Author: mlaier
Date: Wed Sep  1 19:50:03 2010
New Revision: 212112
URL: http://svn.freebsd.org/changeset/base/212112

Log:
  rmlock(9) two additions and one change/fix:
   - add rm_try_rlock().
   - add RM_SLEEPABLE to use sx(9) as the back-end lock in order to sleep while
     holding the write lock.
   - change rm_noreadtoken to a cpu bitmask to indicate which CPUs need to go
     through the lock/unlock in order to synchronize.  As a side effect, this
     also avoids IPI to CPUs without any readers during rm_wlock.
  
  Discussed with:               ups@, rwatson@ on arch@
  Sponsored by:         Isilon Systems, Inc.

Modified:
  head/share/man/man9/Makefile
  head/share/man/man9/locking.9
  head/share/man/man9/rmlock.9
  head/sys/kern/kern_rmlock.c
  head/sys/sys/_rmlock.h
  head/sys/sys/rmlock.h

Modified: head/share/man/man9/Makefile
==============================================================================
--- head/share/man/man9/Makefile        Wed Sep  1 19:49:24 2010        
(r212111)
+++ head/share/man/man9/Makefile        Wed Sep  1 19:50:03 2010        
(r212112)
@@ -986,6 +986,7 @@ MLINKS+=rman.9 rman_activate_resource.9 
 MLINKS+=rmlock.9 rm_destroy.9 \
        rmlock.9 rm_init.9 \
        rmlock.9 rm_rlock.9 \
+       rmlock.9 rm_try_rlock.9 \
        rmlock.9 rm_runlock.9 \
        rmlock.9 RM_SYSINIT.9 \
        rmlock.9 rm_wlock.9 \

Modified: head/share/man/man9/locking.9
==============================================================================
--- head/share/man/man9/locking.9       Wed Sep  1 19:49:24 2010        
(r212111)
+++ head/share/man/man9/locking.9       Wed Sep  1 19:50:03 2010        
(r212112)
@@ -301,7 +301,7 @@ one of the synchronization primitives di
 .It mutex     Ta \&ok Ta \&ok-1 Ta \&no Ta \&ok Ta \&ok Ta \&no-3
 .It sx        Ta \&ok Ta \&ok Ta \&ok-2 Ta \&ok Ta \&ok Ta \&ok-4
 .It rwlock    Ta \&ok Ta \&ok Ta \&no Ta \&ok-2 Ta \&ok Ta \&no-3
-.It rmlock    Ta \&ok Ta \&ok Ta \&no Ta \&ok Ta \&ok-2 Ta \&no
+.It rmlock    Ta \&ok Ta \&ok Ta \&ok-5 Ta \&ok Ta \&ok-2 Ta \&ok-5
 .El
 .Pp
 .Em *1
@@ -326,6 +326,13 @@ Though one can sleep holding an sx lock,
 .Fn sx_sleep
 which will atomically release this primitive when going to sleep and
 reacquire it on wakeup.
+.Pp
+.Em *5
+.Em Read-mostly
+locks can be initialized to support sleeping while holding a write lock.
+See
+.Xr rmlock 9
+for details.
 .Ss Context mode table
 The next table shows what can be used in different contexts.
 At this time this is a rather easy to remember table.

Modified: head/share/man/man9/rmlock.9
==============================================================================
--- head/share/man/man9/rmlock.9        Wed Sep  1 19:49:24 2010        
(r212111)
+++ head/share/man/man9/rmlock.9        Wed Sep  1 19:50:03 2010        
(r212112)
@@ -35,6 +35,7 @@
 .Nm rm_init_flags ,
 .Nm rm_destroy ,
 .Nm rm_rlock ,
+.Nm rm_try_rlock ,
 .Nm rm_wlock ,
 .Nm rm_runlock ,
 .Nm rm_wunlock ,
@@ -53,6 +54,8 @@
 .Fn rm_destroy "struct rmlock *rm"
 .Ft void
 .Fn rm_rlock "struct rmlock *rm"  "struct rm_priotracker* tracker"
+.Ft int
+.Fn rm_try_rlock "struct rmlock *rm"  "struct rm_priotracker* tracker"
 .Ft void
 .Fn rm_wlock "struct rmlock *rm"
 .Ft void
@@ -84,14 +87,16 @@ Although reader/writer locks look very s
 locks, their usage pattern is different.
 Reader/writer locks can be treated as mutexes (see
 .Xr mutex 9 )
-with shared/exclusive semantics.
+with shared/exclusive semantics unless initialized with
+.Dv RM_SLEEPABLE .
 Unlike
 .Xr sx 9 ,
 an
 .Nm
 can be locked while holding a non-spin mutex, and an
 .Nm
-cannot be held while sleeping.
+cannot be held while sleeping, again unless initialized with
+.Dv RM_SLEEPABLE .
 The
 .Nm
 locks have full priority propagation like mutexes.
@@ -135,6 +140,13 @@ to ignore this lock.
 .It Dv RM_RECURSE
 Allow threads to recursively acquire exclusive locks for
 .Fa rm .
+.It Dv RM_SLEEPABLE
+Allow writers to sleep while holding the lock.
+Readers must not sleep while holding the lock and can avoid to sleep on
+taking the lock by using
+.Fn rm_try_rlock
+instead of
+.Fn rm_rlock .
 .El
 .It Fn rm_rlock "struct rmlock *rm" "struct rm_priotracker* tracker"
 Lock
@@ -161,6 +173,16 @@ access on
 .Fa rm .
 This is called
 .Dq "recursing on a lock" .
+.It Fn rm_try_rlock "struct rmlock *rm" "struct rm_priotracker* tracker"
+Try to lock
+.Fa rm
+as a reader.
+.Fn rm_try_rlock
+will return 0 if the lock cannot be acquired immediately;
+otherwise the lock will be acquired and a non-zero value will be returned.
+Note that
+.Fn rm_try_rlock
+may fail even while the lock is not currently held by a writer.
 .It Fn rm_wlock "struct rmlock *rm"
 Lock
 .Fa rm
@@ -217,6 +239,10 @@ The
 .Nm
 implementation is currently not optimized for single processor systems.
 .Pp
+.Fn rm_try_rlock
+can fail transiently even when there is no writer, while another reader
+updates the state on the local CPU.
+.Pp
 The
 .Nm
 implementation uses a single per CPU list shared by all

Modified: head/sys/kern/kern_rmlock.c
==============================================================================
--- head/sys/kern/kern_rmlock.c Wed Sep  1 19:49:24 2010        (r212111)
+++ head/sys/kern/kern_rmlock.c Wed Sep  1 19:50:03 2010        (r212112)
@@ -187,6 +187,8 @@ rm_cleanIPI(void *arg)
        }
 }
 
+CTASSERT((RM_SLEEPABLE & LO_CLASSFLAGS) == RM_SLEEPABLE);
+
 void
 rm_init_flags(struct rmlock *rm, const char *name, int opts)
 {
@@ -197,9 +199,13 @@ rm_init_flags(struct rmlock *rm, const c
                liflags |= LO_WITNESS;
        if (opts & RM_RECURSE)
                liflags |= LO_RECURSABLE;
-       rm->rm_noreadtoken = 1;
+       rm->rm_writecpus = all_cpus;
        LIST_INIT(&rm->rm_activeReaders);
-       mtx_init(&rm->rm_lock, name, "rmlock_mtx", MTX_NOWITNESS);
+       if (opts & RM_SLEEPABLE) {
+               liflags |= RM_SLEEPABLE;
+               sx_init_flags(&rm->rm_lock_sx, "rmlock_sx", SX_RECURSE);
+       } else
+               mtx_init(&rm->rm_lock_mtx, name, "rmlock_mtx", MTX_NOWITNESS);
        lock_init(&rm->lock_object, &lock_class_rm, name, NULL, liflags);
 }
 
@@ -214,7 +220,10 @@ void
 rm_destroy(struct rmlock *rm)
 {
 
-       mtx_destroy(&rm->rm_lock);
+       if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+               sx_destroy(&rm->rm_lock_sx);
+       else
+               mtx_destroy(&rm->rm_lock_mtx);
        lock_destroy(&rm->lock_object);
 }
 
@@ -222,7 +231,10 @@ int
 rm_wowned(struct rmlock *rm)
 {
 
-       return (mtx_owned(&rm->rm_lock));
+       if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+               return (sx_xlocked(&rm->rm_lock_sx));
+       else
+               return (mtx_owned(&rm->rm_lock_mtx));
 }
 
 void
@@ -241,8 +253,8 @@ rm_sysinit_flags(void *arg)
        rm_init_flags(args->ra_rm, args->ra_desc, args->ra_opts);
 }
 
-static void
-_rm_rlock_hard(struct rmlock *rm, struct rm_priotracker *tracker)
+static int
+_rm_rlock_hard(struct rmlock *rm, struct rm_priotracker *tracker, int trylock)
 {
        struct pcpu *pc;
        struct rm_queue *queue;
@@ -252,9 +264,9 @@ _rm_rlock_hard(struct rmlock *rm, struct
        pc = pcpu_find(curcpu);
 
        /* Check if we just need to do a proper critical_exit. */
-       if (0 == rm->rm_noreadtoken) {
+       if (!(pc->pc_cpumask & rm->rm_writecpus)) {
                critical_exit();
-               return;
+               return (1);
        }
 
        /* Remove our tracker from the per-cpu list. */
@@ -265,7 +277,7 @@ _rm_rlock_hard(struct rmlock *rm, struct
                /* Just add back tracker - we hold the lock. */
                rm_tracker_add(pc, tracker);
                critical_exit();
-               return;
+               return (1);
        }
 
        /*
@@ -289,7 +301,7 @@ _rm_rlock_hard(struct rmlock *rm, struct
                                mtx_unlock_spin(&rm_spinlock);
                                rm_tracker_add(pc, tracker);
                                critical_exit();
-                               return;
+                               return (1);
                        }
                }
        }
@@ -297,20 +309,38 @@ _rm_rlock_hard(struct rmlock *rm, struct
        sched_unpin();
        critical_exit();
 
-       mtx_lock(&rm->rm_lock);
-       rm->rm_noreadtoken = 0;
-       critical_enter();
+       if (trylock) {
+               if (rm->lock_object.lo_flags & RM_SLEEPABLE) {
+                       if (!sx_try_xlock(&rm->rm_lock_sx))
+                               return (0);
+               } else {
+                       if (!mtx_trylock(&rm->rm_lock_mtx))
+                               return (0);
+               }
+       } else {
+               if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+                       sx_xlock(&rm->rm_lock_sx);
+               else
+                       mtx_lock(&rm->rm_lock_mtx);
+       }
 
+       critical_enter();
        pc = pcpu_find(curcpu);
+       rm->rm_writecpus &= ~pc->pc_cpumask;
        rm_tracker_add(pc, tracker);
        sched_pin();
        critical_exit();
 
-       mtx_unlock(&rm->rm_lock);
+       if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+               sx_xunlock(&rm->rm_lock_sx);
+       else
+               mtx_unlock(&rm->rm_lock_mtx);
+
+       return (1);
 }
 
-void
-_rm_rlock(struct rmlock *rm, struct rm_priotracker *tracker)
+int
+_rm_rlock(struct rmlock *rm, struct rm_priotracker *tracker, int trylock)
 {
        struct thread *td = curthread;
        struct pcpu *pc;
@@ -337,11 +367,11 @@ _rm_rlock(struct rmlock *rm, struct rm_p
         * Fast path to combine two common conditions into a single
         * conditional jump.
         */
-       if (0 == (td->td_owepreempt | rm->rm_noreadtoken))
-               return;
+       if (0 == (td->td_owepreempt | (rm->rm_writecpus & pc->pc_cpumask)))
+               return (1);
 
        /* We do not have a read token and need to acquire one. */
-       _rm_rlock_hard(rm, tracker);
+       return _rm_rlock_hard(rm, tracker, trylock);
 }
 
 static void
@@ -400,20 +430,26 @@ _rm_wlock(struct rmlock *rm)
 {
        struct rm_priotracker *prio;
        struct turnstile *ts;
+       cpumask_t readcpus;
 
-       mtx_lock(&rm->rm_lock);
+       if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+               sx_xlock(&rm->rm_lock_sx);
+       else
+               mtx_lock(&rm->rm_lock_mtx);
 
-       if (rm->rm_noreadtoken == 0) {
+       if (rm->rm_writecpus != all_cpus) {
                /* Get all read tokens back */
 
-               rm->rm_noreadtoken = 1;
+               readcpus = all_cpus & (all_cpus & ~rm->rm_writecpus);
+               rm->rm_writecpus = all_cpus;
 
                /*
-                * Assumes rm->rm_noreadtoken update is visible on other CPUs
+                * Assumes rm->rm_writecpus update is visible on other CPUs
                 * before rm_cleanIPI is called.
                 */
 #ifdef SMP
-               smp_rendezvous(smp_no_rendevous_barrier,
+               smp_rendezvous_cpus(readcpus,
+                   smp_no_rendevous_barrier,
                    rm_cleanIPI,
                    smp_no_rendevous_barrier,
                    rm);
@@ -439,7 +475,10 @@ void
 _rm_wunlock(struct rmlock *rm)
 {
 
-       mtx_unlock(&rm->rm_lock);
+       if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+               sx_xunlock(&rm->rm_lock_sx);
+       else
+               mtx_unlock(&rm->rm_lock_mtx);
 }
 
 #ifdef LOCK_DEBUG
@@ -454,7 +493,11 @@ void _rm_wlock_debug(struct rmlock *rm, 
 
        LOCK_LOG_LOCK("RMWLOCK", &rm->lock_object, 0, 0, file, line);
 
-       WITNESS_LOCK(&rm->lock_object, LOP_EXCLUSIVE, file, line);
+       if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+               WITNESS_LOCK(&rm->rm_lock_sx.lock_object, LOP_EXCLUSIVE,
+                   file, line);        
+       else
+               WITNESS_LOCK(&rm->lock_object, LOP_EXCLUSIVE, file, line);
 
        curthread->td_locks++;
 
@@ -465,25 +508,35 @@ _rm_wunlock_debug(struct rmlock *rm, con
 {
 
        curthread->td_locks--;
-       WITNESS_UNLOCK(&rm->lock_object, LOP_EXCLUSIVE, file, line);
+       if (rm->lock_object.lo_flags & RM_SLEEPABLE)
+               WITNESS_UNLOCK(&rm->rm_lock_sx.lock_object, LOP_EXCLUSIVE,
+                   file, line);
+       else
+               WITNESS_UNLOCK(&rm->lock_object, LOP_EXCLUSIVE, file, line);
        LOCK_LOG_LOCK("RMWUNLOCK", &rm->lock_object, 0, 0, file, line);
        _rm_wunlock(rm);
 }
 
-void
+int
 _rm_rlock_debug(struct rmlock *rm, struct rm_priotracker *tracker,
-    const char *file, int line)
+    int trylock, const char *file, int line)
 {
-
+       if (!trylock && (rm->lock_object.lo_flags & RM_SLEEPABLE))
+               WITNESS_CHECKORDER(&rm->rm_lock_sx.lock_object, LOP_NEWORDER,
+                   file, line, NULL);
        WITNESS_CHECKORDER(&rm->lock_object, LOP_NEWORDER, file, line, NULL);
 
-       _rm_rlock(rm, tracker);
+       if (_rm_rlock(rm, tracker, trylock)) {
+               LOCK_LOG_LOCK("RMRLOCK", &rm->lock_object, 0, 0, file, line);
 
-       LOCK_LOG_LOCK("RMRLOCK", &rm->lock_object, 0, 0, file, line);
+               WITNESS_LOCK(&rm->lock_object, 0, file, line);
 
-       WITNESS_LOCK(&rm->lock_object, 0, file, line);
+               curthread->td_locks++;
 
-       curthread->td_locks++;
+               return (1);
+       }
+
+       return (0);
 }
 
 void
@@ -517,12 +570,12 @@ _rm_wunlock_debug(struct rmlock *rm, con
        _rm_wunlock(rm);
 }
 
-void
+int
 _rm_rlock_debug(struct rmlock *rm, struct rm_priotracker *tracker,
-    const char *file, int line)
+    int trylock, const char *file, int line)
 {
 
-       _rm_rlock(rm, tracker);
+       return _rm_rlock(rm, tracker, trylock);
 }
 
 void

Modified: head/sys/sys/_rmlock.h
==============================================================================
--- head/sys/sys/_rmlock.h      Wed Sep  1 19:49:24 2010        (r212111)
+++ head/sys/sys/_rmlock.h      Wed Sep  1 19:50:03 2010        (r212112)
@@ -45,11 +45,15 @@ LIST_HEAD(rmpriolist,rm_priotracker);
 
 struct rmlock {
        struct lock_object lock_object; 
-       volatile int    rm_noreadtoken;
+       volatile cpumask_t rm_writecpus;
        LIST_HEAD(,rm_priotracker) rm_activeReaders;
-       struct mtx      rm_lock;
-
+       union {
+               struct mtx _rm_lock_mtx;
+               struct sx _rm_lock_sx;
+       } _rm_lock;
 };
+#define        rm_lock_mtx     _rm_lock._rm_lock_mtx
+#define        rm_lock_sx      _rm_lock._rm_lock_sx
 
 struct rm_priotracker {
        struct rm_queue rmp_cpuQueue; /* Must be first */

Modified: head/sys/sys/rmlock.h
==============================================================================
--- head/sys/sys/rmlock.h       Wed Sep  1 19:49:24 2010        (r212111)
+++ head/sys/sys/rmlock.h       Wed Sep  1 19:50:03 2010        (r212112)
@@ -33,6 +33,7 @@
 #define _SYS_RMLOCK_H_
 
 #include <sys/mutex.h>
+#include <sys/sx.h>
 #include <sys/_lock.h>
 #include <sys/_rmlock.h>
 
@@ -43,6 +44,7 @@
  */
 #define        RM_NOWITNESS    0x00000001
 #define        RM_RECURSE      0x00000002
+#define        RM_SLEEPABLE    0x00000004
 
 void   rm_init(struct rmlock *rm, const char *name);
 void   rm_init_flags(struct rmlock *rm, const char *name, int opts);
@@ -53,14 +55,15 @@ void        rm_sysinit_flags(void *arg);
 
 void   _rm_wlock_debug(struct rmlock *rm, const char *file, int line);
 void   _rm_wunlock_debug(struct rmlock *rm, const char *file, int line);
-void   _rm_rlock_debug(struct rmlock *rm, struct rm_priotracker *tracker,
-           const char *file, int line);
+int    _rm_rlock_debug(struct rmlock *rm, struct rm_priotracker *tracker,
+           int trylock, const char *file, int line);
 void   _rm_runlock_debug(struct rmlock *rm,  struct rm_priotracker *tracker,
            const char *file, int line);
 
 void   _rm_wlock(struct rmlock *rm);
 void   _rm_wunlock(struct rmlock *rm);
-void   _rm_rlock(struct rmlock *rm, struct rm_priotracker *tracker);
+int    _rm_rlock(struct rmlock *rm, struct rm_priotracker *tracker,
+           int trylock);
 void   _rm_runlock(struct rmlock *rm,  struct rm_priotracker *tracker);
 
 /*
@@ -74,14 +77,17 @@ void        _rm_runlock(struct rmlock *rm,  str
 #define        rm_wlock(rm)    _rm_wlock_debug((rm), LOCK_FILE, LOCK_LINE)
 #define        rm_wunlock(rm)  _rm_wunlock_debug((rm), LOCK_FILE, LOCK_LINE)
 #define        rm_rlock(rm,tracker)  \
-    _rm_rlock_debug((rm),(tracker), LOCK_FILE, LOCK_LINE )
+    ((void)_rm_rlock_debug((rm),(tracker), 0, LOCK_FILE, LOCK_LINE ))
+#define        rm_try_rlock(rm,tracker)  \
+    _rm_rlock_debug((rm),(tracker), 1, LOCK_FILE, LOCK_LINE )
 #define        rm_runlock(rm,tracker)  \
     _rm_runlock_debug((rm), (tracker), LOCK_FILE, LOCK_LINE )
 #else
-#define        rm_wlock(rm)            _rm_wlock((rm))
-#define        rm_wunlock(rm)          _rm_wunlock((rm))
-#define        rm_rlock(rm,tracker)    _rm_rlock((rm),(tracker))
-#define        rm_runlock(rm,tracker)  _rm_runlock((rm), (tracker))
+#define        rm_wlock(rm)                    _rm_wlock((rm))
+#define        rm_wunlock(rm)                  _rm_wunlock((rm))
+#define        rm_rlock(rm,tracker)            
((void)_rm_rlock((rm),(tracker), 0))
+#define        rm_try_rlock(rm,tracker)        _rm_rlock((rm),(tracker), 1)
+#define        rm_runlock(rm,tracker)          _rm_runlock((rm), (tracker))
 #endif
 
 struct rm_args {
_______________________________________________
[email protected] mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-head
To unsubscribe, send any mail to "[email protected]"

Reply via email to