remove unneeded includes in arch/i386
ok krw@ miod@

Use the MI interrupt enable/distable API instead of the MD one on i386 and
remove the MD API.

ok deraadt@

Fix a wrong memcmp in k7/k8 powernow code.

ok brynet

Recommit preparation for i386 Meltdown fix after OpenBSD 6.3 release.

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

iBackout the preparations for fixing Meltdown on i386. The task was
only halfway done and the current state does not help anybody. For
OpenBSD 6.3 release go back to the original code before 2018/03/13.
This gives us a stable release and the changes will come back later.
discussed with guenther@ deraadt@ hshoexer@

Preparation for i386 Meltdown fix:

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

If you use sys/param.h, you don't need sys/types.h

sizes for free(); ok semarie

remove uneeded proc.h includes
ok mpi@ kspillner@

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

remove proc.h include from uvm_map.h. This has far reaching effects, as
sysctl.h was reliant on this particular include, and many drivers included
sysctl.h unnecessarily. remove sysctl.h or add proc.h as needed.
ok deraadt

Based upon a diff orignally submitted by devin smith, Do two things run the
check that the maxfid and the highest reported acpi p_state frequency agree only
after actually getting the acpi p_state. Secondly prefer the legacy powernow
table over ACPI because machines of this vintage of are likely to have broken
ACPI implementations.

Tested by devin smith and tilo stritzky.
ok beck@

Adapt powernow-k7.c driver to use the ACPI _PSS table for retreiving
p_state data based on similar work done to powernow-k8.c. This work was
done by Devin Smith <devin at devinsmith dot net> and this commit is a
slightly modified verson of his last diff to tech@, thanks again for your
work.

Choose the state for a request performance level based on calculations
involving the number of states and not the frequencies of the states,
which can lead to strange distributions of the states over the hw.setperf
range (0-100). Tested by many.

ok canacar, tedu

"#ifdef is a tool of the weak!"
Rename pentium_mhz to cpuspeed which is consistant with amd64 making
shared ACPI code less nasty.
ok marco, deraadt

Complete gwk's previous patch to stop setperf methods from returning
errors to userland: make all cpu_setperf functions return void.

Tested by many, ok gwk@

Fix the spellings of Cool'n'Quiet and PowerNow! in a couple of places.

ok deraadt@

s/Mhz/MHz/ in comments and printf() strings

ok jsg@

If we don't find a matching CPU signature using the base CPUID call, try
matching against the extended CPUID (0x80000001) signature. Problem
found by, and patch based on work by jason@.
ok jason@

Martin V�giard <deadbug AT gmail.com> has been kind enough to rescind the
third clause of his license and strike the phrase "All rights reserved", so we
can consolidate both license blocks into one. Thanks!

From k*_powernow_setperf() always return 0, i.e.
Do not return low level errors to userspace where they will not/can not be
coped with.
: ----------------------------------------------------------------------

Change the value returned if the CPU fails to honor a request to
adjust frequency from 1 (EPERM) to EIO.
Suggested by and ok tedu@

Teach powernow about setperf prioritys, add a setperf_prio to amd64.
ok uwe@, "good" tedu@

kill trainling spaces

* Plug memory leak if init fails; ok gwk@

Correct the heuristic to determin Cool`n'Quiet vs. Powernow! spotted in
the linux driver.
Correct the the cpu signature used to determin if the chip is affected
by errata a0, verified in the linux driver.
Some demagification and KNF.
ok dlg@

Ensure that we check if the transition succeeded before we adjust
pentium_mhz on K7, and print a usefull diagnostic message if it does
not on K7 and K8.
"makes sense to me" dlg@

Don't bother setting the frequency entry in the available states table
to -1 we never check for it. (As has been the case on K8 for a long time).
ok dlg@

Eliminate some pointers we dont use, originally for multiprocessor support,
K7 MP parts do not support Cool'N'Quiet.
"also makes sense" dlg@

Rid k7_powernow_setperf() of some bad math, this code expected the values
to be expressed in hertz not megahertz.
"makes sense to me" dlg@

Change the return value in k7_powernow_setperf() from 1 to 0 so that it
does not print "operation not permitted" even though the operation was
successful.
ok dlg@

Loop on the number of PST reported by the bios not a magic number.
Fixes PR 5075.
"go for it" dlg@

Small hex-vs-decimal typo from gwk, and tested by me (with cpusig check
disabled because my BIOS sucks).

a bit of demagification. this moves the checks for the cpu scaling features
out of machdep and into powernow-k7 and -k8. machdep now just figures out
if its the right type of cpu before calling the powernow code which
figures out if the scaling is supported.

from gwk

spacing

another dangling semicolon from Matthias Bauer

some cleanup from gordon

lost a diff from gordon, only attach if we can do something useful

malloc(M_WAITOK) is bad in init code, M_DEVBUF is nicer than M_TEMP,
and if the cpu gets stuck, don't spin too long.

update support for powernow (cool and quiet) on k7, and add support
for k8. preliminary and not well tested yet. from freebsd via
gordon klok.

fix addr range in the comments

unmap bios when done. from form@pdp-11.org.ru via grange

names of frequencies aren't interesting, and require ugly snprintf
manipulation theo doesn't like. just print number of states.
this way doesn't leak the memory for the string either.

setperf driver for powernow in amd k7 cpus. derived from a diff to
tech-i386@netbsd by martin vegiard, and adapted to fit in here.
as yet untested. ok deraadt@

Use the MI interrupt enable/distable API instead of the MD one on i386 and
remove the MD API.

ok deraadt@

Fix a wrong memcmp in k7/k8 powernow code.

ok brynet

Recommit preparation for i386 Meltdown fix after OpenBSD 6.3 release.

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

iBackout the preparations for fixing Meltdown on i386. The task was
only halfway done and the current state does not help anybody. For
OpenBSD 6.3 release go back to the original code before 2018/03/13.
This gives us a stable release and the changes will come back later.
discussed with guenther@ deraadt@ hshoexer@

Preparation for i386 Meltdown fix:

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

If you use sys/param.h, you don't need sys/types.h

sizes for free(); ok semarie

remove uneeded proc.h includes
ok mpi@ kspillner@

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

remove proc.h include from uvm_map.h. This has far reaching effects, as
sysctl.h was reliant on this particular include, and many drivers included
sysctl.h unnecessarily. remove sysctl.h or add proc.h as needed.
ok deraadt

Based upon a diff orignally submitted by devin smith, Do two things run the
check that the maxfid and the highest reported acpi p_state frequency agree only
after actually getting the acpi p_state. Secondly prefer the legacy powernow
table over ACPI because machines of this vintage of are likely to have broken
ACPI implementations.

Tested by devin smith and tilo stritzky.
ok beck@

Adapt powernow-k7.c driver to use the ACPI _PSS table for retreiving
p_state data based on similar work done to powernow-k8.c. This work was
done by Devin Smith <devin at devinsmith dot net> and this commit is a
slightly modified verson of his last diff to tech@, thanks again for your
work.

Choose the state for a request performance level based on calculations
involving the number of states and not the frequencies of the states,
which can lead to strange distributions of the states over the hw.setperf
range (0-100). Tested by many.

ok canacar, tedu

"#ifdef is a tool of the weak!"
Rename pentium_mhz to cpuspeed which is consistant with amd64 making
shared ACPI code less nasty.
ok marco, deraadt

Complete gwk's previous patch to stop setperf methods from returning
errors to userland: make all cpu_setperf functions return void.

Tested by many, ok gwk@

Fix the spellings of Cool'n'Quiet and PowerNow! in a couple of places.

ok deraadt@

s/Mhz/MHz/ in comments and printf() strings

ok jsg@

If we don't find a matching CPU signature using the base CPUID call, try
matching against the extended CPUID (0x80000001) signature. Problem
found by, and patch based on work by jason@.
ok jason@

Martin V�giard <deadbug AT gmail.com> has been kind enough to rescind the
third clause of his license and strike the phrase "All rights reserved", so we
can consolidate both license blocks into one. Thanks!

From k*_powernow_setperf() always return 0, i.e.
Do not return low level errors to userspace where they will not/can not be
coped with.
: ----------------------------------------------------------------------

Change the value returned if the CPU fails to honor a request to
adjust frequency from 1 (EPERM) to EIO.
Suggested by and ok tedu@

Teach powernow about setperf prioritys, add a setperf_prio to amd64.
ok uwe@, "good" tedu@

kill trainling spaces

* Plug memory leak if init fails; ok gwk@

Correct the heuristic to determin Cool`n'Quiet vs. Powernow! spotted in
the linux driver.
Correct the the cpu signature used to determin if the chip is affected
by errata a0, verified in the linux driver.
Some demagification and KNF.
ok dlg@

Ensure that we check if the transition succeeded before we adjust
pentium_mhz on K7, and print a usefull diagnostic message if it does
not on K7 and K8.
"makes sense to me" dlg@

Don't bother setting the frequency entry in the available states table
to -1 we never check for it. (As has been the case on K8 for a long time).
ok dlg@

Eliminate some pointers we dont use, originally for multiprocessor support,
K7 MP parts do not support Cool'N'Quiet.
"also makes sense" dlg@

Rid k7_powernow_setperf() of some bad math, this code expected the values
to be expressed in hertz not megahertz.
"makes sense to me" dlg@

Change the return value in k7_powernow_setperf() from 1 to 0 so that it
does not print "operation not permitted" even though the operation was
successful.
ok dlg@

Loop on the number of PST reported by the bios not a magic number.
Fixes PR 5075.
"go for it" dlg@

Small hex-vs-decimal typo from gwk, and tested by me (with cpusig check
disabled because my BIOS sucks).

a bit of demagification. this moves the checks for the cpu scaling features
out of machdep and into powernow-k7 and -k8. machdep now just figures out
if its the right type of cpu before calling the powernow code which
figures out if the scaling is supported.

from gwk

spacing

another dangling semicolon from Matthias Bauer

some cleanup from gordon

lost a diff from gordon, only attach if we can do something useful

malloc(M_WAITOK) is bad in init code, M_DEVBUF is nicer than M_TEMP,
and if the cpu gets stuck, don't spin too long.

update support for powernow (cool and quiet) on k7, and add support
for k8. preliminary and not well tested yet. from freebsd via
gordon klok.

fix addr range in the comments

unmap bios when done. from form@pdp-11.org.ru via grange

names of frequencies aren't interesting, and require ugly snprintf
manipulation theo doesn't like. just print number of states.
this way doesn't leak the memory for the string either.

setperf driver for powernow in amd k7 cpus. derived from a diff to
tech-i386@netbsd by martin vegiard, and adapted to fit in here.
as yet untested. ok deraadt@

Fix a wrong memcmp in k7/k8 powernow code.

ok brynet

Recommit preparation for i386 Meltdown fix after OpenBSD 6.3 release.

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

iBackout the preparations for fixing Meltdown on i386. The task was
only halfway done and the current state does not help anybody. For
OpenBSD 6.3 release go back to the original code before 2018/03/13.
This gives us a stable release and the changes will come back later.
discussed with guenther@ deraadt@ hshoexer@

Preparation for i386 Meltdown fix:

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

If you use sys/param.h, you don't need sys/types.h

sizes for free(); ok semarie

remove uneeded proc.h includes
ok mpi@ kspillner@

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

remove proc.h include from uvm_map.h. This has far reaching effects, as
sysctl.h was reliant on this particular include, and many drivers included
sysctl.h unnecessarily. remove sysctl.h or add proc.h as needed.
ok deraadt

Based upon a diff orignally submitted by devin smith, Do two things run the
check that the maxfid and the highest reported acpi p_state frequency agree only
after actually getting the acpi p_state. Secondly prefer the legacy powernow
table over ACPI because machines of this vintage of are likely to have broken
ACPI implementations.

Tested by devin smith and tilo stritzky.
ok beck@

Adapt powernow-k7.c driver to use the ACPI _PSS table for retreiving
p_state data based on similar work done to powernow-k8.c. This work was
done by Devin Smith <devin at devinsmith dot net> and this commit is a
slightly modified verson of his last diff to tech@, thanks again for your
work.

Choose the state for a request performance level based on calculations
involving the number of states and not the frequencies of the states,
which can lead to strange distributions of the states over the hw.setperf
range (0-100). Tested by many.

ok canacar, tedu

"#ifdef is a tool of the weak!"
Rename pentium_mhz to cpuspeed which is consistant with amd64 making
shared ACPI code less nasty.
ok marco, deraadt

Complete gwk's previous patch to stop setperf methods from returning
errors to userland: make all cpu_setperf functions return void.

Tested by many, ok gwk@

Fix the spellings of Cool'n'Quiet and PowerNow! in a couple of places.

ok deraadt@

s/Mhz/MHz/ in comments and printf() strings

ok jsg@

If we don't find a matching CPU signature using the base CPUID call, try
matching against the extended CPUID (0x80000001) signature. Problem
found by, and patch based on work by jason@.
ok jason@

Martin V�giard <deadbug AT gmail.com> has been kind enough to rescind the
third clause of his license and strike the phrase "All rights reserved", so we
can consolidate both license blocks into one. Thanks!

From k*_powernow_setperf() always return 0, i.e.
Do not return low level errors to userspace where they will not/can not be
coped with.
: ----------------------------------------------------------------------

Change the value returned if the CPU fails to honor a request to
adjust frequency from 1 (EPERM) to EIO.
Suggested by and ok tedu@

Teach powernow about setperf prioritys, add a setperf_prio to amd64.
ok uwe@, "good" tedu@

kill trainling spaces

* Plug memory leak if init fails; ok gwk@

Correct the heuristic to determin Cool`n'Quiet vs. Powernow! spotted in
the linux driver.
Correct the the cpu signature used to determin if the chip is affected
by errata a0, verified in the linux driver.
Some demagification and KNF.
ok dlg@

Ensure that we check if the transition succeeded before we adjust
pentium_mhz on K7, and print a usefull diagnostic message if it does
not on K7 and K8.
"makes sense to me" dlg@

Don't bother setting the frequency entry in the available states table
to -1 we never check for it. (As has been the case on K8 for a long time).
ok dlg@

Eliminate some pointers we dont use, originally for multiprocessor support,
K7 MP parts do not support Cool'N'Quiet.
"also makes sense" dlg@

Rid k7_powernow_setperf() of some bad math, this code expected the values
to be expressed in hertz not megahertz.
"makes sense to me" dlg@

Change the return value in k7_powernow_setperf() from 1 to 0 so that it
does not print "operation not permitted" even though the operation was
successful.
ok dlg@

Loop on the number of PST reported by the bios not a magic number.
Fixes PR 5075.
"go for it" dlg@

Small hex-vs-decimal typo from gwk, and tested by me (with cpusig check
disabled because my BIOS sucks).

a bit of demagification. this moves the checks for the cpu scaling features
out of machdep and into powernow-k7 and -k8. machdep now just figures out
if its the right type of cpu before calling the powernow code which
figures out if the scaling is supported.

from gwk

spacing

another dangling semicolon from Matthias Bauer

some cleanup from gordon

lost a diff from gordon, only attach if we can do something useful

malloc(M_WAITOK) is bad in init code, M_DEVBUF is nicer than M_TEMP,
and if the cpu gets stuck, don't spin too long.

update support for powernow (cool and quiet) on k7, and add support
for k8. preliminary and not well tested yet. from freebsd via
gordon klok.

fix addr range in the comments

unmap bios when done. from form@pdp-11.org.ru via grange

names of frequencies aren't interesting, and require ugly snprintf
manipulation theo doesn't like. just print number of states.
this way doesn't leak the memory for the string either.

setperf driver for powernow in amd k7 cpus. derived from a diff to
tech-i386@netbsd by martin vegiard, and adapted to fit in here.
as yet untested. ok deraadt@

Recommit preparation for i386 Meltdown fix after OpenBSD 6.3 release.

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

iBackout the preparations for fixing Meltdown on i386. The task was
only halfway done and the current state does not help anybody. For
OpenBSD 6.3 release go back to the original code before 2018/03/13.
This gives us a stable release and the changes will come back later.
discussed with guenther@ deraadt@ hshoexer@

Preparation for i386 Meltdown fix:

- provide a cpu_softc for cpu_attach() etc.
- replace per PCB TSS with per CPU TSS

The first change prepares for cpu_info being embedded in a
cpu_full_info. Therefore during autoconf/cpu_attach we hand down
a softc.

The second change removes the per PCB TSS. We now have one TSS per
CPU, thus in cpu_switchto() we only have to patch the ring 0 stack
pointer instead of loading a new TSS. This also allows for cleaning
up the GDT, so we only have a single slot for the TSS.

from hshoexer@; OK deraadt@

If you use sys/param.h, you don't need sys/types.h

sizes for free(); ok semarie

remove uneeded proc.h includes
ok mpi@ kspillner@

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

remove proc.h include from uvm_map.h. This has far reaching effects, as
sysctl.h was reliant on this particular include, and many drivers included
sysctl.h unnecessarily. remove sysctl.h or add proc.h as needed.
ok deraadt

Based upon a diff orignally submitted by devin smith, Do two things run the
check that the maxfid and the highest reported acpi p_state frequency agree only
after actually getting the acpi p_state. Secondly prefer the legacy powernow
table over ACPI because machines of this vintage of are likely to have broken
ACPI implementations.

Tested by devin smith and tilo stritzky.
ok beck@

Adapt powernow-k7.c driver to use the ACPI _PSS table for retreiving
p_state data based on similar work done to powernow-k8.c. This work was
done by Devin Smith <devin at devinsmith dot net> and this commit is a
slightly modified verson of his last diff to tech@, thanks again for your
work.

Choose the state for a request performance level based on calculations
involving the number of states and not the frequencies of the states,
which can lead to strange distributions of the states over the hw.setperf
range (0-100). Tested by many.

ok canacar, tedu

"#ifdef is a tool of the weak!"
Rename pentium_mhz to cpuspeed which is consistant with amd64 making
shared ACPI code less nasty.
ok marco, deraadt

Complete gwk's previous patch to stop setperf methods from returning
errors to userland: make all cpu_setperf functions return void.

Tested by many, ok gwk@

Fix the spellings of Cool'n'Quiet and PowerNow! in a couple of places.

ok deraadt@

s/Mhz/MHz/ in comments and printf() strings

ok jsg@

If we don't find a matching CPU signature using the base CPUID call, try
matching against the extended CPUID (0x80000001) signature. Problem
found by, and patch based on work by jason@.
ok jason@

Martin V�giard <deadbug AT gmail.com> has been kind enough to rescind the
third clause of his license and strike the phrase "All rights reserved", so we
can consolidate both license blocks into one. Thanks!

From k*_powernow_setperf() always return 0, i.e.
Do not return low level errors to userspace where they will not/can not be
coped with.
: ----------------------------------------------------------------------

Change the value returned if the CPU fails to honor a request to
adjust frequency from 1 (EPERM) to EIO.
Suggested by and ok tedu@

Teach powernow about setperf prioritys, add a setperf_prio to amd64.
ok uwe@, "good" tedu@

kill trainling spaces

* Plug memory leak if init fails; ok gwk@

Correct the heuristic to determin Cool`n'Quiet vs. Powernow! spotted in
the linux driver.
Correct the the cpu signature used to determin if the chip is affected
by errata a0, verified in the linux driver.
Some demagification and KNF.
ok dlg@

Ensure that we check if the transition succeeded before we adjust
pentium_mhz on K7, and print a usefull diagnostic message if it does
not on K7 and K8.
"makes sense to me" dlg@

Don't bother setting the frequency entry in the available states table
to -1 we never check for it. (As has been the case on K8 for a long time).
ok dlg@

Eliminate some pointers we dont use, originally for multiprocessor support,
K7 MP parts do not support Cool'N'Quiet.
"also makes sense" dlg@

Rid k7_powernow_setperf() of some bad math, this code expected the values
to be expressed in hertz not megahertz.
"makes sense to me" dlg@

Change the return value in k7_powernow_setperf() from 1 to 0 so that it
does not print "operation not permitted" even though the operation was
successful.
ok dlg@

Loop on the number of PST reported by the bios not a magic number.
Fixes PR 5075.
"go for it" dlg@

Small hex-vs-decimal typo from gwk, and tested by me (with cpusig check
disabled because my BIOS sucks).

a bit of demagification. this moves the checks for the cpu scaling features
out of machdep and into powernow-k7 and -k8. machdep now just figures out
if its the right type of cpu before calling the powernow code which
figures out if the scaling is supported.

from gwk

spacing

another dangling semicolon from Matthias Bauer

some cleanup from gordon

lost a diff from gordon, only attach if we can do something useful

malloc(M_WAITOK) is bad in init code, M_DEVBUF is nicer than M_TEMP,
and if the cpu gets stuck, don't spin too long.

update support for powernow (cool and quiet) on k7, and add support
for k8. preliminary and not well tested yet. from freebsd via
gordon klok.

fix addr range in the comments

unmap bios when done. from form@pdp-11.org.ru via grange

names of frequencies aren't interesting, and require ugly snprintf
manipulation theo doesn't like. just print number of states.
this way doesn't leak the memory for the string either.

setperf driver for powernow in amd k7 cpus. derived from a diff to
tech-i386@netbsd by martin vegiard, and adapted to fit in here.
as yet untested. ok deraadt@

If you use sys/param.h, you don't need sys/types.h

sizes for free(); ok semarie

remove uneeded proc.h includes
ok mpi@ kspillner@

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

remove proc.h include from uvm_map.h. This has far reaching effects, as
sysctl.h was reliant on this particular include, and many drivers included
sysctl.h unnecessarily. remove sysctl.h or add proc.h as needed.
ok deraadt

Based upon a diff orignally submitted by devin smith, Do two things run the
check that the maxfid and the highest reported acpi p_state frequency agree only
after actually getting the acpi p_state. Secondly prefer the legacy powernow
table over ACPI because machines of this vintage of are likely to have broken
ACPI implementations.

Tested by devin smith and tilo stritzky.
ok beck@

Adapt powernow-k7.c driver to use the ACPI _PSS table for retreiving
p_state data based on similar work done to powernow-k8.c. This work was
done by Devin Smith <devin at devinsmith dot net> and this commit is a
slightly modified verson of his last diff to tech@, thanks again for your
work.

Choose the state for a request performance level based on calculations
involving the number of states and not the frequencies of the states,
which can lead to strange distributions of the states over the hw.setperf
range (0-100). Tested by many.

ok canacar, tedu

"#ifdef is a tool of the weak!"
Rename pentium_mhz to cpuspeed which is consistant with amd64 making
shared ACPI code less nasty.
ok marco, deraadt

Complete gwk's previous patch to stop setperf methods from returning
errors to userland: make all cpu_setperf functions return void.

Tested by many, ok gwk@

Fix the spellings of Cool'n'Quiet and PowerNow! in a couple of places.

ok deraadt@

s/Mhz/MHz/ in comments and printf() strings

ok jsg@

If we don't find a matching CPU signature using the base CPUID call, try
matching against the extended CPUID (0x80000001) signature. Problem
found by, and patch based on work by jason@.
ok jason@

Martin V�giard <deadbug AT gmail.com> has been kind enough to rescind the
third clause of his license and strike the phrase "All rights reserved", so we
can consolidate both license blocks into one. Thanks!

From k*_powernow_setperf() always return 0, i.e.
Do not return low level errors to userspace where they will not/can not be
coped with.
: ----------------------------------------------------------------------

Change the value returned if the CPU fails to honor a request to
adjust frequency from 1 (EPERM) to EIO.
Suggested by and ok tedu@

Teach powernow about setperf prioritys, add a setperf_prio to amd64.
ok uwe@, "good" tedu@

kill trainling spaces

* Plug memory leak if init fails; ok gwk@

Correct the heuristic to determin Cool`n'Quiet vs. Powernow! spotted in
the linux driver.
Correct the the cpu signature used to determin if the chip is affected
by errata a0, verified in the linux driver.
Some demagification and KNF.
ok dlg@

Ensure that we check if the transition succeeded before we adjust
pentium_mhz on K7, and print a usefull diagnostic message if it does
not on K7 and K8.
"makes sense to me" dlg@

Don't bother setting the frequency entry in the available states table
to -1 we never check for it. (As has been the case on K8 for a long time).
ok dlg@

Eliminate some pointers we dont use, originally for multiprocessor support,
K7 MP parts do not support Cool'N'Quiet.
"also makes sense" dlg@

Rid k7_powernow_setperf() of some bad math, this code expected the values
to be expressed in hertz not megahertz.
"makes sense to me" dlg@

Change the return value in k7_powernow_setperf() from 1 to 0 so that it
does not print "operation not permitted" even though the operation was
successful.
ok dlg@

Loop on the number of PST reported by the bios not a magic number.
Fixes PR 5075.
"go for it" dlg@

Small hex-vs-decimal typo from gwk, and tested by me (with cpusig check
disabled because my BIOS sucks).

a bit of demagification. this moves the checks for the cpu scaling features
out of machdep and into powernow-k7 and -k8. machdep now just figures out
if its the right type of cpu before calling the powernow code which
figures out if the scaling is supported.

from gwk

spacing

another dangling semicolon from Matthias Bauer

some cleanup from gordon

lost a diff from gordon, only attach if we can do something useful

malloc(M_WAITOK) is bad in init code, M_DEVBUF is nicer than M_TEMP,
and if the cpu gets stuck, don't spin too long.

update support for powernow (cool and quiet) on k7, and add support
for k8. preliminary and not well tested yet. from freebsd via
gordon klok.

fix addr range in the comments

unmap bios when done. from form@pdp-11.org.ru via grange

names of frequencies aren't interesting, and require ugly snprintf
manipulation theo doesn't like. just print number of states.
this way doesn't leak the memory for the string either.

setperf driver for powernow in amd k7 cpus. derived from a diff to
tech-i386@netbsd by martin vegiard, and adapted to fit in here.
as yet untested. ok deraadt@