Revision c9d398fa237882ea07167e23bcfc5e6847066518 authored by Naoya Horiguchi on 31 March 2017, 22:11:55 UTC, committed by Linus Torvalds on 01 April 2017, 00:13:30 UTC
I found the race condition which triggers the following bug when
move_pages() and soft offline are called on a single hugetlb page
concurrently.

    Soft offlining page 0x119400 at 0x700000000000
    BUG: unable to handle kernel paging request at ffffea0011943820
    IP: follow_huge_pmd+0x143/0x190
    PGD 7ffd2067
    PUD 7ffd1067
    PMD 0
        [61163.582052] Oops: 0000 [#1] SMP
    Modules linked in: binfmt_misc ppdev virtio_balloon parport_pc pcspkr i2c_piix4 parport i2c_core acpi_cpufreq ip_tables xfs libcrc32c ata_generic pata_acpi virtio_blk 8139too crc32c_intel ata_piix serio_raw libata virtio_pci 8139cp virtio_ring virtio mii floppy dm_mirror dm_region_hash dm_log dm_mod [last unloaded: cap_check]
    CPU: 0 PID: 22573 Comm: iterate_numa_mo Tainted: P           OE   4.11.0-rc2-mm1+ #2
    Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
    RIP: 0010:follow_huge_pmd+0x143/0x190
    RSP: 0018:ffffc90004bdbcd0 EFLAGS: 00010202
    RAX: 0000000465003e80 RBX: ffffea0004e34d30 RCX: 00003ffffffff000
    RDX: 0000000011943800 RSI: 0000000000080001 RDI: 0000000465003e80
    RBP: ffffc90004bdbd18 R08: 0000000000000000 R09: ffff880138d34000
    R10: ffffea0004650000 R11: 0000000000c363b0 R12: ffffea0011943800
    R13: ffff8801b8d34000 R14: ffffea0000000000 R15: 000077ff80000000
    FS:  00007fc977710740(0000) GS:ffff88007dc00000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    CR2: ffffea0011943820 CR3: 000000007a746000 CR4: 00000000001406f0
    Call Trace:
     follow_page_mask+0x270/0x550
     SYSC_move_pages+0x4ea/0x8f0
     SyS_move_pages+0xe/0x10
     do_syscall_64+0x67/0x180
     entry_SYSCALL64_slow_path+0x25/0x25
    RIP: 0033:0x7fc976e03949
    RSP: 002b:00007ffe72221d88 EFLAGS: 00000246 ORIG_RAX: 0000000000000117
    RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007fc976e03949
    RDX: 0000000000c22390 RSI: 0000000000001400 RDI: 0000000000005827
    RBP: 00007ffe72221e00 R08: 0000000000c2c3a0 R09: 0000000000000004
    R10: 0000000000c363b0 R11: 0000000000000246 R12: 0000000000400650
    R13: 00007ffe72221ee0 R14: 0000000000000000 R15: 0000000000000000
    Code: 81 e4 ff ff 1f 00 48 21 c2 49 c1 ec 0c 48 c1 ea 0c 4c 01 e2 49 bc 00 00 00 00 00 ea ff ff 48 c1 e2 06 49 01 d4 f6 45 bc 04 74 90 <49> 8b 7c 24 20 40 f6 c7 01 75 2b 4c 89 e7 8b 47 1c 85 c0 7e 2a
    RIP: follow_huge_pmd+0x143/0x190 RSP: ffffc90004bdbcd0
    CR2: ffffea0011943820
    ---[ end trace e4f81353a2d23232 ]---
    Kernel panic - not syncing: Fatal exception
    Kernel Offset: disabled

This bug is triggered when pmd_present() returns true for non-present
hugetlb, so fixing the present check in follow_huge_pmd() prevents it.
Using pmd_present() to determine present/non-present for hugetlb is not
correct, because pmd_present() checks multiple bits (not only
_PAGE_PRESENT) for historical reason and it can misjudge hugetlb state.

Fixes: e66f17ff7177 ("mm/hugetlb: take page table lock in follow_huge_pmd()")
Link: http://lkml.kernel.org/r/1490149898-20231-1-git-send-email-n-horiguchi@ah.jp.nec.com
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: <stable@vger.kernel.org>        [4.0+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent 0cefabd
Raw File
rtmutex-debug.c
/*
 * RT-Mutexes: blocking mutual exclusion locks with PI support
 *
 * started by Ingo Molnar and Thomas Gleixner:
 *
 *  Copyright (C) 2004-2006 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
 *  Copyright (C) 2006 Timesys Corp., Thomas Gleixner <tglx@timesys.com>
 *
 * This code is based on the rt.c implementation in the preempt-rt tree.
 * Portions of said code are
 *
 *  Copyright (C) 2004  LynuxWorks, Inc., Igor Manyilov, Bill Huey
 *  Copyright (C) 2006  Esben Nielsen
 *  Copyright (C) 2006  Kihon Technologies Inc.,
 *			Steven Rostedt <rostedt@goodmis.org>
 *
 * See rt.c in preempt-rt for proper credits and further information
 */
#include <linux/sched.h>
#include <linux/sched/rt.h>
#include <linux/sched/debug.h>
#include <linux/delay.h>
#include <linux/export.h>
#include <linux/spinlock.h>
#include <linux/kallsyms.h>
#include <linux/syscalls.h>
#include <linux/interrupt.h>
#include <linux/rbtree.h>
#include <linux/fs.h>
#include <linux/debug_locks.h>

#include "rtmutex_common.h"

static void printk_task(struct task_struct *p)
{
	if (p)
		printk("%16s:%5d [%p, %3d]", p->comm, task_pid_nr(p), p, p->prio);
	else
		printk("<none>");
}

static void printk_lock(struct rt_mutex *lock, int print_owner)
{
	if (lock->name)
		printk(" [%p] {%s}\n",
			lock, lock->name);
	else
		printk(" [%p] {%s:%d}\n",
			lock, lock->file, lock->line);

	if (print_owner && rt_mutex_owner(lock)) {
		printk(".. ->owner: %p\n", lock->owner);
		printk(".. held by:  ");
		printk_task(rt_mutex_owner(lock));
		printk("\n");
	}
}

void rt_mutex_debug_task_free(struct task_struct *task)
{
	DEBUG_LOCKS_WARN_ON(!RB_EMPTY_ROOT(&task->pi_waiters));
	DEBUG_LOCKS_WARN_ON(task->pi_blocked_on);
}

/*
 * We fill out the fields in the waiter to store the information about
 * the deadlock. We print when we return. act_waiter can be NULL in
 * case of a remove waiter operation.
 */
void debug_rt_mutex_deadlock(enum rtmutex_chainwalk chwalk,
			     struct rt_mutex_waiter *act_waiter,
			     struct rt_mutex *lock)
{
	struct task_struct *task;

	if (!debug_locks || chwalk == RT_MUTEX_FULL_CHAINWALK || !act_waiter)
		return;

	task = rt_mutex_owner(act_waiter->lock);
	if (task && task != current) {
		act_waiter->deadlock_task_pid = get_pid(task_pid(task));
		act_waiter->deadlock_lock = lock;
	}
}

void debug_rt_mutex_print_deadlock(struct rt_mutex_waiter *waiter)
{
	struct task_struct *task;

	if (!waiter->deadlock_lock || !debug_locks)
		return;

	rcu_read_lock();
	task = pid_task(waiter->deadlock_task_pid, PIDTYPE_PID);
	if (!task) {
		rcu_read_unlock();
		return;
	}

	if (!debug_locks_off()) {
		rcu_read_unlock();
		return;
	}

	printk("\n============================================\n");
	printk(  "[ BUG: circular locking deadlock detected! ]\n");
	printk("%s\n", print_tainted());
	printk(  "--------------------------------------------\n");
	printk("%s/%d is deadlocking current task %s/%d\n\n",
	       task->comm, task_pid_nr(task),
	       current->comm, task_pid_nr(current));

	printk("\n1) %s/%d is trying to acquire this lock:\n",
	       current->comm, task_pid_nr(current));
	printk_lock(waiter->lock, 1);

	printk("\n2) %s/%d is blocked on this lock:\n",
		task->comm, task_pid_nr(task));
	printk_lock(waiter->deadlock_lock, 1);

	debug_show_held_locks(current);
	debug_show_held_locks(task);

	printk("\n%s/%d's [blocked] stackdump:\n\n",
		task->comm, task_pid_nr(task));
	show_stack(task, NULL);
	printk("\n%s/%d's [current] stackdump:\n\n",
		current->comm, task_pid_nr(current));
	dump_stack();
	debug_show_all_locks();
	rcu_read_unlock();

	printk("[ turning off deadlock detection."
	       "Please report this trace. ]\n\n");
}

void debug_rt_mutex_lock(struct rt_mutex *lock)
{
}

void debug_rt_mutex_unlock(struct rt_mutex *lock)
{
	DEBUG_LOCKS_WARN_ON(rt_mutex_owner(lock) != current);
}

void
debug_rt_mutex_proxy_lock(struct rt_mutex *lock, struct task_struct *powner)
{
}

void debug_rt_mutex_proxy_unlock(struct rt_mutex *lock)
{
	DEBUG_LOCKS_WARN_ON(!rt_mutex_owner(lock));
}

void debug_rt_mutex_init_waiter(struct rt_mutex_waiter *waiter)
{
	memset(waiter, 0x11, sizeof(*waiter));
	waiter->deadlock_task_pid = NULL;
}

void debug_rt_mutex_free_waiter(struct rt_mutex_waiter *waiter)
{
	put_pid(waiter->deadlock_task_pid);
	memset(waiter, 0x22, sizeof(*waiter));
}

void debug_rt_mutex_init(struct rt_mutex *lock, const char *name)
{
	/*
	 * Make sure we are not reinitializing a held lock:
	 */
	debug_check_no_locks_freed((void *)lock, sizeof(*lock));
	lock->name = name;
}

void
rt_mutex_deadlock_account_lock(struct rt_mutex *lock, struct task_struct *task)
{
}

void rt_mutex_deadlock_account_unlock(struct task_struct *task)
{
}

back to top