Re: [PATCH] mm/gup: restore the ability to pin more than 2GB at a time

[Jason Gunthorpe <jgg@nvidia.com>]

On Tue, Oct 29, 2024 at 09:42:10PM -0700, Christoph Hellwig wrote:
> On Tue, Oct 29, 2024 at 09:39:15PM -0700, John Hubbard wrote:
> > I expect I could piece together something with Nouveau, given enough
> > time and help from Ben Skeggs and Danillo and all...
> > 
> > Yes, this originated with the out of tree driver. But it never occurred
> > to me that upstream be uninterested in an obvious fix to an obvious
> > regression.
> 
> Because pinning down these amounts of memoryt is completely insane.
> I don't mind the switch to kvmalloc, but we need to put in an upper
> bound of what can be pinned.

I have RDMA customers that pin basically the entire machine's
memory, which is 100Gb's of RAM. It is a dedicated server doing some
HPC or Database job and it has been carefully setup to do that.

Many VFIO VM deployments will pin almost all the memory too - just a
little left over for the hypervisor. Again the machine was carefully
setup to have huge pages that cover almost all system ram and are
dedicated to backing memory for VMs.

Pinning almost all memory is a legitimate use case.

Jason