PR/55664: Ruslan Nikolaev: Fix:

1. A race condition (bug) in sys/rump/librump/rumpkern/intr.c since
rumpuser_cv_signal() is called without holding a mutex
2. sleepq is implemented using a single (global) conditional
variable; that should be done per each sleepq separately

- Extend the per-CPU counters matt@ did to include all of the hot counters
in UVM, excluding uvmexp.free, which needs special treatment and will be
done with a separate commit. Cuts system time for a build by 20-25% on
a 48 CPU machine w/DIAGNOSTIC.

- Avoid 64-bit integer divide on every fault (for rnd_add_uint32).

branches: 1.54.18;
Put the kernelside rump kernel headers into <rump-sys> instead of
sprinkling them around the faction directories. Avoids having
to add a CPPFLAGS (or several) to pretty much every component
Makefile.

Leave compat headers around in the old locations.

The commit changes some autogenerated files, but I'll fix the
generators shortly and regen.

Don't use KASSERT() to test for external return values, use panic()

from Robert Millan <rmh@freebsd.org>

Apparently not all ports define struct clockframe in cpu.h, so we cannot
provide our definition that way. Instead, generate the struct clockframe
passed to hardclock() in MD code.

Include kern_clock.c in rump kernels.

Use hardclock_ticks instead of a homegrown variable.

... not that I understand how various kernel algorithms can work
after enough uptime with hardclock_ticks being a signed int.

"extern int hz" was written in a very aesthetically pleasing way in this
file, but let's just be happy with the sys/kernel.h style of writing it.

Execute softints in the order in which they are scheduled (per level).

Implement softint_schedule_cpu() for rump kernels.

While distributing processing all over the place is not relevant for
high-performance rump kernel I/O stacks (and downright counterproductive),
the mechanism is used e.g. to reach a quiescent state when detaching
an interface, and therefore a semantically correct implementation is
required.

Fixes at least an uncommon race in the ifconfig destroy case.
reported & patch tested by Justin Cormack.

branches: 1.46.4;
Initialize cpu_softcpu before creating softint threads. Fixes things
with rump kernel hypervisors which wrap the thread creation hypercall.

pointed out by Justin Cormack

Restore the assert in RUMP's softint_schedule_cpu() and just ensure
curcpu() in the caller.

RUMP's softint_schedule_cpu: comment out an assert for now.

librump: add kpreempt_disabled(9) and softint_schedule_cpu(9).

branches: 1.42.2;
Avoid calling curcpu() when we know the result is constant.

Defer softint thread creation to first softint_establish() for that level.
Speeds up rump kernel bootstrap and saves memory -- very rarely are all
softint levels in a rump kernel used.

branches: 1.40.4;
Ok, maybe using int64 for nanoseconds in the (sec,nsec) tuple was
a bit too future-proof. I think long is enough there (let's just
hope nobody redefines "nano"). Also, make seconds signed just in
case someone wants their clock to be in 1901.

Flip parameter order in the clock hypercalls to make them consistent
with everything else wrt in/out parameters.

Improve the time-related hypercalls so that's it's possible to
sleep until an absolute time on the host's monotonic clock (should
something like that be supported).

roll mutex init hypercalls into one (one of them already took a flag anyway)

branches: 1.36.4; 1.36.14;
Update copyright statements.

no functional change.

branches: 1.35.2;
Track lwp as the rumpuser mutex owner instead of pthread_t (this
is done in rumpuser for simplicity, since on the kernel side things
we assume we have only one pointer of space). As a side-effect,
we can no longer know if the current thread is holding on to a
mutex locked without curlwp context (basically all mutexes inited
outside of mutex_init()). The only thing that called rumpuser_mutex_held()
for a non-kmutex was the giant lock. So, instead implement recursive
locking for the giant lock in the rump kernel and get rid of the
now-unused recursive pthread mutex in the hypercall interface.

Allocate softint vectors for the final number of CPUs, not the
number currently attached. Deals with a SNAFU in my commit earlier
today which would cause softints established early to lack a
softint context on non-bootstrap CPUs.

update comments. no code change.

Implement softints properly: they need to have a schedulable entity
per cpu.

Don't create the percpu clock interrupt threads as softint threads
because they aren't softint threads. This fixes callouts in
situations where there is nothing else happening in the rump kernel
(i.e. no threads executed which would trigger the softints when
they unschedule).

Use correct indices for clock threads and more descriptive ones
for the softint threads.

Add missing include

Make it possible to use the scheduler lock as the rumpuser condvar
interlock. This is applicable in cases where the actual interlock
is the CPU the currently running thread is scheduled on. Borrowing
the scheduler lock as the mutex mandated by pthread_cond_wait()
does away with need to have an additional mutex. This both optimizes
runtime execution and simplifies code, as the extra lock typically
lead to quite some trickeries to avoid the dungeon collapsing due
to zaps from the wand of deadlock.

fix inversion: advance clock on cpu0, not the complement of cpu0

Fix snafu which caused the clock to travel lightspeed.

Fix multiple virtual cpu support.

... or at least on x86. CPU_INFO_FOREACH() still isn't MI, and I
don't want to support 2^n different versions.

Include kern_tc and use a timecounter driver instead of homerolled
kern_tc implementation.

branches: 1.23.2; 1.23.4;
Cast Oh Kath Ra once every second.

Almost there for virtual CPU MP support:
* support bound kernel threads
* bind softint threads to specific virtual cpus
+ remove now-unnecessary locks from softint code

Now, if we only had MI CPU_INFO_FOREACH() .... (hi rmind ;)

Make rumpuser_cv_timedwait take two int64's instead timespec to
uncouple it from the timespec layout. Also, change return value
to zero for "timeout didn't expire" and non-zero for "timeout
expired". This decouples the interface from errno assignments.

Hash out soft interrupts to be a little closer to real softints:
* split them into levels
* allow only one per level to be active at a time
* fire softints only when we are unscheduling from a CPU instead
of immediately in softint_schedule(). this will later morph
into return from interrupt, but that part isn't done yet.

lie that we are never in a softintr (i.e. add stub)

arrr, implement softint_disestablish(). this code be needin' an enema, matey.

fix uninitialized

Fix getnano/microuptime to report actual uptime.

decouple rumpuser gettime from struct timeval

branches: 1.14.2;
Make the clock interrupt tick based on absolute time instead of
relative time. This prevents drifting. Also, keep track of time
within userspace, so we do not have to make a syscall to get the
clock value. This is approximately 7 times cheaper, but on the
negative side is limited to the clock interrupt frequency.

Create the clock interrupt and soft interrupt workers as MPSAFE

misc cleanup, mainly header polish

__KERNEL_RCSID

Separate the timer thread and soft interrut workers. Use a dynamic
size pool of workers to make sure there are enough for processing
the scheduled soft interrupts.

fix compilation and lint warnings.

branches: 1.8.2;
Provide time_uptime and push it forward every now and then if we happen
to have the timer thread. rump uptime - woohoo.

fix brainfart: kthread_create() sets curlwp

Support callouts and call callout_hardclock() from the timer
interrupt thread.

The sleepq implementation required for callouts is horrible, kludged
only for callouts, and generally unacceptable. It needs revisiting,
but I'm not sure yet should rump or kern_timeout be improved. It's
almost untested as of now, but committing this will give me some
maneuvering space while letting application compile.

Rewrite interrupts to provide better softintr support, i.e. not
execute them simply in the context of the scheduling code, as this
does not work for all applications (e.g. networking soft interrupts).
Rather, schedule them and execute them from a separate context.

Also provide a timer which for now executes just hardclock_callout()
(well, at least after callouts are included in librump, which is
soon).

Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

branches: 1.3.2; 1.3.4; 1.3.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.2; 1.2.4; 1.2.6; 1.2.8; 1.2.14; 1.2.16; 1.2.18;
Merge vmlocking2 to head.

branches: 1.1.2;
file intr.c was initially added on branch vmlocking2.

- Extend the per-CPU counters matt@ did to include all of the hot counters
in UVM, excluding uvmexp.free, which needs special treatment and will be
done with a separate commit. Cuts system time for a build by 20-25% on
a 48 CPU machine w/DIAGNOSTIC.

- Avoid 64-bit integer divide on every fault (for rnd_add_uint32).

Put the kernelside rump kernel headers into <rump-sys> instead of
sprinkling them around the faction directories. Avoids having
to add a CPPFLAGS (or several) to pretty much every component
Makefile.

Leave compat headers around in the old locations.

The commit changes some autogenerated files, but I'll fix the
generators shortly and regen.

Don't use KASSERT() to test for external return values, use panic()

from Robert Millan <rmh@freebsd.org>

Apparently not all ports define struct clockframe in cpu.h, so we cannot
provide our definition that way. Instead, generate the struct clockframe
passed to hardclock() in MD code.

Include kern_clock.c in rump kernels.

Use hardclock_ticks instead of a homegrown variable.

... not that I understand how various kernel algorithms can work
after enough uptime with hardclock_ticks being a signed int.

"extern int hz" was written in a very aesthetically pleasing way in this
file, but let's just be happy with the sys/kernel.h style of writing it.

Execute softints in the order in which they are scheduled (per level).

Implement softint_schedule_cpu() for rump kernels.

While distributing processing all over the place is not relevant for
high-performance rump kernel I/O stacks (and downright counterproductive),
the mechanism is used e.g. to reach a quiescent state when detaching
an interface, and therefore a semantically correct implementation is
required.

Fixes at least an uncommon race in the ifconfig destroy case.
reported & patch tested by Justin Cormack.

branches: 1.46.4;
Initialize cpu_softcpu before creating softint threads. Fixes things
with rump kernel hypervisors which wrap the thread creation hypercall.

pointed out by Justin Cormack

Restore the assert in RUMP's softint_schedule_cpu() and just ensure
curcpu() in the caller.

RUMP's softint_schedule_cpu: comment out an assert for now.

librump: add kpreempt_disabled(9) and softint_schedule_cpu(9).

branches: 1.42.2;
Avoid calling curcpu() when we know the result is constant.

Defer softint thread creation to first softint_establish() for that level.
Speeds up rump kernel bootstrap and saves memory -- very rarely are all
softint levels in a rump kernel used.

branches: 1.40.4;
Ok, maybe using int64 for nanoseconds in the (sec,nsec) tuple was
a bit too future-proof. I think long is enough there (let's just
hope nobody redefines "nano"). Also, make seconds signed just in
case someone wants their clock to be in 1901.

Flip parameter order in the clock hypercalls to make them consistent
with everything else wrt in/out parameters.

Improve the time-related hypercalls so that's it's possible to
sleep until an absolute time on the host's monotonic clock (should
something like that be supported).

roll mutex init hypercalls into one (one of them already took a flag anyway)

branches: 1.36.4; 1.36.14;
Update copyright statements.

no functional change.

branches: 1.35.2;
Track lwp as the rumpuser mutex owner instead of pthread_t (this
is done in rumpuser for simplicity, since on the kernel side things
we assume we have only one pointer of space). As a side-effect,
we can no longer know if the current thread is holding on to a
mutex locked without curlwp context (basically all mutexes inited
outside of mutex_init()). The only thing that called rumpuser_mutex_held()
for a non-kmutex was the giant lock. So, instead implement recursive
locking for the giant lock in the rump kernel and get rid of the
now-unused recursive pthread mutex in the hypercall interface.

Allocate softint vectors for the final number of CPUs, not the
number currently attached. Deals with a SNAFU in my commit earlier
today which would cause softints established early to lack a
softint context on non-bootstrap CPUs.

update comments. no code change.

Implement softints properly: they need to have a schedulable entity
per cpu.

Don't create the percpu clock interrupt threads as softint threads
because they aren't softint threads. This fixes callouts in
situations where there is nothing else happening in the rump kernel
(i.e. no threads executed which would trigger the softints when
they unschedule).

Use correct indices for clock threads and more descriptive ones
for the softint threads.

Add missing include

Make it possible to use the scheduler lock as the rumpuser condvar
interlock. This is applicable in cases where the actual interlock
is the CPU the currently running thread is scheduled on. Borrowing
the scheduler lock as the mutex mandated by pthread_cond_wait()
does away with need to have an additional mutex. This both optimizes
runtime execution and simplifies code, as the extra lock typically
lead to quite some trickeries to avoid the dungeon collapsing due
to zaps from the wand of deadlock.

fix inversion: advance clock on cpu0, not the complement of cpu0

Fix snafu which caused the clock to travel lightspeed.

Fix multiple virtual cpu support.

... or at least on x86. CPU_INFO_FOREACH() still isn't MI, and I
don't want to support 2^n different versions.

Include kern_tc and use a timecounter driver instead of homerolled
kern_tc implementation.

branches: 1.23.2; 1.23.4;
Cast Oh Kath Ra once every second.

Almost there for virtual CPU MP support:
* support bound kernel threads
* bind softint threads to specific virtual cpus
+ remove now-unnecessary locks from softint code

Now, if we only had MI CPU_INFO_FOREACH() .... (hi rmind ;)

Make rumpuser_cv_timedwait take two int64's instead timespec to
uncouple it from the timespec layout. Also, change return value
to zero for "timeout didn't expire" and non-zero for "timeout
expired". This decouples the interface from errno assignments.

Hash out soft interrupts to be a little closer to real softints:
* split them into levels
* allow only one per level to be active at a time
* fire softints only when we are unscheduling from a CPU instead
of immediately in softint_schedule(). this will later morph
into return from interrupt, but that part isn't done yet.

lie that we are never in a softintr (i.e. add stub)

arrr, implement softint_disestablish(). this code be needin' an enema, matey.

fix uninitialized

Fix getnano/microuptime to report actual uptime.

decouple rumpuser gettime from struct timeval

branches: 1.14.2;
Make the clock interrupt tick based on absolute time instead of
relative time. This prevents drifting. Also, keep track of time
within userspace, so we do not have to make a syscall to get the
clock value. This is approximately 7 times cheaper, but on the
negative side is limited to the clock interrupt frequency.

Create the clock interrupt and soft interrupt workers as MPSAFE

misc cleanup, mainly header polish

__KERNEL_RCSID

Separate the timer thread and soft interrut workers. Use a dynamic
size pool of workers to make sure there are enough for processing
the scheduled soft interrupts.

fix compilation and lint warnings.

branches: 1.8.2;
Provide time_uptime and push it forward every now and then if we happen
to have the timer thread. rump uptime - woohoo.

fix brainfart: kthread_create() sets curlwp

Support callouts and call callout_hardclock() from the timer
interrupt thread.

The sleepq implementation required for callouts is horrible, kludged
only for callouts, and generally unacceptable. It needs revisiting,
but I'm not sure yet should rump or kern_timeout be improved. It's
almost untested as of now, but committing this will give me some
maneuvering space while letting application compile.

Rewrite interrupts to provide better softintr support, i.e. not
execute them simply in the context of the scheduling code, as this
does not work for all applications (e.g. networking soft interrupts).
Rather, schedule them and execute them from a separate context.

Also provide a timer which for now executes just hardclock_callout()
(well, at least after callouts are included in librump, which is
soon).

Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

branches: 1.3.2; 1.3.4; 1.3.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.2; 1.2.4; 1.2.6; 1.2.8; 1.2.14; 1.2.16; 1.2.18;
Merge vmlocking2 to head.

branches: 1.1.2;
file intr.c was initially added on branch vmlocking2.

Put the kernelside rump kernel headers into <rump-sys> instead of
sprinkling them around the faction directories. Avoids having
to add a CPPFLAGS (or several) to pretty much every component
Makefile.

Leave compat headers around in the old locations.

The commit changes some autogenerated files, but I'll fix the
generators shortly and regen.

Don't use KASSERT() to test for external return values, use panic()

from Robert Millan <rmh@freebsd.org>

Apparently not all ports define struct clockframe in cpu.h, so we cannot
provide our definition that way. Instead, generate the struct clockframe
passed to hardclock() in MD code.

Include kern_clock.c in rump kernels.

Use hardclock_ticks instead of a homegrown variable.

... not that I understand how various kernel algorithms can work
after enough uptime with hardclock_ticks being a signed int.

"extern int hz" was written in a very aesthetically pleasing way in this
file, but let's just be happy with the sys/kernel.h style of writing it.

Execute softints in the order in which they are scheduled (per level).

Implement softint_schedule_cpu() for rump kernels.

While distributing processing all over the place is not relevant for
high-performance rump kernel I/O stacks (and downright counterproductive),
the mechanism is used e.g. to reach a quiescent state when detaching
an interface, and therefore a semantically correct implementation is
required.

Fixes at least an uncommon race in the ifconfig destroy case.
reported & patch tested by Justin Cormack.

branches: 1.46.4;
Initialize cpu_softcpu before creating softint threads. Fixes things
with rump kernel hypervisors which wrap the thread creation hypercall.

pointed out by Justin Cormack

Restore the assert in RUMP's softint_schedule_cpu() and just ensure
curcpu() in the caller.

RUMP's softint_schedule_cpu: comment out an assert for now.

librump: add kpreempt_disabled(9) and softint_schedule_cpu(9).

branches: 1.42.2;
Avoid calling curcpu() when we know the result is constant.

Defer softint thread creation to first softint_establish() for that level.
Speeds up rump kernel bootstrap and saves memory -- very rarely are all
softint levels in a rump kernel used.

branches: 1.40.4;
Ok, maybe using int64 for nanoseconds in the (sec,nsec) tuple was
a bit too future-proof. I think long is enough there (let's just
hope nobody redefines "nano"). Also, make seconds signed just in
case someone wants their clock to be in 1901.

Flip parameter order in the clock hypercalls to make them consistent
with everything else wrt in/out parameters.

Improve the time-related hypercalls so that's it's possible to
sleep until an absolute time on the host's monotonic clock (should
something like that be supported).

roll mutex init hypercalls into one (one of them already took a flag anyway)

branches: 1.36.4; 1.36.14;
Update copyright statements.

no functional change.

branches: 1.35.2;
Track lwp as the rumpuser mutex owner instead of pthread_t (this
is done in rumpuser for simplicity, since on the kernel side things
we assume we have only one pointer of space). As a side-effect,
we can no longer know if the current thread is holding on to a
mutex locked without curlwp context (basically all mutexes inited
outside of mutex_init()). The only thing that called rumpuser_mutex_held()
for a non-kmutex was the giant lock. So, instead implement recursive
locking for the giant lock in the rump kernel and get rid of the
now-unused recursive pthread mutex in the hypercall interface.

Allocate softint vectors for the final number of CPUs, not the
number currently attached. Deals with a SNAFU in my commit earlier
today which would cause softints established early to lack a
softint context on non-bootstrap CPUs.

update comments. no code change.

Implement softints properly: they need to have a schedulable entity
per cpu.

Don't create the percpu clock interrupt threads as softint threads
because they aren't softint threads. This fixes callouts in
situations where there is nothing else happening in the rump kernel
(i.e. no threads executed which would trigger the softints when
they unschedule).

Use correct indices for clock threads and more descriptive ones
for the softint threads.

Add missing include

Make it possible to use the scheduler lock as the rumpuser condvar
interlock. This is applicable in cases where the actual interlock
is the CPU the currently running thread is scheduled on. Borrowing
the scheduler lock as the mutex mandated by pthread_cond_wait()
does away with need to have an additional mutex. This both optimizes
runtime execution and simplifies code, as the extra lock typically
lead to quite some trickeries to avoid the dungeon collapsing due
to zaps from the wand of deadlock.

fix inversion: advance clock on cpu0, not the complement of cpu0

Fix snafu which caused the clock to travel lightspeed.

Fix multiple virtual cpu support.

... or at least on x86. CPU_INFO_FOREACH() still isn't MI, and I
don't want to support 2^n different versions.

Include kern_tc and use a timecounter driver instead of homerolled
kern_tc implementation.

branches: 1.23.2; 1.23.4;
Cast Oh Kath Ra once every second.

Almost there for virtual CPU MP support:
* support bound kernel threads
* bind softint threads to specific virtual cpus
+ remove now-unnecessary locks from softint code

Now, if we only had MI CPU_INFO_FOREACH() .... (hi rmind ;)

Make rumpuser_cv_timedwait take two int64's instead timespec to
uncouple it from the timespec layout. Also, change return value
to zero for "timeout didn't expire" and non-zero for "timeout
expired". This decouples the interface from errno assignments.

Hash out soft interrupts to be a little closer to real softints:
* split them into levels
* allow only one per level to be active at a time
* fire softints only when we are unscheduling from a CPU instead
of immediately in softint_schedule(). this will later morph
into return from interrupt, but that part isn't done yet.

lie that we are never in a softintr (i.e. add stub)

arrr, implement softint_disestablish(). this code be needin' an enema, matey.

fix uninitialized

Fix getnano/microuptime to report actual uptime.

decouple rumpuser gettime from struct timeval

branches: 1.14.2;
Make the clock interrupt tick based on absolute time instead of
relative time. This prevents drifting. Also, keep track of time
within userspace, so we do not have to make a syscall to get the
clock value. This is approximately 7 times cheaper, but on the
negative side is limited to the clock interrupt frequency.

Create the clock interrupt and soft interrupt workers as MPSAFE

misc cleanup, mainly header polish

__KERNEL_RCSID

Separate the timer thread and soft interrut workers. Use a dynamic
size pool of workers to make sure there are enough for processing
the scheduled soft interrupts.

fix compilation and lint warnings.

branches: 1.8.2;
Provide time_uptime and push it forward every now and then if we happen
to have the timer thread. rump uptime - woohoo.

fix brainfart: kthread_create() sets curlwp

Support callouts and call callout_hardclock() from the timer
interrupt thread.

The sleepq implementation required for callouts is horrible, kludged
only for callouts, and generally unacceptable. It needs revisiting,
but I'm not sure yet should rump or kern_timeout be improved. It's
almost untested as of now, but committing this will give me some
maneuvering space while letting application compile.

Rewrite interrupts to provide better softintr support, i.e. not
execute them simply in the context of the scheduling code, as this
does not work for all applications (e.g. networking soft interrupts).
Rather, schedule them and execute them from a separate context.

Also provide a timer which for now executes just hardclock_callout()
(well, at least after callouts are included in librump, which is
soon).

Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

branches: 1.3.2; 1.3.4; 1.3.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.2; 1.2.4; 1.2.6; 1.2.8; 1.2.14; 1.2.16; 1.2.18;
Merge vmlocking2 to head.

branches: 1.1.2;
file intr.c was initially added on branch vmlocking2.