Re: [PATCH] munmap sealed memory cause memory to split (bug)

From: Jeff Xu
Date: Thu Oct 17 2024 - 16:12:58 EST


On Thu, Oct 17, 2024 at 12:14 PM Pedro Falcato <pedro.falcato@xxxxxxxxx> wrote:
>
> On Thu, Oct 17, 2024 at 09:20:20AM -0700, Jeff Xu wrote:
> > On Thu, Oct 17, 2024 at 1:18 AM Lorenzo Stoakes
> > <lorenzo.stoakes@xxxxxxxxxx> wrote:
> > >
> > > NACK. Greg's bot got to it but...
> > >
> > > As per Greg's bot, no signed-off-by line.
> > >
> > Sorry for confusion, I wasn't meant to send this as a PATCH, but
> > reporting the issue.
> > The diff was just sent as reference to repro the bug, and I forgot to
> > remove PATCH from the title. I apologize for the confusion.
> >
>
> Can you explain what the issue is? I don't get it.
>
The issue is there is one VMA that gets splitted after an unmap call fails.
Two splitted VMA share the same attributes. e.g.

- Allocate 12 pages (0-11).
- Seal middle 4 pages (4567)
- munmap (2345) - this will fail due to 4567 being sealed.

The VMA for page (0123) is split as 2 VMAs (01)-(23), those 2 VMA
have the same attribute, and should be merged as one.


> --
> Pedro