1
linux/lib/kernel_lock.c
Linus Torvalds f01eb36403 [BKL] add 'might_sleep()' to the outermost lock taker
As shown by the previous patch (6698e3472: "tty: Fix BKL taken under a
spinlock bug introduced in the BKL split") the BKL removal is prone to
some subtle issues, where removing the BKL in one place may in fact make
a previously nested BKL call the new outer call, and then prone to nasty
deadlocks with other spinlocks.

In general, we should never take the BKL while we're holding a spinlock,
so let's just add a "might_sleep()" to it (even though the BKL doesn't
technically sleep - at least not yet), and we'll get nice warnings the
next time this kind of problem happens during BKL removal.

Acked-and-Tested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-12 14:46:33 -08:00

144 lines
3.2 KiB
C

/*
* lib/kernel_lock.c
*
* This is the traditional BKL - big kernel lock. Largely
* relegated to obsolescence, but used by various less
* important (or lazy) subsystems.
*/
#include <linux/module.h>
#include <linux/kallsyms.h>
#include <linux/semaphore.h>
#include <linux/smp_lock.h>
#define CREATE_TRACE_POINTS
#include <trace/events/bkl.h>
/*
* The 'big kernel lock'
*
* This spinlock is taken and released recursively by lock_kernel()
* and unlock_kernel(). It is transparently dropped and reacquired
* over schedule(). It is used to protect legacy code that hasn't
* been migrated to a proper locking design yet.
*
* Don't use in new code.
*/
static __cacheline_aligned_in_smp DEFINE_SPINLOCK(kernel_flag);
/*
* Acquire/release the underlying lock from the scheduler.
*
* This is called with preemption disabled, and should
* return an error value if it cannot get the lock and
* TIF_NEED_RESCHED gets set.
*
* If it successfully gets the lock, it should increment
* the preemption count like any spinlock does.
*
* (This works on UP too - _raw_spin_trylock will never
* return false in that case)
*/
int __lockfunc __reacquire_kernel_lock(void)
{
while (!_raw_spin_trylock(&kernel_flag)) {
if (need_resched())
return -EAGAIN;
cpu_relax();
}
preempt_disable();
return 0;
}
void __lockfunc __release_kernel_lock(void)
{
_raw_spin_unlock(&kernel_flag);
preempt_enable_no_resched();
}
/*
* These are the BKL spinlocks - we try to be polite about preemption.
* If SMP is not on (ie UP preemption), this all goes away because the
* _raw_spin_trylock() will always succeed.
*/
#ifdef CONFIG_PREEMPT
static inline void __lock_kernel(void)
{
preempt_disable();
if (unlikely(!_raw_spin_trylock(&kernel_flag))) {
/*
* If preemption was disabled even before this
* was called, there's nothing we can be polite
* about - just spin.
*/
if (preempt_count() > 1) {
_raw_spin_lock(&kernel_flag);
return;
}
/*
* Otherwise, let's wait for the kernel lock
* with preemption enabled..
*/
do {
preempt_enable();
while (spin_is_locked(&kernel_flag))
cpu_relax();
preempt_disable();
} while (!_raw_spin_trylock(&kernel_flag));
}
}
#else
/*
* Non-preemption case - just get the spinlock
*/
static inline void __lock_kernel(void)
{
_raw_spin_lock(&kernel_flag);
}
#endif
static inline void __unlock_kernel(void)
{
/*
* the BKL is not covered by lockdep, so we open-code the
* unlocking sequence (and thus avoid the dep-chain ops):
*/
_raw_spin_unlock(&kernel_flag);
preempt_enable();
}
/*
* Getting the big kernel lock.
*
* This cannot happen asynchronously, so we only need to
* worry about other CPU's.
*/
void __lockfunc _lock_kernel(const char *func, const char *file, int line)
{
int depth = current->lock_depth + 1;
trace_lock_kernel(func, file, line);
if (likely(!depth)) {
might_sleep();
__lock_kernel();
}
current->lock_depth = depth;
}
void __lockfunc _unlock_kernel(const char *func, const char *file, int line)
{
BUG_ON(current->lock_depth < 0);
if (likely(--current->lock_depth < 0))
__unlock_kernel();
trace_unlock_kernel(func, file, line);
}
EXPORT_SYMBOL(_lock_kernel);
EXPORT_SYMBOL(_unlock_kernel);