1 .. SPDX-License-Identifier: (GPL-2.0+ OR CC-BY-4.0)
2 .. See the bottom of this file for additional redistribution information.
7 *We don't cause regressions* -- this document describes what this "first rule of
8 Linux kernel development" means in practice for developers. It complements
9 Documentation/admin-guide/reporting-regressions.rst, which covers the topic from a
10 user's point of view; if you never read that text, go and at least skim over it
11 before continuing here.
13 The important bits (aka "The TL;DR")
14 ====================================
16 #. Ensure subscribers of the `regression mailing list <https://lore.kernel.org/regressions/>`_
17 (regressions@lists.linux.dev) quickly become aware of any new regression
20 * When receiving a mailed report that did not CC the list, bring it into the
21 loop by immediately sending at least a brief "Reply-all" with the list
24 * Forward or bounce any reports submitted in bug trackers to the list.
26 #. Make the Linux kernel regression tracking bot "regzbot" track the issue (this
27 is optional, but recommended):
29 * For mailed reports, check if the reporter included a line like ``#regzbot
30 introduced v5.13..v5.14-rc1``. If not, send a reply (with the regressions
31 list in CC) containing a paragraph like the following, which tells regzbot
32 when the issue started to happen::
34 #regzbot ^introduced 1f2e3d4c5b6a
36 * When forwarding reports from a bug tracker to the regressions list (see
37 above), include a paragraph like the following::
39 #regzbot introduced: v5.13..v5.14-rc1
40 #regzbot from: Some N. Ice Human <some.human@example.com>
41 #regzbot monitor: http://some.bugtracker.example.com/ticket?id=123456789
43 #. When submitting fixes for regressions, add "Link:" tags to the patch
44 description pointing to all places where the issue was reported, as
45 mandated by Documentation/process/submitting-patches.rst and
46 :ref:`Documentation/process/5.Posting.rst <development_posting>`.
48 #. Try to fix regressions quickly once the culprit has been identified; fixes
49 for most regressions should be merged within two weeks, but some need to be
50 resolved within two or three days.
53 All the details on Linux kernel regressions relevant for developers
54 ===================================================================
57 The important basics in more detail
58 -----------------------------------
61 What to do when receiving regression reports
62 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
64 Ensure the Linux kernel's regression tracker and others subscribers of the
65 `regression mailing list <https://lore.kernel.org/regressions/>`_
66 (regressions@lists.linux.dev) become aware of any newly reported regression:
68 * When you receive a report by mail that did not CC the list, immediately bring
69 it into the loop by sending at least a brief "Reply-all" with the list CCed;
70 try to ensure it gets CCed again in case you reply to a reply that omitted
73 * If a report submitted in a bug tracker hits your Inbox, forward or bounce it
74 to the list. Consider checking the list archives beforehand, if the reporter
75 already forwarded the report as instructed by
76 Documentation/admin-guide/reporting-issues.rst.
78 When doing either, consider making the Linux kernel regression tracking bot
79 "regzbot" immediately start tracking the issue:
81 * For mailed reports, check if the reporter included a "regzbot command" like
82 ``#regzbot introduced 1f2e3d4c5b6a``. If not, send a reply (with the
83 regressions list in CC) with a paragraph like the following:::
85 #regzbot ^introduced: v5.13..v5.14-rc1
87 This tells regzbot the version range in which the issue started to happen;
88 you can specify a range using commit-ids as well or state a single commit-id
89 in case the reporter bisected the culprit.
91 Note the caret (^) before the "introduced": it tells regzbot to treat the
92 parent mail (the one you reply to) as the initial report for the regression
93 you want to see tracked; that's important, as regzbot will later look out
94 for patches with "Link:" tags pointing to the report in the archives on
97 * When forwarding a regressions reported to a bug tracker, include a paragraph
98 with these regzbot commands::
100 #regzbot introduced: 1f2e3d4c5b6a
101 #regzbot from: Some N. Ice Human <some.human@example.com>
102 #regzbot monitor: http://some.bugtracker.example.com/ticket?id=123456789
104 Regzbot will then automatically associate patches with the report that
105 contain "Link:" tags pointing to your mail or the mentioned ticket.
107 What's important when fixing regressions
108 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
110 You don't need to do anything special when submitting fixes for regression, just
111 remember to do what Documentation/process/submitting-patches.rst,
112 :ref:`Documentation/process/5.Posting.rst <development_posting>`, and
113 Documentation/process/stable-kernel-rules.rst already explain in more detail:
115 * Point to all places where the issue was reported using "Link:" tags::
117 Link: https://lore.kernel.org/r/30th.anniversary.repost@klaava.Helsinki.FI/
118 Link: https://bugzilla.kernel.org/show_bug.cgi?id=1234567890
120 * Add a "Fixes:" tag to specify the commit causing the regression.
122 * If the culprit was merged in an earlier development cycle, explicitly mark
123 the fix for backporting using the ``Cc: stable@vger.kernel.org`` tag.
125 All this is expected from you and important when it comes to regression, as
126 these tags are of great value for everyone (you included) that might be looking
127 into the issue weeks, months, or years later. These tags are also crucial for
128 tools and scripts used by other kernel developers or Linux distributions; one of
129 these tools is regzbot, which heavily relies on the "Link:" tags to associate
130 reports for regression with changes resolving them.
132 Prioritize work on fixing regressions
133 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
135 You should fix any reported regression as quickly as possible, to provide
136 affected users with a solution in a timely manner and prevent more users from
137 running into the issue; nevertheless developers need to take enough time and
138 care to ensure regression fixes do not cause additional damage.
140 In the end though, developers should give their best to prevent users from
141 running into situations where a regression leaves them only three options: "run
142 a kernel with a regression that seriously impacts usage", "continue running an
143 outdated and thus potentially insecure kernel version for more than two weeks
144 after a regression's culprit was identified", and "downgrade to a still
145 supported kernel series that lack required features".
147 How to realize this depends a lot on the situation. Here are a few rules of
148 thumb for you, in order or importance:
150 * Prioritize work on handling regression reports and fixing regression over all
151 other Linux kernel work, unless the latter concerns acute security issues or
152 bugs causing data loss or damage.
154 * Always consider reverting the culprit commits and reapplying them later
155 together with necessary fixes, as this might be the least dangerous and
156 quickest way to fix a regression.
158 * Developers should handle regressions in all supported kernel series, but are
159 free to delegate the work to the stable team, if the issue probably at no
160 point in time occurred with mainline.
162 * Try to resolve any regressions introduced in the current development before
163 its end. If you fear a fix might be too risky to apply only days before a new
164 mainline release, let Linus decide: submit the fix separately to him as soon
165 as possible with the explanation of the situation. He then can make a call
166 and postpone the release if necessary, for example if multiple such changes
167 show up in his inbox.
169 * Address regressions in stable, longterm, or proper mainline releases with
170 more urgency than regressions in mainline pre-releases. That changes after
171 the release of the fifth pre-release, aka "-rc5": mainline then becomes as
172 important, to ensure all the improvements and fixes are ideally tested
173 together for at least one week before Linus releases a new mainline version.
175 * Fix regressions within two or three days, if they are critical for some
176 reason -- for example, if the issue is likely to affect many users of the
177 kernel series in question on all or certain architectures. Note, this
178 includes mainline, as issues like compile errors otherwise might prevent many
179 testers or continuous integration systems from testing the series.
181 * Aim to fix regressions within one week after the culprit was identified, if
182 the issue was introduced in either:
184 * a recent stable/longterm release
186 * the development cycle of the latest proper mainline release
188 In the latter case (say Linux v5.14), try to address regressions even
189 quicker, if the stable series for the predecessor (v5.13) will be abandoned
190 soon or already was stamped "End-of-Life" (EOL) -- this usually happens about
191 three to four weeks after a new mainline release.
193 * Try to fix all other regressions within two weeks after the culprit was
194 found. Two or three additional weeks are acceptable for performance
195 regressions and other issues which are annoying, but don't prevent anyone
196 from running Linux (unless it's an issue in the current development cycle,
197 as those should ideally be addressed before the release). A few weeks in
198 total are acceptable if a regression can only be fixed with a risky change
199 and at the same time is affecting only a few users; as much time is
200 also okay if the regression is already present in the second newest longterm
203 Note: The aforementioned time frames for resolving regressions are meant to
204 include getting the fix tested, reviewed, and merged into mainline, ideally with
205 the fix being in linux-next at least briefly. This leads to delays you need to
208 Subsystem maintainers are expected to assist in reaching those periods by doing
209 timely reviews and quick handling of accepted patches. They thus might have to
210 send git-pull requests earlier or more often than usual; depending on the fix,
211 it might even be acceptable to skip testing in linux-next. Especially fixes for
212 regressions in stable and longterm kernels need to be handled quickly, as fixes
213 need to be merged in mainline before they can be backported to older series.
216 More aspects regarding regressions developers should be aware of
217 ----------------------------------------------------------------
220 How to deal with changes where a risk of regression is known
221 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
223 Evaluate how big the risk of regressions is, for example by performing a code
224 search in Linux distributions and Git forges. Also consider asking other
225 developers or projects likely to be affected to evaluate or even test the
226 proposed change; if problems surface, maybe some solution acceptable for all
229 If the risk of regressions in the end seems to be relatively small, go ahead
230 with the change, but let all involved parties know about the risk. Hence, make
231 sure your patch description makes this aspect obvious. Once the change is
232 merged, tell the Linux kernel's regression tracker and the regressions mailing
233 list about the risk, so everyone has the change on the radar in case reports
234 trickle in. Depending on the risk, you also might want to ask the subsystem
235 maintainer to mention the issue in his mainline pull request.
237 What else is there to known about regressions?
238 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
240 Check out Documentation/admin-guide/reporting-regressions.rst, it covers a lot
241 of other aspects you want might want to be aware of:
243 * the purpose of the "no regressions rule"
245 * what issues actually qualify as regression
247 * who's in charge for finding the root cause of a regression
249 * how to handle tricky situations, e.g. when a regression is caused by a
250 security fix or when fixing a regression might cause another one
252 Whom to ask for advice when it comes to regressions
253 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
255 Send a mail to the regressions mailing list (regressions@lists.linux.dev) while
256 CCing the Linux kernel's regression tracker (regressions@leemhuis.info); if the
257 issue might better be dealt with in private, feel free to omit the list.
260 More about regression tracking and regzbot
261 ------------------------------------------
264 Why the Linux kernel has a regression tracker, and why is regzbot used?
265 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
267 Rules like "no regressions" need someone to ensure they are followed, otherwise
268 they are broken either accidentally or on purpose. History has shown this to be
269 true for the Linux kernel as well. That's why Thorsten Leemhuis volunteered to
270 keep an eye on things as the Linux kernel's regression tracker, who's
271 occasionally helped by other people. Neither of them are paid to do this,
272 that's why regression tracking is done on a best effort basis.
274 Earlier attempts to manually track regressions have shown it's an exhausting and
275 frustrating work, which is why they were abandoned after a while. To prevent
276 this from happening again, Thorsten developed regzbot to facilitate the work,
277 with the long term goal to automate regression tracking as much as possible for
280 How does regression tracking work with regzbot?
281 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
283 The bot watches for replies to reports of tracked regressions. Additionally,
284 it's looking out for posted or committed patches referencing such reports
285 with "Link:" tags; replies to such patch postings are tracked as well.
286 Combined this data provides good insights into the current state of the fixing
289 Regzbot tries to do its job with as little overhead as possible for both
290 reporters and developers. In fact, only reporters are burdened with an extra
291 duty: they need to tell regzbot about the regression report using the ``#regzbot
292 introduced`` command outlined above; if they don't do that, someone else can
293 take care of that using ``#regzbot ^introduced``.
295 For developers there normally is no extra work involved, they just need to make
296 sure to do something that was expected long before regzbot came to light: add
297 "Link:" tags to the patch description pointing to all reports about the issue
300 Do I have to use regzbot?
301 ~~~~~~~~~~~~~~~~~~~~~~~~~
303 It's in the interest of everyone if you do, as kernel maintainers like Linus
304 Torvalds partly rely on regzbot's tracking in their work -- for example when
305 deciding to release a new version or extend the development phase. For this they
306 need to be aware of all unfixed regression; to do that, Linus is known to look
307 into the weekly reports sent by regzbot.
309 Do I have to tell regzbot about every regression I stumble upon?
310 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
312 Ideally yes: we are all humans and easily forget problems when something more
313 important unexpectedly comes up -- for example a bigger problem in the Linux
314 kernel or something in real life that's keeping us away from keyboards for a
315 while. Hence, it's best to tell regzbot about every regression, except when you
316 immediately write a fix and commit it to a tree regularly merged to the affected
319 How to see which regressions regzbot tracks currently?
320 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
322 Check `regzbot's web-interface <https://linux-regtracking.leemhuis.info/regzbot/>`_
323 for the latest info; alternatively, `search for the latest regression report
324 <https://lore.kernel.org/lkml/?q=%22Linux+regressions+report%22+f%3Aregzbot>`_,
325 which regzbot normally sends out once a week on Sunday evening (UTC), which is a
326 few hours before Linus usually publishes new (pre-)releases.
328 What places is regzbot monitoring?
329 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
331 Regzbot is watching the most important Linux mailing lists as well as the git
332 repositories of linux-next, mainline, and stable/longterm.
334 What kind of issues are supposed to be tracked by regzbot?
335 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
337 The bot is meant to track regressions, hence please don't involve regzbot for
338 regular issues. But it's okay for the Linux kernel's regression tracker if you
339 use regzbot to track severe issues, like reports about hangs, corrupted data,
340 or internal errors (Panic, Oops, BUG(), warning, ...).
342 Can I add regressions found by CI systems to regzbot's tracking?
343 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
345 Feel free to do so, if the particular regression likely has impact on practical
346 use cases and thus might be noticed by users; hence, please don't involve
347 regzbot for theoretical regressions unlikely to show themselves in real world
350 How to interact with regzbot?
351 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
353 By using a 'regzbot command' in a direct or indirect reply to the mail with the
354 regression report. These commands need to be in their own paragraph (IOW: they
355 need to be separated from the rest of the mail using blank lines).
357 One such command is ``#regzbot introduced <version or commit>``, which makes
358 regzbot consider your mail as a regressions report added to the tracking, as
359 already described above; ``#regzbot ^introduced <version or commit>`` is another
360 such command, which makes regzbot consider the parent mail as a report for a
361 regression which it starts to track.
363 Once one of those two commands has been utilized, other regzbot commands can be
364 used in direct or indirect replies to the report. You can write them below one
365 of the `introduced` commands or in replies to the mail that used one of them
366 or itself is a reply to that mail:
368 * Set or update the title::
372 * Monitor a discussion or bugzilla.kernel.org ticket where additions aspects of
373 the issue or a fix are discussed -- for example the posting of a patch fixing
376 #regzbot monitor: https://lore.kernel.org/all/30th.anniversary.repost@klaava.Helsinki.FI/
378 Monitoring only works for lore.kernel.org and bugzilla.kernel.org; regzbot
379 will consider all messages in that thread or ticket as related to the fixing
382 * Point to a place with further details of interest, like a mailing list post
383 or a ticket in a bug tracker that are slightly related, but about a different
386 #regzbot link: https://bugzilla.kernel.org/show_bug.cgi?id=123456789
388 * Mark a regression as fixed by a commit that is heading upstream or already
391 #regzbot fixed-by: 1f2e3d4c5d
393 * Mark a regression as a duplicate of another one already tracked by regzbot::
395 #regzbot dup-of: https://lore.kernel.org/all/30th.anniversary.repost@klaava.Helsinki.FI/
397 * Mark a regression as invalid::
399 #regzbot invalid: wasn't a regression, problem has always existed
401 Is there more to tell about regzbot and its commands?
402 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
404 More detailed and up-to-date information about the Linux
405 kernel's regression tracking bot can be found on its
406 `project page <https://gitlab.com/knurd42/regzbot>`_, which among others
407 contains a `getting started guide <https://gitlab.com/knurd42/regzbot/-/blob/main/docs/getting_started.md>`_
408 and `reference documentation <https://gitlab.com/knurd42/regzbot/-/blob/main/docs/reference.md>`_
409 which both cover more details than the above section.
411 Quotes from Linus about regression
412 ----------------------------------
414 Find below a few real life examples of how Linus Torvalds expects regressions to
417 * From `2017-10-26 (1/2)
418 <https://lore.kernel.org/lkml/CA+55aFwiiQYJ+YoLKCXjN_beDVfu38mg=Ggg5LFOcqHE8Qi7Zw@mail.gmail.com/>`_::
420 If you break existing user space setups THAT IS A REGRESSION.
422 It's not ok to say "but we'll fix the user space setup".
430 - we don't cause regressions
432 and the corollary is that when regressions *do* occur, we admit to
433 them and fix them, instead of blaming user space.
435 The fact that you have apparently been denying the regression now for
436 three weeks means that I will revert, and I will stop pulling apparmor
437 requests until the people involved understand how kernel development
440 * From `2017-10-26 (2/2)
441 <https://lore.kernel.org/lkml/CA+55aFxW7NMAMvYhkvz1UPbUTUJewRt6Yb51QAx5RtrWOwjebg@mail.gmail.com/>`_::
443 People should basically always feel like they can update their kernel
444 and simply not have to worry about it.
446 I refuse to introduce "you can only update the kernel if you also
447 update that other program" kind of limitations. If the kernel used to
448 work for you, the rule is that it continues to work for you.
450 There have been exceptions, but they are few and far between, and they
451 generally have some major and fundamental reasons for having happened,
452 that were basically entirely unavoidable, and people _tried_hard_ to
453 avoid them. Maybe we can't practically support the hardware any more
454 after it is decades old and nobody uses it with modern kernels any
455 more. Maybe there's a serious security issue with how we did things,
456 and people actually depended on that fundamentally broken model. Maybe
457 there was some fundamental other breakage that just _had_ to have a
458 flag day for very core and fundamental reasons.
460 And notice that this is very much about *breaking* peoples environments.
462 Behavioral changes happen, and maybe we don't even support some
463 feature any more. There's a number of fields in /proc/<pid>/stat that
464 are printed out as zeroes, simply because they don't even *exist* in
465 the kernel any more, or because showing them was a mistake (typically
466 an information leak). But the numbers got replaced by zeroes, so that
467 the code that used to parse the fields still works. The user might not
468 see everything they used to see, and so behavior is clearly different,
469 but things still _work_, even if they might no longer show sensitive
470 (or no longer relevant) information.
472 But if something actually breaks, then the change must get fixed or
473 reverted. And it gets fixed in the *kernel*. Not by saying "well, fix
474 your user space then". It was a kernel change that exposed the
475 problem, it needs to be the kernel that corrects for it, because we
476 have a "upgrade in place" model. We don't have a "upgrade with new
479 And I seriously will refuse to take code from people who do not
480 understand and honor this very simple rule.
482 This rule is also not going to change.
484 And yes, I realize that the kernel is "special" in this respect. I'm
487 I have seen, and can point to, lots of projects that go "We need to
488 break that use case in order to make progress" or "you relied on
489 undocumented behavior, it sucks to be you" or "there's a better way to
490 do what you want to do, and you have to change to that new better
491 way", and I simply don't think that's acceptable outside of very early
492 alpha releases that have experimental users that know what they signed
493 up for. The kernel hasn't been in that situation for the last two
496 We do API breakage _inside_ the kernel all the time. We will fix
497 internal problems by saying "you now need to do XYZ", but then it's
498 about internal kernel API's, and the people who do that then also
499 obviously have to fix up all the in-kernel users of that API. Nobody
500 can say "I now broke the API you used, and now _you_ need to fix it
501 up". Whoever broke something gets to fix it too.
503 And we simply do not break user space.
506 <https://lore.kernel.org/all/CAHk-=wiVi7mSrsMP=fLXQrXK_UimybW=ziLOwSzFTtoXUacWVQ@mail.gmail.com/>`_::
508 The rules about regressions have never been about any kind of
509 documented behavior, or where the code lives.
511 The rules about regressions are always about "breaks user workflow".
513 Users are literally the _only_ thing that matters.
515 No amount of "you shouldn't have used this" or "that behavior was
516 undefined, it's your own fault your app broke" or "that used to work
517 simply because of a kernel bug" is at all relevant.
519 Now, reality is never entirely black-and-white. So we've had things
520 like "serious security issue" etc that just forces us to make changes
521 that may break user space. But even then the rule is that we don't
522 really have other options that would allow things to continue.
524 And obviously, if users take years to even notice that something
525 broke, or if we have sane ways to work around the breakage that
526 doesn't make for too much trouble for users (ie "ok, there are a
527 handful of users, and they can use a kernel command line to work
528 around it" kind of things) we've also been a bit less strict.
530 But no, "that was documented to be broken" (whether it's because the
531 code was in staging or because the man-page said something else) is
532 irrelevant. If staging code is so useful that people end up using it,
533 that means that it's basically regular kernel code with a flag saying
534 "please clean this up".
536 The other side of the coin is that people who talk about "API
537 stability" are entirely wrong. API's don't matter either. You can make
538 any changes to an API you like - as long as nobody notices.
540 Again, the regression rule is not about documentation, not about
541 API's, and not about the phase of the moon.
543 It's entirely about "we caused problems for user space that used to work".
546 <https://lore.kernel.org/all/CA+55aFzUvbGjD8nQ-+3oiMBx14c_6zOj2n7KLN3UsJ-qsd4Dcw@mail.gmail.com/>`_::
548 And our regression rule has never been "behavior doesn't change".
549 That would mean that we could never make any changes at all.
551 For example, we do things like add new error handling etc all the
552 time, which we then sometimes even add tests for in our kselftest
555 So clearly behavior changes all the time and we don't consider that a
558 The rule for a regression for the kernel is that some real user
559 workflow breaks. Not some test. Not a "look, I used to be able to do
563 <https://lore.kernel.org/all/CA+55aFwWZX=CXmWDTkDGb36kf12XmTehmQjbiMPCqCRG2hi9kw@mail.gmail.com/>`_::
565 YOU ARE MISSING THE #1 KERNEL RULE.
567 We do not regress, and we do not regress exactly because your are 100% wrong.
569 And the reason you state for your opinion is in fact exactly *WHY* you
572 Your "good reasons" are pure and utter garbage.
574 The whole point of "we do not regress" is so that people can upgrade
575 the kernel and never have to worry about it.
577 > Kernel had a bug which has been fixed
579 That is *ENTIRELY* immaterial.
581 Guys, whether something was buggy or not DOES NOT MATTER.
585 Bugs happen. That's a fact of life. Arguing that "we had to break
586 something because we were fixing a bug" is completely insane. We fix
587 tens of bugs every single day, thinking that "fixing a bug" means that
588 we can break something is simply NOT TRUE.
590 So bugs simply aren't even relevant to the discussion. They happen,
591 they get found, they get fixed, and it has nothing to do with "we
594 Because the only thing that matters IS THE USER.
596 How hard is that to understand?
598 Anybody who uses "but it was buggy" as an argument is entirely missing
599 the point. As far as the USER was concerned, it wasn't buggy - it
602 Maybe it worked *because* the user had taken the bug into account,
603 maybe it worked because the user didn't notice - again, it doesn't
604 matter. It worked for the user.
606 Breaking a user workflow for a "bug" is absolutely the WORST reason
607 for breakage you can imagine.
609 It's basically saying "I took something that worked, and I broke it,
610 but now it's better". Do you not see how f*cking insane that statement
613 And without users, your program is not a program, it's a pointless
614 piece of code that you might as well throw away.
616 Seriously. This is *why* the #1 rule for kernel development is "we
617 don't break users". Because "I fixed a bug" is absolutely NOT AN
618 ARGUMENT if that bug fix broke a user setup. You actually introduced a
619 MUCH BIGGER bug by "fixing" something that the user clearly didn't
622 And dammit, we upgrade the kernel ALL THE TIME without upgrading any
623 other programs at all. It is absolutely required, because flag-days
624 and dependencies are horribly bad.
626 And it is also required simply because I as a kernel developer do not
627 upgrade random other tools that I don't even care about as I develop
628 the kernel, and I want any of my users to feel safe doing the same
631 So no. Your rule is COMPLETELY wrong. If you cannot upgrade a kernel
632 without upgrading some other random binary, then we have a problem.
635 <https://lore.kernel.org/all/CAHk-=wiUVqHN76YUwhkjZzwTdjMMJf_zN4+u7vEJjmEGh3recw@mail.gmail.com/>`_::
637 THERE ARE NO VALID ARGUMENTS FOR REGRESSIONS.
639 Honestly, security people need to understand that "not working" is not
640 a success case of security. It's a failure case.
642 Yes, "not working" may be secure. But security in that case is *pointless*.
644 * From `2011-05-06 (1/3)
645 <https://lore.kernel.org/all/BANLkTim9YvResB+PwRp7QTK-a5VNg2PvmQ@mail.gmail.com/>`_::
647 Binary compatibility is more important.
649 And if binaries don't use the interface to parse the format (or just
650 parse it wrongly - see the fairly recent example of adding uuid's to
651 /proc/self/mountinfo), then it's a regression.
653 And regressions get reverted, unless there are security issues or
654 similar that makes us go "Oh Gods, we really have to break things".
656 I don't understand why this simple logic is so hard for some kernel
657 developers to understand. Reality matters. Your personal wishes matter
660 If you made an interface that can be used without parsing the
661 interface description, then we're stuck with the interface. Theory
662 simply doesn't matter.
664 You could help fix the tools, and try to avoid the compatibility
665 issues that way. There aren't that many of them.
667 From `2011-05-06 (2/3)
668 <https://lore.kernel.org/all/BANLkTi=KVXjKR82sqsz4gwjr+E0vtqCmvA@mail.gmail.com/>`_::
670 it's clearly NOT an internal tracepoint. By definition. It's being
673 From `2011-05-06 (3/3)
674 <https://lore.kernel.org/all/BANLkTinazaXRdGovYL7rRVp+j6HbJ7pzhg@mail.gmail.com/>`_::
676 We have programs that use that ABI and thus it's a regression if they break.
678 * From `2012-07-06 <https://lore.kernel.org/all/CA+55aFwnLJ+0sjx92EGREGTWOx84wwKaraSzpTNJwPVV8edw8g@mail.gmail.com/>`_::
680 > Now this got me wondering if Debian _unstable_ actually qualifies as a
681 > standard distro userspace.
683 Oh, if the kernel breaks some standard user space, that counts. Tons
684 of people run Debian unstable
687 <https://lore.kernel.org/lkml/CAHk-=wiP4K8DRJWsCo=20hn_6054xBamGKF2kPgUzpB5aMaofA@mail.gmail.com/>`_::
689 One _particularly_ last-minute revert is the top-most commit (ignoring
690 the version change itself) done just before the release, and while
691 it's very annoying, it's perhaps also instructive.
693 What's instructive about it is that I reverted a commit that wasn't
694 actually buggy. In fact, it was doing exactly what it set out to do,
695 and did it very well. In fact it did it _so_ well that the much
696 improved IO patterns it caused then ended up revealing a user-visible
697 regression due to a real bug in a completely unrelated area.
699 The actual details of that regression are not the reason I point that
700 revert out as instructive, though. It's more that it's an instructive
701 example of what counts as a regression, and what the whole "no
702 regressions" kernel rule means. The reverted commit didn't change any
703 API's, and it didn't introduce any new bugs. But it ended up exposing
704 another problem, and as such caused a kernel upgrade to fail for a
705 user. So it got reverted.
707 The point here being that we revert based on user-reported _behavior_,
708 not based on some "it changes the ABI" or "it caused a bug" concept.
709 The problem was really pre-existing, and it just didn't happen to
710 trigger before. The better IO patterns introduced by the change just
711 happened to expose an old bug, and people had grown to depend on the
712 previously benign behavior of that old issue.
714 And never fear, we'll re-introduce the fix that improved on the IO
715 patterns once we've decided just how to handle the fact that we had a
716 bad interaction with an interface that people had then just happened
717 to rely on incidental behavior for before. It's just that we'll have
718 to hash through how to do that (there are no less than three different
719 patches by three different developers being discussed, and there might
720 be more coming...). In the meantime, I reverted the thing that exposed
721 the problem to users for this release, even if I hope it will be
722 re-introduced (perhaps even backported as a stable patch) once we have
723 consensus about the issue it exposed.
725 Take-away from the whole thing: it's not about whether you change the
726 kernel-userspace ABI, or fix a bug, or about whether the old code
727 "should never have worked in the first place". It's about whether
728 something breaks existing users' workflow.
730 Anyway, that was my little aside on the whole regression thing. Since
731 it's that "first rule of kernel programming", I felt it is perhaps
732 worth just bringing it up every once in a while
737 This text is available under GPL-2.0+ or CC-BY-4.0, as stated at the top
738 of the file. If you want to distribute this text under CC-BY-4.0 only,
739 please use "The Linux kernel developers" for author attribution and link
741 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/plain/Documentation/process/handling-regressions.rst
743 Note: Only the content of this RST file as found in the Linux kernel sources
744 is available under CC-BY-4.0, as versions of this text that were processed
745 (for example by the kernel's build system) might contain content taken from
746 files which use a more restrictive license.