This bring huge amount of changes, I'll enumerate only user-visible changes:
- Delegated Administration
Allows regular users to perform ZFS operations, like file system
creation, snapshot creation, etc.
- L2ARC
Level 2 cache for ZFS - allows to use additional disks for cache.
Huge performance improvements mostly for random read of mostly
static content.
- slog
Allow to use additional disks for ZFS Intent Log to speed up
operations like fsync(2).
- vfs.zfs.super_owner
Allows regular users to perform privileged operations on files stored
on ZFS file systems owned by him. Very careful with this one.
- chflags(2)
Not all the flags are supported. This still needs work.
- ZFSBoot
Support to boot off of ZFS pool. Not finished, AFAIK.
Submitted by: dfr
- Snapshot properties
- New failure modes
Before if write requested failed, system paniced. Now one
can select from one of three failure modes:
- panic - panic on write error
- wait - wait for disk to reappear
- continue - serve read requests if possible, block write requests
- Refquota, refreservation properties
Just quota and reservation properties, but don't count space consumed
by children file systems, clones and snapshots.
- Sparse volumes
ZVOLs that don't reserve space in the pool.
- External attributes
Compatible with extattr(2).
- NFSv4-ACLs
Not sure about the status, might not be complete yet.
Submitted by: trasz
- Creation-time properties
- Regression tests for zpool(8) command.
Obtained from: OpenSolaris
unnecessary, the normal process lock and thread lock are enough. The
spin lock is still needed for process and thread exiting to mimic
single sched_lock.
this eliminates some problems of locking, e.g, a thread lock is needed
but can not be used at that time. Only the process lock is needed now
for new field.
thread_unsuspend_one() needs to optionally wakeup the swapper. Since we
hold the thread lock for that entire function, however, we have to push
that requirement up into the caller.
Found by: rwatson
- Speedup the lock orderings lookup modifying the witness graph from a
linked tree to a matrix. A table lookup caches the lock orderings in
order to make a O(1) access for them. Any witness object has an unique
index withing this lookup cache table.
- Reduce the lock contention on w_mtx acquiring it only when the LOR
actually happens and not in a sane case. In order to do this don't totally
flush lock lists (per-CPU spinlocks list and per-thread sleeplocks list)
but check for ll_count anytime we need to have to verify allocations sanity.
- Introduce the function witness_thread_exit() in the witness namespace which
should verify a thread doesn't hold any witness occurrence why exiting.
- Rename the sysctl debug.witness.graphs into debug.witness.fullgraph and
add debug.witness.badstacks which prints out stacks for LOR revealed.
This is implemented using the stack(9) support, which makes WITNESS to be
dependent by the STACK option or by the DDB (including STACK) option.
- Fix style(9) for src/sys/kern/subr_witness.c
The hash table approach has been developed by Ilya Maykov on the behalf of
Isilon Systems which kindly released the patch.
Jeff Roberson, ported the patch to -CURRENT and fixed w_mtx contention, on the
behalf of Nokia.
Submitted by: Ilya Maykov <ivmaykov at gmail dot com> (Isilon Systems), jeff
Sponsored by: Nokia
routine wakes up proc0 so that proc0 can swap the thread back in.
Historically, this has been done by waking up proc0 directly from
setrunnable() itself via a wakeup(). When waking up a sleeping thread
that was swapped out (the usual case when waking proc0 since only sleeping
threads are eligible to be swapped out), this resulted in a bit of
recursion (e.g. wakeup() -> setrunnable() -> wakeup()).
With sleep queues having separate locks in 6.x and later, this caused a
spin lock LOR (sleepq lock -> sched_lock/thread lock -> sleepq lock).
An attempt was made to fix this in 7.0 by making the proc0 wakeup use
the ithread mechanism for doing the wakeup. However, this required
grabbing proc0's thread lock to perform the wakeup. If proc0 was asleep
elsewhere in the kernel (e.g. waiting for disk I/O), then this degenerated
into the same LOR since the thread lock would be some other sleepq lock.
Fix this by deferring the wakeup of the swapper until after the sleepq
lock held by the upper layer has been locked. The setrunnable() routine
now returns a boolean value to indicate whether or not proc0 needs to be
woken up. The end result is that consumers of the sleepq API such as
*sleep/wakeup, condition variables, sx locks, and lockmgr, have to wakeup
proc0 if they get a non-zero return value from sleepq_abort(),
sleepq_broadcast(), or sleepq_signal().
Discussed with: jeff
Glanced at by: sam
Tested by: Jurgen Weber jurgen - ish com au
MFC after: 2 weeks
variables and sysctl nodes.
- In reset walk the children of kern_sched_stats and reset the counters
via the oid_arg1 pointer. This allows us to add arbitrary counters to
the tree and still reset them properly.
- Define a set of switch types to be passed with flags to mi_switch().
These types are named SWT_*. These types correspond to SCHED_STATS
counters and are automatically handled in this way.
- Make the new SWT_ types more specific than the older switch stats.
There are now stats for idle switches, remote idle wakeups, remote
preemption ithreads idling, etc.
- Add switch statistics for ULE's pickcpu algorithm. These stats include
how much migration there is, how often affinity was successful, how
often threads were migrated to the local cpu on wakeup, etc.
Sponsored by: Nokia
to enter thread_suspend_check().
- Set TDF_ASTPENDING along with TDF_NEEDSUSPCHK so we can move the
thread_suspend_check() to ast() rather than userret().
- Check TDF_NEEDSUSPCHK in the sleepq_catch_signals() optimization so
that we don't miss a suspend request. If this is set use the
expensive signal path.
- Set NEEDSUSPCHK when creating a new thread in thr in case the
creating thread is due to be suspended as well but has not yet.
Reviewed by: davidxu (Authored original patch)
thread_fini(). The schedulers initialize themselves properly during
sched_fork_thread() anyhow. fini is only called when we're returning
the memory to the allocator which surely doesn't care what state the
memory is in.
requiring the per-process spinlock to only requiring the process lock.
- Reflect these changes in the proc.h documentation and consumers throughout
the kernel. This is a substantial reduction in locking cost for these
fields and was made possible by recent changes to threading support.
While the KSE project was quite successful in bringing threading to
FreeBSD, the M:N approach taken by the kse library was never developed
to its full potential. Backwards compatibility will be provided via
libmap.conf for dynamically linked binaries and static binaries will
be broken.
sched_sleep(). This removes extra thread_lock() acquisition and
allows the scheduler to decide what to do with the static boost.
- Change the priority arguments to cv_* to match sleepq/msleep/etc.
where 0 means no priority change. Catch -1 in cv_broadcastpri() and
convert it to 0 for now.
- Set a flag when sleeping in a way that is compatible with swapping
since direct priority comparisons are meaningless now.
- Add a sysctl to ule, kern.sched.static_boost, that defaults to on which
controls the boost behavior. Turning it off gives better performance
in some workloads but needs more investigation.
- While we're modifying sleepq, change signal and broadcast to both
return with the lock held as the lock was held on enter.
Reviewed by: jhb, peter
and assignment.
- Add a reference to a struct cpuset in each thread that is inherited from
the thread that created it.
- Release the reference when the thread is destroyed.
- Add prototypes for syscalls and macros for manipulating cpusets in
sys/cpuset.h
- Add syscalls to create, get, and set new numbered cpusets:
cpuset(), cpuset_{get,set}id()
- Add syscalls for getting and setting affinity masks for cpusets or
individual threads: cpuid_{get,set}affinity()
- Add types for the 'level' and 'which' parameters for the cpuset. This
will permit expansion of the api to cover cpu masks for other objects
identifiable with an id_t integer. For example, IRQs and Jails may be
coming soon.
- The root set 0 contains all valid cpus. All thread initially belong to
cpuset 1. This permits migrating all threads off of certain cpus to
reserve them for special applications.
Sponsored by: Nokia
Discussed with: arch, rwatson, brooks, davidxu, deischen
Reviewed by: antoine
details from consumers.
- Track individual selecters on a per-descriptor basis such that there
are no longer collisions and after sleeping for events only those
descriptors which triggered events must be rescaned.
- Protect the selinfo (per descriptor) structure with a mtx pool mutex.
mtx pool mutexes were chosen to preserve api compatibility with
existing code which does nothing but bzero() to setup selinfo
structures.
- Use a per-thread wait channel rather than a global wait channel.
- Hide select implementation details in a seltd structure which is
opaque to the rest of the kernel.
- Provide a 'selsocket' interface for those kernel consumers who wish to
select on a socket when they have no fd so they no longer have to
be aware of select implementation details.
Tested by: kris
Reviewed on: arch
the ABI when enabled. There is no longer an embedded lock_profile_object
in each lock. Instead a list of lock_profile_objects is kept per-thread
for each lock it may own. The cnt_hold statistic is now always 0 to
facilitate this.
- Support shared locking by tracking individual lock instances and
statistics in the per-thread per-instance lock_profile_object.
- Make the lock profiling hash table a per-cpu singly linked list with a
per-cpu static lock_prof allocator. This removes the need for an array
of spinlocks and reduces cache contention between cores.
- Use a seperate hash for spinlocks and other locks so that only a
critical_enter() is required and not a spinlock_enter() to modify the
per-cpu tables.
- Count time spent spinning in the lock statistics.
- Remove the LOCK_PROFILE_SHARED option as it is always supported now.
- Specifically drop and release the scheduler locks in both schedulers
since we track owners now.
In collaboration with: Kip Macy
Sponsored by: Nokia
process_fini, thread_ctor, thread_dtor, thread_init, thread_fini. This
will allow us to extend dynamically areas in proc/thread for dtrace ;-)
Reviewed by: rwatson
communicate that it relates to (is called by) thread_alloc()
o Add cpu_thread_free() which is called from thread_free()
to counter-act cpu_thread_alloc().
i386: Have cpu_thread_free() call cpu_thread_clean() to
preserve behaviour.
ia64: Have cpu_thread_free() call mtx_destroy() for the
mutex initialized in cpu_thread_alloc().
PR: ia64/118024
silent NULL pointer dereference in the i386 and sparc64 pmap_pinit()
when the kmem_alloc_nofault() failed to allocate address space. Both
functions now return error instead of panicing or dereferencing NULL.
As consequence, vmspace_exec() and vmspace_unshare() returns the errno
int. struct vmspace arg was added to vm_forkproc() to avoid dealing
with failed allocation when most of the fork1() job is already done.
The kernel stack for the thread is now set up in the thread_alloc(),
that itself may return NULL. Also, allocation of the first process
thread is performed in the fork1() to properly deal with stack
allocation failure. proc_linkup() is separated into proc_linkup()
called from fork1(), and proc_linkup0(), that is used to set up the
kernel process (was known as swapper).
In collaboration with: Peter Holm
Reviewed by: jhb
kthread_add() takes the same parameters as the old kthread_create()
plus a pointer to a process structure, and adds a kernel thread
to that process.
kproc_kthread_add() takes the parameters for kthread_add,
plus a process name and a pointer to a pointer to a process instead of just
a pointer, and if the proc * is NULL, it creates the process to the
specifications required, before adding the thread to it.
All other old kthread_xxx() calls return, but act on (struct thread *)
instead of (struct proc *). One reason to change the name is so that
any old kernel modules that are lying around and expect kthread_create()
to make a process will not just accidentally link.
fix top to show kernel threads by their thread name in -SH mode
add a tdnam formatting option to ps to show thread names.
make all idle threads actual kthreads and put them into their own idled process.
make all interrupt threads kthreads and put them in an interd process
(mainly for aesthetic and accounting reasons)
rename proc 0 to be 'kernel' and it's swapper thread is now 'swapper'
man page fixes to follow.
In particular:
- smp_tlb_mtx is no longer used, so it is axed.
- smp rendezvous lock isn't really a leaf spin-mutex. Its bad placement in
the table, however, has been the source of a false positive LOR reporting
with the dt_lock. However, smp rendezvous lock would have had sched_lock
there for older lock, so it wasn't still a leaf lock.
- allpmaps is only used in ia32 architecture, so it is inserted in the
appropriate stub.
Addictionally:
- kse_zombie_lock is no longer present, so its definition is axed out.
- zombie_lock doesn't need to have an exported symbol, so just let's it be
declared as static.
Tested by: kris
Approved by: jeff (mentor)
Approved by: re
- p_sflag was mostly protected by PROC_LOCK rather than the PROC_SLOCK or
previously the sched_lock. These bugs have existed for some time.
- Allow swapout to try each thread in a process individually and then
swapin the whole process if any of these fail. This allows us to move
most scheduler related swap flags into td_flags.
- Keep ki_sflag for backwards compat but change all in source tools to
use the new and more correct location of P_INMEM.
Reported by: pho
Reviewed by: attilio, kib
Approved by: re (kensmith)
should call uma_zfree() with various spinlock helds. Rearranging the
code would not help here because we cannot break atomicity respect
prcess spinlock, so the only one choice we have is to defer the operation.
In order to do this use a global queue synchronized through the kse_lock
spinlock which is freed at any thread_alloc() / thread_wait() through a
call to thread_reap().
Note that this approach is not ideal as we should want a per-process
list of zombie upcalls, but it follows initial guidelines of KSE authors.
Tested by: jkim, pav
Approved by: jeff, julian
Approved by: re
dangerous races.
Fix this problems adding correct locking for the members of 'struct
kse_upcall' and other struct proc/struct thread related members.
For the moment, just leave ku_mflag and ku_flags "lazy" locked.
While here, cleanup the code removing the function kse_GC() (unused),
and merging upcall_link(), upcall_unlink(), upcall_stash() in their
respective callers (static functions, very short and only called in one
place).
Reported by: pav
Tested by: pav (on some pointyhat cluster nodes)
Approved by: jeff
Approved by: re
Sponsorized by: NGX Italy (http://www.ngx.it)
- Remove unused kse fields from struct proc.
- Group remaining fields and #ifdef KSE them.
- Move some kern_kse.c only prototypes out of proc and into kern_kse.
Discussed with: Julian
- Add a count of exiting threads, p_exitthreads, to struct proc.
- Increment p_exithreads when we set the deadthread in thread_exit().
- When we thread_stash() a deadthread use an atomic to drop the count.
- Spin until the p_exithreads count reaches 0 in thread_wait().
- Lock the last exiting thread momentarily to be certain that it has
exited cpu_throw().
- Restructure thread_wait(). It does not need a loop as there will only
ever be one thread.
Tested by: moose@opera.com
Reported by: kris, moose@opera.com
- Unsafeness on ruadd() in thread_exit()
- Unatomicity of thread_exiit() in the exit1() operations
This patch addresses these problems allocating p_fd as part of the
process and modifying the way it is accessed.
A small chunk of this patch, resolves a race about p_state in kern_wait(),
since we have to be sure about the zombif-ing process.
Submitted by: jeff
Approved by: jeff (mentor)
- Use thread_lock() rather than sched_lock for per-thread scheduling
sychronization.
- Use the per-process spinlock rather than the sched_lock for per-process
scheduling synchronization.
- Move some common code into thread_suspend_switch() to handle the
mechanics of suspending a thread. The locking here is incredibly
convoluted and should be simplified.
Tested by: kris, current@
Tested on: i386, amd64, ULE, 4BSD, libthr, libkse, PREEMPTION, etc.
Discussed with: kris, attilio, kmacy, jhb, julian, bde (small parts each)
Now, we assume no more sched_lock protection for some of them and use the
distribuited loads method for vmmeter (distribuited through CPUs).
Reviewed by: alc, bde
Approved by: jeff (mentor)
td_ru. This removes the requirement for per-process synchronization in
statclock() and mi_switch(). This was previously supported by
sched_lock which is going away. All modifications to rusage are now
done in the context of the owning thread. reads proceed without locks.
- Aggregate exiting threads rusage in thread_exit() such that the exiting
thread's rusage is not lost.
- Provide a new routine, rufetch() to fetch an aggregate of all rusage
structures from all threads in a process. This routine must be used
in any place requiring a rusage from a process prior to it's exit. The
exited process's rusage is still available via p_ru.
- Aggregate tick statistics only on demand via rufetch() or when a thread
exits. Tick statistics are kept in the thread and protected by sched_lock
until it exits.
Initial patch by: attilio
Reviewed by: attilio, bde (some objections), arch (mostly silent)
Probabilly, a general approach is not the better solution here, so we should
solve the sched_lock protection problems separately.
Requested by: alc
Approved by: jeff (mentor)
vmcnts. This can be used to abstract away pcpu details but also changes
to use atomics for all counters now. This means sched lock is no longer
responsible for protecting counts in the switch routines.
Contributed by: Attilio Rao <attilio@FreeBSD.org>
always 0. Previously we aligned threads on a minimum of 8-byte boundaries.
Note: This changes the uma zone to no longer cache align threads. We
really want the uma zone to do align threads to MAX(16, cache line size)
but there currently isn't a good way to express that to uma.
Submitted by: attilio
GETATTRs being generated - one from lookup()/namei() and the other
from nfs_open() (for cto consistency). This change eliminates the
GETATTR in nfs_open() if an otw GETATTR was done from the namei()
path. Instead of extending the vop interface, we timestamp each attr
load, and use this to detect whether a GETATTR was done from namei()
for this syscall. Introduces a thread-local variable that counts the
syscalls made by the thread and uses <pid, tid, thread syscalls> as
the attrload timestamp. Thanks to jhb@ and peter@ for a discussion on
thread state that could be used as the timestamp with minimal overhead.
Make part of John Birrell's KSE patch permanent..
Specifically, remove:
Any reference of the ksegrp structure. This feature was
never fully utilised and made things overly complicated.
All code in the scheduler that tried to make threaded programs
fair to unthreaded programs. Libpthread processes will already
do this to some extent and libthr processes already disable it.
Also:
Since this makes such a big change to the scheduler(s), take the opportunity
to rename some structures and elements that had to be moved anyhow.
This makes the code a lot more readable.
The ULE scheduler compiles again but I have no idea if it works.
The 4bsd scheduler still reqires a little cleaning and some functions that now do
ALMOST nothing will go away, but I thought I'd do that as a separate commit.
Tested by David Xu, and Dan Eischen using libthr and libpthread.
mutex structure is added as following:
struct umutex {
__lwpid_t m_owner;
uint32_t m_flags;
uint32_t m_ceilings[2];
uint32_t m_spare[4];
};
The m_owner represents owner thread, it is a thread id, in non-contested
case, userland can simply use atomic_cmpset_int to lock the mutex, if the
mutex is contested, high order bit will be set, and userland should do locking
and unlocking via kernel syscall. Flag UMUTEX_PRIO_INHERIT represents
pthread's PTHREAD_PRIO_INHERIT mutex, which when contention happens, kernel
should do priority propagating. Flag UMUTEX_PRIO_PROTECT indicates it is
pthread's PTHREAD_PRIO_PROTECT mutex, userland should initialize m_owner
to contested state UMUTEX_CONTESTED, then atomic_cmpset_int will be failure
and kernel syscall should be invoked to do locking, this becauses
for such a mutex, kernel should always boost the thread's priority before
it can lock the mutex, m_ceilings is used by PTHREAD_PRIO_PROTECT mutex,
the first element is used to boost thread's priority when it locked the mutex,
second element is used when the mutex is unlocked, the PTHREAD_PRIO_PROTECT
mutex's link list is kept in userland, the m_ceiling[1] is managed by thread
library so kernel needn't allocate memory to keep the link list, when such
a mutex is unlocked, kernel reset m_owner to UMUTEX_CONTESTED.
Flag USYNC_PROCESS_SHARED indicate if the synchronization object is process
shared, if the flag is not set, it saves a vm_map_lookup() call.
The umtx chain is still used as a sleep queue, when a thread is blocked on
PTHREAD_PRIO_INHERIT mutex, a umtx_pi is allocated to support priority
propagating, it is dynamically allocated and reference count is used,
it is not optimized but works well in my tests, while the umtx chain has
its own locking protocol, the priority propagating protocol are all protected
by sched_lock because priority propagating function is called with sched_lock
held from scheduler.
No visible performance degradation is found which these changes. Some parameter
names in _umtx_op syscall are renamed.
