On Wed, Mar 22, 2023 at 02:59:13PM +0100, Jean Delvare wrote: > Hi Jerry, > > On Fri, 17 Mar 2023 16:19:55 -0600, Jerry Hoemann wrote: > > Address an incompatible change to record type 237. Field in offest 0x09 > > "offset" ;-) > > > changed from a STRING to a WORD date. This causes the decode of these > > records to display <BAD INDEX>. > > > > The patch addresses both forms of the record. This is accomplished based > > upon size of the field in offest 0x09. > > You can't actually detect the size of a field in a DMI record. The only > size you can check is the size of the record itself (and your patch is > doing exactly that).
Correct. The code infers the field size by the size of the record. This trick works as long as the old spec/definition is never extended beyond offset 0x09. > > AFAIK, the version of record type 237 that is currently implemented in > > dmidecode never made it outside of HPE. So, patch addressing both forms > > might be overkill. I have implemented and tested a more straightforward > > change to the decoding of the record based upon its new definition. > > So, please let me know if you want the simpler fix. > > Indeed, in my large collection of DMI tables, which includes a good > number of HP(E) systems thanks to you, I see type 237 records of size > 8, 9 and 11 but not 10. > > So I'd be happy to take the simpler fix in the name of, well, > simplicity. Unless it somehow makes your life easier if dmidecode > supports both. I think we'll be fine w/ the simpler approach. If any firmware exists using the old definition, it will be transitory. > > -- > Jean Delvare > SUSE L3 Support -- ----------------------------------------------------------------------------- Jerry Hoemann Software Engineer Hewlett Packard Enterprise -----------------------------------------------------------------------------
