Re: [PATCH v2 0/4] extend vmalloc support for constrained allocations

[Dave Chinner <david@fromorbit.com>]

On Mon, Nov 22, 2021 at 04:32:29PM +0100, Michal Hocko wrote:
> Hi,
> The previous version has been posted here [1] 
> 
> I hope I have addressed all the feedback. There were some suggestions
> for further improvements but I would rather make this smaller as I
> cannot really invest more time and I believe further changes can be done
> on top.
> 
> This version is a rebase on top of the current Linus tree. Except for
> the review feedback and conflicting changes in the area there is only
> one change to filter out __GFP_NOFAIL from the bulk allocator. This is
> not necessary strictly speaking AFAICS but I found it less confusing
> because vmalloc has its fallback strategy and the bulk allocator is
> meant only for the fast path.
> 
> Original cover:
> Based on a recent discussion with Dave and Neil [2] I have tried to
> implement NOFS, NOIO, NOFAIL support for the vmalloc to make
> life of kvmalloc users easier.
> 
> A requirement for NOFAIL support for kvmalloc was new to me but this
> seems to be really needed by the xfs code.
> 
> NOFS/NOIO was a known and a long term problem which was hoped to be
> handled by the scope API. Those scope should have been used at the
> reclaim recursion boundaries both to document them and also to remove
> the necessity of NOFS/NOIO constrains for all allocations within that
> scope. Instead workarounds were developed to wrap a single allocation
> instead (like ceph_kvmalloc).
> 
> First patch implements NOFS/NOIO support for vmalloc. The second one
> adds NOFAIL support and the third one bundles all together into kvmalloc
> and drops ceph_kvmalloc which can use kvmalloc directly now.
> 
> I hope I haven't missed anything in the vmalloc allocator.

Correct __GFP_NOLOCKDEP support is also needed. See:

https://lore.kernel.org/linux-mm/20211119225435.GZ449541@dread.disaster.area/

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com