[PATCH] wlcore: time sync : add support for 64 bit clock

[PATCH] wlcore: time sync : add support for 64 bit clock

[Yaniv Machani]

Changed the configuration to support 64bit instead of 32bit
this in order to offload the driver from handling a wraparound.

Signed-off-by: Yaniv Machani <yanivma@ti.com>
---
 drivers/net/wireless/ti/wl18xx/event.c | 26 +++++++++++++++++---------
 drivers/net/wireless/ti/wl18xx/event.h | 19 +++++++++++++------
 2 files changed, 30 insertions(+), 15 deletions(-)

diff --git a/drivers/net/wireless/ti/wl18xx/event.c b/drivers/net/wireless/ti/wl18xx/event.c
index ef81184..2c5df43 100644
--- a/drivers/net/wireless/ti/wl18xx/event.c
+++ b/drivers/net/wireless/ti/wl18xx/event.c
@@ -112,12 +112,18 @@ static int wlcore_smart_config_decode_event(struct wl1271 *wl,
 	return 0;
 }
 
-static void wlcore_event_time_sync(struct wl1271 *wl, u16 tsf_msb, u16 tsf_lsb)
+static void wlcore_event_time_sync(struct wl1271 *wl,
+				   u16 tsf_high_msb, u16 tsf_high_lsb,
+				   u16 tsf_low_msb, u16 tsf_low_lsb)
 {
-	u32 clock;
-	/* convert the MSB+LSB to a u32 TSF value */
-	clock = (tsf_msb << 16) | tsf_lsb;
-	wl1271_info("TIME_SYNC_EVENT_ID: clock %u", clock);
+	u32 clock_low;
+	u32 clock_high;
+
+	clock_high = (tsf_high_msb << 16) | tsf_high_lsb;
+	clock_low = (tsf_low_msb << 16) | tsf_low_lsb;
+
+	wl1271_info("TIME_SYNC_EVENT_ID: clock_high %u, clock low %u",
+		    clock_high, clock_low);
 }
 
 int wl18xx_process_mailbox_events(struct wl1271 *wl)
@@ -138,8 +144,10 @@ int wl18xx_process_mailbox_events(struct wl1271 *wl)
 
 	if (vector & TIME_SYNC_EVENT_ID)
 		wlcore_event_time_sync(wl,
-				mbox->time_sync_tsf_msb,
-				mbox->time_sync_tsf_lsb);
+			mbox->time_sync_tsf_high_msb,
+			mbox->time_sync_tsf_high_lsb,
+			mbox->time_sync_tsf_low_msb,
+			mbox->time_sync_tsf_low_lsb);
 
 	if (vector & RADAR_DETECTED_EVENT_ID) {
 		wl1271_info("radar event: channel %d type %s",
@@ -187,11 +195,11 @@ int wl18xx_process_mailbox_events(struct wl1271 *wl)
 	 */
 	if (vector & MAX_TX_FAILURE_EVENT_ID)
 		wlcore_event_max_tx_failure(wl,
-				le32_to_cpu(mbox->tx_retry_exceeded_bitmap));
+				le16_to_cpu(mbox->tx_retry_exceeded_bitmap));
 
 	if (vector & INACTIVE_STA_EVENT_ID)
 		wlcore_event_inactive_sta(wl,
-				le32_to_cpu(mbox->inactive_sta_bitmap));
+				le16_to_cpu(mbox->inactive_sta_bitmap));
 
 	if (vector & REMAIN_ON_CHANNEL_COMPLETE_EVENT_ID)
 		wlcore_event_roc_complete(wl);
diff --git a/drivers/net/wireless/ti/wl18xx/event.h b/drivers/net/wireless/ti/wl18xx/event.h
index 070de12..b436bf9 100644
--- a/drivers/net/wireless/ti/wl18xx/event.h
+++ b/drivers/net/wireless/ti/wl18xx/event.h
@@ -74,10 +74,16 @@ struct wl18xx_event_mailbox {
 	__le16 bss_loss_bitmap;
 
 	/* bitmap of stations (by HLID) which exceeded max tx retries */
-	__le32 tx_retry_exceeded_bitmap;
+	__le16 tx_retry_exceeded_bitmap;
+
+	/* time sync high msb*/
+	u16 time_sync_tsf_high_msb;
 
 	/* bitmap of inactive stations (by HLID) */
-	__le32 inactive_sta_bitmap;
+	__le16 inactive_sta_bitmap;
+
+	/* time sync high lsb*/
+	u16 time_sync_tsf_high_lsb;
 
 	/* rx BA win size indicated by RX_BA_WIN_SIZE_CHANGE_EVENT_ID */
 	u8 rx_ba_role_id;
@@ -98,14 +104,15 @@ struct wl18xx_event_mailbox {
 	u8 sc_sync_channel;
 	u8 sc_sync_band;
 
-	/* time sync msb*/
-	u16 time_sync_tsf_msb;
+	/* time sync low msb*/
+	u16 time_sync_tsf_low_msb;
+
 	/* radar detect */
 	u8 radar_channel;
 	u8 radar_type;
 
-	/* time sync lsb*/
-	u16 time_sync_tsf_lsb;
+	/* time sync low lsb*/
+	u16 time_sync_tsf_low_lsb;
 
 } __packed;
 
-- 
2.9.0

Re: [PATCH] wlcore: time sync : add support for 64 bit clock

[Johannes Berg]

On Thu, 2016-06-23 at 14:12 +0300, Yaniv Machani wrote:
> Changed the configuration to support 64bit instead of 32bit
> this in order to offload the driver from handling a wraparound.

[...]

Since you Cc'ed me, and presumably want me to review it, I'll say that
this looks like a terrible idea:

> @@ -74,10 +74,16 @@ struct wl18xx_event_mailbox {

This struct is evidently used for firmware/host communication.

>  	__le16 bss_loss_bitmap;
>  
>  	/* bitmap of stations (by HLID) which exceeded max tx
> retries */
> -	__le32 tx_retry_exceeded_bitmap;
> +	__le16 tx_retry_exceeded_bitmap;
> +
> +	/* time sync high msb*/
> +	u16 time_sync_tsf_high_msb;

So first of all, just using u16 instead of __le16 seems wrong.

Additionally, this looks like it changes the firmware API, so that
older firmware images will no longer work?

johannes

RE: [PATCH] wlcore: time sync : add support for 64 bit clock

[Machani, Yaniv]

On Thu, Jun 23, 2016 at 14:18:00, Johannes Berg wrote:
> linux-wireless@vger.kernel.org; netdev@vger.kernel.org
> Subject: Re: [PATCH] wlcore: time sync : add support for 64 bit clock
> 
> On Thu, 2016-06-23 at 14:12 +0300, Yaniv Machani wrote:
> > Changed the configuration to support 64bit instead of 32bit this in 
> > order to offload the driver from handling a wraparound.
> 
> [...]
> 
> Since you Cc'ed me, and presumably want me to review it, I'll say that 
> this looks like a terrible idea:
> 
> > @@ -74,10 +74,16 @@ struct wl18xx_event_mailbox {
> 
> This struct is evidently used for firmware/host communication.
> 
> >  	__le16 bss_loss_bitmap;
> >
> >  	/* bitmap of stations (by HLID) which exceeded max tx retries */
> > -	__le32 tx_retry_exceeded_bitmap;
> > +	__le16 tx_retry_exceeded_bitmap;
> > +
> > +	/* time sync high msb*/
> > +	u16 time_sync_tsf_high_msb;
> 
> So first of all, just using u16 instead of __le16 seems wrong.

Agree, should be changed.

> 
> Additionally, this looks like it changes the firmware API, so that 
> older firmware images will no longer work?

It is backwards compatible, 
although it changes a API structure, older firmware are using only u16 for the field so there is no impact on that.
Of course that for actually using the 64bit information, you will have to upgrade the firmware.

Yaniv

Re: [PATCH] wlcore: time sync : add support for 64 bit clock

[Johannes Berg]

> > Additionally, this looks like it changes the firmware API, so that 
> > older firmware images will no longer work?
> 
> It is backwards compatible, 
> although it changes a API structure, older firmware are using only
> u16 for the field so there is no impact on that.
> 

Oh, ok. I had also thought that the size changed, but missed that you
replaced a u32 with two u16. Thanks for checking :)

johannes