Message-ID: <CA+EESO6F4=jQujGU35EPp_EVtHRHEG7b7_izLv1-NJW_VFvGRw@mail.gmail.com>
Date: Wed, 6 Aug 2025 10:30:47 -0700
From: Lokesh Gidra <lokeshgidra@...gle.com>
To: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>
Cc: akpm@...ux-foundation.org, aarcange@...hat.com, linux-mm@...ck.org, 
	linux-kernel@...r.kernel.org, 21cnbao@...il.com, ngeoffray@...gle.com, 
	Suren Baghdasaryan <surenb@...gle.com>, Kalesh Singh <kaleshsingh@...gle.com>, 
	Barry Song <v-songbaohua@...o.com>, David Hildenbrand <david@...hat.com>, Peter Xu <peterx@...hat.com>
Subject: Re: [PATCH v2] userfaultfd: opportunistic TLB-flush batching for
 present pages in MOVE

On Wed, Aug 6, 2025 at 10:18 AM Lorenzo Stoakes
<lorenzo.stoakes@...cle.com> wrote:
>
> Andrew - Could we drop this for now, please, it's splatting and has broken
> mm-new.
>
> Lokesh - could you make sure to run the mm self tests with CONFIG_DBUG_VM
> set before you submit please? As this is splat is occurring immediately on
> uffd-unit-tests.

Sincere apologies. Will be extra careful from now on.
>
> On Tue, Aug 05, 2025 at 05:14:10AM -0700, Lokesh Gidra wrote:
> > MOVE ioctl's runtime is dominated by TLB-flush cost, which is required
> > for moving present pages. Mitigate this cost by opportunistically
> > batching present contiguous pages for TLB flushing.
> >
> > Without batching, in our testing on an arm64 Android device with UFFD GC,
> > which uses MOVE ioctl for compaction, we observed that out of the total
> > time spent in move_pages_pte(), over 40% is in ptep_clear_flush(), and
> > ~20% in vm_normal_folio().
> >
> > With batching, the proportion of vm_normal_folio() increases to over
> > 70% of move_pages_pte() without any changes to vm_normal_folio().
> > Furthermore, time spent within move_pages_pte() is only ~20%, which
> > includes TLB-flush overhead.
> >
> > Cc: Suren Baghdasaryan <surenb@...gle.com>
> > Cc: Kalesh Singh <kaleshsingh@...gle.com>
> > Cc: Barry Song <v-songbaohua@...o.com>
> > Cc: David Hildenbrand <david@...hat.com>
> > Cc: Peter Xu <peterx@...hat.com>
> > Signed-off-by: Lokesh Gidra <lokeshgidra@...gle.com>
> > ---
> > Changes since v1 [1]
> > - Removed flush_tlb_batched_pending(), per Barry Song
> > - Unified single and multi page case, per Barry Song
>
> Splat, decoded via scripts/decode_stacktrace.sh:
>
> $ sudo ./uffd-unit-tests
> Testing UFFDIO_API (with syscall)... done
> Testing UFFDIO_API (with /dev/userfaultfd)... done
> Testing register-ioctls on anon... done
> Testing register-ioctls on shmem... done
> Testing register-ioctls on shmem-private... done
> Testing register-ioctls on hugetlb... skipped [reason: memory allocation failed]
> Testing register-ioctls on hugetlb-private... skipped [reason: memory allocation failed]
> Testing zeropage on anon... done
> Testing zeropage on shmem... done
> Testing zeropage on shmem-private... done
> Testing zeropage on hugetlb... skipped [reason: memory allocation failed]
> Testing zeropage on hugetlb-private... skipped [reason: memory allocation failed]
> Testing move on anon... [   12.230740] Kernel panic - not syncing: kernel: panic_on_warn set ...
> [   12.231322] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Arch Linux 1.17.0-1-1 04/01/2014
> [   12.231655] Call Trace:
> [   12.231748]  <TASK>
> [   12.231830] dump_stack_lvl (lib/dump_stack.c:122)
> [   12.231963] vpanic (kernel/panic.c:448)
> [   12.232088] panic (kernel/panic.c:312 kernel/panic.c:303)
> [   12.232199] ? move_pages (mm/userfaultfd.c:1964 (discriminator 2))
>
> Appears to be:
>
>         VM_WARN_ON_ONCE(err > 0);
>
> [   12.232341] check_panic_on_warn.cold (kernel/panic.c:327)
> [   12.232502] __warn.cold (kernel/panic.c:839)
> [   12.232628] ? move_pages (mm/userfaultfd.c:1964 (discriminator 2))
> [   12.232764] report_bug (lib/bug.c:176 lib/bug.c:215)
> [   12.232891] handle_bug (arch/x86/kernel/traps.c:338 (discriminator 1))
> [   12.233034] ? move_pages (mm/userfaultfd.c:1964 (discriminator 2))
> [   12.233174] exc_invalid_op (arch/x86/kernel/traps.c:392 (discriminator 3))
> [   12.233312] asm_exc_invalid_op (./arch/x86/include/asm/idtentry.h:621)
> [   12.233460] RIP: 0010:move_pages (mm/userfaultfd.c:1964 (discriminator 2))
> [ 12.233615] Code: 5e 41 5f c3 cc cc cc cc 49 89 c5 e9 e1 fe ff ff eb c4 e9 6d ff ff ff 90 0f 0b 90 45 31 ff eb cf 90 0f 0b 90 48 85 d2 7e c6 90 <0f> 0b 90 eb b9 90 0f 0b 90 f7 c1 ff 0f 00 00 0f 84 4e fe ff ff 90
> All code
> ========
>    0:   5e                      pop    %rsi
>    1:   41 5f                   pop    %r15
>    3:   c3                      ret
>    4:   cc                      int3
>    5:   cc                      int3
>    6:   cc                      int3
>    7:   cc                      int3
>    8:   49 89 c5                mov    %rax,%r13
>    b:   e9 e1 fe ff ff          jmp    0xfffffffffffffef1
>   10:   eb c4                   jmp    0xffffffffffffffd6
>   12:   e9 6d ff ff ff          jmp    0xffffffffffffff84
>   17:   90                      nop
>   18:   0f 0b                   ud2
>   1a:   90                      nop
>   1b:   45 31 ff                xor    %r15d,%r15d
>   1e:   eb cf                   jmp    0xffffffffffffffef
>   20:   90                      nop
>   21:   0f 0b                   ud2
>   23:   90                      nop
>   24:   48 85 d2                test   %rdx,%rdx
>   27:   7e c6                   jle    0xffffffffffffffef
>   29:   90                      nop
>   2a:*  0f 0b                   ud2             <-- trapping instruction
>   2c:   90                      nop
>   2d:   eb b9                   jmp    0xffffffffffffffe8
>   2f:   90                      nop
>   30:   0f 0b                   ud2
>   32:   90                      nop
>   33:   f7 c1 ff 0f 00 00       test   $0xfff,%ecx
>   39:   0f 84 4e fe ff ff       je     0xfffffffffffffe8d
>   3f:   90                      nop
>
> Code starting with the faulting instruction
> ===========================================
>    0:   0f 0b                   ud2
>    2:   90                      nop
>    3:   eb b9                   jmp    0xffffffffffffffbe
>    5:   90                      nop
>    6:   0f 0b                   ud2
>    8:   90                      nop
>    9:   f7 c1 ff 0f 00 00       test   $0xfff,%ecx
>    f:   0f 84 4e fe ff ff       je     0xfffffffffffffe63
>   15:   90                      nop
> [   12.234294] RSP: 0018:ffffafeb00483d70 EFLAGS: 00010206
> [   12.234484] RAX: 0000000000000000 RBX: 0000000000000001 RCX: 0000000000000002
> [   12.234738] RDX: 0000000000001000 RSI: ffff90afbb433078 RDI: ffff90afbc7f0080
> [   12.234997] RBP: ffff90afbc7f0000 R08: ffff90b037cae540 R09: 0000000000000001
> [   12.235255] R10: ffffffffffffffff R11: 0000000000000003 R12: ffff90af01acb980
> [   12.235508] R13: ffff90afbc7f0240 R14: ffffefc386e80b00 R15: 0000000000001000
> [   12.235764] userfaultfd_ioctl (fs/userfaultfd.c:1925 fs/userfaultfd.c:2046)
> [   12.235917] __x64_sys_ioctl (fs/ioctl.c:51 fs/ioctl.c:598 fs/ioctl.c:584 fs/ioctl.c:584)
> [   12.236065] ? ksys_read (./include/linux/file.h:63 ./include/linux/file.h:80 ./include/linux/file.h:85 fs/read_write.c:706)
> [   12.236202] do_syscall_64 (arch/x86/entry/syscall_64.c:63 (discriminator 1) arch/x86/entry/syscall_64.c:94 (discriminator 1))
> [   12.236345] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:130)
> [   12.236524] RIP: 0033:0x7f457600fecd
> [ 12.236658] Code: 04 25 28 00 00 00 48 89 45 c8 31 c0 48 8d 45 10 c7 45 b0 10 00 00 00 48 89 45 b8 48 8d 45 d0 48 89 45 c0 b8 10 00 00 00 0f 05 <89> c2 3d 00 f0 ff ff 77 1a 48 8b 45 c8 64 48 2b 04 25 28 00 00 00
> All code
> ========
>    0:   04 25                   add    $0x25,%al
>    2:   28 00                   sub    %al,(%rax)
>    4:   00 00                   add    %al,(%rax)
>    6:   48 89 45 c8             mov    %rax,-0x38(%rbp)
>    a:   31 c0                   xor    %eax,%eax
>    c:   48 8d 45 10             lea    0x10(%rbp),%rax
>   10:   c7 45 b0 10 00 00 00    movl   $0x10,-0x50(%rbp)
>   17:   48 89 45 b8             mov    %rax,-0x48(%rbp)
>   1b:   48 8d 45 d0             lea    -0x30(%rbp),%rax
>   1f:   48 89 45 c0             mov    %rax,-0x40(%rbp)
>   23:   b8 10 00 00 00          mov    $0x10,%eax
>   28:   0f 05                   syscall
>   2a:*  89 c2                   mov    %eax,%edx                <-- trapping instruction
>   2c:   3d 00 f0 ff ff          cmp    $0xfffff000,%eax
>   31:   77 1a                   ja     0x4d
>   33:   48 8b 45 c8             mov    -0x38(%rbp),%rax
>   37:   64 48 2b 04 25 28 00    sub    %fs:0x28,%rax
>   3e:   00 00
>
> Code starting with the faulting instruction
> ===========================================
>    0:   89 c2                   mov    %eax,%edx
>    2:   3d 00 f0 ff ff          cmp    $0xfffff000,%eax
>    7:   77 1a                   ja     0x23
>    9:   48 8b 45 c8             mov    -0x38(%rbp),%rax
>    d:   64 48 2b 04 25 28 00    sub    %fs:0x28,%rax
>   14:   00 00
> [   12.237333] RSP: 002b:00007f4569dfed20 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
> [   12.237599] RAX: ffffffffffffffda RBX: 0000000000001000 RCX: 00007f457600fecd
> [   12.237854] RDX: 00007f4569dfeda0 RSI: 00000000c028aa05 RDI: 0000000000000009
> [   12.238129] RBP: 00007f4569dfed70 R08: 0000000000000000 R09: 0000000000000000
> [   12.238384] R10: 0000000000000000 R11: 0000000000000246 R12: 00007fffb974ba60
> [   12.238636] R13: 00007fffb974b830 R14: 00007f4569dffcdc R15: 00007fffb974b937
> [   12.238890]  </TASK>
> [   12.239094] Kernel Offset: 0x33000000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
> [   12.239480] ---[ end Kernel panic - not syncing: kernel: panic_on_warn set ... ]---

Powered by blists - more mailing lists