Gerry_Houlder at notes.seagate.com
Thu Jun 22 14:24:17 PDT 1995
>Our application clients, which collect log parameters, would like to get all
> log parameters without doing a separate LOG SENSE for each page code.
An application client shouldn't care if several LOG SENSE commands need to
be issued. They don't seem to care if a file they want to load up is read by one
long READ command or is broken down into a bunch of smaller READ
>LOG SENSE appears to be patterned after MODE SENSE. MODE SENSE
>uses the page code 3Fh to indicate that all mode pages implemented by
>the target shall be returned.
>If page code 3Fh is defined for LOG SENSE to indicate that all log pages
>implemented by the target shall be returned, then the application client can
>collect all the Log Parameters.
>If page code 3Fh is added, should the PPC bit be set to zero and the
>parameter pointer field contain zeros, or should they be ignored?
The LOG pages can be very long, and the length of individual parameters will
vary from vendor to vendor. Example: vendor A may use an 8 byte field for the
"total bytes read" parameter in log page 3 and vendor B may use a 32 byte field
for that parameter. It has also been anticipated that the size of the parameter
may vary over time. For example, vendor C may use an 8 byte field for the
"total bytes read" parameter until the value becomes larger than 8 bytes, then
the vendor may start returning a 16 byte value for that field. The main point
here is that the application must be smart enough to expect (and parse) a LOT
of variation between vendors. This may be easier to do (and the resulting
application easier to maintain) when the pages are requested individually.
Also the total length of all pages may get very large, larger than a single LOG
SENSE command can return. This would force the application to get the pages
individually anyhow. This situation was anticipated during the development of
the LOG SENSE command and the use of page 0x3F to return all pages was
intentionally not included.
I'd like you to bring in a proposal though; now that the LOG SENSE monster has
been defined for a while, it would be interesting to hear the experiences of a
real end-user trying to use the results!
More information about the T10