Re: [Gumstix-users] overo com reliability

[Dan N. <Dan...@ro...>]

Hi Jeff,

Can you point me to where I can find the Gumstix 2.6.32 source or patch?

Thanks

Dan

-----Original Message-----
From: Jeff DeFouw [mailto:je...@gr...] 
Sent: Tuesday, March 20, 2012 10:03 AM
To: gum...@li...
Subject: Re: [Gumstix-users] overo com reliability

On 3/19/2012 9:26 AM, mmeisner wrote:
> We have taken a prototype to a few trade shows, and other than a few
hangs
> where the overo would stop responding (not acceptable for production),
it
> has fared reasonably well.  We did however talk to a company who got
through
> a full development cycle with the overo, and in production, had major
> reliability problems.  I will be following up this week for more
details.
>
> I have read that finding a very stable linux kernal is key because
certain
> builds cause issues.  Other than the common bluetooth and wifi issues,
has
> anyone experienced any major problems in production?  Were they a
result of
> hardware or software?

Starting in 2010, our first 80 or so Water boards had many hardware
issues 
that were discovered during testing.  The biggest problem was that the
USB 
Host would randomly die (a permanent death).  It might happen before
testing, 
during testing, or some time after shipping to the customer.  Many have
come 
back to be replaced when the customer couldn't load an update.  We've
also had 
many of the early boards freeze up after a while with the customer, some

running for rapidly decreasing periods until finally they won't boot at
all. 
The failure rate in later batches (220+ boards) has been much better.

The main problem we have right now is the OpenGL graphics driver.  We
use 
OpenGL for everything, directly drawing our own user interface, and
share that 
code base with earlier products on different processors and GPU's.  The
driver 
version we're using often locks up right after booting, and sometimes 
completely fails to draw some textures or triangles depending on the
frame and 
how the calls are ordered.  Once the user successfully gets to a main
screen, 
the system is stable.  We do a test during boot to make sure the driver 
renders properly, but it doesn't catch many cases.  The newer drivers
look 
like they might fix those two problems, except that they often fail to
draw 
anything inside a scissor window, and would also require more kernel
work.

I put together our 2.6.32 kernel in January 2010, using linux-omap and
patches 
from OE, mailing lists, and other trees.  The kernel has been very
stable. 
The special (non-mainline) ARM kernels were a huge, ugly mess, with
several 
versions of patches being reverted and reapplied all the time, and
different 
trees didn't have all the same patches, even if they were important.
Mainline 
support is always getting better, but I see this is still a problem.

-- 
Jeff DeFouw <je...@gr...>
Programmer
Grand Rapids Technologies

------------------------------------------------------------------------
------
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here 
http://p.sf.net/sfu/sfd2d-msazure
_______________________________________________
gumstix-users mailing list
gum...@li...
https://lists.sourceforge.net/lists/listinfo/gumstix-users