Matthew Dharm wrote:
>
> A while back, there seemed to be a general agreement that the mechanism for
> determining the setting of the direction flag needed an overhaul. If my
> memory serves, the general consensus seemed to be that either (a) the
> mid-layer should have a direction table for all commands, or (b) there were
> some inferrence rules which could be used to derive the information.
>
> This change was only needed for the scsi-generic interface, as userspace
> programs (i.e. cdrecord, cdparanoia, etc) wound up causing Scsi_Cmnd
> structures with a direction of UNKNOWN to be sent to the HBAs -- this
> causes problems with the USB SCSI emulation.
>
> My question is this: Has anyone managed to implement this?
Matt,
This was fixed (i.e reverted to 2.3.99-pre3 behaviour) in a
2.4.0-test1-ac9 patch to sg. So you should no longer see
UNKNOWN data direction coming from any existing sg based
applications (i.e. SANE, cdrecord, paranoia, etc).
[My post to this newsgroup on 2000/04/21 titled:
"Re: sc_data_direction -- will we require it?" outlined
the reasoning.]
So as far as I can see, UNKNOWN data direction will only
occur when _explicitly_ asked for in 2 situations:
- CDROM_SEND_PACKET ioctl
- sg's new sg_io_hdr interface
In both cases applications have the ability to set specific
direction flags. I am not aware of any applications that
(will) use UNKNOWN data direction but it's there since it
isn't the business of pass-through mechanisms to dictate
policy.
Doug Gilbert
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
