mm/madvise: introduce MADV_COLLAPSE sync hugepage collapse

This idea was introduced by David Rientjes[1].

Introduce a new madvise mode, MADV_COLLAPSE, that allows users to request
a synchronous collapse of memory at their own expense.

The benefits of this approach are:

* CPU is charged to the process that wants to spend the cycles for the
THP
* Avoid unpredictable timing of khugepaged collapse

Semantics

This call is independent of the system-wide THP sysfs settings, but will
fail for memory marked VM_NOHUGEPAGE. If the ranges provided span
multiple VMAs, the semantics of the collapse over each VMA is independent
from the others. This implies a hugepage cannot cross a VMA boundary. If
collapse of a given hugepage-aligned/sized region fails, the operation may
continue to attempt collapsing the remainder of memory specified.

The memory ranges provided must be page-aligned, but are not required to
be hugepage-aligned. If the memory ranges are not hugepage-aligned, the
start/end of the range will be clamped to the first/last hugepage-aligned
address covered by said range. The memory ranges must span at least one
hugepage-sized region.

All non-resident pages covered by the range will first be
swapped/faulted-in, before being internally copied onto a freshly
allocated hugepage. Unmapped pages will have their data directly
initialized to 0 in the new hugepage. However, for every eligible
hugepage aligned/sized region to-be collapsed, at least one page must
currently be backed by memory (a PMD covering the address range must
already exist).

Allocation for the new hugepage may enter direct reclaim and/or
compaction, regardless of VMA flags. When the system has multiple NUMA
nodes, the hugepage will be allocated from the node providing the most
native pages. This operation operates on the current state of the
specified process and makes no persistent changes or guarantees on how
pages will be mapped, constructed, or faulted in the future

Return Value

If all hugepage-sized/aligned regions covered by the provided range were
either successfully collapsed, or were already PMD-mapped THPs, this
operation will be deemed successful. On success, process_madvise(2)
returns the number of bytes advised, and madvise(2) returns 0. Else, -1
is returned and errno is set to indicate the error for the most-recently
attempted hugepage collapse. Note that many failures might have occurred,
since the operation may continue to collapse in the event a single
hugepage-sized/aligned region fails.

ENOMEM Memory allocation failed or VMA not found
EBUSY Memcg charging failed
EAGAIN Required resource temporarily unavailable. Try again
might succeed.
EINVAL Other error: No PMD found, subpage doesn't have Present
bit set, "Special" page no backed by struct page, VMA
incorrectly sized, address not page-aligned, ...

Most notable here is ENOMEM and EBUSY (new to madvise) which are intended
to provide the caller with actionable feedback so they may take an
appropriate fallback measure.

Use Cases

An immediate user of this new functionality are malloc() implementations
that manage memory in hugepage-sized chunks, but sometimes subrelease
memory back to the system in native-sized chunks via MADV_DONTNEED;
zapping the pmd. Later, when the memory is hot, the implementation could
madvise(MADV_COLLAPSE) to re-back the memory by THPs to regain hugepage
coverage and dTLB performance. TCMalloc is such an implementation that
could benefit from this[2].

Only privately-mapped anon memory is supported for now, but additional
support for file, shmem, and HugeTLB high-granularity mappings[2] is
expected. File and tmpfs/shmem support would permit:

* Backing executable text by THPs. Current support provided by
CONFIG_READ_ONLY_THP_FOR_FS may take a long time on a large system which
might impair services from serving at their full rated load after
(re)starting. Tricks like mremap(2)'ing text onto anonymous memory to
immediately realize iTLB performance prevents page sharing and demand
paging, both of which increase steady state memory footprint. With
MADV_COLLAPSE, we get the best of both worlds: Peak upfront performance
and lower RAM footprints.
* Backing guest memory by hugapages after the memory contents have been
migrated in native-page-sized chunks to a new host, in a
userfaultfd-based live-migration stack.

[1] https://lore.kernel.org/linux-mm/[e-mail]/
[2] https://github.com/google/tcmalloc/tree/master/tcmalloc

[[e-mail]: avoid possible memory leak in failure path]
Link: https://lkml.kernel.org/r/[e-mail]
[[e-mail] add missing kfree() to madvise_collapse()]
Link: https://lore.kernel.org/linux-mm/[e-mail]/
Link: https://lkml.kernel.org/r/[e-mail]
[[e-mail]: delay computation of hpage boundaries until use]]
Link: https://lkml.kernel.org/r/[e-mail]
Link: https://lkml.kernel.org/r/[e-mail]
Signed-off-by: Zach O'Keefe <[e-mail]>
Signed-off-by: "Souptick Joarder (HPE)" <[e-mail]>
Suggested-by: David Rientjes <[e-mail]>
Cc: Alex Shi <[e-mail]>
Cc: Andrea Arcangeli <[e-mail]>
Cc: Arnd Bergmann <[e-mail]>
Cc: Axel Rasmussen <[e-mail]>
Cc: Chris Kennelly <[e-mail]>
Cc: Chris Zankel <[e-mail]>
Cc: David Hildenbrand <[e-mail]>
Cc: Helge Deller <[e-mail]>
Cc: Hugh Dickins <[e-mail]>
Cc: Ivan Kokshaysky <[e-mail]>
Cc: James Bottomley <[e-mail]>
Cc: Jens Axboe <[e-mail]>
Cc: "Kirill A. Shutemov" <[e-mail]>
Cc: Matthew Wilcox <[e-mail]>
Cc: Matt Turner <[e-mail]>
Cc: Max Filippov <[e-mail]>
Cc: Miaohe Lin <[e-mail]>
Cc: Michal Hocko <[e-mail]>
Cc: Minchan Kim <[e-mail]>
Cc: Pasha Tatashin <[e-mail]>
Cc: Pavel Begunkov <[e-mail]>
Cc: Peter Xu <[e-mail]>
Cc: Rongwei Wang <[e-mail]>
Cc: SeongJae Park <[e-mail]>
Cc: Song Liu <[e-mail]>
Cc: Thomas Bogendoerfer <[e-mail]>
Cc: Vlastimil Babka <[e-mail]>
Cc: Yang Shi <[e-mail]>
Cc: Zi Yan <[e-mail]>
Cc: Dan Carpenter <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>

arch/alpha/include/uapi/asm/mman.h

···
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 

arch/mips/include/uapi/asm/mman.h

···
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 

arch/parisc/include/uapi/asm/mman.h

···
 #define MADV_WIPEONFORK 71		/* Zero memory on fork, child only */
 #define MADV_KEEPONFORK 72		/* Undo MADV_WIPEONFORK */
 
+#define MADV_COLLAPSE	73		/* Synchronous hugepage collapse */
+
 #define MADV_HWPOISON     100		/* poison a page for testing */
 #define MADV_SOFT_OFFLINE 101		/* soft offline page for testing */
 

arch/xtensa/include/uapi/asm/mman.h

···
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 

include/linux/huge_mm.h

···
 
 int hugepage_madvise(struct vm_area_struct *vma, unsigned long *vm_flags,
 		     int advice);
+int madvise_collapse(struct vm_area_struct *vma,
+		     struct vm_area_struct **prev,
+		     unsigned long start, unsigned long end);
 void vma_adjust_trans_huge(struct vm_area_struct *vma, unsigned long start,
 			   unsigned long end, long adjust_next);
 spinlock_t *__pmd_trans_huge_lock(pmd_t *pmd, struct vm_area_struct *vma);
···
 static inline int hugepage_madvise(struct vm_area_struct *vma,
 				   unsigned long *vm_flags, int advice)
 {
-	BUG();
-	return 0;
+	return -EINVAL;
 }
+
+static inline int madvise_collapse(struct vm_area_struct *vma,
+				   struct vm_area_struct **prev,
+				   unsigned long start, unsigned long end)
+{
+	return -EINVAL;
+}
+
 static inline void vma_adjust_trans_huge(struct vm_area_struct *vma,
 					 unsigned long start,
 					 unsigned long end,

include/uapi/asm-generic/mman-common.h

···
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 

mm/khugepaged.c

···
 			      struct collapse_control *cc)
 {
 	/* Only allocate from the target node */
-	gfp_t gfp = alloc_hugepage_khugepaged_gfpmask() | __GFP_THISNODE;
+	gfp_t gfp = (cc->is_khugepaged ? alloc_hugepage_khugepaged_gfpmask() :
+		     GFP_TRANSHUGE) | __GFP_THISNODE;
 	int node = khugepaged_find_target_node(cc);
 
 	if (!khugepaged_alloc_page(hpage, gfp, node))
···
 	if (hugepage_flags_enabled() && khugepaged_thread)
 		set_recommended_min_free_kbytes();
 	mutex_unlock(&khugepaged_mutex);
+}
+
+static int madvise_collapse_errno(enum scan_result r)
+{
+	/*
+	 * MADV_COLLAPSE breaks from existing madvise(2) conventions to provide
+	 * actionable feedback to caller, so they may take an appropriate
+	 * fallback measure depending on the nature of the failure.
+	 */
+	switch (r) {
+	case SCAN_ALLOC_HUGE_PAGE_FAIL:
+		return -ENOMEM;
+	case SCAN_CGROUP_CHARGE_FAIL:
+		return -EBUSY;
+	/* Resource temporary unavailable - trying again might succeed */
+	case SCAN_PAGE_LOCK:
+	case SCAN_PAGE_LRU:
+		return -EAGAIN;
+	/*
+	 * Other: Trying again likely not to succeed / error intrinsic to
+	 * specified memory range. khugepaged likely won't be able to collapse
+	 * either.
+	 */
+	default:
+		return -EINVAL;
+	}
+}
+
+int madvise_collapse(struct vm_area_struct *vma, struct vm_area_struct **prev,
+		     unsigned long start, unsigned long end)
+{
+	struct collapse_control *cc;
+	struct mm_struct *mm = vma->vm_mm;
+	unsigned long hstart, hend, addr;
+	int thps = 0, last_fail = SCAN_FAIL;
+	bool mmap_locked = true;
+
+	BUG_ON(vma->vm_start > start);
+	BUG_ON(vma->vm_end < end);
+
+	*prev = vma;
+
+	/* TODO: Support file/shmem */
+	if (!vma->anon_vma || !vma_is_anonymous(vma))
+		return -EINVAL;
+
+	if (!hugepage_vma_check(vma, vma->vm_flags, false, false, false))
+		return -EINVAL;
+
+	cc = kmalloc(sizeof(*cc), GFP_KERNEL);
+	if (!cc)
+		return -ENOMEM;
+	cc->is_khugepaged = false;
+	cc->last_target_node = NUMA_NO_NODE;
+
+	mmgrab(mm);
+	lru_add_drain_all();
+
+	hstart = (start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK;
+	hend = end & HPAGE_PMD_MASK;
+
+	for (addr = hstart; addr < hend; addr += HPAGE_PMD_SIZE) {
+		int result = SCAN_FAIL;
+
+		if (!mmap_locked) {
+			cond_resched();
+			mmap_read_lock(mm);
+			mmap_locked = true;
+			result = hugepage_vma_revalidate(mm, addr, &vma, cc);
+			if (result  != SCAN_SUCCEED) {
+				last_fail = result;
+				goto out_nolock;
+			}
+		}
+		mmap_assert_locked(mm);
+		memset(cc->node_load, 0, sizeof(cc->node_load));
+		result = khugepaged_scan_pmd(mm, vma, addr, &mmap_locked, cc);
+		if (!mmap_locked)
+			*prev = NULL;  /* Tell caller we dropped mmap_lock */
+
+		switch (result) {
+		case SCAN_SUCCEED:
+		case SCAN_PMD_MAPPED:
+			++thps;
+			break;
+		/* Whitelisted set of results where continuing OK */
+		case SCAN_PMD_NULL:
+		case SCAN_PTE_NON_PRESENT:
+		case SCAN_PTE_UFFD_WP:
+		case SCAN_PAGE_RO:
+		case SCAN_LACK_REFERENCED_PAGE:
+		case SCAN_PAGE_NULL:
+		case SCAN_PAGE_COUNT:
+		case SCAN_PAGE_LOCK:
+		case SCAN_PAGE_COMPOUND:
+		case SCAN_PAGE_LRU:
+			last_fail = result;
+			break;
+		default:
+			last_fail = result;
+			/* Other error, exit */
+			goto out_maybelock;
+		}
+	}
+
+out_maybelock:
+	/* Caller expects us to hold mmap_lock on return */
+	if (!mmap_locked)
+		mmap_read_lock(mm);
+out_nolock:
+	mmap_assert_locked(mm);
+	mmdrop(mm);
+	kfree(cc);
+
+	return thps == ((hend - hstart) >> HPAGE_PMD_SHIFT) ? 0
+			: madvise_collapse_errno(last_fail);
 }

mm/madvise.c

···
 	case MADV_FREE:
 	case MADV_POPULATE_READ:
 	case MADV_POPULATE_WRITE:
+	case MADV_COLLAPSE:
 		return 0;
 	default:
 		/* be safe, default to 1. list exceptions explicitly */
···
 		if (error)
 			goto out;
 		break;
+	case MADV_COLLAPSE:
+		return madvise_collapse(vma, prev, start, end);
 	}
 
 	anon_name = anon_vma_name(vma);
···
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 	case MADV_HUGEPAGE:
 	case MADV_NOHUGEPAGE:
+	case MADV_COLLAPSE:
 #endif
 	case MADV_DONTDUMP:
 	case MADV_DODUMP:
···
  *  MADV_NOHUGEPAGE - mark the given range as not worth being backed by
  *		transparent huge pages so the existing pages will not be
  *		coalesced into THP and new pages will not be allocated as THP.
+ *  MADV_COLLAPSE - synchronously coalesce pages into new THP.
  *  MADV_DONTDUMP - the application wants to prevent pages in the given range
  *		from being included in its core dump.
  *  MADV_DODUMP - cancel MADV_DONTDUMP: no longer exclude from core dump.

tools/include/uapi/asm-generic/mman-common.h

···
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0