[PATCH 0/7] ath10k: recovery improvements 2014-05-09

[PATCH 0/7] ath10k: recovery improvements 2014-05-09

[Michal Kazior]

Hi,

This depends on my `ath10k: recovery fixes
2014-05-09`.

Hopefully the split makes it a little saner to
review.

Please note: `ath10k: make wmi timeout trigger hw
restart` effectively hides the mgmt tx / htc
credit starvation issue that was discussed
recently. My intention is to make the driver more
robust and as a side effect it makes the
starvation issue less painful. If desired so
the patch can be easily ignored in the series as
nothing depends on it.


Michal Kazior (7):
  ath10k: move restart_work call
  ath10k: introduce core restart symbol
  ath10k: drain wmi quickly upon hw restart
  ath10k: wake up everything on core restart
  ath10k: reduce timeout values
  ath10k: make wmi timeout trigger hw restart
  ath10k: remove old unused defines

 drivers/net/wireless/ath/ath10k/core.c | 27 +++++++++++++++++++++++++--
 drivers/net/wireless/ath/ath10k/core.h |  7 +++++--
 drivers/net/wireless/ath/ath10k/htt.c  |  2 +-
 drivers/net/wireless/ath/ath10k/mac.c  |  8 +++-----
 drivers/net/wireless/ath/ath10k/pci.c  |  3 +--
 drivers/net/wireless/ath/ath10k/txrx.c |  2 +-
 drivers/net/wireless/ath/ath10k/wmi.c  | 17 ++++++++++++++++-
 drivers/net/wireless/ath/ath10k/wmi.h  |  4 ++--
 8 files changed, 54 insertions(+), 16 deletions(-)

-- 
1.8.5.3

[PATCH 1/7] ath10k: move restart_work call

[Michal Kazior]

That call really should not be in
ath10k_pci_hif_dump_area(). It's very misleading
to have a function name which implies just
printing stuff but actually also restarts
firmware.

Suggested-By: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
---
 drivers/net/wireless/ath/ath10k/pci.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index 66b1f30..438d3cf 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -864,8 +864,6 @@ static void ath10k_pci_hif_dump_area(struct ath10k *ar)
 			   reg_dump_values[i + 1],
 			   reg_dump_values[i + 2],
 			   reg_dump_values[i + 3]);
-
-	queue_work(ar->workqueue, &ar->restart_work);
 }
 
 static void ath10k_pci_hif_send_complete_check(struct ath10k *ar, u8 pipe,
@@ -1790,6 +1788,7 @@ static void ath10k_pci_fw_interrupt_handler(struct ath10k *ar)
 
 		if (ar_pci->started) {
 			ath10k_pci_hif_dump_area(ar);
+			queue_work(ar->workqueue, &ar->restart_work);
 		} else {
 			/*
 			 * Probable Target failure before we're prepared
-- 
1.8.5.3

[PATCH 2/7] ath10k: introduce core restart symbol

[Michal Kazior]

This makes it easier to have an extra preparation
code before restart work is scheduled.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
---
 drivers/net/wireless/ath/ath10k/core.c | 10 ++++++++--
 drivers/net/wireless/ath/ath10k/core.h |  1 +
 drivers/net/wireless/ath/ath10k/pci.c  |  2 +-
 3 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
index 7034c72..acc22cc 100644
--- a/drivers/net/wireless/ath/ath10k/core.c
+++ b/drivers/net/wireless/ath/ath10k/core.c
@@ -672,7 +672,7 @@ static int ath10k_init_hw_params(struct ath10k *ar)
 	return 0;
 }
 
-static void ath10k_core_restart(struct work_struct *work)
+static void ath10k_core_restart_work(struct work_struct *work)
 {
 	struct ath10k *ar = container_of(work, struct ath10k, restart_work);
 
@@ -704,6 +704,12 @@ static void ath10k_core_restart(struct work_struct *work)
 	mutex_unlock(&ar->conf_mutex);
 }
 
+void ath10k_core_restart(struct ath10k *ar)
+{
+	queue_work(ar->workqueue, &ar->restart_work);
+}
+EXPORT_SYMBOL(ath10k_core_restart);
+
 struct ath10k *ath10k_core_create(void *hif_priv, struct device *dev,
 				  const struct ath10k_hif_ops *hif_ops)
 {
@@ -749,7 +755,7 @@ struct ath10k *ath10k_core_create(void *hif_priv, struct device *dev,
 	INIT_WORK(&ar->wmi_mgmt_tx_work, ath10k_mgmt_over_wmi_tx_work);
 	skb_queue_head_init(&ar->wmi_mgmt_tx_queue);
 
-	INIT_WORK(&ar->restart_work, ath10k_core_restart);
+	INIT_WORK(&ar->restart_work, ath10k_core_restart_work);
 
 	return ar;
 
diff --git a/drivers/net/wireless/ath/ath10k/core.h b/drivers/net/wireless/ath/ath10k/core.h
index 2c1dfd7..170094d 100644
--- a/drivers/net/wireless/ath/ath10k/core.h
+++ b/drivers/net/wireless/ath/ath10k/core.h
@@ -494,5 +494,6 @@ int ath10k_wait_for_suspend(struct ath10k *ar, u32 suspend_opt);
 void ath10k_core_stop(struct ath10k *ar);
 int ath10k_core_register(struct ath10k *ar, u32 chip_id);
 void ath10k_core_unregister(struct ath10k *ar);
+void ath10k_core_restart(struct ath10k *ar);
 
 #endif /* _CORE_H_ */
diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index 438d3cf..7be5784 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -1788,7 +1788,7 @@ static void ath10k_pci_fw_interrupt_handler(struct ath10k *ar)
 
 		if (ar_pci->started) {
 			ath10k_pci_hif_dump_area(ar);
-			queue_work(ar->workqueue, &ar->restart_work);
+			ath10k_core_restart(ar);
 		} else {
 			/*
 			 * Probable Target failure before we're prepared
-- 
1.8.5.3

[PATCH 3/7] ath10k: drain wmi quickly upon hw restart

[Michal Kazior]

Once restart is started it doesn't make much sense
to process pending requests. Just drop all WMI
commands that are in progress.

This speeds up hw restart because conf_mutex can
be held in some cases while a crash happens. In
that case WMI tx credits are no replenished and 3
second timeout for each WMI command happens. This
could add up to quite a lot.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
---
 drivers/net/wireless/ath/ath10k/core.c |  5 +++++
 drivers/net/wireless/ath/ath10k/core.h |  3 +++
 drivers/net/wireless/ath/ath10k/wmi.c  | 12 +++++++++++-
 3 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
index acc22cc..3f4d28f 100644
--- a/drivers/net/wireless/ath/ath10k/core.c
+++ b/drivers/net/wireless/ath/ath10k/core.c
@@ -706,6 +706,11 @@ static void ath10k_core_restart_work(struct work_struct *work)
 
 void ath10k_core_restart(struct ath10k *ar)
 {
+	spin_lock_bh(&ar->data_lock);
+	ar->wmi.drop = true;
+	wake_up(&ar->wmi.tx_credits_wq);
+	spin_unlock_bh(&ar->data_lock);
+
 	queue_work(ar->workqueue, &ar->restart_work);
 }
 EXPORT_SYMBOL(ath10k_core_restart);
diff --git a/drivers/net/wireless/ath/ath10k/core.h b/drivers/net/wireless/ath/ath10k/core.h
index 170094d..074cb41 100644
--- a/drivers/net/wireless/ath/ath10k/core.h
+++ b/drivers/net/wireless/ath/ath10k/core.h
@@ -113,6 +113,9 @@ struct ath10k_wmi {
 
 	u32 num_mem_chunks;
 	struct ath10k_mem_chunk mem_chunks[ATH10K_MAX_MEM_REQS];
+
+	/* protected by ar->data_lock. used to speed up hw restart */
+	bool drop;
 };
 
 struct ath10k_peer_stat {
diff --git a/drivers/net/wireless/ath/ath10k/wmi.c b/drivers/net/wireless/ath/ath10k/wmi.c
index 72cc4f2..159d744 100644
--- a/drivers/net/wireless/ath/ath10k/wmi.c
+++ b/drivers/net/wireless/ath/ath10k/wmi.c
@@ -621,7 +621,13 @@ static int ath10k_wmi_cmd_send(struct ath10k *ar, struct sk_buff *skb,
 		/* try to send pending beacons first. they take priority */
 		ath10k_wmi_tx_beacons_nowait(ar);
 
-		ret = ath10k_wmi_cmd_send_nowait(ar, skb, cmd_id);
+		spin_lock_bh(&ar->data_lock);
+		if (ar->wmi.drop)
+			ret = -EPERM;
+		else
+			ret = ath10k_wmi_cmd_send_nowait(ar, skb, cmd_id);
+		spin_unlock_bh(&ar->data_lock);
+
 		(ret != -EAGAIN);
 	}), 3*HZ);
 
@@ -2341,6 +2347,10 @@ int ath10k_wmi_attach(struct ath10k *ar)
 	init_completion(&ar->wmi.unified_ready);
 	init_waitqueue_head(&ar->wmi.tx_credits_wq);
 
+	spin_lock_bh(&ar->data_lock);
+	ar->wmi.drop = false;
+	spin_unlock_bh(&ar->data_lock);
+
 	return 0;
 }
 
-- 
1.8.5.3

[PATCH 4/7] ath10k: wake up everything on core restart

[Michal Kazior]

It makes sense to finish completions as they may
be waited for while holding conf_mutex thus
delaying hw restart.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
---
 drivers/net/wireless/ath/ath10k/core.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
index 3f4d28f..bf7c323 100644
--- a/drivers/net/wireless/ath/ath10k/core.c
+++ b/drivers/net/wireless/ath/ath10k/core.c
@@ -709,6 +709,18 @@ void ath10k_core_restart(struct ath10k *ar)
 	spin_lock_bh(&ar->data_lock);
 	ar->wmi.drop = true;
 	wake_up(&ar->wmi.tx_credits_wq);
+
+	complete(&ar->target_suspend);
+	complete(&ar->debug.event_stats_compl);
+	complete(&ar->htt.target_version_received);
+	complete(&ar->install_key_done);
+	complete(&ar->offchan_tx_completed);
+	complete(&ar->scan.started);
+	complete(&ar->scan.on_channel);
+	complete(&ar->vdev_setup_done);
+	complete(&ar->wmi.service_ready);
+	complete(&ar->wmi.unified_ready);
+	complete_all(&ar->scan.completed);
 	spin_unlock_bh(&ar->data_lock);
 
 	queue_work(ar->workqueue, &ar->restart_work);
-- 
1.8.5.3

[PATCH 5/7] ath10k: reduce timeout values

[Michal Kazior]

It doesn't make much sense to wait more than 1
second in most cases.

This speeds up teardown in some failcases.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
---
 drivers/net/wireless/ath/ath10k/core.h | 2 +-
 drivers/net/wireless/ath/ath10k/htt.c  | 2 +-
 drivers/net/wireless/ath/ath10k/mac.c  | 6 +++---
 drivers/net/wireless/ath/ath10k/txrx.c | 2 +-
 drivers/net/wireless/ath/ath10k/wmi.h  | 4 ++--
 5 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/core.h b/drivers/net/wireless/ath/ath10k/core.h
index 074cb41..01f1830 100644
--- a/drivers/net/wireless/ath/ath10k/core.h
+++ b/drivers/net/wireless/ath/ath10k/core.h
@@ -221,7 +221,7 @@ struct ath10k_sta {
 	struct work_struct update_wk;
 };
 
-#define ATH10K_VDEV_SETUP_TIMEOUT_HZ (5*HZ)
+#define ATH10K_VDEV_SETUP_TIMEOUT_HZ (1*HZ)
 
 struct ath10k_vif {
 	struct list_head list;
diff --git a/drivers/net/wireless/ath/ath10k/htt.c b/drivers/net/wireless/ath/ath10k/htt.c
index 69697af5..60ea5f1 100644
--- a/drivers/net/wireless/ath/ath10k/htt.c
+++ b/drivers/net/wireless/ath/ath10k/htt.c
@@ -100,7 +100,7 @@ err_htc_attach:
 	return ret;
 }
 
-#define HTT_TARGET_VERSION_TIMEOUT_HZ (3*HZ)
+#define HTT_TARGET_VERSION_TIMEOUT_HZ (1*HZ)
 
 static int ath10k_htt_verify_version(struct ath10k_htt *htt)
 {
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index 9b14ec8..9e3b16f 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -98,7 +98,7 @@ static int ath10k_install_key(struct ath10k_vif *arvif,
 	if (ret)
 		return ret;
 
-	ret = wait_for_completion_timeout(&ar->install_key_done, 3*HZ);
+	ret = wait_for_completion_timeout(&ar->install_key_done, 1*HZ);
 	if (ret == 0)
 		return -ETIMEDOUT;
 
@@ -2190,7 +2190,7 @@ static int ath10k_abort_scan(struct ath10k *ar)
 		return -EIO;
 	}
 
-	ret = wait_for_completion_timeout(&ar->scan.completed, 3*HZ);
+	ret = wait_for_completion_timeout(&ar->scan.completed, 1*HZ);
 	if (ret == 0)
 		ath10k_warn("timed out while waiting for scan to stop\n");
 
@@ -3604,7 +3604,7 @@ static int ath10k_remain_on_channel(struct ieee80211_hw *hw,
 		goto exit;
 	}
 
-	ret = wait_for_completion_timeout(&ar->scan.on_channel, 3*HZ);
+	ret = wait_for_completion_timeout(&ar->scan.on_channel, 1*HZ);
 	if (ret == 0) {
 		ath10k_warn("failed to switch to channel for roc scan\n");
 		ath10k_abort_scan(ar);
diff --git a/drivers/net/wireless/ath/ath10k/txrx.c b/drivers/net/wireless/ath/ath10k/txrx.c
index 82669a7..9007f2e 100644
--- a/drivers/net/wireless/ath/ath10k/txrx.c
+++ b/drivers/net/wireless/ath/ath10k/txrx.c
@@ -146,7 +146,7 @@ static int ath10k_wait_for_peer_common(struct ath10k *ar, int vdev_id,
 			spin_unlock_bh(&ar->data_lock);
 
 			mapped == expect_mapped;
-		}), 3*HZ);
+		}), 1*HZ);
 
 	if (ret <= 0)
 		return -ETIMEDOUT;
diff --git a/drivers/net/wireless/ath/ath10k/wmi.h b/drivers/net/wireless/ath/ath10k/wmi.h
index ae83822..772dbb6 100644
--- a/drivers/net/wireless/ath/ath10k/wmi.h
+++ b/drivers/net/wireless/ath/ath10k/wmi.h
@@ -1162,8 +1162,8 @@ struct wmi_service_ready_event_10x {
 } __packed;
 
 
-#define WMI_SERVICE_READY_TIMEOUT_HZ (5*HZ)
-#define WMI_UNIFIED_READY_TIMEOUT_HZ (5*HZ)
+#define WMI_SERVICE_READY_TIMEOUT_HZ (1*HZ)
+#define WMI_UNIFIED_READY_TIMEOUT_HZ (1*HZ)
 
 struct wmi_ready_event {
 	__le32 sw_version;
-- 
1.8.5.3

[PATCH 6/7] ath10k: make wmi timeout trigger hw restart

[Michal Kazior]

If tx credits aren't replenished for too long
firmware and driver state get out of sync. It
doesn't matter if firmware recovers a little later
on its own since by then it's already too late.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
---
 drivers/net/wireless/ath/ath10k/wmi.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/net/wireless/ath/ath10k/wmi.c b/drivers/net/wireless/ath/ath10k/wmi.c
index 159d744..e33c811 100644
--- a/drivers/net/wireless/ath/ath10k/wmi.c
+++ b/drivers/net/wireless/ath/ath10k/wmi.c
@@ -634,6 +634,11 @@ static int ath10k_wmi_cmd_send(struct ath10k *ar, struct sk_buff *skb,
 	if (ret)
 		dev_kfree_skb_any(skb);
 
+	if (ret == -EAGAIN) {
+		/* no credits for 3 seconds means fw is practically dead */
+		ath10k_core_restart(ar);
+	}
+
 	return ret;
 }
 
-- 
1.8.5.3

[PATCH 7/7] ath10k: remove old unused defines

[Michal Kazior]

These aren't used anywhere anymore.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
---
 drivers/net/wireless/ath/ath10k/core.h | 1 -
 drivers/net/wireless/ath/ath10k/mac.c  | 2 --
 2 files changed, 3 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/core.h b/drivers/net/wireless/ath/ath10k/core.h
index 01f1830..1128b5d 100644
--- a/drivers/net/wireless/ath/ath10k/core.h
+++ b/drivers/net/wireless/ath/ath10k/core.h
@@ -37,7 +37,6 @@
 #define WO(_f)      ((_f##_OFFSET) >> 2)
 
 #define ATH10K_SCAN_ID 0
-#define WMI_READY_TIMEOUT (5 * HZ)
 #define ATH10K_FLUSH_TIMEOUT_HZ (5*HZ)
 #define ATH10K_NUM_CHANS 38
 
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index 9e3b16f..e9608e7 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -3551,8 +3551,6 @@ exit:
 	return ret;
 }
 
-#define ATH10K_ROC_TIMEOUT_HZ (2*HZ)
-
 static int ath10k_remain_on_channel(struct ieee80211_hw *hw,
 				    struct ieee80211_vif *vif,
 				    struct ieee80211_channel *chan,
-- 
1.8.5.3

Re: [PATCH 0/7] ath10k: recovery improvements 2014-05-09

[Michal Kazior]

On 9 May 2014 14:35, Michal Kazior <michal.kazior@tieto.com> wrote:
> Hi,
>
> This depends on my `ath10k: recovery fixes
> 2014-05-09`.
>
> Hopefully the split makes it a little saner to
> review.
>
> Please note: `ath10k: make wmi timeout trigger hw
> restart` effectively hides the mgmt tx / htc
> credit starvation issue that was discussed
> recently. My intention is to make the driver more
> robust and as a side effect it makes the
> starvation issue less painful. If desired so
> the patch can be easily ignored in the series as
> nothing depends on it.
>
>
> Michal Kazior (7):
>   ath10k: move restart_work call
>   ath10k: introduce core restart symbol


>   ath10k: drain wmi quickly upon hw restart
>   ath10k: wake up everything on core restart

These 2 are harmful as they are. In case of early fw crash
uninitialized structures are accessed. I'll re-spin later.


Michał

Re: [PATCH 0/7] ath10k: recovery improvements 2014-05-09

[Kalle Valo]

Michal Kazior <michal.kazior@tieto.com> writes:

> On 9 May 2014 14:35, Michal Kazior <michal.kazior@tieto.com> wrote:
>> Hi,
>>
>> This depends on my `ath10k: recovery fixes
>> 2014-05-09`.
>>
>> Hopefully the split makes it a little saner to
>> review.
>>
>> Please note: `ath10k: make wmi timeout trigger hw
>> restart` effectively hides the mgmt tx / htc
>> credit starvation issue that was discussed
>> recently. My intention is to make the driver more
>> robust and as a side effect it makes the
>> starvation issue less painful. If desired so
>> the patch can be easily ignored in the series as
>> nothing depends on it.
>>
>>
>> Michal Kazior (7):
>>   ath10k: move restart_work call
>>   ath10k: introduce core restart symbol
>
>
>>   ath10k: drain wmi quickly upon hw restart
>>   ath10k: wake up everything on core restart
>
> These 2 are harmful as they are. In case of early fw crash
> uninitialized structures are accessed. I'll re-spin later.

So I can drop the whole series, right?

-- 
Kalle Valo

Re: [PATCH 0/7] ath10k: recovery improvements 2014-05-09

[Michal Kazior]

On 14 May 2014 15:34, Kalle Valo <kvalo@qca.qualcomm.com> wrote:
> Michal Kazior <michal.kazior@tieto.com> writes:
>
>> On 9 May 2014 14:35, Michal Kazior <michal.kazior@tieto.com> wrote:
>>> Hi,
>>>
>>> This depends on my `ath10k: recovery fixes
>>> 2014-05-09`.
>>>
>>> Hopefully the split makes it a little saner to
>>> review.
>>>
>>> Please note: `ath10k: make wmi timeout trigger hw
>>> restart` effectively hides the mgmt tx / htc
>>> credit starvation issue that was discussed
>>> recently. My intention is to make the driver more
>>> robust and as a side effect it makes the
>>> starvation issue less painful. If desired so
>>> the patch can be easily ignored in the series as
>>> nothing depends on it.
>>>
>>>
>>> Michal Kazior (7):
>>>   ath10k: move restart_work call
>>>   ath10k: introduce core restart symbol
>>
>>
>>>   ath10k: drain wmi quickly upon hw restart
>>>   ath10k: wake up everything on core restart
>>
>> These 2 are harmful as they are. In case of early fw crash
>> uninitialized structures are accessed. I'll re-spin later.
>
> So I can drop the whole series, right?

Correct.


Michał

Re: [PATCH 0/7] ath10k: recovery improvements 2014-05-09

[Kalle Valo]

Michal Kazior <michal.kazior@tieto.com> writes:

> On 14 May 2014 15:34, Kalle Valo <kvalo@qca.qualcomm.com> wrote:
>> Michal Kazior <michal.kazior@tieto.com> writes:
>>
>>> On 9 May 2014 14:35, Michal Kazior <michal.kazior@tieto.com> wrote:
>>>> Hi,
>>>>
>>>> This depends on my `ath10k: recovery fixes
>>>> 2014-05-09`.
>>>>
>>>> Hopefully the split makes it a little saner to
>>>> review.
>>>>
>>>> Please note: `ath10k: make wmi timeout trigger hw
>>>> restart` effectively hides the mgmt tx / htc
>>>> credit starvation issue that was discussed
>>>> recently. My intention is to make the driver more
>>>> robust and as a side effect it makes the
>>>> starvation issue less painful. If desired so
>>>> the patch can be easily ignored in the series as
>>>> nothing depends on it.
>>>>
>>>>
>>>> Michal Kazior (7):
>>>>   ath10k: move restart_work call
>>>>   ath10k: introduce core restart symbol
>>>
>>>
>>>>   ath10k: drain wmi quickly upon hw restart
>>>>   ath10k: wake up everything on core restart
>>>
>>> These 2 are harmful as they are. In case of early fw crash
>>> uninitialized structures are accessed. I'll re-spin later.
>>
>> So I can drop the whole series, right?
>
> Correct.

Thanks. I dropped the series from my queue now.

-- 
Kalle Valo

Re: [PATCH 4/7] ath10k: wake up everything on core restart

[Kalle Valo]

Michal Kazior <michal.kazior@tieto.com> writes:

> It makes sense to finish completions as they may
> be waited for while holding conf_mutex thus
> delaying hw restart.
>
> Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
> ---
>  drivers/net/wireless/ath/ath10k/core.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
>
> diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
> index 3f4d28f..bf7c323 100644
> --- a/drivers/net/wireless/ath/ath10k/core.c
> +++ b/drivers/net/wireless/ath/ath10k/core.c
> @@ -709,6 +709,18 @@ void ath10k_core_restart(struct ath10k *ar)
>  	spin_lock_bh(&ar->data_lock);
>  	ar->wmi.drop = true;
>  	wake_up(&ar->wmi.tx_credits_wq);
> +
> +	complete(&ar->target_suspend);
> +	complete(&ar->debug.event_stats_compl);

We dropped this patchset already, but for the future, this patch
introduces a new warning:

   drivers/net/wireless/ath/ath10k/core.c: In function 'ath10k_core_restart':
>> drivers/net/wireless/ath/ath10k/core.c:714:14: error: 'struct ath10k'
>> has no member named 'debug'
     complete(&ar->debug.event_stats_compl);
                 ^

Maybe that complete call should be in debug.c?

-- 
Kalle Valo

Re: [PATCH 4/7] ath10k: wake up everything on core restart

[Michal Kazior]

On 15 May 2014 15:21, Kalle Valo <kvalo@qca.qualcomm.com> wrote:
> Michal Kazior <michal.kazior@tieto.com> writes:
>
>> It makes sense to finish completions as they may
>> be waited for while holding conf_mutex thus
>> delaying hw restart.
>>
>> Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
>> ---
>>  drivers/net/wireless/ath/ath10k/core.c | 12 ++++++++++++
>>  1 file changed, 12 insertions(+)
>>
>> diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
>> index 3f4d28f..bf7c323 100644
>> --- a/drivers/net/wireless/ath/ath10k/core.c
>> +++ b/drivers/net/wireless/ath/ath10k/core.c
>> @@ -709,6 +709,18 @@ void ath10k_core_restart(struct ath10k *ar)
>>       spin_lock_bh(&ar->data_lock);
>>       ar->wmi.drop = true;
>>       wake_up(&ar->wmi.tx_credits_wq);
>> +
>> +     complete(&ar->target_suspend);
>> +     complete(&ar->debug.event_stats_compl);
>
> We dropped this patchset already, but for the future, this patch
> introduces a new warning:
>
>    drivers/net/wireless/ath/ath10k/core.c: In function 'ath10k_core_restart':
>>> drivers/net/wireless/ath/ath10k/core.c:714:14: error: 'struct ath10k'
>>> has no member named 'debug'
>      complete(&ar->debug.event_stats_compl);
>                  ^

Good catch. I completely forgot ar->debug is inside an #ifdef DEBUG.


> Maybe that complete call should be in debug.c?

Well, either we add an ifdef here (core.c @ L714), or an ifdef in
debug.h + debug.c..


Michał