Help - Search - Members - Calendar
Full Version: sustained writes to SD (is there a solution?)
OESF Forums > Distros, Development, and Model Specific Forums > Model Specific Forums > Cxx0 Hardware
raduga
There have been several threads posted over the past year and a half, from users with various sizes and brands of SD card that for some reason or another, just WONT work with sustained write.

Some suggestions have been offered:

* Patching the kernel to allow async (or FS_SYNC=N) access to SD
* mounting the partition with -o async options
* using "updated" sd drivers (a new mmc_sd.o was posted to ZUG some time ago)
* trying a different brand, different size of SD card
* mounting as a loopback device
* format as fat/ext2

For people with SD write issues, which of these has worked?
In my limited experience, with a new Lexar 256mb SD,
both cacko qt (1.21) and pdaxrom (1.0.5) kernels showed problems
mounting with and without -o async didn't help
the patched mmc_sd.o does not initialize properly, and doesn't load right
both fat and ext2 seem to show similar issues

has anyone found a way (reliably) to get their flaky SD working with sustained writes?

What SD media do you use (brand, size) and have you had any problems so far?

Sorry for the horrible ugly poll sad.gif [there should be a better way to condense this, or at least allow more poll options]
xamindar
Sorry, would you mind explaining what a sustained write is? _ow is it different from a normal write? Or are you just talking about writing large files?
raduga
QUOTE
Sorry, would you mind explaining what a sustained write is?  _ow is it different from a normal write?  Or are you just talking about writing large files?
Yeah, pretty much. Writing data of more than a few megabytes at a time seems to cause some SD cards to generate I/O error, and go offline.

What model of SD card do you use (with what ROM/kernel) and have you had any problems?
xamindar
I have just recently bought a Fujifilm 256Mb sd card and so far it works great. I haven't formatted it yet or anything. I have copied a 78Mb file to and from it without any issues.

I also just fdisked and formatted my friends lexar sd 256 card which he "bricked" by pulling it out while he was flashing the crow rom and now it works again, good as new.

I have a 64Mb mmc card. I wouldn't buy any more mmc cards just because of the way they are physically made.

I think all sd cards and even all types of memory cards are made in the same factory. When I was in England a few months ago I was looking to buy a high capacity (128 or 256) memory stick for my digital camera but had a hard time finding these large capacity sticks. A guy at one of the camera shops told me that the factory that provided the memory had a fire so there was a shortage at that time in all memory cards (memory stick, sc, mmc, and others). I'm not sure if that is true or not but it makes total sence to me considering the sercomstances.

anyway, hope that helps. My sd brand isn't in your poll but they are probably all the same anyway.

by the way, why is this in the c7x0 forumn? is this only a c series issue?
raduga
QUOTE
I have just recently bought a Fujifilm 256Mb sd card and so far it works great.  I haven't formatted it yet or anything.  I have copied a 78Mb file to and from it without any issues.

I also just fdisked and formatted my friends lexar sd 256 card which he "bricked" by pulling it out while he was flashing the crow rom and now it works again, good as new.

I have a 64Mb mmc card.  I wouldn't buy any more mmc cards just because of the way they are physically made.  

I think all sd cards and even all types of memory cards are made in the same factory.  When I was in England a few months ago I was looking to buy a high capacity (128 or 256) memory stick for my digital camera but had a hard time finding these large capacity sticks.  A guy at one of the camera shops told me that the factory that provided the memory had a fire so there was a shortage at that time in all memory cards (memory stick, sc, mmc, and others).  I'm not sure if that is true or not but it makes total sence to me considering the sercomstances.

anyway, hope that helps.  My sd brand isn't in your poll but they are probably all the same anyway.

by the way, why is this in the c7x0 forumn?  is this only a c series issue?


I've only heard of it being an issue with c7x0; possibly its affected other people too. I had to trim a *lot* from the poll to make it fit.

I gather that Lexar boards are rebranded from (a few?) other manufacturers; some are Panasonic and some Sandisk; the Sandisk boards tend to be more often problematic. I'm not sure if Panasonic and Sandisk are manufactured at the same place - possibly.

The problems i'm seeing aren't simply file system corruption; the Lexar formats fine, and seems to handle short file transfers without incident.

Thanks for your input!
systemparadox
LOL.
Panasonic were probably the cause of all the problems with the Thinkpad 600 batteries, now this.
Any hardware you get- take a good look at it- if it says "panasonic" anywhere send it back, because it's probably defective (sorry panasonic- bye bye (shouldn't have made substandard equipment, should you?)).

Kingston SD's seem to all be Toshiba- no problems at all.

Just my 2 cents.
raybert
Boy, this problem sure sounds a lot like a similar problem I've been tracking for a while. I refer to this other problem as the "stalled upload" problem. The difference with "stalled upload" is that it is not limited to SD cards. Could these problems be one and the same?

There have been many threads started about the "stalled upload" problem. I tried to bring everything together in post #4 of this thread: <http://www.oesf.org/forums/index.php?showtopic=5948>

I think we've narrowed it down to a problem in the Z's USB networking subsystem. Unfortunatey, no one seems to know any details about this or have any idea how to fix it. I believe the source code for this subsystem is still closed by Sharp. I've suggested trying an older version but I don't know if anyone has tried this. I further suspect that we might lose USB MSD functionality if we were to move to an older version.

Those of you with this "SD problem", please read the threads I referenced and let us know if this sounds like the same problem.

~ray
lardman
Hmm, I just tried to copy a 200Mb file over to my 256Mb Viking SD card and it stalled and eventually failed after about 20-30Mb (it looks like the usb drivers also crashed on the Z's end as I had to re-insmod them to get my Z to be recognised after this).

I noticed that the kswapd process (iirc) was taking lots (50%) of processor time just before it failed (I was given a warning that it hadn't responded for 30sec and therefore, etc.)

The transfer then worked fine to my Kingston 512Mb CF.

I'm running OZ3.5.2 BTW, no idea whether sync is on or off.


Si
technojunkie
Part of the problem is the 'new' SD driver was for the 5500 (and I think 500d) only. %600 and up were different processors. (if this has changed, well then I'm wrong) See my sig for current specs, and I have no difficulty with sustained writes.
lardman
Ah, I was using my C750 in fact.

Si
Mickeyl
If you're interested in SD performance on C7x0, you should have a look at the OpenZaurus 2.6 kernel. This one contains an open source driver. 2.6 probably comes as an option w/ OZ 3.5.3, if not as the one and only kernel - we haven't decided yet.
lardman
Cheers Mickeyl, I'll take a look.

Si
systemparadox
QUOTE
2.6 probably comes as an option w/ OZ 3.5.3, if not as the one and only kernel - we haven't decided yet.


the 2.6 kernel is ready already?
Mickeyl
On C7x0, yes. Just a few minor things to sort out, but overall pretty usable.
danr
QUOTE(Mickeyl @ Mar 3 2005, 10:48 PM)
On C7x0, yes. Just a few minor things to sort out, but overall pretty usable.
*


Mickey,

Do you happen to know if the 2.6 kernel stops the dreaded eject-CF-card-during-suspend problem?

Cheers,

Dan
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2014 Invision Power Services, Inc.