Archive-Team
/
zfs
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 103 Wiki Activity

Linux 5.3 compat: retire rw_tryupgrade() 

The Linux kernel's rwsem's have never provided an interface to
allow a reader to be upgraded to a writer.  Historically, this
functionality has been implemented by a SPL wrapper function.
However, this approach depends on internal knowledge of the
rw_semaphore and is therefore rather brittle.

Since the ZFS code must always be able to fallback to rw_exit()
and rw_enter() when an rw_tryupgrade() fails; this functionality
isn't critical.  Furthermore, the only potentially performance
sensitive consumer is dmu_zfetch() and no decrease in performance
was observed with this change applied.  See the PR comments for
additional testing details.

Therefore, it is being retired to make the build more robust and
to simplify the rwlock implementation.

Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Reviewed-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #9029
This commit is contained in:
Brian Behlendorf 2019-07-12 14:06:36 -07:00
parent 041205afee
commit e7a99dab2b
2 changed files with 7 additions and 154 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

60
include/spl/sys/rwlock.h
View File

@ -29,43 +29,6 @@
#include <linux/rwsem.h> #include <linux/rwsem.h>
#include <linux/sched.h> #include <linux/sched.h>
/* Linux kernel compatibility */
#if defined(CONFIG_PREEMPT_RT_FULL)
#define SPL_RWSEM_SINGLE_READER_VALUE (1)
#define SPL_RWSEM_SINGLE_WRITER_VALUE (0)
#elif defined(CONFIG_RWSEM_GENERIC_SPINLOCK)
#define SPL_RWSEM_SINGLE_READER_VALUE (1)
#define SPL_RWSEM_SINGLE_WRITER_VALUE (-1)
#elif defined(RWSEM_ACTIVE_MASK)
#define SPL_RWSEM_SINGLE_READER_VALUE (RWSEM_ACTIVE_READ_BIAS)
#define SPL_RWSEM_SINGLE_WRITER_VALUE (RWSEM_ACTIVE_WRITE_BIAS)
#endif
/* Linux 3.16 changed activity to count for rwsem-spinlock */
#if defined(CONFIG_PREEMPT_RT_FULL)
#define RWSEM_COUNT(sem) sem->read_depth
#elif defined(HAVE_RWSEM_ACTIVITY)
#define RWSEM_COUNT(sem) sem->activity
/* Linux 4.8 changed count to an atomic_long_t for !rwsem-spinlock */
#elif defined(HAVE_RWSEM_ATOMIC_LONG_COUNT)
#define RWSEM_COUNT(sem) atomic_long_read(&(sem)->count)
#else
#define RWSEM_COUNT(sem) sem->count
#endif
#if defined(RWSEM_SPINLOCK_IS_RAW)
#define spl_rwsem_lock_irqsave(lk, fl) raw_spin_lock_irqsave(lk, fl)
#define spl_rwsem_unlock_irqrestore(lk, fl) \
raw_spin_unlock_irqrestore(lk, fl)
#define spl_rwsem_trylock_irqsave(lk, fl) raw_spin_trylock_irqsave(lk, fl)
#else
#define spl_rwsem_lock_irqsave(lk, fl) spin_lock_irqsave(lk, fl)
#define spl_rwsem_unlock_irqrestore(lk, fl) spin_unlock_irqrestore(lk, fl)
#define spl_rwsem_trylock_irqsave(lk, fl) spin_trylock_irqsave(lk, fl)
#endif /* RWSEM_SPINLOCK_IS_RAW */
#define spl_rwsem_is_locked(rwsem) rwsem_is_locked(rwsem)
typedef enum { typedef enum {
RW_DRIVER = 2, RW_DRIVER = 2,
RW_DEFAULT = 4, RW_DEFAULT = 4,
@ -133,7 +96,7 @@ spl_rw_lockdep_on_maybe(krwlock_t *rwp) \
static inline int static inline int
RW_LOCK_HELD(krwlock_t *rwp) RW_LOCK_HELD(krwlock_t *rwp)
{ {
return (spl_rwsem_is_locked(SEM(rwp))); return (rwsem_is_locked(SEM(rwp)));
} }
static inline int static inline int
@ -170,6 +133,12 @@ RW_READ_HELD(krwlock_t *rwp)
*/ */
#define rw_destroy(rwp) ((void) 0) #define rw_destroy(rwp) ((void) 0)
/*
* Upgrading a rwsem from a reader to a writer is not supported by the
* Linux kernel. The lock must be dropped and reacquired as a writer.
*/
#define rw_tryupgrade(rwp) RW_WRITE_HELD(rwp)
#define rw_tryenter(rwp, rw) \ #define rw_tryenter(rwp, rw) \
({ \ ({ \
int _rc_ = 0; \ int _rc_ = 0; \
@ -228,24 +197,9 @@ RW_READ_HELD(krwlock_t *rwp)
spl_rw_lockdep_on_maybe(rwp); \ spl_rw_lockdep_on_maybe(rwp); \
}) })
#define rw_tryupgrade(rwp) \
({ \
int _rc_ = 0; \
\
if (RW_WRITE_HELD(rwp)) { \
_rc_ = 1; \
} else { \
spl_rw_lockdep_off_maybe(rwp); \
if ((_rc_ = rwsem_tryupgrade(SEM(rwp)))) \
spl_rw_set_owner(rwp); \
spl_rw_lockdep_on_maybe(rwp); \
} \
_rc_; \
})
/* END CSTYLED */ /* END CSTYLED */
int spl_rw_init(void); int spl_rw_init(void);
void spl_rw_fini(void); void spl_rw_fini(void);
int rwsem_tryupgrade(struct rw_semaphore *rwsem);
#endif /* _SPL_RWLOCK_H */ #endif /* _SPL_RWLOCK_H */

101
module/spl/spl-rwlock.c
View File

@ -24,106 +24,5 @@
* Solaris Porting Layer (SPL) Reader/Writer Lock Implementation. * Solaris Porting Layer (SPL) Reader/Writer Lock Implementation.
*/ */
#include <sys/rwlock.h>
#include <linux/module.h>
#if defined(CONFIG_PREEMPT_RT_FULL)
#include <linux/rtmutex.h>
#define RT_MUTEX_OWNER_MASKALL 1UL
static int
__rwsem_tryupgrade(struct rw_semaphore *rwsem)
{
#if defined(READER_BIAS) && defined(WRITER_BIAS)
/*
* After the 4.9.20-rt16 kernel the realtime patch series lifted the
* single reader restriction. While this could be accommodated by
* adding additional compatibility code assume the rwsem can never
* be upgraded. All caller must already cleanly handle this case.
*/
return (0);
#else
ASSERT((struct task_struct *)
((unsigned long)rwsem->lock.owner & ~RT_MUTEX_OWNER_MASKALL) ==
current);
/*
* Prior to 4.9.20-rt16 kernel the realtime patch series, rwsem is
* implemented as a single mutex held by readers and writers alike.
* However, this implementation would prevent a thread from taking
* a read lock twice, as the mutex would already be locked on
* the second attempt. Therefore the implementation allows a
* single thread to take a rwsem as read lock multiple times
* tracking that nesting as read_depth counter.
*/
if (rwsem->read_depth <= 1) {
/*
* In case, the current thread has not taken the lock
* more than once as read lock, we can allow an
* upgrade to a write lock. rwsem_rt.h implements
* write locks as read_depth == 0.
*/
rwsem->read_depth = 0;
return (1);
}
return (0);
#endif
}
#elif defined(CONFIG_RWSEM_GENERIC_SPINLOCK)
static int
__rwsem_tryupgrade(struct rw_semaphore *rwsem)
{
int ret = 0;
unsigned long flags;
spl_rwsem_lock_irqsave(&rwsem->wait_lock, flags);
if (RWSEM_COUNT(rwsem) == SPL_RWSEM_SINGLE_READER_VALUE &&
list_empty(&rwsem->wait_list)) {
ret = 1;
RWSEM_COUNT(rwsem) = SPL_RWSEM_SINGLE_WRITER_VALUE;
}
spl_rwsem_unlock_irqrestore(&rwsem->wait_lock, flags);
return (ret);
}
#elif defined(RWSEM_ACTIVE_MASK)
#if defined(HAVE_RWSEM_ATOMIC_LONG_COUNT)
static int
__rwsem_tryupgrade(struct rw_semaphore *rwsem)
{
long val;
val = atomic_long_cmpxchg(&rwsem->count, SPL_RWSEM_SINGLE_READER_VALUE,
SPL_RWSEM_SINGLE_WRITER_VALUE);
return (val == SPL_RWSEM_SINGLE_READER_VALUE);
}
#else
static int
__rwsem_tryupgrade(struct rw_semaphore *rwsem)
{
typeof(rwsem->count) val;
val = cmpxchg(&rwsem->count, SPL_RWSEM_SINGLE_READER_VALUE,
SPL_RWSEM_SINGLE_WRITER_VALUE);
return (val == SPL_RWSEM_SINGLE_READER_VALUE);
}
#endif
#else
static int
__rwsem_tryupgrade(struct rw_semaphore *rwsem)
{
return (0);
}
#endif
int
rwsem_tryupgrade(struct rw_semaphore *rwsem)
{
if (__rwsem_tryupgrade(rwsem)) {
rwsem_release(&rwsem->dep_map, 1, _RET_IP_);
rwsem_acquire(&rwsem->dep_map, 0, 1, _RET_IP_);
return (1);
}
return (0);
}
EXPORT_SYMBOL(rwsem_tryupgrade);
int spl_rw_init(void) { return 0; } int spl_rw_init(void) { return 0; }
void spl_rw_fini(void) { } void spl_rw_fini(void) { }