introduce a new kern.autoconf_serial sysctl that can be used by userland
to monitor state changes of the kernel device tree

input from dnd ok dlg@, deraadt@

Fix kernel builds with pseudo-device rd

Make the cf_attach member of struct cfdata const and sprinkle a few
const into subr_autoconf.c to make this work. Fixes the compilation
of sys/dev/rd.c with newly const rd_ca.

ok miod (who had a similar diff)

Improve unhibernate performance (30% on some machines, another upcoming diff
shows gains up to 50%) by skipping attach of irrelevant devices, which are
tagged CD_SKIPHIBERNATE in the per-driver cfdriver. In particular, usb devices
are not attached, so they don't need to detach during the suspend-unpack-resume.
New bootblocks are required (which tell the kernel it's job is unhibernate
before configure runs)
tested by various

convert infinite msleep(9) to msleep_nsec(9)

ok mpi@

free(9) sizes for softcs.

ok tedu@

Change a bunch of (<blah> *)0 to NULL.

ok beck@ deraadt@

Simple free(9) sizes, from Matthieu <naabed AT poolp DOT org>.

Replace mountroothook_establish(9) by config_mountroot(9) a narrower API
similar to config_defer(9).

ok mikeb@, deraadt@

back out refcnt for dv_ref, there's too many hand crafted devices all
over the tree.

much encouragement from l2k15

use refcnts for the device reference counts as an example of how
refcnt(9) can be used.

Remove some includes include-what-you-use claims don't
have any direct symbols used. Tested for indirect use by compiling
amd64/i386/sparc64 kernels.

ok tedu@ deraadt@

you need atomic.h to do atomic things.

pointed out by deraadt@

use atomic ops to increment and decrement the device ref count in
device_ref and device_unref.

ok guenther@ deraadt@

bring back r1.80:

pass the size of the softc to free.

this time we only dereference the cfattach structure if we actually
need it once the refcount has dropped to 0.

tested by krw@ who found the original panic r1.80 was backed out for.
putting this in again so i can find out when else will break.

back it out properly

never tested with a make release

pass the size of the softc to free.

ok deraadt@

Introduce config_suspend_all(9) a function that invokes config_suspend(9)
on all relevant device hierarchies in the appropriate order. For now this
means mpath(4) and mainbus(4), doing mpath(4) before mainbus(4) when
suspending or powering down and doing mpath(4) after mainbus(4) when
resuming such that mpath(4) can realy on the underlying hardware being
in a functional state.

Fixes problems with unflushed disk caches on machines where mpath(4) takes
control of some of your disks.

ok dlg@

remove uneeded proc.h includes
ok mpi@ kspillner@

use mallocarray where arguments are multipled. ok deraadt

add a size argument to free. will be used soon, but for now default to 0.
after discussions with beck deraadt kettenis.

Change config_deactivate(9) to always unset the DVF_ACTIVE flag so
that drivers can check for this flag to know they have to stop doing
any hardware access and release their reference.

While here makes the ca_activate() function optional and propagates
DVACT_DEACTIVATE to the children of a device that do not implement this
interface like it is done during suspend/resume.

ok uebayasi@, deraadt@

It's been a quarter century: we can assume volatile is present with that name.

ok dlg@ mpi@ deraadt@

get rid of the assumption that the head of the alldevs list is the
"mainbus" device. this breaks when mpath is enabled because it
attaches before mainbus and therefore takes the head position.

have autoconf provide device_mainbus() which looks up mainbus_cd,
and use that instead.

discussed with deraadt who just wants mpath stuff to move forward
despite there being many ways to shine this particular turd.

printf(variable) --> printf("%s", variable)

ok deraadt@

Add a DVACT_WAKEUP op to the *_activate() API. This is called after the
kernel resumes normal (non-cold, able to run processes, etc) operation.
Previously we were relying on specific DVACT_RESUME op's in drivers
creating callback/threads themselves, but that has become too common,
indicating the need for a built-in mechanism.
ok dlg kettenis, tested by a sufficient amount of people

msgs can be const

get rid of the goop for maintaining multiple cftables since theres
no current code there to add or remove them. such code was put in
17 years ago, disabled 14 years ago, and removed 11 years ago, so
i suspect we dont really need it or the goop to support it.

very ok miod@

compare pointers to NULL

Enforce ca_activate tree-walks over the entire heirarchy for all events,
cleaning up some shutdown-hook related code on the way.
(A few drivers related to sparc are still skipped at kettenis' request)
ok kettenis mlarkin, tested by many others too

Revamp the sequences for suspend/hibernate -> resume so that the code
paths are reflexive. It is now possible to fail part-way through a
suspend sequence, and recover along the resume code path.
Split DVACT_SUSPEND by adding a new DVACT_POWERDOWN method is used
after hibernate (and suspend too) to finish the job. Some drivers
must be converted at the same time to use this instead of shutdown hooks
(the others will follow at a later time)
ok kettenis mlarkin

Remove config_activate() and DVACT_ACTIVATE. PCMCIA's the only thing
that's ever used it, and it's long since been changed to use
DVACT_{QUIESCE,SUSPEND,RESUME} instead.

ok deraadt@, dlg@; miod@ also agreed with this idea when I brought it
up a few weeks ago

serialize attach and detach of device sub-trees -- only one device
sub-tree may attach or detach at a time. attach and detach will sleep
against each other.
this is fixing (working around?) some bizzare corner cases that have
been seen (but not fully diagnosed) where the device trees, disk registration
subsystem, and other things could get messed up. one could argue though
that this serialization is a very good thing; it is easier than adding piles
of locks in various other places.
ok matthew jsing

Add a few KASSERTs to config_attach() for sanity to make sure we don't
try to reuse device unit numbers and to check that the device pointer
array is allocated and large enough.

Also, improve the panic message generated by config_detach() when we
detect that we're detaching a device that still has children.

Discussed with deraadt@ while trying to brainstorm ways that
interleaving config_attach and config_detach could blow up.

remove a useless comment

Add DVACT_QUIECE support. This is called before splhigh() and before
DVACT_SUSPEND, therefore DVACT_QUIECE can do standard sleeping operations
to get ready.
Discussed quite a while back with kettenis and jakemsr, oga suddenly needed
it as well and wrote half of it, so it was time to finish it.
proofread by miod.

Silence the activation debug reporting. We are starting to get worried
that there are interactions when we print in the various (saved, post-saved,
pre-restored, and restored) vga states, especially with how intel drm is
getting involved. If your machine has issues, you may want to activate
these messages again to help debug things better.
ok mlarkin

Remove leftover debug printf in config_activate_children(); ok deraadt

Make config_activate_children() more useable as a suspend-and-resume
device tree walker, and add config_suspend() as well.
ok mlarkin pirofti, discussion with kettenis

Add basic suspend/resume autoconf functionality.

Okay deraadt@, kettenis@, mlarkin@.

Get rid of devact enum, substitute it with an int and coresponding defines.

This is needed for the addition of further suspend/resume actions.

Okay deraadt@, marco@.

Allow devices attaching at root (such as softraid) to be disabled in UKC.
ok marco@ deraadt@

Since the devices that we detach may have children we cannot rely on the
"previous" device still being there, so if we detach a device, we really
have to start from scratch again (like the comment on this code actually
says we should). Simplifies the code as a bonus.

ok dlg@, deraadt@

get rid of the unused cond_predicate_t typedef.

ok miod@ deraadt@ krw@

Use M_ZERO in a few more places to shave bytes from the kernel.

eyeballed and ok dlg@

the (root) stuff is ugly, so print "at root", so it looks more like
regular things. everyone here buys it, and i'm 90% sure miod will be
ok with it

kill __HAVE_DEVICE_REGISTER by requiring all architectures to have a
device_register() function -- even if it does nothing. reduces the
cpp-based blather different between architectures
idea ok'd by miod; tested on all architectures (except a few miod will
need to cleanup because he has them)

evcnt_attach() is no longer used; ok miod

Remove device id from hotplug events.

ok henning

more fallout from the sensors: Every single ramdisk got broken.

Include device id in hotplug events. This will be used by ntpd to check
sensors on attach/detach. hotplugd changes following in a minute.

ok henning.

cf_unit should have a value between 0 and cd_ndevs. So set cf_unit to 0 when
cd_ndevs is set to 0.

"No kidding!" deraadt@ "Seems to make sense to me." beck@ "Seems right." miod@

Detach children devices in reverse order. This gives better display, and
allows cf_unit to be decremented correctly.

'put it in' deraadt@ "i think it's good' tedu@

When assigning a unit number, treat running off the end of the
cd_devs[] array the same as finding a NULL entry.

Fixes inconsistant assignment of unit numbers to media slots on the
Neodio 8-in-1 USB card reader donated by Manuel Pata. Previously the
unit numbers would vary depending on whether it had attached an even
or odd number of times.

'reads ok' marco@ ok miod@

Remove now unused zeroref member in cfattach structures.

In config_detach_children(), after detaching a device, restart the device
list walk from its parent device, as the device which was following it may
have been one of its children, and thus gone as well. Found the hard way
using the strict queue macros.

Feedback and help toby@, ok deraadt@

ansi and deregister. No binary change.

Use list and queue macros where applicable to make the code easier to read;
no change in compiler assembly output.

Create the init process earlier, before the root filesystem is mounted,
and have it stall on a semaphore. This allows all kthread creations which
could have been requested during autoconf to be processed before root
is mounted as well.

This causes umass devices attached to any usb with flags 1 (such as on macppc)
to configure properly instead of panicing the kernel at mountroot time.

From NetBSD; tested by various.

Fix a minor bug where indirect config crashes the kernel
when match verbose is turend on.

Devices hot plugging support.
The hotplug pseudo-device passes device attachment and detachment events
to userland. When a device attaches or detaches, the corresponding event
is queued. The events can then be obtained from the queue through the
read(2) call on the /dev/hotplug device file. Each event consists of
event type (attach/detach), device class (DV_*) and device name (sd1 e.g.).

We have hotplug pseudo-device on alpha, amd64, i386, macppc and sparc64.
Since it was tested only on i386 other archs has it commented out
in GENERIC until tested.

The idea liked peter@ tedu@ drahn@ millert@ marco@ henning@.
Ok deraadt@.

Remove the advertising clause in the UCB license which Berkeley
rescinded 22 July 1999. Proofed by myself and Theo.

string fixes; tedu ok

Tweak previous sprintf -> snprintf fix to eliminate unnecessary
intermediate variable.

Suggested by deraadt@.

ok henning@ millert@ tdeval@

Use snprintf() to simplify device name construction, removing now
superfluous variables and the function 'number()'.

ok deraadt@ and millert@

Remove some '#if 0' cruft.

Introduce a new file, machine/internal_types.h, to hold that specific arch
type characteristics.

internal_types.h will contain only settings invisible from standard C, e.g.,
in the __* or _[A-Z]* namespace, and be reused by files like limits.h.

This allows us to shorten machine/limits.h greatly, as all the common defines
are now in sys/limits.h, plus a small stub in internal_types.h.

Tested on all arches as far as I know.
Approved after discussion with art, millert, deraadt, and others.

First round of __P removal in sys

branches: 1.29.4;
Add __HAVE_DEVICE_REGISTER to hp300/alpha (sparc64 already had it) and
clean up the arch specific #ifdef's in subr_autoconf.c; from NetBSD.
(tested by me on alpha/sparc64 and miod on hp300)

implement autoconf mechanism using device_register(); from NetBSD.

range error; < cd_ndevs, not <=

Added config_detach_children, config_activate_children.

Added ref counting to devices: device_ref, device_unref and a new method
for devices : ca_zeroref, called when the ref count hits zero.

Note that ca_zeroref may be called from interrupt context.

branches: 1.25.4;
From NetBSD; new device detaching infrastructure.

From NetBSD; config_defer to defer configuration of a device until after all
other siblings.

Remove obsolete detach logic in preparation of a newer one

panic prints a newline for you, don't do it in the panic string

Formerly reuse of unit numbers of detached devices was impossible, with
this change starred devices will search from the first allowed unit for
such devices and up for free units when attaching. This means you have
to rerun config(8) before yopu compile current kernels. FYI, this is
a necessity when doing PCMCIA flashcard reading, otherwise you soon run
out of unit numbers.

I have also added a way to detach all children of a certain device, used
in a new PCMCIA framework to come, so you detach all functions on a
single PCMCIA card when pulling it out.

A few bugs wrt starred devices fixed in the autoconfig_verbose code too.
Last and least, fixed some panic messages.

Please GCC 2.8's harsher view of good style

Prevent a NULL-deref in the autoconf_verbose case

typo

Provide a way to debug probes. set autoconf_verbose via DDB, or
option AUTOCONF_VERBOSE=1
in your kernel config.

indent

hp300 needs device_register(), too.

export struct cfdata cfdata[];

As alpha uses device_register, it needs its proto too

no mknod in chroot

Remove unnecessary prototype

pretty

-Wall & -Wstrict-prototype fixes

Added support for user modifiable kernel at boot (-c) /maja

sick netbsd alpha hack (why not do what other ports do?)

Make sure config_init gets called before consinit, as at least one port
actually uses config_* functions for console attachment. Document
config_init should not call malloc as the move was over the mem-init
function calls too.

Pull in John Kohl's [jtk@netbsd.org] most recent (15Apr96) APM and PCMCIA work
(original PCMCIA framework by Stefan Grefen [grefen@convex.com]).

partial sync with netbsd 960418, more to come

NetBSD 960317 merge

From NetBSD: 960217 merge

branches: 1.1.1;
Initial revision

Fix kernel builds with pseudo-device rd

Make the cf_attach member of struct cfdata const and sprinkle a few
const into subr_autoconf.c to make this work. Fixes the compilation
of sys/dev/rd.c with newly const rd_ca.

ok miod (who had a similar diff)

Improve unhibernate performance (30% on some machines, another upcoming diff
shows gains up to 50%) by skipping attach of irrelevant devices, which are
tagged CD_SKIPHIBERNATE in the per-driver cfdriver. In particular, usb devices
are not attached, so they don't need to detach during the suspend-unpack-resume.
New bootblocks are required (which tell the kernel it's job is unhibernate
before configure runs)
tested by various

convert infinite msleep(9) to msleep_nsec(9)

ok mpi@

free(9) sizes for softcs.

ok tedu@

Change a bunch of (<blah> *)0 to NULL.

ok beck@ deraadt@

Simple free(9) sizes, from Matthieu <naabed AT poolp DOT org>.

Replace mountroothook_establish(9) by config_mountroot(9) a narrower API
similar to config_defer(9).

ok mikeb@, deraadt@

back out refcnt for dv_ref, there's too many hand crafted devices all
over the tree.

much encouragement from l2k15

use refcnts for the device reference counts as an example of how
refcnt(9) can be used.

Remove some includes include-what-you-use claims don't
have any direct symbols used. Tested for indirect use by compiling
amd64/i386/sparc64 kernels.

ok tedu@ deraadt@

you need atomic.h to do atomic things.

pointed out by deraadt@

use atomic ops to increment and decrement the device ref count in
device_ref and device_unref.

ok guenther@ deraadt@

bring back r1.80:

pass the size of the softc to free.

this time we only dereference the cfattach structure if we actually
need it once the refcount has dropped to 0.

tested by krw@ who found the original panic r1.80 was backed out for.
putting this in again so i can find out when else will break.

back it out properly

never tested with a make release

pass the size of the softc to free.

ok deraadt@

Introduce config_suspend_all(9) a function that invokes config_suspend(9)
on all relevant device hierarchies in the appropriate order. For now this
means mpath(4) and mainbus(4), doing mpath(4) before mainbus(4) when
suspending or powering down and doing mpath(4) after mainbus(4) when
resuming such that mpath(4) can realy on the underlying hardware being
in a functional state.

Fixes problems with unflushed disk caches on machines where mpath(4) takes
control of some of your disks.

ok dlg@

remove uneeded proc.h includes
ok mpi@ kspillner@

use mallocarray where arguments are multipled. ok deraadt

add a size argument to free. will be used soon, but for now default to 0.
after discussions with beck deraadt kettenis.

Change config_deactivate(9) to always unset the DVF_ACTIVE flag so
that drivers can check for this flag to know they have to stop doing
any hardware access and release their reference.

While here makes the ca_activate() function optional and propagates
DVACT_DEACTIVATE to the children of a device that do not implement this
interface like it is done during suspend/resume.

ok uebayasi@, deraadt@

It's been a quarter century: we can assume volatile is present with that name.

ok dlg@ mpi@ deraadt@

get rid of the assumption that the head of the alldevs list is the
"mainbus" device. this breaks when mpath is enabled because it
attaches before mainbus and therefore takes the head position.

have autoconf provide device_mainbus() which looks up mainbus_cd,
and use that instead.

discussed with deraadt who just wants mpath stuff to move forward
despite there being many ways to shine this particular turd.

printf(variable) --> printf("%s", variable)

ok deraadt@

Add a DVACT_WAKEUP op to the *_activate() API. This is called after the
kernel resumes normal (non-cold, able to run processes, etc) operation.
Previously we were relying on specific DVACT_RESUME op's in drivers
creating callback/threads themselves, but that has become too common,
indicating the need for a built-in mechanism.
ok dlg kettenis, tested by a sufficient amount of people

msgs can be const

get rid of the goop for maintaining multiple cftables since theres
no current code there to add or remove them. such code was put in
17 years ago, disabled 14 years ago, and removed 11 years ago, so
i suspect we dont really need it or the goop to support it.

very ok miod@

compare pointers to NULL

Enforce ca_activate tree-walks over the entire heirarchy for all events,
cleaning up some shutdown-hook related code on the way.
(A few drivers related to sparc are still skipped at kettenis' request)
ok kettenis mlarkin, tested by many others too

Revamp the sequences for suspend/hibernate -> resume so that the code
paths are reflexive. It is now possible to fail part-way through a
suspend sequence, and recover along the resume code path.
Split DVACT_SUSPEND by adding a new DVACT_POWERDOWN method is used
after hibernate (and suspend too) to finish the job. Some drivers
must be converted at the same time to use this instead of shutdown hooks
(the others will follow at a later time)
ok kettenis mlarkin

Remove config_activate() and DVACT_ACTIVATE. PCMCIA's the only thing
that's ever used it, and it's long since been changed to use
DVACT_{QUIESCE,SUSPEND,RESUME} instead.

ok deraadt@, dlg@; miod@ also agreed with this idea when I brought it
up a few weeks ago

serialize attach and detach of device sub-trees -- only one device
sub-tree may attach or detach at a time. attach and detach will sleep
against each other.
this is fixing (working around?) some bizzare corner cases that have
been seen (but not fully diagnosed) where the device trees, disk registration
subsystem, and other things could get messed up. one could argue though
that this serialization is a very good thing; it is easier than adding piles
of locks in various other places.
ok matthew jsing

Add a few KASSERTs to config_attach() for sanity to make sure we don't
try to reuse device unit numbers and to check that the device pointer
array is allocated and large enough.

Also, improve the panic message generated by config_detach() when we
detect that we're detaching a device that still has children.

Discussed with deraadt@ while trying to brainstorm ways that
interleaving config_attach and config_detach could blow up.

remove a useless comment

Add DVACT_QUIECE support. This is called before splhigh() and before
DVACT_SUSPEND, therefore DVACT_QUIECE can do standard sleeping operations
to get ready.
Discussed quite a while back with kettenis and jakemsr, oga suddenly needed
it as well and wrote half of it, so it was time to finish it.
proofread by miod.

Silence the activation debug reporting. We are starting to get worried
that there are interactions when we print in the various (saved, post-saved,
pre-restored, and restored) vga states, especially with how intel drm is
getting involved. If your machine has issues, you may want to activate
these messages again to help debug things better.
ok mlarkin

Remove leftover debug printf in config_activate_children(); ok deraadt

Make config_activate_children() more useable as a suspend-and-resume
device tree walker, and add config_suspend() as well.
ok mlarkin pirofti, discussion with kettenis

Add basic suspend/resume autoconf functionality.

Okay deraadt@, kettenis@, mlarkin@.

Get rid of devact enum, substitute it with an int and coresponding defines.

This is needed for the addition of further suspend/resume actions.

Okay deraadt@, marco@.

Allow devices attaching at root (such as softraid) to be disabled in UKC.
ok marco@ deraadt@

Since the devices that we detach may have children we cannot rely on the
"previous" device still being there, so if we detach a device, we really
have to start from scratch again (like the comment on this code actually
says we should). Simplifies the code as a bonus.

ok dlg@, deraadt@

get rid of the unused cond_predicate_t typedef.

ok miod@ deraadt@ krw@

Use M_ZERO in a few more places to shave bytes from the kernel.

eyeballed and ok dlg@

the (root) stuff is ugly, so print "at root", so it looks more like
regular things. everyone here buys it, and i'm 90% sure miod will be
ok with it

kill __HAVE_DEVICE_REGISTER by requiring all architectures to have a
device_register() function -- even if it does nothing. reduces the
cpp-based blather different between architectures
idea ok'd by miod; tested on all architectures (except a few miod will
need to cleanup because he has them)

evcnt_attach() is no longer used; ok miod

Remove device id from hotplug events.

ok henning

more fallout from the sensors: Every single ramdisk got broken.

Include device id in hotplug events. This will be used by ntpd to check
sensors on attach/detach. hotplugd changes following in a minute.

ok henning.

cf_unit should have a value between 0 and cd_ndevs. So set cf_unit to 0 when
cd_ndevs is set to 0.

"No kidding!" deraadt@ "Seems to make sense to me." beck@ "Seems right." miod@

Detach children devices in reverse order. This gives better display, and
allows cf_unit to be decremented correctly.

'put it in' deraadt@ "i think it's good' tedu@

When assigning a unit number, treat running off the end of the
cd_devs[] array the same as finding a NULL entry.

Fixes inconsistant assignment of unit numbers to media slots on the
Neodio 8-in-1 USB card reader donated by Manuel Pata. Previously the
unit numbers would vary depending on whether it had attached an even
or odd number of times.

'reads ok' marco@ ok miod@

Remove now unused zeroref member in cfattach structures.

In config_detach_children(), after detaching a device, restart the device
list walk from its parent device, as the device which was following it may
have been one of its children, and thus gone as well. Found the hard way
using the strict queue macros.

Feedback and help toby@, ok deraadt@

ansi and deregister. No binary change.

Use list and queue macros where applicable to make the code easier to read;
no change in compiler assembly output.

Create the init process earlier, before the root filesystem is mounted,
and have it stall on a semaphore. This allows all kthread creations which
could have been requested during autoconf to be processed before root
is mounted as well.

This causes umass devices attached to any usb with flags 1 (such as on macppc)
to configure properly instead of panicing the kernel at mountroot time.

From NetBSD; tested by various.

Fix a minor bug where indirect config crashes the kernel
when match verbose is turend on.

Devices hot plugging support.
The hotplug pseudo-device passes device attachment and detachment events
to userland. When a device attaches or detaches, the corresponding event
is queued. The events can then be obtained from the queue through the
read(2) call on the /dev/hotplug device file. Each event consists of
event type (attach/detach), device class (DV_*) and device name (sd1 e.g.).

We have hotplug pseudo-device on alpha, amd64, i386, macppc and sparc64.
Since it was tested only on i386 other archs has it commented out
in GENERIC until tested.

The idea liked peter@ tedu@ drahn@ millert@ marco@ henning@.
Ok deraadt@.

Remove the advertising clause in the UCB license which Berkeley
rescinded 22 July 1999. Proofed by myself and Theo.

string fixes; tedu ok

Tweak previous sprintf -> snprintf fix to eliminate unnecessary
intermediate variable.

Suggested by deraadt@.

ok henning@ millert@ tdeval@

Use snprintf() to simplify device name construction, removing now
superfluous variables and the function 'number()'.

ok deraadt@ and millert@

Remove some '#if 0' cruft.

Introduce a new file, machine/internal_types.h, to hold that specific arch
type characteristics.

internal_types.h will contain only settings invisible from standard C, e.g.,
in the __* or _[A-Z]* namespace, and be reused by files like limits.h.

This allows us to shorten machine/limits.h greatly, as all the common defines
are now in sys/limits.h, plus a small stub in internal_types.h.

Tested on all arches as far as I know.
Approved after discussion with art, millert, deraadt, and others.

First round of __P removal in sys

branches: 1.29.4;
Add __HAVE_DEVICE_REGISTER to hp300/alpha (sparc64 already had it) and
clean up the arch specific #ifdef's in subr_autoconf.c; from NetBSD.
(tested by me on alpha/sparc64 and miod on hp300)

implement autoconf mechanism using device_register(); from NetBSD.

range error; < cd_ndevs, not <=

Added config_detach_children, config_activate_children.

Added ref counting to devices: device_ref, device_unref and a new method
for devices : ca_zeroref, called when the ref count hits zero.

Note that ca_zeroref may be called from interrupt context.

branches: 1.25.4;
From NetBSD; new device detaching infrastructure.

From NetBSD; config_defer to defer configuration of a device until after all
other siblings.

Remove obsolete detach logic in preparation of a newer one

panic prints a newline for you, don't do it in the panic string

Formerly reuse of unit numbers of detached devices was impossible, with
this change starred devices will search from the first allowed unit for
such devices and up for free units when attaching. This means you have
to rerun config(8) before yopu compile current kernels. FYI, this is
a necessity when doing PCMCIA flashcard reading, otherwise you soon run
out of unit numbers.

I have also added a way to detach all children of a certain device, used
in a new PCMCIA framework to come, so you detach all functions on a
single PCMCIA card when pulling it out.

A few bugs wrt starred devices fixed in the autoconfig_verbose code too.
Last and least, fixed some panic messages.

Please GCC 2.8's harsher view of good style

Prevent a NULL-deref in the autoconf_verbose case

typo

Provide a way to debug probes. set autoconf_verbose via DDB, or
option AUTOCONF_VERBOSE=1
in your kernel config.

indent

hp300 needs device_register(), too.

export struct cfdata cfdata[];

As alpha uses device_register, it needs its proto too

no mknod in chroot

Remove unnecessary prototype

pretty

-Wall & -Wstrict-prototype fixes

Added support for user modifiable kernel at boot (-c) /maja

sick netbsd alpha hack (why not do what other ports do?)

Make sure config_init gets called before consinit, as at least one port
actually uses config_* functions for console attachment. Document
config_init should not call malloc as the move was over the mem-init
function calls too.

Pull in John Kohl's [jtk@netbsd.org] most recent (15Apr96) APM and PCMCIA work
(original PCMCIA framework by Stefan Grefen [grefen@convex.com]).

partial sync with netbsd 960418, more to come

NetBSD 960317 merge

From NetBSD: 960217 merge

branches: 1.1.1;
Initial revision

Improve unhibernate performance (30% on some machines, another upcoming diff
shows gains up to 50%) by skipping attach of irrelevant devices, which are
tagged CD_SKIPHIBERNATE in the per-driver cfdriver. In particular, usb devices
are not attached, so they don't need to detach during the suspend-unpack-resume.
New bootblocks are required (which tell the kernel it's job is unhibernate
before configure runs)
tested by various

convert infinite msleep(9) to msleep_nsec(9)

ok mpi@

free(9) sizes for softcs.

ok tedu@

Change a bunch of (<blah> *)0 to NULL.

ok beck@ deraadt@

Simple free(9) sizes, from Matthieu <naabed AT poolp DOT org>.

Replace mountroothook_establish(9) by config_mountroot(9) a narrower API
similar to config_defer(9).

ok mikeb@, deraadt@

back out refcnt for dv_ref, there's too many hand crafted devices all
over the tree.

much encouragement from l2k15

use refcnts for the device reference counts as an example of how
refcnt(9) can be used.

Remove some includes include-what-you-use claims don't
have any direct symbols used. Tested for indirect use by compiling
amd64/i386/sparc64 kernels.

ok tedu@ deraadt@

you need atomic.h to do atomic things.

pointed out by deraadt@

use atomic ops to increment and decrement the device ref count in
device_ref and device_unref.

ok guenther@ deraadt@

bring back r1.80:

pass the size of the softc to free.

this time we only dereference the cfattach structure if we actually
need it once the refcount has dropped to 0.

tested by krw@ who found the original panic r1.80 was backed out for.
putting this in again so i can find out when else will break.

back it out properly

never tested with a make release

pass the size of the softc to free.

ok deraadt@

Introduce config_suspend_all(9) a function that invokes config_suspend(9)
on all relevant device hierarchies in the appropriate order. For now this
means mpath(4) and mainbus(4), doing mpath(4) before mainbus(4) when
suspending or powering down and doing mpath(4) after mainbus(4) when
resuming such that mpath(4) can realy on the underlying hardware being
in a functional state.

Fixes problems with unflushed disk caches on machines where mpath(4) takes
control of some of your disks.

ok dlg@

remove uneeded proc.h includes
ok mpi@ kspillner@

use mallocarray where arguments are multipled. ok deraadt

add a size argument to free. will be used soon, but for now default to 0.
after discussions with beck deraadt kettenis.

Change config_deactivate(9) to always unset the DVF_ACTIVE flag so
that drivers can check for this flag to know they have to stop doing
any hardware access and release their reference.

While here makes the ca_activate() function optional and propagates
DVACT_DEACTIVATE to the children of a device that do not implement this
interface like it is done during suspend/resume.

ok uebayasi@, deraadt@

It's been a quarter century: we can assume volatile is present with that name.

ok dlg@ mpi@ deraadt@

get rid of the assumption that the head of the alldevs list is the
"mainbus" device. this breaks when mpath is enabled because it
attaches before mainbus and therefore takes the head position.

have autoconf provide device_mainbus() which looks up mainbus_cd,
and use that instead.

discussed with deraadt who just wants mpath stuff to move forward
despite there being many ways to shine this particular turd.

printf(variable) --> printf("%s", variable)

ok deraadt@

Add a DVACT_WAKEUP op to the *_activate() API. This is called after the
kernel resumes normal (non-cold, able to run processes, etc) operation.
Previously we were relying on specific DVACT_RESUME op's in drivers
creating callback/threads themselves, but that has become too common,
indicating the need for a built-in mechanism.
ok dlg kettenis, tested by a sufficient amount of people

msgs can be const

get rid of the goop for maintaining multiple cftables since theres
no current code there to add or remove them. such code was put in
17 years ago, disabled 14 years ago, and removed 11 years ago, so
i suspect we dont really need it or the goop to support it.

very ok miod@

compare pointers to NULL

Enforce ca_activate tree-walks over the entire heirarchy for all events,
cleaning up some shutdown-hook related code on the way.
(A few drivers related to sparc are still skipped at kettenis' request)
ok kettenis mlarkin, tested by many others too

Revamp the sequences for suspend/hibernate -> resume so that the code
paths are reflexive. It is now possible to fail part-way through a
suspend sequence, and recover along the resume code path.
Split DVACT_SUSPEND by adding a new DVACT_POWERDOWN method is used
after hibernate (and suspend too) to finish the job. Some drivers
must be converted at the same time to use this instead of shutdown hooks
(the others will follow at a later time)
ok kettenis mlarkin

Remove config_activate() and DVACT_ACTIVATE. PCMCIA's the only thing
that's ever used it, and it's long since been changed to use
DVACT_{QUIESCE,SUSPEND,RESUME} instead.

ok deraadt@, dlg@; miod@ also agreed with this idea when I brought it
up a few weeks ago

serialize attach and detach of device sub-trees -- only one device
sub-tree may attach or detach at a time. attach and detach will sleep
against each other.
this is fixing (working around?) some bizzare corner cases that have
been seen (but not fully diagnosed) where the device trees, disk registration
subsystem, and other things could get messed up. one could argue though
that this serialization is a very good thing; it is easier than adding piles
of locks in various other places.
ok matthew jsing

Add a few KASSERTs to config_attach() for sanity to make sure we don't
try to reuse device unit numbers and to check that the device pointer
array is allocated and large enough.

Also, improve the panic message generated by config_detach() when we
detect that we're detaching a device that still has children.

Discussed with deraadt@ while trying to brainstorm ways that
interleaving config_attach and config_detach could blow up.

remove a useless comment

Add DVACT_QUIECE support. This is called before splhigh() and before
DVACT_SUSPEND, therefore DVACT_QUIECE can do standard sleeping operations
to get ready.
Discussed quite a while back with kettenis and jakemsr, oga suddenly needed
it as well and wrote half of it, so it was time to finish it.
proofread by miod.

Silence the activation debug reporting. We are starting to get worried
that there are interactions when we print in the various (saved, post-saved,
pre-restored, and restored) vga states, especially with how intel drm is
getting involved. If your machine has issues, you may want to activate
these messages again to help debug things better.
ok mlarkin

Remove leftover debug printf in config_activate_children(); ok deraadt

Make config_activate_children() more useable as a suspend-and-resume
device tree walker, and add config_suspend() as well.
ok mlarkin pirofti, discussion with kettenis

Add basic suspend/resume autoconf functionality.

Okay deraadt@, kettenis@, mlarkin@.

Get rid of devact enum, substitute it with an int and coresponding defines.

This is needed for the addition of further suspend/resume actions.

Okay deraadt@, marco@.

Allow devices attaching at root (such as softraid) to be disabled in UKC.
ok marco@ deraadt@

Since the devices that we detach may have children we cannot rely on the
"previous" device still being there, so if we detach a device, we really
have to start from scratch again (like the comment on this code actually
says we should). Simplifies the code as a bonus.

ok dlg@, deraadt@

get rid of the unused cond_predicate_t typedef.

ok miod@ deraadt@ krw@

Use M_ZERO in a few more places to shave bytes from the kernel.

eyeballed and ok dlg@

the (root) stuff is ugly, so print "at root", so it looks more like
regular things. everyone here buys it, and i'm 90% sure miod will be
ok with it

kill __HAVE_DEVICE_REGISTER by requiring all architectures to have a
device_register() function -- even if it does nothing. reduces the
cpp-based blather different between architectures
idea ok'd by miod; tested on all architectures (except a few miod will
need to cleanup because he has them)

evcnt_attach() is no longer used; ok miod

Remove device id from hotplug events.

ok henning

more fallout from the sensors: Every single ramdisk got broken.

Include device id in hotplug events. This will be used by ntpd to check
sensors on attach/detach. hotplugd changes following in a minute.

ok henning.

cf_unit should have a value between 0 and cd_ndevs. So set cf_unit to 0 when
cd_ndevs is set to 0.

"No kidding!" deraadt@ "Seems to make sense to me." beck@ "Seems right." miod@

Detach children devices in reverse order. This gives better display, and
allows cf_unit to be decremented correctly.

'put it in' deraadt@ "i think it's good' tedu@

When assigning a unit number, treat running off the end of the
cd_devs[] array the same as finding a NULL entry.

Fixes inconsistant assignment of unit numbers to media slots on the
Neodio 8-in-1 USB card reader donated by Manuel Pata. Previously the
unit numbers would vary depending on whether it had attached an even
or odd number of times.

'reads ok' marco@ ok miod@

Remove now unused zeroref member in cfattach structures.

In config_detach_children(), after detaching a device, restart the device
list walk from its parent device, as the device which was following it may
have been one of its children, and thus gone as well. Found the hard way
using the strict queue macros.

Feedback and help toby@, ok deraadt@

ansi and deregister. No binary change.

Use list and queue macros where applicable to make the code easier to read;
no change in compiler assembly output.

Create the init process earlier, before the root filesystem is mounted,
and have it stall on a semaphore. This allows all kthread creations which
could have been requested during autoconf to be processed before root
is mounted as well.

This causes umass devices attached to any usb with flags 1 (such as on macppc)
to configure properly instead of panicing the kernel at mountroot time.

From NetBSD; tested by various.

Fix a minor bug where indirect config crashes the kernel
when match verbose is turend on.

Devices hot plugging support.
The hotplug pseudo-device passes device attachment and detachment events
to userland. When a device attaches or detaches, the corresponding event
is queued. The events can then be obtained from the queue through the
read(2) call on the /dev/hotplug device file. Each event consists of
event type (attach/detach), device class (DV_*) and device name (sd1 e.g.).

We have hotplug pseudo-device on alpha, amd64, i386, macppc and sparc64.
Since it was tested only on i386 other archs has it commented out
in GENERIC until tested.

The idea liked peter@ tedu@ drahn@ millert@ marco@ henning@.
Ok deraadt@.

Remove the advertising clause in the UCB license which Berkeley
rescinded 22 July 1999. Proofed by myself and Theo.

string fixes; tedu ok

Tweak previous sprintf -> snprintf fix to eliminate unnecessary
intermediate variable.

Suggested by deraadt@.

ok henning@ millert@ tdeval@

Use snprintf() to simplify device name construction, removing now
superfluous variables and the function 'number()'.

ok deraadt@ and millert@

Remove some '#if 0' cruft.

Introduce a new file, machine/internal_types.h, to hold that specific arch
type characteristics.

internal_types.h will contain only settings invisible from standard C, e.g.,
in the __* or _[A-Z]* namespace, and be reused by files like limits.h.

This allows us to shorten machine/limits.h greatly, as all the common defines
are now in sys/limits.h, plus a small stub in internal_types.h.

Tested on all arches as far as I know.
Approved after discussion with art, millert, deraadt, and others.

First round of __P removal in sys

branches: 1.29.4;
Add __HAVE_DEVICE_REGISTER to hp300/alpha (sparc64 already had it) and
clean up the arch specific #ifdef's in subr_autoconf.c; from NetBSD.
(tested by me on alpha/sparc64 and miod on hp300)

implement autoconf mechanism using device_register(); from NetBSD.

range error; < cd_ndevs, not <=

Added config_detach_children, config_activate_children.

Added ref counting to devices: device_ref, device_unref and a new method
for devices : ca_zeroref, called when the ref count hits zero.

Note that ca_zeroref may be called from interrupt context.

branches: 1.25.4;
From NetBSD; new device detaching infrastructure.

From NetBSD; config_defer to defer configuration of a device until after all
other siblings.

Remove obsolete detach logic in preparation of a newer one

panic prints a newline for you, don't do it in the panic string

Formerly reuse of unit numbers of detached devices was impossible, with
this change starred devices will search from the first allowed unit for
such devices and up for free units when attaching. This means you have
to rerun config(8) before yopu compile current kernels. FYI, this is
a necessity when doing PCMCIA flashcard reading, otherwise you soon run
out of unit numbers.

I have also added a way to detach all children of a certain device, used
in a new PCMCIA framework to come, so you detach all functions on a
single PCMCIA card when pulling it out.

A few bugs wrt starred devices fixed in the autoconfig_verbose code too.
Last and least, fixed some panic messages.

Please GCC 2.8's harsher view of good style

Prevent a NULL-deref in the autoconf_verbose case

typo

Provide a way to debug probes. set autoconf_verbose via DDB, or
option AUTOCONF_VERBOSE=1
in your kernel config.

indent

hp300 needs device_register(), too.

export struct cfdata cfdata[];

As alpha uses device_register, it needs its proto too

no mknod in chroot

Remove unnecessary prototype

pretty

-Wall & -Wstrict-prototype fixes

Added support for user modifiable kernel at boot (-c) /maja

sick netbsd alpha hack (why not do what other ports do?)

Make sure config_init gets called before consinit, as at least one port
actually uses config_* functions for console attachment. Document
config_init should not call malloc as the move was over the mem-init
function calls too.

Pull in John Kohl's [jtk@netbsd.org] most recent (15Apr96) APM and PCMCIA work
(original PCMCIA framework by Stefan Grefen [grefen@convex.com]).

partial sync with netbsd 960418, more to come

NetBSD 960317 merge

From NetBSD: 960217 merge

branches: 1.1.1;
Initial revision

convert infinite msleep(9) to msleep_nsec(9)

ok mpi@

free(9) sizes for softcs.

ok tedu@

Change a bunch of (<blah> *)0 to NULL.

ok beck@ deraadt@

Simple free(9) sizes, from Matthieu <naabed AT poolp DOT org>.

Replace mountroothook_establish(9) by config_mountroot(9) a narrower API
similar to config_defer(9).

ok mikeb@, deraadt@

back out refcnt for dv_ref, there's too many hand crafted devices all
over the tree.

much encouragement from l2k15

use refcnts for the device reference counts as an example of how
refcnt(9) can be used.

Remove some includes include-what-you-use claims don't
have any direct symbols used. Tested for indirect use by compiling
amd64/i386/sparc64 kernels.

ok tedu@ deraadt@

you need atomic.h to do atomic things.

pointed out by deraadt@

use atomic ops to increment and decrement the device ref count in
device_ref and device_unref.

ok guenther@ deraadt@

bring back r1.80:

pass the size of the softc to free.

this time we only dereference the cfattach structure if we actually
need it once the refcount has dropped to 0.

tested by krw@ who found the original panic r1.80 was backed out for.
putting this in again so i can find out when else will break.

back it out properly

never tested with a make release

pass the size of the softc to free.

ok deraadt@

Introduce config_suspend_all(9) a function that invokes config_suspend(9)
on all relevant device hierarchies in the appropriate order. For now this
means mpath(4) and mainbus(4), doing mpath(4) before mainbus(4) when
suspending or powering down and doing mpath(4) after mainbus(4) when
resuming such that mpath(4) can realy on the underlying hardware being
in a functional state.

Fixes problems with unflushed disk caches on machines where mpath(4) takes
control of some of your disks.

ok dlg@

remove uneeded proc.h includes
ok mpi@ kspillner@

use mallocarray where arguments are multipled. ok deraadt

add a size argument to free. will be used soon, but for now default to 0.
after discussions with beck deraadt kettenis.

Change config_deactivate(9) to always unset the DVF_ACTIVE flag so
that drivers can check for this flag to know they have to stop doing
any hardware access and release their reference.

While here makes the ca_activate() function optional and propagates
DVACT_DEACTIVATE to the children of a device that do not implement this
interface like it is done during suspend/resume.

ok uebayasi@, deraadt@

It's been a quarter century: we can assume volatile is present with that name.

ok dlg@ mpi@ deraadt@

get rid of the assumption that the head of the alldevs list is the
"mainbus" device. this breaks when mpath is enabled because it
attaches before mainbus and therefore takes the head position.

have autoconf provide device_mainbus() which looks up mainbus_cd,
and use that instead.

discussed with deraadt who just wants mpath stuff to move forward
despite there being many ways to shine this particular turd.

printf(variable) --> printf("%s", variable)

ok deraadt@

Add a DVACT_WAKEUP op to the *_activate() API. This is called after the
kernel resumes normal (non-cold, able to run processes, etc) operation.
Previously we were relying on specific DVACT_RESUME op's in drivers
creating callback/threads themselves, but that has become too common,
indicating the need for a built-in mechanism.
ok dlg kettenis, tested by a sufficient amount of people

msgs can be const

get rid of the goop for maintaining multiple cftables since theres
no current code there to add or remove them. such code was put in
17 years ago, disabled 14 years ago, and removed 11 years ago, so
i suspect we dont really need it or the goop to support it.

very ok miod@

compare pointers to NULL

Enforce ca_activate tree-walks over the entire heirarchy for all events,
cleaning up some shutdown-hook related code on the way.
(A few drivers related to sparc are still skipped at kettenis' request)
ok kettenis mlarkin, tested by many others too

Revamp the sequences for suspend/hibernate -> resume so that the code
paths are reflexive. It is now possible to fail part-way through a
suspend sequence, and recover along the resume code path.
Split DVACT_SUSPEND by adding a new DVACT_POWERDOWN method is used
after hibernate (and suspend too) to finish the job. Some drivers
must be converted at the same time to use this instead of shutdown hooks
(the others will follow at a later time)
ok kettenis mlarkin

Remove config_activate() and DVACT_ACTIVATE. PCMCIA's the only thing
that's ever used it, and it's long since been changed to use
DVACT_{QUIESCE,SUSPEND,RESUME} instead.

ok deraadt@, dlg@; miod@ also agreed with this idea when I brought it
up a few weeks ago

serialize attach and detach of device sub-trees -- only one device
sub-tree may attach or detach at a time. attach and detach will sleep
against each other.
this is fixing (working around?) some bizzare corner cases that have
been seen (but not fully diagnosed) where the device trees, disk registration
subsystem, and other things could get messed up. one could argue though
that this serialization is a very good thing; it is easier than adding piles
of locks in various other places.
ok matthew jsing

Add a few KASSERTs to config_attach() for sanity to make sure we don't
try to reuse device unit numbers and to check that the device pointer
array is allocated and large enough.

Also, improve the panic message generated by config_detach() when we
detect that we're detaching a device that still has children.

Discussed with deraadt@ while trying to brainstorm ways that
interleaving config_attach and config_detach could blow up.

remove a useless comment

Add DVACT_QUIECE support. This is called before splhigh() and before
DVACT_SUSPEND, therefore DVACT_QUIECE can do standard sleeping operations
to get ready.
Discussed quite a while back with kettenis and jakemsr, oga suddenly needed
it as well and wrote half of it, so it was time to finish it.
proofread by miod.

Silence the activation debug reporting. We are starting to get worried
that there are interactions when we print in the various (saved, post-saved,
pre-restored, and restored) vga states, especially with how intel drm is
getting involved. If your machine has issues, you may want to activate
these messages again to help debug things better.
ok mlarkin

Remove leftover debug printf in config_activate_children(); ok deraadt

Make config_activate_children() more useable as a suspend-and-resume
device tree walker, and add config_suspend() as well.
ok mlarkin pirofti, discussion with kettenis

Add basic suspend/resume autoconf functionality.

Okay deraadt@, kettenis@, mlarkin@.

Get rid of devact enum, substitute it with an int and coresponding defines.

This is needed for the addition of further suspend/resume actions.

Okay deraadt@, marco@.

Allow devices attaching at root (such as softraid) to be disabled in UKC.
ok marco@ deraadt@

Since the devices that we detach may have children we cannot rely on the
"previous" device still being there, so if we detach a device, we really
have to start from scratch again (like the comment on this code actually
says we should). Simplifies the code as a bonus.

ok dlg@, deraadt@

get rid of the unused cond_predicate_t typedef.

ok miod@ deraadt@ krw@

Use M_ZERO in a few more places to shave bytes from the kernel.

eyeballed and ok dlg@

the (root) stuff is ugly, so print "at root", so it looks more like
regular things. everyone here buys it, and i'm 90% sure miod will be
ok with it

kill __HAVE_DEVICE_REGISTER by requiring all architectures to have a
device_register() function -- even if it does nothing. reduces the
cpp-based blather different between architectures
idea ok'd by miod; tested on all architectures (except a few miod will
need to cleanup because he has them)

evcnt_attach() is no longer used; ok miod

Remove device id from hotplug events.

ok henning

more fallout from the sensors: Every single ramdisk got broken.

Include device id in hotplug events. This will be used by ntpd to check
sensors on attach/detach. hotplugd changes following in a minute.

ok henning.

cf_unit should have a value between 0 and cd_ndevs. So set cf_unit to 0 when
cd_ndevs is set to 0.

"No kidding!" deraadt@ "Seems to make sense to me." beck@ "Seems right." miod@

Detach children devices in reverse order. This gives better display, and
allows cf_unit to be decremented correctly.

'put it in' deraadt@ "i think it's good' tedu@

When assigning a unit number, treat running off the end of the
cd_devs[] array the same as finding a NULL entry.

Fixes inconsistant assignment of unit numbers to media slots on the
Neodio 8-in-1 USB card reader donated by Manuel Pata. Previously the
unit numbers would vary depending on whether it had attached an even
or odd number of times.

'reads ok' marco@ ok miod@

Remove now unused zeroref member in cfattach structures.

In config_detach_children(), after detaching a device, restart the device
list walk from its parent device, as the device which was following it may
have been one of its children, and thus gone as well. Found the hard way
using the strict queue macros.

Feedback and help toby@, ok deraadt@

ansi and deregister. No binary change.

Use list and queue macros where applicable to make the code easier to read;
no change in compiler assembly output.

Create the init process earlier, before the root filesystem is mounted,
and have it stall on a semaphore. This allows all kthread creations which
could have been requested during autoconf to be processed before root
is mounted as well.

This causes umass devices attached to any usb with flags 1 (such as on macppc)
to configure properly instead of panicing the kernel at mountroot time.

From NetBSD; tested by various.

Fix a minor bug where indirect config crashes the kernel
when match verbose is turend on.

Devices hot plugging support.
The hotplug pseudo-device passes device attachment and detachment events
to userland. When a device attaches or detaches, the corresponding event
is queued. The events can then be obtained from the queue through the
read(2) call on the /dev/hotplug device file. Each event consists of
event type (attach/detach), device class (DV_*) and device name (sd1 e.g.).

We have hotplug pseudo-device on alpha, amd64, i386, macppc and sparc64.
Since it was tested only on i386 other archs has it commented out
in GENERIC until tested.

The idea liked peter@ tedu@ drahn@ millert@ marco@ henning@.
Ok deraadt@.

Remove the advertising clause in the UCB license which Berkeley
rescinded 22 July 1999. Proofed by myself and Theo.

string fixes; tedu ok

Tweak previous sprintf -> snprintf fix to eliminate unnecessary
intermediate variable.

Suggested by deraadt@.

ok henning@ millert@ tdeval@

Use snprintf() to simplify device name construction, removing now
superfluous variables and the function 'number()'.

ok deraadt@ and millert@

Remove some '#if 0' cruft.

Introduce a new file, machine/internal_types.h, to hold that specific arch
type characteristics.

internal_types.h will contain only settings invisible from standard C, e.g.,
in the __* or _[A-Z]* namespace, and be reused by files like limits.h.

This allows us to shorten machine/limits.h greatly, as all the common defines
are now in sys/limits.h, plus a small stub in internal_types.h.

Tested on all arches as far as I know.
Approved after discussion with art, millert, deraadt, and others.

First round of __P removal in sys

branches: 1.29.4;
Add __HAVE_DEVICE_REGISTER to hp300/alpha (sparc64 already had it) and
clean up the arch specific #ifdef's in subr_autoconf.c; from NetBSD.
(tested by me on alpha/sparc64 and miod on hp300)

implement autoconf mechanism using device_register(); from NetBSD.

range error; < cd_ndevs, not <=

Added config_detach_children, config_activate_children.

Added ref counting to devices: device_ref, device_unref and a new method
for devices : ca_zeroref, called when the ref count hits zero.

Note that ca_zeroref may be called from interrupt context.

branches: 1.25.4;
From NetBSD; new device detaching infrastructure.

From NetBSD; config_defer to defer configuration of a device until after all
other siblings.

Remove obsolete detach logic in preparation of a newer one

panic prints a newline for you, don't do it in the panic string

Formerly reuse of unit numbers of detached devices was impossible, with
this change starred devices will search from the first allowed unit for
such devices and up for free units when attaching. This means you have
to rerun config(8) before yopu compile current kernels. FYI, this is
a necessity when doing PCMCIA flashcard reading, otherwise you soon run
out of unit numbers.

I have also added a way to detach all children of a certain device, used
in a new PCMCIA framework to come, so you detach all functions on a
single PCMCIA card when pulling it out.

A few bugs wrt starred devices fixed in the autoconfig_verbose code too.
Last and least, fixed some panic messages.

Please GCC 2.8's harsher view of good style

Prevent a NULL-deref in the autoconf_verbose case

typo

Provide a way to debug probes. set autoconf_verbose via DDB, or
option AUTOCONF_VERBOSE=1
in your kernel config.

indent

hp300 needs device_register(), too.

export struct cfdata cfdata[];

As alpha uses device_register, it needs its proto too

no mknod in chroot

Remove unnecessary prototype

pretty

-Wall & -Wstrict-prototype fixes

Added support for user modifiable kernel at boot (-c) /maja

sick netbsd alpha hack (why not do what other ports do?)

Make sure config_init gets called before consinit, as at least one port
actually uses config_* functions for console attachment. Document
config_init should not call malloc as the move was over the mem-init
function calls too.

Pull in John Kohl's [jtk@netbsd.org] most recent (15Apr96) APM and PCMCIA work
(original PCMCIA framework by Stefan Grefen [grefen@convex.com]).

partial sync with netbsd 960418, more to come

NetBSD 960317 merge

From NetBSD: 960217 merge

branches: 1.1.1;
Initial revision

free(9) sizes for softcs.

ok tedu@

Change a bunch of (<blah> *)0 to NULL.

ok beck@ deraadt@

Simple free(9) sizes, from Matthieu <naabed AT poolp DOT org>.

Replace mountroothook_establish(9) by config_mountroot(9) a narrower API
similar to config_defer(9).

ok mikeb@, deraadt@

back out refcnt for dv_ref, there's too many hand crafted devices all
over the tree.

much encouragement from l2k15

use refcnts for the device reference counts as an example of how
refcnt(9) can be used.

Remove some includes include-what-you-use claims don't
have any direct symbols used. Tested for indirect use by compiling
amd64/i386/sparc64 kernels.

ok tedu@ deraadt@

you need atomic.h to do atomic things.

pointed out by deraadt@

use atomic ops to increment and decrement the device ref count in
device_ref and device_unref.

ok guenther@ deraadt@

bring back r1.80:

pass the size of the softc to free.

this time we only dereference the cfattach structure if we actually
need it once the refcount has dropped to 0.

tested by krw@ who found the original panic r1.80 was backed out for.
putting this in again so i can find out when else will break.

back it out properly

never tested with a make release

pass the size of the softc to free.

ok deraadt@

Introduce config_suspend_all(9) a function that invokes config_suspend(9)
on all relevant device hierarchies in the appropriate order. For now this
means mpath(4) and mainbus(4), doing mpath(4) before mainbus(4) when
suspending or powering down and doing mpath(4) after mainbus(4) when
resuming such that mpath(4) can realy on the underlying hardware being
in a functional state.

Fixes problems with unflushed disk caches on machines where mpath(4) takes
control of some of your disks.

ok dlg@

remove uneeded proc.h includes
ok mpi@ kspillner@

use mallocarray where arguments are multipled. ok deraadt

add a size argument to free. will be used soon, but for now default to 0.
after discussions with beck deraadt kettenis.

Change config_deactivate(9) to always unset the DVF_ACTIVE flag so
that drivers can check for this flag to know they have to stop doing
any hardware access and release their reference.

While here makes the ca_activate() function optional and propagates
DVACT_DEACTIVATE to the children of a device that do not implement this
interface like it is done during suspend/resume.

ok uebayasi@, deraadt@

It's been a quarter century: we can assume volatile is present with that name.

ok dlg@ mpi@ deraadt@

get rid of the assumption that the head of the alldevs list is the
"mainbus" device. this breaks when mpath is enabled because it
attaches before mainbus and therefore takes the head position.

have autoconf provide device_mainbus() which looks up mainbus_cd,
and use that instead.

discussed with deraadt who just wants mpath stuff to move forward
despite there being many ways to shine this particular turd.

printf(variable) --> printf("%s", variable)

ok deraadt@

Add a DVACT_WAKEUP op to the *_activate() API. This is called after the
kernel resumes normal (non-cold, able to run processes, etc) operation.
Previously we were relying on specific DVACT_RESUME op's in drivers
creating callback/threads themselves, but that has become too common,
indicating the need for a built-in mechanism.
ok dlg kettenis, tested by a sufficient amount of people

msgs can be const

get rid of the goop for maintaining multiple cftables since theres
no current code there to add or remove them. such code was put in
17 years ago, disabled 14 years ago, and removed 11 years ago, so
i suspect we dont really need it or the goop to support it.

very ok miod@

compare pointers to NULL

Enforce ca_activate tree-walks over the entire heirarchy for all events,
cleaning up some shutdown-hook related code on the way.
(A few drivers related to sparc are still skipped at kettenis' request)
ok kettenis mlarkin, tested by many others too

Revamp the sequences for suspend/hibernate -> resume so that the code
paths are reflexive. It is now possible to fail part-way through a
suspend sequence, and recover along the resume code path.
Split DVACT_SUSPEND by adding a new DVACT_POWERDOWN method is used
after hibernate (and suspend too) to finish the job. Some drivers
must be converted at the same time to use this instead of shutdown hooks
(the others will follow at a later time)
ok kettenis mlarkin

Remove config_activate() and DVACT_ACTIVATE. PCMCIA's the only thing
that's ever used it, and it's long since been changed to use
DVACT_{QUIESCE,SUSPEND,RESUME} instead.

ok deraadt@, dlg@; miod@ also agreed with this idea when I brought it
up a few weeks ago

serialize attach and detach of device sub-trees -- only one device
sub-tree may attach or detach at a time. attach and detach will sleep
against each other.
this is fixing (working around?) some bizzare corner cases that have
been seen (but not fully diagnosed) where the device trees, disk registration
subsystem, and other things could get messed up. one could argue though
that this serialization is a very good thing; it is easier than adding piles
of locks in various other places.
ok matthew jsing

Add a few KASSERTs to config_attach() for sanity to make sure we don't
try to reuse device unit numbers and to check that the device pointer
array is allocated and large enough.

Also, improve the panic message generated by config_detach() when we
detect that we're detaching a device that still has children.

Discussed with deraadt@ while trying to brainstorm ways that
interleaving config_attach and config_detach could blow up.

remove a useless comment

Add DVACT_QUIECE support. This is called before splhigh() and before
DVACT_SUSPEND, therefore DVACT_QUIECE can do standard sleeping operations
to get ready.
Discussed quite a while back with kettenis and jakemsr, oga suddenly needed
it as well and wrote half of it, so it was time to finish it.
proofread by miod.

Silence the activation debug reporting. We are starting to get worried
that there are interactions when we print in the various (saved, post-saved,
pre-restored, and restored) vga states, especially with how intel drm is
getting involved. If your machine has issues, you may want to activate
these messages again to help debug things better.
ok mlarkin

Remove leftover debug printf in config_activate_children(); ok deraadt