From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <84d01c3bb61f8ed6970ed1c22364038a@quanstro.net>
From: erik quanstrom <quanstro@quanstro.net>
Date: Sat, 12 Sep 2009 14:23:20 -0400
To: 9fans@9fans.net
In-Reply-To: <140e7ec30909121108l411080bam3a34bac1168f911f@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: Re: [9fans] sd(3) and concurrent readers/writers?
Topicbox-Message-UUID: 6de07072-ead5-11e9-9d60-3106f5b1d025

On Sat Sep 12 14:10:20 EDT 2009, sqweek@gmail.com wrote:
>  I have a fairly simple task - I'm trying to mirror a disk. So I run
> the obvious:
>
> dd -if /dev/sdE0/data -of /dev/sdF0/data -bs 1048576
>
>  That's sitting there on it's merry way, and I get curious as to how
> much progress it has made. I think about acid briefly but I don't have
> much experience with it, and figure I can get the answer without
> touching the procs:

you can cat /proc/$pid/fd and look at the offsets instead.

>  With this crude tool I can easily find a block that differs between
> the disks and binary search to gauge how far through the process is.
>  However, something seems a little fragile. If I keep running chk
> eventually the dd falls over with an error of some sort. It just gave
> me "write: " (that's an empty error string by the looks of it) 3 times
> in a row, and the fourth time around dd didn't report an error itself
> but I get this in kmesg:
>
> atagenioretry: disabling dma
> sdE0: retry: dma 00000000 rwm 0000

the lba of the error should be printed here.  have
you gotten a chance to check the contrib sd stuff?
i have added debugging to my version on sources, if you are
interested.

i know there are some fixes that would only be noticable
for a few lbas.

it's also possible that there are some known jmicron bugs
that we need to be working around.  i'll take a look.

>  System details: SIL3112 sata controller... is there an easy way to
> tell if I'm in AHCI mode or not? can't see anything obvious in the
> bios. During boot I get these relevant looking lines:
>
> dev A0 port 170 config 85C0 capabilities 0F00 mwdma 0007 udma 0407
> dev A0 port 9400 config 0C5A capabilities 2F00 mwdma 0007 udma 207F
>   LLBA sectors 586072368
> dev A0 port 9C00 config 0C5A capabilities 2F00 mwdma 0007 udma 207F
>   LLBA sectors 586072368

you wouldn't get these messages in ahci mode.  also
/dev/sdctl will list each driver name and the drive letter,
one drive letter per line.

- erik