Hi keith,

sorry to not clearly giving details.
I am trying to communicate with target through HBA. I am collecting data of
target through HBA.
I want to detect how many targets are attched to HBA.
I have done a research on all that you have told. But with these analysis I
am not still getting the positive result

I will try further.

Thanks for your valuable guidance!!


Regards,

Aakash




On Mon, Feb 25, 2008 at 11:15 PM, Keith M Wesolowski <
[EMAIL PROTECTED]> wrote:

> On Mon, Feb 25, 2008 at 10:52:34PM +0530, aakash berde wrote:
>
> > I am creating an application which should communicate with the different
> > HBAs.
>
> Yes, but what does this actually mean?  Are you trying to tell the HBA
> to do something (e.g., create a RAID0 array), or are you trying to
> tell some target attached to the HBA to do something (e.g., move a
> tape around)?  HBA internals tend to be heavily vendor-specific and
> are usually proprietary.
>
> > I am getting problem with the adaptec HBA. I tried today to use uscsi
> for
> > communicating with adaptec HBA but not succeeded.
>
> That's not surprising; an HBA is not a SCSI target.  It is usually a
> SCSI initiator.
>
> > 1. which directory/file is being used for scsi devices list. I tried
> > /dev/cfg, dev/dsk, dev/rdsk but getting error as " No such file or
> > directory".
>
> SCSI devices may be found, depending on which driver is attached, in
> /dev/rdsk/*, /dev/rmt/*, /dev/scsi/*/*, and other driver-specific
> directories.  You need to read the man page for the driver attached to
> the targets of interest.  scsi(4) has some general information of
> interest.
>
> > 2. I am getting an error "inappropriate ioctl for device" or
> "Inappropriate
> > argument" when trying to communicate with HBA. can I get the root cause
> > behind this?
>
> HBAs are not SCSI targets and therefore do not support uscsi(7I).
>
> > 3. In Linux there is sg driver resonsible for SCSI ioctls. On solaris is
> > there any kind of driver is available?
>
> sgen(7D) serves the same purpose.  sd(7D), st(7D), and ses(7D) among
> others also support uscsi(7I) although ses(7D) in particular provides
> no value beyond that provided by sgen(7D).
>
> > 4. Is there any congiguration to be doe before using uscsi
>  functionality.
>
> Other than ensuring that the appropriate drivers are attached to the
> targets of interest, no.  You open(2) the device and then you may
> ioctl(fd, USCSICMD, &ucmd) thereafter.
>
> > 5. Besides uscsi is there any different way to communicate with HBAs on
> SCSI
> > ioctls.
>
> Again, HBAs are not targets.  I still do not know whether you are
> trying to communicate with an HBA or a target.  The two are COMPLETELY
> different.
>
> > As these might be question to be raised by new persons next to me,
> kindly
> > give solution if anyone have, as it will be useful to all solaris system
> > Learners otherwise I might have to stuck with the Windows OS.
>
> I do not see how using a different OS will help you; talking to HBAs
> is always vendor-specific and usually proprietary.  For that reason,
> it should be abstracted into thin drivers that provide nothing more
> than transport and attachement points.  Talking to targets works just
> fine on Solaris, as I'm sure it does on Windows as well.  You still
> have to know which you want to do, though.  You seem to be confusing
> the two, which may be an unfortunate side effect of an extremely
> broken Windows-centric paradigm in which all non-I/O operations are
> approached in an HBA-centric way, even though the operations
> themselves apply to SCSI targets or SAS, FC, etc. targets of other
> types out in the fabric.  That is the approach taken by t11's SM-HBA;
> I consider it to be a byproduct of a "we're stuck with the frameworks
> Windows provides so let's just make the lowest common denominator
> available on all systems" approach that denies the reality that OS
> innovation is both possible and necessary.
>
> Don't buy into that.  Targets are targets and all manipulation of
> targets should be done through a device node representing that target.
> HBAs just provide attachment points for those target nodes, and
> transport data to and from them.  The need to talk explicitly to or
> even be aware of HBAs should be minimal or nonexistent for any sane
> application.
>
> Hope this helps.
>
> --
> Keith M Wesolowski              "Sir, we're surrounded!"
> Fishworks                       "Excellent; we can attack in any
> direction!"
>
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code

Reply via email to