought the idea was to use tty_port struct without a tty_struct. >> Next up after this are moving some functions to the tty_port ops. I've >> got some WIP patches for some of that, but nothing ready to send out >> quite yet. > > I think before this lot happens you need to decide where these structures > belong. Termios and termios_locked for example could live in the tty_port > as the physical tty is incapable of having multiple sets of terminal data > at once. I was planning to keep termios out of tty_port and make clients of tty_port carry it if for nothing else not quite understanding all the details around the lifetime, init and locking of it. If there's always a tty_struct then there's not much point moving it other than which struct makes more sense. But that would cause some churn. Rob
Re: [PATCH 0/9] tty: tty_struct dependency clean-ups
On Fri, 9 Sep 2016 17:37:01 -0500 Rob Herring wrote: > This patch series removes or prepares to remove some of the dependencies > on tty_struct within tty_port drivers. This will allow using tty_ports > directly for so called UART slave devices. You can create a tty_struct kernel side with the two tiny changes I posted before. Why do you want to do invasive tree wide changes when you can do simple ones ? > Next up after this are moving some functions to the tty_port ops. I've > got some WIP patches for some of that, but nothing ready to send out > quite yet. I think before this lot happens you need to decide where these structures belong. Termios and termios_locked for example could live in the tty_port as the physical tty is incapable of having multiple sets of terminal data at once. Really though this looks to me like *MASSIVE* churn for now purpose. Create a tty_struct kernel side, and use that, the needed patch is then tiny. so IMHO NAK Alan
