Compact Flash IDE chip loses content

We use an Advantech single board CPU card on which we install a 48M compact
flash chip. The chip appears as the primary drive on the secondary IDE
channel. QNX 4.25 is installed along with Photon 1.14. Once everything is
installed, the “disk” appears to retain its content, however, a power outage
has caused files to become corrupted (pdm.cfg, sysinit). Furthermore, using
vedit to edit any file on the flash disk produces a final edited file
containing nothing but "N"s (probably the fill character used when the disk
is created).

We suspect that either Fsys.eide or the flash bios is misbehaving. Has
anyone experienced this phenomenon and found the solution? Until then we’re
back to installing hard drives although using as 10G hard disk where a 48M
flash chip would do seems a bit ridiculous.

Previously, Harry Qualls wrote in qdn.public.qnx4:

We use an Advantech single board CPU card on which we install a 48M compact
flash chip. The chip appears as the primary drive on the secondary IDE
channel. QNX 4.25 is installed along with Photon 1.14. Once everything is
installed, the “disk” appears to retain its content, however, a power outage
has caused files to become corrupted (pdm.cfg, sysinit). Furthermore, using
vedit to edit any file on the flash disk produces a final edited file
containing nothing but "N"s (probably the fill character used when the disk
is created).

We suspect that either Fsys.eide or the flash bios is misbehaving. Has
anyone experienced this phenomenon and found the solution? Until then we’re
back to installing hard drives although using as 10G hard disk where a 48M
flash chip would do seems a bit ridiculous.

Do you run “chkfsys” every time you boot? It can help clean up problems like corrupted files.

  • Pete

Harry Qualls wrote:

We use an Advantech single board CPU card on which we install a 48M compact
flash chip. The chip appears as the primary drive on the secondary IDE
channel. QNX 4.25 is installed along with Photon 1.14. Once everything is
installed, the “disk” appears to retain its content, however, a power outage
has caused files to become corrupted (pdm.cfg, sysinit). Furthermore, using
vedit to edit any file on the flash disk produces a final edited file
containing nothing but "N"s (probably the fill character used when the disk
is created).

We suspect that either Fsys.eide or the flash bios is misbehaving. Has
anyone experienced this phenomenon and found the solution? Until then we’re
back to installing hard drives although using as 10G hard disk where a 48M
flash chip would do seems a bit ridiculous.

We use the Fsys.vpm50 (link to Fsys.ide) on a compact flash card. There is
some
but not much disk writing during operation. The system is
stopped by unplugging the power. Once in a while the filesystem gets corrupt.

Though we always do a chkfsys -Pqrs at startup, the filesystem cannot be
repaired.

So we have also experienced your problems but we haven’t found any solution,
but
we are VERY interrested in any .

regards


Per Åkesson
Carmenta AB
SWEDEN

Pete D. <peted@ifspurity.com> wrote:

Not to hijack the thread, but does everyone here realize
that Pete D. makes beer pumping systems?

Just thought he deserved some recognition…

Keep integrating that flow Pete…

Previously, pete@spam-yourself.com wrote in qdn.public.qnx4:

Not to hijack the thread, but does everyone here realize
that Pete D. makes beer pumping systems?

http://www.ifspurity.com/SemiSW2000_02.html

But that’s just a side-line until we get into our main business: making giant killer robots.


Keep integrating that flow Pete…

I take it that it’s your turn to wear the mascot outfit at QSSL HQ? It must be tough trying to type when you’re dressed like a giant neutrino…




\

  • Pete D.

Hi,

I have similar problems with SanDisk CF cards on an Advantech pc/104
embedded PC.

I suspect this is caused by some form of buffering in the CF card to
minimisize the number of write operations as each card only supports a
finite number of block erases in its lifetime. This may be nonsense
though…

Anyway, on my development system I avoid any corruption problems by
unmounting the disk and performing a proper shutdown before cycling the
power.
On the target hardware I mount the CF card partitions read only as the
application doesn’t require write access.
If you must have write access on your target use a seperate partition for
reading / writing data and store the application on a read only partition.
That way, if you lose some data you at least haven’t corrupted your code!!!
Again, if your data is mission critical this approach is not going to cut it
either.

It seems from my limited research that there is no 100% way to ensure a
write enabled partition on a CF card will never be corrupted without the use
of a UPS of some sort.
Hope this is useful,

Liam.



“Harry Qualls” <hfq@tree-d.com> schrieb im Newsbeitrag
news:99advo$c8g$1@inn.qnx.com

We use an Advantech single board CPU card on which we install a 48M
compact
flash chip. The chip appears as the primary drive on the secondary IDE
channel. QNX 4.25 is installed along with Photon 1.14. Once everything is
installed, the “disk” appears to retain its content, however, a power
outage
has caused files to become corrupted (pdm.cfg, sysinit). Furthermore,
using
vedit to edit any file on the flash disk produces a final edited file
containing nothing but "N"s (probably the fill character used when the
disk
is created).

We suspect that either Fsys.eide or the flash bios is misbehaving. Has
anyone experienced this phenomenon and found the solution? Until then
we’re
back to installing hard drives although using as 10G hard disk where a 48M
flash chip would do seems a bit ridiculous.

Hmmn,

I’ve hit this problem too. The buffer theory sounds reasonable, but how do
these cards then work in digital cameras? You can take your picture and then
just pull the card out without problem.

Do any Fsys gurus out there have any input to this problem?

thanks

Julian Thornhill

“Liam Power” <liam.power@hofmann-ge.com> wrote in message
news:9ahon7$ad8$1@inn.qnx.com

Hi,

I have similar problems with SanDisk CF cards on an Advantech pc/104
embedded PC.

I suspect this is caused by some form of buffering in the CF card to
minimisize the number of write operations as each card only supports a
finite number of block erases in its lifetime. This may be nonsense
though…

Anyway, on my development system I avoid any corruption problems by
unmounting the disk and performing a proper shutdown before cycling the
power.
On the target hardware I mount the CF card partitions read only as the
application doesn’t require write access.
If you must have write access on your target use a seperate partition for
reading / writing data and store the application on a read only partition.
That way, if you lose some data you at least haven’t corrupted your
code!!!
Again, if your data is mission critical this approach is not going to cut
it
either.

It seems from my limited research that there is no 100% way to ensure a
write enabled partition on a CF card will never be corrupted without the
use
of a UPS of some sort.
Hope this is useful,

Liam.



“Harry Qualls” <> hfq@tree-d.com> > schrieb im Newsbeitrag
news:99advo$c8g$> 1@inn.qnx.com> …
We use an Advantech single board CPU card on which we install a 48M
compact
flash chip. The chip appears as the primary drive on the secondary IDE
channel. QNX 4.25 is installed along with Photon 1.14. Once everything
is
installed, the “disk” appears to retain its content, however, a power
outage
has caused files to become corrupted (pdm.cfg, sysinit). Furthermore,
using
vedit to edit any file on the flash disk produces a final edited file
containing nothing but "N"s (probably the fill character used when the
disk
is created).

We suspect that either Fsys.eide or the flash bios is misbehaving. Has
anyone experienced this phenomenon and found the solution? Until then
we’re
back to installing hard drives although using as 10G hard disk where a
48M
flash chip would do seems a bit ridiculous.
\