[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] Block Devices support for JFFS2



At 07:50 AM 8/21/2001 +0200, willy tarreau wrote:
>  >  From experience, I can state that these "IDE"
>flash
> > devices (I have tested 2 different manufactures of
>CF and
> > 3 different models of 1 major IDE flash disk
> > manufacturer) are NOT reliable under power fail
> > conditions- if the system is writing to them.
>
>I totally agree with you, same experience here.


I won't say that I'm happy that you had the same (poor performance) 
experience as mine, but at least the experience was consistent :)



>Indeed, my
>firewall's CompactFlash is a 16 MB I had to partition
>after
>hardware errors due to poor power supply. The power
>wire
>has been cut, and the only power that reached the
>device
>was drawn through the IDE signals (which was enough
>for reads, but unfortunately not for writes). And
>after that,
>the device returns "CRC error", "sector not found",
>and such
>standard hard disks errors which are absolutely
>unrecoverable (even after attempting to reformat it).


Yup! All those beasts. As a matter of fact, I had better luck with
a "standard" desktop hard drive. Imagine that!

But, of course those suffer from shock and vibration and temp issues- so 
don't know if they are any better to use in a shipping embedded system.



> > If one is really interested in building a reliable
>system,
> > that you are going to ship in quantity, I would
>strongly
> > encourage power fail testing on these devices.
>
>in fact, it would be good if the IDE "low level
>format"
>command was completely implemented so that it sets the
>device back in factory configuration.

That would be nice. At least it would let me recover my 32MB CF disk so 
that I can use it in my digital camera again :)

I'm not quite sure that this will help build a reliable system- even with 
JFFS2 on it. JFFS2, AFAIK is built with the assumption that the underlying 
flash layer does not loose "blocks" (in the IDE case- 512 bytes worth) of 
data on its own accord. JFFS2 will break in this situation (where we had to 
low level format the IDE sector to recover it).

Vipin



To unsubscribe from this list: send the line "unsubscribe jffs-dev" in
the body of a message to majordomo@xxxxxxx.com