[Raw Msg Headers][Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: .48-970407.patch1b is OK
> > .48-970407.patch1b seems to work correctly so far. Non-delivery
> > reports are good. @a:b@c addresses resolved correctly.
> > Even long broken identd code works now.
> > I'll see if the problems with crasy smtp processes are over or not.
>
> Although, I noticed a big (18Meg) smtp process now and started
> investigation. It revealed that the process is trying to deliver
> a hughe message (17M). Matti, does it mean that smtp agent keeps
> the whole message in memory? Is it expected behavior?
No, it PAGES the message in. If you try to send a message
that is way too large (2-3 GB?) then average 32-bit system
can't send it, but it should not be needed either..
The message file (and the transport-descriptor file) are
opened RDONLY (and RW respectively), and then mmap()ed
MAP_FILE|MAP_SHARED.
It gives process often enlarged apparent size, which really
is not that big in memory/swap.
> (No, in my previous reports, smtp processes grew to 50Meg while
> delivering average several Kb size messages).
Propably an error in lib/selfaddr.c. I spotted it even in
my home machine... "weird, the machine went to a swap
frenzy, what is going on ? smtp growing rapidly ? Oooo..."
Fixed in patch1b
> Eugene
/Matti Aarnio