This is the final step to get rid of the one of the structures.  A
further cleanup will follow. And I struct serial_state deserves cease
to exist after a switch to tty_port too.

While changing the lines, it removes also pointless tty->driver_data
casts.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

They used to work as a storage for 'info' pointer used in ISRs.  They
are not really needed. Just pass the pointer through request_irq to
the handlers.

It was set to NULL and tested in the ISRs, but we do not need the
tests as we disable all the interrupts at the same places where NULL
sets were.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

We do not set ASYNC_SHARE_IRQ anywhere. And since IRQF_DISABLED is a
noop, pass zero to request_irq directly instead of this ugly macro.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It never worked there. The ISR was never written for that kind of
stuff. So remove all that crap with a hash of linked lists and pass
the pointer directly to the ISR.

BTW this answers the question there:
  * I don't know exactly why they don't use the dev_id opaque data
  * pointer instead of this extra lookup table

-> Because they thought they will support more devices bound to a
single interrupt w/o IRQF_SHARED. They would need exactly the hash
there.

What I don't understand is rebinding of the interrupt in the shutdown
path. They perhaps meant to do just synchronize_irq? In any case, this
is all gone and free_irq there properly.

By removing the hash we save some bits (exactly NR_IRQS * 8 bytes of
.bss and over a kilo of .text):
before:
   text    data     bss     dec     hex filename
  19600     320    8227   28147    6df3 ../a/ia64/arch/ia64/hp/sim/simserial.o
after:
   text    data     bss     dec     hex filename
  18568     320      28   18916    49e4 ../a/ia64/arch/ia64/hp/sim/simserial.o

Note that a shared interrupt could not work too. request_irq requires
data parameter to be non-NULL. So the whole IRQ_T exercise was
pointless.

Finally, this helps us remove another two members of async_struct :).
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This means:
* close_delay
* closing_wait
* line
* port
* xmit_fifo_size

This actually fixes a bug in amiserial. It initializes one and uses
the other of the close delays. Yes, duplicating structure members is
evil.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The same as for amiserial. Use only one instance of the flags.

Also remove them from async_struct now. Nobody else uses them.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Without this, the code succeeds when the port is opened by root and we
get unwanted interrupts storm on the first key stroke.

Instead of that, tell the user we failed and that we won't continue. I
suppose, the code was copied from the serial layer where we may want
to change the irq number, so we must allow open even of the failing
port. This is not the case for this driver at all.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Currently, when assign_irq_vector is called and the irq connected in
the simulator, the irq is not ready. request_irq will return ENOSYS
immediately. It is because the irq chip is unset.

Hence set the chip properly to irq_type_hp_sim. And make sure this is
done from both users of simulated interrupts.

Also we have to set handler here, otherwise we end up in
handle_bad_int resulting in spam in logs and no irqs handled. We use
handle_simple_irq as these are SW interrupts that need no ACK or
anything.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

And remove declarations which are already in the headers.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The switch-cases of SAL_FREQ_BASE generate non-relocatable code. The
same as for the ifs one level upper. This causes oopses early in boot
because the kernel jumps to the hell instead of the offset in sal
callback.

So use ifs here for SAL_FREQ_BASE decision too.

Isn't there any compiler directive or settings to solve that cleanly?
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Huh, why would one want to store two copies of them? Get rid of the
one from async_struct. That structure is going away as a whole soon.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

tty_wakeup is safe to be called from all contexts. No need to schedule
a tasklet for that. Let's call it directly like in other drivers.

This allows us to kill another member of async_struct structure. (If
we remove the dummy uses in simserial.)
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

First, remove unused macro and rs_multiport_struct structure. Nobody
uses them at all.

Further, the 2 drivers (they are below) which use the rest of
structures from serialP.h (async_struct and serial_state) do not use
all the members. Remove the members:
* which are unused or
* which are only initialized and never used for something real.

Everybody should avoid the structures with a looong distance.

Finally, remove the ALPHA kludge MCR quirks. They are 1:1 copy from
8250.h. No need to redefine them here.

The 2 promised users of the structures:
arch/ia64/hp/sim/simserial.c
drivers/tty/amiserial.c
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The structures there are going away. And speakup has enough troubles
already.

So define a structure similar to what 8250 does: old_serial_port.
There define an array of speed, port base and so on needed for
configuration. Then use this structure instead of serial_state defined
in serialP.h.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: William Hubbs <w.d.hubbs@gmail.com>
Cc: Chris Brannon <chris@the-brannons.com>
Cc: Kirk Reiser <kirk@braille.uwo.ca>
Cc: Samuel Thibault <samuel.thibault@ens-lyon.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

All of them do not use the ugly interface defined in that header.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It uses pointers to pci_dev, but compiler complains it doesn't know
it:
In file included from .../m32r_sio.c:53:
.../m32r_sio.h:21: warning: "struct pci_dev" declared inside parameter list
.../m32r_sio.h:21: warning: its scope is only this definition or declaration, which is probably not what you want
.../m32r_sio.h:22: warning: "struct pci_dev" declared inside parameter list
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

We want to know the value of the atomic variable in intr_connect after
the increment. But atomic_inc doesn't, per definition, return the
value.  It is just a pure coincidence that ia64 defines atomic_inc as
atomic_inc_return.

So fix this mistake by using atomic_inc_return properly.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Even though the port is not used for anything real there yet, this
will change as tty buffers will be in tty_port in the near future. So
the port will be needed in all drivers.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

If the timer ticks while we are holding the spinlock, the system
deadlocks. It is due to synchronous del_timer.

So to fix that, use spinlocks that properly disable bottom halves.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Use setup_timer instead of explicit assignments.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Again, no need to do that from the pci probe function.

Hmm, I noticed this driver is marked as BROKEN. Won't touch it more,
it has to be converted to dynamic tty driver allocation first.

Perhaps it is time to move it to staging?
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: netdev@vger.kernel.org
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

They are in .bss which is initialized to zeros when the module is
loaded/kernel booted.

What a strange way to do the initialization once in the pci probe
routine...
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: netdev@vger.kernel.org
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

* do not test if tty->index is in bounds. It is always.
* tty->index is not a minor! Fix that.

>From now on, let's assume that the parameter of the function is tty
index with base being zero. This makes also the code more readable.

Factually, there is no real change as tty_driver->minor_start is zero,
so the tests are equivalent. But it did not make sense. And if this
had changed eventually, it would have caused troubles.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Acked-by: NJiri Kosina <jkosina@suse.cz>
Cc: David Sterba <dsterba@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Checking if tty->index is in bounds is not needed. The tty has the
index set in the initial open. This is done in get_tty_driver. And it
can be only in interval <0,driver->num).

So remove the tests which check exactly this interval. Some are
left untouched as they check against the current backing device count.
(Leaving apart that the check is racy in most of the cases.)
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The macro is always defined now. This was there only for historical
reasons.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This is needed because the tty buffer will become a tty_port member
later. That will help us to wipe out most of the races and checks for
the tty pointer in hot paths.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The test and the assignment were racy. Make it really a singleton.
This is achieved by one global variable initialized at the module
init.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It makes the code more readable. We move the setup to the allocation
location because we need to initialize timers only once.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Note that tty->ops->shutdown is called from whatever context the user
drops the last tty reference from. E.g. if one takes a reference in
an ISR, tty close happens on other CPU and the final tty put is from
the ISR, tty->ops->shutdown will be called from that hard irq context.

We would have a problem in vt if we start using tty refcounting from
other contexts than user there. It is because vt's shutdown uses
mutexes. This is yet to be fixed.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Reported-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

By using ASYNC_SPD_MASK instead of the single speed bits.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This is from tty_reopen:
    struct tty_driver *driver = tty->driver;
    ...
    tty->driver = driver;
and it doesn't make sense at all. The driver is intended to be set in
initialize_tty_struct from tty_init_dev (initial open). So this set in
tty_reopen is not needed.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Remove the useless local variable and return the value itself.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

All num, magic and owner are set by alloc_tty_driver. No need to
re-set them on each allocation site.

pti driver sets something different to what it passes to
alloc_tty_driver. It is not a bug, since we don't use the lines
parameter in any way. Anyway this is fixed, and now we do the right
thing.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Acked-by: NTilman Schmidt <tilman@imap.cc>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It was added back in 2004 and never used for anything real. Remove the
only assignment in the tree as well.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Like the rest of the kernel, make a stub from alloc_tty_driver which
calls __alloc_tty_driver with proper owner. This will save us one more
assignment on the driver side.

Also this fixes some drivers which didn't set the owner. This allowed
user to remove the module from the system even though a tty from the
driver is still open.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

TTY buffer head and tail are initialized in tty_buffer_init. No need
to do it once again in initialize_tty_struct where tty_buffer_init is
called.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This is a piece I missed the last time.

Do not copy the functionality all over the tree. Instead, use the
helper the tty layer provides us with.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Acked-by: NOliver Neukum <oneukum@suse.de>
Acked-by: NAlan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

We leave the existing paste mess alone and just fix up the vt side of
things.
Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Some of this ventures into selection which is still a complete lost cause. We
are not making it any worse. It's completely busted anyway.
Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>