This message is from the T13 list server.
[ BC [EMAIL PROTECTED] ]
Correction:
Sorry, I see now I didn't write quite all that I meant, which now forces me to be less
concise than I like, in order to be accurate.
Personally I have almost no idea how x05 CdRom devices report block sizes, I didn't
mean to say I did. I've been told that a decode of certain particular variations on
op x5A ModeSense(10) is the least unreliable method documented in public for
discovering how many bytes a Scsi-over-whatever Cd-rom device will try to copy In if
provoked with a Cdb like the one block read near Lba 0 that is xA8 00 00:00:00:00 00
00:01 00.
Me, my personal history of pain centres largely in peripheralDeviceType x00
DirectAccess. Some of that pain teaches me that Cd-rom drives more than occasionally
copy In 8 bytes in response to the command -x 25 00 00:00:00:00 00 00:00 0 -i x8 and
that the last 4 bytes commonly then are 00:00:08:00. If the Cd-rom drive were of
peripheralDeviceType x00, then this would be a way of saying the block size of the
media present was x800 (2048) bytes/block.
As ever with removeable media devices, to reproduce this result, you may find you
first have to get past the SK ASC = x 6 28 NotReadyToReady (aka "MediaChange")
UnitAttention.
> the command -x 25 00 00:00:00:00 00 00:00 0 -i x8
The -x $cdb here - an op followed by all zeroes - is a popular example of a Cdb that
gives no plain indication of how many bytes to copy which way. This command works
only because "everyone knows" op x25 for peripheralDeviceType's x 00 05 0E ... always
should copy In 8 bytes, provided media is present, no unit attentions are pending, the
media capacity in fact does not exceed 2TiB @ 0.5KiB/block, etc. etc. etc.
Pat LaVarre
-----Original Message-----
From: Pat LaVarre
Sent: Mon 4/29/2002 8:49 AM
To: [EMAIL PROTECTED]
Cc:
Subject: RE: [t13] ATAPI Questions
This message is from the T13 list server.
[ BC [EMAIL PROTECTED] ]
As far as I know ...
1) SwDma and MwDma do not let data burst or stream more quickly than Pio4 does
- all share the 17e+6 byte/s limit. However, Wintel motherboards by design do
commonly waste a lot more cpu time streaming data in via Pio than in via SwDma or via
MwDma. The (insignificant) competition is commonly less broken. However, Ata Pio, if
distinguished from Atapi Pio, then has no protocol for reporting an error after
copying the last block in, whereas Ata SwDma and Ata SwDma do.
2) The standard of x200 (512) bytes/block is more de facto than de jure. Scsi
devices, including the Scsi over Ide that is Atapi, vary in how they report block
size. Devices whose op x12 Inquiry data byte 0 says they are of type x00
DirectAccess, x05 CdRom, x0E DirectAccessBastardisation (aka RBC), share the quality
of reporting via op x25 Read Capacity what their block size is, whenever the block
size in question is reasonably small and the media in question contains less than
xFFFF:FFFF Lba's (2TiB @ 0.5KiB/block).
3) Scsi over Ide, unlike Scsi over Usb and Scsi over 1394 and ..., omits to
pass over the bus the "specific field(s) that explicitly define - (a) if the command
contains data and if it does then (b) What is the length of the data associate to the
packet command?". Whether that omission is consequential or not has been the subject
of extensive, largely fruitless debate here since last November (2001). The
whitepaper giving my perspective is:
http://members.aol.com/plscsi/cdbcomplete.html
Pat LaVarre
-----Original Message-----
From: Yaniv Shapira [mailto:[EMAIL PROTECTED]]
Sent: Sun 4/28/2002 10:23 AM
To: [EMAIL PROTECTED]
Cc:
Subject: [t13] ATAPI Questions
This message is from the T13 list server.
Dear T13,
Can you help with the following questions:
- Which ATA/ATAPI devices (if at all) being used today requires
multiword DMA protocol?
- The ATA protocol defines a sector size of 512 bytes. Right? What is
the sector size used by ATAPI devices?
- Looking at the 12/16 bytes packet command can I find a specific
field(s) that explicitly define -
(a) if the command contains data and if it does then
(b) What is the length of the data associate to the packet command?
Thanks,
Yaniv Shapira
Product Definition Group
Galileo - Marvell Company
Email - [EMAIL PROTECTED]
Tel - 972 8 9247555 ext. 398
FAX - 972 8 9247554
