On 21.10.20 15:34, David Hildenbrand wrote:
> On 21.10.20 15:11, David Hildenbrand wrote:
>> On 21.10.20 14:57, Michal Privoznik wrote:
>>> On 10/21/20 5:35 AM, Mike Kravetz wrote:
>>>> On 10/20/20 6:38 AM, David Hildenbrand wrote:
>>>>>
>>>>> I'm bisecting the warning right now. Looks like it was introduced in v5.7.
>>>>
>>>> I found the following bugs in the cgroup reservation accounting. The ones
>>>> in region_del are pretty obvious as the number of pages to uncharge would
>>>> always be zero. The one on alloc_huge_page needs racing code to expose.
>>>>
>>>> With these fixes, my testing is showing consistent/correct results for
>>>> hugetlb reservation cgroup accounting.
>>>>
>>>> It would be good if Mina (at least) would look these over. Would also
>>>> be interesting to know if these fixes address the bug seen with the qemu
>>>> use case.
>>>>
>>>> I'm still doing more testing and code inspection to look for other issues.
>>>>
>>>> From 861bcd7d0443f18a5fed3c3ddc5f1c71e78c4ef4 Mon Sep 17 00:00:00 2001
>>>> From: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
>>>> Date: Tue, 20 Oct 2020 20:21:42 -0700
>>>> Subject: [PATCH] hugetlb_cgroup: fix reservation accounting
>>>>
>>>> Signed-off-by: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
>>>> ---
>>>> mm/hugetlb.c | 15 +++++++++------
>>>> 1 file changed, 9 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>>> index 67fc6383995b..c92366313780 100644
>>>> --- a/mm/hugetlb.c
>>>> +++ b/mm/hugetlb.c
>>>> @@ -685,17 +685,17 @@ static long region_del(struct resv_map *resv, long f, long t)
>>>> }
>>>>
>>>> if (f <= rg->from) { /* Trim beginning of region */
>>>> - del += t - rg->from;
>>>> - rg->from = t;
>>>> -
>>>> hugetlb_cgroup_uncharge_file_region(resv, rg,
>>>> t - rg->from);
>>>> - } else { /* Trim end of region */
>>>> - del += rg->to - f;
>>>> - rg->to = f;
>>>>
>>>> + del += t - rg->from;
>>>> + rg->from = t;
>>>> + } else { /* Trim end of region */
>>>> hugetlb_cgroup_uncharge_file_region(resv, rg,
>>>> rg->to - f);
>>>> +
>>>> + del += rg->to - f;
>>>> + rg->to = f;
>>>> }
>>>> }
>>>>
>>>> @@ -2454,6 +2454,9 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>>>>
>>>> rsv_adjust = hugepage_subpool_put_pages(spool, 1);
>>>> hugetlb_acct_memory(h, -rsv_adjust);
>>>> + if (deferred_reserve)
>>>> + hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
>>>> + pages_per_huge_page(h), page);
>>>> }
>>>> return page;
>>>>
>>>>
>>>
>>> I've applied, rebuilt and tested, but unfortunately I still hit the problem:
>>> [ 6472.719047] ------------[ cut here ]------------
>>> [ 6472.719052] WARNING: CPU: 6 PID: 11773 at mm/page_counter.c:57
>>> page_counter_uncharge+0x33/0x40
>>> [ 6472.719052] Modules linked in: kvm_amd amdgpu kvm btusb sp5100_tco
>>> btrtl watchdog k10temp btbcm btintel mfd_core gpu_sched ttm
>>> [ 6472.719057] CPU: 6 PID: 11773 Comm: CPU 3/KVM Not tainted
>>> 5.9.1-gentoo-x86_64 #1
>>> [ 6472.719057] Hardware name: System manufacturer System Product
>>> Name/PRIME X570-PRO, BIOS 1005 08/01/2019
>>> [ 6472.719059] RIP: 0010:page_counter_uncharge+0x33/0x40
>>> [ 6472.719060] Code: 48 85 ff 74 24 4c 89 c8 f0 48 0f c1 07 4c 29 c0 48
>>> 89 c1 48 89 c6 e8 7c fe ff ff 48 85 c9 78 0a 48 8b 7f 28 48 85 ff 75 dc
>>> c3 <0f> 0b eb f2 66 0f 1f 84 00 00 00 00 00 48 8b 17 48 39 d6 72 41 41
>>> [ 6472.719061] RSP: 0018:ffffc90000b77b40 EFLAGS: 00010286
>>> [ 6472.719061] RAX: fffffffffffe9200 RBX: ffff888fb3b97b40 RCX:
>>> fffffffffffe9200
>>> [ 6472.719062] RDX: 0000000000000221 RSI: fffffffffffe9200 RDI:
>>> ffff888fd8451dd0
>>> [ 6472.719062] RBP: ffff888fb6990420 R08: 0000000000044200 R09:
>>> fffffffffffbbe00
>>> [ 6472.719062] R10: ffff888fb3b97b40 R11: 000000000000000a R12:
>>> 0000000000000001
>>> [ 6472.719063] R13: 00000000000005df R14: 00000000000005de R15:
>>> ffff888fb3b97b40
>>> [ 6472.719063] FS: 00007fbd175fe700(0000) GS:ffff888fde980000(0000)
>>> knlGS:0000000000000000
>>> [ 6472.719064] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>> [ 6472.719064] CR2: 00007fbd825101f0 CR3: 0000000fb5e41000 CR4:
>>> 0000000000350ee0
>>> [ 6472.719065] Call Trace:
>>> [ 6472.719067] hugetlb_cgroup_uncharge_file_region+0x46/0x70
>>> [ 6472.719069] region_del+0x1ae/0x270
>>> [ 6472.719070] hugetlb_unreserve_pages+0x32/0xa0
>>> [ 6472.719072] remove_inode_hugepages+0x19d/0x3a0
>>> [ 6472.719079] ? writeback_registers+0x45/0x60 [kvm]
>>> [ 6472.719080] hugetlbfs_fallocate+0x3f2/0x4a0
>>> [ 6472.719081] ? __mod_lruvec_state+0x1d/0x40
>>> [ 6472.719081] ? __mod_memcg_lruvec_state+0x1b/0xe0
>>> [ 6472.719083] ? __seccomp_filter+0x75/0x6a0
>>> [ 6472.719084] vfs_fallocate+0x122/0x260
>>> [ 6472.719085] __x64_sys_fallocate+0x39/0x60
>>> [ 6472.719086] do_syscall_64+0x2d/0x40
>>> [ 6472.719088] entry_SYSCALL_64_after_hwframe+0x44/0xa9
>>> [ 6472.719089] RIP: 0033:0x7fbe3cefcde7
>>> [ 6472.719089] Code: 89 7c 24 08 48 89 4c 24 18 e8 45 fc f8 ff 41 89 c0
>>> 4c 8b 54 24 18 48 8b 54 24 10 b8 1d 01 00 00 8b 74 24 0c 8b 7c 24 08 0f
>>> 05 <48> 3d 00 f0 ff ff 77 41 44 89 c7 89 44 24 08 e8 75 fc f8 ff 8b 44
>>> [ 6472.719090] RSP: 002b:00007fbd175fc7a0 EFLAGS: 00000293 ORIG_RAX:
>>> 000000000000011d
>>> [ 6472.719090] RAX: ffffffffffffffda RBX: 00000000bbe00000 RCX:
>>> 00007fbe3cefcde7
>>> [ 6472.719091] RDX: 00000000bbc00000 RSI: 0000000000000003 RDI:
>>> 000000000000001d
>>> [ 6472.719091] RBP: 00007fbd175fc800 R08: 0000000000000000 R09:
>>> 0000000000000000
>>> [ 6472.719091] R10: 0000000000200000 R11: 0000000000000293 R12:
>>> 00007ffeea066d2e
>>> [ 6472.719092] R13: 00007ffeea066d2f R14: 00007fbd175fe700 R15:
>>> 00007fbd175fcdc0
>>> [ 6472.719092] ---[ end trace c97dc6281a861980 ]---
>>
>> Agreed, same over here. :(
>>
>
> I *think* the following on top makes it fly
>
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 67fc6383995b..5cf7f6a6c1a6 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -656,6 +656,9 @@ static long region_del(struct resv_map *resv, long
> f, long t)
>
> del += t - f;
>
> + hugetlb_cgroup_uncharge_file_region(
> + resv, rg, t - f);
> +
> /* New entry for end of split region */
> nrg->from = t;
> nrg->to = rg->to;
> @@ -667,9 +670,6 @@ static long region_del(struct resv_map *resv, long
> f, long t)
> /* Original entry is trimmed */
> rg->to = f;
>
> - hugetlb_cgroup_uncharge_file_region(
> - resv, rg, nrg->to - nrg->from);
> -
> list_add(&nrg->link, &rg->link);
> nrg = NULL;
> break;
>
>
(sorry for the nasty line wrapping - Thunderbird update keeps breaking
helpful plugins - like Toggle Word Wrap in this case ... argh)
--
Thanks,
David / dhildenb
