Re: [PATCH] mac80211: Keep CoDel stats per txq, export them in debugfs.

[Toke Høiland-Jørgensen]

Johannes Berg  writes:

>> @@ -137,18 +137,20 @@ static int aqm_open(struct inode *inode, struct
>> file *file)
>>  len += scnprintf(info->buf + len,
>>   info->size - len,
>>   "* vif\n"
>> - "ifname addr ac backlog-bytes backlog-
>> packets flows overlimit collisions tx-bytes tx-packets\n");
>> + "ifname addr ac backlog-bytes backlog-
>> packets flows drops marks overlimit collisions tx-bytes tx-
>> packets\n");
>
> It seems to me that you have to change the buffer length to take this
> into account?

Haven't run into issues with running out of buffer space in my testing.
But yeah, guess that could become an issue.

>>  list_for_each_entry_rcu(sdata, >interfaces, list) {
>>  txqi = to_txq_info(sdata->vif.txq);
>>  len += scnprintf(info->buf + len, info->size - len,
>> - "%s %pM %u %u %u %u %u %u %u %u\n",
>> + "%s %pM %u %u %u %u %u %u %u %u %u
>> %u\n",
>>   sdata->name,
>
> Why is it this way anyway? It'd seem nicer to move the content of this
> into the per-netdev subdirectories, and then it becomes a lot simpler
> code too (yes, at the expense of some userspace, but still)

Yeah, makes sense. Can do a larger reorg moving things into the
per-netdev and per-station directories instead.

>>   txqi->txq.ac,
>>   txqi->tin.backlog_bytes,
>>   txqi->tin.backlog_packets,
>>   txqi->tin.flows,
>> + txqi->cstats.drop_count,
>> + txqi->cstats.ecn_mark,
>>   txqi->tin.overlimit,
>>   txqi->tin.collisions,
>>   txqi->tin.tx_bytes,
>
> Do you really want to add these in the middle? Seems that if you add
> them at the end, you at least have *some* way of keeping this working
> with older versions?

Well I though they should be logically grouped with overlimits, and was
counting on no one actually parsing these yet. Guess if the information
is moved that becomes moot.


Will re-send; thanks for the feedback :)

-Toke
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] mac80211: Keep CoDel stats per txq, export them in debugfs.

[Johannes Berg]

> @@ -137,18 +137,20 @@ static int aqm_open(struct inode *inode, struct
> file *file)
>   len += scnprintf(info->buf + len,
>    info->size - len,
>    "* vif\n"
> -  "ifname addr ac backlog-bytes backlog-
> packets flows overlimit collisions tx-bytes tx-packets\n");
> +  "ifname addr ac backlog-bytes backlog-
> packets flows drops marks overlimit collisions tx-bytes tx-
> packets\n");

It seems to me that you have to change the buffer length to take this
into account?
 
>   list_for_each_entry_rcu(sdata, >interfaces, list) {
>   txqi = to_txq_info(sdata->vif.txq);
>   len += scnprintf(info->buf + len, info->size - len,
> -  "%s %pM %u %u %u %u %u %u %u %u\n",
> +  "%s %pM %u %u %u %u %u %u %u %u %u
> %u\n",
>    sdata->name,

Why is it this way anyway? It'd seem nicer to move the content of this
into the per-netdev subdirectories, and then it becomes a lot simpler
code too (yes, at the expense of some userspace, but still)

>    sdata->vif.addr,

This is also kinda pointless since it's easy to get elsewhere.

>    txqi->txq.ac,
>    txqi->tin.backlog_bytes,
>    txqi->tin.backlog_packets,
>    txqi->tin.flows,
> +  txqi->cstats.drop_count,
> +  txqi->cstats.ecn_mark,
>    txqi->tin.overlimit,
>    txqi->tin.collisions,
>    txqi->tin.tx_bytes,

Do you really want to add these in the middle? Seems that if you add
them at the end, you at least have *some* way of keeping this working
with older versions?

johannes
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html