Re: swap space leak in 5.0_STABLE

[Jed Davis <jld%panix.com@localhost>]

Manuel Bouyer <bouyer%antioche.eu.org@localhost> writes:

> On Mon, Aug 10, 2009 at 02:43:50PM +0000, YAMAMOTO Takashi wrote:
>> hi,
>> 
>> > Any message in dmesg or /var/log/message ?
>> > uvm can leak swap slots when under high memory pressure, but in this
>> > case a printf() fires up.
>> 
>> can you explain?
>
> I remember seeing a message saying that the systen couldn't
> alocate memory for an internal structure and as a result of this some
> swap pages were lost, but I couldn't locate this message in
> the sources. Maybe it was in an older version of NetBSD ...

I believe that was from when swap management used the extent(9)
routines, which can require allocating memory in order to free space in
the resource being managed.  It was changed to <sys/blist.h>, which is a
flat bitmap, back in 2005 (r1.90 of uvm_swap.c).

-- 
(let ((C call-with-current-continuation)) (apply (lambda (x y) (x y)) (map
((lambda (r) ((C C) (lambda (s) (r (lambda l (apply (s s) l))))))  (lambda
(f) (lambda (l) (if (null? l) C (lambda (k) (display (car l)) ((f (cdr l))
(C k)))))))    '((#\J #\d #\D #\v #\s) (#\e #\space #\a #\i #\newline)))))