mm: account pmd page tables to the process

Dave noticed that unprivileged process can allocate significant amount of
memory -- >500 MiB on x86_64 -- and stay unnoticed by oom-killer and
memory cgroup.  The trick is to allocate a lot of PMD page tables.  Linux
kernel doesn't account PMD tables to the process, only PTE.

The use-cases below use few tricks to allocate a lot of PMD page tables
while keeping VmRSS and VmPTE low.  oom_score for the process will be 0.

	#include <errno.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <unistd.h>
	#include <sys/mman.h>
	#include <sys/prctl.h>

	#define PUD_SIZE (1UL << 30)
	#define PMD_SIZE (1UL << 21)

	#define NR_PUD 130000

	int main(void)
	{
		char *addr = NULL;
		unsigned long i;

		prctl(PR_SET_THP_DISABLE);
		for (i = 0; i < NR_PUD ; i++) {
			addr = mmap(addr + PUD_SIZE, PUD_SIZE, PROT_WRITE|PROT_READ,
					MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
			if (addr == MAP_FAILED) {
				perror("mmap");
				break;
			}
			*addr = 'x';
			munmap(addr, PMD_SIZE);
			mmap(addr, PMD_SIZE, PROT_WRITE|PROT_READ,
					MAP_ANONYMOUS|MAP_PRIVATE|MAP_FIXED, -1, 0);
			if (addr == MAP_FAILED)
				perror("re-mmap"), exit(1);
		}
		printf("PID %d consumed %lu KiB in PMD page tables\n",
				getpid(), i * 4096 >> 10);
		return pause();
	}

The patch addresses the issue by account PMD tables to the process the
same way we account PTE.

The main place where PMD tables is accounted is __pmd_alloc() and
free_pmd_range(). But there're few corner cases:

 - HugeTLB can share PMD page tables. The patch handles by accounting
   the table to all processes who share it.

 - x86 PAE pre-allocates few PMD tables on fork.

 - Architectures with FIRST_USER_ADDRESS > 0. We need to adjust sanity
   check on exit(2).

Accounting only happens on configuration where PMD page table's level is
present (PMD is not folded).  As with nr_ptes we use per-mm counter.  The
counter value is used to calculate baseline for badness score by
oom-killer.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Reported-by: Dave Hansen <[email protected]>
Cc: Hugh Dickins <[email protected]>
Reviewed-by: Cyrill Gorcunov <[email protected]>
Cc: Pavel Emelyanov <[email protected]>
Cc: David Rientjes <[email protected]>
Tested-by: Sedat Dilek <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Documentation/sysctl/vm.txt

@@ -555,12 +555,12 @@ this is causing problems for your system/application.
 
 oom_dump_tasks
 
-Enables a system-wide task dump (excluding kernel threads) to be
-produced when the kernel performs an OOM-killing and includes such
-information as pid, uid, tgid, vm size, rss, nr_ptes, swapents,
-oom_score_adj score, and name.  This is helpful to determine why the
-OOM killer was invoked, to identify the rogue task that caused it,
-and to determine why the OOM killer chose the task it did to kill.
+Enables a system-wide task dump (excluding kernel threads) to be produced
+when the kernel performs an OOM-killing and includes such information as
+pid, uid, tgid, vm size, rss, nr_ptes, nr_pmds, swapents, oom_score_adj
+score, and name.  This is helpful to determine why the OOM killer was
+invoked, to identify the rogue task that caused it, and to determine why
+the OOM killer chose the task it did to kill.
 
 If this is set to zero, this information is suppressed.  On very
 large systems with thousands of tasks it may not be feasible to dump

arch/x86/mm/pgtable.c

@@ -190,18 +190,19 @@ void pud_populate(struct mm_struct *mm, pud_t *pudp, pmd_t *pmd)
 
 #endif	/* CONFIG_X86_PAE */
 
-static void free_pmds(pmd_t *pmds[])
+static void free_pmds(struct mm_struct *mm, pmd_t *pmds[])
 {
 	int i;
 
 	for(i = 0; i < PREALLOCATED_PMDS; i++)
 		if (pmds[i]) {
 			pgtable_pmd_page_dtor(virt_to_page(pmds[i]));
 			free_page((unsigned long)pmds[i]);
+			mm_dec_nr_pmds(mm);
 		}
 }
 
-static int preallocate_pmds(pmd_t *pmds[])
+static int preallocate_pmds(struct mm_struct *mm, pmd_t *pmds[])
 {
 	int i;
 	bool failed = false;
@@ -215,11 +216,13 @@ static int preallocate_pmds(pmd_t *pmds[])
 			pmd = NULL;
 			failed = true;
 		}
+		if (pmd)
+			mm_inc_nr_pmds(mm);
 		pmds[i] = pmd;
 	}
 
 	if (failed) {
-		free_pmds(pmds);
+		free_pmds(mm, pmds);
 		return -ENOMEM;
 	}
 
@@ -246,6 +249,7 @@ static void pgd_mop_up_pmds(struct mm_struct *mm, pgd_t *pgdp)
 
 			paravirt_release_pmd(pgd_val(pgd) >> PAGE_SHIFT);
 			pmd_free(mm, pmd);
+			mm_dec_nr_pmds(mm);
 		}
 	}
 }
@@ -283,7 +287,7 @@ pgd_t *pgd_alloc(struct mm_struct *mm)
 
 	mm->pgd = pgd;
 
-	if (preallocate_pmds(pmds) != 0)
+	if (preallocate_pmds(mm, pmds) != 0)
 		goto out_free_pgd;
 
 	if (paravirt_pgd_alloc(mm) != 0)
@@ -304,7 +308,7 @@ pgd_t *pgd_alloc(struct mm_struct *mm)
 	return pgd;
 
 out_free_pmds:
-	free_pmds(pmds);
+	free_pmds(mm, pmds);
 out_free_pgd:
 	free_page((unsigned long)pgd);
 out:

fs/proc/task_mmu.c

@@ -21,7 +21,7 @@
 
 void task_mem(struct seq_file *m, struct mm_struct *mm)
 {
-	unsigned long data, text, lib, swap;
+	unsigned long data, text, lib, swap, ptes, pmds;
 	unsigned long hiwater_vm, total_vm, hiwater_rss, total_rss;
 
 	/*
@@ -42,6 +42,8 @@ void task_mem(struct seq_file *m, struct mm_struct *mm)
 	text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) >> 10;
 	lib = (mm->exec_vm << (PAGE_SHIFT-10)) - text;
 	swap = get_mm_counter(mm, MM_SWAPENTS);
+	ptes = PTRS_PER_PTE * sizeof(pte_t) * atomic_long_read(&mm->nr_ptes);
+	pmds = PTRS_PER_PMD * sizeof(pmd_t) * mm_nr_pmds(mm);
 	seq_printf(m,
 		"VmPeak:\t%8lu kB\n"
 		"VmSize:\t%8lu kB\n"
@@ -54,6 +56,7 @@ void task_mem(struct seq_file *m, struct mm_struct *mm)
 		"VmExe:\t%8lu kB\n"
 		"VmLib:\t%8lu kB\n"
 		"VmPTE:\t%8lu kB\n"
+		"VmPMD:\t%8lu kB\n"
 		"VmSwap:\t%8lu kB\n",
 		hiwater_vm << (PAGE_SHIFT-10),
 		total_vm << (PAGE_SHIFT-10),
@@ -63,8 +66,8 @@ void task_mem(struct seq_file *m, struct mm_struct *mm)
 		total_rss << (PAGE_SHIFT-10),
 		data << (PAGE_SHIFT-10),
 		mm->stack_vm << (PAGE_SHIFT-10), text, lib,
-		(PTRS_PER_PTE * sizeof(pte_t) *
-		 atomic_long_read(&mm->nr_ptes)) >> 10,
+		ptes >> 10,
+		pmds >> 10,
 		swap << (PAGE_SHIFT-10));
 }
 

include/linux/mm.h

@@ -1438,8 +1438,32 @@ static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
 {
 	return 0;
 }
+
+static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
+{
+	return 0;
+}
+
+static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
+static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
+
 #else
 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
+
+static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
+{
+	return atomic_long_read(&mm->nr_pmds);
+}
+
+static inline void mm_inc_nr_pmds(struct mm_struct *mm)
+{
+	atomic_long_inc(&mm->nr_pmds);
+}
+
+static inline void mm_dec_nr_pmds(struct mm_struct *mm)
+{
+	atomic_long_dec(&mm->nr_pmds);
+}
 #endif
 
 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,

include/linux/mm_types.h

@@ -363,7 +363,8 @@ struct mm_struct {
 	pgd_t * pgd;
 	atomic_t mm_users;			/* How many users with user space? */
 	atomic_t mm_count;			/* How many references to "struct mm_struct" (users count as 1) */
-	atomic_long_t nr_ptes;			/* Page table pages */
+	atomic_long_t nr_ptes;			/* PTE page table pages */
+	atomic_long_t nr_pmds;			/* PMD page table pages */
 	int map_count;				/* number of VMAs */
 
 	spinlock_t page_table_lock;		/* Protects page tables and some counters */

kernel/fork.c

@@ -555,6 +555,9 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
 	INIT_LIST_HEAD(&mm->mmlist);
 	mm->core_state = NULL;
 	atomic_long_set(&mm->nr_ptes, 0);
+#ifndef __PAGETABLE_PMD_FOLDED
+	atomic_long_set(&mm->nr_pmds, 0);
+#endif
 	mm->map_count = 0;
 	mm->locked_vm = 0;
 	mm->pinned_vm = 0;

mm/debug.c

@@ -173,7 +173,7 @@ void dump_mm(const struct mm_struct *mm)
 		"get_unmapped_area %p\n"
 #endif
 		"mmap_base %lu mmap_legacy_base %lu highest_vm_end %lu\n"
-		"pgd %p mm_users %d mm_count %d nr_ptes %lu map_count %d\n"
+		"pgd %p mm_users %d mm_count %d nr_ptes %lu nr_pmds %lu map_count %d\n"
 		"hiwater_rss %lx hiwater_vm %lx total_vm %lx locked_vm %lx\n"
 		"pinned_vm %lx shared_vm %lx exec_vm %lx stack_vm %lx\n"
 		"start_code %lx end_code %lx start_data %lx end_data %lx\n"
@@ -206,6 +206,7 @@ void dump_mm(const struct mm_struct *mm)
 		mm->pgd, atomic_read(&mm->mm_users),
 		atomic_read(&mm->mm_count),
 		atomic_long_read((atomic_long_t *)&mm->nr_ptes),
+		mm_nr_pmds((struct mm_struct *)mm),
 		mm->map_count,
 		mm->hiwater_rss, mm->hiwater_vm, mm->total_vm, mm->locked_vm,
 		mm->pinned_vm, mm->shared_vm, mm->exec_vm, mm->stack_vm,

mm/hugetlb.c

@@ -3598,6 +3598,7 @@ pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
 		if (saddr) {
 			spte = huge_pte_offset(svma->vm_mm, saddr);
 			if (spte) {
+				mm_inc_nr_pmds(mm);
 				get_page(virt_to_page(spte));
 				break;
 			}
@@ -3609,11 +3610,13 @@ pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
 
 	ptl = huge_pte_lockptr(hstate_vma(vma), mm, spte);
 	spin_lock(ptl);
-	if (pud_none(*pud))
+	if (pud_none(*pud)) {
 		pud_populate(mm, pud,
 				(pmd_t *)((unsigned long)spte & PAGE_MASK));
-	else
+	} else {
 		put_page(virt_to_page(spte));
+		mm_inc_nr_pmds(mm);
+	}
 	spin_unlock(ptl);
 out:
 	pte = (pte_t *)pmd_alloc(mm, pud, addr);
@@ -3644,6 +3647,7 @@ int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
 
 	pud_clear(pud);
 	put_page(virt_to_page(ptep));
+	mm_dec_nr_pmds(mm);
 	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
 	return 1;
 }

mm/memory.c

@@ -428,6 +428,7 @@ static inline void free_pmd_range(struct mmu_gather *tlb, pud_t *pud,
 	pmd = pmd_offset(pud, start);
 	pud_clear(pud);
 	pmd_free_tlb(tlb, pmd, start);
+	mm_dec_nr_pmds(tlb->mm);
 }
 
 static inline void free_pud_range(struct mmu_gather *tlb, pgd_t *pgd,
@@ -3322,15 +3323,17 @@ int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 
 	spin_lock(&mm->page_table_lock);
 #ifndef __ARCH_HAS_4LEVEL_HACK
-	if (pud_present(*pud))		/* Another has populated it */
-		pmd_free(mm, new);
-	else
+	if (!pud_present(*pud)) {
+		mm_inc_nr_pmds(mm);
 		pud_populate(mm, pud, new);
-#else
-	if (pgd_present(*pud))		/* Another has populated it */
+	} else	/* Another has populated it */
 		pmd_free(mm, new);
-	else
+#else
+	if (!pgd_present(*pud)) {
+		mm_inc_nr_pmds(mm);
 		pgd_populate(mm, pud, new);
+	} else /* Another has populated it */
+		pmd_free(mm, new);
 #endif /* __ARCH_HAS_4LEVEL_HACK */
 	spin_unlock(&mm->page_table_lock);
 	return 0;

mm/mmap.c

@@ -2853,7 +2853,9 @@ void exit_mmap(struct mm_struct *mm)
 	vm_unacct_memory(nr_accounted);
 
 	WARN_ON(atomic_long_read(&mm->nr_ptes) >
-			(FIRST_USER_ADDRESS+PMD_SIZE-1)>>PMD_SHIFT);
+			round_up(FIRST_USER_ADDRESS, PMD_SIZE) >> PMD_SHIFT);
+	WARN_ON(mm_nr_pmds(mm) >
+			round_up(FIRST_USER_ADDRESS, PUD_SIZE) >> PUD_SHIFT);
 }
 
 /* Insert vm structure into process list sorted by address

mm/oom_kill.c

@@ -169,8 +169,8 @@ unsigned long oom_badness(struct task_struct *p, struct mem_cgroup *memcg,
 	 * The baseline for the badness score is the proportion of RAM that each
 	 * task's rss, pagetable and swap space use.
 	 */
-	points = get_mm_rss(p->mm) + atomic_long_read(&p->mm->nr_ptes) +
-		 get_mm_counter(p->mm, MM_SWAPENTS);
+	points = get_mm_rss(p->mm) + get_mm_counter(p->mm, MM_SWAPENTS) +
+		atomic_long_read(&p->mm->nr_ptes) + mm_nr_pmds(p->mm);
 	task_unlock(p);
 
 	/*
@@ -351,7 +351,7 @@ static void dump_tasks(struct mem_cgroup *memcg, const nodemask_t *nodemask)
 	struct task_struct *p;
 	struct task_struct *task;
 
-	pr_info("[ pid ]   uid  tgid total_vm      rss nr_ptes swapents oom_score_adj name\n");
+	pr_info("[ pid ]   uid  tgid total_vm      rss nr_ptes nr_pmds swapents oom_score_adj name\n");
 	rcu_read_lock();
 	for_each_process(p) {
 		if (oom_unkillable_task(p, memcg, nodemask))
@@ -367,10 +367,11 @@ static void dump_tasks(struct mem_cgroup *memcg, const nodemask_t *nodemask)
 			continue;
 		}
 
-		pr_info("[%5d] %5d %5d %8lu %8lu %7ld %8lu         %5hd %s\n",
+		pr_info("[%5d] %5d %5d %8lu %8lu %7ld %7ld %8lu         %5hd %s\n",
 			task->pid, from_kuid(&init_user_ns, task_uid(task)),
 			task->tgid, task->mm->total_vm, get_mm_rss(task->mm),
 			atomic_long_read(&task->mm->nr_ptes),
+			mm_nr_pmds(task->mm),
 			get_mm_counter(task->mm, MM_SWAPENTS),
 			task->signal->oom_score_adj, task->comm);
 		task_unlock(task);