Re: More problems in dup_mmap

Philip Gladstone (philip@raptor.com)
Wed, 06 Aug 1997 10:11:01 -0400

Messages sorted by: [ date ][ thread ][ subject ][ author ]
Next message: Linus Torvalds: "Re: rmdir(".") works, and causes havoc"
Previous message: Fyodor Yarochkin: "SV: how to write module (fwd)"
Maybe in reply to: Bill Hawes: "More problems in dup_mmap"

Bill Hawes wrote:
>
> It appears that there are yet more problems in dup_mmap -- as far as I
> can see, an error return from copy_page_range leaves the current vma
> struct not linked into the mmap list, so it won't be released. The
> memory leak is the lesser problem, as it also means that there could be
> many pages left stuck from elevated use counts. This would certainly
> exacerbate a low-memory situation ...

I was thinking last night when I was driving home -- things like
verify_area examine the mm, and this can be happening concurrently
with other functions manipulating it. It seems that reader/writer
locks are really required on this (and probably other) structures.
This adds a whole new level of complexity which should be
addressed in the 2.1.x series of kernels. I'm not convinced
that the level of upheaval is worth it in 2.0.3x

Philip

-- 
Philip Gladstone                           +1 617 487 7700
Raptor Systems, Waltham, MA         http://www.raptor.com/

Next message: Linus Torvalds: "Re: rmdir(".") works, and causes havoc"
Previous message: Fyodor Yarochkin: "SV: how to write module (fwd)"
Maybe in reply to: Bill Hawes: "More problems in dup_mmap"