[PATCH 1/2] soc: qcom: aoss: Don't wait for IRQ if we might be in suspend/resume noirq

From: Douglas Anderson
Date: Wed Aug 05 2020 - 12:30:50 EST

Running suspend/resume tests on a sc7180-based board with a modem I
found that both system suspend and system resume would hang for 1
second. These messages indicate where:

genpd genpd:0:4080000.remoteproc: calling genpd_suspend_noirq+0x0/0x2c @ 18659, parent: none
genpd genpd:0:4080000.remoteproc: genpd_suspend_noirq+0x0/0x2c returned 0 after 987917 usecs

Adding a printout, I found that we were working with the power domain
where "res->pd.name" was "modem".

I found that we were hanging on the wait_event_interruptible_timeout()
call in qmp_send(). Specifically we'd wait for the whole 1 second
timeout to hit, then we'd notice that our condition was true and would
continue on our merry way. Sure enough, I could confirm that
wait_event_interruptible_timeout() was returning "1" which indicates
that the condition evaluated to true and we also timed out.

Dumping the stack at the time of the failure made the problem clear.
Specifically the stack looked like:
As you can see we're running from the "noirq" callback. Looking at
what was supposed to wake us up, it was qmp_intr() (our IRQ handler).

I believe that the correct fix here is to assume that our power_off /
power_on functions might be called at "noirq" time and just always
poll if we're called via that path. Other paths can continue to wait
for the IRQ.

Fixes: 2209481409b7 ("soc: qcom: Add AOSS QMP driver")
Signed-off-by: Douglas Anderson <dianders@xxxxxxxxxxxx>
This problem was observed on the Chrome OS 5.4 tree which has some
extra patches in it compared to mainline. The top of the current
Qualcomm tree didn't have the delay, but that's probably because
everything isn't fully enabled there yet. I at least confirmed that
this patch doesn't actually _break_ anything on mainline, though.

drivers/soc/qcom/qcom_aoss.c | 26 +++++++++++++++++++-------
1 file changed, 19 insertions(+), 7 deletions(-)

diff --git a/drivers/soc/qcom/qcom_aoss.c b/drivers/soc/qcom/qcom_aoss.c
index ed2c687c16b3..818cdf74a267 100644
--- a/drivers/soc/qcom/qcom_aoss.c
+++ b/drivers/soc/qcom/qcom_aoss.c
@@ -6,6 +6,7 @@
#include <linux/clk-provider.h>
#include <linux/interrupt.h>
#include <linux/io.h>
+#include <linux/iopoll.h>
#include <linux/mailbox_client.h>
#include <linux/module.h>
#include <linux/platform_device.h>
@@ -215,6 +216,8 @@ static bool qmp_message_empty(struct qmp *qmp)
* @qmp: qmp context
* @data: message to be sent
* @len: length of the message
+ * @noirq: If true we might have been called from the "noirq" suspend/resume
+ * callbacks, so fall back to polling mode for waiting for completion.
* Transmit @data to AOSS and wait for the AOSS to acknowledge the message.
* @len must be a multiple of 4 and not longer than the mailbox size. Access is
@@ -222,11 +225,12 @@ static bool qmp_message_empty(struct qmp *qmp)
* Return: 0 on success, negative errno on failure
-static int qmp_send(struct qmp *qmp, const void *data, size_t len)
+static int qmp_send(struct qmp *qmp, const void *data, size_t len, bool noirq)
long time_left;
size_t tlen;
int ret;
+ bool is_empty;

if (WARN_ON(len + sizeof(u32) > qmp->size))
return -EINVAL;
@@ -245,8 +249,16 @@ static int qmp_send(struct qmp *qmp, const void *data, size_t len)
tlen = readl(qmp->msgram + qmp->offset);

- time_left = wait_event_interruptible_timeout(qmp->event,
- qmp_message_empty(qmp), HZ);
+ /*
+ * We may be called from a suspend/resume "noirq" context. In such
+ * a case we have no choice but to poll.
+ */
+ if (noirq)
+ time_left = readx_poll_timeout_atomic(qmp_message_empty, qmp,
+ is_empty, is_empty, 1U, 1000000U);
+ else
+ time_left = wait_event_interruptible_timeout(qmp->event,
+ qmp_message_empty(qmp), HZ);
if (!time_left) {
dev_err(qmp->dev, "ucore did not ack channel\n");
@@ -267,7 +279,7 @@ static int qmp_qdss_clk_prepare(struct clk_hw *hw)
static const char buf[QMP_MSG_LEN] = "{class: clock, res: qdss, val: 1}";
struct qmp *qmp = container_of(hw, struct qmp, qdss_clk);

- return qmp_send(qmp, buf, sizeof(buf));
+ return qmp_send(qmp, buf, sizeof(buf), false);

static void qmp_qdss_clk_unprepare(struct clk_hw *hw)
@@ -275,7 +287,7 @@ static void qmp_qdss_clk_unprepare(struct clk_hw *hw)
static const char buf[QMP_MSG_LEN] = "{class: clock, res: qdss, val: 0}";
struct qmp *qmp = container_of(hw, struct qmp, qdss_clk);

- qmp_send(qmp, buf, sizeof(buf));
+ qmp_send(qmp, buf, sizeof(buf), false);

static const struct clk_ops qmp_qdss_clk_ops = {
@@ -321,7 +333,7 @@ static int qmp_pd_power_toggle(struct qmp_pd *res, bool enable)
snprintf(buf, sizeof(buf),
"{class: image, res: load_state, name: %s, val: %s}",
res->pd.name, enable ? "on" : "off");
- return qmp_send(res->qmp, buf, sizeof(buf));
+ return qmp_send(res->qmp, buf, sizeof(buf), true);

static int qmp_pd_power_on(struct generic_pm_domain *domain)
@@ -438,7 +450,7 @@ static int qmp_cdev_set_cur_state(struct thermal_cooling_device *cdev,
cdev_state ? "on" : "off");

- ret = qmp_send(qmp_cdev->qmp, buf, sizeof(buf));
+ ret = qmp_send(qmp_cdev->qmp, buf, sizeof(buf), false);

if (!ret)
qmp_cdev->state = cdev_state;