linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Yunsheng Lin <linyunsheng@huawei.com>
To: <davem@davemloft.net>, <kuba@kernel.org>, <pabeni@redhat.com>
Cc: <somnath.kotur@broadcom.com>, <liuyonglong@huawei.com>,
	<fanghaiqing@huawei.com>, <zhangkun09@huawei.com>,
	Yunsheng Lin <linyunsheng@huawei.com>,
	Alexander Lobakin <aleksander.lobakin@intel.com>,
	Robin Murphy <robin.murphy@arm.com>,
	Alexander Duyck <alexander.duyck@gmail.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	IOMMU <iommu@lists.linux.dev>, MM <linux-mm@kvack.org>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Jesper Dangaard Brouer <hawk@kernel.org>,
	John Fastabend <john.fastabend@gmail.com>,
	Matthias Brugger <matthias.bgg@gmail.com>,
	AngeloGioacchino Del Regno
	<angelogioacchino.delregno@collabora.com>,
	<netdev@vger.kernel.org>, <intel-wired-lan@lists.osuosl.org>,
	<bpf@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
	<linux-arm-kernel@lists.infradead.org>,
	<linux-mediatek@lists.infradead.org>
Subject: [PATCH RFCv5 0/8] fix two bugs related to page_pool
Date: Fri, 13 Dec 2024 20:27:31 +0800	[thread overview]
Message-ID: <20241213122739.4050137-1-linyunsheng@huawei.com> (raw)

This version is mainly to see if using page_pool_item metadata
to keep track of all pages is the correct way to fix the dma
API misuse problem.

Note, it is not based on the latest net-next tree yet, but based
on the below commit in net-next:
commit da4fa00abe56 ("Merge branch 'mitigate-the-two-reallocations-issue-for-iptunnels'")

From the below performance data, the overhead is avoided as much as possible
for time_bench_page_pool01_fast_path() and time_bench_page_pool02_ptr_ring,
and there is about 20ns overhead for time_bench_page_pool03_slow() for fixing
the bug.

Before this patchset:
root@(none)$ insmod bench_page_pool_simple.ko
root@(none)$ insmod bench_page_pool_simple.ko
[   67.667196] bench_page_pool_simple: Loaded
[   67.748321] time_bench: Type:for_loop Per elem: 0 cycles(tsc) 0.769 ns (step:0) - (measurement period time:0.076977910 sec time_interval:76977910) - (invoke count:100000000 tsc_interval:7697783)
[   69.851812] time_bench: Type:atomic_inc Per elem: 2 cycles(tsc) 20.862 ns (step:0) - (measurement period time:2.086207700 sec time_interval:2086207700) - (invoke count:100000000 tsc_interval:208620764)
[   70.019852] time_bench: Type:lock Per elem: 1 cycles(tsc) 15.015 ns (step:0) - (measurement period time:0.150151600 sec time_interval:150151600) - (invoke count:10000000 tsc_interval:15015154)
[   70.691100] time_bench: Type:rcu Per elem: 0 cycles(tsc) 6.541 ns (step:0) - (measurement period time:0.654142450 sec time_interval:654142450) - (invoke count:100000000 tsc_interval:65414239)
[   70.708119] bench_page_pool_simple: time_bench_page_pool01_fast_path(): Cannot use page_pool fast-path
[   70.975262] time_bench: Type:no-softirq-page_pool01 Per elem: 2 cycles(tsc) 25.785 ns (step:0) - (measurement period time:0.257850110 sec time_interval:257850110) - (invoke count:10000000 tsc_interval:25785005)
[   70.993931] bench_page_pool_simple: time_bench_page_pool02_ptr_ring(): Cannot use page_pool fast-path
[   71.575053] time_bench: Type:no-softirq-page_pool02 Per elem: 5 cycles(tsc) 57.191 ns (step:0) - (measurement period time:0.571916900 sec time_interval:571916900) - (invoke count:10000000 tsc_interval:57191684)
[   71.593722] bench_page_pool_simple: time_bench_page_pool03_slow(): Cannot use page_pool fast-path
[   73.384560] time_bench: Type:no-softirq-page_pool03 Per elem: 17 cycles(tsc) 178.197 ns (step:0) - (measurement period time:1.781979820 sec time_interval:1781979820) - (invoke count:10000000 tsc_interval:178197975)
[   73.403581] bench_page_pool_simple: pp_tasklet_handler(): in_serving_softirq fast-path
[   73.411485] bench_page_pool_simple: time_bench_page_pool01_fast_path(): in_serving_softirq fast-path
[   73.678410] time_bench: Type:tasklet_page_pool01_fast_path Per elem: 2 cycles(tsc) 25.780 ns (step:0) - (measurement period time:0.257807630 sec time_interval:257807630) - (invoke count:10000000 tsc_interval:25780758)
[   73.697686] bench_page_pool_simple: time_bench_page_pool02_ptr_ring(): in_serving_softirq fast-path
[   74.242807] time_bench: Type:tasklet_page_pool02_ptr_ring Per elem: 5 cycles(tsc) 53.608 ns (step:0) - (measurement period time:0.536089620 sec time_interval:536089620) - (invoke count:10000000 tsc_interval:53608957)
[   74.261996] bench_page_pool_simple: time_bench_page_pool03_slow(): in_serving_softirq fast-path
[   76.115762] time_bench: Type:tasklet_page_pool03_slow Per elem: 18 cycles(tsc) 184.508 ns (step:0) - (measurement period time:1.845082100 sec time_interval:1845082100) - (invoke count:10000000 tsc_interval:184508203)

After this patchset:
root@(none)$ insmod bench_page_pool_simple.ko
[   88.665991] bench_page_pool_simple: Loaded
[   88.747105] time_bench: Type:for_loop Per elem: 0 cycles(tsc) 0.769 ns (step:0) - (measurement period time:0.076981170 sec time_interval:76981170) - (invoke count:100000000 tsc_interval:7698109)
[   91.585102] time_bench: Type:atomic_inc Per elem: 2 cycles(tsc) 28.206 ns (step:0) - (measurement period time:2.820699360 sec time_interval:2820699360) - (invoke count:100000000 tsc_interval:282069929)
[   91.753048] time_bench: Type:lock Per elem: 1 cycles(tsc) 15.005 ns (step:0) - (measurement period time:0.150057320 sec time_interval:150057320) - (invoke count:10000000 tsc_interval:15005727)
[   92.424306] time_bench: Type:rcu Per elem: 0 cycles(tsc) 6.541 ns (step:0) - (measurement period time:0.654151520 sec time_interval:654151520) - (invoke count:100000000 tsc_interval:65415145)
[   92.441325] bench_page_pool_simple: time_bench_page_pool01_fast_path(): Cannot use page_pool fast-path
[   92.696225] time_bench: Type:no-softirq-page_pool01 Per elem: 2 cycles(tsc) 24.560 ns (step:0) - (measurement period time:0.245607210 sec time_interval:245607210) - (invoke count:10000000 tsc_interval:24560715)
[   92.714893] bench_page_pool_simple: time_bench_page_pool02_ptr_ring(): Cannot use page_pool fast-path
[   93.336550] time_bench: Type:no-softirq-page_pool02 Per elem: 6 cycles(tsc) 61.245 ns (step:0) - (measurement period time:0.612451380 sec time_interval:612451380) - (invoke count:10000000 tsc_interval:61245127)
[   93.355219] bench_page_pool_simple: time_bench_page_pool03_slow(): Cannot use page_pool fast-path
[   95.402370] time_bench: Type:no-softirq-page_pool03 Per elem: 20 cycles(tsc) 203.828 ns (step:0) - (measurement period time:2.038286740 sec time_interval:2038286740) - (invoke count:10000000 tsc_interval:203828660)
[   95.421395] bench_page_pool_simple: pp_tasklet_handler(): in_serving_softirq fast-path
[   95.429301] bench_page_pool_simple: time_bench_page_pool01_fast_path(): in_serving_softirq fast-path
[   95.684025] time_bench: Type:tasklet_page_pool01_fast_path Per elem: 2 cycles(tsc) 24.560 ns (step:0) - (measurement period time:0.245605490 sec time_interval:245605490) - (invoke count:10000000 tsc_interval:24560544)
[   95.703301] bench_page_pool_simple: time_bench_page_pool02_ptr_ring(): in_serving_softirq fast-path
[   96.401877] time_bench: Type:tasklet_page_pool02_ptr_ring Per elem: 6 cycles(tsc) 68.954 ns (step:0) - (measurement period time:0.689544160 sec time_interval:689544160) - (invoke count:10000000 tsc_interval:68954410)
[   96.421065] bench_page_pool_simple: time_bench_page_pool03_slow(): in_serving_softirq fast-path
[   98.496283] time_bench: Type:tasklet_page_pool03_slow Per elem: 20 cycles(tsc) 206.653 ns (step:0) - (measurement period time:2.066533210 sec time_interval:2066533210) - (invoke count:10000000 tsc_interval:206653316)

1. https://lore.kernel.org/lkml/8067f204-1380-4d37-8ffd-007fc6f26738@kernel.org/T/

CC: Alexander Lobakin <aleksander.lobakin@intel.com>
CC: Robin Murphy <robin.murphy@arm.com>
CC: Alexander Duyck <alexander.duyck@gmail.com>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: IOMMU <iommu@lists.linux.dev>
CC: MM <linux-mm@kvack.org>

Change log:
V5:
  1. Support unlimit inflight pages.
  2. Add some optimization to avoid the overhead of fixing bug.

V4:
  1. use scanning to do the unmapping
  2. spilt dma sync skipping into separate patch

V3:
  1. Target net-next tree instead of net tree.
  2. Narrow the rcu lock as the discussion in v2.
  3. Check the ummapping cnt against the inflight cnt.

V2:
  1. Add a item_full stat.
  2. Use container_of() for page_pool_to_pp().

Yunsheng Lin (8):
  page_pool: introduce page_pool_to_pp() API
  page_pool: fix timing for checking and disabling napi_local
  page_pool: fix IOMMU crash when driver has already unbound
  page_pool: support unlimited number of inflight pages
  page_pool: skip dma sync operation for inflight pages
  page_pool: use list instead of ptr_ring for ring cache
  page_pool: batch refilling pages to reduce atomic operation
  page_pool: use list instead of array for alloc cache

 drivers/net/ethernet/freescale/fec_main.c     |   8 +-
 .../ethernet/google/gve/gve_buffer_mgmt_dqo.c |   2 +-
 drivers/net/ethernet/intel/iavf/iavf_txrx.c   |   6 +-
 drivers/net/ethernet/intel/idpf/idpf_txrx.c   |  14 +-
 drivers/net/ethernet/intel/libeth/rx.c        |   2 +-
 .../net/ethernet/mellanox/mlx5/core/en/xdp.c  |   3 +-
 drivers/net/netdevsim/netdev.c                |   6 +-
 drivers/net/wireless/mediatek/mt76/mt76.h     |   2 +-
 include/linux/mm_types.h                      |   2 +-
 include/linux/skbuff.h                        |   1 +
 include/net/libeth/rx.h                       |   3 +-
 include/net/netmem.h                          |  10 +-
 include/net/page_pool/helpers.h               |  11 +
 include/net/page_pool/types.h                 |  57 +-
 net/core/devmem.c                             |   4 +-
 net/core/netmem_priv.h                        |   5 +-
 net/core/page_pool.c                          | 660 ++++++++++++++----
 net/core/page_pool_priv.h                     |  12 +-
 net/core/skbuff.c                             |   3 +-
 net/core/xdp.c                                |   3 +-
 20 files changed, 650 insertions(+), 164 deletions(-)

-- 
2.33.0



             reply	other threads:[~2024-12-13 12:34 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-12-13 12:27 Yunsheng Lin [this message]
2024-12-13 12:27 ` [PATCH RFCv5 3/8] page_pool: fix IOMMU crash when driver has already unbound Yunsheng Lin

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=20241213122739.4050137-1-linyunsheng@huawei.com \
    --to=linyunsheng@huawei.com \
    --cc=akpm@linux-foundation.org \
    --cc=aleksander.lobakin@intel.com \
    --cc=alexander.duyck@gmail.com \
    --cc=angelogioacchino.delregno@collabora.com \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=davem@davemloft.net \
    --cc=fanghaiqing@huawei.com \
    --cc=hawk@kernel.org \
    --cc=intel-wired-lan@lists.osuosl.org \
    --cc=iommu@lists.linux.dev \
    --cc=john.fastabend@gmail.com \
    --cc=kuba@kernel.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mediatek@lists.infradead.org \
    --cc=linux-mm@kvack.org \
    --cc=liuyonglong@huawei.com \
    --cc=matthias.bgg@gmail.com \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=robin.murphy@arm.com \
    --cc=somnath.kotur@broadcom.com \
    --cc=zhangkun09@huawei.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox