This message is from the T13 list server.
Ditto! On 3/12/02 12:25 PM, "[EMAIL PROTECTED]" <[EMAIL PROTECTED]> wrote: > This message is from the T13 list server. > > > Instead of defining "proprietary vendor-specific extension" to deliver this > function by each manufacturer for above 128GB drives, will it make more > sense for T13 to "unify" this efforts? > > Raymond Liu > > -----Original Message----- > From: Pat LaVarre [mailto:[EMAIL PROTECTED]] > Sent: Tuesday, March 12, 2002 11:53 AM > To: [EMAIL PROTECTED] > Subject: [t13] but already we did drop Read/Write Long > > > This message is from the T13 list server. > > > [ BC [EMAIL PROTECTED] ] > >> Harlan Andrews 03/12/02 10:15AM >> Although Read/Write long >> may have been deleted from the spec, >> it is STILL a requirement >> for several Host vendors. Drive suppliers >> should be very careful not to actually remove >> the Read/Write long commands. > > But we Ansi T13 the committee actually have flatly removed Read/Write Long > above 128GiB @ 0.5KiB/block (and Ansi T10 has removed them above 2TiB) ... > > ... no? > > Up above 128GiB, there's nothing an Ata drive vendor can do to bring these > back, except provide some reasonable proprietary vendor-specific extension > to deliver the same function. Ditto for Scsi/Atapi work up above 2TiB. > > The legit uses for these ops known to me are to make a Read respond slowly > without an error and to make a Read report an error. This is how I most > easily observe, for example, that some drives have an entirely creative idea > of what the correct Pio protocol is for reporting a read error. > > I can imagine being limited to below 128GiB to perform error injection like > this won't much matter ... I guess soon enough we'll all find out. > > Slow writes and Write errors have always been harder to observe. Scsi > ModeSelect sometimes will give you some reasonably repeatable Write errors, > but I've never had a good way to get individual Write's to be slow. > >> Thomas Colligan 03/12/02 12:06PM >> Gee! yet another use for the W/R long commands. > > Nothing like removing a feature to dig out how it actually is helping > people. > > Pat LaVarre > >> RE: [t13] Read/Write Long >>>> [EMAIL PROTECTED] 03/12/02 11:42AM >>> > This message is from the T13 list server. > > > In the ATA RAID1 (disk mirroring) application, R/W Long has been used to > perform forced error on the rebuild drive in certain LBA (when the host > cannot get valid data from the corresponding LBA in source drive during the > mirror rebuild process). This will force the OS to recognize the error > situation when OS is trying to access the data from that LBA. Without R/W > Long, it will require a much more complicated error handling algorithm to > handle the situation. > > Raymond Liu >
