3.14 development statistics

Normally, by the time the -rc6 prepatch is released in a given development cycle, the final release is thought to be not too far away. This time around, Linus is making noises about needing to go as far as -rc8 or even -rc9 due to the number of outstanding problems. Even if we are still a few weeks from release, the situation with 3.14 will be stable enough that a look at where the changes in this release came from makes sense. The picture that emerges can be described as "business as usual," with the continuation of some longstanding trends.

As of this writing, the 3.14 development cycle has seen the addition of just over 12,000 non-merge changesets from exactly 1,400 developers. These changes added 591,000 lines of code while removing 250,000 lines, for a net growth of 341,000 lines. The most active developers this time around were:

Most active 3.14 developers

By changesets H Hartley Sweeten 278 2.3% Laurent Pinchart 232 1.9% Jingoo Han 174 1.4% Rashika Kheria 162 1.3% Geert Uytterhoeven 138 1.1% Tejun Heo 123 1.0% Sachin Kamat 122 1.0% Kuninori Morimoto 110 0.9% Sujith Manoharan 97 0.8% Linus Walleij 89 0.7% Wei Yongjun 86 0.7% Alex Deucher 82 0.7% Stephen Warren 81 0.7% Lars-Peter Clausen 79 0.7% Ville Syrjälä 78 0.6% Namhyung Kim 76 0.6% Thierry Reding 74 0.6% Johannes Berg 73 0.6% Christoph Hellwig 73 0.6% Ding Tianhong 71 0.6%

By changed lines Greg Kroah-Hartman 73035 10.5% Micky Ching 23657 3.4% Stephen Boyd 17511 2.5% Paul Gortmaker 17511 2.5% Greg Rose 16428 2.4% Tero Kristo 14509 2.1% Ben Skeggs 13320 1.9% Sergio Aguirre 8388 1.2% Ben Hutchings 8002 1.2% Tejun Heo 7975 1.2% Laurent Pinchart 7799 1.1% Frank Haverkamp 7424 1.1% Takashi Iwai 6247 0.9% Thomas Hellstrom 6148 0.9% Tom Lendacky 6103 0.9% Upinder Malhi 6012 0.9% Sujith Manoharan 5837 0.8% Peter De Schrijver 5680 0.8% H Hartley Sweeten 5586 0.8% Rob Clark 5345 0.8%

After a brief absence, H. Hartley Sweeten has returned to the top of the "by changesets" list with more work on the seemingly interminable project of fixing up the Comedi drivers in the staging tree. Laurent Pinchart did a lot of work in the Video4Linux and ARM architecture trees, while Jingoo Han continues to work on cleaning up the driver tree. Rashika Kheria, an Outreach Program for Women participant, contributed lots of driver cleanup patches, and Geert Uytterhoeven did a lot of work in the m68k architecture and embedded driver subsystems.

Greg Kroah-Hartman often appears near the top of the "lines changed" column; this time, it is as the result of the addition of the Realtek 8821 PCI WIFI driver and more than the usual number of reverted patches. Micky Ching added the rts5208 and rts5288 drivers to the staging tree, Stephen Boyd added a bunch of Qualcomm hardware support, Paul Gortmaker did (among other things) a bunch of header file cleanups, and Greg Rose contributed the Intel i40evf network driver.

At least 210 employers supported work on the 3.14 kernel. The most active of these employers were:

Most active 3.14 employers

By changesets Intel 1233 10.2% (None) 1075 8.9% Red Hat 877 7.3% (Unknown) 701 5.8% Linaro 528 4.4% Samsung 523 4.3% SUSE 396 3.3% IBM 351 2.9% Renesas Electronics 339 2.8% Google 324 2.7% Texas Instruments 288 2.4% Vision Engraving Systems 278 2.3% (Consultant) 268 2.2% NVIDIA 248 2.1% FOSS Outreach Program for Women 237 2.0% Huawei Technologies 211 1.8% Freescale 210 1.7% Qualcomm 157 1.3% Oracle 152 1.3% Broadcom 144 1.2%

By lines changed Linux Foundation 78675 11.4% Intel 69526 10.0% (None) 47083 6.8% Red Hat 46371 6.7% Texas Instruments 28274 4.1% (Unknown) 25716 3.7% IBM 25427 3.7% Realsil Microelectronics 23676 3.4% SUSE 22686 3.3% NVIDIA 20720 3.0% Samsung 19988 2.9% Wind River 19946 2.9% Code Aurora Forum 17878 2.6% Google 13452 1.9% Linaro 12945 1.9% Cisco 12747 1.8% (Consultant) 12301 1.8% Qualcomm 10806 1.6% Renesas Electronics 9655 1.4% Freescale 9533 1.4%

There are few surprises here; instead, this table shows the continuation of some trends we have been seeing for a while. After a short-lived jump in 3.13, the number of contributions from volunteers is back to its long-term decline. Intel seems to have taken a permanent place at the top of the list of changeset contributors. Contributions from the mobile and embedded industry continue to grow. It's tempting to call out the fact that 3.14 will contain a fix to the nouveau driver that came directly from NVIDIA, but this turns out to be the second time that has happened; the first fix from NVIDIA was quietly merged for 3.9 in early 2013.

A slightly different picture emerges if one looks at non-author signoffs — Signed-off-by tags applied to patches by developers other than the author. Such tags are applied by subsystem maintainers as they accept patches; the statistics can thus indicate who the gatekeepers to the kernel are. Associating signoffs with employers leads to these results:

Most signoffs in 3.14

By developer Greg Kroah-Hartman 1516 13.1% David S. Miller 1128 9.7% Mark Brown 502 4.3% Andrew Morton 465 4.0% Mauro Carvalho Chehab 370 3.2% John W. Linville 352 3.0% Simon Horman 256 2.2% Arnaldo Carvalho de Melo 237 2.0% Daniel Vetter 225 1.9% Rafael J. Wysocki 173 1.5% Jeff Kirsher 166 1.4% Chris Mason 160 1.4% Linus Walleij 148 1.3% Ingo Molnar 146 1.3% Brian Norris 140 1.2% John Crispin 129 1.1% Jesse Brandeburg 127 1.1% Josef Bacik 126 1.1% Johannes Berg 125 1.1% Benjamin Herrenschmidt 121 1.0%

By employer Red Hat 2336 20.2% Linux Foundation 1548 13.4% Intel 1367 11.8% Linaro 1093 9.4% Google 649 5.6% Samsung 647 5.6% (None) 417 3.6% Facebook 283 2.4% SUSE 270 2.3% Renesas Electronics 265 2.3% IBM 262 2.3% Texas Instruments 200 1.7% Broadcom 183 1.6% (Unknown) 159 1.4% Fon 129 1.1% NVIDIA 97 0.8% Cisco 91 0.8% Pure Storage 91 0.8% (Consultant) 82 0.7% University of Cambridge 77 0.7%

The concentration of companies here is reduced from what it once was, but it is still true that more than half of the patches that were merged into 3.14 went by way of the employees of just four organizations. Linaro and Facebook are moving up the list; in both cases, this has mostly been done by hiring developers who were already working as subsystem maintainers. Here, too, the presence of volunteer developers has been slowly declining over time.

All told, the kernel development machine appears to be in a relatively steady state, with things running smoothly and changes happening over relatively long periods of time. Your editor is thus curious to know whether these reports remain useful on a per-release basis. Perhaps it makes sense to scale them back to, say, an annual exercise where the long-term trends might be more pronounced. This is a question that will be considered over the course of the 3.15 development cycle.

Index entries for this article
Kernel	Releases/3.14

---

to post comments