[Tfug] Question tres: Overbuffering overkill

[John Gruenenfelder]

On Thu, Sep 04, 2014 at 01:59:48AM -0700, Nick Lopez wrote:
>
>On Sep 4, 2014 12:48 AM, John Gruenenfelder <jetpackjohn at gmail.com> wrote:
>>
>> And now for my third question.  This has to do with, I believe, an absurdly 
>> excessive amount of I/O buffering being done by the system as I attempt to 
>> copy approximately 9 GB of data from my super fast SSD to a microSD card 
>> (plugged into my laptops SD card slot via an adapter). 
>>
>  Sounds like the same criminally atrocious SD controller/driver I have in
>one of my laptops. Any time I write to a card the whole system chokes,
>sputters, and eventually completely freezes up until the IO is done. I think
>the last time I made the mistake of using it was to write a 500M n00bs image
>for a Raspberry Pi, and it took 10+ minutes in which the system was
>unresponsive. USB card readers are never a problem but this built-in PCI one
>has a knack for stalling the PCI bus for it seems like seconds at a time.

Hi Nick,

This made me curious as to who made the controller in my laptop (Samsung
Series 9 - np900x3b).  I couldn't find it in the lspci output so I checked
with lsusb, but it wasn't there either.

So, I inserted a card, and, ta-da!  It immediately appeared as a new
high-speed USB device attached to the ehci-pci hub.  Now, from lsusb I get
something useful:

Bus 004 Device 006: ID 05e3:0727 Genesys Logic, Inc. microSD Reader/Writer

That would certainly give a clue as to its lousy performance, but not to the
system unresponsiveness.  Both my wired and wifi are attached to the PCI bus
so I wouldn't think they should be adversely effected.  The Bluetooth 4.0
radio *is* attached internally via USB, but I wasn't using it at the time.

A bit of Googling didn't reveal anything Earth shattering.  A few people
complaining, a few people saying product X works fine and as a Genesys
controller chip, etc.  Most of the questions/chatter revolved around trying to
locate the thing in the first place since it doesn't appear on the system
until you put a card in it.

At any rate, USB traffic of *any* capacity should not be able to effect the
system in that manner.  And even if it could, why wasn't it the  *whole*
system?  It seems to preferentially hate my Intel/Centrino (Advanced-N 6230)
wifi chipset.  I've had some other random issues with the wifi though, too.
On rare occasions the connection will just sort of die out.  I have two SSIDs:
Bebop on 2.4 GHz and BebopCola on 5.5 GHz.  Net manager is set to prefer the
5GHz wifi-n network.  Sometimes, though, for reasons I don't know, it will
drop that connection and continue on its way connected to the 2.4 GHz
network.  It even ceases to see that it exists.  I have to turn wifi off, then
back on.  The both SSIDs appear and I can manually tell it to use the right
one.

I suspect this latter issue is probably some kernel regression and possibly
unrelated to the buffering/latency issue.  I'm using:

    Linux 3.14-2-amd64 #1 SMP Debian 3.14.15-2

as supplied in Debian/testing.


-- 
--John Gruenenfelder    Systems Manager, MKS Imaging Technology, LLC.
Try Weasel Reader for PalmOS  --  http://weaselreader.org
"This is the most fun I've had without being drenched in the blood
of my enemies!"
        --Sam of Sam & Max
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://tfug.org/pipermail/tfug_tfug.org/attachments/20140904/c583ad02/attachment.asc>