[3488] in linux-scsi channel archive
Re: [RU.LINUX] =?ISO-8859-1?Q?=CF=D0=D1=D4=D8...?= =?ISO-8859-1?Q?=CE=D5,?= =?ISO-8859-1?Q?=D3=CB=C1=D6=C5=CD,?= =?ISO-8859-1?Q?=CF?= =?ISO-8859-1?Q?=C4=C5=D7=CF=DE=CB=C1=C8.?=
daemon@ATHENA.MIT.EDU (Eric Youngdale)
Sun Mar 8 23:57:45 1998
Date: Sun, 8 Mar 1998 23:54:25 -0500 (EST)
From: Eric Youngdale <eric@andante.jic.com>
To: Boris Tobotras <boris@xtalk.msk.su>
cc: linux-scsi@vger.rutgers.edu
In-Reply-To: <m24t1hvh68.fsf@xtalk.msk.su>
On 2 Mar 1998, Boris Tobotras wrote:
> I've already wrote -- stupid linux allows for doing umount
> /external-scsi-drive ; echo 'scsi-remove-single-device 0 0 0 0' >
> /proc/scsi/scsi, turn its power off... and loose up to meg of data. Because
> it have no idea about cache flush. Especially pleasing is that root dir of
> device is usually goes to hell.
>
> Patch is not in exact place where it actually belongs, but I feel it even
> better -- I don't feel like having unflushed cache even on non-removable
> device.
>
> Those who're willing can translate this to linux-scsi.
>
> * * *
>
> So I do. One more question for me: what about write cache of SCSI
> drive itself? What have one do to do robust flush of it befure power off?
This is a known problem, unfortunately.
One way of handling this is to flush all data before the command
to deregister the device takes effect. This would seem like the most
idiot-proof solution, I guess, but the trick is that for each device type
you would need to go back from the ID/LUN and get to a disk number. Not
that hard, I guess.
-Eric
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
