Re: re-detect SCSi bus?

Doug Crompton (wa3dsp@marconi.crompton.com)
Wed, 18 Mar 1998 01:24:45 -0500 (EST)


On Tue, 17 Mar 1998, Kevin Lentin wrote:

> On Sun, Mar 15, 1998 at 09:43:42PM -0500, Richard B. Johnson wrote:
> > This is not a kernel issue at all. Just get a copy of "mount" and
> > add any features you want. The mount program can certainly read
> > the /proc file system and force a re-scan of SCSI if, and only if,
> > a SCSI interface is detected.
>
> HOW? How do you make Linux rescan the scsi bus? Manually do a remove-device
> and add-device for every possible combo?
>
> The only way I know is to unmount everything, remove all modules and then
> start from scratch. And the chances of doing that on any non-toy system is
> zero.
>

Not being an expert I can only use logic. Mount deals with devices,
linked to device drivers. I don't think mount should be messing with I/O
directly or make assumptions with /proc/scsi/scsi.

Hard drives are cheap - removeable frames are cheap. Installing
removeable hard drives (made up of a standard HD in a removeable frame)
are becoming more and more common. If the HD is not installed and turned
on (key on) at boot then the drive is not recognized.

The /proc/scsi/scsi is a hack to get it done but it seems dangerous.
Maybe we are pushing the SCSI spec using it in this way (hot removeable)
but it works reliably in many systems here.

Certainly RAID does this at the hardware level - between board and drives.

Windows95 does this "rescan" by looking at the installed devices and
rechecking for their presence. If such as list were kept in Linux - at
least for SCSI, then I suspect the "rescan" could be accomplished.

I am making a lot of assumptions here - but I think it is important to
visit this issue. SCSI is used far more extensively in Linux then Win95.
It also is becoming more of a plug and play bus - with auto termination,
auto ID, and the new single connector (power and data) connections.

Shouldn't this be done in a common SCSI module (non card specific) ?

Doug

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu