USB - the Elephant in our Room

[abishur]

Gordon, first of all thank you for having taken the time to write a detailed summary.

It is strange that my pi developed the symptoms of the split transaction after running for 2 months
See (http://www.raspberrypi.org/phpBB3/viewt ... 41#p195941).
NOTE: To spare everyone to read, it I use this pi as headless.. no usb other than internelt ethernet card.. ethernet dies...
But you are already aware of this since you asked me to RMA it in a private message so you can test it.
I had noticed that you have requested for help since back in July. Hence my suggetions.
People with the peculiar skill set like that you are after, are generaly busy, you need to look for them. Or at least advertise better... Definitely not in a 25 pages discussion. Sorry Jamesh but I have worked in support for many years and if there is a constant amongst ITs, consumers, engineers ... No one read long threads. emails... Because everyone is busy !

Also, I was just reading this excellent magazine the MagPI and ...
I think you should all talk to the head of the foundation, there is a clear divergence in views.

[snip]

tl;dr (kidding, kidding)

If you legitimately want to discuss... well, whatever your point was attempting to be, *please* make a new thread where it can appropriately be discussed.

[rmistero]

For the people interested in the point I raised, please go here :

http://www.raspberrypi.org/phpBB3/viewt ... 83#p199183

Sorry again for the disturbance of this thread.

[abishur]

For the people interested in the point I raised, please go here :

http://www.raspberrypi.org/phpBB3/viewt ... 83#p199183

Sorry again for the disturbance of this thread.

Thanks!

[chrisw2]

... I've many times asked for people to come forward with expertise in USB hosts or USB hubs or split transactions or EHCI and have so far had a flood of '0' people to my door...

Hi Gordon,

First, many thanks for all the work you have put into this.

Regarding expertise (and at the risk of teaching my grandmother to suck eggs): have you tried asking on mailing lists where that sort of expertise from the open source community might hang out?

eg: http://lists.infradead.org/pipermail/linux-rpi-kernel/
or maybe http://www.linux-usb.org/mailing.html

[insanityideas]

Hey guys thought this thread could do with some good news!! My keyboard seems to be behaving itself much better, as evidenced if you compare this post with my previous ones with missing letters and stuck keys all over the place!! I think I still loose the odd mouse click and occasional key, but its much more usable now. Can't say which update fixed it, as none of the apt-get upgrades seemed to have files that sounded particularly "USB" related, but one of them was an improvement.

Whilst I know Gordon is still working hard on a "better" long term fix I would like to hank him for the improvements thus far. It is now usable again, and the other optimisations keep speeding the pi up on a regular basis.

One of the reservations I have about using the Pi in a @Codeclub teaching context was the potential for keyboard screwups and other system responsiveness issues to be very distracting for the kids. I think its almost ready for primetime with that, especially compared with getting python put on school computers in a working state!!

[thradtke]

[...] I would like to hank him [...]

Hey!



You *could* have a trouble-free keyboard already. Since I strikly have seperated high and full speed devices, I haven't noticed any glitches (and I should have noticed as I'm using the rootfs on a USB HDD). But then, the whole matter is unfortunately beyond me. E.g. I don't understand one word of Gordons explanation.

[mic34]

Greetings to all, new here. Many thanks to Gordon for his great work trying to fix the Pi's USB.

[Since I strikly have seperated high and full speed devices, I haven't noticed any glitches (and I should have noticed as I'm using the rootfs on a USB HDD). But then, the whole matter is unfortunately beyond me.

Thradtke or anyone, am I correct in understanding that separating high speed (480 Mbps) and full speed (12 Mbps) devices consists in inserting an extra high speed hub between the Pi's integrated (high speed) hub and the full speed devices?

And if so, will any high speed hub do?

[thradtke]

Thradtke or anyone, am I correct in understanding that separating high speed (480 Mbps) and full speed (12 Mbps) devices consists in inserting an extra high speed hub between the Pi's integrated (high speed) hub and the full speed devices?

And if so, will any high speed hub do?

Sorry, I have not tried that, but think a full speed hub will do also. My setup consists currently of a fullspeed hub connected to one port on the Pi and a 1.1 mouse directly connected to the other port. The external hub is populated with keyboard and USB HDD, both USB 2.0.

[mic34]

Thanks, can you share the ref of your fullspeed hub?

[thradtke]

It's a LogiLink 4-port hub, 2A. Its adapter powers everything in my setup.

[mic34]

It's a LogiLink 4-port hub, 2A. Its adapter powers everything in my setup.

Are you sure it's full speed ie usb 1.1? I googled that and only found usb 2.0 (presumably high speed) devices.

[gsh]

Using a full speed hub will not improve the problem at all... The problem is that split transactions are going awry and this is done when a high speed port talks to a low or full speed hub.

The 9512 chip on the RPi contains a high speed hub so the only way to get rid of the split transactions is to switch that to full speed (i.e. using the speed=1 cmdline argument)

It'll make things slow but will fix the issues with split transaction scheduling...

Gordon

[mic34]

Using a full speed hub will not improve the problem at all... The problem is that split transactions are going awry and this is done when a high speed port talks to a low or full speed hub.

The 9512 chip on the RPi contains a high speed hub so the only way to get rid of the split transactions is to switch that to full speed (i.e. using the speed=1 cmdline argument)

It'll make things slow but will fix the issues with split transaction scheduling...

Gordon

Thanks Gordon I tried the speed=1 trick, it did fix the crackling noise for stereo WAVs on my cm6206 based usb sound card, but then it wouldn't play more than two audio channels

I tried instead to insert a high speed hub, hoping it would talk high speed to the Pi and full speed to the soundcard but the crackling was still there even on stereo files. Why didn't this work, does the Pi still have to do the splitting in this case?

[thradtke]

Are you sure it's full speed ie usb 1.1?

No. As you might have guessed, I mixed up the terms. It's a 2.0 hub connecting to 2.0 devices and a 1.1 mouse straight on the Pi's hub. That should be correct now.

@Gordon - why does it work for me? It even worked with the 1.1 mouse on the 2.0 hub and the 2.0 keyboard on the Pi. Those devices just cannot be mixed on a single hub. I even have a weak setup currently with the Pi also powered through the hub.

[insanityideas]

@Gordon - why does it work for me? It even worked with the 1.1 mouse on the 2.0 hub and the 2.0 keyboard on the Pi. Those devices just cannot be mixed on a single hub. I even have a weak setup currently with the Pi also powered through the hub.

I think you probably just got lucky.... computer gremlins are funny like that, sometimes they come, sometimes they go.

I havn't used the speed=1 fix, but have noticed an improvement recently after some updates.

[vassilij123]

well after reading a lot of posts,
this is non other than a design fault in the usb circuit
i have just today bought a belkin top line 4port hub powered,5v 2.1a,
when i plug in the usb when powered into the pi.all hell breaks loose
it has stated to corrupt the config file under boot conditions,
so i am constantly re-installing the OS
to RPI,i say this,in your considerable haste to get this great concept into production,you forgot to assign
a decent electrical engineer to design,you need to shoot whoever let it go into full production
without considerable testing first,
and yes £30.00 is a fortune to some(certainly not geeks who seem these days not to have a normal life and are able to save squillions of squids) people on these forums listen,stop trying to sound like you have all the answers.we(the general public)dont need or are able to desolder the board and start again from scratch.pi people get it right,and stop the back slapping,the rest of the world is watching and have just taken a large breath to start laughing there socks off at a major fail,yet again in the U K,

[itimpi]

It sounds to me as if your Belkin hub is faulty and is back-feeding power to the Pi (which is against the USB spec).

[MaxK1]

Yes, but faulty top of the line Belkin hub...

I have 3 IOgear hubs, all work fine. (And an ancient Belkin USB 2.0 hub which also works fine. Must not be "top of the line"... wifi, hdd, wired and wireless keyboards, Flash drives all work fine. FTDI serial adapter may or may not be working, need to look at that some more one of these days...

[jamesh]

well after reading a lot of posts,
this is non other than a design fault in the usb circuit
i have just today bought a belkin top line 4port hub powered,5v 2.1a,
when i plug in the usb when powered into the pi.all hell breaks loose
it has stated to corrupt the config file under boot conditions,
so i am constantly re-installing the OS
to RPI,i say this,in your considerable haste to get this great concept into production,you forgot to assign
a decent electrical engineer to design,you need to shoot whoever let it go into full production
without considerable testing first,
and yes £30.00 is a fortune to some(certainly not geeks who seem these days not to have a normal life and are able to save squillions of squids) people on these forums listen,stop trying to sound like you have all the answers.we(the general public)dont need or are able to desolder the board and start again from scratch.pi people get it right,and stop the back slapping,the rest of the world is watching and have just taken a large breath to start laughing there socks off at a major fail,yet again in the U K,

Why, thank you for your constructive criticism. However, there doesn't appear to be a design fault in the USB circuitry - its more of a SW problem at this stage - so unfortunately we won't be able to shoot anyone right now. If we do find a fault, we will of course take the appropriate person outside and shoot him. And his children, just in case the problem is genetic.

As for geeks and their money, and their lives, wuh? I think its a bit insulting to geeks for you to say that. Please apologise to geekkind.

In precis - chill dude. We are fixing the problems! Most people don't even have problems, depending on the combination of USB devices used.

[jamesh]

Had to delete a reply post for quite vitriolic abuse. Seem English humour doesn't translate well.

Can I suggest that posters refrain from being really REALLY abusive on their first or seconds posts. We welcome reports of faults and criticism when presented in a civil manner. We DO NOT allow abusive posts, or diatribes. There are ways to get your point across; being abusive, insulting, rude and isn't one of them. I would also suggest that people try and get their facts straight before launching off on nasty posts.

[hermanhermitage]

Had to delete a reply post for abuse.

I caught a glimpse of it. I can understand the fellows frustrations. I picked up 5 of the 512MB R PI's for xmas gifts for nieces and nephews and having had no USB issues before they have suddenly come home to roost with a particular brand of keyboard and mouse. Its a pain but I'll live Younger kids I can imagine might be frustrated by it.

I've been burnt by the realities of USB standard versus the real world before on a production run of 100k+ devices in an almost career ending debacle.

The most positive suggestion I can think of is - How about a "Tested for Pi", or "Approved for Pi" certification program? Structure the right way it could be another revenue stream for the foundation.

[jamesh]

Whilst frustrations are fine, that level of rudeness isn't. Even his first post was bordering on unacceptable, which is a bizarre way of approaching someone you want to fix something for you.

We do have a list of tested equipment on the WIki (not sure how up to date though), and we are working hard on fixing the remaining issues. It's difficult doing a lot of device testing without any full time employees dedicated to the task, and it's also expensive if you want lots of coverage, so the majority effort is in to the fix itself.

Which brand keyboard/mouse just for future reference?

[thradtke]

If it is of interest: My second Pi connects to a Hama RF2200 wireless desktop w/o problems.

Re deleted post: Haven't read such a personal attack for a long time.

[MaxK1]

Then there are devices in combination with what other devices and what engineering changes were/were not present on the "test" device. A pretty daunting task....

[jamesh]

Then there are devices in combination with what other devices and what engineering changes were/were not present on the "test" device. A pretty daunting task....

Indeed, which is why we think a big effort to fix as much of the underlying issues is the best approach; meanwhile we are forced to reply on community involvement to document what does and doesn't work. After all, there's lots more people out there than in here, with access to a vast array of USB equipment.