From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-10.7 required=3.0 tests=BAYES_00, HEADER_FROM_DIFFERENT_DOMAINS,HTML_MESSAGE,INCLUDES_CR_TRAILER,INCLUDES_PATCH, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,UNPARSEABLE_RELAY,URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id D85A0C433F5 for ; Thu, 9 Sep 2021 01:33:19 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id 4B5276113E for ; Thu, 9 Sep 2021 01:33:19 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org 4B5276113E Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linux.alibaba.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=kvack.org Received: by kanga.kvack.org (Postfix) id A9F476B0071; Wed, 8 Sep 2021 21:33:18 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id A4DC76B0072; Wed, 8 Sep 2021 21:33:18 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id 914DC900002; Wed, 8 Sep 2021 21:33:18 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0033.hostedemail.com [216.40.44.33]) by kanga.kvack.org (Postfix) with ESMTP id 7CE286B0071 for ; Wed, 8 Sep 2021 21:33:18 -0400 (EDT) Received: from smtpin01.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay05.hostedemail.com (Postfix) with ESMTP id 2B82618033E79 for ; Thu, 9 Sep 2021 01:33:18 +0000 (UTC) X-FDA: 78566312076.01.5F4A4F3 Received: from out30-43.freemail.mail.aliyun.com (out30-43.freemail.mail.aliyun.com [115.124.30.43]) by imf23.hostedemail.com (Postfix) with ESMTP id BE96F900009B for ; Thu, 9 Sep 2021 01:33:16 +0000 (UTC) X-Alimail-AntiSpam:AC=PASS;BC=-1|-1;BR=01201311R661e4;CH=green;DM=||false|;DS=||;FP=0|-1|-1|-1|0|-1|-1|-1;HT=e01e04423;MF=rongwei.wang@linux.alibaba.com;NM=1;PH=DS;RN=9;SR=0;TI=SMTPD_---0Unjpf29_1631151192; Received: from 30.240.104.109(mailfrom:rongwei.wang@linux.alibaba.com fp:SMTPD_---0Unjpf29_1631151192) by smtp.aliyun-inc.com(127.0.0.1); Thu, 09 Sep 2021 09:33:13 +0800 From: Rongwei Wang Message-Id: Content-Type: multipart/alternative; boundary="Apple-Mail=_4F7DA573-C26A-45CA-ADAB-BF011527DD24" Mime-Version: 1.0 (Mac OS X Mail 14.0 \(3654.60.0.2.21\)) Subject: Re: [PATCH 2/2] mm, thp: bail out early in collapse_file for writeback page Date: Thu, 9 Sep 2021 09:33:11 +0800 In-Reply-To: Cc: Linux MM , Linux Kernel Mailing List , Andrew Morton , cfijalkovich@google.com, song@kernel.org, william.kucharski@oracle.com, Hugh Dickins , "Kirill A. Shutemov" To: Yang Shi References: <20210906121200.57905-1-rongwei.wang@linux.alibaba.com> <20210906121200.57905-3-rongwei.wang@linux.alibaba.com> <44BE85B4-692C-41E8-B5A0-C1E0B0272ACD@linux.alibaba.com> X-Mailer: Apple Mail (2.3654.60.0.2.21) X-Rspamd-Server: rspam01 X-Rspamd-Queue-Id: BE96F900009B Authentication-Results: imf23.hostedemail.com; dkim=none; dmarc=pass (policy=none) header.from=alibaba.com; spf=pass (imf23.hostedemail.com: domain of rongwei.wang@linux.alibaba.com designates 115.124.30.43 as permitted sender) smtp.mailfrom=rongwei.wang@linux.alibaba.com X-Stat-Signature: f5omrfa9wtjzei1z9u3wjzdsqgt9n7fe X-HE-Tag: 1631151196-273047 X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: --Apple-Mail=_4F7DA573-C26A-45CA-ADAB-BF011527DD24 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=us-ascii > On Sep 9, 2021, at 5:51 AM, Yang Shi wrote: >=20 > On Tue, Sep 7, 2021 at 7:16 PM Rongwei Wang > > wrote: >>=20 >>=20 >>=20 >> On Sep 8, 2021, at 12:56 AM, Yang Shi wrote: >>=20 >> On Mon, Sep 6, 2021 at 5:12 AM Rongwei Wang >> wrote: >>=20 >>=20 >> Currently collapse_file does not explicitly check PG_writeback, = instead, >> page_has_private and try_to_release_page are used to filter writeback >> pages. This does not work for xfs with blocksize equal to or larger >>=20 >>=20 >> I'm not quite sure if using try_to_release() to check the writeback >> flag is on purpose or not, but it should be better to check the flag >> more explicitly. >>=20 >> Uh-huh. It is sure that try_to_release_page will check the writeback = flag >> of page. But the place where check the writeback flag using: >>=20 >> If (page_has_private(page) && >> !try_to_release_page(page, GFP_KERNEL)) { >> ... >> } >>=20 >> The original code will miss the page which in writeback status but = has no private data. >> The XFS case that above mentioned belongs to. >>=20 >> Thanks! >>=20 >>=20 >> than pagesize, because in such case xfs has no page->private. >>=20 >> This makes collapse_file bail out early for writeback page. = Otherwise, >> xfs end_page_writeback will panic as follows. >>=20 >> [ 6411.448211] page:fffffe00201bcc80 refcount:0 mapcount:0 = mapping:ffff0003f88c86a8 index:0x0 pfn:0x84ef32 >> [ 6411.448304] aops:xfs_address_space_operations [xfs] ino:30000b7 = dentry name:"libtest.so" >> [ 6411.448312] flags: = 0x57fffe0000008027(locked|referenced|uptodate|active|writeback) >> [ 6411.448317] raw: 57fffe0000008027 ffff80001b48bc28 = ffff80001b48bc28 ffff0003f88c86a8 >> [ 6411.448321] raw: 0000000000000000 0000000000000000 = 00000000ffffffff ffff0000c3e9a000 >> [ 6411.448324] page dumped because: VM_BUG_ON_PAGE(((unsigned int) = page_ref_count(page) + 127u <=3D 127u)) >> [ 6411.448327] page->mem_cgroup:ffff0000c3e9a000 >> [ 6411.448340] ------------[ cut here ]------------ >> [ 6411.448343] kernel BUG at include/linux/mm.h:1212! >> [ 6411.449288] Internal error: Oops - BUG: 0 [#1] SMP >> [ 6411.449786] Modules linked in: >> [ 6411.449790] BUG: Bad page state in process khugepaged pfn:84ef32 >> [ 6411.450143] xfs(E) >> [ 6411.450459] page:fffffe00201bcc80 refcount:0 mapcount:0 mapping:0 = index:0x0 pfn:0x84ef32 >> [ 6411.451361] libcrc32c(E) rfkill(E) aes_ce_blk(E) crypto_simd(E) = cryptd(E) aes_ce_cipher(E) crct10dif_ce(E) ghash_ce(E) sha1_ce(E) = uio_pdrv_genirq(E) uio(E) vfat(E) nfsd(E) fat(E) auth_rpcgss(E) = nfs_acl(E) lockd(E) grace(E) sunrpc(E) sch_fq_codel(E) ip_tables(E) = ext4(E) mbcache(E) jbd2(E) virtio_net(E) net_failover(E) virtio_blk(E) = failover(E) sha2_ce(E) sha256_arm64(E) virtio_mmio(E) virtio_pci(E) = virtio_ring(E) virtio(E) >> [ 6411.451387] CPU: 25 PID: 0 Comm: swapper/25 Kdump: loaded Tainted: = G W E >> [ 6411.451389] pstate: 60400005 (nZCv daif +PAN -UAO -TCO BTYPE=3D--) >> [ 6411.451393] pc : end_page_writeback+0x1c0/0x214 >> [ 6411.451394] lr : end_page_writeback+0x1c0/0x214 >> [ 6411.451395] sp : ffff800011ce3cc0 >> [ 6411.451396] x29: ffff800011ce3cc0 x28: 0000000000000000 >> [ 6411.451398] x27: ffff000c04608040 x26: 0000000000000000 >> [ 6411.451399] x25: ffff000c04608040 x24: 0000000000001000 >> [ 6411.451401] x23: ffff0003f88c8530 x22: 0000000000001000 >> [ 6411.451403] x21: ffff0003f88c8530 x20: 0000000000000000 >> [ 6411.451404] x19: fffffe00201bcc80 x18: 0000000000000030 >> [ 6411.451406] x17: 0000000000000000 x16: 0000000000000000 >> [ 6411.451407] x15: ffff000c018f9760 x14: ffffffffffffffff >> [ 6411.451409] x13: ffff8000119d72b0 x12: ffff8000119d6ee3 >> [ 6411.451410] x11: ffff8000117b69b8 x10: 00000000ffff8000 >> [ 6411.451412] x9 : ffff800010617534 x8 : 0000000000000000 >> [ 6411.451413] x7 : ffff8000114f69b8 x6 : 000000000000000f >> [ 6411.451415] x5 : 0000000000000000 x4 : 0000000000000000 >> [ 6411.451416] x3 : 0000000000000400 x2 : 0000000000000000 >> [ 6411.451418] x1 : 0000000000000000 x0 : 0000000000000000 >> [ 6411.451420] Call trace: >> [ 6411.451421] end_page_writeback+0x1c0/0x214 >> [ 6411.451424] iomap_finish_page_writeback+0x13c/0x204 >> [ 6411.451425] iomap_finish_ioend+0xe8/0x19c >> [ 6411.451426] iomap_writepage_end_bio+0x38/0x50 >> [ 6411.451427] bio_endio+0x168/0x1ec >> [ 6411.451430] blk_update_request+0x278/0x3f0 >> [ 6411.451432] blk_mq_end_request+0x34/0x15c >> [ 6411.451435] virtblk_request_done+0x38/0x74 [virtio_blk] >> [ 6411.451437] blk_done_softirq+0xc4/0x110 >> [ 6411.451439] __do_softirq+0x128/0x38c >> [ 6411.451441] __irq_exit_rcu+0x118/0x150 >> [ 6411.451442] irq_exit+0x1c/0x30 >> [ 6411.451445] __handle_domain_irq+0x8c/0xf0 >> [ 6411.451446] gic_handle_irq+0x84/0x108 >> [ 6411.451447] el1_irq+0xcc/0x180 >> [ 6411.451448] arch_cpu_idle+0x18/0x40 >> [ 6411.451450] default_idle_call+0x4c/0x1a0 >> [ 6411.451453] cpuidle_idle_call+0x168/0x1e0 >> [ 6411.451454] do_idle+0xb4/0x104 >> [ 6411.451455] cpu_startup_entry+0x30/0x9c >> [ 6411.451458] secondary_start_kernel+0x104/0x180 >> [ 6411.451460] Code: d4210000 b0006161 910c8021 94013f4d (d4210000) >> [ 6411.451462] ---[ end trace 4a88c6a074082f8c ]--- >> [ 6411.451464] Kernel panic - not syncing: Oops - BUG: Fatal = exception in interrupt >>=20 >> Fixes: eb6ecbed0aa2 ("mm, thp: relax the VM_DENYWRITE constraint on = file-backed THPs") >> Signed-off-by: Xu Yu >> Signed-off-by: Rongwei Wang >> --- >> mm/khugepaged.c | 3 ++- >> 1 file changed, 2 insertions(+), 1 deletion(-) >>=20 >> diff --git a/mm/khugepaged.c b/mm/khugepaged.c >> index 045cc57..529b117 100644 >> --- a/mm/khugepaged.c >> +++ b/mm/khugepaged.c >> @@ -1798,7 +1798,8 @@ static void collapse_file(struct mm_struct *mm, >> goto out_unlock; >> } >>=20 >> - if (!is_shmem && PageDirty(page)) { >> + if (!is_shmem && (PageDirty(page) || >> + PageWriteback(page))) { >>=20 >>=20 >> Why not check the writeback flag in the place before taking page lock = too? >>=20 >> Thanks! Your words remind me. I will double check the writeback flag = before >> and after taking page lock. Likes: >>=20 >> + } else if (PageWriteback(page)) { >> + xas_unlock_irq(&xas); >> + result =3D SCAN_FAIL; >> + goto xa_unlocked; >> } else if (trylock_page(page)) { >> get_page(page); >> xas_unlock_irq(&xas); >> @@ -1798,7 +1802,8 @@ static void collapse_file(struct mm_struct *mm, >> goto out_unlock; >> } >>=20 >> - if (!is_shmem && PageDirty(page)) { >> + if (!is_shmem && (PageDirty(page) || >> + PageWriteback(page))) { >=20 > Yeah, this is what I meant. Thanks. OK, I will update these two patches and send PATCH v2 after receiving = comments from Matthew. =20 Thanks! >=20 >>=20 >> Thanks! >>=20 >>=20 >> /* >> * khugepaged only works on read-only fd, so = this >> * page is dirty because it hasn't been flushed >> -- >> 1.8.3.1 --Apple-Mail=_4F7DA573-C26A-45CA-ADAB-BF011527DD24 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=us-ascii

On Sep 9, 2021, at 5:51 AM, Yang Shi <shy828301@gmail.com>= wrote:

On Tue, Sep 7, 2021 at 7:16 PM Rongwei Wang
<rongwei.wang@linux.alibaba.com> wrote:



On Sep 8, 2021, at 12:56 AM, Yang Shi <shy828301@gmail.com>= wrote:

On Mon, Sep 6, 2021 at 5:12 AM = Rongwei Wang
<rongwei.wang@linux.alibaba.com> wrote:

Currently collapse_file does not explicitly = check PG_writeback, instead,
page_has_private and = try_to_release_page are used to filter writeback
pages. = This does not work for xfs with blocksize equal to or larger


I'm not quite sure if using = try_to_release() to check the writeback
flag is on purpose = or not, but it should be better to check the flag
more = explicitly.

Uh-huh. It is sure that = try_to_release_page will check the writeback flag
of page. = But the place where check the writeback flag using:

If (page_has_private(page) &&
!try_to_release_page(page, GFP_KERNEL)) {
...
}

The original code will miss = the page which in writeback status but has no private data.
The XFS case that above mentioned belongs to.
Thanks!


than = pagesize, because in such case xfs has no page->private.

This makes collapse_file bail out early for = writeback page. Otherwise,
xfs end_page_writeback will = panic as follows.

[ 6411.448211] = page:fffffe00201bcc80 refcount:0 mapcount:0 mapping:ffff0003f88c86a8 = index:0x0 pfn:0x84ef32
[ 6411.448304] = aops:xfs_address_space_operations [xfs] ino:30000b7 dentry = name:"libtest.so"
[ 6411.448312] flags: = 0x57fffe0000008027(locked|referenced|uptodate|active|writeback)
[ 6411.448317] raw: 57fffe0000008027 ffff80001b48bc28 = ffff80001b48bc28 ffff0003f88c86a8
[ 6411.448321] raw: = 0000000000000000 0000000000000000 00000000ffffffff ffff0000c3e9a000
[ 6411.448324] page dumped because: VM_BUG_ON_PAGE(((unsigned = int) page_ref_count(page) + 127u <=3D 127u))
[ = 6411.448327] page->mem_cgroup:ffff0000c3e9a000
[ = 6411.448340] ------------[ cut here ]------------
[ = 6411.448343] kernel BUG at include/linux/mm.h:1212!
[ = 6411.449288] Internal error: Oops - BUG: 0 [#1] SMP
[ = 6411.449786] Modules linked in:
[ 6411.449790] BUG: Bad = page state in process khugepaged  pfn:84ef32
[ = 6411.450143]  xfs(E)
[ 6411.450459] = page:fffffe00201bcc80 refcount:0 mapcount:0 mapping:0 index:0x0 = pfn:0x84ef32
[ 6411.451361]  libcrc32c(E) rfkill(E) = aes_ce_blk(E) crypto_simd(E) cryptd(E) aes_ce_cipher(E) crct10dif_ce(E) = ghash_ce(E) sha1_ce(E) uio_pdrv_genirq(E) uio(E) vfat(E) nfsd(E) fat(E) = auth_rpcgss(E) nfs_acl(E) lockd(E) grace(E) sunrpc(E) sch_fq_codel(E) = ip_tables(E) ext4(E) mbcache(E) jbd2(E) virtio_net(E) net_failover(E) = virtio_blk(E) failover(E) sha2_ce(E) sha256_arm64(E) virtio_mmio(E) = virtio_pci(E) virtio_ring(E) virtio(E)
[ 6411.451387] CPU: = 25 PID: 0 Comm: swapper/25 Kdump: loaded Tainted: G =        W   E
[ = 6411.451389] pstate: 60400005 (nZCv daif +PAN -UAO -TCO BTYPE=3D--)
[ 6411.451393] pc : end_page_writeback+0x1c0/0x214
[ 6411.451394] lr : end_page_writeback+0x1c0/0x214
[ 6411.451395] sp : ffff800011ce3cc0
[ = 6411.451396] x29: ffff800011ce3cc0 x28: 0000000000000000
[ = 6411.451398] x27: ffff000c04608040 x26: 0000000000000000
[ = 6411.451399] x25: ffff000c04608040 x24: 0000000000001000
[ = 6411.451401] x23: ffff0003f88c8530 x22: 0000000000001000
[ = 6411.451403] x21: ffff0003f88c8530 x20: 0000000000000000
[ = 6411.451404] x19: fffffe00201bcc80 x18: 0000000000000030
[ = 6411.451406] x17: 0000000000000000 x16: 0000000000000000
[ = 6411.451407] x15: ffff000c018f9760 x14: ffffffffffffffff
[ = 6411.451409] x13: ffff8000119d72b0 x12: ffff8000119d6ee3
[ = 6411.451410] x11: ffff8000117b69b8 x10: 00000000ffff8000
[ = 6411.451412] x9 : ffff800010617534 x8 : 0000000000000000
[ = 6411.451413] x7 : ffff8000114f69b8 x6 : 000000000000000f
[ = 6411.451415] x5 : 0000000000000000 x4 : 0000000000000000
[ = 6411.451416] x3 : 0000000000000400 x2 : 0000000000000000
[ = 6411.451418] x1 : 0000000000000000 x0 : 0000000000000000
[ = 6411.451420] Call trace:
[ 6411.451421] =  end_page_writeback+0x1c0/0x214
[ 6411.451424] =  iomap_finish_page_writeback+0x13c/0x204
[ = 6411.451425]  iomap_finish_ioend+0xe8/0x19c
[ = 6411.451426]  iomap_writepage_end_bio+0x38/0x50
[ = 6411.451427]  bio_endio+0x168/0x1ec
[ 6411.451430] =  blk_update_request+0x278/0x3f0
[ 6411.451432] =  blk_mq_end_request+0x34/0x15c
[ 6411.451435] =  virtblk_request_done+0x38/0x74 [virtio_blk]
[ = 6411.451437]  blk_done_softirq+0xc4/0x110
[ = 6411.451439]  __do_softirq+0x128/0x38c
[ 6411.451441] =  __irq_exit_rcu+0x118/0x150
[ 6411.451442] =  irq_exit+0x1c/0x30
[ 6411.451445] =  __handle_domain_irq+0x8c/0xf0
[ 6411.451446] =  gic_handle_irq+0x84/0x108
[ 6411.451447] =  el1_irq+0xcc/0x180
[ 6411.451448] =  arch_cpu_idle+0x18/0x40
[ 6411.451450] =  default_idle_call+0x4c/0x1a0
[ 6411.451453] =  cpuidle_idle_call+0x168/0x1e0
[ 6411.451454] =  do_idle+0xb4/0x104
[ 6411.451455] =  cpu_startup_entry+0x30/0x9c
[ 6411.451458] =  secondary_start_kernel+0x104/0x180
[ 6411.451460] = Code: d4210000 b0006161 910c8021 94013f4d (d4210000)
[ = 6411.451462] ---[ end trace 4a88c6a074082f8c ]---
[ = 6411.451464] Kernel panic - not syncing: Oops - BUG: Fatal exception in = interrupt

Fixes: eb6ecbed0aa2 ("mm, thp: = relax the VM_DENYWRITE constraint on file-backed THPs")
Signed-off-by: Xu Yu <xuyu@linux.alibaba.com>
Signed-off-by: = Rongwei Wang <rongwei.wang@linux.alibaba.com>
---
mm/khugepaged.c | 3 ++-
1 file changed, 2 = insertions(+), 1 deletion(-)

diff --git = a/mm/khugepaged.c b/mm/khugepaged.c
index 045cc57..529b117 = 100644
--- a/mm/khugepaged.c
+++ = b/mm/khugepaged.c
@@ -1798,7 +1798,8 @@ static void = collapse_file(struct mm_struct *mm,
          &nb= sp;           goto = out_unlock;
          &nb= sp;   }

- =             &n= bsp; if (!is_shmem && PageDirty(page)) {
+ =             &n= bsp; if (!is_shmem && (PageDirty(page) ||
+ =             &n= bsp;           &nbs= p;       PageWriteback(page))) {


Why not check the writeback = flag in the place before taking page lock too?

Thanks! Your words remind me. I will double check the = writeback flag before
and after taking page lock. = Likes:

+ =             &n= bsp;         } else if = (PageWriteback(page)) {
+ =             &n= bsp;           &nbs= p;     xas_unlock_irq(&xas);
+ =             &n= bsp;           &nbs= p;     result =3D SCAN_FAIL;
+ =             &n= bsp;           &nbs= p;     goto xa_unlocked;
          &nb= sp;            = ;} else if (trylock_page(page)) {
          &nb= sp;            = ;        get_page(page);
          &nb= sp;            = ;        xas_unlock_irq(&xas);=
@@ -1798,7 +1802,8 @@ static void collapse_file(struct = mm_struct *mm,
          &nb= sp;            = ;goto out_unlock;
          &nb= sp;    }

- =             &n= bsp; if (!is_shmem && PageDirty(page)) {
+ =             &n= bsp; if (!is_shmem && (PageDirty(page) ||
+ =             &n= bsp;           &nbs= p;       PageWriteback(page))) {

Yeah, this is what I meant. Thanks.
OK, I will update these two patches = and send PATCH v2 after receiving comments from = Matthew.
 
Thanks!


Thanks!


          &nb= sp;           /*          &nb= sp;            = ;* khugepaged only works on read-only fd, so this
          &nb= sp;            = ;* page is dirty because it hasn't been flushed
--
1.8.3.1

= --Apple-Mail=_4F7DA573-C26A-45CA-ADAB-BF011527DD24--