Re: GBDE + md() + ccd() = corruption?

From: matt <matt_at_grogged.dyndns.org>
Date: Mon, 14 Apr 2003 20:09:31 -0500 (CDT)
Well, the good news is that your changes to ccd.c seemed to fix the ENOMEM
messages.

As for corruption problem, I had a little time today to play with it.
Just in case there were any relevant changes I used the kernel sources
from April 14th - I would have done a make world but things were "angry"
after I cvsup'ed.  I'll wait a couple days and see if the problem goes
away...  Anyhow, I nuked the files md() was using, and whipped up some new
ones, same size as the old ones, 1175000000b - I tried things in a
different order this time, I used md() as before with the 4x files, then I
used ccd() to concatenate the md devices together, and then I used gdbe to
init/attach ccd().  I've experienced the same problems since then
however...

To try and facilitate the discovery of what was happening to the files, I
started creating big text files (yes 'echo "abcdefg123456789" >>
testfile.txt'|sh), The smallest one that had md5sum issues so far was
140mb or so, though the bigger the files are the more likely they seem to
get corrupted.  Unfortunately, I discovered diff seems prone to spit out
"memory exhausted" errors on larger text files (unless --brief flag is
applied). <sigh>

I guess I'll see if I can whip up a python script or something to "diff"
the problem files so as to provide output more descriptive than "files foo
and foo2 differ".


More later as things develop...

-m



On Mon, 14 Apr 2003, Poul-Henning Kamp wrote:

> In message <20030413174853.F57816-100000_at_grogged.dyndns.org>, matt writes:
> >
> >Quick question (please CC me, as I'm not on the list yet) -
> >
> >I have 4x files setup through the md device (md1, md2, md3, mde4), each of
> >which I have init'd and attached gbde to (successfully), for md1.bde,
> >md2.bde, etc...  I've used ccdconfig to make these 4 encrypted devices
> >a concatenated disk (ccd0c), where upon I newfs'd /dev/ccd0c and
> >successfully mounted it.
>
> Ok, I just found and fixed one bad bug in ccd with respect to ENOMEM:
> it would never return the I/O request in question.
>
> You want to pull in version 1.133 of sys/dev/ccd/ccd.c
>
> --
> Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
> phk_at_FreeBSD.ORG         | TCP/IP since RFC 956
> FreeBSD committer       | BSD since 4.3-tahoe
> Never attribute to malice what can adequately be explained by incompetence.
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
>
Received on Mon Apr 14 2003 - 16:19:38 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:03 UTC