wifi driver + kernel updated - please test


Yes its 290MB...i wrongly written it...the most important thing is that downloaded pnd works. :)
 
Just tested without powersaving using this

sudo iwconfig wlan0 power off rate 54M
and successfully downloaded a 250MB file (naev.pnd).
You and Dunny can try
Code:
echo 10000 > '/sys/bus/mmc/devices/mmc2:0001/mmc2:0001:1/ps_rate_threshold'
This will automatically disable powersave when the rate goes above 10K/s (instead of default 100K/s) and turn it back on when idle, so it doesn't drain your battery for nothing.
 
Last edited by a moderator:
CC Pandora here. Wifi was fine for me as far as I can remember, never made any lan checks, but it managed to get the full 3.5 Mbit/s of that crappy internet I have.
Now, after updating I can connect fine and like instantly to the network, but I get barely any data out of it. Only managed to load like half of a page slowly and then nothing, not even in lan (connetion to the rouer was never interrupted, though).

Everything works fine with power saving turned off, but I had it on before updating.

The 10K/s based powersaving disabling seems to work for me too.

If of relevance, my FritzBox said the Pandora was connected with 1 Mbit/s after the updated and 36 Mbit/s with power saving turned off/the rate fix.

edit: Or it only seemed so. Didn't last long, kinda reverted to not loading anything. I don't know either. As things stand, only disabled power saving seems to work reliably for me after the update.
 
Last edited by a moderator:
Just tested without powersaving using this

sudo iwconfig wlan0 power off rate 54M
and successfully downloaded a 250MB file (naev.pnd).
You and Dunny can try

echo 10000 > '/sys/bus/mmc/devices/mmc2:0001/mmc2:0001:1/ps_rate_threshold'

This will automatically disable powersave when the rate goes above 10K/s (instead of default 100K/s) and turn it back on when idle, so it doesn't drain your battery for nothing.
Another issue addressed by the command is the rate which seems to change randomly unless the wifi signal is really strong. Is it possible to tune the rate changing algorithm somehow so it doesn't pick 1Mb speed after a minor signal strength drop?

This leave out a question which rate value is optimal. Setting rate to 54M improves my wifi a lot when I use it outside, but I suppose there is an intermediate value between 1 and 54 which offers a better compromise. Do you know anything about it?
 
Last edited by a moderator:
Just tested without powersaving using this

sudo iwconfig wlan0 power off rate 54M
and successfully downloaded a 250MB file (naev.pnd).
You and Dunny can try
echo 10000 > '/sys/bus/mmc/devices/mmc2:0001/mmc2:0001:1/ps_rate_threshold'
This will automatically disable powersave when the rate goes above 10K/s (instead of default 100K/s) and turn it back on when idle, so it doesn't drain your battery for nothing.
Tried both - the first executes without errors, but the second throws a "No such file or directory" error. Time to reflash again? Can't use internet at all now, PNDManager complains that sync fails every time and can't download PNDs.

Edit: Just tried the "Upgrade Pandora OS" option again and there's no updates. I guess the repo must be down :)

Edit2: Repo is fine, just ran PNDManager again and it's working now. Just leaves the file/directory not found issue then!

D.
 
Last edited by a moderator:
^ did you try to run the second command with sudo ? 
 
^ did you try to run the second command with sudo ?
I did indeed.

I'll try with the backslash escapes.

Edit: Worked by using 'cd /sys', 'cd bus' etc etc. Then

echo 10000 > ps_rate_threshold

So now off to test download speeds.

Edit2: Ok, no noticeable change in speed - though the speed seems constant at 80-ish KB/sec. Still not gobbling up all the available wifi though, so that's better. It's a fibre connection btw :)

D.
 
Last edited by a moderator:
Hmm quite mixed feedback on the latest changes. But it gives an impression that latest changes help everyone except (some?) CC users?
 

edit: Or it only seemed so. Didn't last long, kinda reverted to not loading anything. I don't know either. As things stand, only disabled power saving seems to work reliably for me after the update.
I think you, Farox and Dunny are missing the resistor fix, or have the fix improperly applied by Craig's team, so it's a hardware issue. See the image MWeston sent me back in 2010:
wifi_fix_stage1.JPG


I could make another file in /sys/bus/... (or maybe /proc/pandora/ for easier use) that would enable the old behavior of toggling powersave on each packet. I don't have other ideas as the missing/bad resistor can't be detected in software.
 

Another issue addressed by the command is the rate which seems to change randomly unless the wifi signal is really strong. Is it possible to tune the rate changing algorithm somehow so it doesn't pick 1Mb speed after a minor signal strength drop?
 
This leave out a question which rate value is optimal. Setting rate to 54M improves my wifi a lot when I use it outside, but I suppose there is an intermediate value between 1 and 54 which offers a better compromise. Do you know anything about it?
The rate is fully controlled by binary blob. Are you sure it's not power saving off alone doing this?
 

Is it normal that the kernel git still on 3.2.64 ?
Fixed.
 
@Notaz, I'm having a hard time ignoring that mess... my OCD is going into overdrive.
 
Well it looks like the resistor wasn't replaced there.

Here is how my CC looks, I think it was replaced by ED himself:

(edit: actually it's a stacked resistor on top, like MWeston recommended. These things are hard to see and photograph)

pandora_cc_wifi_resistor.jpg


They should have switched to correct resistor for population at some point, but I don't know when.

I'll try to ask ED.
 
Last edited by a moderator:
Update seems to work great for me. Connection has been very strong, and as mentioned by others, my Pandora does not seem to turn other devices' connections to a crawl anymore. (It wasn't all devices, but just my wife's computer)

Thank you, Notaz!
 
Just wanted to add my datapoint to the anecdotal evidence: on my original CC Pandora's, #008 and #045, after this update - things 'seem' a lot smoother and the connection is less likely to be interrupted.  I've sort of gotten into the habit of hitting a key on my Pandora every few minutes when I'm connected to it (I ssh in to use cdevtools or codeblocks onboard as a dev environment) .. well it seems like I don't need to do this any more, which means the connection is being kept alive.  I don't know any further details than that, just wanted to report: thank you notaz, and seems like this made things a lot more stable/smooth/easier to depend on during those big compiles ..
 
I ssh in to use cdevtools or codeblocks onboard as a dev environment
I do pretty much the same, though I use g_ether driver and have a stable 480Mb/s connection via USB2 instead of a less stable 54 Mb/s wifi which is additionally busy with other traffic. This doesn't change a lot if all you use is a terminal, but most apps I work with have GUIs, so it's tempting to run them in a remote X session. That's where additional bandwith comes in handy.
 
Another issue addressed by the command is the rate which seems to change randomly unless the wifi signal is really strong. Is it possible to tune the rate changing algorithm somehow so it doesn't pick 1Mb speed after a minor signal strength drop?


This leave out a question which rate value is optimal. Setting rate to 54M improves my wifi a lot when I use it outside, but I suppose there is an intermediate value between 1 and 54 which offers a better compromise. Do you know anything about it?
The rate is fully controlled by binary blob. Are you sure it's not power saving off alone doing this?
I'm pretty sure that increasing the rate alone helped a lot. Might have been still related to power saving though, since 1Mb/s is not enough to go above the PS threshold, which in turn results in weak signal, keeping the rate at 1 Mb/s forever.
 
Back
Top