** This is the quasi-official and semi-temporary T13 email list server. **
> > ... 28-bit commands that span the 28-bit boundary? > ... the drive should error at whatever 60,61 reports... Interesting difference from Scsi/Atapi here, whose analogues to the 28 and 48 bit Lba's here are 21 and 32 bit Lba's. Me, I wasn't working in storage back when Scsi2 added the way to report an Lba-in-error, but I've heard that report always included room for a 32 bit Lba. So 21 bit Lba ops tend to get defined in terms of the 32 bit Lba ops, leading to utterly thoughtless, but workable, support for spanning accesses across the 21 bit boundary. /// Likewise here for Ata, I'd guess the people who design devices to get away with as little work as possible would say no, any host rude enough to access across the 28 bit boundary deserves what it gets. The thinking would be to convert 28-bit requests to 48-bit requests on the front end, with most of the drive seeing only 48-bit requests. A spanning access is an "ambiguous condition", which might cause an obvious error, or a partition-sector-corrupting wraparound, or work just fine. This can get emotional: unless we let the transfer begin and choke later, we're talking about one of the metrics that matters most in the real world. We're talking inline command overhead - apparent additional seek time - in block reads and writes. /// If and when an error occurred that couldn't be pinpointed in a 28 bit Lba, presuming the error is not an unreportable error that Ansi makes flatly unreportable like trouble seen while transferring back the last pio read block, people might report the error occurred at the largest Lba they could report. I've been told people generally accept the idea that a host should retry all of a write command that failed, rather than assuming everything went well up from the first Lba up to the only error Lba reported. Pat LaVarre >>> [EMAIL PROTECTED] 09/06/01 11:08AM >>> ... I think we agree 100% on this ... >>> [EMAIL PROTECTED] 09/06/01 10:40AM >>> ... the drive should error at whatever 60,61 reports... One issue with allowing the access to be greater than 60,61 or to cross the 28 bit boundary is that errors can not be properly reported. ... On Thu, 6 Sep 2001 09:32:25 -0600, Jim Castleberry wrote: >1. On a huge drive, is a READ SECTORS starting at LBA 2^28-5 and going > for 10 sectors guaranteed to give the same result as a READ SECTORS > EXT with the same start and length, i.e. is the device required to > support 28-bit commands that span the 28-bit boundary? I think the drive should error (ERR=1 IDNF=1) the 28-bit READ SECTORS because it is attempting to read LBAs that are not addressable by a 28-bit command. This is no different than trying to read past the end of the drive when the drive is only a 40GB drive, it is an "LBA range error" that should get an IDNF error. ... -- 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]
