From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pavel Roskin Date: Wed, 15 Feb 2012 13:06:09 -0500 Subject: [ath9k-devel] kernel panic with ath9k In-Reply-To: References: <20120204104848.28186td6kg8w88cg-cebfxv@webmail.spamcop.net> <20120207170009.612f8d65@mj> <4F32CCB8.8060600@openwrt.org> <20120208175400.40c00ca8@mj> <20120210230615.hc9q4avo74w0w808-cebfxv@webmail.spamcop.net> <20120215092313.u26y5kweo0w4o0c4-cebfxv@webmail.spamcop.net> Message-ID: <20120215130609.476e4e0e@mj> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ath9k-devel@lists.ath9k.org On Wed, 15 Feb 2012 21:18:36 +0530 Mohammed Shafi wrote: > if we refer rate_control_get_rate in rate.c of mac80211 > > for (i = 0; i < IEEE80211_TX_MAX_RATES; i++) { > info->control.rates[i].idx = -1; > info->control.rates[i].flags = 0; > info->control.rates[i].count = 1; > } Thanks for finding this code! That's another sign that my patch is not hiding an underlying problem, such as memory corruption. > why count = 1 not 0 ? if count has to be 0 then we should fix in > mac80211 also. though the doc in mac80211 says idx -1 means invalid > rate, did not say if count =0 invalid > but anyway the rate is tried once ? > > *now the drivers assumption of count=0 for invalid is not correct i > think ministrel checks for both index and count in > 'minstrel_ht_txstat_valid' while ath9k does not before your patch We should remove the checks for count=0 eventually. But it's not a change for stable kernels. -- Regards, Pavel Roskin