Re: Two scenarios leading to filesystem corruption under NetBSD-4.0, anyone seen this?

[buhrow%lothlorien.nfbcal.org@localhost (Brian Buhrow)]

        Hello.  Yes, I've confirmed that scenario 2 fails predictably on
multiple computers with multiple flash disks.  What I'm not sure is if
there is a timing element involved yet.
        What's interesting is that I can induce the same problem under NetBSD-3,
but only if I write to flash disks under ehci(4) control.  uhci(4) or
ohci(4) devices work fine.  I think  this still  might be true under
NetBSD-4, meaning that scenarios 1 and 2  in my description are not related
to each other after all.
        I suspect that this problem is exaserbated by writing to slow media,
at least in the msdos case, because I imagine it would be hard to get two
processes writing to a small msdos image simultaneously when writing is
fast.

-Brian

On Apr 15,  6:23pm, "Greg A. Woods; Planix, Inc." wrote:
} Subject: Re: Two scenarios leading to filesystem corruption under NetBSD-4
} 
} On 15-Apr-08, at 11:48 AM, Brian Buhrow wrote:
} >     Hello.  I'm not sure if this is a related issue or not.  I'm not
} > seeing filesystem corruption in the sense that fsck doesn't pass the
} > filesystem after a change.  What I'm seeing is that the files on the
} > filesystem do not contain the data which they're supposed to.  this  
} > applies
} > to both MSDOS and FFS filesystems connected via USB.  In the MSDOS  
} > case, I
} > can get the corect results by not having two processes write to the
} > filesystem at once.
} 
} Ah, I see -- it looks like what I've seen is not related to your  
} observations.
} 
} Can you confirm that your USB device is reliable?
} 
} For example can you write verifiable data to the raw device and read  
} it all back reliably and repeatedly?
} 
} -- 
}                                       Greg A. Woods; Planix, Inc.
}                                       <woods%planix.ca@localhost>
} 
>-- End of excerpt from "Greg A. Woods; Planix, Inc."