** This is the quasi-official and semi-temporary T13 email list server. **
Jim, just because we accept it does not make it correct.
Most of the HOST-Drivers have to do outragous HACKS to fix firmware
KLUDGES that should not exist, period. The simple fact is that most ATAPI
is junk and HOST's suffer the burden of cleaning up the mess.
Also the HOST puts the device into DSC overlap mode for a reason.
We call this ATAPI starvation because the bandwidth overhead wrt an ATA
device is unacceptable. If the HOST is doing two operations on a given
channel, the HOST will decide to grant the majority of the bandwidth to
the ATA and when there is time we go back an clean up the ATAPI device
that supports DSC.
Now if you are unfortunate to mix an ATA with an ATAPI that does not
support DSC you are 0xdeadbeef. The pure performance gain of executing
and force DSC feature sets on ATAPI's is the all the ATA devices pound the
data down the pipes.
The axiom problem the BSY being randomly set is the silly devices sending
more than the requested information. This is like an unwanted opinion.
Even though you ignore it, it still comes. Thus one has to do something
to dump it, when it should not be there in the first place.
Junk Mail delivered to you house ranks in the same annoiance level.
Andre Hedrick
CTO ASL, Inc.
Linux ATA Development
PS spellings are creative thanks to public education.
On Wed, 11 Jul 2001 [EMAIL PROTECTED] wrote:
> Date: Wed, 11 Jul 2001 17:15:08 -0600
> From: [EMAIL PROTECTED]
> Reply-To: [EMAIL PROTECTED]
> To: [EMAIL PROTECTED]
> Subject: [temp t13] Re: Are devices allowed to go busy on their own?
>
> ** This is the quasi-official and semi-temporary T13 email list server. **
>
>
> Actually, there is a very GOOD reason that you may see this happen.
> This is a normal part of ATAPI behavior that hosts must 'just accept'.
> -------------
> Part of the ATAPI protocol (e.g. QIC-157 for tape, ??? for CD) involves the
> <asynchronous> setting of the DSC status bit when a device has
> 'logically disconnnected' from the host during some
> long operation (usually involving media). Please see examples below.
>
> Assume that an ATAPI device has a status of RDY only (x40)
> and a media-access command is still in process (e.g. LOAD or REWIND).
> When the <media> is ready (again), the device is required to set the
> DSC status bit.
>
> - - - but according to ATA, when BSY=0, the device does not 'own' the bus.
> Proper ATAPI controller chips and firmware implement and respect this.
>
> Device firmware that wants to set DSC must first set BSY=1 in order to get
> control of the bus, set the desired new status, and then clear BSY=0
> AS FAST AS POSSIBLE. Some devices do this in hardware
> (and it happens very quickly), and some do it in firmware (slowly).
> -----------------
> Now, part of the problem is that the processor speeds of devices and hosts
> used to be closer to each other than now. Currently, hosts are many times
> faster than the processors in many devices, so a host may read the STATUS
> register many times during a transition. Transitions are not seen so much
> on
> slower hosts.
>
> For our own ATAPI tape drives, I have seen this many, many times.
> It happens less frequently when hardware does the transition, and
> when the host is slow.
> ----------------------
> Example:
>
> Media access command that takes a long time:
>
> Status Event
> ---------------------------
> ------------------------------------------------------------
> RDY+DSC device is idle
>
> RDY+DSC+BSY device receives command
> (e.g. REWIND,
> LOAD/UNLOAD, SPACE,
> LOCATE,
> etc.)
>
> device decodes command
> and starts
> processing it.
>
> RDY device logically disconnects
> from bus
> to allow host to
> access other devices
> during a long media
> operation
>
> (time passes)
> device operation is
> completed
>
> RDY+BSY device 'steals' control of bus
>
> RDY+DSC+BSY device changes status
>
> RDY+DSC device is idle, and command is
> complete
> ======================================================================
> - Jim Hatfield
> Sr. Firmware Engineer
> Overland Data Inc.
>
>
>----------------------------------------------------------------------------------------------------------------
>
>
>
>
> On Wed, 11 Jul 2001 14:47:06 -0700, Larry Barras wrote:
> >I've run into a number of devices (atapi usually) which transition
> >from BSY=0 to BSY=1 on their own.
>
> I have seen this problem on many ATAPI devices.
>
> THIS IS ABSOLUTELY AND COMPLETELY ILLEGAL! Such devices should never
> pass a vendor qualification test (period, no excuses)!
>
> >This is a rather vexing situation
> >for the host, who looks at the status register and sees the device is
> >not busy, starts to issue a command based on that status which fails
> >because the device decides to set busy on its own.
>
> You are correct. Related to this are the ATAPI devices that decide
> all on their own that a host is too slow during command execution and
> just "reset" themselves. This too is absolutely and completely
> illegal.
>
>
> *** Hale Landis *** [EMAIL PROTECTED] ***
> *** Niwot, CO USA *** www.ata-atapi.com ***
>
>
>
>
>
>
>
> ********************************************************************
>
> Get it all in Overland's revolutionary new LibraryXpress Neo Series.
>
> Visit us at CA World July 8 - 11, Orlando, FL booth #124 where you'll
> find everything you always wanted in a tape library.
>
> For more information on Neo:
> http://www.overlanddata.com/mktg.nsf/lookup/lxn2000.pdf/$file/lxn2000.pdf
>
> For more information on Overland:
> http://www.overlanddata.com
>
> ********************************************************************
> --
> If you have any questions or wish to unsubscribe send a
> message to Hale Landis, [EMAIL PROTECTED] To post to
> this list server send your message to [EMAIL PROTECTED]
>
> For questions concerning Thistle Grove Industries or TGI's
> list services please send email to [EMAIL PROTECTED]
>
>
>
--
If you have any questions or wish to unsubscribe send a
message to Hale Landis, [EMAIL PROTECTED] To post to
this list server send your message to [EMAIL PROTECTED]
For questions concerning Thistle Grove Industries or TGI's
list services please send email to [EMAIL PROTECTED]