The Wifi Problem Will Be Found To Be Caused By:

[mali]

^ If I had known this before, I'd have put on my Sundays avatar, oh well...

 

[xyta]

I don't know if these links are helpful or known yet:


http://greekclaud.blogspot.com/2009_08_01_archive.html
http://groups.google.com/group/beagleboard/browse_thread/thread/25fcddf411c5da57/ec89bda9a40c0d2c?lnk=raot
http://markmail.org/message/h3i55qgkxcfqejpo
http://permalink.gmane.org/gmane.comp.hardware.beagleboard.user/1952

In my case , on OMAP3530 the SDIO xfer fails in 2 cases.

1. source buffer is not in DMA:able space.
2. source buffer address is not 4 buf aligned.

BTW, I replaced both the omap_hsmmc and SDIO driver with drivers I ported from an omap 2430 implementation and solved all performance problems (4-bit mode with interrupts now working). This is on the 2.6.29-rc3 kernel with B7 beagleboard. I haven't looked yet but I'll bet this implementation uses pre- allocated kmalloced buffers as you suggested.

Using kmalloced DMA:able buffer before calling sdio_memcpy_fromio/sdio_memcpy_toio is a workaround.

drivers/mmc/host/omap.c switches to PIO when DMA is not possible due to constraints.

drivers/mmc/host/omap_hsmmc.c always attempts DMA. I tried host->use_dma = 0, but it didn't work.

compare definition of function mmc_omap_prepare_data in both the drivers.

Finially I get a workable WL1251 driver with beagleboard. The tcp performance ~10 Mbps with iperf. The DMA workaround is work like Kalpesh's hint. But the size of DMA need to adjust.

 

[zevdawg]

MCP
End of line.

 

[ben_dash]

Finially I get a workable WL1251 driver with beagleboard. The tcp performance ~10 Mbps with iperf. The DMA workaround is work like Kalpesh's hint. But the size of DMA need to adjust.

If this is indeed applicable then it totally kicks ass!

 

[mhadf]

So from what those links say, it looks like it's things happening in the wrong order in the software and kernel with regards to this specific wifi chipset.

 

[kloplop321]

I love how the poll results look like an approximate normal curve/t distribution

 

[hdonk]

I love how the poll results look like an approximate normal curve/t distribution

*lol* Maybe that's because noone knows and is randomly guessing... It was either going to be an even distribution or a bell

 

[multiplex]

So is it currently slow but without packet loss, slow and has packet loss, or slow because of packet loss?
I'm curious what other tests have been done. Has the performance impacts of adding noise to the signal, or of a congested network been measured?

 

[droneb]

So is it currently slow but without packet loss, slow and has packet loss, or slow because of packet loss?
I'm curious what other tests have been done. Has the performance impacts of adding noise to the signal, or of a congested network been measured?

IIRC

it can be either slow with normal operation thru another connection method, or thru SDIO trying for fast but all it gets is a unacceptable amounts of packet loss leading to slow operation

 

[hdonk]

So now we know - the correct answer is 'PCB layout - Trace impedance'
Time to close this thread?

 

[Polossatik]

I'm still not convinced there where no Gremlins involved...

 

[darionapole]

so at the end noone won

 

[gahndalf]

So now we know - the correct answer is 'PCB layout - Trace impedance'
Time to close this thread?

No it was not the PCB layout.
It was caused by the gremlins that work for the company populating the boards.
They decided to make changes that caused the problem.

 

[Sphinxter]

Everyone who guessed, 'Too much sand in the vagina.', please collect your prize.

 

[dontspam2]

What is Open Pandora's WLAN chip's speed?


Even with protocol loss the minimum speed would be at minimum 500 kilobytes/second. Is everything OK with the hardware and/or drivers or what's the deal here?

The last information I read from Pandora Press was that someone could transfer stuff over the Pandora's WLAN at 500 kilobytes/second.
The wiki says pandora has 802.11 b/g supporting WLAN chip, which would mean 11 megabits/second and 54 megabits/second respectively.


I didn't want to start a new topic, because I think my question snugly fits here.

 

[WizardStan]

The wiki says pandora has 802.11 b/g supporting WLAN chip, which would mean 11 megabits/second and 54 megabits/second respectively.

I have yet to find any wireless device that gets anywhere close to the theoretical maximum. Most wifi G sticks I've tried top out at about 10Mb/s at best. The most recent information we have is that he started downloading a file at about 3Mb/s and it was still climbing when the download finished. This was through the internet. No one has done a local area test. It should be safe to say that you will get a sustained minimum of 3Mb/s if the network can handle it, although probably much higher than that even.

 

[notaz]

No one has done a local area test.

I did plenty of those tests, the speed is ~500+-200KiB. This is without using optimal resistor value (as those are difficult to solder for me).

 

[dontspam2]

The wiki says pandora has 802.11 b/g supporting WLAN chip, which would mean 11 megabits/second and 54 megabits/second respectively.

I have yet to find any wireless device that gets anywhere close to the theoretical maximum. Most wifi G sticks I've tried top out at about 10Mb/s at best. The most recent information we have is that he started downloading a file at about 3Mb/s and it was still climbing when the download finished. This was through the internet. No one has done a local area test. It should be safe to say that you will get a sustained minimum of 3Mb/s if the network can handle it, although probably much higher than that even.

I just tested the speed suboptimally (meaning it could be better) by starting a download of a large file from my WWW-server inside my network. The download stabilized at 1.9 - 2 megabytes/second. Test was done with Ubuntu computer downloading through 802.11g WLAN-connection. Client program was wget and server program was nginx.

I did plenty of those tests, the speed is ~500+-200KiB. This is without using optimal resistor value (as those are difficult to solder for me).

So the speed you witnessed was without "the resistor mod". And "the resistor mod" is the solution they came up with after the WLAN-issue was resolved, correct?

What WLAN-speed will the final Pandora have. The one that a customer receives?

 

[Tor]

The wiki says pandora has 802.11 b/g supporting WLAN chip, which would mean 11 megabits/second and 54 megabits/second respectively.

No, that's not what it means. It means 5 megabits/s and 22 megabits/s. Read all about it on wikipedia.
Besides, sticks seem to be even slower than that.
But I seem to get slightly more speed out of my N900 than what's been said about the wi-fi testing on the Pandora. We'll see later when things have been optimized.

-Tor

 

[dontspam2]

That post did very little to help solving this matter. Please, not like this!

A mobile device like N900 demonstrate low WLAN speeds because they have been put under heavy power management. I did a test and got ~650 kilobytes/second on my N900. Woo-hoo. Very informative. The bottom line is this: Since I get constant speed of ~2.5+ megabytes/second on my computer over WLAN 802.11g, I don't see the Pandora doing a very good job with the WLAN chip.


So I'm sorry for my bluntness, but what really is the problem here? The Pandoras are being built this very moment with a possibly (STILL) faulty hardware or drivers. Give some goddamn clarifications on the matter, please.