[PATCH] mm/migrate: Fix ref-count handling when !hugepage_migration_supported()

[Naoya Horiguchi]

On Wed, May 24, 2017 at 04:47:28PM +0100, Punit Agrawal wrote:
> On failing to migrate a page, soft_offline_huge_page() performs the
> necessary update to the hugepage ref-count. When
> !hugepage_migration_supported() , unmap_and_move_hugepage() also
> decrements the page ref-count for the hugepage. The combined behaviour
> leaves the ref-count in an inconsistent state.
> 
> This leads to soft lockups when running the overcommitted hugepage test
> from mce-tests suite.
> 
> Soft offlining pfn 0x83ed600 at process virtual address 0x400000000000
> soft offline: 0x83ed600: migration failed 1, type
> 1fffc00000008008 (uptodate|head)
> INFO: rcu_preempt detected stalls on CPUs/tasks:
>  Tasks blocked on level-0 rcu_node (CPUs 0-7): P2715
>   (detected by 7, t=5254 jiffies, g=963, c=962, q=321)
>   thugetlb_overco R  running task        0  2715   2685 0x00000008
>   Call trace:
>   [<ffff000008089f90>] dump_backtrace+0x0/0x268
>   [<ffff00000808a2d4>] show_stack+0x24/0x30
>   [<ffff000008100d34>] sched_show_task+0x134/0x180
>   [<ffff0000081c90fc>] rcu_print_detail_task_stall_rnp+0x54/0x7c
>   [<ffff00000813cfd4>] rcu_check_callbacks+0xa74/0xb08
>   [<ffff000008143a3c>] update_process_times+0x34/0x60
>   [<ffff0000081550e8>] tick_sched_handle.isra.7+0x38/0x70
>   [<ffff00000815516c>] tick_sched_timer+0x4c/0x98
>   [<ffff0000081442e0>] __hrtimer_run_queues+0xc0/0x300
>   [<ffff000008144fa4>] hrtimer_interrupt+0xac/0x228
>   [<ffff0000089a56d4>] arch_timer_handler_phys+0x3c/0x50
>   [<ffff00000812f1bc>] handle_percpu_devid_irq+0x8c/0x290
>   [<ffff0000081297fc>] generic_handle_irq+0x34/0x50
>   [<ffff000008129f00>] __handle_domain_irq+0x68/0xc0
>   [<ffff0000080816b4>] gic_handle_irq+0x5c/0xb0
> 
> Fix this by dropping the ref-count decrement in
> unmap_and_move_hugepage() when !hugepage_migration_supported().
> 
> Fixes: 32665f2bbfed ("mm/migrate: correct failure handling if !hugepage_migration_support()")
> Reported-by: Manoj Iyer <manoj.iyer at canonical.com>
> Signed-off-by: Punit Agrawal <punit.agrawal at arm.com>
> Cc: Joonsoo Kim <iamjoonsoo.kim at lge.com>
> Cc: Naoya Horiguchi <n-horiguchi at ah.jp.nec.com>
> Cc: Wanpeng Li <liwanp at linux.vnet.ibm.com>
> Cc: Christoph Lameter <cl at linux.com>
> 
> --
> Hi Andrew,
> 
> We ran into this bug when working towards enabling memory corruption
> on arm64. The patch was tested on an arm64 platform running v4.12-rc2
> with the series to enable memory corruption handling[0].
> 
> Please consider merging as a fix for the 4.12 release.
> 
> Thanks,
> Punit
> 
> [0] https://www.spinics.net/lists/arm-kernel/msg581657.html
> ---
>  mm/migrate.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 89a0a1707f4c..187abd1526df 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1201,10 +1201,8 @@ static int unmap_and_move_huge_page(new_page_t get_new_page,
>  	 * tables or check whether the hugepage is pmd-based or not before
>  	 * kicking migration.
>  	 */
> -	if (!hugepage_migration_supported(page_hstate(hpage))) {
> -		putback_active_hugepage(hpage);

Thank you for reporting and suggestion, Punit, Manoj.

Simply dropping this putback_active_hugepage() may resume the failure
counting issue addressed in 32665f2bbfed, so I would recommend to call
putback_movable_pages() in failure path in soft_offline_huge_page().

@@ -1600,7 +1600,8 @@ static int soft_offline_huge_page(struct page *page, int flags)
 		 * only one hugepage pointed to by hpage, so we need not
 		 * run through the pagelist here.
 		 */
-		putback_active_hugepage(hpage);
+		if (!list_empty(&pagelist))
+			putback_movable_pages(&pagelist);
 		if (ret > 0)
 			ret = -EIO;
 	} else {

Could you check this works for you?

Thanks,
Naoya Horiguchi