From: "Yin, Fengwei" <fengwei.yin@intel.com>
To: Ryan Roberts <ryan.roberts@arm.com>,
"Matthew Wilcox (Oracle)" <willy@infradead.org>,
<linux-arch@vger.kernel.org>
Cc: <linux-mm@kvack.org>, <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v4 34/36] rmap: add folio_add_file_rmap_range()
Date: Fri, 17 Mar 2023 21:28:57 +0800 [thread overview]
Message-ID: <c7c7f5ed-ef80-ef6f-4a73-806b21e7c65d@intel.com> (raw)
In-Reply-To: <2b41a6dd-130b-acf3-72fd-b996272c1710@arm.com>
On 3/17/2023 8:46 PM, Ryan Roberts wrote:
> On 17/03/2023 08:23, Yin, Fengwei wrote:
> [...]
>
>>>>>> FYI, I'm seeing a perf regression of about 1% when compiling the kernel on
>>>>>> Ampere Altra (arm64) with this whole series on top of v6.3-rc1 (In a VM using
>>>>>> ext4 filesystem). Looks like instruction aborts are taking much longer and a
>>>>>> selection of syscalls are a bit slower. Still hunting down the root cause. Will
>>>>>> report once I have conclusive diagnosis.
>>>>>
>>>>> I'm sorry - I'm struggling to find the exact cause. But its spending over 2x the
>>>>> amount of time in the instruction abort handling code once patches 32-36 are
>>>>> included. Everything in the flame graph is just taking longer. Perhaps we are
>>>>> getting more instruction aborts somehow? I have the flamegraphs if anyone wants
>>>>> them - just shout and I'll email them separately.
>>>> Thanks a lot to Ryan for sharing the flamegraphs to me. I found the __do_fault()
>>>> is called with patch 32-36 while no __do_fault() just with first 31 patches. I
>>>> suspect the folio_add_file_rmap_range() missed some PTEs population. Please give
>>>> me few days to find the root cause and fix. Sorry for this.
>>>
>>> You're welcome. Give me a shout once you have a re-spin and I'll rerun the tests.
>> Could you please help to try following changes? Thanks in advance.
>>
>> diff --git a/mm/filemap.c b/mm/filemap.c
>> index 40be33b5ee46..137011320c68 100644
>> --- a/mm/filemap.c
>> +++ b/mm/filemap.c
>> @@ -3504,15 +3504,16 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
>> if (!pte_none(vmf->pte[count]))
>> goto skip;
>>
>> - if (vmf->address == addr)
>> - ret = VM_FAULT_NOPAGE;
>> -
>> count++;
>> continue;
>> skip:
>> if (count) {
>> set_pte_range(vmf, folio, page, count, addr);
>> folio_ref_add(folio, count);
>> + if ((vmf->address < (addr + count * PAGE_SIZE)) &&
>> + (vmf->address >= addr))
>> + ret = VM_FAULT_NOPAGE;
>> +
>> }
>>
>> count++;
>> @@ -3525,6 +3526,9 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
>> if (count) {
>> set_pte_range(vmf, folio, page, count, addr);
>> folio_ref_add(folio, count);
>> + if ((vmf->address < (addr + count * PAGE_SIZE)) &&
>> + (vmf->address >= addr))
>> + ret = VM_FAULT_NOPAGE;
>> }
>>
>> vmf->pte = old_ptep;
>>
>
> I'm afraid this hasn't fixed it, and I still see __do_fault(). I'll send the
> flame graph over separately.
>
> Given I'm running on ext4, I wasn't expecting to see any large page cache
> folios? So I don't think we would have expected this patch to help anyway? (or
> perhaps there are still THP folios? But I think they will get PMD mapped?).
OK. I will try to reproduce the issue on my local env to see whether I could
reproduce it on x86_64 env.
Regards
Yin, Fengwei
>
>
>>
>> Regards
>> Yin, Fengwei
>>
>>>
>>>>
>>>>
>>>> Regards
>>>> Yin, Fengwei
>>>>
>>>>>
>>>>>>
>>>>>> Thanks,
>>>>>> Ryan
>>>>>>
>>>>>>
>>>>>>> + }
>>>>>>> +
>>>>>>> + if (first)
>>>>>>> + nr++;
>>>>>>> + } while (page++, --nr_pages > 0);
>>>>>>> } else if (folio_test_pmd_mappable(folio)) {
>>>>>>> /* That test is redundant: it's for safety or to optimize out */
>>>>>>>
>>>>>>> @@ -1354,6 +1362,30 @@ void page_add_file_rmap(struct page *page, struct vm_area_struct *vma,
>>>>>>> mlock_vma_folio(folio, vma, compound);
>>>>>>> }
>>>>>>>
>>>>>>> +/**
>>>>>>> + * page_add_file_rmap - add pte mapping to a file page
>>>>>>> + * @page: the page to add the mapping to
>>>>>>> + * @vma: the vm area in which the mapping is added
>>>>>>> + * @compound: charge the page as compound or small page
>>>>>>> + *
>>>>>>> + * The caller needs to hold the pte lock.
>>>>>>> + */
>>>>>>> +void page_add_file_rmap(struct page *page, struct vm_area_struct *vma,
>>>>>>> + bool compound)
>>>>>>> +{
>>>>>>> + struct folio *folio = page_folio(page);
>>>>>>> + unsigned int nr_pages;
>>>>>>> +
>>>>>>> + VM_WARN_ON_ONCE_PAGE(compound && !PageTransHuge(page), page);
>>>>>>> +
>>>>>>> + if (likely(!compound))
>>>>>>> + nr_pages = 1;
>>>>>>> + else
>>>>>>> + nr_pages = folio_nr_pages(folio);
>>>>>>> +
>>>>>>> + folio_add_file_rmap_range(folio, page, nr_pages, vma, compound);
>>>>>>> +}
>>>>>>> +
>>>>>>> /**
>>>>>>> * page_remove_rmap - take down pte mapping from a page
>>>>>>> * @page: page to remove mapping from
>>>>>>
>>>>>
>>>
>>>
>
next prev parent reply other threads:[~2023-03-17 13:29 UTC|newest]
Thread overview: 138+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-15 5:14 [PATCH v4 00/36] New page table range API Matthew Wilcox (Oracle)
2023-03-15 5:14 ` [PATCH v4 01/36] mm: Convert page_table_check_pte_set() to page_table_check_ptes_set() Matthew Wilcox (Oracle)
2023-03-15 9:21 ` Mike Rapoport
2023-03-23 18:36 ` Pasha Tatashin
2023-05-25 2:16 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 02/36] mm: Add generic flush_icache_pages() and documentation Matthew Wilcox (Oracle)
2023-03-15 9:27 ` Mike Rapoport
2023-05-25 2:23 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 03/36] mm: Add folio_flush_mapping() Matthew Wilcox (Oracle)
2023-03-15 9:28 ` Mike Rapoport
2023-05-25 2:35 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 04/36] mm: Remove ARCH_IMPLEMENTS_FLUSH_DCACHE_FOLIO Matthew Wilcox (Oracle)
2023-03-15 9:28 ` Mike Rapoport
2023-05-25 2:43 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 05/36] mm: Add default definition of set_ptes() Matthew Wilcox (Oracle)
2023-03-15 9:34 ` Mike Rapoport
2023-05-25 3:01 ` Anshuman Khandual
2023-05-25 4:06 ` Matthew Wilcox
2023-03-15 5:14 ` [PATCH v4 06/36] alpha: Implement the new page table range API Matthew Wilcox (Oracle)
2023-03-15 9:41 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 07/36] arc: " Matthew Wilcox (Oracle)
2023-03-15 9:44 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 08/36] arm: " Matthew Wilcox (Oracle)
2023-03-15 9:48 ` Mike Rapoport
2023-03-15 10:56 ` Russell King (Oracle)
2023-03-15 5:14 ` [PATCH v4 09/36] arm64: " Matthew Wilcox (Oracle)
2023-03-15 9:49 ` Mike Rapoport
2023-05-25 3:35 ` Anshuman Khandual
2023-05-25 4:05 ` Matthew Wilcox
2023-05-25 4:43 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 10/36] csky: " Matthew Wilcox (Oracle)
2023-03-15 9:50 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 11/36] hexagon: " Matthew Wilcox (Oracle)
2023-03-15 9:54 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 12/36] ia64: " Matthew Wilcox (Oracle)
2023-03-15 9:55 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 13/36] loongarch: " Matthew Wilcox (Oracle)
2023-03-15 10:07 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 14/36] m68k: " Matthew Wilcox (Oracle)
2023-03-15 7:43 ` Geert Uytterhoeven
2023-03-16 16:32 ` Geert Uytterhoeven
2023-03-15 10:07 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 15/36] microblaze: " Matthew Wilcox (Oracle)
2023-03-15 10:07 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 16/36] mips: " Matthew Wilcox (Oracle)
2023-03-15 10:08 ` Mike Rapoport
2023-03-15 10:50 ` Thomas Bogendoerfer
2023-03-15 20:33 ` Matthew Wilcox
2023-03-17 15:29 ` Thomas Bogendoerfer
2023-03-19 18:45 ` Thomas Bogendoerfer
2023-03-19 20:16 ` Matthew Wilcox
2023-03-21 11:30 ` Thomas Bogendoerfer
2023-03-15 5:14 ` [PATCH v4 17/36] nios2: " Matthew Wilcox (Oracle)
2023-03-15 10:08 ` Mike Rapoport
2023-06-13 22:45 ` Dinh Nguyen
2023-07-10 20:18 ` Matthew Wilcox
2023-07-10 23:10 ` Dinh Nguyen
2023-03-15 5:14 ` [PATCH v4 18/36] openrisc: " Matthew Wilcox (Oracle)
2023-03-15 10:09 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 19/36] parisc: " Matthew Wilcox (Oracle)
2023-03-15 10:09 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 20/36] powerpc: " Matthew Wilcox (Oracle)
2023-03-15 9:43 ` Christophe Leroy
2023-03-15 10:18 ` Christophe Leroy
2023-03-17 3:47 ` Matthew Wilcox
2023-03-18 9:19 ` Christophe Leroy
2023-07-10 20:24 ` Matthew Wilcox
2023-07-11 4:40 ` Christophe Leroy
2023-03-15 10:09 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 21/36] riscv: " Matthew Wilcox (Oracle)
2023-03-15 10:10 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 22/36] s390: " Matthew Wilcox (Oracle)
2023-03-15 10:10 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 23/36] superh: " Matthew Wilcox (Oracle)
2023-03-15 7:22 ` John Paul Adrian Glaubitz
2023-03-15 7:36 ` John Paul Adrian Glaubitz
2023-03-15 10:10 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 24/36] sparc32: " Matthew Wilcox (Oracle)
2023-03-15 10:11 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 25/36] sparc64: " Matthew Wilcox (Oracle)
2023-03-15 10:11 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 26/36] um: " Matthew Wilcox (Oracle)
2023-03-15 10:12 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 27/36] x86: " Matthew Wilcox (Oracle)
2023-03-15 10:12 ` Mike Rapoport
2023-03-15 10:34 ` Peter Zijlstra
2023-03-15 11:16 ` Mike Rapoport
2023-03-15 11:19 ` Peter Zijlstra
2023-03-15 16:12 ` Matthew Wilcox
2023-03-15 5:14 ` [PATCH v4 28/36] xtensa: " Matthew Wilcox (Oracle)
2023-03-15 10:12 ` Mike Rapoport
2023-03-15 5:14 ` [PATCH v4 29/36] mm: Remove page_mapping_file() Matthew Wilcox (Oracle)
2023-05-25 3:50 ` Anshuman Khandual
2023-05-25 4:03 ` Matthew Wilcox
2023-05-25 4:46 ` Anshuman Khandual
2023-05-25 5:37 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 30/36] mm: Rationalise flush_icache_pages() and flush_icache_page() Matthew Wilcox (Oracle)
2023-03-15 5:14 ` [PATCH v4 31/36] mm: Tidy up set_ptes definition Matthew Wilcox (Oracle)
2023-05-25 6:20 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 32/36] mm: Use flush_icache_pages() in do_set_pmd() Matthew Wilcox (Oracle)
2023-05-25 6:31 ` Anshuman Khandual
2023-03-15 5:14 ` [PATCH v4 33/36] filemap: Add filemap_map_folio_range() Matthew Wilcox (Oracle)
2023-03-15 5:14 ` [PATCH v4 34/36] rmap: add folio_add_file_rmap_range() Matthew Wilcox (Oracle)
2023-03-15 13:34 ` Ryan Roberts
2023-03-15 16:08 ` Ryan Roberts
2023-03-16 16:27 ` Yin, Fengwei
2023-03-16 16:34 ` Ryan Roberts
2023-03-17 8:23 ` Yin, Fengwei
2023-03-17 12:46 ` Ryan Roberts
2023-03-17 13:28 ` Yin, Fengwei [this message]
2023-03-15 5:14 ` [PATCH v4 35/36] mm: Convert do_set_pte() to set_pte_range() Matthew Wilcox (Oracle)
2023-03-15 15:26 ` Ryan Roberts
2023-03-16 16:23 ` Yin, Fengwei
2023-03-16 16:38 ` Ryan Roberts
2023-03-16 16:41 ` Yin, Fengwei
2023-03-16 16:50 ` Ryan Roberts
2023-03-16 17:52 ` Matthew Wilcox
2023-03-17 1:58 ` Yin, Fengwei
2023-03-17 3:44 ` Matthew Wilcox
2023-03-17 6:33 ` Yin, Fengwei
2023-03-17 8:00 ` Ryan Roberts
2023-03-17 8:19 ` Yin, Fengwei
2023-03-17 13:00 ` Ryan Roberts
2023-03-17 13:44 ` Yin, Fengwei
2023-03-24 14:58 ` Will Deacon
2023-03-24 15:11 ` Matthew Wilcox
2023-03-24 17:23 ` Will Deacon
2023-03-27 1:23 ` Yin Fengwei
2023-03-20 13:38 ` Yin, Fengwei
2023-03-20 14:08 ` Matthew Wilcox
2023-03-21 1:58 ` Yin, Fengwei
2023-03-21 5:13 ` Yin Fengwei
2023-05-30 8:07 ` [PATCH 0/4] New page table range API fixup patches Yin Fengwei
2023-05-30 8:07 ` [PATCH 1/4] filemap: avoid interfere with xas.xa_index Yin Fengwei
2023-05-30 8:07 ` [PATCH 2/4] rmap: fix typo in folio_add_file_rmap_range() Yin Fengwei
2023-05-30 8:07 ` [PATCH 3/4] mm: mark PTEs referencing the accessed folio young Yin Fengwei
2023-05-30 8:07 ` [PATCH 4/4] filemap: Check address range in filemap_map_folio_range() Yin Fengwei
2023-03-15 5:14 ` [PATCH v4 36/36] filemap: Batch PTE mappings Matthew Wilcox (Oracle)
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=c7c7f5ed-ef80-ef6f-4a73-806b21e7c65d@intel.com \
--to=fengwei.yin@intel.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=ryan.roberts@arm.com \
--cc=willy@infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).