Lewis Thompson wrote: > Hi Pramod > > On 17 February 2010 18:37, Pramod Batni <Pramod.Batni at sun.com> wrote: > >>> Lewis Thompson wrote: >>> >>>>> ::lastran >>>>> would print the time a thread last ran in walltime and ticks >>>>> >>> ::thread -d >>> This prints the last time [lbolt] the thread ran on a cpu . >>> Printing the time since it last ran [t - X] would be useful like it is done >>> for ::cpuinfo wrt to the lbolt. >>> > This was exactly what I was thinking... time since last ran along with > a wallclock version in hours, minutes, seconds. Both of these > readouts can be very handy at times. > Right. This would be useful. > >>>>> ::ts_chain >>>>> would look up the ts_chain for a given sobj. we already have sobj2ts >>>>> which does this work so it would make sense to update sobj2ts to use >>>>> ts_chain and expose both functionalities. >>>>> >>> I am not quite sure what you are proposing. can you please provide an >>> example. >>> > Certainly... > > >>> sobjptr::ts_chain >>> > would return the turnstile_chain_t from the turnstile_table bucket > sobj is associated with. We can check whether the dispatch lock is > held for the turnstile_chain_t with this. >
I think I get the idea now. :-) In most dumps where there was a need to find the turnstile_chain bucket, it was done by doing the math [for TURNSTILE_SOBJ_HASH] in mdb and getting the actual entry. > We already have > > >>> sobjptr::sobj2ts >>> > which returns the turnstile_t associated with sobj. The current > sobj2ts dcmd performs the lookup that I propose ::ts_chain will do, > and then goes further by finding the turnstile_t off the > turnstile_chain_t for sobj. > > It would make sense to implement sobj2ts with a call to ts_chain and a > walk rather than implementing the same lookup twice. > Makes sense. > >>>>> My question is really around the inclusion of these dcmds in mdb. I'd >>>>> like to see them included and not maintain a personal set of modules. Can >>>>> anybody comment on whether this is likely. I work at Sun, if that makes >>>>> any >>>>> difference. >>>>> > Any thoughts on this? > I think these are useful extensions. +1 from my side. Pramod
