Some numbers from the 4.19 development cycle

As of this writing, 13,657 non-merge changesets have found their way into the mainline for 4.19. That makes this development cycle the busiest since 4.15, but only by a little bit; the patch volume in recent cycles has been remarkably constant:

Cycle	Changesets
4.15	14,866
4.16	13,630
4.17	13,541
4.18	13,283
4.19	13,657	(so far)

The changes in 4.19 were contributed by 1,710 developers, again a typical number; 253 of those developers were contributing to the kernel for the first time. The last two development cycles both removed more lines of code from the kernel than they added; that trend has come to a screeching halt in 4.19, which added 307,000 lines.

The most active 4.19 developers were:

Most active 4.19 developers

By changesets John Whitmore 222 1.6% Chris Wilson 208 1.5% Gustavo A. R. Silva 205 1.5% Colin Ian King 178 1.3% Arnd Bergmann 155 1.1% Christoph Hellwig 132 1.0% Takashi Iwai 124 0.9% Todd Poynor 116 0.8% Bart Van Assche 110 0.8% Ville Syrjälä 104 0.8% Paul E. McKenney 101 0.7% Michael Straube 101 0.7% Brian Foster 90 0.7% Hans de Goede 86 0.6% Jason Gunthorpe 86 0.6% Boris Brezillon 85 0.6% Geert Uytterhoeven 83 0.6% Jerome Brunet 79 0.6% Jakub Kicinski 78 0.6% YueHaibing 77 0.6%

By changed lines Jeykumar Sankaran 32138 4.8% Richard Fitzgerald 14390 2.2% Jason Cooper 11415 1.7% Steven J. Hill 10008 1.5% Stanislaw Gruszka 8686 1.3% Darrick J. Wong 8396 1.3% Christoph Hellwig 8366 1.3% Simon Que 8083 1.2% Jerome Brunet 7702 1.2% Jiri Pirko 6597 1.0% Gao Xiang 6464 1.0% Jason Gunthorpe 6333 1.0% Rob Clark 6220 0.9% Lorenzo Bianconi 6032 0.9% Chris Wilson 5970 0.9% Linus Walleij 5642 0.9% Srinivas Kandagatla 5170 0.8% Benjamin Herrenschmidt 5150 0.8% Jordan Crouse 5114 0.8% David Lechner 5063 0.8%

The ranks of the top contributors include some new names this time around. John Whitmore's work was entirely focused on improving two Realtek drivers in the staging tree. Chris Wilson made many changes to the i915 graphics driver, Gustavo A. R. Silva and Colin Ian King made small cleanups all over the tree, and Arnd Bergmann made significant fixes all over, many as part of the larger year-2038 readiness effort.

On the "lines changed" side, Jeykumar Sankaran added a graphics driver for SDM845 chipsets. Richard Fitzgerald added support for some Cirrus Logic codecs, Jason Cooper removed the unloved Skein and Threefish crypto algorithms from the staging tree, Steven Hill cleaned up the MIPS Octeon architecture code, and Stanislaw Gruszka contributed the MediaTek mt76x0 driver.

Work on 4.19 was supported by a minimum of 230 employers, the most active of which were:

Most active 4.19 employers

By changesets Intel 1294 9.5% (None) 1180 8.6% Red Hat 970 7.1% IBM 674 4.9% (Unknown) 662 4.8% Linaro 604 4.4% Mellanox 561 4.1% AMD 546 4.0% Google 541 4.0% SUSE 495 3.6% Huawei Technologies 390 2.9% (Consultant) 309 2.3% Renesas Electronics 294 2.2% Bootlin 291 2.1% ARM 272 2.0% Oracle 250 1.8% Linux Foundation 235 1.7% Canonical 225 1.6% NXP Semiconductors 204 1.5% Code Aurora Forum 190 1.4%

By lines changed (None) 56201 8.5% Code Aurora Forum 53644 8.1% Intel 52937 8.0% Red Hat 44222 6.7% Mellanox 35693 5.4% Linaro 35591 5.4% IBM 26092 3.9% Google 24996 3.8% AMD 20602 3.1% (Unknown) 19136 2.9% Huawei Technologies 17230 2.6% (Consultant) 16199 2.4% Cirrus Logic 14565 2.2% SUSE 13685 2.1% Cavium 13338 2.0% Oracle 13309 2.0% BayLibre 11854 1.8% ARM 10897 1.6% Renesas Electronics 10703 1.6% Facebook 10082 1.5%

One thing that jumps out is the amount of work that came from developers who were not working for anybody else. That number is still low by long-term historic standards (it was 12% for 3.0, for example), but is higher than it has been in recent times. There were 116 developers known to be working on their own time, or just under 7% of the developers working on the kernel; they contributed 8.5% of the total work.

With regard to testing and reviewing, the numbers this time around look like this:

Test and review credits in 4.19

Tested-by Pavel Machek 40 5.5% Andrew Bowers 27 3.7% Alexandre Courbot 26 3.6% Arnaldo Carvalho de Melo 26 3.6% Joel Stanley 23 3.2% Shakeel Butt 17 2.4% Neil Brown 15 2.1% Hans de Goede 14 1.9% Tony Brelinski 13 1.8% Stan Johnson 12 1.7% Jiri Kosina 11 1.5% Song Liu 10 1.4% Randy Dunlap 10 1.4% Peter Rosin 9 1.2% Matthias Kaehlcke 7 1.0% Lucas Stach 7 1.0% Hanjun Guo 7 1.0% Ganapatrao Kulkarni 7 1.0% Dave Penkler 7 1.0% Aaron Brown 7 1.0%

Reviewed-by Rob Herring 185 3.8% Darrick J. Wong 144 2.9% Christoph Hellwig 134 2.7% Christian König 101 2.1% Andrew Morton 91 1.8% Alex Deucher 85 1.7% Geert Uytterhoeven 77 1.6% Simon Horman 66 1.3% David Sterba 63 1.3% Boris Brezillon 62 1.3% Tony Cheng 61 1.2% Andy Shevchenko 57 1.2% Tvrtko Ursulin 57 1.2% Daniel Vetter 53 1.1% Quentin Monnet 53 1.1% Ville Syrjälä 52 1.1% Rodrigo Vivi 49 1.0% Harry Wentland 49 1.0% Fabio Estevam 49 1.0%

Of the 13,657 patches merged for 4.19, 659 carried Tested-by tags — about 5% of the total. 4,085 (30%) carried Reviewed-by tags. The following chart shows the trends in the use of these tags in recent years:

As can be seen in this plot, the percentage of patches carrying Tested-by tags is at best flat over the 4.x series (which began in 2015). If one looks further back to 3.1 in late 2011, though, 3.6% of patches carried those tags, so there has been a bit of growth over that period. Reviewed-by tags, instead, are clearly appearing in more patches over time as more subsystems require them. That number, too, is up significantly from 3.1, where only 8.7% of patches had such tags. Kernel developers may be slow to adopt change, but some things do eventually change over time.

What doesn't seem to change is that the kernel-development machine continues to crank out releases integrating vast amounts of change on a predictable basis. The community is in a period of change in a number of ways, but it is to be expected that this record, which has held steady for many years now, will continue into the future.