Re: Deadlock during page-in and thread_suspend

Hello,

Michael Kelly, le dim. 28 sept. 2025 09:33:32 +0100, a ecrit:
> It seems to me that there are many inter-dependencies across areas
> of code throughout gnumach that are difficult to find without a very
> thorough knowledge. That in itself makes it hard for newcomers to
> contribute.

Memory management is well-known to be the hardest thing that operating
systems have to take care of. So, yew, it's really not an area for
newcomers, whatever the operating system :)

> A method that might require less functional change would be to somehow transfer
> the responsibility to complete page-in to another thread although I cannot see
> how that could be efficiently managed.

Re-reading the documentation for thread_suspend, it's only saying it's
meant to prevent userland code execution, not kernelland code execution.
So the actual suspension can be defered to the point of return to
userland, instead of halting the kernel code itself, so that the latter
can continue what it was doing.

Samuel