msm: kgsl: Don't use 'wait_timeout' in the idle functions
There are two distinct parts of the code that may need to loop waiting
for the GPU to complete a task: waiting for a timestamp and waiting
for the entire core to go idle. Waiting for a timestamp technically
doesn't need a timeout since the only downside is a process that sleeps
forever with an interruptible timeout. Waiting for the core to go idle
is more problematic because it is a busy wait and it is the last point
we can safely detect a GPU hang.
Beacuse we can (and will) not use a timeout in wait for timestamp, we
need to institute a new timeout value to be used in idle. Nowhere the
idle function is called uses a custom value for the timeout, so remove
that parameter from the calls and use a static timeout value in the
core specific functions.
Change-Id: Ic0dedbad9ecd2044c34e4cec551dc7f53b253f3d
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
diff --git a/drivers/gpu/msm/adreno.c b/drivers/gpu/msm/adreno.c
index 6a894c8..1963fcd 100644
--- a/drivers/gpu/msm/adreno.c
+++ b/drivers/gpu/msm/adreno.c
@@ -1481,7 +1481,7 @@
* them to pass */
adreno_ringbuffer_restore(rb, rec_data->bad_rb_buffer,
rec_data->bad_rb_size);
- idle_ret = adreno_idle(device, KGSL_TIMEOUT_DEFAULT);
+ idle_ret = adreno_idle(device);
if (idle_ret) {
ret = adreno_stop(device);
if (ret) {
@@ -1524,7 +1524,7 @@
if (ret || !rec_data->bad_rb_size) {
adreno_ringbuffer_restore(rb, rec_data->rb_buffer,
rec_data->rb_size);
- ret = adreno_idle(device, KGSL_TIMEOUT_DEFAULT);
+ ret = adreno_idle(device);
if (ret) {
/* If we fail here we can try to invalidate another
* context and try recovering again */
@@ -1806,61 +1806,74 @@
adreno_regwrite(device, REG_CP_RB_WPTR, adreno_dev->ringbuffer.wptr);
}
-/* Caller must hold the device mutex. */
-int adreno_idle(struct kgsl_device *device, unsigned int timeout)
+static int adreno_ringbuffer_drain(struct kgsl_device *device,
+ unsigned int *regs)
{
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
struct adreno_ringbuffer *rb = &adreno_dev->ringbuffer;
+ unsigned long wait;
+ unsigned long timeout = jiffies + msecs_to_jiffies(ADRENO_IDLE_TIMEOUT);
+
+ if (!(rb->flags & KGSL_FLAGS_STARTED))
+ return 0;
+
+ /*
+ * The first time into the loop, wait for 100 msecs and kick wptr again
+ * to ensure that the hardware has updated correctly. After that, kick
+ * it periodically every KGSL_TIMEOUT_PART msecs until the timeout
+ * expires
+ */
+
+ wait = jiffies + msecs_to_jiffies(100);
+
+ adreno_poke(device);
+
+ do {
+ if (time_after(jiffies, wait)) {
+ adreno_poke(device);
+
+ /* Check to see if the core is hung */
+ if (adreno_hang_detect(device, regs))
+ return -ETIMEDOUT;
+
+ wait = jiffies + msecs_to_jiffies(KGSL_TIMEOUT_PART);
+ }
+ GSL_RB_GET_READPTR(rb, &rb->rptr);
+
+ if (time_after(jiffies, timeout)) {
+ KGSL_DRV_ERR(device, "rptr: %x, wptr: %x\n",
+ rb->rptr, rb->wptr);
+ return -ETIMEDOUT;
+ }
+ } while (rb->rptr != rb->wptr);
+
+ return 0;
+}
+
+/* Caller must hold the device mutex. */
+int adreno_idle(struct kgsl_device *device)
+{
+ struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
unsigned int rbbm_status;
- unsigned long wait_timeout =
- msecs_to_jiffies(adreno_dev->wait_timeout);
unsigned long wait_time;
unsigned long wait_time_part;
- unsigned int msecs;
- unsigned int msecs_first;
- unsigned int msecs_part = KGSL_TIMEOUT_PART;
unsigned int prev_reg_val[hang_detect_regs_count];
memset(prev_reg_val, 0, sizeof(prev_reg_val));
- /* Restrict timeout value between adreno_dev->wait_timeout and 0 */
- if ((timeout == 0) || (timeout > adreno_dev->wait_timeout))
- msecs = adreno_dev->wait_timeout;
- else
- msecs = timeout;
-
kgsl_cffdump_regpoll(device->id,
adreno_dev->gpudev->reg_rbbm_status << 2,
0x00000000, 0x80000000);
- /* first, wait until the CP has consumed all the commands in
- * the ring buffer
- */
+
retry:
- if (rb->flags & KGSL_FLAGS_STARTED) {
- msecs_first = (msecs <= 100) ? ((msecs + 4) / 5) : 100;
- wait_time = jiffies + wait_timeout;
- wait_time_part = jiffies + msecs_to_jiffies(msecs_first);
- adreno_poke(device);
- do {
- if (time_after(jiffies, wait_time_part)) {
- adreno_poke(device);
- wait_time_part = jiffies +
- msecs_to_jiffies(msecs_part);
- if ((adreno_hang_detect(device, prev_reg_val)))
- goto err;
- }
- GSL_RB_GET_READPTR(rb, &rb->rptr);
- if (time_after(jiffies, wait_time)) {
- KGSL_DRV_ERR(device, "rptr: %x, wptr: %x\n",
- rb->rptr, rb->wptr);
- goto err;
- }
- } while (rb->rptr != rb->wptr);
- }
+ /* First, wait for the ringbuffer to drain */
+ if (adreno_ringbuffer_drain(device, prev_reg_val))
+ goto err;
/* now, wait for the GPU to finish its operations */
- wait_time = jiffies + wait_timeout;
- wait_time_part = jiffies + msecs_to_jiffies(msecs_part);
+ wait_time = jiffies + ADRENO_IDLE_TIMEOUT;
+ wait_time_part = jiffies + msecs_to_jiffies(KGSL_TIMEOUT_PART);
+
while (time_before(jiffies, wait_time)) {
adreno_regread(device, adreno_dev->gpudev->reg_rbbm_status,
&rbbm_status);
@@ -1876,7 +1889,7 @@
*/
if (time_after(jiffies, wait_time_part)) {
wait_time_part = jiffies +
- msecs_to_jiffies(msecs_part);
+ msecs_to_jiffies(KGSL_TIMEOUT_PART);
if ((adreno_hang_detect(device, prev_reg_val)))
goto err;
}
@@ -1887,7 +1900,7 @@
KGSL_DRV_ERR(device, "spun too long waiting for RB to idle\n");
if (KGSL_STATE_DUMP_AND_RECOVER != device->state &&
!adreno_dump_and_recover(device)) {
- wait_time = jiffies + wait_timeout;
+ wait_time = jiffies + ADRENO_IDLE_TIMEOUT;
goto retry;
}
return -ETIMEDOUT;
@@ -1934,7 +1947,7 @@
/* switch to NULL ctxt */
if (adreno_dev->drawctxt_active != NULL) {
adreno_drawctxt_switch(adreno_dev, NULL, 0);
- status = adreno_idle(device, KGSL_TIMEOUT_DEFAULT);
+ status = adreno_idle(device);
}
return status;