Re: BUG: sleeping function called from invalid context in __alloc_pages_nodemask

From: Tetsuo Handa
Date: Sat Nov 30 2019 - 06:51:14 EST


On 2019/11/30 17:32, Hillf Danton wrote:
>
> On Fri, 29 Nov 2019 23:35:08 -0800
>> Hello,
>>
>> syzbot found the following crash on:
>>
>> HEAD commit: 419593da Add linux-next specific files for 20191129
>> git tree: linux-next
>> console output: https://syzkaller.appspot.com/x/log.txt?x=12cc369ce00000
>> kernel config: https://syzkaller.appspot.com/x/.config?x=7c04b0959e75c206
>> dashboard link: https://syzkaller.appspot.com/bug?extid=4925d60532bf4c399608
>> compiler: gcc (GCC) 9.0.0 20181231 (experimental)
>>
>> Unfortunately, I don't have any reproducer for this crash yet.
>>
>> IMPORTANT: if you fix the bug, please add the following tag to the commit:
>> Reported-by: syzbot+4925d60532bf4c399608@xxxxxxxxxxxxxxxxxxxxxxxxx
>>
>> BUG: sleeping function called from invalid context at mm/page_alloc.c:4681
>> in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 2710, name:
>> kworker/0:2
>> 4 locks held by kworker/0:2/2710:
>> #0: ffff8880aa026d28 ((wq_completion)events){+.+.}, at: __write_once_size
>> include/linux/compiler.h:247 [inline]
>> #0: ffff8880aa026d28 ((wq_completion)events){+.+.}, at: arch_atomic64_set
>> arch/x86/include/asm/atomic64_64.h:34 [inline]
>> #0: ffff8880aa026d28 ((wq_completion)events){+.+.}, at: atomic64_set
>> include/asm-generic/atomic-instrumented.h:868 [inline]
>> #0: ffff8880aa026d28 ((wq_completion)events){+.+.}, at: atomic_long_set
>> include/asm-generic/atomic-long.h:40 [inline]
>> #0: ffff8880aa026d28 ((wq_completion)events){+.+.}, at: set_work_data
>> kernel/workqueue.c:615 [inline]
>> #0: ffff8880aa026d28 ((wq_completion)events){+.+.}, at:
>> set_work_pool_and_clear_pending kernel/workqueue.c:642 [inline]
>> #0: ffff8880aa026d28 ((wq_completion)events){+.+.}, at:
>> process_one_work+0x88b/0x1740 kernel/workqueue.c:2235
>> #1: ffffc9000802fdc0 (pcpu_balance_work){+.+.}, at:
>> process_one_work+0x8c1/0x1740 kernel/workqueue.c:2239
>> #2: ffffffff8983ff20 (pcpu_alloc_mutex){+.+.}, at:
>> pcpu_balance_workfn+0xb7/0x1310 mm/percpu.c:1845
>> #3: ffffffff89851b18 (vmap_area_lock){+.+.}, at: spin_lock
>> include/linux/spinlock.h:338 [inline]
>> #3: ffffffff89851b18 (vmap_area_lock){+.+.}, at:
>> pcpu_get_vm_areas+0x3b27/0x3f00 mm/vmalloc.c:3431
>> Preemption disabled at:
>> [<ffffffff81a89ce7>] spin_lock include/linux/spinlock.h:338 [inline]
>> [<ffffffff81a89ce7>] pcpu_get_vm_areas+0x3b27/0x3f00 mm/vmalloc.c:3431
>> CPU: 0 PID: 2710 Comm: kworker/0:2 Not tainted
>> 5.4.0-next-20191129-syzkaller #0
>> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS
>> Google 01/01/2011
>> Workqueue: events pcpu_balance_workfn
>> Call Trace:
>> __dump_stack lib/dump_stack.c:77 [inline]
>> dump_stack+0x197/0x210 lib/dump_stack.c:118
>> ___might_sleep.cold+0x1fb/0x23e kernel/sched/core.c:6800
>> __might_sleep+0x95/0x190 kernel/sched/core.c:6753
>> prepare_alloc_pages mm/page_alloc.c:4681 [inline]
>> __alloc_pages_nodemask+0x523/0x910 mm/page_alloc.c:4730
>> alloc_pages_current+0x107/0x210 mm/mempolicy.c:2211
>> alloc_pages include/linux/gfp.h:532 [inline]
>> __get_free_pages+0xc/0x40 mm/page_alloc.c:4786
>> kasan_populate_vmalloc_pte mm/kasan/common.c:762 [inline]
>> kasan_populate_vmalloc_pte+0x2f/0x1c0 mm/kasan/common.c:753
>> apply_to_pte_range mm/memory.c:2041 [inline]
>> apply_to_pmd_range mm/memory.c:2068 [inline]
>> apply_to_pud_range mm/memory.c:2088 [inline]
>> apply_to_p4d_range mm/memory.c:2108 [inline]
>> apply_to_page_range+0x445/0x700 mm/memory.c:2133
>> kasan_populate_vmalloc+0x68/0x90 mm/kasan/common.c:791
>> pcpu_get_vm_areas+0x3c77/0x3f00 mm/vmalloc.c:3439
>> pcpu_create_chunk+0x24e/0x7f0 mm/percpu-vm.c:340
>> pcpu_balance_workfn+0xf1b/0x1310 mm/percpu.c:1934
>> process_one_work+0x9af/0x1740 kernel/workqueue.c:2264
>> worker_thread+0x98/0xe40 kernel/workqueue.c:2410
>> kthread+0x361/0x430 kernel/kthread.c:255
>> ret_from_fork+0x24/0x30 arch/x86/entry/entry_64.S:352
>
> Replace the blocking gfp mask with a non-blocking one to survive
> checks like might_sleep.
>
> --- a/mm/kasan/common.c
> +++ b/mm/kasan/common.c
> @@ -759,7 +759,7 @@ static int kasan_populate_vmalloc_pte(pt
> if (likely(!pte_none(*ptep)))
> return 0;
>
> - page = __get_free_page(GFP_KERNEL);
> + page = __get_free_page(GFP_NOWAIT);
> if (!page)
> return -ENOMEM;
>

Nope. This change would survive might_sleep() check, but the caller is
expecting that this is __GFP_NOFAIL allocation. Even if the caller can
tolerate allocation failures, __GFP_NOWARN should be added in order to
avoid flooding of allocation failure messages.

/* insert all vm's */
spin_lock(&vmap_area_lock);
for (area = 0; area < nr_vms; area++) {
insert_vmap_area(vas[area], &vmap_area_root, &vmap_area_list);

setup_vmalloc_vm_locked(vms[area], vas[area], VM_ALLOC,
pcpu_get_vm_areas);

/* assume success here */
kasan_populate_vmalloc(sizes[area], vms[area]);
}
spin_unlock(&vmap_area_lock);