netdev development stats for 6.9
Subsystems (2)
Jakub Kicinski
maintainerIntro
-----
We have posted our pull requests with networking changes for the 6.9 kernel release last night. As is tradition here are the development statistics based on mailing list traffic on netdev@vger.
These stats are somewhat like LWN stats: https://lwn.net/Articles/956765/
but more focused on review participation.
Previous stats (for 6.8): https://lore.kernel.org/all/[email protected]/
Testing
-------
The selftest runner went live early during this release.
I don't have any great ideas on how to quantify the progress but, first, I'd like to thank everyone who contributed to improving and fixing the tests, and those who contributed to NIPA directly. We are reporting results from 243 tests to patchwork (not counting kunit tests), each test on two kernel flavors and with many sub-cases.
There's still work to be done fixing some of the tests but we made more progress than I anticipated! There are 66 test + kernel config combinations which we currently ignore, either because they permanently skip / fail or are too flaky. I should mention that those tests do matter, recently the ignored pmtu test would have caught a last minute xfrm regression :(
Speaking of catching regressions, I do not have an objective count but subjectively our tests do in fact catch bugs. In fact the signal to noise ratio is higher than I initially expected. I wish it was easier to write driver tests and use YNL in the tests, but we'll get there..
General stats
-------------
Cycle started on Tue, 09 Jan, ended Mon, 11 Mar. That's 63 days, one week shorter than the previous cycle.
We have seen 264 msg/day on the list, and 24 commits/day. The number of messages is back to the level we have seen in 6.6, and has recovered after couple of slower releases (last one being particularly slow due to the winter break). The number of commits applied directly by netdev maintainers is 9% higher than in 6.6 and highest on the record.
The total number of review tags in git history have dipped again, 61% of commits contain a review tag (down from 69%). The number of commits with a review tag from an email domain different than the author, however, dropped only by 1% to 53%. Similarly our statistic of series which received at least one ack / review tag on the list did not change much, increasing by 1% to 67%.
Rankings
--------
As promised last time, the left side of the stats is now in "change sets" (cs) rather than threads, IOW trying to track multiple revisions of the same series as one.
Top reviewers (cs): Top reviewers (msg):
1 ( +1) [30] Jakub Kicinski 1 ( ) [66] Jakub Kicinski 2 ( -1) [26] Simon Horman 2 ( ) [45] Simon Horman 3 ( ) [14] Andrew Lunn 3 ( ) [44] Andrew Lunn 4 ( +1) [12] Paolo Abeni 4 ( +3) [25] Jiri Pirko 5 ( +4) [11] Jiri Pirko 5 ( ) [23] Eric Dumazet 6 ( -2) [10] Eric Dumazet 6 ( +3) [20] Paolo Abeni 7 ( ) [ 5] David Ahern 7 ( +1) [15] Krzysztof Kozlowski 8 ( +6) [ 4] Stephen Hemminger 8 ( +2) [11] David Ahern 9 ( +2) [ 4] Willem de Bruijn 9 ( +4) [10] Florian Fainelli 10 ( ) [ 4] Krzysztof Kozlowski 10 ( -6) [10] Vladimir Oltean 11 ( +1) [ 4] Florian Fainelli 11 ( -5) [ 9] Russell King 12 ( -6) [ 4] Russell King 12 ( ) [ 9] Willem de Bruijn 13 ( +3) [ 2] Jacob Keller 13 ( -2) [ 8] Sergey Shtylyov 14 ( +1) [ 2] Rob Herring 14 ( +6) [ 7] Jacob Keller 15 ( +2) [ 2] Jamal Hadi Salim 15 (+16) [ 7] Jason Wang
Jiri jumps into top 5, reviewing various driver and netlink changes. Stephen largely works on iproute2 reviews. Russell and Vladimir have indicated that they are occupied outside of netdev, and slip by a few positions.
Thank you all for your work!
Top authors (cs): Top authors (msg):
1 ( +1) [7] Eric Dumazet 1 (+35) [40] Eric Dumazet 2 ( -1) [7] Jakub Kicinski 2 ( +5) [26] Jakub Kicinski 3 (+46) [4] Breno Leitao 3 ( +1) [20] Saeed Mahameed 4 ( ) [3] Heiner Kallweit 4 (+18) [17] Xuan Zhuo 5 ( +8) [2] Stephen Hemminger 5 (***) [15] Jason Xing 6 (+24) [2] Paolo Abeni 6 (***) [15] Matthieu Baerts 7 (+11) [2] Kuniyuki Iwashima 7 (***) [14] Breno Leitao 8 (***) [2] Maks Mishin 8 ( -3) [14] Tony Nguyen 9 ( +6) [2] Kunwu Chan 9 ( -3) [13] Kuniyuki Iwashima
10 (***) [2] Matthieu Baerts 10 ( -8) [11] Christian Marangi
Thanks to switching from threads to changes sets we see Eric rightfully claim the #1 spot, previously occupied by yours truly. Breno added missing MODULE_DESCRIPTION()s and refactored drivers using per-cpu stats. Paolo contributed a number of MPTCP changes and many selftest improvements. Maks Mishin sent a lot of individual iproute2 patches. Xuan Zhuo is working on vhost / virtio changes for AF_XDP.
Top scores (positive): Top scores (negative):
1 ( +1) [398] Jakub Kicinski 1 ( +2) [80] Saeed Mahameed 2 ( -1) [371] Simon Horman 2 (+17) [60] Xuan Zhuo 3 ( ) [243] Andrew Lunn 3 (***) [46] Jason Xing 4 ( +8) [160] Jiri Pirko 4 (***) [43] Yang Xiwen via B4 Relay 5 ( +2) [152] Paolo Abeni 5 (***) [42] Matthieu Baerts 6 ( +2) [ 78] Krzysztof Kozlowski 6 ( -2) [40] Tony Nguyen 7 ( +2) [ 76] David Ahern 7 ( -6) [38] David Howells 8 ( +2) [ 68] Willem de Bruijn 8 ( -6) [37] Christian Marangi 9 ( +2) [ 60] Florian Fainelli 9 (+30) [37] Kory Maincent
10 ( -5) [ 54] Russell King 10 (***) [36] Breno Leitao
Corporate stats
---------------
Top reviewers (cs): Top reviewers (msg):
1 ( ) [42] RedHat 1 ( ) [98] RedHat 2 ( ) [32] Meta 2 ( ) [74] Meta 3 ( +2) [16] Google 3 ( ) [44] Andrew Lunn 4 ( ) [14] Andrew Lunn 4 ( +1) [43] Google 5 ( -2) [13] Intel 5 ( -1) [36] Intel 6 ( +1) [13] nVidia 6 ( +2) [31] Linaro 7 ( +1) [ 8] Linaro 7 ( ) [31] nVidia
Top authors (cs): Top authors (msg):
1 ( +1) [12] RedHat 1 ( ) [69] Intel 2 ( +2) [12] Meta 2 ( +2) [60] Meta 3 ( -2) [10] Intel 3 ( +2) [55] Google 4 ( -1) [10] Google 4 ( -1) [50] nVidia 5 ( ) [ 7] nVidia 5 ( -3) [47] RedHat 6 ( +2) [ 3] Huawei 6 ( +4) [32] Bootlin 7 ( +7) [ 3] Wirecard 7 ( -1) [23] Alibaba
Top scores (positive): Top scores (negative):
1 ( ) [496] RedHat 1 ( +7) [89] Bootlin 2 ( ) [303] Meta 2 ( ) [79] Alibaba 3 ( ) [243] Andrew Lunn 3 (***) [46] Tencent 4 ( +2) [138] Linaro 4 (***) [43] Yang Xiwen 5 ( ) [ 79] Google 5 (***) [42] Tessares 6 ( +2) [ 76] Enfabrica 6 ( +6) [38] AMD 7 ( -3) [ 60] Oracle 7 ( -6) [37] Christian Marangi
RedHat remains unbeatable with the combined powers of Simon and Paolo, as well as high participation of the less active authors. Alibaba maintains its strongly net-negative review score. Tencent (Jason Xing) joins them (Tencent doesn't use their email domain so it's likely under-counted). Yang Xiwen makes the negative list as well, cost of reposting large series too often...
Signature · 2 lines
Code: https://github.com/kuba-moo/ml-stat Raw output: https://netdev.bots.linux.dev/static/nipa/stats-6.9/stdout
Quoted 147 lines · Jakub Kicinski
Intro
-----We have posted our pull requests with networking changes for the 6.9 kernel release last night. As is tradition here are the development statistics based on mailing list traffic on netdev@vger.
These stats are somewhat like LWN stats: https://lwn.net/Articles/956765/
but more focused on review participation.Previous stats (for 6.8): https://lore.kernel.org/all/[email protected]/
Testing
-------The selftest runner went live early during this release.
I don't have any great ideas on how to quantify the progress but, first, I'd like to thank everyone who contributed to improving and fixing the tests, and those who contributed to NIPA directly. We are reporting results from 243 tests to patchwork (not counting kunit tests), each test on two kernel flavors and with many sub-cases.There's still work to be done fixing some of the tests but we made more progress than I anticipated! There are 66 test + kernel config combinations which we currently ignore, either because they permanently skip / fail or are too flaky. I should mention that those tests do matter, recently the ignored pmtu test would have caught a last minute xfrm regression :(
Speaking of catching regressions, I do not have an objective count but subjectively our tests do in fact catch bugs. In fact the signal to noise ratio is higher than I initially expected. I wish it was easier to write driver tests and use YNL in the tests, but we'll get there..
General stats
-------------Cycle started on Tue, 09 Jan, ended Mon, 11 Mar. That's 63 days, one week shorter than the previous cycle.
We have seen 264 msg/day on the list, and 24 commits/day. The number of messages is back to the level we have seen in 6.6, and has recovered after couple of slower releases (last one being particularly slow due to the winter break). The number of commits applied directly by netdev maintainers is 9% higher than in 6.6 and highest on the record.
The total number of review tags in git history have dipped again, 61% of commits contain a review tag (down from 69%). The number of commits with a review tag from an email domain different than the author, however, dropped only by 1% to 53%. Similarly our statistic of series which received at least one ack / review tag on the list did not change much, increasing by 1% to 67%.
Rankings
--------As promised last time, the left side of the stats is now in "change sets" (cs) rather than threads, IOW trying to track multiple revisions of the same series as one.
Top reviewers (cs): Top reviewers (msg):
1 ( +1) [30] Jakub Kicinski 1 ( ) [66] Jakub Kicinski 2 ( -1) [26] Simon Horman 2 ( ) [45] Simon Horman 3 ( ) [14] Andrew Lunn 3 ( ) [44] Andrew Lunn 4 ( +1) [12] Paolo Abeni 4 ( +3) [25] Jiri Pirko 5 ( +4) [11] Jiri Pirko 5 ( ) [23] Eric Dumazet 6 ( -2) [10] Eric Dumazet 6 ( +3) [20] Paolo Abeni 7 ( ) [ 5] David Ahern 7 ( +1) [15] Krzysztof Kozlowski 8 ( +6) [ 4] Stephen Hemminger 8 ( +2) [11] David Ahern 9 ( +2) [ 4] Willem de Bruijn 9 ( +4) [10] Florian Fainelli 10 ( ) [ 4] Krzysztof Kozlowski 10 ( -6) [10] Vladimir Oltean 11 ( +1) [ 4] Florian Fainelli 11 ( -5) [ 9] Russell King 12 ( -6) [ 4] Russell King 12 ( ) [ 9] Willem de Bruijn 13 ( +3) [ 2] Jacob Keller 13 ( -2) [ 8] Sergey Shtylyov 14 ( +1) [ 2] Rob Herring 14 ( +6) [ 7] Jacob Keller 15 ( +2) [ 2] Jamal Hadi Salim 15 (+16) [ 7] Jason WangJiri jumps into top 5, reviewing various driver and netlink changes. Stephen largely works on iproute2 reviews. Russell and Vladimir have indicated that they are occupied outside of netdev, and slip by a few positions.
Thank you all for your work!
Top authors (cs): Top authors (msg):
1 ( +1) [7] Eric Dumazet 1 (+35) [40] Eric Dumazet 2 ( -1) [7] Jakub Kicinski 2 ( +5) [26] Jakub Kicinski 3 (+46) [4] Breno Leitao 3 ( +1) [20] Saeed Mahameed 4 ( ) [3] Heiner Kallweit 4 (+18) [17] Xuan Zhuo 5 ( +8) [2] Stephen Hemminger 5 (***) [15] Jason Xing 6 (+24) [2] Paolo Abeni 6 (***) [15] Matthieu Baerts 7 (+11) [2] Kuniyuki Iwashima 7 (***) [14] Breno Leitao 8 (***) [2] Maks Mishin 8 ( -3) [14] Tony Nguyen 9 ( +6) [2] Kunwu Chan 9 ( -3) [13] Kuniyuki Iwashima10 (***) [2] Matthieu Baerts 10 ( -8) [11] Christian Marangi
Thanks to switching from threads to changes sets we see Eric rightfully claim the #1 spot, previously occupied by yours truly. Breno added missing MODULE_DESCRIPTION()s and refactored drivers using per-cpu stats. Paolo contributed a number of MPTCP changes and many selftest improvements. Maks Mishin sent a lot of individual iproute2 patches. Xuan Zhuo is working on vhost / virtio changes for
AF_XDP.Top scores (positive): Top scores (negative):
1 ( +1) [398] Jakub Kicinski 1 ( +2) [80] Saeed Mahameed 2 ( -1) [371] Simon Horman 2 (+17) [60] Xuan Zhuo 3 ( ) [243] Andrew Lunn 3 (***) [46] Jason Xing 4 ( +8) [160] Jiri Pirko 4 (***) [43] Yang Xiwen via B4 Relay 5 ( +2) [152] Paolo Abeni 5 (***) [42] Matthieu Baerts 6 ( +2) [ 78] Krzysztof Kozlowski 6 ( -2) [40] Tony Nguyen 7 ( +2) [ 76] David Ahern 7 ( -6) [38] David Howells 8 ( +2) [ 68] Willem de Bruijn 8 ( -6) [37] Christian Marangi 9 ( +2) [ 60] Florian Fainelli 9 (+30) [37] Kory Maincent10 ( -5) [ 54] Russell King 10 (***) [36] Breno Leitao
Corporate stats
---------------Top reviewers (cs): Top reviewers (msg):
1 ( ) [42] RedHat 1 ( ) [98] RedHat 2 ( ) [32] Meta 2 ( ) [74] Meta 3 ( +2) [16] Google 3 ( ) [44] Andrew Lunn 4 ( ) [14] Andrew Lunn 4 ( +1) [43] Google 5 ( -2) [13] Intel 5 ( -1) [36] Intel 6 ( +1) [13] nVidia 6 ( +2) [31] Linaro 7 ( +1) [ 8] Linaro 7 ( ) [31] nVidiaTop authors (cs): Top authors (msg):
1 ( +1) [12] RedHat 1 ( ) [69] Intel 2 ( +2) [12] Meta 2 ( +2) [60] Meta 3 ( -2) [10] Intel 3 ( +2) [55] Google 4 ( -1) [10] Google 4 ( -1) [50] nVidia 5 ( ) [ 7] nVidia 5 ( -3) [47] RedHat 6 ( +2) [ 3] Huawei 6 ( +4) [32] Bootlin 7 ( +7) [ 3] Wirecard 7 ( -1) [23] AlibabaTop scores (positive): Top scores (negative):
1 ( ) [496] RedHat 1 ( +7) [89] Bootlin 2 ( ) [303] Meta 2 ( ) [79] Alibaba 3 ( ) [243] Andrew Lunn 3 (***) [46] Tencent 4 ( +2) [138] Linaro 4 (***) [43] Yang Xiwen 5 ( ) [ 79] Google 5 (***) [42] Tessares 6 ( +2) [ 76] Enfabrica 6 ( +6) [38] AMD 7 ( -3) [ 60] Oracle 7 ( -6) [37] Christian Marangi
Quoted 5 lines
RedHat remains unbeatable with the combined powers of Simon and Paolo, as well as high participation of the less active authors. Alibaba maintains its strongly net-negative review score. Tencent (Jason Xing) joins them (Tencent doesn't use their email domain so it's likely under-counted). Yang Xiwen makes the negative
Interesting numbers.
Well, actually, I always send patches by using the company address and review patches by using my personal address, which probably makes my scores negative :S
Next time I'll try to use the same address. Hope I will not appear on the top of the negative scores list.
Quoted 5 lines
list as well, cost of reposting large series too often... --
Code: https://github.com/kuba-moo/ml-statRaw output: https://netdev.bots.linux.dev/static/nipa/stats-6.9/stdout
Jakub Kicinski
maintainerQuoted 6 lines · On Wed, 13 Mar 2024 10:13:43 +0800 Jason Xing wrote:
Quoted 5 lines
RedHat remains unbeatable with the combined powers of Simon and Paolo, as well as high participation of the less active authors. Alibaba maintains its strongly net-negative review score. Tencent (Jason Xing) joins them (Tencent doesn't use their email domain so it's likely under-counted). Yang Xiwen makes the negative
Quoted 8 lines
Interesting numbers.
Well, actually, I always send patches by using the company address and review patches by using my personal address, which probably makes my scores negative :S
Next time I'll try to use the same address. Hope I will not appear on the top of the negative scores list.
I can add an alias to map multiple addresses to the same person locally. Feel free to send me the addresses (on- or off-list).
Some goes for everyone else, please feel free to send me your mappings and company affiliations. I don't have access to the Greg KH/LWN gitdm db and Linux Foundation remains unwilling to help.