* [PATCH 1/3] dma-buf: Prettify typecasts for dma-fence-chain
@ 2020-04-09 11:00 ` Chris Wilson
0 siblings, 0 replies; 9+ messages in thread
From: Chris Wilson @ 2020-04-09 11:00 UTC (permalink / raw)
To: dri-devel; +Cc: Mika Kuoppala, intel-gfx, Chris Wilson
Inside dma-fence-chain, we use a cmpxchg on an RCU-protected pointer. To
avoid the sparse warning for using the RCU pointer directly, we have to
cast away the __rcu annotation. However, we don't need to use void*
everywhere and can stick to the dma_fence*.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
---
drivers/dma-buf/dma-fence-chain.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/dma-buf/dma-fence-chain.c b/drivers/dma-buf/dma-fence-chain.c
index 44a741677d25..3d123502ff12 100644
--- a/drivers/dma-buf/dma-fence-chain.c
+++ b/drivers/dma-buf/dma-fence-chain.c
@@ -62,7 +62,8 @@ struct dma_fence *dma_fence_chain_walk(struct dma_fence *fence)
replacement = NULL;
}
- tmp = cmpxchg((void **)&chain->prev, (void *)prev, (void *)replacement);
+ tmp = cmpxchg((struct dma_fence __force **)&chain->prev,
+ prev, replacement);
if (tmp == prev)
dma_fence_put(tmp);
else
--
2.20.1
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [Intel-gfx] [PATCH 1/3] dma-buf: Prettify typecasts for dma-fence-chain
@ 2020-04-09 11:00 ` Chris Wilson
0 siblings, 0 replies; 9+ messages in thread
From: Chris Wilson @ 2020-04-09 11:00 UTC (permalink / raw)
To: dri-devel; +Cc: intel-gfx, Chris Wilson
Inside dma-fence-chain, we use a cmpxchg on an RCU-protected pointer. To
avoid the sparse warning for using the RCU pointer directly, we have to
cast away the __rcu annotation. However, we don't need to use void*
everywhere and can stick to the dma_fence*.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
---
drivers/dma-buf/dma-fence-chain.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/dma-buf/dma-fence-chain.c b/drivers/dma-buf/dma-fence-chain.c
index 44a741677d25..3d123502ff12 100644
--- a/drivers/dma-buf/dma-fence-chain.c
+++ b/drivers/dma-buf/dma-fence-chain.c
@@ -62,7 +62,8 @@ struct dma_fence *dma_fence_chain_walk(struct dma_fence *fence)
replacement = NULL;
}
- tmp = cmpxchg((void **)&chain->prev, (void *)prev, (void *)replacement);
+ tmp = cmpxchg((struct dma_fence __force **)&chain->prev,
+ prev, replacement);
if (tmp == prev)
dma_fence_put(tmp);
else
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH 2/3] dma-buf: Report signaled links inside dma-fence-chain
2020-04-09 11:00 ` [Intel-gfx] " Chris Wilson
@ 2020-04-09 11:01 ` Chris Wilson
-1 siblings, 0 replies; 9+ messages in thread
From: Chris Wilson @ 2020-04-09 11:01 UTC (permalink / raw)
To: dri-devel; +Cc: intel-gfx, Venkata Sandeep Dhanalakota, Chris Wilson
Whenever we walk along the dma-fence-chain, we prune signaled links to
keep the chain nice and tidy. This leads to situations where we can
prune a link and report the earlier fence as the target seqno --
violating our own consistency checks that the seqno is not more advanced
than the last element in a dma-fence-chain.
Report a NULL fence and success if the seqno has already been signaled.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
Reviewed-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
---
drivers/dma-buf/dma-fence-chain.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/dma-buf/dma-fence-chain.c b/drivers/dma-buf/dma-fence-chain.c
index 3d123502ff12..c435bbba851c 100644
--- a/drivers/dma-buf/dma-fence-chain.c
+++ b/drivers/dma-buf/dma-fence-chain.c
@@ -99,6 +99,12 @@ int dma_fence_chain_find_seqno(struct dma_fence **pfence, uint64_t seqno)
return -EINVAL;
dma_fence_chain_for_each(*pfence, &chain->base) {
+ if ((*pfence)->seqno < seqno) { /* already signaled */
+ dma_fence_put(*pfence);
+ *pfence = NULL;
+ break;
+ }
+
if ((*pfence)->context != chain->base.context ||
to_dma_fence_chain(*pfence)->prev_seqno < seqno)
break;
@@ -222,6 +228,7 @@ EXPORT_SYMBOL(dma_fence_chain_ops);
* @chain: the chain node to initialize
* @prev: the previous fence
* @fence: the current fence
+ * @seqno: the sequence number (syncpt) of the fence within the chain
*
* Initialize a new chain node and either start a new chain or add the node to
* the existing chain of the previous fence.
--
2.20.1
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [Intel-gfx] [PATCH 2/3] dma-buf: Report signaled links inside dma-fence-chain
@ 2020-04-09 11:01 ` Chris Wilson
0 siblings, 0 replies; 9+ messages in thread
From: Chris Wilson @ 2020-04-09 11:01 UTC (permalink / raw)
To: dri-devel; +Cc: intel-gfx, Chris Wilson
Whenever we walk along the dma-fence-chain, we prune signaled links to
keep the chain nice and tidy. This leads to situations where we can
prune a link and report the earlier fence as the target seqno --
violating our own consistency checks that the seqno is not more advanced
than the last element in a dma-fence-chain.
Report a NULL fence and success if the seqno has already been signaled.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
Reviewed-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
---
drivers/dma-buf/dma-fence-chain.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/dma-buf/dma-fence-chain.c b/drivers/dma-buf/dma-fence-chain.c
index 3d123502ff12..c435bbba851c 100644
--- a/drivers/dma-buf/dma-fence-chain.c
+++ b/drivers/dma-buf/dma-fence-chain.c
@@ -99,6 +99,12 @@ int dma_fence_chain_find_seqno(struct dma_fence **pfence, uint64_t seqno)
return -EINVAL;
dma_fence_chain_for_each(*pfence, &chain->base) {
+ if ((*pfence)->seqno < seqno) { /* already signaled */
+ dma_fence_put(*pfence);
+ *pfence = NULL;
+ break;
+ }
+
if ((*pfence)->context != chain->base.context ||
to_dma_fence_chain(*pfence)->prev_seqno < seqno)
break;
@@ -222,6 +228,7 @@ EXPORT_SYMBOL(dma_fence_chain_ops);
* @chain: the chain node to initialize
* @prev: the previous fence
* @fence: the current fence
+ * @seqno: the sequence number (syncpt) of the fence within the chain
*
* Initialize a new chain node and either start a new chain or add the node to
* the existing chain of the previous fence.
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH 3/3] dma-buf: Exercise dma-fence-chain under selftests
2020-04-09 11:00 ` [Intel-gfx] " Chris Wilson
@ 2020-04-09 11:01 ` Chris Wilson
-1 siblings, 0 replies; 9+ messages in thread
From: Chris Wilson @ 2020-04-09 11:01 UTC (permalink / raw)
To: dri-devel; +Cc: intel-gfx, Venkata Sandeep Dhanalakota, Chris Wilson
A few very simple testcases to exercise the dma-fence-chain API.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
---
drivers/dma-buf/Makefile | 3 +-
drivers/dma-buf/selftests.h | 1 +
drivers/dma-buf/st-dma-fence-chain.c | 713 +++++++++++++++++++++++++++
3 files changed, 716 insertions(+), 1 deletion(-)
create mode 100644 drivers/dma-buf/st-dma-fence-chain.c
diff --git a/drivers/dma-buf/Makefile b/drivers/dma-buf/Makefile
index 9c190026bfab..995e05f609ff 100644
--- a/drivers/dma-buf/Makefile
+++ b/drivers/dma-buf/Makefile
@@ -9,6 +9,7 @@ obj-$(CONFIG_UDMABUF) += udmabuf.o
dmabuf_selftests-y := \
selftest.o \
- st-dma-fence.o
+ st-dma-fence.o \
+ st-dma-fence-chain.o
obj-$(CONFIG_DMABUF_SELFTESTS) += dmabuf_selftests.o
diff --git a/drivers/dma-buf/selftests.h b/drivers/dma-buf/selftests.h
index 5320386f02e5..55918ef9adab 100644
--- a/drivers/dma-buf/selftests.h
+++ b/drivers/dma-buf/selftests.h
@@ -11,3 +11,4 @@
*/
selftest(sanitycheck, __sanitycheck__) /* keep first (igt selfcheck) */
selftest(dma_fence, dma_fence)
+selftest(dma_fence_chain, dma_fence_chain)
diff --git a/drivers/dma-buf/st-dma-fence-chain.c b/drivers/dma-buf/st-dma-fence-chain.c
new file mode 100644
index 000000000000..bd08ba67b03b
--- /dev/null
+++ b/drivers/dma-buf/st-dma-fence-chain.c
@@ -0,0 +1,713 @@
+// SPDX-License-Identifier: MIT
+
+/*
+ * Copyright © 2019 Intel Corporation
+ */
+
+#include <linux/delay.h>
+#include <linux/dma-fence.h>
+#include <linux/dma-fence-chain.h>
+#include <linux/kernel.h>
+#include <linux/kthread.h>
+#include <linux/mm.h>
+#include <linux/sched/signal.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/random.h>
+
+#include "selftest.h"
+
+static struct kmem_cache *slab_fences;
+
+static inline struct mock_fence {
+ struct dma_fence base;
+ spinlock_t lock;
+} *to_mock_fence(struct dma_fence *f) {
+ return container_of(f, struct mock_fence, base);
+}
+
+static const char *mock_name(struct dma_fence *f)
+{
+ return "mock";
+}
+
+static void mock_fence_release(struct dma_fence *f)
+{
+ kmem_cache_free(slab_fences, to_mock_fence(f));
+}
+
+static const struct dma_fence_ops mock_ops = {
+ .get_driver_name = mock_name,
+ .get_timeline_name = mock_name,
+ .release = mock_fence_release,
+};
+
+static struct dma_fence *mock_fence(void)
+{
+ struct mock_fence *f;
+
+ f = kmem_cache_alloc(slab_fences, GFP_KERNEL);
+ if (!f)
+ return NULL;
+
+ spin_lock_init(&f->lock);
+ dma_fence_init(&f->base, &mock_ops, &f->lock, 0, 0);
+
+ return &f->base;
+}
+
+static inline struct mock_chain {
+ struct dma_fence_chain base;
+} *to_mock_chain(struct dma_fence *f) {
+ return container_of(f, struct mock_chain, base.base);
+}
+
+static struct dma_fence *mock_chain(struct dma_fence *prev,
+ struct dma_fence *fence,
+ u64 seqno)
+{
+ struct mock_chain *f;
+
+ f = kmalloc(sizeof(*f), GFP_KERNEL);
+ if (!f)
+ return NULL;
+
+ dma_fence_chain_init(&f->base,
+ dma_fence_get(prev),
+ dma_fence_get(fence),
+ seqno);
+
+ return &f->base.base;
+}
+
+static int sanitycheck(void *arg)
+{
+ struct dma_fence *f, *chain;
+ int err = 0;
+
+ f = mock_fence();
+ if (!f)
+ return -ENOMEM;
+
+ chain = mock_chain(NULL, f, 1);
+ if (!chain)
+ err = -ENOMEM;
+
+ dma_fence_signal(f);
+ dma_fence_put(f);
+
+ dma_fence_put(chain);
+
+ return err;
+}
+
+struct fence_chains {
+ unsigned int chain_length;
+ struct dma_fence **fences;
+ struct dma_fence **chains;
+
+ struct dma_fence *tail;
+};
+
+static uint64_t seqno_inc(unsigned int i)
+{
+ return i + 1;
+}
+
+static int fence_chains_init(struct fence_chains *fc, unsigned int count,
+ uint64_t (*seqno_fn)(unsigned int))
+{
+ unsigned int i;
+ int err = 0;
+
+ fc->chains = kvmalloc_array(count, sizeof(*fc->chains),
+ GFP_KERNEL | __GFP_ZERO);
+ if (!fc->chains)
+ return -ENOMEM;
+
+ fc->fences = kvmalloc_array(count, sizeof(*fc->fences),
+ GFP_KERNEL | __GFP_ZERO);
+ if (!fc->fences) {
+ err = -ENOMEM;
+ goto err_chains;
+ }
+
+ fc->tail = NULL;
+ for (i = 0; i < count; i++) {
+ fc->fences[i] = mock_fence();
+ if (!fc->fences[i]) {
+ err = -ENOMEM;
+ goto unwind;
+ }
+
+ fc->chains[i] = mock_chain(fc->tail,
+ fc->fences[i],
+ seqno_fn(i));
+ if (!fc->chains[i]) {
+ err = -ENOMEM;
+ goto unwind;
+ }
+
+ fc->tail = fc->chains[i];
+ }
+
+ fc->chain_length = i;
+ return 0;
+
+unwind:
+ for (i = 0; i < count; i++) {
+ dma_fence_put(fc->fences[i]);
+ dma_fence_put(fc->chains[i]);
+ }
+ kvfree(fc->fences);
+err_chains:
+ kvfree(fc->chains);
+ return err;
+}
+
+static void fence_chains_fini(struct fence_chains *fc)
+{
+ unsigned int i;
+
+ for (i = 0; i < fc->chain_length; i++) {
+ dma_fence_signal(fc->fences[i]);
+ dma_fence_put(fc->fences[i]);
+ }
+ kvfree(fc->fences);
+
+ for (i = 0; i < fc->chain_length; i++)
+ dma_fence_put(fc->chains[i]);
+ kvfree(fc->chains);
+}
+
+static int find_seqno(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc);
+ if (err)
+ return err;
+
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 0);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno(0)!\n", err);
+ goto err;
+ }
+
+ for (i = 0; i < fc.chain_length; i++) {
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, i + 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno(%d:%d)!\n",
+ err, fc.chain_length + 1, i + 1);
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence reported by find_seqno(%d:%d)\n",
+ fc.chain_length + 1, i + 1);
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, i + 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Error reported for finding self\n");
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence reported by find self\n");
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, i + 2);
+ dma_fence_put(fence);
+ if (!err) {
+ pr_err("Error not reported for future fence: find_seqno(%d:%d)!\n",
+ i + 1, i + 2);
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, i);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Error reported for previous fence!\n");
+ goto err;
+ }
+ if (i > 0 && fence != fc.chains[i - 1]) {
+ pr_err("Incorrect fence reported by find_seqno(%d:%d)\n",
+ i + 1, i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int find_signaled(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+
+ err = fence_chains_init(&fc, 2, seqno_inc);
+ if (err)
+ return err;
+
+ dma_fence_signal(fc.fences[0]);
+
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno()!\n", err);
+ goto err;
+ }
+
+ if (fence && fence != fc.chains[0]) {
+ pr_err("Incorrect chain-fence.seqno:%lld reported for completed seqno:1\n",
+ fence->seqno);
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, 1);
+ dma_fence_put(fence);
+ if (err)
+ pr_err("Reported %d for finding self!\n", err);
+
+ err = -EINVAL;
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int find_out_of_order(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+
+ err = fence_chains_init(&fc, 3, seqno_inc);
+ if (err)
+ return err;
+
+ dma_fence_signal(fc.fences[1]);
+
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 2);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno()!\n", err);
+ goto err;
+ }
+
+ if (fence && fence != fc.chains[1]) {
+ pr_err("Incorrect chain-fence.seqno:%lld reported for completed seqno:2\n",
+ fence->seqno);
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, 2);
+ dma_fence_put(fence);
+ if (err)
+ pr_err("Reported %d for finding self!\n", err);
+
+ err = -EINVAL;
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static uint64_t seqno_inc2(unsigned int i)
+{
+ return 2 * i + 2;
+}
+
+static int find_gap(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc2);
+ if (err)
+ return err;
+
+ for (i = 0; i < fc.chain_length; i++) {
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 2 * i + 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno(%d:%d)!\n",
+ err, fc.chain_length + 1, 2 * i + 1);
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence.seqno:%lld reported by find_seqno(%d:%d)\n",
+ fence->seqno,
+ fc.chain_length + 1,
+ 2 * i + 1);
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, 2 * i + 2);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Error reported for finding self\n");
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence reported by find self\n");
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+struct find_race {
+ struct fence_chains fc;
+ atomic_t children;
+};
+
+static int __find_race(void *arg)
+{
+ struct find_race *data = arg;
+ int err = 0;
+
+ while (!kthread_should_stop()) {
+ struct dma_fence *fence = dma_fence_get(data->fc.tail);
+ int seqno;
+
+ seqno = prandom_u32_max(data->fc.chain_length) + 1;
+
+ err = dma_fence_chain_find_seqno(&fence, seqno);
+ if (err) {
+ pr_err("Failed to find fence seqno:%d\n",
+ seqno);
+ dma_fence_put(fence);
+ break;
+ }
+ if (!fence)
+ goto signal;
+
+ err = dma_fence_chain_find_seqno(&fence, seqno);
+ if (err) {
+ pr_err("Reported an invalid fence for find-self:%d\n",
+ seqno);
+ dma_fence_put(fence);
+ break;
+ }
+
+ if (fence->seqno < seqno) {
+ pr_err("Reported an earlier fence.seqno:%lld for seqno:%d\n",
+ fence->seqno, seqno);
+ err = -EINVAL;
+ dma_fence_put(fence);
+ break;
+ }
+
+ dma_fence_put(fence);
+
+signal:
+ seqno = prandom_u32_max(data->fc.chain_length - 1);
+ dma_fence_signal(data->fc.fences[seqno]);
+ cond_resched();
+ }
+
+ if (atomic_dec_and_test(&data->children))
+ wake_up_var(&data->children);
+ return err;
+}
+
+static int find_race(void *arg)
+{
+ struct find_race data;
+ int ncpus = num_online_cpus();
+ struct task_struct **threads;
+ unsigned long count;
+ int err;
+ int i;
+
+ err = fence_chains_init(&data.fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ threads = kmalloc_array(ncpus, sizeof(*threads), GFP_KERNEL);
+ if (!threads) {
+ err = -ENOMEM;
+ goto err;
+ }
+
+ atomic_set(&data.children, 0);
+ for (i = 0; i < ncpus; i++) {
+ threads[i] = kthread_run(__find_race, &data, "dmabuf/%d", i);
+ if (IS_ERR(threads[i])) {
+ ncpus = i;
+ break;
+ }
+ atomic_inc(&data.children);
+ get_task_struct(threads[i]);
+ }
+
+ wait_var_event_timeout(&data.children,
+ !atomic_read(&data.children),
+ 5 * HZ);
+
+ for (i = 0; i < ncpus; i++) {
+ int ret;
+
+ ret = kthread_stop(threads[i]);
+ if (ret && !err)
+ err = ret;
+ put_task_struct(threads[i]);
+ }
+ kfree(threads);
+
+ count = 0;
+ for (i = 0; i < data.fc.chain_length; i++)
+ if (dma_fence_is_signaled(data.fc.fences[i]))
+ count++;
+ pr_info("Completed %lu cycles\n", count);
+
+err:
+ fence_chains_fini(&data.fc);
+ return err;
+}
+
+static int signal_forward(void *arg)
+{
+ struct fence_chains fc;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc);
+ if (err)
+ return err;
+
+ for (i = 0; i < fc.chain_length; i++) {
+ dma_fence_signal(fc.fences[i]);
+
+ if (!dma_fence_is_signaled(fc.chains[i])) {
+ pr_err("chain[%d] not signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+
+ if (i + 1 < fc.chain_length &&
+ dma_fence_is_signaled(fc.chains[i + 1])) {
+ pr_err("chain[%d] is signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int signal_backward(void *arg)
+{
+ struct fence_chains fc;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc);
+ if (err)
+ return err;
+
+ for (i = fc.chain_length; i--; ) {
+ dma_fence_signal(fc.fences[i]);
+
+ if (i > 0 && dma_fence_is_signaled(fc.chains[i])) {
+ pr_err("chain[%d] is signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+ for (i = 0; i < fc.chain_length; i++) {
+ if (!dma_fence_is_signaled(fc.chains[i])) {
+ pr_err("chain[%d] was not signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int __wait_fence_chains(void *arg)
+{
+ struct fence_chains *fc = arg;
+
+ if (dma_fence_wait(fc->tail, false))
+ return -EIO;
+
+ return 0;
+}
+
+static int wait_forward(void *arg)
+{
+ struct fence_chains fc;
+ struct task_struct *tsk;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ tsk = kthread_run(__wait_fence_chains, &fc, "dmabuf/wait");
+ if (IS_ERR(tsk)) {
+ err = PTR_ERR(tsk);
+ goto err;
+ }
+ get_task_struct(tsk);
+ yield_to(tsk, true);
+
+ for (i = 0; i < fc.chain_length; i++)
+ dma_fence_signal(fc.fences[i]);
+
+ err = kthread_stop(tsk);
+ put_task_struct(tsk);
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int wait_backward(void *arg)
+{
+ struct fence_chains fc;
+ struct task_struct *tsk;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ tsk = kthread_run(__wait_fence_chains, &fc, "dmabuf/wait");
+ if (IS_ERR(tsk)) {
+ err = PTR_ERR(tsk);
+ goto err;
+ }
+ get_task_struct(tsk);
+ yield_to(tsk, true);
+
+ for (i = fc.chain_length; i--; )
+ dma_fence_signal(fc.fences[i]);
+
+ err = kthread_stop(tsk);
+ put_task_struct(tsk);
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static void randomise_fences(struct fence_chains *fc)
+{
+ unsigned int count = fc->chain_length;
+
+ /* Fisher-Yates shuffle courtesy of Knuth */
+ while (--count) {
+ unsigned int swp;
+
+ swp = prandom_u32_max(count + 1);
+ if (swp == count)
+ continue;
+
+ swap(fc->fences[count], fc->fences[swp]);
+ }
+}
+
+static int wait_random(void *arg)
+{
+ struct fence_chains fc;
+ struct task_struct *tsk;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ randomise_fences(&fc);
+
+ tsk = kthread_run(__wait_fence_chains, &fc, "dmabuf/wait");
+ if (IS_ERR(tsk)) {
+ err = PTR_ERR(tsk);
+ goto err;
+ }
+ get_task_struct(tsk);
+ yield_to(tsk, true);
+
+ for (i = 0; i < fc.chain_length; i++)
+ dma_fence_signal(fc.fences[i]);
+
+ err = kthread_stop(tsk);
+ put_task_struct(tsk);
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+int dma_fence_chain(void)
+{
+ static const struct subtest tests[] = {
+ SUBTEST(sanitycheck),
+ SUBTEST(find_seqno),
+ SUBTEST(find_signaled),
+ SUBTEST(find_out_of_order),
+ SUBTEST(find_gap),
+ SUBTEST(find_race),
+ SUBTEST(signal_forward),
+ SUBTEST(signal_backward),
+ SUBTEST(wait_forward),
+ SUBTEST(wait_backward),
+ SUBTEST(wait_random),
+ };
+ int ret;
+
+ pr_info("sizeof(dma_fence_chain)=%zu\n",
+ sizeof(struct dma_fence_chain));
+
+ slab_fences = KMEM_CACHE(mock_fence,
+ SLAB_TYPESAFE_BY_RCU |
+ SLAB_HWCACHE_ALIGN);
+ if (!slab_fences)
+ return -ENOMEM;
+
+ ret = subtests(tests, NULL);
+
+ kmem_cache_destroy(slab_fences);
+ return ret;
+}
--
2.20.1
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [Intel-gfx] [PATCH 3/3] dma-buf: Exercise dma-fence-chain under selftests
@ 2020-04-09 11:01 ` Chris Wilson
0 siblings, 0 replies; 9+ messages in thread
From: Chris Wilson @ 2020-04-09 11:01 UTC (permalink / raw)
To: dri-devel; +Cc: intel-gfx, Chris Wilson
A few very simple testcases to exercise the dma-fence-chain API.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
---
drivers/dma-buf/Makefile | 3 +-
drivers/dma-buf/selftests.h | 1 +
drivers/dma-buf/st-dma-fence-chain.c | 713 +++++++++++++++++++++++++++
3 files changed, 716 insertions(+), 1 deletion(-)
create mode 100644 drivers/dma-buf/st-dma-fence-chain.c
diff --git a/drivers/dma-buf/Makefile b/drivers/dma-buf/Makefile
index 9c190026bfab..995e05f609ff 100644
--- a/drivers/dma-buf/Makefile
+++ b/drivers/dma-buf/Makefile
@@ -9,6 +9,7 @@ obj-$(CONFIG_UDMABUF) += udmabuf.o
dmabuf_selftests-y := \
selftest.o \
- st-dma-fence.o
+ st-dma-fence.o \
+ st-dma-fence-chain.o
obj-$(CONFIG_DMABUF_SELFTESTS) += dmabuf_selftests.o
diff --git a/drivers/dma-buf/selftests.h b/drivers/dma-buf/selftests.h
index 5320386f02e5..55918ef9adab 100644
--- a/drivers/dma-buf/selftests.h
+++ b/drivers/dma-buf/selftests.h
@@ -11,3 +11,4 @@
*/
selftest(sanitycheck, __sanitycheck__) /* keep first (igt selfcheck) */
selftest(dma_fence, dma_fence)
+selftest(dma_fence_chain, dma_fence_chain)
diff --git a/drivers/dma-buf/st-dma-fence-chain.c b/drivers/dma-buf/st-dma-fence-chain.c
new file mode 100644
index 000000000000..bd08ba67b03b
--- /dev/null
+++ b/drivers/dma-buf/st-dma-fence-chain.c
@@ -0,0 +1,713 @@
+// SPDX-License-Identifier: MIT
+
+/*
+ * Copyright © 2019 Intel Corporation
+ */
+
+#include <linux/delay.h>
+#include <linux/dma-fence.h>
+#include <linux/dma-fence-chain.h>
+#include <linux/kernel.h>
+#include <linux/kthread.h>
+#include <linux/mm.h>
+#include <linux/sched/signal.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/random.h>
+
+#include "selftest.h"
+
+static struct kmem_cache *slab_fences;
+
+static inline struct mock_fence {
+ struct dma_fence base;
+ spinlock_t lock;
+} *to_mock_fence(struct dma_fence *f) {
+ return container_of(f, struct mock_fence, base);
+}
+
+static const char *mock_name(struct dma_fence *f)
+{
+ return "mock";
+}
+
+static void mock_fence_release(struct dma_fence *f)
+{
+ kmem_cache_free(slab_fences, to_mock_fence(f));
+}
+
+static const struct dma_fence_ops mock_ops = {
+ .get_driver_name = mock_name,
+ .get_timeline_name = mock_name,
+ .release = mock_fence_release,
+};
+
+static struct dma_fence *mock_fence(void)
+{
+ struct mock_fence *f;
+
+ f = kmem_cache_alloc(slab_fences, GFP_KERNEL);
+ if (!f)
+ return NULL;
+
+ spin_lock_init(&f->lock);
+ dma_fence_init(&f->base, &mock_ops, &f->lock, 0, 0);
+
+ return &f->base;
+}
+
+static inline struct mock_chain {
+ struct dma_fence_chain base;
+} *to_mock_chain(struct dma_fence *f) {
+ return container_of(f, struct mock_chain, base.base);
+}
+
+static struct dma_fence *mock_chain(struct dma_fence *prev,
+ struct dma_fence *fence,
+ u64 seqno)
+{
+ struct mock_chain *f;
+
+ f = kmalloc(sizeof(*f), GFP_KERNEL);
+ if (!f)
+ return NULL;
+
+ dma_fence_chain_init(&f->base,
+ dma_fence_get(prev),
+ dma_fence_get(fence),
+ seqno);
+
+ return &f->base.base;
+}
+
+static int sanitycheck(void *arg)
+{
+ struct dma_fence *f, *chain;
+ int err = 0;
+
+ f = mock_fence();
+ if (!f)
+ return -ENOMEM;
+
+ chain = mock_chain(NULL, f, 1);
+ if (!chain)
+ err = -ENOMEM;
+
+ dma_fence_signal(f);
+ dma_fence_put(f);
+
+ dma_fence_put(chain);
+
+ return err;
+}
+
+struct fence_chains {
+ unsigned int chain_length;
+ struct dma_fence **fences;
+ struct dma_fence **chains;
+
+ struct dma_fence *tail;
+};
+
+static uint64_t seqno_inc(unsigned int i)
+{
+ return i + 1;
+}
+
+static int fence_chains_init(struct fence_chains *fc, unsigned int count,
+ uint64_t (*seqno_fn)(unsigned int))
+{
+ unsigned int i;
+ int err = 0;
+
+ fc->chains = kvmalloc_array(count, sizeof(*fc->chains),
+ GFP_KERNEL | __GFP_ZERO);
+ if (!fc->chains)
+ return -ENOMEM;
+
+ fc->fences = kvmalloc_array(count, sizeof(*fc->fences),
+ GFP_KERNEL | __GFP_ZERO);
+ if (!fc->fences) {
+ err = -ENOMEM;
+ goto err_chains;
+ }
+
+ fc->tail = NULL;
+ for (i = 0; i < count; i++) {
+ fc->fences[i] = mock_fence();
+ if (!fc->fences[i]) {
+ err = -ENOMEM;
+ goto unwind;
+ }
+
+ fc->chains[i] = mock_chain(fc->tail,
+ fc->fences[i],
+ seqno_fn(i));
+ if (!fc->chains[i]) {
+ err = -ENOMEM;
+ goto unwind;
+ }
+
+ fc->tail = fc->chains[i];
+ }
+
+ fc->chain_length = i;
+ return 0;
+
+unwind:
+ for (i = 0; i < count; i++) {
+ dma_fence_put(fc->fences[i]);
+ dma_fence_put(fc->chains[i]);
+ }
+ kvfree(fc->fences);
+err_chains:
+ kvfree(fc->chains);
+ return err;
+}
+
+static void fence_chains_fini(struct fence_chains *fc)
+{
+ unsigned int i;
+
+ for (i = 0; i < fc->chain_length; i++) {
+ dma_fence_signal(fc->fences[i]);
+ dma_fence_put(fc->fences[i]);
+ }
+ kvfree(fc->fences);
+
+ for (i = 0; i < fc->chain_length; i++)
+ dma_fence_put(fc->chains[i]);
+ kvfree(fc->chains);
+}
+
+static int find_seqno(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc);
+ if (err)
+ return err;
+
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 0);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno(0)!\n", err);
+ goto err;
+ }
+
+ for (i = 0; i < fc.chain_length; i++) {
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, i + 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno(%d:%d)!\n",
+ err, fc.chain_length + 1, i + 1);
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence reported by find_seqno(%d:%d)\n",
+ fc.chain_length + 1, i + 1);
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, i + 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Error reported for finding self\n");
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence reported by find self\n");
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, i + 2);
+ dma_fence_put(fence);
+ if (!err) {
+ pr_err("Error not reported for future fence: find_seqno(%d:%d)!\n",
+ i + 1, i + 2);
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, i);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Error reported for previous fence!\n");
+ goto err;
+ }
+ if (i > 0 && fence != fc.chains[i - 1]) {
+ pr_err("Incorrect fence reported by find_seqno(%d:%d)\n",
+ i + 1, i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int find_signaled(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+
+ err = fence_chains_init(&fc, 2, seqno_inc);
+ if (err)
+ return err;
+
+ dma_fence_signal(fc.fences[0]);
+
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno()!\n", err);
+ goto err;
+ }
+
+ if (fence && fence != fc.chains[0]) {
+ pr_err("Incorrect chain-fence.seqno:%lld reported for completed seqno:1\n",
+ fence->seqno);
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, 1);
+ dma_fence_put(fence);
+ if (err)
+ pr_err("Reported %d for finding self!\n", err);
+
+ err = -EINVAL;
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int find_out_of_order(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+
+ err = fence_chains_init(&fc, 3, seqno_inc);
+ if (err)
+ return err;
+
+ dma_fence_signal(fc.fences[1]);
+
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 2);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno()!\n", err);
+ goto err;
+ }
+
+ if (fence && fence != fc.chains[1]) {
+ pr_err("Incorrect chain-fence.seqno:%lld reported for completed seqno:2\n",
+ fence->seqno);
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, 2);
+ dma_fence_put(fence);
+ if (err)
+ pr_err("Reported %d for finding self!\n", err);
+
+ err = -EINVAL;
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static uint64_t seqno_inc2(unsigned int i)
+{
+ return 2 * i + 2;
+}
+
+static int find_gap(void *arg)
+{
+ struct fence_chains fc;
+ struct dma_fence *fence;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc2);
+ if (err)
+ return err;
+
+ for (i = 0; i < fc.chain_length; i++) {
+ fence = dma_fence_get(fc.tail);
+ err = dma_fence_chain_find_seqno(&fence, 2 * i + 1);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Reported %d for find_seqno(%d:%d)!\n",
+ err, fc.chain_length + 1, 2 * i + 1);
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence.seqno:%lld reported by find_seqno(%d:%d)\n",
+ fence->seqno,
+ fc.chain_length + 1,
+ 2 * i + 1);
+ err = -EINVAL;
+ goto err;
+ }
+
+ dma_fence_get(fence);
+ err = dma_fence_chain_find_seqno(&fence, 2 * i + 2);
+ dma_fence_put(fence);
+ if (err) {
+ pr_err("Error reported for finding self\n");
+ goto err;
+ }
+ if (fence != fc.chains[i]) {
+ pr_err("Incorrect fence reported by find self\n");
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+struct find_race {
+ struct fence_chains fc;
+ atomic_t children;
+};
+
+static int __find_race(void *arg)
+{
+ struct find_race *data = arg;
+ int err = 0;
+
+ while (!kthread_should_stop()) {
+ struct dma_fence *fence = dma_fence_get(data->fc.tail);
+ int seqno;
+
+ seqno = prandom_u32_max(data->fc.chain_length) + 1;
+
+ err = dma_fence_chain_find_seqno(&fence, seqno);
+ if (err) {
+ pr_err("Failed to find fence seqno:%d\n",
+ seqno);
+ dma_fence_put(fence);
+ break;
+ }
+ if (!fence)
+ goto signal;
+
+ err = dma_fence_chain_find_seqno(&fence, seqno);
+ if (err) {
+ pr_err("Reported an invalid fence for find-self:%d\n",
+ seqno);
+ dma_fence_put(fence);
+ break;
+ }
+
+ if (fence->seqno < seqno) {
+ pr_err("Reported an earlier fence.seqno:%lld for seqno:%d\n",
+ fence->seqno, seqno);
+ err = -EINVAL;
+ dma_fence_put(fence);
+ break;
+ }
+
+ dma_fence_put(fence);
+
+signal:
+ seqno = prandom_u32_max(data->fc.chain_length - 1);
+ dma_fence_signal(data->fc.fences[seqno]);
+ cond_resched();
+ }
+
+ if (atomic_dec_and_test(&data->children))
+ wake_up_var(&data->children);
+ return err;
+}
+
+static int find_race(void *arg)
+{
+ struct find_race data;
+ int ncpus = num_online_cpus();
+ struct task_struct **threads;
+ unsigned long count;
+ int err;
+ int i;
+
+ err = fence_chains_init(&data.fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ threads = kmalloc_array(ncpus, sizeof(*threads), GFP_KERNEL);
+ if (!threads) {
+ err = -ENOMEM;
+ goto err;
+ }
+
+ atomic_set(&data.children, 0);
+ for (i = 0; i < ncpus; i++) {
+ threads[i] = kthread_run(__find_race, &data, "dmabuf/%d", i);
+ if (IS_ERR(threads[i])) {
+ ncpus = i;
+ break;
+ }
+ atomic_inc(&data.children);
+ get_task_struct(threads[i]);
+ }
+
+ wait_var_event_timeout(&data.children,
+ !atomic_read(&data.children),
+ 5 * HZ);
+
+ for (i = 0; i < ncpus; i++) {
+ int ret;
+
+ ret = kthread_stop(threads[i]);
+ if (ret && !err)
+ err = ret;
+ put_task_struct(threads[i]);
+ }
+ kfree(threads);
+
+ count = 0;
+ for (i = 0; i < data.fc.chain_length; i++)
+ if (dma_fence_is_signaled(data.fc.fences[i]))
+ count++;
+ pr_info("Completed %lu cycles\n", count);
+
+err:
+ fence_chains_fini(&data.fc);
+ return err;
+}
+
+static int signal_forward(void *arg)
+{
+ struct fence_chains fc;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc);
+ if (err)
+ return err;
+
+ for (i = 0; i < fc.chain_length; i++) {
+ dma_fence_signal(fc.fences[i]);
+
+ if (!dma_fence_is_signaled(fc.chains[i])) {
+ pr_err("chain[%d] not signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+
+ if (i + 1 < fc.chain_length &&
+ dma_fence_is_signaled(fc.chains[i + 1])) {
+ pr_err("chain[%d] is signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int signal_backward(void *arg)
+{
+ struct fence_chains fc;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64, seqno_inc);
+ if (err)
+ return err;
+
+ for (i = fc.chain_length; i--; ) {
+ dma_fence_signal(fc.fences[i]);
+
+ if (i > 0 && dma_fence_is_signaled(fc.chains[i])) {
+ pr_err("chain[%d] is signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+ for (i = 0; i < fc.chain_length; i++) {
+ if (!dma_fence_is_signaled(fc.chains[i])) {
+ pr_err("chain[%d] was not signaled!\n", i);
+ err = -EINVAL;
+ goto err;
+ }
+ }
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int __wait_fence_chains(void *arg)
+{
+ struct fence_chains *fc = arg;
+
+ if (dma_fence_wait(fc->tail, false))
+ return -EIO;
+
+ return 0;
+}
+
+static int wait_forward(void *arg)
+{
+ struct fence_chains fc;
+ struct task_struct *tsk;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ tsk = kthread_run(__wait_fence_chains, &fc, "dmabuf/wait");
+ if (IS_ERR(tsk)) {
+ err = PTR_ERR(tsk);
+ goto err;
+ }
+ get_task_struct(tsk);
+ yield_to(tsk, true);
+
+ for (i = 0; i < fc.chain_length; i++)
+ dma_fence_signal(fc.fences[i]);
+
+ err = kthread_stop(tsk);
+ put_task_struct(tsk);
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static int wait_backward(void *arg)
+{
+ struct fence_chains fc;
+ struct task_struct *tsk;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ tsk = kthread_run(__wait_fence_chains, &fc, "dmabuf/wait");
+ if (IS_ERR(tsk)) {
+ err = PTR_ERR(tsk);
+ goto err;
+ }
+ get_task_struct(tsk);
+ yield_to(tsk, true);
+
+ for (i = fc.chain_length; i--; )
+ dma_fence_signal(fc.fences[i]);
+
+ err = kthread_stop(tsk);
+ put_task_struct(tsk);
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+static void randomise_fences(struct fence_chains *fc)
+{
+ unsigned int count = fc->chain_length;
+
+ /* Fisher-Yates shuffle courtesy of Knuth */
+ while (--count) {
+ unsigned int swp;
+
+ swp = prandom_u32_max(count + 1);
+ if (swp == count)
+ continue;
+
+ swap(fc->fences[count], fc->fences[swp]);
+ }
+}
+
+static int wait_random(void *arg)
+{
+ struct fence_chains fc;
+ struct task_struct *tsk;
+ int err;
+ int i;
+
+ err = fence_chains_init(&fc, 64 << 10, seqno_inc);
+ if (err)
+ return err;
+
+ randomise_fences(&fc);
+
+ tsk = kthread_run(__wait_fence_chains, &fc, "dmabuf/wait");
+ if (IS_ERR(tsk)) {
+ err = PTR_ERR(tsk);
+ goto err;
+ }
+ get_task_struct(tsk);
+ yield_to(tsk, true);
+
+ for (i = 0; i < fc.chain_length; i++)
+ dma_fence_signal(fc.fences[i]);
+
+ err = kthread_stop(tsk);
+ put_task_struct(tsk);
+
+err:
+ fence_chains_fini(&fc);
+ return err;
+}
+
+int dma_fence_chain(void)
+{
+ static const struct subtest tests[] = {
+ SUBTEST(sanitycheck),
+ SUBTEST(find_seqno),
+ SUBTEST(find_signaled),
+ SUBTEST(find_out_of_order),
+ SUBTEST(find_gap),
+ SUBTEST(find_race),
+ SUBTEST(signal_forward),
+ SUBTEST(signal_backward),
+ SUBTEST(wait_forward),
+ SUBTEST(wait_backward),
+ SUBTEST(wait_random),
+ };
+ int ret;
+
+ pr_info("sizeof(dma_fence_chain)=%zu\n",
+ sizeof(struct dma_fence_chain));
+
+ slab_fences = KMEM_CACHE(mock_fence,
+ SLAB_TYPESAFE_BY_RCU |
+ SLAB_HWCACHE_ALIGN);
+ if (!slab_fences)
+ return -ENOMEM;
+
+ ret = subtests(tests, NULL);
+
+ kmem_cache_destroy(slab_fences);
+ return ret;
+}
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for series starting with [1/3] dma-buf: Prettify typecasts for dma-fence-chain
2020-04-09 11:00 ` [Intel-gfx] " Chris Wilson
` (2 preceding siblings ...)
(?)
@ 2020-04-09 12:14 ` Patchwork
-1 siblings, 0 replies; 9+ messages in thread
From: Patchwork @ 2020-04-09 12:14 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [1/3] dma-buf: Prettify typecasts for dma-fence-chain
URL : https://patchwork.freedesktop.org/series/75743/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
5fe27cc9dd40 dma-buf: Prettify typecasts for dma-fence-chain
38f765418273 dma-buf: Report signaled links inside dma-fence-chain
9118ea143e62 dma-buf: Exercise dma-fence-chain under selftests
-:34: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#34:
new file mode 100644
-:62: CHECK:UNCOMMENTED_DEFINITION: spinlock_t definition without comment
#62: FILE: drivers/dma-buf/st-dma-fence-chain.c:24:
+ spinlock_t lock;
-:236: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'find_seqno', this function's name, in a string
#236: FILE: drivers/dma-buf/st-dma-fence-chain.c:198:
+ pr_err("Reported %d for find_seqno(0)!\n", err);
-:245: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'find_seqno', this function's name, in a string
#245: FILE: drivers/dma-buf/st-dma-fence-chain.c:207:
+ pr_err("Reported %d for find_seqno(%d:%d)!\n",
-:250: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'find_seqno', this function's name, in a string
#250: FILE: drivers/dma-buf/st-dma-fence-chain.c:212:
+ pr_err("Incorrect fence reported by find_seqno(%d:%d)\n",
-:273: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'find_seqno', this function's name, in a string
#273: FILE: drivers/dma-buf/st-dma-fence-chain.c:235:
+ pr_err("Error not reported for future fence: find_seqno(%d:%d)!\n",
-:287: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'find_seqno', this function's name, in a string
#287: FILE: drivers/dma-buf/st-dma-fence-chain.c:249:
+ pr_err("Incorrect fence reported by find_seqno(%d:%d)\n",
-:738: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'dma_fence_chain', this function's name, in a string
#738: FILE: drivers/dma-buf/st-dma-fence-chain.c:700:
+ pr_info("sizeof(dma_fence_chain)=%zu\n",
total: 0 errors, 7 warnings, 1 checks, 725 lines checked
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 9+ messages in thread
* [Intel-gfx] ✗ Fi.CI.BAT: failure for series starting with [1/3] dma-buf: Prettify typecasts for dma-fence-chain
2020-04-09 11:00 ` [Intel-gfx] " Chris Wilson
` (3 preceding siblings ...)
(?)
@ 2020-04-09 12:43 ` Patchwork
-1 siblings, 0 replies; 9+ messages in thread
From: Patchwork @ 2020-04-09 12:43 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [1/3] dma-buf: Prettify typecasts for dma-fence-chain
URL : https://patchwork.freedesktop.org/series/75743/
State : failure
== Summary ==
CI Bug Log - changes from CI_DRM_8283 -> Patchwork_17267
====================================================
Summary
-------
**FAILURE**
Serious unknown changes coming with Patchwork_17267 absolutely need to be
verified manually.
If you think the reported changes have nothing to do with the changes
introduced in Patchwork_17267, please notify your bug team to allow them
to document this new failure mode, which will reduce false positives in CI.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17267/index.html
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in Patchwork_17267:
### IGT changes ###
#### Possible regressions ####
* igt@i915_selftest@live@execlists:
- fi-icl-y: [PASS][1] -> [INCOMPLETE][2]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8283/fi-icl-y/igt@i915_selftest@live@execlists.html
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17267/fi-icl-y/igt@i915_selftest@live@execlists.html
New tests
---------
New tests have been introduced between CI_DRM_8283 and Patchwork_17267:
### New IGT tests (1) ###
* igt@dmabuf@all@dma_fence_chain:
- Statuses : 43 pass(s)
- Exec time: [7.47, 31.83] s
Known issues
------------
Here are the changes found in Patchwork_17267 that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@i915_selftest@live@requests:
- fi-icl-dsi: [PASS][3] -> [INCOMPLETE][4] ([i915#1531] / [i915#1581])
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8283/fi-icl-dsi/igt@i915_selftest@live@requests.html
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17267/fi-icl-dsi/igt@i915_selftest@live@requests.html
[i915#1531]: https://gitlab.freedesktop.org/drm/intel/issues/1531
[i915#1581]: https://gitlab.freedesktop.org/drm/intel/issues/1581
Participating hosts (52 -> 46)
------------------------------
Additional (1): fi-skl-6770hq
Missing (7): fi-ilk-m540 fi-hsw-4200u fi-byt-squawks fi-bsw-cyan fi-ctg-p8600 fi-kbl-7560u fi-byt-clapper
Build changes
-------------
* CI: CI-20190529 -> None
* Linux: CI_DRM_8283 -> Patchwork_17267
CI-20190529: 20190529
CI_DRM_8283: a6f4f55d343fea03e11e754b1094dda8cf2538ac @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_5585: 13c0be2fe8669fef08c0d1c44b147c43d1f53d2b @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_17267: 9118ea143e62d62cfa3d9e4a254b40dddcec845b @ git://anongit.freedesktop.org/gfx-ci/linux
== Linux commits ==
9118ea143e62 dma-buf: Exercise dma-fence-chain under selftests
38f765418273 dma-buf: Report signaled links inside dma-fence-chain
5fe27cc9dd40 dma-buf: Prettify typecasts for dma-fence-chain
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17267/index.html
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 1/3] dma-buf: Prettify typecasts for dma-fence-chain
@ 2019-10-20 16:43 Chris Wilson
0 siblings, 0 replies; 9+ messages in thread
From: Chris Wilson @ 2019-10-20 16:43 UTC (permalink / raw)
To: dri-devel; +Cc: intel-gfx
Inside dma-fence-chain, we use a cmpxchg on an RCU-protected pointer. To
avoid the sparse warning for using the RCU pointer directly, we have to
cast away the __rcu annotation. However, we don't need to use void*
everywhere and can stick to the dma_fence*.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/dma-buf/dma-fence-chain.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/dma-buf/dma-fence-chain.c b/drivers/dma-buf/dma-fence-chain.c
index 44a741677d25..3d123502ff12 100644
--- a/drivers/dma-buf/dma-fence-chain.c
+++ b/drivers/dma-buf/dma-fence-chain.c
@@ -62,7 +62,8 @@ struct dma_fence *dma_fence_chain_walk(struct dma_fence *fence)
replacement = NULL;
}
- tmp = cmpxchg((void **)&chain->prev, (void *)prev, (void *)replacement);
+ tmp = cmpxchg((struct dma_fence __force **)&chain->prev,
+ prev, replacement);
if (tmp == prev)
dma_fence_put(tmp);
else
--
2.24.0.rc0
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
^ permalink raw reply related [flat|nested] 9+ messages in thread
end of thread, other threads:[~2020-04-09 12:43 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-04-09 11:00 [PATCH 1/3] dma-buf: Prettify typecasts for dma-fence-chain Chris Wilson
2020-04-09 11:00 ` [Intel-gfx] " Chris Wilson
2020-04-09 11:01 ` [PATCH 2/3] dma-buf: Report signaled links inside dma-fence-chain Chris Wilson
2020-04-09 11:01 ` [Intel-gfx] " Chris Wilson
2020-04-09 11:01 ` [PATCH 3/3] dma-buf: Exercise dma-fence-chain under selftests Chris Wilson
2020-04-09 11:01 ` [Intel-gfx] " Chris Wilson
2020-04-09 12:14 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for series starting with [1/3] dma-buf: Prettify typecasts for dma-fence-chain Patchwork
2020-04-09 12:43 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork
-- strict thread matches above, loose matches on Subject: below --
2019-10-20 16:43 [PATCH 1/3] " Chris Wilson
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.