Revision f6b57101a6b31277a4bde1d8028c46e898bd2ff2 authored by Dom Cobley on 21 January 2021, 10:57:59 UTC, committed by Maxime Ripard on 25 January 2021, 10:53:44 UTC
Fixes failure with 4096x1080 resolutions

[  284.315379] WARNING: CPU: 1 PID: 901 at drivers/gpu/drm/vc4/vc4_plane.c:981 vc4_plane_mode_set+0x1374/0x13c4
[  284.315385] Modules linked in: ir_rc5_decoder rpivid_hevc(C) bcm2835_codec(C) bcm2835_isp(C) bcm2835_mmal_vchiq(C) bcm2835_gpiomem v4l2_mem2mem videobuf2_dma_contig videobuf2_memops videobuf2_v4l2 videobuf2_common videodev mc cdc_acm xpad ir_rc6_decoder rc_rc6_mce gpio_ir_recv fuse
[  284.315509] CPU: 1 PID: 901 Comm: kodi.bin Tainted: G         C        5.10.7 #1
[  284.315514] Hardware name: BCM2711
[  284.315518] Backtrace:
[  284.315533] [<c0cc5ca0>] (dump_backtrace) from [<c0cc6014>] (show_stack+0x20/0x24)
[  284.315540]  r7:ffffffff r6:00000000 r5:68000013 r4:c18ecf1c
[  284.315549] [<c0cc5ff4>] (show_stack) from [<c0cca638>] (dump_stack+0xc4/0xf0)
[  284.315558] [<c0cca574>] (dump_stack) from [<c022314c>] (__warn+0xfc/0x158)
[  284.315564]  r9:00000000 r8:00000009 r7:000003d5 r6:00000009 r5:c08cc7dc r4:c0fd09b8
[  284.315572] [<c0223050>] (__warn) from [<c0cc67ec>] (warn_slowpath_fmt+0x74/0xe4)
[  284.315577]  r7:c08cc7dc r6:000003d5 r5:c0fd09b8 r4:00000000
[  284.315584] [<c0cc677c>] (warn_slowpath_fmt) from [<c08cc7dc>] (vc4_plane_mode_set+0x1374/0x13c4)
[  284.315589]  r8:00000000 r7:00000000 r6:00001000 r5:c404c600 r4:c2e34600
[  284.315596] [<c08cb468>] (vc4_plane_mode_set) from [<c08cc984>] (vc4_plane_atomic_check+0x40/0x1c0)
[  284.315601]  r10:00000001 r9:c2e34600 r8:c0e67068 r7:c0fc44e0 r6:c2ce3640 r5:c3d636c0
[  284.315605]  r4:c2e34600
[  284.315614] [<c08cc944>] (vc4_plane_atomic_check) from [<c0860504>] (drm_atomic_helper_check_planes+0xec/0x1ec)
[  284.315620]  r9:c2e34600 r8:c0e67068 r7:c0fc44e0 r6:c2ce3640 r5:c3d636c0 r4:00000006
[  284.315627] [<c0860418>] (drm_atomic_helper_check_planes) from [<c0860658>] (drm_atomic_helper_check+0x54/0x9c)
[  284.315633]  r9:c2e35400 r8:00000006 r7:00000000 r6:c2ba7800 r5:c3d636c0 r4:00000000
[  284.315641] [<c0860604>] (drm_atomic_helper_check) from [<c08b7ca8>] (vc4_atomic_check+0x25c/0x454)
[  284.315645]  r7:00000000 r6:c2ba7800 r5:00000001 r4:c3d636c0
[  284.315652] [<c08b7a4c>] (vc4_atomic_check) from [<c0881278>] (drm_atomic_check_only+0x5cc/0x7e0)
[  284.315658]  r10:c404c6c8 r9:ffffffff r8:c472c480 r7:00000003 r6:c3d636c0 r5:00000000
[  284.315662]  r4:0000003c r3:c08b7a4c
[  284.315670] [<c0880cac>] (drm_atomic_check_only) from [<c089ba60>] (drm_mode_atomic_ioctl+0x758/0xa7c)
[  284.315675]  r10:c3d46000 r9:c3d636c0 r8:c2ce8a70 r7:027e3a54 r6:00000043 r5:c1fbb800
[  284.315679]  r4:0281a858
[  284.315688] [<c089b308>] (drm_mode_atomic_ioctl) from [<c086e9f8>] (drm_ioctl_kernel+0xc4/0x108)
[  284.315693]  r10:c03864bc r9:c1fbb800 r8:c3d47e64 r7:c089b308 r6:00000002 r5:c2ba7800
[  284.315697]  r4:00000000
[  284.315705] [<c086e934>] (drm_ioctl_kernel) from [<c086ee28>] (drm_ioctl+0x1e8/0x3a0)
[  284.315711]  r9:c1fbb800 r8:000000bc r7:c3d47e64 r6:00000038 r5:c0e59570 r4:00000038
[  284.315719] [<c086ec40>] (drm_ioctl) from [<c041f354>] (sys_ioctl+0x35c/0x914)
[  284.315724]  r10:c2d08200 r9:00000000 r8:c36fa300 r7:befdd870 r6:c03864bc r5:c36fa301
[  284.315728]  r4:c03864bc
[  284.315735] [<c041eff8>] (sys_ioctl) from [<c0200040>] (ret_fast_syscall+0x0/0x28)
[  284.315739] Exception stack(0xc3d47fa8 to 0xc3d47ff0)
[  284.315745] 7fa0:                   027eb750 befdd870 00000000 c03864bc befdd870 00000000
[  284.315750] 7fc0: 027eb750 befdd870 c03864bc 00000036 027e3948 0281a640 0281a850 027e3a50
[  284.315756] 7fe0: b4b64100 befdd844 b4b5ba2c b49c994c
[  284.315762]  r10:00000036 r9:c3d46000 r8:c0200204 r7:00000036 r6:c03864bc r5:befdd870
[  284.315765]  r4:027eb750

Fixes: c54619b0bfb3 ("drm/vc4: Add support for the BCM2711 HVS5")
Signed-off-by: Dom Cobley <popcornmix@gmail.com>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
Reviewed-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Tested-By: Lucas Nussbaum <lucas@debian.org>
Tested-By: Ryutaroh Matsumoto <ryutaroh@ict.e.titech.ac.jp>
Link: https://patchwork.freedesktop.org/patch/msgid/20210121105759.1262699-2-maxime@cerno.tech
1 parent 78e5330
Raw File
percpu-refcount.c
// SPDX-License-Identifier: GPL-2.0-only
#define pr_fmt(fmt) "%s: " fmt, __func__

#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/wait.h>
#include <linux/slab.h>
#include <linux/percpu-refcount.h>

/*
 * Initially, a percpu refcount is just a set of percpu counters. Initially, we
 * don't try to detect the ref hitting 0 - which means that get/put can just
 * increment or decrement the local counter. Note that the counter on a
 * particular cpu can (and will) wrap - this is fine, when we go to shutdown the
 * percpu counters will all sum to the correct value
 *
 * (More precisely: because modular arithmetic is commutative the sum of all the
 * percpu_count vars will be equal to what it would have been if all the gets
 * and puts were done to a single integer, even if some of the percpu integers
 * overflow or underflow).
 *
 * The real trick to implementing percpu refcounts is shutdown. We can't detect
 * the ref hitting 0 on every put - this would require global synchronization
 * and defeat the whole purpose of using percpu refs.
 *
 * What we do is require the user to keep track of the initial refcount; we know
 * the ref can't hit 0 before the user drops the initial ref, so as long as we
 * convert to non percpu mode before the initial ref is dropped everything
 * works.
 *
 * Converting to non percpu mode is done with some RCUish stuff in
 * percpu_ref_kill. Additionally, we need a bias value so that the
 * atomic_long_t can't hit 0 before we've added up all the percpu refs.
 */

#define PERCPU_COUNT_BIAS	(1LU << (BITS_PER_LONG - 1))

static DEFINE_SPINLOCK(percpu_ref_switch_lock);
static DECLARE_WAIT_QUEUE_HEAD(percpu_ref_switch_waitq);

static unsigned long __percpu *percpu_count_ptr(struct percpu_ref *ref)
{
	return (unsigned long __percpu *)
		(ref->percpu_count_ptr & ~__PERCPU_REF_ATOMIC_DEAD);
}

/**
 * percpu_ref_init - initialize a percpu refcount
 * @ref: percpu_ref to initialize
 * @release: function which will be called when refcount hits 0
 * @flags: PERCPU_REF_INIT_* flags
 * @gfp: allocation mask to use
 *
 * Initializes @ref.  @ref starts out in percpu mode with a refcount of 1 unless
 * @flags contains PERCPU_REF_INIT_ATOMIC or PERCPU_REF_INIT_DEAD.  These flags
 * change the start state to atomic with the latter setting the initial refcount
 * to 0.  See the definitions of PERCPU_REF_INIT_* flags for flag behaviors.
 *
 * Note that @release must not sleep - it may potentially be called from RCU
 * callback context by percpu_ref_kill().
 */
int percpu_ref_init(struct percpu_ref *ref, percpu_ref_func_t *release,
		    unsigned int flags, gfp_t gfp)
{
	size_t align = max_t(size_t, 1 << __PERCPU_REF_FLAG_BITS,
			     __alignof__(unsigned long));
	unsigned long start_count = 0;
	struct percpu_ref_data *data;

	ref->percpu_count_ptr = (unsigned long)
		__alloc_percpu_gfp(sizeof(unsigned long), align, gfp);
	if (!ref->percpu_count_ptr)
		return -ENOMEM;

	data = kzalloc(sizeof(*ref->data), gfp);
	if (!data) {
		free_percpu((void __percpu *)ref->percpu_count_ptr);
		return -ENOMEM;
	}

	data->force_atomic = flags & PERCPU_REF_INIT_ATOMIC;
	data->allow_reinit = flags & PERCPU_REF_ALLOW_REINIT;

	if (flags & (PERCPU_REF_INIT_ATOMIC | PERCPU_REF_INIT_DEAD)) {
		ref->percpu_count_ptr |= __PERCPU_REF_ATOMIC;
		data->allow_reinit = true;
	} else {
		start_count += PERCPU_COUNT_BIAS;
	}

	if (flags & PERCPU_REF_INIT_DEAD)
		ref->percpu_count_ptr |= __PERCPU_REF_DEAD;
	else
		start_count++;

	atomic_long_set(&data->count, start_count);

	data->release = release;
	data->confirm_switch = NULL;
	data->ref = ref;
	ref->data = data;
	return 0;
}
EXPORT_SYMBOL_GPL(percpu_ref_init);

static void __percpu_ref_exit(struct percpu_ref *ref)
{
	unsigned long __percpu *percpu_count = percpu_count_ptr(ref);

	if (percpu_count) {
		/* non-NULL confirm_switch indicates switching in progress */
		WARN_ON_ONCE(ref->data && ref->data->confirm_switch);
		free_percpu(percpu_count);
		ref->percpu_count_ptr = __PERCPU_REF_ATOMIC_DEAD;
	}
}

/**
 * percpu_ref_exit - undo percpu_ref_init()
 * @ref: percpu_ref to exit
 *
 * This function exits @ref.  The caller is responsible for ensuring that
 * @ref is no longer in active use.  The usual places to invoke this
 * function from are the @ref->release() callback or in init failure path
 * where percpu_ref_init() succeeded but other parts of the initialization
 * of the embedding object failed.
 */
void percpu_ref_exit(struct percpu_ref *ref)
{
	struct percpu_ref_data *data = ref->data;
	unsigned long flags;

	__percpu_ref_exit(ref);

	if (!data)
		return;

	spin_lock_irqsave(&percpu_ref_switch_lock, flags);
	ref->percpu_count_ptr |= atomic_long_read(&ref->data->count) <<
		__PERCPU_REF_FLAG_BITS;
	ref->data = NULL;
	spin_unlock_irqrestore(&percpu_ref_switch_lock, flags);

	kfree(data);
}
EXPORT_SYMBOL_GPL(percpu_ref_exit);

static void percpu_ref_call_confirm_rcu(struct rcu_head *rcu)
{
	struct percpu_ref_data *data = container_of(rcu,
			struct percpu_ref_data, rcu);
	struct percpu_ref *ref = data->ref;

	data->confirm_switch(ref);
	data->confirm_switch = NULL;
	wake_up_all(&percpu_ref_switch_waitq);

	if (!data->allow_reinit)
		__percpu_ref_exit(ref);

	/* drop ref from percpu_ref_switch_to_atomic() */
	percpu_ref_put(ref);
}

static void percpu_ref_switch_to_atomic_rcu(struct rcu_head *rcu)
{
	struct percpu_ref_data *data = container_of(rcu,
			struct percpu_ref_data, rcu);
	struct percpu_ref *ref = data->ref;
	unsigned long __percpu *percpu_count = percpu_count_ptr(ref);
	unsigned long count = 0;
	int cpu;

	for_each_possible_cpu(cpu)
		count += *per_cpu_ptr(percpu_count, cpu);

	pr_debug("global %lu percpu %lu\n",
		 atomic_long_read(&data->count), count);

	/*
	 * It's crucial that we sum the percpu counters _before_ adding the sum
	 * to &ref->count; since gets could be happening on one cpu while puts
	 * happen on another, adding a single cpu's count could cause
	 * @ref->count to hit 0 before we've got a consistent value - but the
	 * sum of all the counts will be consistent and correct.
	 *
	 * Subtracting the bias value then has to happen _after_ adding count to
	 * &ref->count; we need the bias value to prevent &ref->count from
	 * reaching 0 before we add the percpu counts. But doing it at the same
	 * time is equivalent and saves us atomic operations:
	 */
	atomic_long_add((long)count - PERCPU_COUNT_BIAS, &data->count);

	WARN_ONCE(atomic_long_read(&data->count) <= 0,
		  "percpu ref (%ps) <= 0 (%ld) after switching to atomic",
		  data->release, atomic_long_read(&data->count));

	/* @ref is viewed as dead on all CPUs, send out switch confirmation */
	percpu_ref_call_confirm_rcu(rcu);
}

static void percpu_ref_noop_confirm_switch(struct percpu_ref *ref)
{
}

static void __percpu_ref_switch_to_atomic(struct percpu_ref *ref,
					  percpu_ref_func_t *confirm_switch)
{
	if (ref->percpu_count_ptr & __PERCPU_REF_ATOMIC) {
		if (confirm_switch)
			confirm_switch(ref);
		return;
	}

	/* switching from percpu to atomic */
	ref->percpu_count_ptr |= __PERCPU_REF_ATOMIC;

	/*
	 * Non-NULL ->confirm_switch is used to indicate that switching is
	 * in progress.  Use noop one if unspecified.
	 */
	ref->data->confirm_switch = confirm_switch ?:
		percpu_ref_noop_confirm_switch;

	percpu_ref_get(ref);	/* put after confirmation */
	call_rcu(&ref->data->rcu, percpu_ref_switch_to_atomic_rcu);
}

static void __percpu_ref_switch_to_percpu(struct percpu_ref *ref)
{
	unsigned long __percpu *percpu_count = percpu_count_ptr(ref);
	int cpu;

	BUG_ON(!percpu_count);

	if (!(ref->percpu_count_ptr & __PERCPU_REF_ATOMIC))
		return;

	if (WARN_ON_ONCE(!ref->data->allow_reinit))
		return;

	atomic_long_add(PERCPU_COUNT_BIAS, &ref->data->count);

	/*
	 * Restore per-cpu operation.  smp_store_release() is paired
	 * with READ_ONCE() in __ref_is_percpu() and guarantees that the
	 * zeroing is visible to all percpu accesses which can see the
	 * following __PERCPU_REF_ATOMIC clearing.
	 */
	for_each_possible_cpu(cpu)
		*per_cpu_ptr(percpu_count, cpu) = 0;

	smp_store_release(&ref->percpu_count_ptr,
			  ref->percpu_count_ptr & ~__PERCPU_REF_ATOMIC);
}

static void __percpu_ref_switch_mode(struct percpu_ref *ref,
				     percpu_ref_func_t *confirm_switch)
{
	struct percpu_ref_data *data = ref->data;

	lockdep_assert_held(&percpu_ref_switch_lock);

	/*
	 * If the previous ATOMIC switching hasn't finished yet, wait for
	 * its completion.  If the caller ensures that ATOMIC switching
	 * isn't in progress, this function can be called from any context.
	 */
	wait_event_lock_irq(percpu_ref_switch_waitq, !data->confirm_switch,
			    percpu_ref_switch_lock);

	if (data->force_atomic || (ref->percpu_count_ptr & __PERCPU_REF_DEAD))
		__percpu_ref_switch_to_atomic(ref, confirm_switch);
	else
		__percpu_ref_switch_to_percpu(ref);
}

/**
 * percpu_ref_switch_to_atomic - switch a percpu_ref to atomic mode
 * @ref: percpu_ref to switch to atomic mode
 * @confirm_switch: optional confirmation callback
 *
 * There's no reason to use this function for the usual reference counting.
 * Use percpu_ref_kill[_and_confirm]().
 *
 * Schedule switching of @ref to atomic mode.  All its percpu counts will
 * be collected to the main atomic counter.  On completion, when all CPUs
 * are guaraneed to be in atomic mode, @confirm_switch, which may not
 * block, is invoked.  This function may be invoked concurrently with all
 * the get/put operations and can safely be mixed with kill and reinit
 * operations.  Note that @ref will stay in atomic mode across kill/reinit
 * cycles until percpu_ref_switch_to_percpu() is called.
 *
 * This function may block if @ref is in the process of switching to atomic
 * mode.  If the caller ensures that @ref is not in the process of
 * switching to atomic mode, this function can be called from any context.
 */
void percpu_ref_switch_to_atomic(struct percpu_ref *ref,
				 percpu_ref_func_t *confirm_switch)
{
	unsigned long flags;

	spin_lock_irqsave(&percpu_ref_switch_lock, flags);

	ref->data->force_atomic = true;
	__percpu_ref_switch_mode(ref, confirm_switch);

	spin_unlock_irqrestore(&percpu_ref_switch_lock, flags);
}
EXPORT_SYMBOL_GPL(percpu_ref_switch_to_atomic);

/**
 * percpu_ref_switch_to_atomic_sync - switch a percpu_ref to atomic mode
 * @ref: percpu_ref to switch to atomic mode
 *
 * Schedule switching the ref to atomic mode, and wait for the
 * switch to complete.  Caller must ensure that no other thread
 * will switch back to percpu mode.
 */
void percpu_ref_switch_to_atomic_sync(struct percpu_ref *ref)
{
	percpu_ref_switch_to_atomic(ref, NULL);
	wait_event(percpu_ref_switch_waitq, !ref->data->confirm_switch);
}
EXPORT_SYMBOL_GPL(percpu_ref_switch_to_atomic_sync);

/**
 * percpu_ref_switch_to_percpu - switch a percpu_ref to percpu mode
 * @ref: percpu_ref to switch to percpu mode
 *
 * There's no reason to use this function for the usual reference counting.
 * To re-use an expired ref, use percpu_ref_reinit().
 *
 * Switch @ref to percpu mode.  This function may be invoked concurrently
 * with all the get/put operations and can safely be mixed with kill and
 * reinit operations.  This function reverses the sticky atomic state set
 * by PERCPU_REF_INIT_ATOMIC or percpu_ref_switch_to_atomic().  If @ref is
 * dying or dead, the actual switching takes place on the following
 * percpu_ref_reinit().
 *
 * This function may block if @ref is in the process of switching to atomic
 * mode.  If the caller ensures that @ref is not in the process of
 * switching to atomic mode, this function can be called from any context.
 */
void percpu_ref_switch_to_percpu(struct percpu_ref *ref)
{
	unsigned long flags;

	spin_lock_irqsave(&percpu_ref_switch_lock, flags);

	ref->data->force_atomic = false;
	__percpu_ref_switch_mode(ref, NULL);

	spin_unlock_irqrestore(&percpu_ref_switch_lock, flags);
}
EXPORT_SYMBOL_GPL(percpu_ref_switch_to_percpu);

/**
 * percpu_ref_kill_and_confirm - drop the initial ref and schedule confirmation
 * @ref: percpu_ref to kill
 * @confirm_kill: optional confirmation callback
 *
 * Equivalent to percpu_ref_kill() but also schedules kill confirmation if
 * @confirm_kill is not NULL.  @confirm_kill, which may not block, will be
 * called after @ref is seen as dead from all CPUs at which point all
 * further invocations of percpu_ref_tryget_live() will fail.  See
 * percpu_ref_tryget_live() for details.
 *
 * This function normally doesn't block and can be called from any context
 * but it may block if @confirm_kill is specified and @ref is in the
 * process of switching to atomic mode by percpu_ref_switch_to_atomic().
 *
 * There are no implied RCU grace periods between kill and release.
 */
void percpu_ref_kill_and_confirm(struct percpu_ref *ref,
				 percpu_ref_func_t *confirm_kill)
{
	unsigned long flags;

	spin_lock_irqsave(&percpu_ref_switch_lock, flags);

	WARN_ONCE(ref->percpu_count_ptr & __PERCPU_REF_DEAD,
		  "%s called more than once on %ps!", __func__,
		  ref->data->release);

	ref->percpu_count_ptr |= __PERCPU_REF_DEAD;
	__percpu_ref_switch_mode(ref, confirm_kill);
	percpu_ref_put(ref);

	spin_unlock_irqrestore(&percpu_ref_switch_lock, flags);
}
EXPORT_SYMBOL_GPL(percpu_ref_kill_and_confirm);

/**
 * percpu_ref_is_zero - test whether a percpu refcount reached zero
 * @ref: percpu_ref to test
 *
 * Returns %true if @ref reached zero.
 *
 * This function is safe to call as long as @ref is between init and exit.
 */
bool percpu_ref_is_zero(struct percpu_ref *ref)
{
	unsigned long __percpu *percpu_count;
	unsigned long count, flags;

	if (__ref_is_percpu(ref, &percpu_count))
		return false;

	/* protect us from being destroyed */
	spin_lock_irqsave(&percpu_ref_switch_lock, flags);
	if (ref->data)
		count = atomic_long_read(&ref->data->count);
	else
		count = ref->percpu_count_ptr >> __PERCPU_REF_FLAG_BITS;
	spin_unlock_irqrestore(&percpu_ref_switch_lock, flags);

	return count == 0;
}
EXPORT_SYMBOL_GPL(percpu_ref_is_zero);

/**
 * percpu_ref_reinit - re-initialize a percpu refcount
 * @ref: perpcu_ref to re-initialize
 *
 * Re-initialize @ref so that it's in the same state as when it finished
 * percpu_ref_init() ignoring %PERCPU_REF_INIT_DEAD.  @ref must have been
 * initialized successfully and reached 0 but not exited.
 *
 * Note that percpu_ref_tryget[_live]() are safe to perform on @ref while
 * this function is in progress.
 */
void percpu_ref_reinit(struct percpu_ref *ref)
{
	WARN_ON_ONCE(!percpu_ref_is_zero(ref));

	percpu_ref_resurrect(ref);
}
EXPORT_SYMBOL_GPL(percpu_ref_reinit);

/**
 * percpu_ref_resurrect - modify a percpu refcount from dead to live
 * @ref: perpcu_ref to resurrect
 *
 * Modify @ref so that it's in the same state as before percpu_ref_kill() was
 * called. @ref must be dead but must not yet have exited.
 *
 * If @ref->release() frees @ref then the caller is responsible for
 * guaranteeing that @ref->release() does not get called while this
 * function is in progress.
 *
 * Note that percpu_ref_tryget[_live]() are safe to perform on @ref while
 * this function is in progress.
 */
void percpu_ref_resurrect(struct percpu_ref *ref)
{
	unsigned long __percpu *percpu_count;
	unsigned long flags;

	spin_lock_irqsave(&percpu_ref_switch_lock, flags);

	WARN_ON_ONCE(!(ref->percpu_count_ptr & __PERCPU_REF_DEAD));
	WARN_ON_ONCE(__ref_is_percpu(ref, &percpu_count));

	ref->percpu_count_ptr &= ~__PERCPU_REF_DEAD;
	percpu_ref_get(ref);
	__percpu_ref_switch_mode(ref, NULL);

	spin_unlock_irqrestore(&percpu_ref_switch_lock, flags);
}
EXPORT_SYMBOL_GPL(percpu_ref_resurrect);
back to top