From: Stanislav Fomichev <stfomichev@gmail.com>
To: Gilad Naaman <gnaaman@drivenets.com>
Cc: netdev <netdev@vger.kernel.org>,
"David S. Miller" <davem@davemloft.net>,
Eric Dumazet <edumazet@google.com>,
Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
Kuniyuki Iwashima <kuniyu@amazon.com>
Subject: Re: [PATCH net-next v7 0/6] neighbour: Improve neigh_flush_dev performance
Date: Tue, 22 Oct 2024 16:48:40 -0700 [thread overview]
Message-ID: <Zxg52Ccb8FF8MCt0@mini-arch> (raw)
In-Reply-To: <20241022134343.3354111-1-gnaaman@drivenets.com>
On 10/22, Gilad Naaman wrote:
> This patchsets improves the performance of neigh_flush_dev.
>
> Currently, the only way to implement it requires traversing
> all neighbours known to the kernel, across all network-namespaces.
>
> This means that some flows are slowed down as a function of neighbour-scale,
> even if the specific link they're handling has little to no neighbours.
>
> In order to solve this, this patchset adds a netdev->neighbours list,
> as well as making the original linked-list doubly-, so that it is
> possible to unlink neighbours without traversing the hash-bucket to
> obtain the previous neighbour.
>
> The original use-case we encountered was mass-deletion of links (12K
> VLANs) while there are 50K ARPs and 50K NDPs in the system; though the
> slowdowns would also appear when the links are set down.
>
> Changes in v7:
>
> - Fix crash due to use of poisoned hlist_node
> - Apply samx-tree formatting
>
> Gilad Naaman (6):
> neighbour: Add hlist_node to struct neighbour
> neighbour: Define neigh_for_each_in_bucket
> neighbour: Convert seq_file functions to use hlist
> neighbour: Convert iteration to use hlist+macro
> neighbour: Remove bare neighbour::next pointer
> neighbour: Create netdev->neighbour association
>
> .../networking/net_cachelines/net_device.rst | 1 +
> include/linux/netdevice.h | 7 +
> include/net/neighbour.h | 24 +-
> include/net/neighbour_tables.h | 12 +
> net/core/neighbour.c | 337 ++++++++----------
> net/ipv4/arp.c | 2 +-
> 6 files changed, 174 insertions(+), 209 deletions(-)
> create mode 100644 include/net/neighbour_tables.h
Looks like the test is still unhappy. Can you try to run it on your side
before reposting? Or does it look good?
[ 110.442590][ C2] page: refcount:1 mapcount:0 mapping:0000000000000000 index:0x1 pfn:0x5191
[ 110.443219][ C2] head: order:3 mapcount:0 entire_mapcount:0 nr_pages_mapped:0 pincount:0
[ 110.443498][ C2] flags: 0x80000000000040(head|node=0|zone=1)
[ 110.443752][ C2] page_type: f5(slab)
[ 110.443897][ C2] raw: 0080000000000003 ffffea0000146401 ffffffffffffffff 0000000000000000
[ 110.444236][ C2] raw: 0000000000000008 0000000000000000 00000000ffffffff 0000000000000000
[ 110.444546][ C2] head: 0080000000000040 ffff8880010433c0 ffffea0000256410 ffff8880010410e8
[ 110.444862][ C2] head: 0000000000000000 0000000000020002 00000001f5000000 0000000000000000
[ 110.445175][ C2] head: 0080000000000003 ffffea0000146401 ffffffffffffffff 0000000000000000
[ 110.445890][ C2] head: 0000000000000008 0000000000000000 00000000ffffffff 0000000000000000
[ 110.446197][ C2] page dumped because: VM_BUG_ON_PAGE(page_ref_count(page) == 0)
[ 110.446558][ C2] ------------[ cut here ]------------
[ 110.446754][ C2] kernel BUG at include/linux/mm.h:1140!
[ 110.446972][ C2] Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN NOPTI
[ 110.447210][ C2] CPU: 2 UID: 0 PID: 29 Comm: ksoftirqd/2 Not tainted 6.12.0-rc3-virtme #1
[ 110.447528][ C2] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.16.3-0-ga6ed6b701f0a-prebuilt.qemu.org 04/01/2014
[ 110.447928][ C2] RIP: 0010:__free_pages+0x1e4/0x220
[ 110.448128][ C2] Code: 0f 94 c3 e9 ba fe ff ff 48 c7 c6 a0 18 58 92 4c 89 e7 e8 df bf f4 ff 90 0f 0b 48 c7 c6 40 29 58 92 4c 89 e7 e8 cd bf f4 ff 90 <0f> 0b 48 89 ef e8 72 03 09 00 e9 c5 fe ff ff e8 98 03 09 00 e9 35
[ 110.448803][ C2] RSP: 0018:ffffc90000217cb0 EFLAGS: 00010246
[ 110.449040][ C2] RAX: 000000000000003e RBX: 0000000000000000 RCX: 1ffffffff263b43c
[ 110.449304][ C2] RDX: 0000000000000000 RSI: 0000000000000004 RDI: 0000000000000001
[ 110.449565][ C2] RBP: ffffea0000146474 R08: 0000000000000000 R09: fffffbfff263b43c
[ 110.449840][ C2] R10: 0000000000000003 R11: 205d324320202020 R12: ffffea0000146440
[ 110.450101][ C2] R13: ffffc90000217d78 R14: 0000000000000000 R15: 0000000000000008
[ 110.450399][ C2] FS: 0000000000000000(0000) GS:ffff888036100000(0000) knlGS:0000000000000000
[ 110.450787][ C2] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 110.451008][ C2] CR2: 00007f9289e72270 CR3: 000000000a73a005 CR4: 0000000000772ef0
[ 110.451274][ C2] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 110.451564][ C2] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 110.451860][ C2] PKRU: 55555554
[ 110.452015][ C2] Call Trace:
[ 110.452167][ C2] <TASK>
[ 110.452282][ C2] ? die+0x37/0x90
[ 110.452440][ C2] ? do_trap+0x1a3/0x260
[ 110.452589][ C2] ? __free_pages+0x1e4/0x220
[ 110.452786][ C2] ? do_error_trap+0xbe/0x180
[ 110.452970][ C2] ? __free_pages+0x1e4/0x220
[ 110.453152][ C2] ? __free_pages+0x1e4/0x220
[ 110.453342][ C2] ? handle_invalid_op+0x2c/0x40
[ 110.453527][ C2] ? __free_pages+0x1e4/0x220
[ 110.453709][ C2] ? exc_invalid_op+0x30/0x50
[ 110.453934][ C2] ? asm_exc_invalid_op+0x1a/0x20
[ 110.454126][ C2] ? __free_pages+0x1e4/0x220
[ 110.454321][ C2] ? rcu_do_batch+0x34d/0xf20
[ 110.454528][ C2] neigh_hash_free_rcu+0xb7/0xe0
[ 110.454728][ C2] rcu_do_batch+0x34f/0xf20
[ 110.454913][ C2] ? __pfx___lock_release+0x10/0x10
[ 110.455108][ C2] ? __pfx_rcu_do_batch+0x10/0x10
[ 110.455350][ C2] ? lockdep_hardirqs_on_prepare+0x12b/0x410
[ 110.455604][ C2] rcu_core+0x2bd/0x4f0
[ 110.455773][ C2] handle_softirqs+0x1f6/0x5c0
[ 110.455965][ C2] ? __pfx_run_ksoftirqd+0x10/0x10
[ 110.456152][ C2] run_ksoftirqd+0x33/0x60
[ 110.456342][ C2] smpboot_thread_fn+0x306/0x850
[ 110.456533][ C2] ? __pfx_smpboot_thread_fn+0x10/0x10
[ 110.456719][ C2] ? __pfx_smpboot_thread_fn+0x10/0x10
[ 110.456903][ C2] kthread+0x28a/0x350
[ 110.457041][ C2] ? __pfx_kthread+0x10/0x10
[ 110.457369][ C2] ret_from_fork+0x31/0x70
[ 110.457566][ C2] ? __pfx_kthread+0x10/0x10
[ 110.457748][ C2] ret_from_fork_asm+0x1a/0x30
[ 110.457944][ C2] </TASK>
---
pw-bot: cr
next prev parent reply other threads:[~2024-10-22 23:48 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-10-22 13:43 [PATCH net-next v7 0/6] neighbour: Improve neigh_flush_dev performance Gilad Naaman
2024-10-22 13:43 ` [PATCH net-next v7 1/6] neighbour: Add hlist_node to struct neighbour Gilad Naaman
2024-10-22 14:40 ` Eric Dumazet
2024-10-23 5:16 ` Gilad Naaman
2024-10-22 13:43 ` [PATCH net-next v7 2/6] neighbour: Define neigh_for_each_in_bucket Gilad Naaman
2024-10-22 13:43 ` [PATCH net-next v7 3/6] neighbour: Convert seq_file functions to use hlist Gilad Naaman
2024-10-22 13:43 ` [PATCH net-next v7 4/6] neighbour: Convert iteration to use hlist+macro Gilad Naaman
2024-10-22 13:43 ` [PATCH net-next v7 5/6] neighbour: Remove bare neighbour::next pointer Gilad Naaman
2024-10-22 13:43 ` [PATCH net-next v7 6/6] neighbour: Create netdev->neighbour association Gilad Naaman
2024-10-22 23:48 ` Stanislav Fomichev [this message]
2024-10-23 5:01 ` [PATCH net-next v7 0/6] neighbour: Improve neigh_flush_dev performance Gilad Naaman
2024-10-23 5:18 ` Kuniyuki Iwashima
2024-10-23 5:38 ` Gilad Naaman
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=Zxg52Ccb8FF8MCt0@mini-arch \
--to=stfomichev@gmail.com \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=gnaaman@drivenets.com \
--cc=kuba@kernel.org \
--cc=kuniyu@amazon.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.