index_to_rate[] (for @bcopeland fork)

[VanLaser]

@bcopeland

Hello,

Sorry to put the issue here, but I can't open one at https://github.com/bcopeland/wmediumd fork. Using kernel 4.9.8-1 (Arch Linux) built with mac802_hwsim, I encounter the following issues. I'm not very familiar with wmediumd, so I don't know exactly what the problem is, but, here are the "symptoms":

Grepping 'dmesg', the wlan interfaces, created by loading mac80211_hwsim, use rates configured by minstrel_ht. Here's a rc_stats sample from /sys/kernel/debug/ieee80211/phy7/netdev:wlan0/stations/02:00:00:00:01:00 (where phy7 is of course the id associated with wlan0):

best   __________rate_________    ________statistics________    ________last_______    ______sum-of________
rate  [name idx airtime max_tp]  [avg(tp) avg(prob) sd(prob)]  [prob.|retry|suc|att]  [#success | #attempts]
        1     0    9738    0.9       0.9      99.9      0.0     100.0   1     1 1           654   1282     
        2     1    4922    1.6       1.6      86.8      1.8     100.0   1     0 0            47   97       
        5.5   2    1858    4.7       4.5      86.9      1.8     100.0   2     0 0           126   203      
       11     3     982    9.1       8.3      82.4      1.8     100.0   4     0 0          4299   4412     
        6     4    1648    5.3       5.1      86.8      1.8     100.0   3     0 0           203   268      
        9     5    1112    8.0       7.3      82.3      1.8     100.0   4     0 0            61   190      
       12     6     844   10.5      10.5      92.4      1.7     100.0   5     0 0            54   183      
   D   18     7     576   15.5      14.9      86.6      1.6     100.0   5     0 0            35   187      
  C    24     8     440   20.4      18.6      82.2      1.7     100.0   6     0 0          1292   1464     
A   P  36     9     308   29.1      29.1      99.9      0.0     100.0   6     0 0          5392   7739     
 B     48    10     240   37.3      19.9      48.0      0.8      33.3   6     0 0           915   1648     
       54    11     216    0.0       0.0       0.0      0.0       0.0   6     0 0             0   222   

As you can see, the rates and indexes used are those from the first 2 columns.

Now, when inspecting the function queue_frame() from wmediumd.c (bcopeland fork), it seems to me that the array index_to_rate[], that gets consulted each time an index must be translated to a rate, is out of place - by logging the index value, I get for example values such as 9 etc. which are higher than the dimension of the array. So, my question is, is the following modification the 'right' one to make?

static int index_to_rate[] = {
	10, 20, 55, 110, 60, 90, 120, 180, 240, 360, 480, 540
};

Since I'm not familiar with wmediumd, I have no clue about what causes this new set of rates to kick in, instead of the 'old' one. When does this happen - all the time? or does it depend on minstrel_ht (vs minstrel)? Even in this case, I'm not entirely sure the index is not 'offset' by 1 - if I log the rate index for an acknowledged frame (rate_idx in queue_frame(), after the loop), it seems to be higher (by 1) than the one selected as ideal (A P) in rc_stats (e.g. I get index 10 as the most used, but rc_stats shows that index 9 was the ideal one).

Thanks

[bcopeland]

No idea why you cannot open an issue there, but in any case: Yes, your modification looks correct, *if* you are operating a 2.4 GHz network. The rateset that wmediumd uses by default assumes a 5 GHz 802.11a rateset, the reason being that at the time I originally wrote it, I didn't have any good way of estimating anything . The other reason is that I'm not aware of an easy way to tell from userspace what the rateset actually is. So another option is just to pick 5 GHz frequencies in your test environment.

…

On Wed, Feb 15, 2017 at 04:21:53AM -0800, VanLaser wrote: @bcopeland Hello, Sorry to put the issue *here*, but I can't open one at https://github.com/bcopeland/wmediumd fork. Using kernel 4.9.8-1 (Arch Linux) built with mac802_hwsim, I encounter the following issues. I'm not very familiar with **wmediumd**, so I don't know exactly what the problem is, but, here are the "symptoms": Grepping 'dmesg', the wlan interfaces, created by loading mac80211_hwsim, use rates configured by **minstrel_ht**. Here's a **rc_stats** sample from _/sys/kernel/debug/ieee80211/phy7/netdev:wlan0/stations/02:00:00:00:01:00_ (where **phy7** is of course the id associated with **wlan0**): As you can see, the rates and indexes used are those from the 2 columns. Now, when inspecting the function **queue_frame()** from _wmediumd.c_ (**bcopeland** fork), it seems to me that the array **index_to_rate[]**, that gets consulted each time an index must be translated to a rate, is out of place - by logging the index value, I get for example values such as 9 etc. which are higher than the dimension of the array. So, my question is, is the following modification the 'right' one to make? ``` static int index_to_rate[] = { 10, 20, 55, 110, 60, 90, 120, 180, 240, 360, 480, 540 }; ```

-- Bob Copeland %% http://bobcopeland.com/