[Intel-gfx] [PATCH] mei: Abort writes if incomplete after 1s
Chris Wilson
chris at chris-wilson.co.uk
Tue Jul 23 11:19:13 UTC 2019
During i915 unload, it appears that it may get stuck waiting on a
workqueue being hogged by mei:
<7> [212.666912] i915 0000:00:02.0: [drm:drm_client_release] drm_fb_helper
<3> [308.544943] INFO: task i915_module_loa:2612 blocked for more than 61 seconds.
<3> [308.545047] Tainted: G U W 5.3.0-rc1-CI-CI_DRM_6537+ #1
<3> [308.545085] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
<6> [308.545128] i915_module_loa D13256 2612 960 0x00004004
<4> [308.545137] Call Trace:
<4> [308.545150] ? __schedule+0x326/0x890
<4> [308.545159] ? wait_for_common+0x116/0x1f0
<4> [308.545164] schedule+0x2b/0xb0
<4> [308.545169] schedule_timeout+0x219/0x3c0
<4> [308.545176] ? wait_for_common+0x132/0x1f0
<4> [308.545183] ? _raw_spin_unlock_irq+0x24/0x30
<4> [308.545189] ? wait_for_common+0x116/0x1f0
<4> [308.545193] wait_for_common+0x13a/0x1f0
<4> [308.545200] ? wake_up_q+0x80/0x80
<4> [308.545209] flush_workqueue+0x19d/0x540
<4> [308.545334] ? intel_modeset_driver_remove+0xb3/0x140 [i915]
<4> [308.545407] intel_modeset_driver_remove+0xb3/0x140 [i915]
<4> [308.545464] i915_driver_remove+0xae/0x110 [i915]
<4> [308.545522] i915_pci_remove+0x19/0x30 [i915]
<4> [308.545529] pci_device_remove+0x36/0xb0
<6> [308.565422] Showing busy workqueues and worker pools:
<6> [308.565425] workqueue events: flags=0x0
<6> [308.565572] pwq 2: cpus=1 node=0 flags=0x0 nice=0 active=2/256
<6> [308.565635] in-flight: 441:mei_cl_bus_rescan_work [mei]
<6> [308.565641] pending: dbs_work_handler
<6> [308.565686] pool 2: cpus=1 node=0 flags=0x0 nice=0 hung=0s workers=5 idle: 2248 21 17 169
<6> [308.553788] Workqueue: events mei_cl_bus_rescan_work [mei]
<4> [308.553792] Call Trace:
<4> [308.553799] ? __schedule+0x326/0x890
<4> [308.553808] schedule+0x2b/0xb0
<4> [308.553815] mei_cl_write+0x430/0x5a0 [mei]
<4> [308.553820] ? __kmalloc+0x2b6/0x330
<4> [308.553824] ? wait_woken+0xa0/0xa0
<4> [308.553835] __mei_cl_send+0x1f4/0x240 [mei]
<4> [308.553848] mei_mkhi_fix+0x91/0x280 [mei]
<4> [308.553859] mei_cl_bus_dev_fixup+0xba/0x100 [mei]
<4> [308.553868] ? device_add+0x156/0x670
<4> [308.553889] ? mei_cl_bus_rescan_work+0x1bc/0x350 [mei]
<4> [308.553896] mei_cl_bus_rescan_work+0x1bc/0x350 [mei]
<4> [308.553905] process_one_work+0x245/0x5f0
<4> [308.553915] worker_thread+0x37/0x380
<4> [308.553921] ? process_one_work+0x5f0/0x5f0
<4> [308.553924] kthread+0x119/0x130
<4> [308.553928] ? kthread_park+0xa0/0xa0
<4> [308.553934] ret_from_fork+0x3a/0x50
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Alexander Usyskin <alexander.usyskin at intel.com>
Cc: Tomas Winkler <tomas.winkler at intel.com>
---
drivers/misc/mei/bus.c | 9 +++++++--
drivers/misc/mei/client.c | 5 +++--
drivers/misc/mei/main.c | 18 ++++++++++++++----
3 files changed, 24 insertions(+), 8 deletions(-)
diff --git a/drivers/misc/mei/bus.c b/drivers/misc/mei/bus.c
index 985bd4fd3328..5b2db77d48db 100644
--- a/drivers/misc/mei/bus.c
+++ b/drivers/misc/mei/bus.c
@@ -66,9 +66,10 @@ ssize_t __mei_cl_send(struct mei_cl *cl, u8 *buf, size_t length,
while (cl->tx_cb_queued >= bus->tx_queue_limit) {
mutex_unlock(&bus->device_lock);
- rets = wait_event_interruptible(cl->tx_wait,
+ rets = wait_event_interruptible_timeout(cl->tx_wait,
cl->writing_state == MEI_WRITE_COMPLETE ||
- (!mei_cl_is_connected(cl)));
+ !mei_cl_is_connected(cl),
+ HZ);
mutex_lock(&bus->device_lock);
if (rets) {
if (signal_pending(current))
@@ -79,6 +80,10 @@ ssize_t __mei_cl_send(struct mei_cl *cl, u8 *buf, size_t length,
rets = -ENODEV;
goto out;
}
+ if (cl->writing_state != MEI_WRITE_COMPLETE) {
+ rets = -EFAULT;
+ goto out;
+ }
}
cb = mei_cl_alloc_cb(cl, length, MEI_FOP_WRITE, NULL);
diff --git a/drivers/misc/mei/client.c b/drivers/misc/mei/client.c
index 1e3edbbacb1e..e7acc8aa9b15 100644
--- a/drivers/misc/mei/client.c
+++ b/drivers/misc/mei/client.c
@@ -1767,9 +1767,10 @@ ssize_t mei_cl_write(struct mei_cl *cl, struct mei_cl_cb *cb)
if (blocking && cl->writing_state != MEI_WRITE_COMPLETE) {
mutex_unlock(&dev->device_lock);
- rets = wait_event_interruptible(cl->tx_wait,
+ rets = wait_event_interruptible_timeout(cl->tx_wait,
cl->writing_state == MEI_WRITE_COMPLETE ||
- (!mei_cl_is_connected(cl)));
+ !mei_cl_is_connected(cl),
+ HZ);
mutex_lock(&dev->device_lock);
/* wait_event_interruptible returns -ERESTARTSYS */
if (rets) {
diff --git a/drivers/misc/mei/main.c b/drivers/misc/mei/main.c
index f894d1f8a53e..0eb7bfd89a90 100644
--- a/drivers/misc/mei/main.c
+++ b/drivers/misc/mei/main.c
@@ -294,9 +294,10 @@ static ssize_t mei_write(struct file *file, const char __user *ubuf,
goto out;
}
mutex_unlock(&dev->device_lock);
- rets = wait_event_interruptible(cl->tx_wait,
+ rets = wait_event_interruptible_timeout(cl->tx_wait,
cl->writing_state == MEI_WRITE_COMPLETE ||
- (!mei_cl_is_connected(cl)));
+ !mei_cl_is_connected(cl),
+ HZ);
mutex_lock(&dev->device_lock);
if (rets) {
if (signal_pending(current))
@@ -307,6 +308,10 @@ static ssize_t mei_write(struct file *file, const char __user *ubuf,
rets = -ENODEV;
goto out;
}
+ if (cl->writing_state != MEI_WRITE_COMPLETE) {
+ rets = -EFAULT;
+ goto out;
+ }
}
cb = mei_cl_alloc_cb(cl, length, MEI_FOP_WRITE, file);
@@ -658,9 +663,10 @@ static int mei_fsync(struct file *fp, loff_t start, loff_t end, int datasync)
while (mei_cl_is_write_queued(cl)) {
mutex_unlock(&dev->device_lock);
- rets = wait_event_interruptible(cl->tx_wait,
+ rets = wait_event_interruptible_timeout(cl->tx_wait,
cl->writing_state == MEI_WRITE_COMPLETE ||
- !mei_cl_is_connected(cl));
+ !mei_cl_is_connected(cl),
+ HZ);
mutex_lock(&dev->device_lock);
if (rets) {
if (signal_pending(current))
@@ -671,6 +677,10 @@ static int mei_fsync(struct file *fp, loff_t start, loff_t end, int datasync)
rets = -ENODEV;
goto out;
}
+ if (cl->writing_state != MEI_WRITE_COMPLETE) {
+ rets = -EFAULT;
+ goto out;
+ }
}
rets = 0;
out:
--
2.22.0
More information about the Intel-gfx
mailing list