Re: Flash filesystem

Tech-Archive recommends: Fix windows errors by optimizing your registry

Hi!

We come to conclusion that we should try other filesystems and drivers. Can
you please advise us which will be good for our conditions.

PS: We tried to power down (suspend) platform at least 5 seconds before
reset hits and during stress tests. The filesystem also crashes after a
number of this cycles. We are using standard Stratad driver (in public
directory). Can we rule out software problems and concentrate on hardware?

Thanks,
Jernej Turnsek

"Bruce Eitman (eMVP)" wrote:

> Zhongwei:
>
> As we both know all to well, you left something out of the theoretical
> discussion. The TFAT is only as stable as the media that it sits on, which
> can be made to be very unstable by the driver that manages it. Don't
> misunderstand me, flash is very stable, but can be corrupted if interrupted
> while erasing or writing to a block, BUT TFAT can manage to survive that as
> Zhongwei describes.
>
> The problem can happen if the media driver moves bits around on the disk to
> **without** going through the TFAT, or taking other precautions to prevent
> corruption of data. I believe that MSFlash does compact the data on the
> Flash to make room for new data, flash failures during that can be fatal.
> As Zhongwei points out this is all just theoretical though. I have not been
> able to prove this with MSFlash, mainly due to a shortage of source code and
> time. I have been able to prove this on other Flash drivers though. I have
> also found that suspend/resume can make it worse. Resets can be managed
> better because the disk can be scanned when the driver is loaded and
> repaired.
>
> --
> Bruce Eitman (eMVP)
> Senior Engineer
> beitman AT applieddata DOT net
>
> Applied Data Systems
> www.applieddata.net
> An ISO 9001:2000 Registered Company
> Microsoft WEP Gold-level Member
>
>
> "Zhongwei Wang" <MWang@applieddata[dot]_net> wrote in message
> news:OiHxm2IiFHA.2560@xxxxxxxxxxxxxxxxxxxxxxx
> > Transaction safe file systems can reliably survive the hardware reset,
> > theoretically, if it does it right. That is what they are advertised.
> TFAT
> > is not reliable enough right now then it only means it is not doing
> > everything correct and it needs improvement.
> >
> > There is advertised idea of having two state file systems, working state
> and
> > saved state, saved state is known good state. After a working state
> > transaction is fully finished, working state is saved or converted to
> saved
> > state. And any working state transaction will only write to unused
> blocks,
> > i.e. blocks not being used by saved state, so saved state will not be
> > touched. Does that sound perfect enough theoretically?
> >
> > --
> > Zhongwei Wang
> > Applied Data Systems
> > www.applieddata.net
> > Application - Ready embedded systems
> > Microsoft Windows Embedded Partner
> > Gold Level Member
> >
> >
>
>
>
.

Relevant Pages

  • Re: Flash filesystem
    ... >> can be made to be very unstable by the driver that manages it. ... >> Senior Engineer ... >> Applied Data Systems ... >>> saved state, saved state is known good state. ...
    (microsoft.public.windowsce.platbuilder)
  • Re: Flash filesystem
    ... I don't want to endorse any flash products publicly for several reasons. ... >> can be made to be very unstable by the driver that manages it. ... >> Applied Data Systems ... >>> saved state, saved state is known good state. ...
    (microsoft.public.windowsce.platbuilder)
  • Re: Bug in PCI core
    ... driver calls pci_set_power_state. ... routine calls pci_save_stateagain. ... This overwrites the saved state ...
    (Linux-Kernel)
  • Re: Flash filesystem
    ... while erasing or writing to a block, BUT TFAT can manage to survive that as ... > There is advertised idea of having two state file systems, working state ... > saved state, saved state is known good state. ... > Applied Data Systems ...
    (microsoft.public.windowsce.platbuilder)
  • Re: PCCARD driver question
    ... beitman AT applieddata DOT net ... Applied Data Systems ... What I understand from this is that the device.exe first loads a driver ...
    (microsoft.public.windowsce.embedded)