Re: [Bug #12490] ath5k related kernel panic in 2.6.29-rc1

[Johannes Berg <johannes@sipsolutions.net>]

[-- Attachment #1: Type: text/plain, Size: 1498 bytes --]

On Tue, 2009-04-07 at 20:46 -0400, Bob Copeland wrote:
> On Tue, Apr 07, 2009 at 12:03:45PM +0200, Johannes Berg wrote:
> > 
> > > Anyway here's a patch to buy some time.  Johannes, would this be
> > > an acceptable band-aid for now?
> > 
> > This is going to make it print two warnings for one occurrence of the
> > problem though. I'd much prefer to put something into minstrel.c to
> > catch when _it_ returns bogus values and in that case return 0, that way
> > at least the bandaid is restricted to the code it patches up.
> 
> Ok, then maybe something like this?
> 
> I think this covers most of the bases: if the driver is returning crap or
> the ->cb gets corrupted, it'll get caught in tx_status.  If minstrel
> stuffs a -1 in the first slot for any other reason we'll get a warning
> before get_tx_rate and a hopefully valid index gets substituted.

Yeah, that looks nicer, thanks.

> diff --git a/net/mac80211/rc80211_minstrel.c b/net/mac80211/rc80211_minstrel.c
> index 3824990..1cf7152 100644
> --- a/net/mac80211/rc80211_minstrel.c
> +++ b/net/mac80211/rc80211_minstrel.c
> @@ -181,6 +181,11 @@ minstrel_tx_status(void *priv, struct ieee80211_supported_band *sband,
>  		if (ar[i].idx < 0)
>  			break;
>  
> +		if (WARN(ar[i].idx >= mi->n_rates,
> +		    "minstrel: invalid rate report %d (n=%d)\n",
> +		    ar[i].idx, mi->n_rates))
> +			break;

It might be easier to print this in hex -- that way things like 0x6b are
more apparent?

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]