The following commit has been merged into the locking/urgent branch of tip:
Commit-ID: 1c0908d8e441631f5b8ba433523cf39339ee2ba0
Gitweb: https://git.kernel.org/tip/1c0908d8e441631f5b8ba433523cf39339ee2ba0
Author: Mel Gorman <mgorman(a)techsingularity.net>
AuthorDate: Fri, 02 Dec 2022 10:02:23
Committer: Thomas Gleixner <tglx(a)linutronix.de>
CommitterDate: Mon, 12 Dec 2022 19:55:56 +01:00
rtmutex: Add acquire semantics for rtmutex lock acquisition slow path
Jan Kara reported the following bug triggering on 6.0.5-rt14 running dbench
on XFS on arm64.
kernel BUG at fs/inode.c:625!
Internal error: Oops - BUG: 0 [#1] PREEMPT_RT SMP
CPU: 11 PID: 6611 Comm: dbench Tainted: G E 6.0.0-rt14-rt+ #1
pc : clear_inode+0xa0/0xc0
lr : clear_inode+0x38/0xc0
Call trace:
clear_inode+0xa0/0xc0
evict+0x160/0x180
iput+0x154/0x240
do_unlinkat+0x184/0x300
__arm64_sys_unlinkat+0x48/0xc0
el0_svc_common.constprop.4+0xe4/0x2c0
do_el0_svc+0xac/0x100
el0_svc+0x78/0x200
el0t_64_sync_handler+0x9c/0xc0
el0t_64_sync+0x19c/0x1a0
It also affects 6.1-rc7-rt5 and affects a preempt-rt fork of 5.14 so this
is likely a bug that existed forever and only became visible when ARM
support was added to preempt-rt. The same problem does not occur on x86-64
and he also reported that converting sb->s_inode_wblist_lock to
raw_spinlock_t makes the problem disappear indicating that the RT spinlock
variant is the problem.
Which in turn means that RT mutexes on ARM64 and any other weakly ordered
architecture are affected by this independent of RT.
Will Deacon observed:
"I'd be more inclined to be suspicious of the slowpath tbh, as we need to
make sure that we have acquire semantics on all paths where the lock can
be taken. Looking at the rtmutex code, this really isn't obvious to me
-- for example, try_to_take_rt_mutex() appears to be able to return via
the 'takeit' label without acquire semantics and it looks like we might
be relying on the caller's subsequent _unlock_ of the wait_lock for
ordering, but that will give us release semantics which aren't correct."
Sebastian Andrzej Siewior prototyped a fix that does work based on that
comment but it was a little bit overkill and added some fences that should
not be necessary.
The lock owner is updated with an IRQ-safe raw spinlock held, but the
spin_unlock does not provide acquire semantics which are needed when
acquiring a mutex.
Adds the necessary acquire semantics for lock owner updates in the slow path
acquisition and the waiter bit logic.
It successfully completed 10 iterations of the dbench workload while the
vanilla kernel fails on the first iteration.
[ bigeasy(a)linutronix.de: Initial prototype fix ]
Fixes: 700318d1d7b38 ("locking/rtmutex: Use acquire/release semantics")
Fixes: 23f78d4a03c5 ("[PATCH] pi-futex: rt mutex core")
Reported-by: Jan Kara <jack(a)suse.cz>
Signed-off-by: Mel Gorman <mgorman(a)techsingularity.net>
Signed-off-by: Thomas Gleixner <tglx(a)linutronix.de>
Cc: stable(a)vger.kernel.org
Link: https://lore.kernel.org/r/20221202100223.6mevpbl7i6x5udfd@techsingularity.n…
---
kernel/locking/rtmutex.c | 55 +++++++++++++++++++++++++++++------
kernel/locking/rtmutex_api.c | 6 ++--
2 files changed, 49 insertions(+), 12 deletions(-)
diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index 7779ee8..010cf4e 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -89,15 +89,31 @@ static inline int __ww_mutex_check_kill(struct rt_mutex *lock,
* set this bit before looking at the lock.
*/
-static __always_inline void
-rt_mutex_set_owner(struct rt_mutex_base *lock, struct task_struct *owner)
+static __always_inline struct task_struct *
+rt_mutex_owner_encode(struct rt_mutex_base *lock, struct task_struct *owner)
{
unsigned long val = (unsigned long)owner;
if (rt_mutex_has_waiters(lock))
val |= RT_MUTEX_HAS_WAITERS;
- WRITE_ONCE(lock->owner, (struct task_struct *)val);
+ return (struct task_struct *)val;
+}
+
+static __always_inline void
+rt_mutex_set_owner(struct rt_mutex_base *lock, struct task_struct *owner)
+{
+ /*
+ * lock->wait_lock is held but explicit acquire semantics are needed
+ * for a new lock owner so WRITE_ONCE is insufficient.
+ */
+ xchg_acquire(&lock->owner, rt_mutex_owner_encode(lock, owner));
+}
+
+static __always_inline void rt_mutex_clear_owner(struct rt_mutex_base *lock)
+{
+ /* lock->wait_lock is held so the unlock provides release semantics. */
+ WRITE_ONCE(lock->owner, rt_mutex_owner_encode(lock, NULL));
}
static __always_inline void clear_rt_mutex_waiters(struct rt_mutex_base *lock)
@@ -106,7 +122,8 @@ static __always_inline void clear_rt_mutex_waiters(struct rt_mutex_base *lock)
((unsigned long)lock->owner & ~RT_MUTEX_HAS_WAITERS);
}
-static __always_inline void fixup_rt_mutex_waiters(struct rt_mutex_base *lock)
+static __always_inline void
+fixup_rt_mutex_waiters(struct rt_mutex_base *lock, bool acquire_lock)
{
unsigned long owner, *p = (unsigned long *) &lock->owner;
@@ -172,8 +189,21 @@ static __always_inline void fixup_rt_mutex_waiters(struct rt_mutex_base *lock)
* still set.
*/
owner = READ_ONCE(*p);
- if (owner & RT_MUTEX_HAS_WAITERS)
- WRITE_ONCE(*p, owner & ~RT_MUTEX_HAS_WAITERS);
+ if (owner & RT_MUTEX_HAS_WAITERS) {
+ /*
+ * See rt_mutex_set_owner() and rt_mutex_clear_owner() on
+ * why xchg_acquire() is used for updating owner for
+ * locking and WRITE_ONCE() for unlocking.
+ *
+ * WRITE_ONCE() would work for the acquire case too, but
+ * in case that the lock acquisition failed it might
+ * force other lockers into the slow path unnecessarily.
+ */
+ if (acquire_lock)
+ xchg_acquire(p, owner & ~RT_MUTEX_HAS_WAITERS);
+ else
+ WRITE_ONCE(*p, owner & ~RT_MUTEX_HAS_WAITERS);
+ }
}
/*
@@ -208,6 +238,13 @@ static __always_inline void mark_rt_mutex_waiters(struct rt_mutex_base *lock)
owner = *p;
} while (cmpxchg_relaxed(p, owner,
owner | RT_MUTEX_HAS_WAITERS) != owner);
+
+ /*
+ * The cmpxchg loop above is relaxed to avoid back-to-back ACQUIRE
+ * operations in the event of contention. Ensure the successful
+ * cmpxchg is visible.
+ */
+ smp_mb__after_atomic();
}
/*
@@ -1243,7 +1280,7 @@ static int __sched __rt_mutex_slowtrylock(struct rt_mutex_base *lock)
* try_to_take_rt_mutex() sets the lock waiters bit
* unconditionally. Clean this up.
*/
- fixup_rt_mutex_waiters(lock);
+ fixup_rt_mutex_waiters(lock, true);
return ret;
}
@@ -1604,7 +1641,7 @@ static int __sched __rt_mutex_slowlock(struct rt_mutex_base *lock,
* try_to_take_rt_mutex() sets the waiter bit
* unconditionally. We might have to fix that up.
*/
- fixup_rt_mutex_waiters(lock);
+ fixup_rt_mutex_waiters(lock, true);
trace_contention_end(lock, ret);
@@ -1719,7 +1756,7 @@ static void __sched rtlock_slowlock_locked(struct rt_mutex_base *lock)
* try_to_take_rt_mutex() sets the waiter bit unconditionally.
* We might have to fix that up:
*/
- fixup_rt_mutex_waiters(lock);
+ fixup_rt_mutex_waiters(lock, true);
debug_rt_mutex_free_waiter(&waiter);
trace_contention_end(lock, 0);
diff --git a/kernel/locking/rtmutex_api.c b/kernel/locking/rtmutex_api.c
index 9002209..cb9fdff 100644
--- a/kernel/locking/rtmutex_api.c
+++ b/kernel/locking/rtmutex_api.c
@@ -267,7 +267,7 @@ void __sched rt_mutex_init_proxy_locked(struct rt_mutex_base *lock,
void __sched rt_mutex_proxy_unlock(struct rt_mutex_base *lock)
{
debug_rt_mutex_proxy_unlock(lock);
- rt_mutex_set_owner(lock, NULL);
+ rt_mutex_clear_owner(lock);
}
/**
@@ -382,7 +382,7 @@ int __sched rt_mutex_wait_proxy_lock(struct rt_mutex_base *lock,
* try_to_take_rt_mutex() sets the waiter bit unconditionally. We might
* have to fix that up.
*/
- fixup_rt_mutex_waiters(lock);
+ fixup_rt_mutex_waiters(lock, true);
raw_spin_unlock_irq(&lock->wait_lock);
return ret;
@@ -438,7 +438,7 @@ bool __sched rt_mutex_cleanup_proxy_lock(struct rt_mutex_base *lock,
* try_to_take_rt_mutex() sets the waiter bit unconditionally. We might
* have to fix that up.
*/
- fixup_rt_mutex_waiters(lock);
+ fixup_rt_mutex_waiters(lock, false);
raw_spin_unlock_irq(&lock->wait_lock);
Hello Rafael,
On Tue, Oct 11, 2022 at 10:20:50AM +0100, Mel Gorman wrote:
> On Mon, Oct 10, 2022 at 08:29:05PM +0200, Rafael J. Wysocki wrote:
> > > That's less than the previous 5/10 failures but I
> > > cannot be certain it helped without running a lot more boot tests. The
> > > failure happens in the same function as before.
> >
> > I've overlooked the fact that acpi_install_fixed_event_handler()
> > enables the event on success, so it is a bug to call it when the
> > handler is not ready.
> >
> > It should help to only enable the event after running cmos_do_probe()
> > where the driver data pointer is set, so please try the attached
> > patch.
I'm hitting this issue on the 6.0 stable releases (aka 6.0.y) and
looking at the stable tree I see this hasn't been merged... I just got
bitten by this on 6.0.12.
Greg, if Rafael agrees, I think you should apply 4919d3eb2ec0 and
0782b66ed2fb to the 6.0.y tree.
Thank you in advance.
Cheers,
--
Mathieu Chouquet-Stringer me(a)mathieu.digital
The sun itself sees not till heaven clears.
-- William Shakespeare --
It seems we can have one or more framebuffers that are still pinned when
suspending lmem, in such a case we end up creating a shmem backup
object, instead of evicting the object directly, but this will skip
copying the CCS aux state, since we don't allocate the extra storage for
the CCS pages as part of the ttm_tt construction. Since we can already
deal with pinned objects just fine, it doesn't seem too nasty to just
extend to support dealing with the CCS aux state, if the object is a
pinned framebuffer. This fixes display corruption (like in gnome-shell)
seen on DG2 when returning from suspend.
Fixes: da0595ae91da ("drm/i915/migrate: Evict and restore the flatccs capable lmem obj")
Signed-off-by: Matthew Auld <matthew.auld(a)intel.com>
Cc: Ville Syrjälä <ville.syrjala(a)linux.intel.com>
Cc: Nirmoy Das <nirmoy.das(a)intel.com>
Cc: Andrzej Hajda <andrzej.hajda(a)intel.com>
Cc: Shuicheng Lin <shuicheng.lin(a)intel.com>
Cc: <stable(a)vger.kernel.org> # v5.19+
---
drivers/gpu/drm/i915/gem/i915_gem_object.c | 3 +++
.../gpu/drm/i915/gem/i915_gem_object_types.h | 10 ++++++----
drivers/gpu/drm/i915/gem/i915_gem_ttm_pm.c | 18 +++++++++++++++++-
3 files changed, 26 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 733696057761..1a0886b8aaa1 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -785,6 +785,9 @@ bool i915_gem_object_needs_ccs_pages(struct drm_i915_gem_object *obj)
if (!HAS_FLAT_CCS(to_i915(obj->base.dev)))
return false;
+ if (obj->flags & I915_BO_ALLOC_CCS_AUX)
+ return true;
+
for (i = 0; i < obj->mm.n_placements; i++) {
/* Compression is not allowed for the objects with smem placement */
if (obj->mm.placements[i]->type == INTEL_MEMORY_SYSTEM)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index a7b70701617a..19c9bdd8f905 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -327,16 +327,18 @@ struct drm_i915_gem_object {
* dealing with userspace objects the CPU fault handler is free to ignore this.
*/
#define I915_BO_ALLOC_GPU_ONLY BIT(6)
+#define I915_BO_ALLOC_CCS_AUX BIT(7)
#define I915_BO_ALLOC_FLAGS (I915_BO_ALLOC_CONTIGUOUS | \
I915_BO_ALLOC_VOLATILE | \
I915_BO_ALLOC_CPU_CLEAR | \
I915_BO_ALLOC_USER | \
I915_BO_ALLOC_PM_VOLATILE | \
I915_BO_ALLOC_PM_EARLY | \
- I915_BO_ALLOC_GPU_ONLY)
-#define I915_BO_READONLY BIT(7)
-#define I915_TILING_QUIRK_BIT 8 /* unknown swizzling; do not release! */
-#define I915_BO_PROTECTED BIT(9)
+ I915_BO_ALLOC_GPU_ONLY | \
+ I915_BO_ALLOC_CCS_AUX)
+#define I915_BO_READONLY BIT(8)
+#define I915_TILING_QUIRK_BIT 9 /* unknown swizzling; do not release! */
+#define I915_BO_PROTECTED BIT(10)
/**
* @mem_flags - Mutable placement-related flags
*
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm_pm.c
index 07e49f22f2de..7e67742bc65e 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_ttm_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm_pm.c
@@ -50,6 +50,7 @@ static int i915_ttm_backup(struct i915_gem_apply_to_region *apply,
container_of(bo->bdev, typeof(*i915), bdev);
struct drm_i915_gem_object *backup;
struct ttm_operation_ctx ctx = {};
+ unsigned int flags;
int err = 0;
if (bo->resource->mem_type == I915_PL_SYSTEM || obj->ttm.backup)
@@ -65,7 +66,22 @@ static int i915_ttm_backup(struct i915_gem_apply_to_region *apply,
if (obj->flags & I915_BO_ALLOC_PM_VOLATILE)
return 0;
- backup = i915_gem_object_create_shmem(i915, obj->base.size);
+ /*
+ * It seems that we might have some framebuffers still pinned at this
+ * stage, but for such objects we might also need to deal with the CCS
+ * aux state. Make sure we force the save/restore of the CCS state,
+ * otherwise we might observe display corruption, when returning from
+ * suspend.
+ */
+ flags = 0;
+ if (i915_gem_object_needs_ccs_pages(obj)) {
+ WARN_ON_ONCE(!i915_gem_object_is_framebuffer(obj));
+ WARN_ON_ONCE(!pm_apply->allow_gpu);
+
+ flags = I915_BO_ALLOC_CCS_AUX;
+ }
+ backup = i915_gem_object_create_region(i915->mm.regions[INTEL_REGION_SMEM],
+ obj->base.size, 0, flags);
if (IS_ERR(backup))
return PTR_ERR(backup);
--
2.38.1