Optimize anonymous swapbacked large folio unmapping

Speed up unmapping of anonymous swapbacked large folios by clearing the ptes, and setting swap ptes, in one go.

The following benchmark (stolen from Barry) is used to measure the time taken to swapout 256M worth of memory backed by 64K large folios:

#define _GNU_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <string.h>
#include <time.h>
#include <unistd.h>
#include <errno.h>

#define SIZE_MB 256
#define SIZE_BYTES (SIZE_MB * 1024 * 1024)

 int main() {
     void *addr = mmap(NULL, SIZE_BYTES, PROT_READ | PROT_WRITE,
                       MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
     if (addr == MAP_FAILED) {
         perror("mmap failed");
         return 1;
     }

     memset(addr, 0, SIZE_BYTES);

     struct timespec start, end;
     clock_gettime(CLOCK_MONOTONIC, &start);

     if (madvise(addr, SIZE_BYTES, MADV_PAGEOUT) != 0) {
         perror("madvise(MADV_PAGEOUT) failed");
         munmap(addr, SIZE_BYTES);
         return 1;
     }

     clock_gettime(CLOCK_MONOTONIC, &end);

     long duration_ns = (end.tv_sec - start.tv_sec) * 1e9 +
                        (end.tv_nsec - start.tv_nsec);
     printf("madvise(MADV_PAGEOUT) took %ld ns (%.3f ms)\n",
            duration_ns, duration_ns / 1e6);

     munmap(addr, SIZE_BYTES);
     return 0;

}

Performance as measured on a Linux VM on Apple M3 (arm64):

Vanilla - Mean: 37401913 ns, std dev: 12%
Patched - Mean: 17420282 ns, std dev: 11%

resulting in more than 2x speedup.

No regression observed on 4K folios.

Performance as measured on bare metal x86:

Vanilla - mean: 54986286 ns, std dev: 1.5%
Patched - mean: 51930795 ns, std dev: 3%

I tried magnifying the difference on x86 by using 1M large folios, but can't spot an obvious improvement (looks like my system is too fast to benefit from batched atomic operations!), hinting that the benefit lies mainly in the reduction of ptep_get() calls and the reduction of TLB flushes during contpte-unfolding, on arm64.

No regression is observed on 4K folios on x86 too.

---
Applies on mm-unstable (24231596edf2).

mm-selftests pass.

This breakout patchset is the final completion of
https://lore.kernel.org/all/[email protected]/

The extra addition is the generic set_softleaf_ptes() helper.

v3->v4:
- Make page_anon_exclusive_batch return the exclusivity instead of computing that in the caller
- Make ttu_anon_folio() return the number of unmapped pages, to simplify the code

v2->v3:
- page_anon_exclusive_batch returns unsigned long, shift to rmap.h
- Drop a para in description of patch 5, use softleaf_is_swap to gate pte_swp_exclusive
- finish_folio_unmap -> finish_folio_unmap_batch
- Add patch 9 to do batching for sparc

v1->v2:
- Add WARNs for folio span in folio_*_swap_pages
- Trivial changes to page_anon_exclusive_batch
- Add two sanity checks in patch 4

Dev Jain (9):
mm/swapfile: add batched version of folio_dup_swap mm/swapfile: add batched version of folio_put_swap
mm: move anon-exclusive batch helper to rmap.h mm/rmap: Add batched version of folio_try_share_anon_rmap_pte mm/internal: rename swap offset helpers to softleaf offset mm/internal: add set_softleaf_ptes
mm/memory: use set_softleaf_ptes for uffd-wp markers mm/rmap: batch unmap anonymous swap-backed large folios mm, sparc: batch arch_unmap_one()

 arch/sparc/include/asm/pgtable_64.h |  10 ++-
 arch/sparc/kernel/adi_64.c          |  21 +++++
 include/linux/pgtable.h             |   6 +-
 include/linux/rmap.h                |  76 ++++++++++++++----
 mm/internal.h                       |  64 +++++++++++----
 mm/memory.c                         |  23 +++---
 mm/mprotect.c                       |  25 +-----
 mm/rmap.c                           | 119 ++++++++++++++++++++--------
 mm/shmem.c                          |   8 +-
 mm/swap.h                           |  35 +++++++-
 mm/swapfile.c                       |  39 +++++----
 11 files changed, 289 insertions(+), 137 deletions(-)

2.43.0