On 3/17/20 8:40 AM, Kalle Valo wrote:
Tony Chuang <yhchuang@xxxxxxxxxxx> writes:
// Add Johannes for commenting on adding another nl80211 commands
Kalle Valo <kvalo@xxxxxxxxxxxxxx> writes:>
Tony Chuang <yhchuang@xxxxxxxxxxx> writes:
Kalle Valo <kvalo@xxxxxxxxxxxxxx> writes:
<yhchuang@xxxxxxxxxxx> writes:
From: Yan-Hsuan Chuang <yhchuang@xxxxxxxxxxx>
It is useful to fix the bit rate of TX packets. For example, if
someone is measuring the TX power, or debugging with the issues
of the TX throughput on the field.
To set the value of fixed rate, one should input corresponding
desc rate index (ex, 0x0b for DESC_RATE54M to fix at 54 Mbps).
Set a value larger than DESC_RATE_MAX will disable fix rate, so
the rate adaptive mechanism can resume to work.
Example,
To fix rate at MCS 1:
echo 0x0d > /sys/kernel/debug/ieee80211/phy0/rtw88/fix_rate
To not to fix rate:
echo 0xff > /sys/kernel/debug/ieee80211/phy0/rtw88/fix_rate
To know which rate was fixed at:
cat /sys/kernel/debug/ieee80211/phy0/rtw88/fix_rate
Signed-off-by: Yan-Hsuan Chuang <yhchuang@xxxxxxxxxxx>
No way, debugfs is not a method for working around nl80211 and doing
whatever idea you come up with. The goal is that we have a generic
nl80211 command for all generic actions, like this one. And I think we
already have an nl80211 command for fixing the tx rate, right?
No, as I can see, there's no suitable nl80211 command that can achieve
what I want. If you are saying about
NL80211_CMD_SET_TX_BITRATE_MASK,
it's used to allow some rates. But actually the firmware has its own rate
adaptive mechanism, so mask out the other rates does not mean the rate
left will be chosen. Moreover, the hardware will choose a lower bit rate
when retry, then the TX rate is not fixed at all. So the debugfs can disable
the firmware's RA mechanism, also disable the TX rate fall back when retry.
Both of them cannot be done by setting TX bitrate mask.
I'm confused, here you talk about firmware implementation etc but I'm
just talking about replacing the fix_rate debugfs file to an nl80211
command (for providing the fix_rate value). Can you clarify more why you
think nl80211 is not suitable?
Oops, I thought that you wanted me to use the existing nl80211
command.
Either use an existing nl80211 command or add a new one if needed. For
me most important is that we don't add hacks to debugfs just for
avoiding using nl80211.
Now I know that you think we can add a new nl80211 command to help
drivers to fix the TX bitrate if necessary. If adding another nl80211
command for that is acceptable, I can work on this. But I need
Johannes's comment if it's better to add a new nl80211 command or to
expand the existing command (ex. NL80211_CMD_SET_TX_BITRATE_MASK).
_Why_ is NL80211_CMD_SET_TX_BITRATE_MASK not suitable for you? You keep
saying that but I have still figured out why exactly you think so.
Please clarify this in detail.
It looks like that adding a new nl80211 command will be better for me
as expanding the existing one would have great impact on the already
distributed drivers/user-tools.
What kind of great impact are you talking about? Please be specific so
that we don't need to guess.
At least with ath10k, the issues I found were that nl80211 doesn't like it
when you try to disable all legacy rates (and force frames out at 54Mbps
encoding, for instance).
I'm not even sure upstream ath10k will even let you set a single rate
using normal API now. Have you tried it?
Another problem is that to keep a connection alive, you probably want mgt
and null-func frames to go out normal and only have the firmware use a particular MCS
for data frames.
Lots of reasons to want a low-level hack for this sort of thing.
Thanks,
Ben
--
Ben Greear <greearb@xxxxxxxxxxxxxxx>
Candela Technologies Inc http://www.candelatech.com
