2010-05-28 21:09:12 -06:00
|
|
|
/*
|
|
|
|
* Copyright 2010 Tilera Corporation. All Rights Reserved.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License
|
|
|
|
* as published by the Free Software Foundation, version 2.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful, but
|
|
|
|
* WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
|
|
|
|
* NON INFRINGEMENT. See the GNU General Public License for
|
|
|
|
* more details.
|
|
|
|
*
|
|
|
|
* 32-bit SMP spinlocks.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#ifndef _ASM_TILE_SPINLOCK_32_H
|
|
|
|
#define _ASM_TILE_SPINLOCK_32_H
|
|
|
|
|
2011-07-26 17:09:06 -06:00
|
|
|
#include <linux/atomic.h>
|
2010-05-28 21:09:12 -06:00
|
|
|
#include <asm/page.h>
|
|
|
|
#include <linux/compiler.h>
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We only use even ticket numbers so the '1' inserted by a tns is
|
|
|
|
* an unambiguous "ticket is busy" flag.
|
|
|
|
*/
|
|
|
|
#define TICKET_QUANTUM 2
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* SMP ticket spinlocks, allowing only a single CPU anywhere
|
|
|
|
*
|
|
|
|
* (the type definitions are in asm/spinlock_types.h)
|
|
|
|
*/
|
|
|
|
static inline int arch_spin_is_locked(arch_spinlock_t *lock)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Note that even if a new ticket is in the process of being
|
|
|
|
* acquired, so lock->next_ticket is 1, it's still reasonable
|
|
|
|
* to claim the lock is held, since it will be momentarily
|
|
|
|
* if not already. There's no need to wait for a "valid"
|
|
|
|
* lock->next_ticket to become available.
|
2015-04-28 11:00:42 -06:00
|
|
|
* Use READ_ONCE() to ensure that calling this in a loop is OK.
|
2010-05-28 21:09:12 -06:00
|
|
|
*/
|
2015-04-28 11:00:42 -06:00
|
|
|
int curr = READ_ONCE(lock->current_ticket);
|
|
|
|
int next = READ_ONCE(lock->next_ticket);
|
|
|
|
|
|
|
|
return next != curr;
|
2010-05-28 21:09:12 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
void arch_spin_lock(arch_spinlock_t *lock);
|
|
|
|
|
|
|
|
/* We cannot take an interrupt after getting a ticket, so don't enable them. */
|
|
|
|
#define arch_spin_lock_flags(lock, flags) arch_spin_lock(lock)
|
|
|
|
|
|
|
|
int arch_spin_trylock(arch_spinlock_t *lock);
|
|
|
|
|
|
|
|
static inline void arch_spin_unlock(arch_spinlock_t *lock)
|
|
|
|
{
|
|
|
|
/* For efficiency, overlap fetching the old ticket with the wmb(). */
|
|
|
|
int old_ticket = lock->current_ticket;
|
|
|
|
wmb(); /* guarantee anything modified under the lock is visible */
|
|
|
|
lock->current_ticket = old_ticket + TICKET_QUANTUM;
|
|
|
|
}
|
|
|
|
|
|
|
|
void arch_spin_unlock_wait(arch_spinlock_t *lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read-write spinlocks, allowing multiple readers
|
|
|
|
* but only one writer.
|
|
|
|
*
|
|
|
|
* We use a "tns/store-back" technique on a single word to manage
|
|
|
|
* the lock state, looping around to retry if the tns returns 1.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* Internal layout of the word; do not use. */
|
|
|
|
#define _WR_NEXT_SHIFT 8
|
|
|
|
#define _WR_CURR_SHIFT 16
|
|
|
|
#define _WR_WIDTH 8
|
|
|
|
#define _RD_COUNT_SHIFT 24
|
|
|
|
#define _RD_COUNT_WIDTH 8
|
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_read_can_lock() - would read_trylock() succeed?
|
|
|
|
*/
|
|
|
|
static inline int arch_read_can_lock(arch_rwlock_t *rwlock)
|
|
|
|
{
|
|
|
|
return (rwlock->lock << _RD_COUNT_WIDTH) == 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_write_can_lock() - would write_trylock() succeed?
|
|
|
|
*/
|
|
|
|
static inline int arch_write_can_lock(arch_rwlock_t *rwlock)
|
|
|
|
{
|
|
|
|
return rwlock->lock == 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_read_lock() - acquire a read lock.
|
|
|
|
*/
|
arch/tile: fix deadlock bugs in rwlock implementation
The first issue fixed in this patch is that pending rwlock write locks
could lock out new readers; this could cause a deadlock if a read lock was
held on cpu 1, a write lock was then attempted on cpu 2 and was pending,
and cpu 1 was interrupted and attempted to re-acquire a read lock.
The write lock code was modified to not lock out new readers.
The second issue fixed is that there was a narrow race window where a tns
instruction had been issued (setting the lock value to "1") and the store
instruction to reset the lock value correctly had not yet been issued.
In this case, if an interrupt occurred and the same cpu then tried to
manipulate the lock, it would find the lock value set to "1" and spin
forever, assuming some other cpu was partway through updating it. The fix
is to enforce an interrupt critical section around the tns/store pair.
In addition, this change now arranges to always validate that after
a readlock we have not wrapped around the count of readers, which
is only eight bits.
Since these changes make the rwlock "fast path" code heavier weight,
I decided to move all the rwlock code all out of line, leaving only the
conventional spinlock code with fastpath inlines. Since the read_lock
and read_trylock implementations ended up very similar, I just expressed
read_lock in terms of read_trylock.
As part of this change I also eliminate support for the now-obsolete
tns_atomic mode.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
2011-03-01 11:30:15 -07:00
|
|
|
void arch_read_lock(arch_rwlock_t *rwlock);
|
2010-05-28 21:09:12 -06:00
|
|
|
|
|
|
|
/**
|
arch/tile: fix deadlock bugs in rwlock implementation
The first issue fixed in this patch is that pending rwlock write locks
could lock out new readers; this could cause a deadlock if a read lock was
held on cpu 1, a write lock was then attempted on cpu 2 and was pending,
and cpu 1 was interrupted and attempted to re-acquire a read lock.
The write lock code was modified to not lock out new readers.
The second issue fixed is that there was a narrow race window where a tns
instruction had been issued (setting the lock value to "1") and the store
instruction to reset the lock value correctly had not yet been issued.
In this case, if an interrupt occurred and the same cpu then tried to
manipulate the lock, it would find the lock value set to "1" and spin
forever, assuming some other cpu was partway through updating it. The fix
is to enforce an interrupt critical section around the tns/store pair.
In addition, this change now arranges to always validate that after
a readlock we have not wrapped around the count of readers, which
is only eight bits.
Since these changes make the rwlock "fast path" code heavier weight,
I decided to move all the rwlock code all out of line, leaving only the
conventional spinlock code with fastpath inlines. Since the read_lock
and read_trylock implementations ended up very similar, I just expressed
read_lock in terms of read_trylock.
As part of this change I also eliminate support for the now-obsolete
tns_atomic mode.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
2011-03-01 11:30:15 -07:00
|
|
|
* arch_write_lock() - acquire a write lock.
|
2010-05-28 21:09:12 -06:00
|
|
|
*/
|
arch/tile: fix deadlock bugs in rwlock implementation
The first issue fixed in this patch is that pending rwlock write locks
could lock out new readers; this could cause a deadlock if a read lock was
held on cpu 1, a write lock was then attempted on cpu 2 and was pending,
and cpu 1 was interrupted and attempted to re-acquire a read lock.
The write lock code was modified to not lock out new readers.
The second issue fixed is that there was a narrow race window where a tns
instruction had been issued (setting the lock value to "1") and the store
instruction to reset the lock value correctly had not yet been issued.
In this case, if an interrupt occurred and the same cpu then tried to
manipulate the lock, it would find the lock value set to "1" and spin
forever, assuming some other cpu was partway through updating it. The fix
is to enforce an interrupt critical section around the tns/store pair.
In addition, this change now arranges to always validate that after
a readlock we have not wrapped around the count of readers, which
is only eight bits.
Since these changes make the rwlock "fast path" code heavier weight,
I decided to move all the rwlock code all out of line, leaving only the
conventional spinlock code with fastpath inlines. Since the read_lock
and read_trylock implementations ended up very similar, I just expressed
read_lock in terms of read_trylock.
As part of this change I also eliminate support for the now-obsolete
tns_atomic mode.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
2011-03-01 11:30:15 -07:00
|
|
|
void arch_write_lock(arch_rwlock_t *rwlock);
|
2010-05-28 21:09:12 -06:00
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_read_trylock() - try to acquire a read lock.
|
|
|
|
*/
|
arch/tile: fix deadlock bugs in rwlock implementation
The first issue fixed in this patch is that pending rwlock write locks
could lock out new readers; this could cause a deadlock if a read lock was
held on cpu 1, a write lock was then attempted on cpu 2 and was pending,
and cpu 1 was interrupted and attempted to re-acquire a read lock.
The write lock code was modified to not lock out new readers.
The second issue fixed is that there was a narrow race window where a tns
instruction had been issued (setting the lock value to "1") and the store
instruction to reset the lock value correctly had not yet been issued.
In this case, if an interrupt occurred and the same cpu then tried to
manipulate the lock, it would find the lock value set to "1" and spin
forever, assuming some other cpu was partway through updating it. The fix
is to enforce an interrupt critical section around the tns/store pair.
In addition, this change now arranges to always validate that after
a readlock we have not wrapped around the count of readers, which
is only eight bits.
Since these changes make the rwlock "fast path" code heavier weight,
I decided to move all the rwlock code all out of line, leaving only the
conventional spinlock code with fastpath inlines. Since the read_lock
and read_trylock implementations ended up very similar, I just expressed
read_lock in terms of read_trylock.
As part of this change I also eliminate support for the now-obsolete
tns_atomic mode.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
2011-03-01 11:30:15 -07:00
|
|
|
int arch_read_trylock(arch_rwlock_t *rwlock);
|
2010-05-28 21:09:12 -06:00
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_write_trylock() - try to acquire a write lock.
|
|
|
|
*/
|
arch/tile: fix deadlock bugs in rwlock implementation
The first issue fixed in this patch is that pending rwlock write locks
could lock out new readers; this could cause a deadlock if a read lock was
held on cpu 1, a write lock was then attempted on cpu 2 and was pending,
and cpu 1 was interrupted and attempted to re-acquire a read lock.
The write lock code was modified to not lock out new readers.
The second issue fixed is that there was a narrow race window where a tns
instruction had been issued (setting the lock value to "1") and the store
instruction to reset the lock value correctly had not yet been issued.
In this case, if an interrupt occurred and the same cpu then tried to
manipulate the lock, it would find the lock value set to "1" and spin
forever, assuming some other cpu was partway through updating it. The fix
is to enforce an interrupt critical section around the tns/store pair.
In addition, this change now arranges to always validate that after
a readlock we have not wrapped around the count of readers, which
is only eight bits.
Since these changes make the rwlock "fast path" code heavier weight,
I decided to move all the rwlock code all out of line, leaving only the
conventional spinlock code with fastpath inlines. Since the read_lock
and read_trylock implementations ended up very similar, I just expressed
read_lock in terms of read_trylock.
As part of this change I also eliminate support for the now-obsolete
tns_atomic mode.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
2011-03-01 11:30:15 -07:00
|
|
|
int arch_write_trylock(arch_rwlock_t *rwlock);
|
2010-05-28 21:09:12 -06:00
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_read_unlock() - release a read lock.
|
|
|
|
*/
|
arch/tile: fix deadlock bugs in rwlock implementation
The first issue fixed in this patch is that pending rwlock write locks
could lock out new readers; this could cause a deadlock if a read lock was
held on cpu 1, a write lock was then attempted on cpu 2 and was pending,
and cpu 1 was interrupted and attempted to re-acquire a read lock.
The write lock code was modified to not lock out new readers.
The second issue fixed is that there was a narrow race window where a tns
instruction had been issued (setting the lock value to "1") and the store
instruction to reset the lock value correctly had not yet been issued.
In this case, if an interrupt occurred and the same cpu then tried to
manipulate the lock, it would find the lock value set to "1" and spin
forever, assuming some other cpu was partway through updating it. The fix
is to enforce an interrupt critical section around the tns/store pair.
In addition, this change now arranges to always validate that after
a readlock we have not wrapped around the count of readers, which
is only eight bits.
Since these changes make the rwlock "fast path" code heavier weight,
I decided to move all the rwlock code all out of line, leaving only the
conventional spinlock code with fastpath inlines. Since the read_lock
and read_trylock implementations ended up very similar, I just expressed
read_lock in terms of read_trylock.
As part of this change I also eliminate support for the now-obsolete
tns_atomic mode.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
2011-03-01 11:30:15 -07:00
|
|
|
void arch_read_unlock(arch_rwlock_t *rwlock);
|
2010-05-28 21:09:12 -06:00
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_write_unlock() - release a write lock.
|
|
|
|
*/
|
arch/tile: fix deadlock bugs in rwlock implementation
The first issue fixed in this patch is that pending rwlock write locks
could lock out new readers; this could cause a deadlock if a read lock was
held on cpu 1, a write lock was then attempted on cpu 2 and was pending,
and cpu 1 was interrupted and attempted to re-acquire a read lock.
The write lock code was modified to not lock out new readers.
The second issue fixed is that there was a narrow race window where a tns
instruction had been issued (setting the lock value to "1") and the store
instruction to reset the lock value correctly had not yet been issued.
In this case, if an interrupt occurred and the same cpu then tried to
manipulate the lock, it would find the lock value set to "1" and spin
forever, assuming some other cpu was partway through updating it. The fix
is to enforce an interrupt critical section around the tns/store pair.
In addition, this change now arranges to always validate that after
a readlock we have not wrapped around the count of readers, which
is only eight bits.
Since these changes make the rwlock "fast path" code heavier weight,
I decided to move all the rwlock code all out of line, leaving only the
conventional spinlock code with fastpath inlines. Since the read_lock
and read_trylock implementations ended up very similar, I just expressed
read_lock in terms of read_trylock.
As part of this change I also eliminate support for the now-obsolete
tns_atomic mode.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
2011-03-01 11:30:15 -07:00
|
|
|
void arch_write_unlock(arch_rwlock_t *rwlock);
|
2010-05-28 21:09:12 -06:00
|
|
|
|
|
|
|
#define arch_read_lock_flags(lock, flags) arch_read_lock(lock)
|
|
|
|
#define arch_write_lock_flags(lock, flags) arch_write_lock(lock)
|
|
|
|
|
|
|
|
#endif /* _ASM_TILE_SPINLOCK_32_H */
|