Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Another UVM wire page count underflow panic

> On Mar 15, 2019, at 2:56 PM, Robert Elz <kre%munnari.OZ.AU@localhost> wrote:
> Actually, on a bit more reflection, perhaps the best way forward
> is to work out a better method of implementing the shm() stuff
> puerly out of uvm facilities (with minimal, less than is there now,
> extra data structs) - so shmctl(SHM_LOCK) is just mlock() but perhaps
> with an extra parameter to the underying function which does the work
> to indicate the semantics.

Actually, I don't think the way SVID shared memory is implemented is particularly bad ... and in fact, the whole reason we have the "anonymous memory object" (an amusing contradiction in terms :-) is to support SVID shared memory.

SVID shared memory has a namespace, and that's basically what sysv_shm.c implements.  The underlying support for mapping the shared memory objects and wiring their pages, etc. is indeed provided by UVM.

-- thorpej

Home | Main Index | Thread Index | Old Index