lkml.org 
[lkml]   [2022]   [Oct]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v7 1/4] mm,hwpoison,hugetlb,memory_hotplug: hotremove memory section with hwpoisoned hugepage
From
Date
On 2022/10/24 14:20, Naoya Horiguchi wrote:
> From: Naoya Horiguchi <naoya.horiguchi@nec.com>
>
> HWPoisoned page is not supposed to be accessed once marked, but currently
> such accesses can happen during memory hotremove because do_migrate_range()
> can be called before dissolve_free_huge_pages() is called.
>
> Clear HPageMigratable for hwpoisoned hugepages to prevent them from being
> migrated. This should be done in hugetlb_lock to avoid race against
> isolate_hugetlb().
>
> get_hwpoison_huge_page() needs to have a flag to show it's called from
> unpoison to take refcount of hwpoisoned hugepages, so add it.
>
> Reported-by: Miaohe Lin <linmiaohe@huawei.com>
> Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
> Reviewed-by: Oscar Salvador <osalvador@suse.de>
> Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
> ---
> ChangeLog v3 -> v7:
> - introduce TESTCLEARHPAGEFLAG() to determine the value of migratable_cleared

Many thanks for update, Naoya. I'm sorry but TestClearHPageMigratable() might be somewhat
overkill. As we discussed in previous thread:

"""
I think I might be nitpicking... But it seems ClearHPageMigratable is not enough here.
1. In MF_COUNT_INCREASED case, we don't know whether HPageMigratable is set.
2. Even if HPageMigratable is set, there might be a race window before we clear HPageMigratable?
So "*migratable_cleared = TestClearHPageMigratable" might be better? But I might be wrong.
"""

The case 2 should be a dumb problem(sorry about it). HPageMigratable() is always cleared while holding
the hugetlb_lock which is already held by get_huge_page_for_hwpoison(). So the only case we should care
about is case 1 and that can be handled by below more efficient pattern:
if (HPageMigratable)
ClearHPageMigratable()

So the overhead of test and clear atomic ops can be avoided. But this is trival.

Anyway, this patch still looks good to me. And my Reviewed-by tag still applies. Many thanks.

Thanks,
Miaohe Lin

\
 
 \ /
  Last update: 2022-10-25 04:38    [W:0.103 / U:0.400 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site