* [ANN] bpf development stats for 7.4
@ 2026-10-23 22:49 Stanislav Fomichev
0 siblings, 0 replies; only message in thread
From: Stanislav Fomichev @ 2026-10-23 22:49 UTC (permalink / raw)
To: bpf; +Cc: kuba
Regular development statistics update, similar to Jakub's netdev post:
https://lore.kernel.org/netdev/20261022135723.07412855@kernel.org/
This cycle, because Jakub added a few new stats, I used this as an opportunity
to run my own generation scripts through an LLM, which obviously found
a few bugs: the git commit ranges have been off since at least 6.14, and
since the MAINTAINERS update for the general BPF entry, the commits
applied by the new maintainers were left out. See "Corrected history"
below for the details and the fixed numbers.
Previous cycle:
15 Jun to 17 Aug: 8618 mailing list messages, 63 days, 137 messages per day
433 repo commits (7 commits/day, 36.26% of these are selftests)
https://lore.kernel.org/bpf/aoen96RONJgf18E5@devvm7509.cco0.facebook.com/
Current cycle:
17 Aug to 21 Oct: 12765 mailing list messages, 65 days, 196 messages per day
815 repo commits (13 commits/day, 41.6% of these are selftests)
Highlights
----------
This part is written by an LLM, I'm extending my workflow to compare
the previous and the current cycle to find interesting tidbits:
- Review coverage dropped sharply: only 26% of the commits carry
a Reviewed-by / Acked-by (64% in 7.3, 44-64% in every cycle since
6.14), 7% ahead of the maintainer's sign-off. Maintainers wrote more
of the commits themselves (44% vs 24%), but patches from everyone
else are down too, from 67% to 33% with a review tag.
- Everything grew: 44% more messages per day, 76% more patch postings,
almost twice as many commits (815 vs 433). Bots sent 25% of the
messages (20% in 7.3), netdev's AI reviewer now also replies to the
patches cross-posted to netdev (301 replies).
- 70 authors had no earlier commit in the kernel, 27% of all authors
(46 and 22% in 7.3), and 57 people contributed selftests (43).
- Alexei is back as the top reviewer (365 review messages, 61 in 7.3),
and more reviews come from the neighboring subsystems, e.g. David
Hildenbrand (18 -> 85) and Tejun Heo (17 -> 68).
Testing
-------
Top contributors of selftests:
1 [ 53] Kumar Kartikeya Dwivedi
2 [ 44] Eduard Zingerman
3 [ 32] Alexei Starovoitov
4 [ 24] Daniel Borkmann
5 [ 20] Yonghong Song
6 [ 15] Amery Hung
7 [ 13] Emil Tsalapatis
8 [ 11] Qiliang Yuan
9 [ 11] Ihor Solodrai
10 [ 10] Jiayuan Chen
Developer rankings
------------------
Top reviewers (cs): Top reviewers (msg):
1 ( +4) [19] Alexei Starovoitov 1 ( +4) [41] Alexei Starovoitov
2 ( +2) [ 5] Andrii Nakryiko 2 ( -1) [15] Eduard Zingerman
3 ( ) [ 4] Kumar Kartikeya Dwivedi 3 ( -1) [11] Andrii Nakryiko
4 ( -3) [ 4] Eduard Zingerman 4 (+23) [ 9] David Hildenbrand (Arm)
5 ( +1) [ 4] Jakub Kicinski 5 ( -1) [ 8] Kumar Kartikeya Dwivedi
6 ( -4) [ 3] Emil Tsalapatis 6 (+22) [ 8] Tejun Heo
7 ( +5) [ 3] Tejun Heo 7 ( +1) [ 7] Jakub Kicinski
8 ( +7) [ 2] Amery Hung 8 ( +6) [ 7] Amery Hung
9 (+14) [ 2] Quentin Monnet 9 ( +1) [ 6] Stanislav Fomichev
10 ( -1) [ 2] Jiri Olsa 10 (+28) [ 5] Quentin Monnet
11 ( -1) [ 2] Ihor Solodrai 11 ( -8) [ 5] Emil Tsalapatis
12 ( -4) [ 2] Daniel Borkmann 12 ( -1) [ 4] Jiri Olsa
13 (+19) [ 2] Simon Horman 13 (+39) [ 4] Simon Horman
14 ( -3) [ 2] Jiayuan Chen 14 (+15) [ 4] Björn Töpel
15 (+20) [ 1] Mykyta Yatsenko 15 ( -8) [ 4] Leon Hwang
Top authors (cs): Top authors (msg):
1 ( ) [3] Kumar Kartikeya Dwivedi 1 ( ) [49] Kumar Kartikeya Dwivedi
2 (***) [2] Alexei Starovoitov 2 ( +1) [48] Yonghong Song
3 (***) [2] Qiliang Yuan 3 (+41) [27] Lorenzo Stoakes (ARM)
4 ( +3) [2] Jiayuan Chen 4 (***) [26] Alexei Starovoitov
5 (***) [2] Eduard Zingerman 5 (+13) [22] Eduard Zingerman
6 ( +8) [2] Yonghong Song 6 ( -1) [15] Daniel Borkmann
7 ( -1) [2] Daniel Borkmann 7 (+12) [13] Emil Tsalapatis
8 (***) [1] Karl Mehltretter 8 (***) [12] Josef Bacik
9 (+10) [1] <chenyuan_fl@163.com> 9 (***) [11] Alan Maguire
10 (***) [1] Lorenzo Bianconi 10 ( -6) [11] Amery Hung
Top scores (positive): Top scores (negative):
1 ( +3) [210] Alexei Starovoitov 1 ( ) [191] Yonghong Song
2 ( ) [ 79] Andrii Nakryiko 2 (+20) [127] Kumar Kartikeya Dwivedi
3 ( +2) [ 54] Jakub Kicinski 3 (***) [101] Lorenzo Stoakes (ARM)
4 ( +7) [ 50] Tejun Heo 4 (***) [ 49] Josef Bacik
5 (+17) [ 44] Quentin Monnet 5 (***) [ 37] Tao Cui
6 (+15) [ 35] David Hildenbrand (Arm) 6 ( +5) [ 34] Daniel Borkmann
7 ( +1) [ 29] Stanislav Fomichev 7 ( +7) [ 29] Harshitha Ramamurthy
Company rankings
----------------
Top reviewers (cs): Top reviewers (msg):
1 ( ) [23] Meta 1 ( ) [63] Meta
2 ( +3) [19] Anthropic 2 ( +4) [39] Anthropic
3 ( +3) [ 5] RedHat 3 ( +2) [19] RedHat
4 ( -2) [ 4] Isovalent 4 ( -2) [10] Google
5 ( -2) [ 3] Google 5 ( -2) [ 8] Isovalent
6 ( +1) [ 3] Intel 6 ( +3) [ 7] Intel
7 ( -3) [ 3] Shopee 7 ( -3) [ 7] Shopee
Top authors (cs): Top authors (msg):
1 ( ) [11] Meta 1 ( ) [164] Meta
2 (***) [ 3] Anthropic 2 (***) [ 40] Anthropic
3 ( +3) [ 3] Isovalent 3 (+13) [ 32] ARM
4 ( +1) [ 2] Google 4 ( -2) [ 26] Isovalent
5 ( -2) [ 2] Shopee 5 ( +2) [ 26] Google
6 (***) [ 2] Qiliang Yuan 6 ( +2) [ 17] Microsoft
7 ( +1) [ 2] Microsoft 7 ( +4) [ 15] Oracle
Top scores (positive): Top scores (negative):
1 ( +1) [142] Anthropic 1 (***) [249] Meta
2 ( +4) [ 49] RedHat 2 (+35) [114] ARM
3 ( +7) [ 44] Hedgehog 3 ( -2) [ 48] Kylin Software
4 ( +3) [ 24] nVidia 4 (***) [ 44] Google
5 (+10) [ 20] Eric Dumazet 5 ( -1) [ 40] Amazon
6 ( +3) [ 16] Linux Foundation 6 ( -1) [ 40] Microsoft
7 ( -3) [ 13] SUSE 7 ( -4) [ 38] Isovalent
Posting efficiency
------------------
New this time: how many patch postings it takes to get a commit merged.
4748 postings turned into 1011 commits this cycle, 4.70 postings per
commit (2.23 counting only the postings which got merged). Companies
best / worst at getting code into the tree:
Lowest postings/commit (min 10): Highest postings/commit (min 10):
1 [123@ 2.72] Anthropic 1 [ 14@ 8.29] RedHat
2 [ 78@ 2.74] Isovalent 2 [ 34@ 8.03] ARM
3 [ 49@ 2.88] Microsoft 3 [ 17@ 5.18] Kylin Software
4 [ 10@ 3.70] <chenyuan_fl@163.com> 4 [ 24@ 5.12] Oracle
5 [ 11@ 3.73] Qiliang Yuan 5 [ 13@ 4.92] Huawei
6 [362@ 3.73] Meta 6 [ 24@ 4.71] Shopee
7 [ 15@ 3.87] Microchip 7 [ 50@ 4.14] Google
^ ^
number of postings per
commits 1 commit
Nothing merged, people (min 30): Nothing merged, companies (min 30):
1 (***) [ 63] Justin Suess 1 (***) [ 63] Justin Suess
2 (***) [ 37] Jay Wang
3 (***) [ 35] Gabriele Monaco
4 (***) [ 33] Andrey Grodzovsky
5 ( -1) [ 32] Ankur Arora
AI review replies
-----------------
Also new: share of each author's postings the review bots replied to
(replies to cover letters count too, so it can go over 100%).
Replies from sashiko-bot@kernel.org: 1427
Per posting, people (min 10): Per posting, companies (min 30):
1 [100% of 27] Ricardo Robaina 1 [81% of 273] ARM
2 [100% of 26] Mauricio Faria de Oliveira 2 [70% of 141] Microsoft
3 [100% of 10] Mickaël Salaün 3 [50% of 116] RedHat
4 [93% of 14] Matt Turner 4 [47% of 34] Crowdstrike
5 [90% of 20] Ratheesh Kannoth 5 [42% of 67] Tao Cui
6 [84% of 237] Lorenzo Stoakes (ARM) 6 [42% of 84] Amazon
7 [83% of 12] Jason Xing 7 [39% of 41] Vineet Gupta
8 [74% of 19] Kees Cook 8 [38% of 88] Kylin Software
9 [73% of 26] Wei Hu 9 [38% of 40] Qualcomm
10 [72% of 18] Zhen Ni 10 [35% of 37] <chenyuan_fl@163.com>
Replies from bot+bpf-ci@kernel.org: 1041
Per posting, people (min 10): Per posting, companies (min 30):
1 [93% of 14] Fuyu Zhao 1 [59% of 34] Crowdstrike
2 [75% of 16] Xu Kuohai 2 [54% of 37] <chenyuan_fl@163.com>
3 [61% of 33] Andrey Grodzovsky 3 [46% of 39] Tianyi Chen
4 [58% of 12] Florent Revest 4 [37% of 41] Vineet Gupta
5 [56% of 16] Ziyang Men 5 [35% of 214] Isovalent
6 [54% of 37] <chenyuan_fl@163.com> 6 [35% of 123] Oracle
7 [54% of 28] Hui Zhu 7 [35% of 113] Shopee
8 [50% of 14] Mykyta Yatsenko 8 [33% of 64] Huawei
9 [50% of 12] Mingpei CAO 9 [27% of 84] Amazon
10 [50% of 10] Thiébaud Weksteen 10 [27% of 63] Justin Suess
Corrected history
-----------------
What was wrong with the previous reports:
- the git stats used wrong commit ranges: the cutoffs were picked by
date in the bpf-next history, which has mainline merged in, so they
landed on random commits from other trees (realistically, was not
affecting the bottom line); they now cover exactly the work between
the bpf-next pull requests to Linus,
- commits applied by the new maintainers were left out of the counts,
- the linux-next size used to normalize the 7.2 and 7.3 trends was
slightly off,
- bots and stable backports weren't fully filtered out of the mailing
list stats (e.g. kernel test robot replies counted as Intel reviews).
7.2 and 7.3 were regenerated completely, for the older cycles only the
git based numbers below were recomputed:
Release commits selftests reviewed x-company any review
(posted) (posted) tag
6.14 222 (378) 45.0% 32.0% (23.5%) 25.7% 52.7%
6.15 205 (337) 51.7% 21.5% (25.5%) 20.5% 52.2%
6.16 183 (183) 34.4% 32.2% (23.4%) 31.1% 52.5%
6.17 206 (376) 41.7% 37.4% (31.9%) 28.2% 55.8%
6.18 227 (437) 37.0% 24.7% (30.7%) 19.8% 49.8%
6.19 218 (415) 44.5% 46.3% (36.4%) 33.9% 59.2%
7.0 309 (205) 39.2% 37.2% (42.4%) 34.3% 52.1%
7.1 320 (587) 50.6% 48.1% (45.3%) 41.6% 55.9%
7.2 398 (301) 42.0% 36.4% (44.9%) 31.7% 44.0%
7.3 433 (377) 36.3% 36.7% (36.9%) 27.7% 64.4%
7.4 815 41.6% 7.0% 5.4% 25.5%
"reviewed" is the share of commits with a review tag ahead of the
maintainer's sign-off, "any review tag" counts a tag anywhere in the
commit message.
With the fixes, some of the 7.3 trends turned out to be artifacts:
commits per day grew 9% (not 25%), review coverage stayed flat at 36.7%
(rather than dropping 8%), and patch sets needed 12% fewer revisions
(not 31% more).
More raw stats
--------------
Prev: start: Mon, 15 Jun 2026 11:20:10 -0700
end: Mon, 17 Aug 2026 23:26:22 +0200
Prev: messages: 8618 days: 63 (137 msg/day)
Prev: bot messages: 1719 (19.95%)
Prev: direct commits: 433 (7 commits/day)
Prev: test commits: 157 (36.26%)
Prev: people: 338 {'author': 153, 'commenter': 128, 'both': 57, 'none': 9}
Prev: people pct: {'author': 0.441, 'commenter': 0.369, 'both': 0.164, 'none': 0.026}
Prev: review pct: 36.72% x-corp pct: 27.71%
Prev: avg revisions: single patch: 1.65 patch set: 2.98
Prev: postings per merged commit: 4.82 (merged postings only: 2.20)
Prev: postings: 2691 merged: 1230 commits: 558
Prev: postings with Fixes: 677 (25.16%) merged: 357 (29.02%)
Curr: start: Mon, 17 Aug 2026 16:27:29 -0500
end: Wed, 21 Oct 2026 21:01:45 +0000
Curr: messages: 12765 days: 65 (196 msg/day)
Curr: bot messages: 3124 (24.47%)
Curr: direct commits: 815 (13 commits/day)
Curr: test commits: 339 (41.6%)
Curr: people: 417 {'author': 177, 'commenter': 155, 'both': 85, 'none': 12}
Curr: people pct: {'author': 0.413, 'commenter': 0.361, 'both': 0.198, 'none': 0.028}
Curr: review pct: 6.99% x-corp pct: 5.4%
Curr: avg revisions: single patch: 1.82 patch set: 3.09
Curr: postings per merged commit: 4.70 (merged postings only: 2.23)
Curr: postings: 4748 merged: 2258 commits: 1011
Curr: postings with Fixes: 1025 (21.59%) merged: 597 (26.44%)
Diff: +16.3% linux-next size
Diff: +43.6% (+23.4%) msg/day
Diff: +82.4% (+56.8%) commits/day
Diff: +19.6% ( +2.8%) people/day
Diff: review pct: -29.7%
x-corp pct: -22.3%
Diff: avg revisions: single patch: +10.8% patch set: +3.7%
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2026-10-23 22:49 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-10-23 22:49 [ANN] bpf development stats for 7.4 Stanislav Fomichev
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox