Skip to contentSkip to conversation

wmi: Handle failure to start scan.

RFC In review [email protected] 3 participants 6 messages Feb 13, 2014 – Feb 17, 2014 Qualcomm Atheros ATH10K wireless driver

lore
Share & tools
lore
#0 lore raw

RFC 1/1 wmi: Handle failure to start scan.

From: Ben Greear <[email protected]>

Properly clean up driver state in case firmware fails to start scan for some reason.

---
 drivers/net/wireless/ath/ath10k/wmi.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)
drivers/net/wireless/ath/ath10k/wmi.c +19 -1
index 20f7c79..a5be0d3 100644
@@ -770,7 +770,25 @@ static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb)
 		ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_PREEMPTED\n");
 		break;
 	case WMI_SCAN_EVENT_START_FAILED:
-		ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_START_FAILED\n");
+		ath10k_warn("WMI_SCAN_EVENT_START_FAILED, reason: %i\n", reason);
+
+		ar->scan_channel = NULL;
+		if (!ar->scan.in_progress) {
+			ath10k_warn("scan-start-failed: no scan requested, ignoring\n");
+			break;
+		}
+
+		if (ar->scan.is_roc) {
+			ath10k_offchan_tx_purge(ar);
+
+			if (!ar->scan.aborting)
+				ieee80211_remain_on_channel_expired(ar->hw);
+		} else {
+			ieee80211_scan_completed(ar->hw, ar->scan.aborting);
+		}
+
+		del_timer(&ar->scan.timeout);
+		ar->scan.in_progress = false;
 		break;
 	default:
 		break;
Signature · 7 lines
1.7.11.7


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#1 lore raw
Quoted 6 lines · [email protected] writes:

From: Ben Greear <[email protected]>

Properly clean up driver state in case firmware fails to start scan for some reason.

Why just RFC?

And "ath10k:" prefix missing.

drivers/net/wireless/ath/ath10k/wmi.c static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb) · line 773

Quoted 8 lines
drivers/net/wireless/ath/ath10k/wmi.c +1 -1
@@ -770,7 +770,25 @@ static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb)
 		ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_PREEMPTED\n");
 		break;
 	case WMI_SCAN_EVENT_START_FAILED:
-		ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_START_FAILED\n");
+		ath10k_warn("WMI_SCAN_EVENT_START_FAILED, reason: %i\n", reason);

"scan failed to start: %i\n"

drivers/net/wireless/ath/ath10k/wmi.c static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb) · line 777

Quoted 3 lines
drivers/net/wireless/ath/ath10k/wmi.c +3 -0
+		ar->scan_channel = NULL;
+		if (!ar->scan.in_progress) {
+			ath10k_warn("scan-start-failed: no scan requested, ignoring\n");

"scan failed to start but no scan requested, ignoring\n"

Signature · 1 line
Kalle Valo
#2 lore raw
In reply to [email protected]
Quoted 31 lines · On 13 February 2014 20:09, <[email protected]> wrote:

From: Ben Greear <[email protected]>

Properly clean up driver state in case firmware fails to start scan for some reason.

---
 drivers/net/wireless/ath/ath10k/wmi.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)
drivers/net/wireless/ath/ath10k/wmi.c +12 -1
index 20f7c79..a5be0d3 100644
@@ -770,7 +770,25 @@ static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb)
                ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_PREEMPTED\n");
                break;
        case WMI_SCAN_EVENT_START_FAILED:
-               ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_START_FAILED\n");
+               ath10k_warn("WMI_SCAN_EVENT_START_FAILED, reason: %i\n", reason);
+
+               ar->scan_channel = NULL;
+               if (!ar->scan.in_progress) {
+                       ath10k_warn("scan-start-failed: no scan requested, ignoring\n");
+                       break;
+               }
+
+               if (ar->scan.is_roc) {
+                       ath10k_offchan_tx_purge(ar);
+
+                       if (!ar->scan.aborting)

drivers/net/wireless/ath/ath10k/wmi.c static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb) · line 794

Quoted 10 lines
drivers/net/wireless/ath/ath10k/wmi.c +7 -0
+                               ieee80211_remain_on_channel_expired(ar->hw);
+               } else {
+                       ieee80211_scan_completed(ar->hw, ar->scan.aborting);
+               }
+
+               del_timer(&ar->scan.timeout);
+               ar->scan.in_progress = false;
                break;
        default:
                break;

We already wait for EVENT_STARTED in mac.c (see ath10k_start_scan) and clean up stuff (ath10k_abort_scan). Why not add the missing bits in there? Or is it possible to get EVENT_START_FAILED *after* EVENT_STARTED? Or am I missing something else here?

Michał

author
#3 lore raw
Quoted 39 lines · Michal Kazior
Quoted 38 lines · On 13 February 2014 20:09, <[email protected]> wrote:

From: Ben Greear <[email protected]>

Properly clean up driver state in case firmware fails to start scan for some reason.

---
  drivers/net/wireless/ath/ath10k/wmi.c | 20 +++++++++++++++++++-
  1 file changed, 19 insertions(+), 1 deletion(-)
drivers/net/wireless/ath/ath10k/wmi.c +19 -1
index 20f7c79..a5be0d3 100644
@@ -770,7 +770,25 @@ static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb)
                 ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_PREEMPTED\n");
                 break;
         case WMI_SCAN_EVENT_START_FAILED:
-               ath10k_dbg(ATH10K_DBG_WMI, "WMI_SCAN_EVENT_START_FAILED\n");
+               ath10k_warn("WMI_SCAN_EVENT_START_FAILED, reason: %i\n", reason);
+
+               ar->scan_channel = NULL;
+               if (!ar->scan.in_progress) {
+                       ath10k_warn("scan-start-failed: no scan requested, ignoring\n");
+                       break;
+               }
+
+               if (ar->scan.is_roc) {
+                       ath10k_offchan_tx_purge(ar);
+
+                       if (!ar->scan.aborting)
+                               ieee80211_remain_on_channel_expired(ar->hw);
+               } else {
+                       ieee80211_scan_completed(ar->hw, ar->scan.aborting);
+               }
+
+               del_timer(&ar->scan.timeout);
+               ar->scan.in_progress = false;
Quoted 8 lines

drivers/net/wireless/ath/ath10k/wmi.c static int ath10k_wmi_event_scan(struct ath10k *ar, struct sk_buff *skb) · line 794

Quoted 3 lines
drivers/net/wireless/ath/ath10k/wmi.c +0 -0
                 break;
         default:
                 break;

We already wait for EVENT_STARTED in mac.c (see ath10k_start_scan) and clean up stuff (ath10k_abort_scan). Why not add the missing bits in there? Or is it possible to get EVENT_START_FAILED *after* EVENT_STARTED? Or am I missing something else here?

I think a lot of this would be firmware dependent, and might change between various versions of the firmware.

It seems to me we should handle this case and do cleanup just to be safe, but maybe cleanup is needed in failure case of ath10k_start_scan as well?

Thanks,
Ben

Signature · 8 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#4 lore raw
Quoted 61 lines · Ben Greear
Quoted 50 lines · Michal Kazior
Quoted 48 lines · On 13 February 2014 20:09, <[email protected]> wrote:

From: Ben Greear <[email protected]>

Properly clean up driver state in case firmware fails to start scan for some reason.

---
  drivers/net/wireless/ath/ath10k/wmi.c | 20 +++++++++++++++++++-
  1 file changed, 19 insertions(+), 1 deletion(-)
drivers/net/wireless/ath/ath10k/wmi.c +1 -1
index 20f7c79..a5be0d3 100644
@@ -770,7 +770,25 @@ static int ath10k_wmi_event_scan(struct ath10k *ar,
 struct sk_buff *skb)
                 ath10k_dbg(ATH10K_DBG_WMI,
 "WMI_SCAN_EVENT_PREEMPTED\n");
                 break;
         case WMI_SCAN_EVENT_START_FAILED:
-               ath10k_dbg(ATH10K_DBG_WMI,
 "WMI_SCAN_EVENT_START_FAILED\n");
+               ath10k_warn("WMI_SCAN_EVENT_START_FAILED, reason: %i\n",

reason);

drivers/net/wireless/ath/ath10k/wmi.c +4 -0
+
+               ar->scan_channel = NULL;
+               if (!ar->scan.in_progress) {
+                       ath10k_warn("scan-start-failed: no scan

requested, ignoring\n");

drivers/net/wireless/ath/ath10k/wmi.c +8 -0
+                       break;
+               }
+
+               if (ar->scan.is_roc) {
+                       ath10k_offchan_tx_purge(ar);
+
+                       if (!ar->scan.aborting)
+

ieee80211_remain_on_channel_expired(ar->hw);

drivers/net/wireless/ath/ath10k/wmi.c +2 -0
+               } else {
+                       ieee80211_scan_completed(ar->hw,

ar->scan.aborting);

drivers/net/wireless/ath/ath10k/wmi.c +4 -0
+               }
+
+               del_timer(&ar->scan.timeout);
+               ar->scan.in_progress = false;
                 break;
Quoted 8 lines
Quoted 2 lines
         default:
                 break;

We already wait for EVENT_STARTED in mac.c (see ath10k_start_scan) and clean up stuff (ath10k_abort_scan). Why not add the missing bits in there? Or is it possible to get EVENT_START_FAILED *after* EVENT_STARTED? Or am I missing something else here?

Quoted 2 lines

I think a lot of this would be firmware dependent, and might change between various versions of the firmware.

It doesn't make any sense. That would suggest a really ugly firmware bug. Did you see this (i.e. START_FAILED after STARTED) on 636 or 10.1.467?

Quoted 2 lines

It seems to me we should handle this case and do cleanup just to be safe, but maybe cleanup is needed in failure case of ath10k_start_scan as well?

If you really get START_FAILED then you shouldn't have received STARTED before that. ath10k_start_scan() already waits for the STARTED event with a timeout and if it fails it triggers a cleanup. If it doesn't work for you then what perhaps needs to be fixed is the current cleanup code?

Michał

author
#5 lore raw
Quoted 8 lines · Michal Kazior
Quoted 6 lines · Ben Greear
Quoted 4 lines

We already wait for EVENT_STARTED in mac.c (see ath10k_start_scan) and clean up stuff (ath10k_abort_scan). Why not add the missing bits in there? Or is it possible to get EVENT_START_FAILED *after* EVENT_STARTED? Or am I missing something else here?

Quoted 6 lines
Quoted 2 lines

I think a lot of this would be firmware dependent, and might change between various versions of the firmware.

It doesn't make any sense. That would suggest a really ugly firmware bug. Did you see this (i.e. START_FAILED after STARTED) on 636 or 10.1.467?

I am working on a 10.1.389 release currently...will move forward when I can get access to newer source code.

Quoted 8 lines
Quoted 2 lines

It seems to me we should handle this case and do cleanup just to be safe, but maybe cleanup is needed in failure case of ath10k_start_scan as well?

If you really get START_FAILED then you shouldn't have received STARTED before that. ath10k_start_scan() already waits for the STARTED event with a timeout and if it fails it triggers a cleanup. If it doesn't work for you then what perhaps needs to be fixed is the current cleanup code?

I am not certain the patch is needed. I was looking at my firmware and it appeared that I could hit the START_FAILED case, but perhaps it was not really possible, and maybe newer firmware keeps it from happening entirely.

Thanks,
Ben

Signature · 8 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k