2019-03-02 21:34:24

by Gustavo A. R. Silva

[permalink] [raw]
Subject: [mt76/mt7603/mac] Question about missing variable assignment

Hi all,

The following piece of code in drivers/net/wireless/mediatek/mt76/mt7603/mac.c
is missing a variable assignment before line 1058. Notice that there
is a potential execution path in which variable *i* is compared against
magic number 15 at line 1075 without being initialized previously
(this was reported by Coverity):

1055 out:
1056 final_rate_flags = info->status.rates[final_idx].flags;
1057
1058 switch (FIELD_GET(MT_TX_RATE_MODE, final_rate)) {
1059 case MT_PHY_TYPE_CCK:
1060 cck = true;
1061 /* fall through */
1062 case MT_PHY_TYPE_OFDM:
1063 if (dev->mt76.chandef.chan->band == NL80211_BAND_5GHZ)
1064 sband = &dev->mt76.sband_5g.sband;
1065 else
1066 sband = &dev->mt76.sband_2g.sband;
1067 final_rate &= GENMASK(5, 0);
1068 final_rate = mt7603_get_rate(dev, sband, final_rate, cck);
1069 final_rate_flags = 0;
1070 break;
1071 case MT_PHY_TYPE_HT_GF:
1072 case MT_PHY_TYPE_HT:
1073 final_rate_flags |= IEEE80211_TX_RC_MCS;
1074 final_rate &= GENMASK(5, 0);
1075 if (i > 15)
1076 return false;
1077 break;
1078 default:
1079 return false;
1080 }

My guess is that such missing assignment should be something similar
to the one at line 566:

i = FIELD_GET(MT_RXV1_TX_RATE, rxdg0);

but I'm not sure what the proper arguments for macro FIELD_GET should
be.

This code was introduced by commit c8846e1015022d2531ac4c895783e400b3e5babe

What do you think?

Thanks
--
Gustavo



2019-03-03 10:06:13

by Felix Fietkau

[permalink] [raw]
Subject: Re: [mt76/mt7603/mac] Question about missing variable assignment

On 2019-03-02 22:10, Gustavo A. R. Silva wrote:
> Hi all,
>
> The following piece of code in drivers/net/wireless/mediatek/mt76/mt7603/mac.c
> is missing a variable assignment before line 1058. Notice that there
> is a potential execution path in which variable *i* is compared against
> magic number 15 at line 1075 without being initialized previously
> (this was reported by Coverity):
>
> 1055 out:
> 1056 final_rate_flags = info->status.rates[final_idx].flags;
> 1057
> 1058 switch (FIELD_GET(MT_TX_RATE_MODE, final_rate)) {
> 1059 case MT_PHY_TYPE_CCK:
> 1060 cck = true;
> 1061 /* fall through */
> 1062 case MT_PHY_TYPE_OFDM:
> 1063 if (dev->mt76.chandef.chan->band == NL80211_BAND_5GHZ)
> 1064 sband = &dev->mt76.sband_5g.sband;
> 1065 else
> 1066 sband = &dev->mt76.sband_2g.sband;
> 1067 final_rate &= GENMASK(5, 0);
> 1068 final_rate = mt7603_get_rate(dev, sband, final_rate, cck);
> 1069 final_rate_flags = 0;
> 1070 break;
> 1071 case MT_PHY_TYPE_HT_GF:
> 1072 case MT_PHY_TYPE_HT:
> 1073 final_rate_flags |= IEEE80211_TX_RC_MCS;
> 1074 final_rate &= GENMASK(5, 0);
> 1075 if (i > 15)
> 1076 return false;
> 1077 break;
> 1078 default:
> 1079 return false;
> 1080 }
>
> My guess is that such missing assignment should be something similar
> to the one at line 566:
>
> i = FIELD_GET(MT_RXV1_TX_RATE, rxdg0);
>
> but I'm not sure what the proper arguments for macro FIELD_GET should
> be.
>
> This code was introduced by commit c8846e1015022d2531ac4c895783e400b3e5babe
>
> What do you think?
Thanks for reporting this. The fix is simpler than that, the check
should be: if (final_rate > 15)
I will send a fix.

- Felix

2019-03-03 15:06:53

by Gustavo A. R. Silva

[permalink] [raw]
Subject: Re: [mt76/mt7603/mac] Question about missing variable assignment



On 3/3/19 4:05 AM, Felix Fietkau wrote:
> On 2019-03-02 22:10, Gustavo A. R. Silva wrote:
>> Hi all,
>>
>> The following piece of code in drivers/net/wireless/mediatek/mt76/mt7603/mac.c
>> is missing a variable assignment before line 1058. Notice that there
>> is a potential execution path in which variable *i* is compared against
>> magic number 15 at line 1075 without being initialized previously
>> (this was reported by Coverity):
>>
>> 1055 out:
>> 1056 final_rate_flags = info->status.rates[final_idx].flags;
>> 1057
>> 1058 switch (FIELD_GET(MT_TX_RATE_MODE, final_rate)) {
>> 1059 case MT_PHY_TYPE_CCK:
>> 1060 cck = true;
>> 1061 /* fall through */
>> 1062 case MT_PHY_TYPE_OFDM:
>> 1063 if (dev->mt76.chandef.chan->band == NL80211_BAND_5GHZ)
>> 1064 sband = &dev->mt76.sband_5g.sband;
>> 1065 else
>> 1066 sband = &dev->mt76.sband_2g.sband;
>> 1067 final_rate &= GENMASK(5, 0);
>> 1068 final_rate = mt7603_get_rate(dev, sband, final_rate, cck);
>> 1069 final_rate_flags = 0;
>> 1070 break;
>> 1071 case MT_PHY_TYPE_HT_GF:
>> 1072 case MT_PHY_TYPE_HT:
>> 1073 final_rate_flags |= IEEE80211_TX_RC_MCS;
>> 1074 final_rate &= GENMASK(5, 0);
>> 1075 if (i > 15)
>> 1076 return false;
>> 1077 break;
>> 1078 default:
>> 1079 return false;
>> 1080 }
>>
>> My guess is that such missing assignment should be something similar
>> to the one at line 566:
>>
>> i = FIELD_GET(MT_RXV1_TX_RATE, rxdg0);
>>
>> but I'm not sure what the proper arguments for macro FIELD_GET should
>> be.
>>
>> This code was introduced by commit c8846e1015022d2531ac4c895783e400b3e5babe
>>
>> What do you think?
> Thanks for reporting this. The fix is simpler than that, the check
> should be: if (final_rate > 15)
> I will send a fix.
>

Great. Glad to help. :)

Thanks
--
Gustavo