From: "Liam R. Howlett" <Liam.Howlett@Oracle.com>
To: Suren Baghdasaryan <surenb@google.com>
Cc: linux-mm@kvack.org, willy@infradead.org,
Laurent Dufour <ldufour@linux.ibm.com>,
Michel Lespinasse <michel@lespinasse.org>,
Jerome Glisse <jglisse@google.com>,
Vlastimil Babka <vbabka@suse.cz>,
"Paul E. McKenney" <paulmck@kernel.org>
Subject: Re: lockdep issue with per-vma locking
Date: Wed, 12 Jul 2023 11:30:50 -0400 [thread overview]
Message-ID: <20230712153050.gm57uee23ykm3srk@revolver> (raw)
In-Reply-To: <CAJuCfpGdjW+zprwC2uhGTLhayintz8-0jWUS-Wky1Dz+ipFphg@mail.gmail.com>
* Suren Baghdasaryan <surenb@google.com> [230712 11:15]:
> On Tue, Jul 11, 2023 at 7:26 PM Liam R. Howlett <Liam.Howlett@oracle.com> wrote:
> >
> > Suren,
> >
> > When running kselftest mm, I believe I've come across a lockdep issue
> > with the per-vma locking pagefault:
> >
> > [ 226.105499] WARNING: CPU: 1 PID: 1907 at include/linux/mmap_lock.h:71 handle_userfault+0x34d/0xff0
> > [ 226.106517] Modules linked in:
> > [ 226.107060] CPU: 1 PID: 1907 Comm: uffd-unit-tests Not tainted 6.5.0-rc1+ #636
> > [ 226.108099] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.0-debian-1.16.0-5 04/01/2014
> > [ 226.109626] RIP: 0010:handle_userfault+0x34d/0xff0
> > [ 226.113056] Code: 00 48 85 c0 0f 85 d4 fe ff ff 4c 89 f7 e8 bb 58 ea ff 0f 0b 31 f6 49 8d be a0 01 00 00 e8 0b 8b 53 01 85 c0 0f 85 00 fe ff ff <0f> 0b e9 f9 fd ff ff 49 8d be a0 01 00 00 be ff ff ff ff e8 eb 8a
> > [ 226.115798] RSP: 0000:ffff888113a8fbf0 EFLAGS: 00010246
> > [ 226.116570] RAX: 0000000000000000 RBX: ffff888113a8fdc8 RCX: 0000000000000001
> > [ 226.117630] RDX: 0000000000000000 RSI: ffffffff97a70220 RDI: ffffffff97c316e0
> > [ 226.118654] RBP: ffff88811de7c1e0 R08: 0000000000000000 R09: ffffed1022991400
> > [ 226.119508] R10: ffff888114c8a003 R11: 0000000000000000 R12: 0000000000000200
> > [ 226.120471] R13: ffff88811de7c1f0 R14: ffff888106ebec00 R15: 0000000000001000
> > [ 226.121521] FS: 00007f226ec0f740(0000) GS:ffff88836f280000(0000) knlGS:0000000000000000
> > [ 226.122543] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > [ 226.123242] CR2: 00007f226ac0f028 CR3: 00000001088a4001 CR4: 0000000000370ee0
> > [ 226.124075] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > [ 226.125073] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> > [ 226.126308] Call Trace:
> > [ 226.127473] <TASK>
> > [ 226.128001] ? __warn+0x9c/0x1f0
> > [ 226.129005] ? handle_userfault+0x34d/0xff0
> > [ 226.129940] ? report_bug+0x1f2/0x220
> > [ 226.130700] ? handle_bug+0x3c/0x70
> > [ 226.131234] ? exc_invalid_op+0x13/0x40
> > [ 226.131827] ? asm_exc_invalid_op+0x16/0x20
> > [ 226.132516] ? handle_userfault+0x34d/0xff0
> > [ 226.133193] ? __pfx_do_raw_spin_lock+0x10/0x10
> > [ 226.133862] ? find_held_lock+0x83/0xa0
> > [ 226.134602] ? do_anonymous_page+0x81f/0x870
> > [ 226.135314] ? __pfx_handle_userfault+0x10/0x10
> > [ 226.136226] ? __pte_offset_map_lock+0xd4/0x160
> > [ 226.136958] ? do_raw_spin_unlock+0x92/0xf0
> > [ 226.137547] ? preempt_count_sub+0xf/0xc0
> > [ 226.138011] ? _raw_spin_unlock+0x24/0x40
> > [ 226.138594] ? do_anonymous_page+0x81f/0x870
> > [ 226.139239] __handle_mm_fault+0x40a/0x470
> > [ 226.139749] ? __pfx___handle_mm_fault+0x10/0x10
> > [ 226.140516] handle_mm_fault+0xe9/0x270
> > [ 226.141015] do_user_addr_fault+0x1a9/0x810
> > [ 226.141638] exc_page_fault+0x58/0xe0
> > [ 226.142101] asm_exc_page_fault+0x22/0x30
> > [ 226.142713] RIP: 0033:0x561107c4967e
> > [ 226.143391] Code: 48 89 85 18 ff ff ff e9 e2 00 00 00 48 8b 15 49 a0 00 00 48 8b 05 2a a0 00 00 48 0f af 45 f8 48 83 c0 2f 48 01 d0 48 83 e0 f8 <48> 8b 00 48 89 45 c8 48 8b 05 54 a0 00 00 48 8b 55 f8 48 c1 e2 03
> > [ 226.145946] RSP: 002b:00007ffee4f22120 EFLAGS: 00010206
> > [ 226.146745] RAX: 00007f226ac0f028 RBX: 00007ffee4f22448 RCX: 00007f226eca1bb4
> > [ 226.147912] RDX: 00007f226ac0f000 RSI: 0000000000000001 RDI: 0000000000000000
> > [ 226.149093] RBP: 00007ffee4f22220 R08: 0000000000000000 R09: 0000000000000000
> > [ 226.150218] R10: 0000000000000008 R11: 0000000000000246 R12: 0000000000000000
> > [ 226.151313] R13: 00007ffee4f22458 R14: 0000561107c52dd8 R15: 00007f226ee34020
> > [ 226.152464] </TASK>
> > [ 226.152802] irq event stamp: 3177751
> > [ 226.153348] hardirqs last enabled at (3177761): [<ffffffff95d9fa69>] __up_console_sem+0x59/0x80
> > [ 226.154679] hardirqs last disabled at (3177772): [<ffffffff95d9fa4e>] __up_console_sem+0x3e/0x80
> > [ 226.155998] softirqs last enabled at (3177676): [<ffffffff95ccea54>] irq_exit_rcu+0x94/0xf0
> > [ 226.157364] softirqs last disabled at (3177667): [<ffffffff95ccea54>] irq_exit_rcu+0x94/0xf0
> > [ 226.158721] ---[ end trace 0000000000000000 ]---
> >
> >
> > CONFIG_PER_VMA_LOCK calls handle_mm_fault() in mm/memory.c
> > handle_mm_fault() may have an outdated comment, depending on what "mm
> > semaphore" means:
> >
> > * By the time we get here, we already hold the mm semaphore
> >
> > __handle_mm_fault+0x40a/0x470:
> > do_pte_missing at mm/memory.c:3672
> > (inlined by) handle_pte_fault at mm/memory.c:4955
> > (inlined by) __handle_mm_fault at mm/memory.c:5095
> >
> > handle_userfault+0x34d/0xff0:
> > mmap_assert_write_locked at include/linux/mmap_lock.h:71
> > (inlined by) __is_vma_write_locked at include/linux/mm.h:673
> > (inlined by) vma_assert_locked at include/linux/mm.h:714
> > (inlined by) assert_fault_locked at include/linux/mm.h:747
> > (inlined by) handle_userfault at fs/userfaultfd.c:440
> >
> > It looks like vma_assert_locked() is causing a problem if the mmap write
> > lock is not held in write mode.
> >
> > It looks to be an easy fix of checking the mmap_lock is held in write
> > mode in every other call location BUT the vma_assert_locked() path?
>
> Thanks Liam! Yes, the fix is indeed very simple. I missed the fact
> that __is_vma_write_locked() generates an assertion, which should
> probably be changed. I believe the same assertion is found by syzbot
> here: https://lore.kernel.org/all/0000000000002db68f05ffb791bc@google.com/#t
Yeah, looks the same. Sorry for the noise.
> I'll post a fix shortly.
> BTW, this is happening only in mm-unstable, right?
Well, I tested it only in mm-unstable..
It came up while i was testing an unrelated fix for another kselftest:mm
that I broke.
Thanks,
Liam
next prev parent reply other threads:[~2023-07-12 15:31 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-07-12 2:26 Liam R. Howlett
2023-07-12 15:15 ` Suren Baghdasaryan
2023-07-12 15:30 ` Liam R. Howlett [this message]
2023-07-12 15:42 ` Suren Baghdasaryan
2023-07-12 19:59 ` Suren Baghdasaryan
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=20230712153050.gm57uee23ykm3srk@revolver \
--to=liam.howlett@oracle.com \
--cc=jglisse@google.com \
--cc=ldufour@linux.ibm.com \
--cc=linux-mm@kvack.org \
--cc=michel@lespinasse.org \
--cc=paulmck@kernel.org \
--cc=surenb@google.com \
--cc=vbabka@suse.cz \
--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