com(4): Note ttylock in comment, not tty_lock.

com(4): Note timecounter_lock in lock order comments.

Revert com.c:r1.381 because i386/qemu cannot boot. Pointed out by gson@n.o and martin@n.o.

Fix hang up writing /dev/console rarely in specific environments.

Some BMC seems to require these syncronous operations. If not,
it does not send transmit completion interrupts for some reason.

KNF a comment

ddb/db_active.h: New home for extern db_active.

This can be included unconditionally, and db_active can then be
queried unconditionally; if DDB is not in the kernel, then db_active
is a constant zero. Reduces need for #include opt_ddb.h, #ifdef DDB.

com(4): Nix quirky `integrate' macro.

Just use `static inline' like everything else.

com(4): Omit needless spltty in comstart.

This is called either via tp->t_oproc, which is done with tty_lock
held (thus, at IPL_VM = IPL_TTY), or from comparam which is called at
IPL_TTY, either via comopen or tp->t_param.

com(4): Nix dead code.

This was introduced in rev. 1.292, whose commit message was...
`Delete dead code.' I guess December 6, 2009 was Opposite Day?

com(4): Comment on lock order.

com(4): Update confusing comment from decades ago to reflect now.

No functional change.

com: Add support for 32-bit IO accesses.

For the DW APB busy bit workaround, only attempt to re-apply LCR and DLB
settings for non-console devices. In the console case, simply clear the
busy bit and continue. Fixes an interrupt storm observed on Macchiatobin.

Restore 16-byte FIFO for ns16550a, accidentally lost in r1.344.

Spotted by msaitoh. Thanks!

- microtime -> microuptime
- avoid kpause with timeo=0

- use microtime instead of getmicrotime, suggested by thorpej
- use ttclos for wchan even though we are sleeping in comopen now

Note that while sleeping (HUPCL), time passes.
Try to avoid the otherwise infinite loop.

Now Jason has made this build, avoid a 50+ year
sleep to implement HUPCL.

com: speed up close with HUPCL set

Instead of incurring a 1s penalty on close of a com device with HUPCL set,
defer the sleep until the next open, and only sleep if necessary.

This has a side effect of making `ttyflags -a` with a default install not
pause for 1s for every non-console com device, which happens every boot
via /etc/rc.d/ttys.

only read cpr register if we're going to use the value

com(4): fix FIFO for DW_APB on Allwinner A20 (got broken by com.c 1.360)

Older DesignWare UARTs do not advertise their FIFO length so we must
provide it via device properties.

branches: 1.363.2;
Introduce COM_HW_BROKEN_ETXRDY bit in sc_hwflags; push TX queue if
possible in comintr() even if IIR_NOPEND is asserted.

Combined with sc_poll_ticks, this flag works around HW bug, by which
ETXRDY interrupts are (sometimes) lost.

Introduce sc_poll_ticks and obsolete COM_HW_POLL bit in sc_hwflags.
Polling is scheduled at every sc_poll_ticks ticks.

This is useful to work around H/W bug, by which interrupts are lost
*sometimes*; interrupt-based I/O mostly works and no need for polling
every counter ticks.

branches: 1.361.2; 1.361.4;
Set sc_fifolen=1 for the no fifo case

Auto-detect DW APB UART FIFO length and print the FIFO length when enabled
instead of just "working fifo".

Make the readahead-while-sending-output code conditional in
com_common_putc(), only erquest readahead before interrupts are enabled
and when called from comcnputc().
Fixes PR kern/55286.

Get rid of needless __predict_true() and move a comment slightly.

Use designated initializers

branches: 1.356.2;
in many device attach paths, allocate memory with M_WAITOK instead of M_NOWAIT
and remove code to handle failures that can no longer happen.

Simplify regmap initialization, and fix an regmap issue that
affected TI OMAP (LCR register would get clobbered due to
using the wrong offset for the MDR1 register) reported by Lwazi Dube
(who also found the root cause).

Add a convenience function, com_init_regs_stride(), that shifts the register
offsets and size by the specified amount. Use in front-ends as appropriate.

sparc64 doesn't have a scalar bus_space_handle_t, so cope with this.
Also, add a comment about why we're using the dummy / nil space handle
in the first place.

Remove the COM_REGMAP option -- just use it all the time. While here,
garbage-collect the COM_FUNCMAP and COM_AU1X00 options, as there are
not used anywhere.

Clean up initialization of com_regs structure, in preparation for
some additional changers.

Add support for com(4) without an interrupt by setting the COM_HW_POLL flag

Add support for a "force_console" property, where when set, reuses the console dev configuration

branches: 1.348.2;
Rename COM_TYPE_SUNXI -> COM_TYPE_DW_APB

Remove COM_AWIN option

branches: 1.346.2;
Put back SUNXI definitions in the COM_REGMAP case; com.c uses them
without #ifdef COM_AWIN.
To be safe, expand the register map to 42 entries.

As discussed on tech-kern: define a new tty internal state flag: TS_KERN_ONLY

Implement it in a few tty drivers. If this flag is set, the underlying
hardware is used by another driver and userland has no right to open
it. A few uses will appear soon in sys/dev/sun/sun{kbd,ms}.c.

Make all of the COM_xxx type options runtime selectable. Kernel configs
with the existing options (COM_16650, COM_16750, COM_AWIN, COM_HAYESP, and
COM_PXA2X0) will select the correct type in com_attach_subr. New code
should specify the com type by passing COM_TYPE_xxx to comcnattach and/or
setting sc_type.

Kill some more extern struct cfdriver declarations.

Down with externs in .c!

Report AUXUART fifo size on rpi3 to 1 byte. This prevents bluetooth
errors using the h5 protocol.

Broadcom AUX UART doesn't seem to set LSR RXRDY bit when data is available
in the RX FIFO.

Add COM_TYPE_BCMAUXUART type for BCM2835 AUX UART.

branches: 1.339.10;
The UART in the allwiner SoCs is not full-compatible with the 16550, and
it's not a 16750 either. Like the 16750 it has the IIR_BUSY interrupt,
which is triggered when writing to LCR while the chip
can't accept it. But unlike the 16750, it has a specific register,
HALT, to allow writing to the LCR and divisor registers, and then
commit the changes.
Tested on an A20 SoC, changing the baud rate while keeping the
tty device open and incoming data.

Fix TX FIFO on Tegra

PR/50395: Vicente Chaves de Melo: Loss of characters in serial port
communication between gdb and kgdb_stub: switch to line break interrupt
both when entering KGDB and also on shutdown so we can re-enter the debugger
with BREAK.

For Tegra K1, set IE_RX_TIMEOUT (bit 4) in IER register. RX_TIMEOUT occurs
when data has been sitting in the Rx FIFO for more than 4 character times
without being read because there is not enough data to reach the trigger
level. With this change, enable FIFO usage for Tegra UARTs.

fix pasto, use SET() and CLR()
thanks jmcneill@

add COM_TYPE_TEGRA

Convert sys/dev to use <sys/rndsource.h>.

KNF

Fix build.

Hi mipsallan

more Ingenic support:
- make sure we always set FIFO_UART_ON
- deal with the absence of DCD support
- enable the TX FIFO timeout interrupt
- set COM_HW_NOIEN
- pretend the FIFO is only 16 bytes deep ( supposed to be 64 but I get
overruns with that )
now this works as CI20 console

branches: 1.329.2;
deal with quirk in Ingenic UARTs
( they have a bit in the FIFO control register which turns the entire
port off if not set )

centralize the dialout/call unit macros.

branches: 1.327.2;
Merge tls-earlyentropy branch into HEAD.

Add d_discard to all struct cdevsw instances I could find.

All have been set to "nodiscard"; some should get a real implementation.

Clear dtr on transition from !0 to 0, and set it on the opposite.
This leaves dtr alone in the normal case.

PR/48720: John Kelly: com driver does not allow B0 to be set. The code to
handle B0 was commented out for no reason given at revision 1.99. POSIX
mandates to hangup on B0:
http://pubs.opengroup.org/onlinepubs/9699919799/functions/tcsetattr.html

branches: 1.323.2;
Change (mostly mechanically) every cdevsw/bdevsw I can find to use
designated initializers.

I have not built every extant kernel so I have probably broken at
least one build; however I've also found and fixed some wrong
cdevsw/bdevsw entries so even if so I think we come out ahead.

Now that cngetc supports -1 as a return value from the cn_getc routine,
let cngetc loop until a character is available so it can call the critpoll
hooks.

Use #if defined in previous.

Fix build.

lcr is only used when COM_16650 is defined; unbreak the build

Only attempt to set 64byte fifo for 16750 when COM_16750 is defined.

Hi kiyohara.

PR/48359 NetBSD/cobalt 6.99.25 GENERIC kernel freeze at initializing com0

Revirt 'Move the Marvell extension to com_mv.c' at Sun Sep 1 04:51:24 UTC 2013.
build test only.

Minor tweaks for newer gcc

no EFR on tl16c750, use AFE bit on MCR reg instead for flow control

Add support 16750 64Byte FIFO. But not test.

Move the Marvell extension to com_mv.c.

As for the console of ns16750, comintr() may be called before comopen().
If sc_lcr is not initialized at this time, console will be in an ugly state.

Remove white-spaces, null-line and new-line.

branches: 1.310.4;
For a polled console incoming characters are buffered on output.
The buffer is never passed to the regular tty routines and can
survive from early boot to halt where it is read by the "press
any continue to reboot" loops.
As a workaround, just kill the read-ahead buffer when switching
from and to polled mode.

Add support for 16750 style UARTs. Activated by defining COM_16750.

Obtained from Marvell, Semihalf.

Initialize t_softc.

don't try to become console if we get is_console=false as device property

branches: 1.306.2;
Remove empty line.

don't include <sys/rnd.h> before we define cn_trap, because it might bring
in <sys/systm.h>

branches: 1.304.2;
Entropy-pool implementation move and cleanup.

1) Move core entropy-pool code and source/sink/sample management code
to sys/kern from sys/dev.

2) Remove use of NRND as test for presence of entropy-pool code throughout
source tree.

3) Remove use of RND_ENABLED in device drivers as microoptimization to
avoid expensive operations on disabled entropy sources; make the
rnd_add calls do this directly so all callers benefit.

4) Fix bug in recent rnd_add_data()/rnd_add_uint32() changes that might
have lead to slight entropy overestimation for some sources.

5) Add new source types for environmental sensors, power sensors, VM
system events, and skew between clocks, with a sample implementation
for each.

ok releng to go in before the branch due to the difficulty of later
pullup (widespread #ifdef removal and moved files). Tested with release
builds on amd64 and evbarm and live testing on amd64.

branches: 1.303.2;
more whitespace and comment spelling fixes

whitespace consistency adjustment

branches: 1.301.4;
Allow COM_TOLERANCE to be tweakable. If comspeed returns an invalid
rate, don't use that error value to set the speed.

Rename ttymalloc() to tty_alloc(), and ttyfree() to tty_free() for
consistency. Remove some unnecessary malloc.h inclusions as well.

Revert part of changes in rev 1.298:
- it breaks cobalt's serial console as mentioned in PR port-cobalt/44292
- MCR_PRESCALE doesn't affect unless EFR_EFCR is set in the EFR register
- even if MCR_PRESCALE is enabled we should define appropriate sc_type
variants and BRG values should be adjusted in comspeed() per sc_type
- sc_frequency should be adjusted in MD attachment if necessary
Tested on cobalt by several people, ok from jklos@

branches: 1.298.2; 1.298.4;
Add support for 16650's prescaler from Frank Wille. Tested with
IOblix on Amiga.

Add default implementations for bus_space_is_equal(9),
bus_space_tag_create(9), and bus_space_tag_destroy(9). Use
bus_space_is_equal(9) throughout the kernel to compare
bus_space_tag_t's. Tested on i386 and on sparc64.

MI code should not compare bus_space_tag_t! Introduce tags_are_equal()
for "comparing" two bus_space_tag_t's. It is always true.

Everywhere that com(4) compares two tags, it compares to I/O base
addresses, too; comparing the base addresses should suffice.

TBD Clean this up more thoroughly.

branches: 1.295.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.294.2;
To print bus_addr_t value, cast it to u_long rather than size_t.

Expand PMF_FN_* macros.

Delete dead code.

Remove some unecessary includes sys/user.h header.

Remove some dead code.

Move the code in com_activate() to com_detach(), where it always
belonged. Remove com_activate().

Consolidate information about the console on com(4) in a new struct
comcons_info.

Support detachment & re-attachment of a system console on com(4).
Re-attachment is somehow incomplete. Ideally, if some other device
could take over as console, it would, but we're not quite there,
yet.

add KASSERT to identify when PR kern/40734 happens again.
It happened to me twice and to Emmanuel Dreyfus (reported on port-amd64 ML)

branches: 1.287.2;
remove extra semicolons.

branches: 1.286.2; 1.286.10; 1.286.14;
Simplify NOERS case. Revert most changes back to 1.283.

Fix a botch in an FIFO check for 16650.

Replace a "type" with "sc->sc_type" to make this build again for cobalt
(and I'm sure a few others as well).

Add support for 16550 chips without an Enhanced Register Set.

Remove OMAP gunk in com_activate, it's not needed.

Spotted by Matthias Drochner.

branches: 1.281.2; 1.281.4; 1.281.6;
More forgotten changes from matt-armv6.

Remove clause 3 and 4 from TNF licenses

branches: 1.279.2;
Make ntp, pmc, reboot, sysarch, time syscalls MPSAFE.

branches: 1.278.2;
If it's a high speed, trim the trigger level down to prevent overflows,
as comment says. The piece of code was removed in rev 1.230,
but it looks an error on reverting of rev 1.228.

Fix compilation under options RND_COM and COM_DEBUG.
Reported by Kurt Schreiner on current-users@.

Split device_t and softc for all com(4) devices (well, everything that
uses a com_softc backend). Use proper types and ansify where appropriate.

In com_detach(), don't destroy the lock twice.

Use device_t and its accessors.

Use pmf_device_register1() instead of shutdownhook_establish() to
register com_cleanup() as the shutdown hook.

Add a generic suspend routine. Suspend and resume com@isa.

Protect against dereferencing a NULL softc in comioctl().

Destroy both a mutex and a callout in com_detach().

Cosmetic: use aprint_*_dev(). Use PMF_FN_ARGS, PMF_FN_PROTO.

branches: 1.272.2; 1.272.6;
Do not allow the system console to detach, because that will lead
to a panic.

Destroy mutex when detaching. PR kern/37819.

Now that __HAVE_TIMECOUNTER and __HAVE_GENERIC_TODR are invariants,
remove the conditionals and the code associated with the undef case.

Work around com hangs with pcmcia. PR kern/37462.

Add a resume handler for com@isa, using Jared McNeill's suggested
code from current-users. com@acpi, et cetera, remain to be done.

branches: 1.267.4;
Match the docs: MUTEX_DRIVER/SPIN are now only for porting code written
for Solaris.

branches: 1.266.2;
- Factor out too many copies of the same bit of tty code.
- Fix another tty signalling/wakeup problem.

branches: 1.265.2;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.264.2;
Reapply rev 1.262 and add a comment about it.

Merge from vmlocking: use mutexes and the softint API.

branches: 1.262.2; 1.262.4;
Remove all locking from com_enable_debugport() - it is either called at
attach time, where we don't need further locking, or from comopen(),
where the caller does proper locking already.

branches: 1.261.2; 1.261.6;
Generic soft interrupts are mandatory.

Merge some of the less invasive changes from the vmlocking branch:

- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements

branches: 1.259.2; 1.259.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.258.2;
Make COM_HAYESP compile again. It assumes no register mapping is needed.

aprint_naive("\n") to terminate "Found" line for quiet boot.
Use aprint_error to report that we are unable to establish power hook.

__unused removal on arguments; approved by core.

- sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386

More from Matt Fleming:

Adapt to KAUTH_DEVICE_TTY_PRIVSET and KAUTH_DEVICE_TTY_OPEN.

Adapt MD code to KAUTH_DEVICE_TTY_OPEN, batch #2 from Matt Fleming, thanks!

Also, add forgotten splx() calls in some places.

Add "name" parameter to powerhook_establish, to aid debugging. No objections
on tech-kern@

branches: 1.251.2; 1.251.4;
fix a GCC uninitialised variable warning only seen at -O3.

branches: 1.250.2;
oops, no 'p' here. from xtraeme@.

don't call kauth_authorize_generic() when holding a lock, from yamt@.

- Use the LWP cached credentials where sane.
- Minor cosmetic changes.

Add an option COM_REGMAP to allow com(4) to use an array of register indices.
This allows us to convert aucom to just another com attachment, and cleanup
some code in the com_arbus.c.

Additionally, we use a common com_cleanup routine rather than having a
zillion copies of it in the attachment points.

This has been tested on a number architectures, and it has been shown to get
close to comparable performance when COM_REGMAP is defined, and comparable
when it is not defined.

Approved by core@. Fixes PR port-evbmips/32362.

Make sure that a powerhook was successfully established before trying to
disestablish it on detach.

Add a power hook for the com device.

branches: 1.244.2;
merge FreeBSD timecounters from branch simonb-timecounters
- struct timeval time is gone
time.tv_sec -> time_second
- struct timeval mono_time is gone
mono_time.tv_sec -> time_uptime
- access to time via
{get,}{micro,nano,bin}time()
get* versions are fast but less precise
- support NTP nanokernel implementation (NTP API 4)
- further reading:
Timecounter Paper: http://phk.freebsd.dk/pubs/timecounter.pdf
NTP Nanokernel: http://www.eecis.udel.edu/~mills/ntp/html/kern.html

branches: 1.243.2;
integrate kauth.

Use device_unit().

branches: 1.241.2; 1.241.4; 1.241.6;
Use device_is_active() rather than testing dv_flags for DVF_ACTIVE
directly.

branches: 1.240.2; 1.240.4;
Save console baud rate as both the input and output rates.
This is what I thought I'd done in rev 1.237

branches: 1.239.2;
remove the COM_MPLOCK option. always include the spinlock in the softc
and always call the simple_* locking functions. the locking functions
are compiled out if they are not needed anyway, so a separate option
for this doesn't gain anything.

this also fixes the serial console on my alpha ES40 (which doesn't make much
sense since the com driver should still be under the big lock on alpha,
but whatever).

merge ktrace-lwp.

Explicitly default the input baud rate (c_ispeed) to the output baud rate (c_ospeed)
instead of setting it to zero. Otherwise nothing ever sets it unless some code
explicitly changes the baud rate. For a serial console (in particular) we want
to use the baud rate set by the bios (or whatever) and used by theboot code.
This is the way it was before a 'new version of com driver' was added in 1997 (rev 1.99)

Change the driver open function's conditional for overriding exclusive tty
use from checking the proc's uid to suser(9), and account for the use of
privileges. Noted by David Holland in PR kern/31126.

Make COM_DEBUG compile again in the wake of -Wwrite-strings;
from Nicolas Joly in PR kern/31141.

branches: 1.234.2;
PR-30566: Poll must not return <sys/errno.h> values.
Start with those places I can easily test.

Revert the part of revision 1.228 that made HW_NO_TXPRELOAD the default:
it appears to break output on the Soekris net4501, which is a rather
popular platform.

This should fix PR kern/29612 -- if not, I will probably revert it again.

branches: 1.232.2;
nuke trailing whitespace

branches: 1.231.4; 1.231.6;
Do an extra read of the IIR during initialization, just to ack any pending
interrupt.

Back off the trigger level a little. Testing shows that it's not terribly
reliable at 14.

Minor simplification.

Incorporate 4 changes that have been suggested:
* At high speed, set the FIFO trigger to the maximum value to reduce interrupt
load. PR 15448.
* Handler the case of IIR=RXRDY and LSR=!LSR_RCV_MASK differently, to avoid a
hang with some chips. PR 21184, and possibly PR 10974 and PR 8248.
* Remove the extra wait at the end of com_common_putc(), after writing a
character out, so that we get another TXRDY interrupt. This prevents stalls
when normal output and kernel output are mixed. PR 4263.
* Do not "preload" the TX FIFO; wait for a TXRDY interrupt to come in after
enabling it.

Rename the COM16650 option to COM_16650, for consistency with other
com variant options.

Wait just a little longer for console output to finish when we're
attaching the console uart.

Fix race condition introduced in rev 1.189; after the change, if there is
THRE interrupt occurs between the LSR read and IIR read, we won't see the
LSR_TXRDY bit when testing it in the variable `lsr' and we don't interrupted
again (as the corresponding bit in the IIR is cleared by reading, except
for some broken device).

Tested by Matthias Scheler and me, reviewed by Allen Briggs.
Closes PR#25010.

branches: 1.224.2;
Fix NTP PPSAPI support (enabled with "options PPS_SYNC"):

From PR kern/13702 from Charles Carvalho. Tested on alpha and
i386 with a Laipac TF10 PPS-capable GPS. The com.c change was
copied wholesale from Charles' z8530tty.c patch.

Use just "type" and not "sc->sc_type" in cominit().

Pointed out by Shoichi Miyake in private mail.

Make comsoft_callout static; it's not referenced outside this file.

Be consistent with other COM_TYPE_PXA2x0 checks.

White space nit.

Add missing semicolon.

fix debugging printfs

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

branches: 1.216.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.

Restore parentheses around return value that were removed with the
previous commit.

Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.

Bump the kernel rev up to 1.6V

Make sure to include opt_foo.h if a defflag option FOO is used.

in cominit(), turn on UUE bit in IER register if it is PXA2x0's
built-in UART.

Wrap some long lines.

Also pass a type argument to comcnattach() and com_kgdb_attach().
comspeed() (and thus cominit()) may need this information.

Introduce a new "type" field in the softc which is used to indicate
the general UART type. Currently, we have "normal", "hayesp" and
"pxa2x0" types. Replace the corresponding hw flags with the new type
values.

Add support, via a kernel config file option, for the 16550-compatible
com ports found on Intel PXA2x0 processors.

Re-arrange the code which detects the FIFO configuration to defer
printing until _after_ com_fifo is cleared. Some supposedly compatible
chips clear the Tx/Rx FIFOs when this happens, resulting in previous
output being lost.

Inialize sc_tty->t_dev early for serial console, so that cn_check_magic
will match in the case of a break before the console was opened for the
first time. Makes BREAKing into DDB work (again) as soon as interrupts
are enabled.

Allow the console to be used for kgdb if both DDB and KGDB are defined.

Ignore input when the port is not open. This change is functionally
identical to the patch in kern/6502 (which is hereby fixed), but
modelled after the corresponding fix to the 8530 driver in
z8530tty.c revision 1.60 to maintain consistency with that driver.

Fix some white-space niggles.

Add cn_halt and cn_flush entries to consdevs. (needed for dma-only console
devices).

Use aprint_*().

Fix signed/unsigned comparison warnings.

merge kqueue branch into -current

kqueue provides a stateful and efficient event notification framework
currently supported events include socket, file, directory, fifo,
pipe, tty and device changes, and monitoring of processes and signals

kqueue is supported by all writable filesystems in NetBSD tree
(with exception of Coda) and all device drivers supporting poll(2)

based on work done by Jonathan Lemon for FreeBSD
initial NetBSD port done by Luke Mewburn and Jason Thorpe

Merge the gehenna-devsw branch into the trunk.

This merge changes the device switch tables from static array to
dynamically generated by config(8).

- All device switches is defined as a constant structure in device drivers.

- The new grammer ``device-major'' is introduced to ``files''.

device-major <prefix> char <num> [block <num>] [<rules>]

- All device major numbers must be listed up in port dependent majors.<arch>
by using this grammer.

- Added the new naming convention.
The name of the device switch must be <prefix>_[bc]devsw for auto-generation
of device switch tables.

- The backward compatibility of loading block/character device
switch by LKM framework is broken. This is necessary to convert
from block/character device major to device name in runtime and vice versa.

- The restriction to assign device major by LKM is completely removed.
We don't need to reserve LKM entries for dynamic loading of device switch.

- In compile time, device major numbers list is packed into the kernel and
the LKM framework will refer it to assign device major number dynamically.

Don't name parameters in a function declaration.

ANSIfy, KNF.

branches: 1.196.2; 1.196.4;
PR/16058: Tero Kivinen: Don't free rbuf or detach tty if rbuf allocation
failed during attach phase.

* Add a new hwflag, COM_HW_NO_TXPRELOAD, which tells comstart() to
not pre-load the chip's Tx buffer, but instead waits for the Tx Ready
interrupt to transmit the first chunk of data.
* On the IOP310, set COM_HW_NO_TXPRELOAD, rather than COM_HW_TXFIFO_DISABLE.

This solves the "UART hangs" problem on the Npwr in a nicer way (i.e. we
get to use the FIFO, whee). The COM_HW_NO_TXPRELOAD happens to match the
Linux 16550 driver's Tx algorithm, and the "UART hang" was never observed
on the Npwr running Linux.

Eventually, we might want to eliminate the COM_HW_NO_TXPRELOAD, and simply
always use its algorithm. But it should be tested on more 16x50 variants
before we do that.

Kudos to Valeriy Ushakov <uwe@netbsd.org> for pointing out this solution
(which also happens to fix the stray UART interrupt issue on the Krups
Javastation), and to Allen Briggs <briggs@netbsd.org> for experimenting
with various methods of fixing this.

Convert ioctl code to use EPASSTHROUGH instead of -1 or ENOTTY for
indicating an unhandled "command". ERESTART is -1, which can lead to
confusion. ERESTART has been moved to -3 and EPASSTHROUGH has been
placed at -4. No ioctl code should now return -1 anywhere. The
ioctl() system call is now properly restartable.

- pull in opt_kgdb.h where necessary
- replace opt_kgdb_machdep.h with opt_kgdb.h
- defparam opt_kgdb.h:
KGDB_DEV KGDB_DEVNAME KGDB_DEVADDR KGDB_DEVRATE KGDB_DEVMODE
- move from opt_ddbparam.h to opt_ddb.h:
DDB_FROMCONSOLE DDB_ONPANIC DDB_HISTORY_SIZE DDB_BREAK_CHAR SYMTAB_SPACE
- replace KGDBDEV with KGDB_DEV
- replace KGDBADDR with KGDB_DEVADDR
- replace KGDBMODE with KGDB_DEVMODE
- replace KGDBRATE with KGDB_DEVRATE
- use `9600' instead of `0x2580' for 9600 baud rate
- use correct quotes for options KGDB_DEVNAME="\"com\""
- use correct quotes for options KGDB_DEV="17*256+0"
- remove unnecessary dependancy on Makefile for kgdb_stub.o
- minor whitespace cleanup

don't need <sys/types.h> when including <sys/param.h>

add/cleanup RCSID

Initialize the cn_magic goo for kgdb (bah, that should really not
be necessary).

branches: 1.189.2;
Patch to handle odd behavior from exar ST16C1550 UART. It does not set
IIR_NOPEND if the TXRDY condition is true. Apparently, other chips do
not take TXRDY into account in the IIR_NOPEND case.

Also remove useless assignment (iir = IIR_NOPEND;) in one case.

No response from patch posted on tech-kern 29Aug2001.

branches: 1.188.2;
For hayes esp card, set prescaler if specified rate is grater than 115200.

Cosmetic cleanups:
- cominit() is local to com.c, remove prototype from comvar.h.
- Don't prototype comstop(), cdev_decl(com) does this.
- Don't use a block device switch table declaration(!).
- No need to include "com.h".

branches: 1.186.2;
Convert to use cnmagic(9). Unifdef DDB and DDB_BREAK_CHAR.
While I'm here, s/ungetc/readahead/ in com_common_* because that's
what it _really_ is, "ungetc" is a very confusing name.

add missing #include "opt_kgdb.h"

Add `l_poll' to `struct linesw' and provide an xxxpoll() entry point
in each tty driver to indirect through it.

This allows tty line-disciplines to handle poll(2) system calls.

branches: 1.183.2;
Rename __GENERIC_SOFT_INTERRUPTS to __HAVE_GENERIC_SOFT_INTERRUPTS,
and place the definition in <machine/types.h>. This can now be used
as a flag to indicate whether or not <machine/intr.h> can be included
to get the generic soft interrupt API.

Initialize ring buffer pointers when the ring buffer is allocated.
(comsoft() can be invoked before comopen() on serial consoles; a character
received before the console is opened "for real" can result in a fatal trap
unless the ring buffer pointers are initialized early)

Adapt to the new line discipline scheme.

Add missing '&& defined(COM_MPLOCK)'

Add `COM_MPLOCK' option to use a device-instance-specific spinlocks
when running at splserial(). This is a temporary measure (until
there's a MP-safe interrupt handling structure); until then, it should
be used when MULTIPROCESSOR and IPL_SERIAL > IPL_SCHED.

paddr_t -> bus_addr_t.

Support for SPARC machines with `su' devices.

Change tiocm_to_{com,zs}'s second argument to u_long to match with
the type in their caller.

Avoid calling tsleep when running above splhigh()

Add options DDB_BREAK_CHAR. This overrides break on the serial console
break character with the supplied one. This is useful for cases where
break is hard to generate, or you are connected to a PC that "sends"
breaks when power cycled. For this mode in com, interpret break char
in the polling section, which allows entry into the debugger before
the tty is opened. Only supported in the com driver currently.

Use device_lookup().

Remove unnecessary junk on #undef line that some C preprocessors
get cranky about.

Delete redundant decl of com_attach_subr(), it's in comvar.h.

New callout mechanism with two major improvements over the old
timeout()/untimeout() API:
- Clients supply callout handle storage, thus eliminating problems of
resource allocation.
- Insertion and removal of callouts is constant time, important as
this facility is used quite a lot in the kernel.

The old timeout()/untimeout() API has been removed from the kernel.

- Implement cnbell() -- ring the console bell. The cn_bell entrypoint
is optional.
- Add cn_bell to statically allocated consdevs as appropriate.

Fix thinko in previous revision: it wouldn't catch deassert.
Revert the code inside the new test to the previous-but-one revision.

Check for DCD status change interrupts which signal PPS events, even
if the line discipline is ignoring carrier (e.g., via comparam()'s
setting of sc->sc_msr_mask). Move PPS timestamp outside the normal
status-change processing, and ignore sc_msr_mask when testing
for PPS events.

kluge to make this work with NEC M403 (arc/magnum).

set carrier early if used as console or kgdb line, fixes PR kern/6547
by Andreas Gustafsson, gson@iengines.com

defopt COM_DEBUG, COM_HAYESP, and COM16650

Update user-level PPS-API to match -05 draft.
Change PPS_<xxx> ioctls to PPS_IOC_<xxx>.

branches: 1.162.2; 1.162.4;
Fix nanosecond/microsecond botch in PPSAPI:
remove bogus TIMESPEC_TO_TIMEVAL() after applying any {assert,clear} offset.

branches: 1.161.2;
Get rid of delay() in com_common_putc() because the Shark delay() is borked
and can't be called until after the timers are setup. (It should not use
a timer, or it should have a fallback algorithm.)

Revert revision 1.159. The correct answer was not to re-break the
serial console support, but rather to fix the Shark delay().

branches: 1.159.2;
back date this file to version 1.155 which works on the shark.

Nuke bogus extra variable, too.

Fix bogons in previous change:
* The fact that IIR_NOPEND was not set on entry does *not* mean that no
transmission was in progress. Besides, we don't want to throw away receive
interrupts either.
* In the !clearirq case, we didn't splx().

Mods to com_common_putc(). When doing kprintf() or kgdb output:
1. don't clear the irq unless it was clear before transmitting
2. also do various bus_space_barrier() ops
Stops console from freezing when kprintf interrupts tty driver output.

Update to slightly altered rnd_attach_source() api

make the probe stricter to catch cases where the chipset doesn't pull
the data lines to "high" for not-existant devices

Separate the tiocm code into functions, like the z8530 code, and fix the same
DTR bug.

Fix an inconsistency: PPS was turned off on first open, but the mode may have
still claimed it was enabled.

Slight redux in PPS code.

Debugger() -> console_debugger() to honour the ddb_fromconsole sysctl

Add support for detaching "com" instances.

Garbage collect unused 3rd parameter to comprobe1().

Don't use M_WAITOK to allocate the ring buffer; we could be in interrupt
context. Reported by Lennart Augustsson.

Assign my copyrights to TNF.

defopt DDB.

PPS support from April, reworked and extended to match the PPS-API
Internet Draft (draft-mogul-pps-api-02.txt).

Move the code to wait for carrier on a tty into a common function, since it
depends only on device-independent state bits.
Implement SunOS-style `dialout' devices.

Remove a bogus initializer.

Replace TS_WOPEN with t_wopen, per mail on tech-kern.

Add dummy }} after comsoft() to keep ctags(1) bashing on.

Disable the device if we fail to open and nobody else has the device open.
(Thanks Charles and Marc for your advices).

Fix bonehead error in previous change.

Remove stray label from last change.

Arrange to call com_break() and com_modem() at splserial().
It's simpler this way. Really.

Fix serial console support, which was broken by the previous revision:
- In the console getc routine, block until a character becomes ready
(no, really, we mean it). This routine should _never_ time out.
- In the console putc routine, if the UART has trouble, do NOT proceed
to print a diagnostic message, which would recursively invoke the
console putc routine ad nauseum.

Restore system priority level before return from comopen().

change #ifdefs keying on "alpha" to key on "__NO_SOFT_SERIAL_INTERRUPT"
instead, and define that on NetBSD/alpha and NetBSD/arm32.

Implement a flag to indicate that the driver attached correctly.
This is useful in the case where an attachment's probe routine
verifies that there is indeed hardware present but something goes
"wrong" in the attach causing the device to be unusable. (Without
keeping track of this, in that case incorrect ports could be
accessed or uninitted pointers could be deferenced on open or at
other times.)

split com_attach_subr into attach stuff and per-open initialization stuff
add timeouts to all splserial() while loops so bugs elsewhere don't cause
the com driver to wedge the machine
add power management support

Adjust for config changes.

Don't do anything in comparam() if the line speed and flags are the same.

Slight optimization in soft input routine.

Mostly cosmetic and performance changes:
* Make the ring buffer size and water marks patchable, and allocate the buffer
separately.
* Do the ttymalloc() at attach time.
* Reorganize the receive buffer so the status and data pair are next to each
other. This is slightly faster.
* Make sure we actually do turn off interrupts in comclose() if we have DDB
configured and it's not the console. (D'oh!!!!)
* When we exhaust the current transmit run, turn off transmit interrupts in
comintr(), so we're fairly sure we don't get another one.
* Nuke the silly lsrmap[] idea; it's slower in the normal case.
* Cache the l_rint pointer in the soft interrupt routine.

Slight code rearrangement.

Slight code rearrangement.

Rearrange *param() slightly to eliminate a spl.
Also fix another comment regarding DTR.

Add another splserial()/splzs().

Minimize some spl regions slightly.

Correct a comment regarding our behaviour with DTR.

Add additional splserial()s in comopen() and comclose(). These are not
necessary on the i386 port, but may be on others.

Print probe result after the LCR is restored. Otherwise, bad
things happen if we are the console.
Restore to the prevoius value (not to hardwired 8N1) because this
could be set by the serial console initialization.
Closed PR kern/4373 (Dave Huang)

branches: 1.118.2;
Make 16650 probe optional on COM16650 define. It breaks on some
multiport Startech chips.

Initialize the IER to 0 in com_attach_subr().

Add 16650 hardware flowcontrol support. Basically PR 3844 from
Dave Huang <khym@bga.com>, with added check for broken early versions
of the 16650, taken from the Linux driver.

This should be extended to use, for example, higher trigger levels for
the bigger 16650 FIFO, and its capability for using a smaller divisor
and thus higher speeds. But this patch is very useful for 16650 users
already.

Add hooks for entropy collection. Right now, you must place
options RND_COM
in your kernel config file (along with the
pseudo-device rnd
line) to enable this. If results are positive, I will make this enabled
by default.

Use long long in bitrate tolerance computation, else we'll overflow for
higher than standard-pc chip clocks.

branches: 1.113.2;
"com" driver files copied by repository copy magick to sys/dev/ic, and
adjusted for their new home.

fix missing com_kgdb_attached() argdecl, from Michael C. Richardson in PR 4196

Add TIOCM{BI{C,S},{S,G}ET} from revision 1.98. As Charles says these
interfaces are slightly broken and should be obsoleted, but some programs
are using them and also they are useful for debugging.

Cosmetic Changes; make coding style closer to other part of this file.
(This version is identical with version 1.109 except whitespace changes.)

Support for the upcoming NetBSD/Amiga Hypercom driver family:
* support chip clocks != COM_FREQ, by introducing sc_frequency (for the
mainline code) and adding a frequency parameter right after the rate
parameter to comcnattach() and com_kgdb_attach().
- Make com_isa and com_multi initialize sc_frequency to COM_FREQ.
- Make i386/machdep.c and alpha/dec_xxx.c call com*attach() with the freq.
parameter.
* supio_attach_args get two more fields: a sc_ipl and a sc_arg, both ints.
- com_supio uses the first for interupt establishment (all childs will, as
soon as they exist) and the 2nd for sc_frequency.
- drsupio passes sc_ipl alway as 5, and for the "com"s, sc_arg as 16*115200
- hyper will pass sc_ipl as 6, and sc_arg as 16 * 460800

Make this work with __GENERIC_SOFT_INTERRUPTS.

Check for KGDB port == console port in a consistent way.

clean up the serial console / KGDB attachment:
-put all early console / KGDB initialization into 1 exported function
(com_*_attach()) each, dont use global variables anymore
-use the passed tcflag_t for port settings instead of hardwiring 8N1
-at autoconfiguration attach time, decide if the attaching device is
already console / KGDB by comparing bus tag and base addr (cgd's wish)
-export a function "com_is_console()" for use by driver frontends for
this comparision
-delay setting of cn_tab->cn_dev until autoconfiguration attach
to get the minor number right
-delete unused comcnprobe() and comcninit()

Bring back the "comconsattached" flag. It is necessary for starred
com devices (on non "__BROKEN_INDIRECT_CONFIG" ISA).

-Export variables needed for system console initialization.
-Enable line break interrupt for DDB on a serial console.
-Share initialization code between DDB and KGDB.

KGDB improvements:
-Separate KGDB port initialization and softc related stuff to allow
KGDB to be attached in early system startup, before autoconfiguration.
-Export the variables needed by md code to hand-craft bus tag/handle.
-Fix initialization to enable interrupt by line break.
-Call DDB/KGDB at line break (move DDB call from the softirq handler
to the hard handler because it should work without a tty attached too).

branches: 1.102.2;
Add support for remote serial KGDB.

Implement a better backpressure mechanism so that data is collected
in the driver's buffer if the tty buffer is full. If the driver's buffer
becomes full, turn off receive interrupts until it drains a bit.

Remove unneeded call to comstop() in commsrint(). This fixes a problem
where the driver would sometimes retransmit bytes.

New version of com driver. Uses a different queueing mechanism and a split
hardware/software interrupt mechanism for improved performance. Many odd
protocols bugs also fixed.

comdefaultrate -> comconsrate
comconsbah -> comconsioh
Only use comconsrate on the console.

don't clobber IER in cominit(). (suggested by mycroft.)

If attaching the console, reinitialize it immediately.

Turn off the baud rate changing optimization.

update these so they compile whether or not __BROKEN_INDIRECT_CONFIG
is defined.

move cominit() prototype to comvar.h, so that ports which use 'com'
for serial console, but don't use the (i386-specific)
comcninit()/comcnprobe() functions, can do the right thing.

add missing argument to bus_space_map in ESP code

New bus.h implementation/interface:
- No more distinction between i/o-mapped and memory-mapped
devices. It's all "bus space" now, and space tags
differentiate the space with finer grain than the
bus chipset tag.
- Add memory barrier methods.
- Implement space alloc/free methods.
- Implement region read/write methods (like memcpy to/from
bus space).
This interface provides a better abstraction for dealing with
machine-independent chipset drivers.

backout kprintf changes

printf -> kprintf, sprintf -> ksprintf

If we get a RXRDY interrupt, but RXRDY is not set in the LSR, briefly set IER
to 0. This fixes a condition where some UARTs send an infinite stream of
RXRDY interrupts.

Serial console changes:
* Enable FIFO with trigger level 1.
* Set DTR and RTS so terminals are happy.

compoll() -> comsoft()

tty stop functions really should return void, not int, and certainly not both.

move DDB-only label inside an #ifdef, so that -Wall works

add tty_attach() where appropriate.

branches: 1.82.4;
Use intr.h.

Move comintr() prototype to comvar.h; needed by the multi-port cards.

- prototype fixes

define and export the global variable "comconscflag", the default tty
'cflag' for the console. Normally set to TTYDEF_CFLAG, may be
overridden by machine-dependent console attachment code, as necessary.
(Alpha uses it to set cs8 -parenb.) Files including comvar.h now
need to include <sys/termios.h>, because comconscflag is of type
tcflag_t.

update for addition of a machine-dependent cookie as the first argument
to isa_intr_{,dis}establish().

Add #ifdef's at the right places, on NCOM_ISA and NCOM_COMMULTI, to
only include the relevant code in the probe & attach functions. Still
one probe and one attach function, with #ifdefs, but this is a step
in the right direction and saves a few hundred bytes (ooh, ahh!).

New device attachment scheme:

- split softc size and match/attach out from cfdriver into
a new struct cfattach.

- new "attach" directive for files.*. May specify the name of
the cfattach structure, so that devices may be easily attached
to parents with different autoconfiguration semantics.

convert these to use new <machine/bus.h> interfaces. This involved
substantial reworking of the multi-port drivers, as they need to frob
bits in the io-port spaces of their children. As a result, the
commulti->com attachment interface is substantially more complex.
(This may be fixable in the future by making some of the code common,
but as long as io-port allocation checking is planned, it's necessary.)

include comvar.h for comprobe1() prototype

Fix typo (PR 2123, from Michael VanLoon)

Add more debugging code.

Update copyright notice.

Set IER_ETXRDY before outputting any bytes, to prevent race condition.

Handle break more correctly.
Make IER_ETXRDY track TS_BUSY. (Prevents a loop in comintr() if no output
is available.)
Add a bunch of debugging code.

Fix a race condition where we might stop outputting if a write follows a
flush too closely.

Add some text describing the UART lossage worked around in the previous.

Drain the transmitter FIFO before changing the baud rate or FIFO depth
registers. Fixes PR 2046.
Also, if COMCONSOLE is defined, expect CONSPEED to be the baud rate.

Hayes ESP patches [from PR database]

Hayes ESP support from Michael Van Loon, with minor cleanups. Tested on
a single port ESP card; works great!

The IST_* and IPL_* constants are not bus-specific; don't treat them as such.
Change splimp -> splnet in Ethernet, ARCnet, and FDDI drivers.

Use SET(), CLR(), and ISSET() macros.
Rename CFCR to LCR, and cache its value.
Check MSR before transmitting, for better flow control.
Preserve LCR_SBREAK inside comparam().

Fix a condition where ttywait() wouldn't be woken up. Also, short-circuit
the case of no pending input in compoll().

note that most of dev/ic's contents have changed names

Various:
* Check for CTS in the correct register.
* Only do a selwakeup() if we output something.
* Don't make assumptions about what TTYDISC does.

Handle BREAK correctly when not entering DDB.

Add an input FIFO, and fix several flow control problems. Based on code from
Felix A. Croes.

Changed ns82450 to ns8250 and dropped ns82550 from probe console messages.

Only use the `no pending interrupt' bit in the IIR; the other bits
are too unreliable on clone chips.

(1) conditionalize debugger usage in comeint with DDB, not COMCONSOLE;
latter would lead to undefined symbols if DDB not defined.
(2) check for break on console, and therefore debugger entry (if ddb
in kernel) earlier, so that the device doesn't need to be open.
(3) return immediately after breaking into the debugger in comeint().
(4) only do the normal character input routine in comintr if receive
mask yeilds _EXACLTY_ LSR_RXRDY. if there's only a receive
error, or there's a receive error _and_ a received character,
do comeint().
(former two by me. latter two from Bob Baron <rvb@cs.cmu.edu>.)

<break> --> ddb if defined(comconsole).
(if you have a serial console and you hit break, you get ddb).

Fix thinko in previous change.

Garbage collect #includes.

Implement comtty().

clean up several ISA device interfaces: autoconfiguration, header
inclusion, and interrupt configuration. more work still needs to be done,
but it's getting better...

Most of the console functions return void. Also change interrupt logic
slightly.

lip service to making ISA support machine-independent. This is the first
round: moving the drivers into a machine-independent directory.
Some drivers (e.g. fd.c) not moved because they use other pc features (e.g.
CMOS settings), and none of the non-driver files moved, because they're
still pretty much PC specific. eventually (when other ports with ISA
busses really start using this code), more 'high-level' ISA support will
live here.

use chip headers in /dev/ic

Explicitly test the RXRDY status bit on incoming characters.

Fix oversight in previous.

Add interrupt sharing types.

Minor update for new autoconfig.

Convert port, IRQ, and DRQ numbers to ints.

Update to match autoconfig code.

be more careful with types, also pull in headers where necessary.

new RCS ID format.

Implement *cnpollc().

Only reset FIFO if changing baud rate. From John Kohl (slightly edited).

Set the FIFO threshold based on the receive speed, per Mark Weaver.

Look for MDMBUF in cflag, not lflag.

Count up the silo overflows and only log a warning at most once per minute.

Bug fixes from Mark Weaver.

branches: 1.31.2;
allow MDMBUF flags to be set (oops)

Implement dynamic IRQ configuration and IRQ sharing. Inline spl*() calls.
Reorganize and clean up the relevant code.

Updates for new autoconfig.

Fix off-by-one error in comopen() unit number sanity check. From Brian de
Alwis.

Some stylistic cleanup, and a very minor speedup.

add AST-style serial multiport support, from Roland McGrath
<roland@@frob.com>. Needs light clean by Mr. I386, but mostly
OK. For some reason, Bad Things (TM) happened on the last cvs commit.

add reasonable support for MDMBUF output flow control. input to be done later

copy appropriate hardware config bits from config's flags

fix some of the last; thanks to charles for looking it over...

support new ioctls. fix the way CRTSCTS is used, etc. Seperate 'hardware'
and 'software' flags. beginnings of multiport support.

Major cleanup and many bugs fixed; based in part on Brad Huntting's version
for BSD/386. More to be done when the low-level interrupt system is replaced.

DELAY() --> delay(). This is not a macro.

All ioctl routines take a struct proc * now.

oops

new cons.h location, etc.

Traditional behaviour is to reset all modes if device isn't open already

Junk comselect(); it's the same at ttselect().

Canonicalize all #includes, and add pio.h where appropriate.

ifdef out the ttyfree()'s, so that fill_eproc doesn't panic when
a process's session still holds a ref to a tty which has been deallocated
and reused.

branches: 1.12.2;
tty XXstart() routines return void

Change tty code to use clist interface, but with ring buffer implementation.
Also, fix a couple of bugs in tty.c and pccons.c, and some gross kluginess
in the hp300 stuff.

pccons.c now dynamically allocates it's "struct tty"
cons.c's "struct tty *cn_tty" wasn't used by any of the kernel, and goes away.

make getc() and ungetc() be rb{un,}getc(), so getc() and ungetc()
don't conflict w/ansi prototypes...

tty dynamic allocation

make kernel select interface be one-stop shopping & clean it all up.

fixed to be compliant, subservient, and to take advantage of the newly
hacked config(8)

have probe return size of io space on successful return, rather
than simply 1.

dmesg output at boottime now tries to print out information as
soon as it is available. The output looks much more like Sunos.

moved closing #endif to include comselect

after 0.2.2 "stable" patches applied

branches: 1.1.1;
Initial revision

Revert com.c:r1.381 because i386/qemu cannot boot. Pointed out by gson@n.o and martin@n.o.

Fix hang up writing /dev/console rarely in specific environments.

Some BMC seems to require these syncronous operations. If not,
it does not send transmit completion interrupts for some reason.

KNF a comment

ddb/db_active.h: New home for extern db_active.

This can be included unconditionally, and db_active can then be
queried unconditionally; if DDB is not in the kernel, then db_active
is a constant zero. Reduces need for #include opt_ddb.h, #ifdef DDB.

com(4): Nix quirky `integrate' macro.

Just use `static inline' like everything else.

com(4): Omit needless spltty in comstart.

This is called either via tp->t_oproc, which is done with tty_lock
held (thus, at IPL_VM = IPL_TTY), or from comparam which is called at
IPL_TTY, either via comopen or tp->t_param.

com(4): Nix dead code.

This was introduced in rev. 1.292, whose commit message was...
`Delete dead code.' I guess December 6, 2009 was Opposite Day?

com(4): Comment on lock order.

com(4): Update confusing comment from decades ago to reflect now.

No functional change.

com: Add support for 32-bit IO accesses.

For the DW APB busy bit workaround, only attempt to re-apply LCR and DLB
settings for non-console devices. In the console case, simply clear the
busy bit and continue. Fixes an interrupt storm observed on Macchiatobin.

Restore 16-byte FIFO for ns16550a, accidentally lost in r1.344.

Spotted by msaitoh. Thanks!

- microtime -> microuptime
- avoid kpause with timeo=0

- use microtime instead of getmicrotime, suggested by thorpej
- use ttclos for wchan even though we are sleeping in comopen now

Note that while sleeping (HUPCL), time passes.
Try to avoid the otherwise infinite loop.

Now Jason has made this build, avoid a 50+ year
sleep to implement HUPCL.

com: speed up close with HUPCL set

Instead of incurring a 1s penalty on close of a com device with HUPCL set,
defer the sleep until the next open, and only sleep if necessary.

This has a side effect of making `ttyflags -a` with a default install not
pause for 1s for every non-console com device, which happens every boot
via /etc/rc.d/ttys.

only read cpr register if we're going to use the value

com(4): fix FIFO for DW_APB on Allwinner A20 (got broken by com.c 1.360)

Older DesignWare UARTs do not advertise their FIFO length so we must
provide it via device properties.

branches: 1.363.2;
Introduce COM_HW_BROKEN_ETXRDY bit in sc_hwflags; push TX queue if
possible in comintr() even if IIR_NOPEND is asserted.

Combined with sc_poll_ticks, this flag works around HW bug, by which
ETXRDY interrupts are (sometimes) lost.

Introduce sc_poll_ticks and obsolete COM_HW_POLL bit in sc_hwflags.
Polling is scheduled at every sc_poll_ticks ticks.

This is useful to work around H/W bug, by which interrupts are lost
*sometimes*; interrupt-based I/O mostly works and no need for polling
every counter ticks.

branches: 1.361.2; 1.361.4;
Set sc_fifolen=1 for the no fifo case

Auto-detect DW APB UART FIFO length and print the FIFO length when enabled
instead of just "working fifo".

Make the readahead-while-sending-output code conditional in
com_common_putc(), only erquest readahead before interrupts are enabled
and when called from comcnputc().
Fixes PR kern/55286.

Get rid of needless __predict_true() and move a comment slightly.

Use designated initializers

branches: 1.356.2;
in many device attach paths, allocate memory with M_WAITOK instead of M_NOWAIT
and remove code to handle failures that can no longer happen.

Simplify regmap initialization, and fix an regmap issue that
affected TI OMAP (LCR register would get clobbered due to
using the wrong offset for the MDR1 register) reported by Lwazi Dube
(who also found the root cause).

Add a convenience function, com_init_regs_stride(), that shifts the register
offsets and size by the specified amount. Use in front-ends as appropriate.

sparc64 doesn't have a scalar bus_space_handle_t, so cope with this.
Also, add a comment about why we're using the dummy / nil space handle
in the first place.

Remove the COM_REGMAP option -- just use it all the time. While here,
garbage-collect the COM_FUNCMAP and COM_AU1X00 options, as there are
not used anywhere.

Clean up initialization of com_regs structure, in preparation for
some additional changers.

Add support for com(4) without an interrupt by setting the COM_HW_POLL flag

Add support for a "force_console" property, where when set, reuses the console dev configuration

branches: 1.348.2;
Rename COM_TYPE_SUNXI -> COM_TYPE_DW_APB

Remove COM_AWIN option

branches: 1.346.2;
Put back SUNXI definitions in the COM_REGMAP case; com.c uses them
without #ifdef COM_AWIN.
To be safe, expand the register map to 42 entries.

As discussed on tech-kern: define a new tty internal state flag: TS_KERN_ONLY

Implement it in a few tty drivers. If this flag is set, the underlying
hardware is used by another driver and userland has no right to open
it. A few uses will appear soon in sys/dev/sun/sun{kbd,ms}.c.

Make all of the COM_xxx type options runtime selectable. Kernel configs
with the existing options (COM_16650, COM_16750, COM_AWIN, COM_HAYESP, and
COM_PXA2X0) will select the correct type in com_attach_subr. New code
should specify the com type by passing COM_TYPE_xxx to comcnattach and/or
setting sc_type.

Kill some more extern struct cfdriver declarations.

Down with externs in .c!

Report AUXUART fifo size on rpi3 to 1 byte. This prevents bluetooth
errors using the h5 protocol.

Broadcom AUX UART doesn't seem to set LSR RXRDY bit when data is available
in the RX FIFO.

Add COM_TYPE_BCMAUXUART type for BCM2835 AUX UART.

branches: 1.339.10;
The UART in the allwiner SoCs is not full-compatible with the 16550, and
it's not a 16750 either. Like the 16750 it has the IIR_BUSY interrupt,
which is triggered when writing to LCR while the chip
can't accept it. But unlike the 16750, it has a specific register,
HALT, to allow writing to the LCR and divisor registers, and then
commit the changes.
Tested on an A20 SoC, changing the baud rate while keeping the
tty device open and incoming data.

Fix TX FIFO on Tegra

PR/50395: Vicente Chaves de Melo: Loss of characters in serial port
communication between gdb and kgdb_stub: switch to line break interrupt
both when entering KGDB and also on shutdown so we can re-enter the debugger
with BREAK.

For Tegra K1, set IE_RX_TIMEOUT (bit 4) in IER register. RX_TIMEOUT occurs
when data has been sitting in the Rx FIFO for more than 4 character times
without being read because there is not enough data to reach the trigger
level. With this change, enable FIFO usage for Tegra UARTs.

fix pasto, use SET() and CLR()
thanks jmcneill@

add COM_TYPE_TEGRA

Convert sys/dev to use <sys/rndsource.h>.

KNF

Fix build.

Hi mipsallan

more Ingenic support:
- make sure we always set FIFO_UART_ON
- deal with the absence of DCD support
- enable the TX FIFO timeout interrupt
- set COM_HW_NOIEN
- pretend the FIFO is only 16 bytes deep ( supposed to be 64 but I get
overruns with that )
now this works as CI20 console

branches: 1.329.2;
deal with quirk in Ingenic UARTs
( they have a bit in the FIFO control register which turns the entire
port off if not set )

centralize the dialout/call unit macros.

branches: 1.327.2;
Merge tls-earlyentropy branch into HEAD.

Add d_discard to all struct cdevsw instances I could find.

All have been set to "nodiscard"; some should get a real implementation.

Clear dtr on transition from !0 to 0, and set it on the opposite.
This leaves dtr alone in the normal case.

PR/48720: John Kelly: com driver does not allow B0 to be set. The code to
handle B0 was commented out for no reason given at revision 1.99. POSIX
mandates to hangup on B0:
http://pubs.opengroup.org/onlinepubs/9699919799/functions/tcsetattr.html

branches: 1.323.2;
Change (mostly mechanically) every cdevsw/bdevsw I can find to use
designated initializers.

I have not built every extant kernel so I have probably broken at
least one build; however I've also found and fixed some wrong
cdevsw/bdevsw entries so even if so I think we come out ahead.