> YAMAMOTO Takashi wrote:
> >> YAMAMOTO Takashi wrote:
> >>
> >>> an example which i told you some time ago:
> >>> 1. thread A calls fileassoc_file_add and sleeps in malloc.
> >>> 2. thread B calls fileassoc_file_add and successfully allocates and inserts
> >>>    a new entry.
> >>> 3. thread A wakes up and inserts a duplicated entry.
> >> can this scenario happen in netbsd-4 as a result of veriexecctl?
> >>
> >> -e.
> > 
> > i don't know.
> 
> then what are we discussing, really? :)
> 
> -e.

the lack of locking in fileassoc.  are you kidding?

as you know, veriexec is not only user of fileassoc.
do you mean you have investigated all of users and know fileassoc
doesn't need to have locking on netbsd-4?  otherwise, why do you claim
it isn't a problem?  to me, these users seem to work basically independently,
so i think nothing serializes users and thus fileassoc itself needs to
have locking internally.
am i missing something obvious?  (i guess i am, given that you are
an author of all of these fileassoc users.)

YAMAMOTO Takashi