Skip to contentSkip to conversation

rt2x00: rt2800: update initial SIFS values

PATCH In review Helmut Schaa 3 participants 7 messages May 6, 2010 Networking drivers (wireless)

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 rt2x00: rt2800: update initial SIFS values

Currently the CCK and OFDM SIFS value is set to 32us. This value is neither used by the Ralink driver nor specified in 802.11.

Instead of using 10us for CCK SIFS (as defined in 802.11) use 16us like in the Ralink drivers. And indeed using a SIFS value of 10us breaks connectivity with 11g + CTS protected connections. Add a comment to the code why we don't use 10us for CCK SIFS value.

The OFDM SIFS value is set to 16us (as defined in 802.11 and also used by the Ralink drivers).

---
 drivers/net/wireless/rt2x00/rt2800lib.c |   11 +++++++++--
 1 files changed, 9 insertions(+), 2 deletions(-)
drivers/net/wireless/rt2x00/rt2800lib.c +9 -2
index e37bbea..f786068 100644
@@ -1415,9 +1415,16 @@ int rt2800_init_registers(struct rt2x00_dev *rt2x00dev)
 
 	rt2800_register_write(rt2x00dev, EXP_ACK_TIME, 0x002400ca);
 
+	/*
+	 * Usually the CCK SIFS time should be set to 10 and the OFDM SIFS
+	 * time should be set to 16. However, the original Ralink driver uses
+	 * 16 for both and indeed using a value of 10 for CCK SIFS results in
+	 * connection problems with 11g + CTS protection. Hence, use the same
+	 * defaults as the Ralink driver: 16 for both, CCK and OFDM SIFS.
+	 */
 	rt2800_register_read(rt2x00dev, XIFS_TIME_CFG, &reg);
-	rt2x00_set_field32(&reg, XIFS_TIME_CFG_CCKM_SIFS_TIME, 32);
-	rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_SIFS_TIME, 32);
+	rt2x00_set_field32(&reg, XIFS_TIME_CFG_CCKM_SIFS_TIME, 16);
+	rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_SIFS_TIME, 16);
 	rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_XIFS_TIME, 4);
 	rt2x00_set_field32(&reg, XIFS_TIME_CFG_EIFS, 314);
 	rt2x00_set_field32(&reg, XIFS_TIME_CFG_BB_RXEND_ENABLE, 1);
Signature · 1 line
1.6.4.2
Acked-by on v1 1/3
#1 lore raw
Quoted 8 lines · Helmut Schaa

Currently the CCK and OFDM SIFS value is set to 32us. This value is neither used by the Ralink driver nor specified in 802.11.

Instead of using 10us for CCK SIFS (as defined in 802.11) use 16us like in the Ralink drivers. And indeed using a SIFS value of 10us breaks connectivity with 11g + CTS protected connections. Add a comment to the code why we don't use 10us for CCK SIFS value.

Quoted 4 lines

The OFDM SIFS value is set to 16us (as defined in 802.11 and also used by the Ralink drivers).

Good catch.

Quoted 27 lines
---
 drivers/net/wireless/rt2x00/rt2800lib.c |   11 +++++++++--
 1 files changed, 9 insertions(+), 2 deletions(-)
drivers/net/wireless/rt2x00/rt2800lib.c +9 -2
index e37bbea..f786068 100644
@@ -1415,9 +1415,16 @@ int rt2800_init_registers(struct rt2x00_dev *rt2x00dev)
 
 	rt2800_register_write(rt2x00dev, EXP_ACK_TIME, 0x002400ca);
 
+	/*
+	 * Usually the CCK SIFS time should be set to 10 and the OFDM SIFS
+	 * time should be set to 16. However, the original Ralink driver uses
+	 * 16 for both and indeed using a value of 10 for CCK SIFS results in
+	 * connection problems with 11g + CTS protection. Hence, use the same
+	 * defaults as the Ralink driver: 16 for both, CCK and OFDM SIFS.
+	 */
 	rt2800_register_read(rt2x00dev, XIFS_TIME_CFG, &reg);
-	rt2x00_set_field32(&reg, XIFS_TIME_CFG_CCKM_SIFS_TIME, 32);
-	rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_SIFS_TIME, 32);
+	rt2x00_set_field32(&reg, XIFS_TIME_CFG_CCKM_SIFS_TIME, 16);
+	rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_SIFS_TIME, 16);
 	rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_XIFS_TIME, 4);
 	rt2x00_set_field32(&reg, XIFS_TIME_CFG_EIFS, 314);
 	rt2x00_set_field32(&reg, XIFS_TIME_CFG_BB_RXEND_ENABLE, 1);
Acked-by on v1 1/3
#2 lore raw
In reply to Helmut Schaa
Quoted 10 lines · Helmut Schaa

Currently the CCK and OFDM SIFS value is set to 32us. This value is neither used by the Ralink driver nor specified in 802.11.

Instead of using 10us for CCK SIFS (as defined in 802.11) use 16us like in the Ralink drivers. And indeed using a SIFS value of 10us breaks connectivity with 11g + CTS protected connections. Add a comment to the code why we don't use 10us for CCK SIFS value.

The OFDM SIFS value is set to 16us (as defined in 802.11 and also used by the Ralink drivers).

Just wondering, but we hardcode the SIFS value in the rt2x00.h file. Perhaps we should remove it in there, and no longer pass it from rt2x00lib. That way there can't be any confusion about which drivers uses the sifs field and whcih do not.

Quoted 1 line

But that is really a different issue then this fix. So:

Quoted 31 lines

---
 drivers/net/wireless/rt2x00/rt2800lib.c |   11 +++++++++--

 1 files changed, 9 insertions(+), 2 deletions(-)
drivers/net/wireless/rt2x00/rt2800lib.c +9 -3
index e37bbea..f786068 100644
@@ -1415,9 +1415,16 @@ int rt2800_init_registers(struct rt2x00_dev *rt2x00dev)
 
        rt2800_register_write(rt2x00dev, EXP_ACK_TIME, 0x002400ca);
 
+       /*
+        * Usually the CCK SIFS time should be set to 10 and the OFDM SIFS
+        * time should be set to 16. However, the original Ralink driver uses
+        * 16 for both and indeed using a value of 10 for CCK SIFS results in
+        * connection problems with 11g + CTS protection. Hence, use the same
+        * defaults as the Ralink driver: 16 for both, CCK and OFDM SIFS.
+        */
        rt2800_register_read(rt2x00dev, XIFS_TIME_CFG, &reg);
-       rt2x00_set_field32(&reg, XIFS_TIME_CFG_CCKM_SIFS_TIME, 32);
-       rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_SIFS_TIME, 32);
+       rt2x00_set_field32(&reg, XIFS_TIME_CFG_CCKM_SIFS_TIME, 16);
+       rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_SIFS_TIME, 16);
        rt2x00_set_field32(&reg, XIFS_TIME_CFG_OFDM_XIFS_TIME, 4);
        rt2x00_set_field32(&reg, XIFS_TIME_CFG_EIFS, 314);
        rt2x00_set_field32(&reg, XIFS_TIME_CFG_BB_RXEND_ENABLE, 1);
--

1.6.4.2

#3 lore raw

Am Donnerstag 06 Mai 2010 schrieb Ivo Van Doorn:

Quoted 10 lines
Quoted 8 lines · Helmut Schaa

Currently the CCK and OFDM SIFS value is set to 32us. This value is neither used by the Ralink driver nor specified in 802.11.

Instead of using 10us for CCK SIFS (as defined in 802.11) use 16us like in the Ralink drivers. And indeed using a SIFS value of 10us breaks connectivity with 11g + CTS protected connections. Add a comment to the code why we don't use 10us for CCK SIFS value.

Quoted 7 lines
Quoted 2 lines

The OFDM SIFS value is set to 16us (as defined in 802.11 and also used by the Ralink drivers).

Just wondering, but we hardcode the SIFS value in the rt2x00.h file. Perhaps we should remove it in there, and no longer pass it from rt2x00lib. That way there can't be any confusion about which drivers uses the sifs field and whcih do not.

Yes, wouldn't be too bad I guess. But all non 2800 drivers use the same sifs value, so we could just leave the define in rt2x00.h and remove the sifs value from the config_erp calback and use the define in all other places?

Nevertheless, I'm not sure why the rt2800 devices don't like the 10us CCK SIFS value (which is defined in 802.11) but I wasn't able to get CTS to self working correctly with it set to 10 (the actual data frame was sent out way too late after the CTS frame, somtimes with delays >100us). Using the 16us also for CCK (as the ralink drivers do) results in perfect CTS & data frame timing. Maybe it's a hardware issue and the (ralink) driver just works around it?

Helmut

#4 lore raw
Quoted 14 lines · Helmut Schaa

Am Donnerstag 06 Mai 2010 schrieb Ivo Van Doorn:

Quoted 13 lines
Quoted 10 lines · Helmut Schaa

Currently the CCK and OFDM SIFS value is set to 32us. This value is neither used by the Ralink driver nor specified in 802.11.

Instead of using 10us for CCK SIFS (as defined in 802.11) use 16us like in the Ralink drivers. And indeed using a SIFS value of 10us breaks connectivity with 11g + CTS protected connections. Add a comment to the code why we don't use 10us for CCK SIFS value.

The OFDM SIFS value is set to 16us (as defined in 802.11 and also used by the Ralink drivers).

Quoted 8 lines
Quoted 4 lines

Just wondering, but we hardcode the SIFS value in the rt2x00.h file. Perhaps we should remove it in there, and no longer pass it from rt2x00lib. That way there can't be any confusion about which drivers uses the sifs field and whcih do not.

Yes, wouldn't be too bad I guess. But all non 2800 drivers use the same sifs value, so we could just leave the define in rt2x00.h and remove the sifs value from the config_erp calback and use the define in all other places?

Well I would still remove the define in that case. I think some of the legacy drivers
work with different SIFS values but always accepted the value which we used in the define. So I would still remove the define, and then each driver is free to set
the value as used in the legacy drivers.

Quoted 7 lines

Nevertheless, I'm not sure why the rt2800 devices don't like the 10us CCK SIFS value (which is defined in 802.11) but I wasn't able to get CTS to self working correctly with it set to 10 (the actual data frame was sent out way too late after the CTS frame, somtimes with delays >100us). Using the 16us also for CCK (as the ralink drivers do) results in perfect CTS & data frame timing. Maybe it's a hardware issue and the (ralink) driver just works around it?

Yes, I expect it to be case. This kind of odd values are used in the legacy drivers
for other settings as well.

Ivo

#5 lore raw

Am Donnerstag 06 Mai 2010 schrieb Ivo Van Doorn:

Quoted 25 lines
Quoted 14 lines · Helmut Schaa

Am Donnerstag 06 Mai 2010 schrieb Ivo Van Doorn:

Quoted 13 lines
Quoted 10 lines · Helmut Schaa

Currently the CCK and OFDM SIFS value is set to 32us. This value is neither used by the Ralink driver nor specified in 802.11.

Instead of using 10us for CCK SIFS (as defined in 802.11) use 16us like in the Ralink drivers. And indeed using a SIFS value of 10us breaks connectivity with 11g + CTS protected connections. Add a comment to the code why we don't use 10us for CCK SIFS value.

The OFDM SIFS value is set to 16us (as defined in 802.11 and also used by the Ralink drivers).

Quoted 8 lines
Quoted 4 lines

Just wondering, but we hardcode the SIFS value in the rt2x00.h file. Perhaps we should remove it in there, and no longer pass it from rt2x00lib. That way there can't be any confusion about which drivers uses the sifs field and whcih do not.

Yes, wouldn't be too bad I guess. But all non 2800 drivers use the same sifs value, so we could just leave the define in rt2x00.h and remove the sifs value from the config_erp calback and use the define in all other places?

Quoted 6 lines

Well I would still remove the define in that case. I think some of the legacy drivers
work with different SIFS values but always accepted the value which we used in the define. So I would still remove the define, and then each driver is free to set
the value as used in the legacy drivers.

Sure. No objections from my side.

Helmut

#6 lore raw
Quoted 26 lines · Helmut Schaa

Am Donnerstag 06 Mai 2010 schrieb Ivo Van Doorn:

Quoted 25 lines
Quoted 22 lines · Helmut Schaa

Am Donnerstag 06 Mai 2010 schrieb Ivo Van Doorn:

Quoted 10 lines
Quoted 8 lines · Helmut Schaa

Currently the CCK and OFDM SIFS value is set to 32us. This value is neither used by the Ralink driver nor specified in 802.11.

Instead of using 10us for CCK SIFS (as defined in 802.11) use 16us like in the Ralink drivers. And indeed using a SIFS value of 10us breaks connectivity with 11g + CTS protected connections. Add a comment to the code why we don't use 10us for CCK SIFS value.

Quoted 7 lines
Quoted 2 lines

The OFDM SIFS value is set to 16us (as defined in 802.11 and also used by the Ralink drivers).

Just wondering, but we hardcode the SIFS value in the rt2x00.h file. Perhaps we should remove it in there, and no longer pass it from rt2x00lib. That way there can't be any confusion about which drivers uses the sifs field and whcih do not.

Yes, wouldn't be too bad I guess. But all non 2800 drivers use the same sifs value, so we could just leave the define in rt2x00.h and remove the sifs value from the config_erp calback and use the define in all other places?

Quoted 8 lines
Quoted 6 lines

Well I would still remove the define in that case. I think some of the legacy drivers
work with different SIFS values but always accepted the value which we used in the define. So I would still remove the define, and then each driver is free to set
the value as used in the legacy drivers.

Sure. No objections from my side.

I'll write a patch for this, this weekend along with some other updates. :)

Ivo