Re: [PATCH -next] mm/hotplug: skip bad PFNs from pfn_to_online_page()

[Qian Cai <cai@lca.pw>]

> On Jun 16, 2019, at 11:42 AM, Dan Williams <dan.j.williams@intel.com> wrote:
> 
> On Fri, Jun 14, 2019 at 1:43 PM Qian Cai <cai@lca.pw> wrote:
>> 
>> On Fri, 2019-06-14 at 12:48 -0700, Dan Williams wrote:
>>> On Fri, Jun 14, 2019 at 12:40 PM Qian Cai <cai@lca.pw> wrote:
>>>> 
>>>> On Fri, 2019-06-14 at 11:57 -0700, Dan Williams wrote:
>>>>> On Fri, Jun 14, 2019 at 11:03 AM Dan Williams <dan.j.williams@intel.com>
>>>>> wrote:
>>>>>> 
>>>>>> On Fri, Jun 14, 2019 at 7:59 AM Qian Cai <cai@lca.pw> wrote:
>>>>>>> 
>>>>>>> On Fri, 2019-06-14 at 14:28 +0530, Aneesh Kumar K.V wrote:
>>>>>>>> Qian Cai <cai@lca.pw> writes:
>>>>>>>> 
>>>>>>>> 
>>>>>>>>> 1) offline is busted [1]. It looks like test_pages_in_a_zone()
>>>>>>>>> missed
>>>>>>>>> the
>>>>>>>>> same
>>>>>>>>> pfn_section_valid() check.
>>>>>>>>> 
>>>>>>>>> 2) powerpc booting is generating endless warnings [2]. In
>>>>>>>>> vmemmap_populated() at
>>>>>>>>> arch/powerpc/mm/init_64.c, I tried to change PAGES_PER_SECTION to
>>>>>>>>> PAGES_PER_SUBSECTION, but it alone seems not enough.
>>>>>>>>> 
>>>>>>>> 
>>>>>>>> Can you check with this change on ppc64.  I haven't reviewed this
>>>>>>>> series
>>>>>>>> yet.
>>>>>>>> I did limited testing with change . Before merging this I need to go
>>>>>>>> through the full series again. The vmemmap poplulate on ppc64 needs
>>>>>>>> to
>>>>>>>> handle two translation mode (hash and radix). With respect to vmemap
>>>>>>>> hash doesn't setup a translation in the linux page table. Hence we
>>>>>>>> need
>>>>>>>> to make sure we don't try to setup a mapping for a range which is
>>>>>>>> arleady convered by an existing mapping.
>>>>>>> 
>>>>>>> It works fine.
>>>>>> 
>>>>>> Strange... it would only change behavior if valid_section() is true
>>>>>> when pfn_valid() is not or vice versa. They "should" be identical
>>>>>> because subsection-size == section-size on PowerPC, at least with the
>>>>>> current definition of SUBSECTION_SHIFT. I suspect maybe
>>>>>> free_area_init_nodes() is too late to call subsection_map_init() for
>>>>>> PowerPC.
>>>>> 
>>>>> Can you give the attached incremental patch a try? This will break
>>>>> support for doing sub-section hot-add in a section that was only
>>>>> partially populated early at init, but that can be repaired later in
>>>>> the series. First things first, don't regress.
>>>>> 
>>>>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>>>>> index 874eb22d22e4..520c83aa0fec 100644
>>>>> --- a/mm/page_alloc.c
>>>>> +++ b/mm/page_alloc.c
>>>>> @@ -7286,12 +7286,10 @@ void __init free_area_init_nodes(unsigned long
>>>>> *max_zone_pfn)
>>>>> 
>>>>>        /* Print out the early node map */
>>>>>        pr_info("Early memory node ranges\n");
>>>>> -       for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn,
>>>>> &nid) {
>>>>> +       for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn,
>>>>> &nid)
>>>>>                pr_info("  node %3d: [mem %#018Lx-%#018Lx]\n", nid,
>>>>>                        (u64)start_pfn << PAGE_SHIFT,
>>>>>                        ((u64)end_pfn << PAGE_SHIFT) - 1);
>>>>> -               subsection_map_init(start_pfn, end_pfn - start_pfn);
>>>>> -       }
>>>>> 
>>>>>        /* Initialise every node */
>>>>>        mminit_verify_pageflags_layout();
>>>>> diff --git a/mm/sparse.c b/mm/sparse.c
>>>>> index 0baa2e55cfdd..bca8e6fa72d2 100644
>>>>> --- a/mm/sparse.c
>>>>> +++ b/mm/sparse.c
>>>>> @@ -533,6 +533,7 @@ static void __init sparse_init_nid(int nid,
>>>>> unsigned long pnum_begin,
>>>>>                }
>>>>>                check_usemap_section_nr(nid, usage);
>>>>>                sparse_init_one_section(__nr_to_section(pnum), pnum,
>>>>> map, usage);
>>>>> +               subsection_map_init(section_nr_to_pfn(pnum),
>>>>> PAGES_PER_SECTION);
>>>>>                usage = (void *) usage + mem_section_usage_size();
>>>>>        }
>>>>>        sparse_buffer_fini();
>>>> 
>>>> It works fine except it starts to trigger slab debugging errors during boot.
>>>> Not
>>>> sure if it is related yet.
>>> 
>>> If you want you can give this branch a try if you suspect something
>>> else in -next is triggering the slab warning.
>>> 
>>> https://git.kernel.org/pub/scm/linux/kernel/git/djbw/nvdimm.git/log/?h=subsect
>>> ion-v9
>>> 
>>> It's the original v9 patchset + dependencies backported to v5.2-rc4.
>>> 
>>> I otherwise don't see how subsections would effect slab caches.
>> 
>> It works fine there.
> 
> Much appreciated Qian!
> 
> Does this change modulate the x86 failures?

Yes, it also fix the kmemleak_scan() and offline issues on x86.