Date: Sat, 1 May 2010 16:36:15 +0000 (UTC) From: Warner Losh <imp@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r207472 - in head/sys: conf dev/ath/ath_hal/ar5212 Message-ID: <201005011636.o41GaFsK084343@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: imp Date: Sat May 1 16:36:14 2010 New Revision: 207472 URL: http://svn.freebsd.org/changeset/base/207472 Log: The Atheros AR71xx CPUs, when paired with the AR5212 parts, has a bug that generates a fatal bus trap. Normally, the chips are setup to do 128 byte DMA bursts, but when on this CPU, they can only safely due 4-byte DMA bursts due to this bug. Details of the exact nature of the bug are sketchy, but some can be found at https://forum.openwrt.org/viewtopic.php?pid=70060 on pages 4, 5 and 6. There's a small performance penalty associated with this workaround, so it is only enabled when needed on the Atheros AR71xx platforms. Unfortunately, this condition is impossible to detect at runtime without MIPS specific ifdefs. Rather than cast an overly-broad net like Linux/OpenWRT dues (which enables this workaround all the time on MIPS32 platforms), we put this option in the kernel for just the affected machines. Sam didn't like this aspect of the patch when he reviewed it, and I'd love to hear sane proposals on how to fix it :) Reviewed by: sam@ Modified: head/sys/conf/NOTES head/sys/conf/options head/sys/dev/ath/ath_hal/ar5212/ar5212_reset.c Modified: head/sys/conf/NOTES ============================================================================== --- head/sys/conf/NOTES Sat May 1 15:00:00 2010 (r207471) +++ head/sys/conf/NOTES Sat May 1 16:36:14 2010 (r207472) @@ -2091,6 +2091,15 @@ device ath_hal # pci/cardbus chip supp #device ath_rf5413 #device ath_ar5416 # AR5416 chips options AH_SUPPORT_AR5416 # enable AR5416 tx/rx descriptors +# All of the AR5212 parts have a problem when paired with the AR71xx +# CPUS. These parts have a bug that triggers a fatal bus error on the AR71xx +# only. Details of the exact nature of the bug are sketchy, but some can be +# found at https://forum.openwrt.org/viewtopic.php?pid=70060 on pages 4, 5 and +# 6. This option enables this workaround. There is a performance penalty +# for this work around, but without it things don't work at all. The DMA +# from the card usually bursts 128 bytes, but on the affected CPUs, only +# 4 are safe. +options AH_RXCFG_SDMAMW_4BYTES #device ath_ar9160 # AR9160 chips #device ath_ar9280 # AR9280 chips #device ath_ar9285 # AR9285 chips Modified: head/sys/conf/options ============================================================================== --- head/sys/conf/options Sat May 1 15:00:00 2010 (r207471) +++ head/sys/conf/options Sat May 1 16:36:14 2010 (r207472) @@ -771,6 +771,7 @@ AH_PRIVATE_DIAG opt_ah.h AH_NEED_DESC_SWAP opt_ah.h AH_USE_INIPDGAIN opt_ah.h AH_MAXCHAN opt_ah.h +AH_RXCFG_SDMAMW_4BYTES opt_ah.h # options for the Broadcom BCM43xx driver (bwi) BWI_DEBUG opt_bwi.h Modified: head/sys/dev/ath/ath_hal/ar5212/ar5212_reset.c ============================================================================== --- head/sys/dev/ath/ath_hal/ar5212/ar5212_reset.c Sat May 1 15:00:00 2010 (r207471) +++ head/sys/dev/ath/ath_hal/ar5212/ar5212_reset.c Sat May 1 16:36:14 2010 (r207472) @@ -283,6 +283,14 @@ ar5212Reset(struct ath_hal *ah, HAL_OPMO regWrites = ath_hal_ini_write(ah, &ahp->ah_ini_modes, modesIndex, 0); regWrites = write_common(ah, &ahp->ah_ini_common, bChannelChange, regWrites); +#ifdef AH_RXCFG_SDMAMW_4BYTES + /* + * Nala doesn't work with 128 byte bursts on pb42(hydra) (ar71xx), + * use 4 instead. Enabling it on all platforms would hurt performance, + * so we only enable it on the ones that are affected by it. + */ + OS_REG_WRITE(ah, AR_RXCFG, 0); +#endif ahp->ah_rfHal->writeRegs(ah, modesIndex, freqIndex, regWrites); OS_MARK(ah, AH_MARK_RESET_LINE, __LINE__);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201005011636.o41GaFsK084343>