Return-path: Received: from mail-qy0-f181.google.com ([209.85.216.181]:46492 "EHLO mail-qy0-f181.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754484Ab0KXBfS convert rfc822-to-8bit (ORCPT ); Tue, 23 Nov 2010 20:35:18 -0500 Received: by qyk12 with SMTP id 12so3191618qyk.19 for ; Tue, 23 Nov 2010 17:35:17 -0800 (PST) MIME-Version: 1.0 In-Reply-To: References: <20101123190911.GN4303@makis.mantri> From: Jonathan Guerin Date: Wed, 24 Nov 2010 11:35:02 +1000 Message-ID: Subject: Re: [ath5k-devel] [PATCH 14/30] ath5k: Extend get_default_sifs/slot_time To: Nick Kossifidis Cc: ath5k-devel@lists.ath5k.org, linux-wireless@vger.kernel.org, linville@tuxdriver.com, me@bobcopeland.com, mcgrof@gmail.com, jirislaby@gmail.com, nbd@openwrt.org, br1@einfach.org Content-Type: text/plain; charset=windows-1252 Sender: linux-wireless-owner@vger.kernel.org List-ID: On Wed, Nov 24, 2010 at 11:04 AM, Nick Kossifidis wrote: > 2010/11/24 Nick Kossifidis : >> 2010/11/24 Jonathan Guerin : >>> On Wed, Nov 24, 2010 at 5:09 AM, Nick Kossifidis wrote: >>>> ?* Extend get_default_sifs/slot_time to include timings for turbo >>>> ?half and quarter rate modes. >>>> >>>> ?* AR5210 code for now uses timings already on core clock units >>>> ?instead of usecs so rename them (we 'll clean it up later). >>>> >>>> ?Signed-off-by: Nick Kossifidis >>>> --- >>>> ?drivers/net/wireless/ath/ath5k/ath5k.h | ? 24 ++++++++++++-- >>>> ?drivers/net/wireless/ath/ath5k/pcu.c ? | ? 52 ++++++++++++++++++++++++-------- >>>> ?drivers/net/wireless/ath/ath5k/qcu.c ? | ? 16 +++++---- >>>> ?3 files changed, 68 insertions(+), 24 deletions(-) >>>> >>>> diff --git a/drivers/net/wireless/ath/ath5k/ath5k.h b/drivers/net/wireless/ath/ath5k/ath5k.h >>>> index 005cad0..e11fc8f 100644 >>>> --- a/drivers/net/wireless/ath/ath5k/ath5k.h >>>> +++ b/drivers/net/wireless/ath/ath5k/ath5k.h >>>> @@ -226,16 +226,16 @@ >>>> ?#define AR5K_INIT_USEC ? ? ? ? ? ? ? ? ? ? ? ? 39 >>>> ?#define AR5K_INIT_USEC_TURBO ? ? ? ? ? ? ? ? ? 79 >>>> ?#define AR5K_INIT_USEC_32 ? ? ? ? ? ? ? ? ? ? ?31 >>>> -#define AR5K_INIT_SLOT_TIME ? ? ? ? ? ? ? ? ? ?396 >>>> -#define AR5K_INIT_SLOT_TIME_TURBO ? ? ? ? ? ? ?480 >>>> +#define AR5K_INIT_SLOT_TIME_CLOCK ? ? ? ? ? ? ?396 >>> >>> This comes out as 9.9us (with a clock multiplier of 40), rather than 9 >>> - should be 360? >>> >>>> +#define AR5K_INIT_SLOT_TIME_TURBO_CLOCK ? ? ? ? ? ? ? ?480 >>>> ?#define AR5K_INIT_ACK_CTS_TIMEOUT ? ? ? ? ? ? ?1024 >>> >>> From a previous post I'd made: >>> >>> According to the 802.11-2007 spec document, the ACKTimeout value is >>> (Section 9.2.8 ACK procedure): >>> ACKTimeout = aSIFSTime + aSlotTime + aPHY-RX-START-Delay >>> >>> From Table 17-15?OFDM PHY characteristics, the values are: >>> aSIFSTime = 16 >>> aSlotTime = 9 >>> aPHY-RX-START-Delay = 25 >>> >>> Therefore, ACKTimeout = 50 >>> >>> With a value of 1024, the ACKTIMEOUT is actually 25. >>> >>>> ?#define AR5K_INIT_ACK_CTS_TIMEOUT_TURBO ? ? ? ? ? ? ? ?0x08000800 >>>> ?#define AR5K_INIT_PROG_IFS ? ? ? ? ? ? ? ? ? ? 920 >>>> ?#define AR5K_INIT_PROG_IFS_TURBO ? ? ? ? ? ? ? 960 >>>> ?#define AR5K_INIT_EIFS ? ? ? ? ? ? ? ? ? ? ? ? 3440 >>> >>> From the 802.11-2007 Spec (9.2.10 DCF timing relations): >>> The EIFS is derived from the SIFS and the DIFS and the length of time >>> it takes to transmit an ACK Control frame at the lowest PHY mandatory >>> rate by the following equation: >>> EIFS = aSIFSTime + DIFS + ACKTxTime >>> where ACKTxTime is the time expressed in microseconds required to >>> transmit an ACK frame, including >>> preamble, PLCP header and any additional PHY dependent information, at >>> the lowest PHY mandatory rate. >>> >>> So, EIFS = 16 + 32 + 44 (ACKTX @ 24Mbps) = 92 >>> or 16 + 32 + 60 (ACKTX @ 6Mbps) = 108 >>> >>> The current value is 86, which I guess is a nice midpoint, but is >>> there any reasoning for this value? It should really be set to the >>> longest value of 108, as an EIFS is only used when we have no >>> information about a frame because it failed the checksum. >>> >>>> ?#define AR5K_INIT_EIFS_TURBO ? ? ? ? ? ? ? ? ? 6880 >>>> -#define AR5K_INIT_SIFS ? ? ? ? ? ? ? ? ? ? ? ? 560 >>>> -#define AR5K_INIT_SIFS_TURBO ? ? ? ? ? ? ? ? ? 480 >>>> +#define AR5K_INIT_SIFS_CLOCK ? ? ? ? ? ? ? ? ? 560 >>> >>> This is wrong, when you divide by the clock rate, you get a SIFS of 14, not 16. >>> >>> Cheers, >>> >>> Jonathan >>> >> >> All _CLOCK variables are left here temporarily until we clean up >> AR5210 stuff later, i know it's wrong ;-) >> Check out patch 16 that adds the new set_ifs_intervals, there we set >> them property, i just left them here so that code can still compile. >> > > Also have in mind that clock on g is not 40 but 44 (for b > compatibility) so 396 is correct here (396/44 = 9). > > Anyway we clean it up + these are only for AR5210 that we still don't > support, i just put a _CLOCK prefix so that they don't get mixed up > with the correct definitions. Check out patch 16, it should be fine > ;-) Oops, my bad on the clock rate. So, for A mode, it's 40, and G mode, 44, or 44 for both? If that's the case, then some of the values are still wrong (they're also wrong in patch 16): EIFS = 3440 / 44 = 78, which is way shorter than it should be! This gives ath5k an advantage over other 802.11-2007-compliant stations. ACK/CTS TIMEOUT = 1024/44 = 23, which is still much, much shorter than the required 50us, once again giving ath5k an advantage for medium contention. SIFS = 560/44 = 12, which is shorter again than the required 16us! Not saying your work is bad, by the way, but just that the values have been wrong in ath5k for a while. :) Cheers, Jonathan > > > -- > GPG ID: 0xD21DB2DB > As you read this post global entropy rises. Have Fun ;-) > Nick >