From nobody Tue Jun 30 04:30:21 2026 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 4369CC433EF for ; Tue, 25 Jan 2022 11:58:52 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1382615AbiAYL6r (ORCPT ); Tue, 25 Jan 2022 06:58:47 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:56104 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1383761AbiAYLpZ (ORCPT ); Tue, 25 Jan 2022 06:45:25 -0500 Received: from wp530.webpack.hosteurope.de (wp530.webpack.hosteurope.de [IPv6:2a01:488:42:1000:50ed:8234::]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id CD213C06175D; Tue, 25 Jan 2022 03:44:17 -0800 (PST) Received: from ip4d173d02.dynamic.kabel-deutschland.de ([77.23.61.2] helo=truhe.fritz.box); authenticated by wp530.webpack.hosteurope.de running ExIM with esmtpsa (TLS1.3:ECDHE_RSA_AES_256_GCM_SHA384:256) id 1nCKFD-0001tA-KW; Tue, 25 Jan 2022 12:44:15 +0100 From: Thorsten Leemhuis To: linux-doc@vger.kernel.org, Linus Torvalds Cc: workflows@vger.kernel.org, Linux Kernel Mailing List , Jonathan Corbet , Randy Dunlap , regressions@lists.linux.dev, Greg Kroah-Hartman , Lukas Bulwahn Subject: [PATCH v3 1/2] docs: add a document about regression handling Date: Tue, 25 Jan 2022 12:44:13 +0100 Message-Id: X-Mailer: git-send-email 2.31.1 In-Reply-To: References: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable X-bounce-key: webpack.hosteurope.de;linux@leemhuis.info;1643111057;57976a3c; X-HE-SMSGID: 1nCKFD-0001tA-KW Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Type: text/plain; charset="utf-8" Create a document explaining various aspects around regression handling and tracking both for users and developers. Among others describe the first rule of Linux kernel development and what it means in practice. Also explain what a regression actually is and how to report one properly. The text additionally provides a brief introduction to the bot the kernel's regression tracker uses to facilitate his work. To sum things up, provide a few quotes from Linus to show how serious he takes regressions. Signed-off-by: Thorsten Leemhuis Acked-by: Greg Kroah-Hartman --- Documentation/admin-guide/index.rst | 1 + Documentation/admin-guide/regressions.rst | 911 ++++++++++++++++++++++ MAINTAINERS | 1 + 3 files changed, 913 insertions(+) create mode 100644 Documentation/admin-guide/regressions.rst diff --git a/Documentation/admin-guide/index.rst b/Documentation/admin-guid= e/index.rst index 1bedab498104..17157ee5a416 100644 --- a/Documentation/admin-guide/index.rst +++ b/Documentation/admin-guide/index.rst @@ -36,6 +36,7 @@ problems and bugs in particular. =20 reporting-issues security-bugs + regressions bug-hunting bug-bisect tainted-kernels diff --git a/Documentation/admin-guide/regressions.rst b/Documentation/admi= n-guide/regressions.rst new file mode 100644 index 000000000000..837b1658d149 --- /dev/null +++ b/Documentation/admin-guide/regressions.rst @@ -0,0 +1,911 @@ +.. SPDX-License-Identifier: (GPL-2.0+ OR CC-BY-4.0) +.. + If you want to distribute this text under CC-BY-4.0 only, please use 'T= he + Linux kernel developers' for author attribution and link this as source: + https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/plai= n/Documentation/admin-guide/regressions.rst +.. + Note: Only the content of this RST file as found in the Linux kernel so= urces + is available under CC-BY-4.0, as versions of this text that were proces= sed + (for example by the kernel's build system) might contain content taken = from + files which use a more restrictive license. + + +Regressions ++++++++++++ + +The first rule of Linux kernel development: '*We don't cause regressions*'. +Linux founder and lead developer Linus Torvalds strictly enforces the rule +himself. This document describes what it means in practice and how the Lin= ux +kernel's development model ensures all reported regressions are addressed. +The text covers aspects relevant for both users and developers. + +The important bits for people affected by regressions +=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D + +It's a regression if something running fine with one Linux kernel works wo= rse or +not at all with a newer version. Note, the newer kernel has to be compiled= using +a similar configuration -- for this and other fine print, check out below +section "What is a 'regression' and what is the 'no regressions rule'?". + +Report your regression as outlined in +`Documentation/admin-guide/reporting-issues.rst`, it already covers all as= pects +important for regressions. Below section "How do I report a regression?" +highlights them for convenience. + +The most important aspect: CC or forward the report to `the regression mai= ling +list `_ (regressions@lists.linux.dev= ). +When doing so, consider mentioning the version range where the regression +started using a paragraph like this:: + + #regzbot introduced v5.13..v5.14-rc1 + +The Linux kernel regression tracking bot 'regzbot' will then start to trac= k the +issue. This is in your interest, as it brings the report on the radar of p= eople +ensuring all regressions are acted upon in a timely manner. + +The important bits for people fixing regressions +=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D + +When submitting fixes for regressions, add "Link:" tags pointing to all pl= aces +where the issue was reported, as tools like the Linux kernel regression bot +'regzbot' heavily rely on them. These pointers are also of great value when +looking into the issue months or years later, that's why +`Documentation/process/submitting-patches.rst` and +:ref:`Documentation/process/5.Posting.rst ` mandate t= heir +use:: + + Link: https://lore.kernel.org/r/30th.anniversary.repost@klaava.Hels= inki.FI/ + Link: https://bugzilla.kernel.org/show_bug.cgi?id=3D1234567890 + +Let the Linux kernel's regression tracker and all other subscribers of the +`regression mailing list `_ +(regressions@lists.linux.dev) quickly know about newly reported regression= s: + + * When you receive a mailed report that did not CC the list, immediately = send + at least a brief "Reply-all" which get the list into the loop; also ens= ure + it's CCed on all future replies. + + * If you get a report from a bug tracker, forward or bounce the report to= the + list, unless the reporter did that already as outlined by + `Documentation/admin-guide/reporting-issues.rst`. + +Ensure regzbot tracks the issue (this is optional, but recommended): + + * For mailed reports, check if the reporter included a 'regzbot command' = like + the ``#regzbot introduced v5.13..v5.14-rc1`` described above. If not, s= end a + reply (with the regressions list in CC) with a paragraph like the follo= wing, + which brings regzbot into the loop by specifying the version range or c= ommit + when the issue started to happen:: + + #regzbot ^introduced 1f2e3d4c5b6a + + * When receiving a report from a bug tracker and forwarding it to the + regressions list (see above), include a paragraph like the following, w= hich + brings regzbot into the loop by specifying the version range or commit = when + the issue started to happen:: + + #regzbot introduced: v5.13..v5.14-rc1 + #regzbot from: Some N. Ice Human + #regzbot monitor: http://some.bugtracker.example.com/ticket?id=3D12= 3456789 + +All the details on handling Linux kernel regressions +=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D + +The important basics +-------------------- + +What is a 'regression' and what is the 'no regressions rule'? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +It's a regression if some application or practical use case running fine w= ith +one Linux kernel works worse or not at all with a newer version compiled u= sing a +similar configuration. The 'no regressions rule' forbids this to take plac= e; if +it happens by accident, developers that caused it are expected to quickly = fix +the issue. + +It thus is a regression when a WiFi driver from Linux 5.13 works fine, but= with +5.14 doesn't work at all, works significantly slower, or misbehaves someho= w. +It's also a regression if a perfectly working application suddenly shows e= rratic +behavior with a newer kernel version, which can be caused by changes in pr= ocfs, +sysfs, or one of the many other interfaces Linux provides to userland soft= ware. +But keep in mind, as mentioned earlier: 5.14 in this example needs to be b= uilt +from a configuration similar to the one from 5.13. This can be achieved us= ing +``make olddefconfig``, as explained in more detail below. + +Note the 'practical use case' in the first sentence of this section: devel= opers +despite the 'no regressions' rule are free to change any aspect of the ker= nel +and even APIs or ABIs to userland, as long as no existing application or u= se +case breaks. + +Also be aware the 'no regressions' rule covers only interfaces the kernel +provides to the userland. It thus does not apply to kernel-internal interf= aces +like the module API, which some externally developed drivers use to hook i= nto +the kernel. + +What is the goal of the 'no regressions rule'? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Users should feel safe when updating kernel versions and not have to worry +something might break. This is in the interest of the kernel developers to= make +updating attractive: they don't want users to stay on stable or longterm L= inux +series that are either abandoned or more than one and a half year old. Tha= t's in +everybody's interest, as `those series might have known bugs, security iss= ues, +or other problematic aspects already fixed in later versions +`_. +Additionally, the kernel developers want to make it simple and appealing f= or +users to test the latest pre-release or regular release. That's also in +everybody's interest, as it's a lot easier to track down and fix problems,= if +they are reported shortly after being introduced. + + +How hard is the rule enforced? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Extraordinarily strict, as can be seen by many mailing list posts from Lin= ux +creator and lead developer Linus Torvalds, some of which are quoted at the= end +of this document. + +Exceptions to this rule are extremely rare; in the past developers almost = always +turned out to be wrong when they assumed a particular situation was warran= ting +an exception. + +How is the rule enforced? +~~~~~~~~~~~~~~~~~~~~~~~~~ + +It's the duty of the subsystem maintainers, which are watched and supporte= d by +Linus Torvalds for mainline or stable/longterm tree maintainers like Greg +Kroah-Hartman. All of them are supported by Thorsten Leemhuis: he's acting= as +'regressions tracker' for the Linux kernel and trying to ensure all regres= sion +reports are acted upon in a timely manner. + +The distributed and slightly unstructured nature of the Linux kernel's +development makes tracking regressions hard. That's why Thorsten relies on= the +help of his Linux kernel regression tracking robot 'regzbot'. It watches m= ailing +lists and git trees to semi-automatically associate regression reports with +patch submissions and commits fixing the issue, as this provides all neces= sary +insights into the fixing progress. + +Note, the regression tracker can only ensure no regression falls through t= he +cracks, if someone tells him or his bot about every regression found. That= 's why +each report needs to be CCed or forwarded to the regressions mailing list +(ideally with a 'regzbot command' in the mail), as explained in the next +section. + +How do I report a regression? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Just report the issue as outlined in +`Documentation/admin-guide/reporting-issues.rst`, it already describes the +important points. The following aspects described there are especially rel= evant +for regressions: + + * When checking for existing reports to join, first check the `archives o= f the + Linux regressions mailing list `_= and + `regzbot's web-interface `_. + + * In your report, mention the last kernel version that worked fine and the + first broken one. Even better: try to find the commit causing the regre= ssion + using a bisection. + + * Remember to let the Linux regressions mailing list + (regressions@lists.linux.dev) know about your report: + + * If you report the regression by mail, CC the regressions list. + + * If you report your regression to some bug tracker, forward the filed r= eport + by mail to the regressions list while CCing the maintainer and the mai= ling + list for the subsystem in question. + +Additionally, you in both cases should directly tell the aforementioned Li= nux +kernel regression tracking bot about your report. To do that, include a +paragraph like this in your report to tell the bot when the regression sta= rted +to happen:: + + #regzbot introduced: v5.13..v5.14-rc1 + +In this example, v5.13 was the last version that worked, while v5.14-rc1 w= as the +first broken one. The smaller the range, the better, as that makes it easi= er to +find out what's wrong and who's responsible. That's why you ideally should +perform a bisection to find the commit causing the regression (the 'culpri= t'). +If you did, specify it instead:: + + #regzbot introduced: 1f2e3d4c5d + +Placing such a 'regzbot command' is in your interest, as it will ensure the +report won't fall through the cracks unnoticed. If you omit this, the Linux +kernel's regressions tracker will take care of telling regzbot about your +regression, as long as you send a copy to the regressions mailing lists. B= ut the +regression tracker is just one human which sometimes has to rest or occasi= onally +might even enjoy some time away from computers (as crazy as that might sou= nd). +Relying on this person thus will result in an unnecessary delay before the +regressions becomes mentioned `on the list of tracked and unresolved Linux +kernel regressions `_ an= d the +weekly regression reports sent by regzbot. Such delays can result in Linus +Torvalds being unaware of important regressions when deciding between 'con= tinue +development or call this finished and release the final?'. + +How to add a regression to regzbot's tracking somebody else reported? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +It depends on the report: + + * If the regression was reported by mail, reply using your mailers 'Reply= -all' + function with the regressions mailing list (regressions@lists.linux.dev= ) in + CC. In your reply, include a paragraph with a regzbot command like this= :: + + #regzbot ^introduced: v5.13..v5.14-rc1 + + The caret (^) before the 'introduced' tells regzbot to treat the parent= mail + (the one you reply to) as the initial report for the regression you wan= t to + see tracked; regzbot then will automatically associate any patches with= this + regression that point to the report using 'Link:' tags. + + * If the regressions was reported to a bug tracker, forward it to the + regressions list and include a paragraph with these regzbot commands:: + + #regzbot introduced: v5.13..v5.14-rc1 + #regzbot from: Some N. Ice Human + #regzbot monitor: http://some.bugtracker.example.com/ticket?id=3D12= 3456789 + + Regzbot will automatically associate patches with the report that use '= Link:' + tags pointing to your mail or the mentioned ticket. + +In both cases you can specify a commit-id instead of a version range, as t= he +previous section outlines in more detail. + +In case you are having trouble, simply forward the report or a pointer to = it +without further ado to Thorsten Leemhuis (regressions@leemhuis.info), he t= hen +will handle things. + +Are really all regressions fixed? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Nearly all of them are, as long as the change causing the regression (the +'culprit commit') is reliably identified. Some regressions can be fixed wi= thout +this, but often it's required. + +Who needs to find the commit causing a regression? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +It's the reporter's duty to find the culprit, but developers of the affect= ed +subsystem should offer advice and reasonably help where they can. + +How can I find the change causing a regression? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Perform a bisection, as roughly outlined in `Documentation/admin-guide/rep= orting-issues.rst` +and described in more detail by `Documentation/admin-guide/bug-bisect.rst`. +It might sound like a lot of work, but in many cases finds the culprit +relatively quickly. If it's hard or time-consuming to reliably reproduce t= he +issue, consider teaming up with others affected by the problem to narrow d= own +the search range together. + +Who can I ask for advice when it comes to regressions? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Send a mail to the regressions mailing list (regressions@lists.linux.dev) = while +CCing the Linux kernel's regression tracker (regressions@leemhuis.info); i= f the +issue might better be dealt with in private, feel free to omit the list. + + +More details about regressions relevant for reporters +----------------------------------------------------- + +Does a regression need to be fixed, if it can be avoided by updating some = other software? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~~~~~ + +Almost always: yes. If a developer tells you otherwise, ask the regression +tracker for advice as outlined above. + +Does it qualify as a regression if a newer kernel works slower or makes th= e system consume more energy? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +It does, but the difference has to be significant. A five percent slow-dow= n in a +micro-benchmark thus is unlikely to qualify as regression, unless it also +influences the results of a broad benchmark by more than one percent. If in +doubt, ask for advice. + +Is it a regression, if an externally developed kernel module is incompatib= le with a newer kernel? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~~~~~~~~~~~~~ + +No, as the 'no regression' rule is about interfaces and services the Linux +kernel provides to the userland. It thus does not cover building or running +externally developed kernel modules, as they run in kernel-space and hook = into +the kernel using internal interfaces occasionally changed. + +How are regressions handled that are caused by a fix for security vulnerab= ility? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~ + +In extremely rare situations security issues can't be fixed without causing +regressions; those are given way, as they are the lesser evil in the end. +Luckily this almost always can be avoided, as key developers for the affec= ted +area and often Linus Torvalds himself try very hard to fix security issues +without causing regressions. + +If you nevertheless face such a case, check the mailing list archives if p= eople +tried their best to avoid the regression; if in doubt, ask for advice as +outlined above. + +What happens if fixing a regression is impossible without causing another = regression? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~ + +Sadly these things happen, but luckily not very often; if they occur, expe= rt +developers of the affected code area should look into the issue to find a = fix +that avoids regressions or at least their impact. If you run into such a +situation, do what was outlined already for regressions caused by security +fixes: check earlier discussions if people already tried their best and as= k for +advice if in doubt. + +A quick note while at it: these situations could be avoided, if people wou= ld +regularly give mainline pre-releases (say v5.15-rc1 or -rc3) from each cyc= le a +test run. This is best explained by imagining a change integrated between = Linux +v5.14 and v5.15-rc1 which causes a regression, but at the same time is a h= ard +requirement for some other improvement applied for 5.15-rc1. All these cha= nges +often can simply be reverted and the regression thus solved, if someone fi= nds +and reports it before 5.15 is released. A few days or weeks later this sol= ution +can become impossible, as some software might have started to rely on aspe= cts +introduced by one of the follow-up changes: reverting all changes would th= en +cause a regression for users of said software and thus is out of the quest= ion. + +A feature I relied on was removed months ago, but I only noticed now. Does= that qualify as regression? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +It does, but often it's hard to fix them due to the aspects outlined in the +previous section. It hence needs to be dealt with on a case-by-case basis.= This +is another reason why it's in everybody's interest to regularly test mainl= ine +pre-releases. + +Does the 'no regression' rule apply if I seem to be the only person in the= world that is affected by a regression? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +It does, but only for practical usage: the Linux developers want to be fre= e to +remove support for hardware only to be found in attics and museums anymore. + +Note, sometimes regressions can't be avoided to make progress -- and the l= atter +is needed to prevent Linux from stagnation. Hence, if only very few users = seem +to be affected by a regression, it for the greater good might be in their = and +everyone else's interest to not insist on the rule. Especially if there is= an +easy way to circumvent the regression somehow, for example by updating some +software or using a kernel parameter created just for this purpose. + +Does the regression rule apply for code in the staging tree as well? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Not according to the `help text for the configuration option covering all +staging code `_, +which since its early days states:: + + Please note that these drivers are under heavy development, may or + may not work, and may contain userspace interfaces that most likely + will be changed in the near future. + +The staging developers nevertheless often adhere to the 'no regressions' r= ule, +but sometimes bend it to make progress. That's for example why some users = had to +deal with (often negligible) regressions when a WiFi driver from the stagi= ng +tree was replaced by a totally different one written from scratch. + +Why do later versions have to be 'compiled with a similar configuration'? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Because the Linux kernel developers sometimes integrate changes known to c= ause +regressions, but make them optional and disable them in the kernel's defau= lt +configuration. This trick allows progress, as the 'no regressions' rule +otherwise would lead to stagnation. + +Consider for example a new security feature blocking access to some kernel +interfaces often abused by malware, which at the same time are required to= run a +few rarely used applications. The outlined approach makes both camps happy: +people using these applications can leave the new security feature off, wh= ile +everyone else can enable it without running into trouble. + +How to create a configuration similar to the one of an older kernel? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Start a known-good kernel and configure the newer Linux version with ``make +olddefconfig``. This makes the kernel's build scripts pick up the configur= ation +file (the `.config` file) from the running kernel as base for the new one = you +are about to compile; afterwards they set all new configuration options to= their +default value, which should disable new features that might cause regressi= ons. + +Can I report a regression to the upstream developers I found in a pre-comp= iled vanilla kernel? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~~~~~~~~~~ + +You need to ensure the newer kernel was compiled with a similar configurat= ion +file as the older one (see above), as the one that built them might have e= nabled +some known-to-be incompatible feature for the newer kernel. If in a doubt, +report this problem to the kernel's provider and ask for advice. + + +More details about regressions relevant for developers +------------------------------------------------------ + +What should I do, if I suspect a change I'm working on might cause regress= ions? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~ + +Evaluate how big the risk of regressions is, for example by performing a c= ode +search in Linux distributions and Git forges. Also consider asking other +developers or projects likely to be affected to evaluate or even test the +proposed change; if problems surface, maybe some middle ground acceptable = for +all can be found. + +If the risk of regressions in the end seems to be relatively small, go ahe= ad +with the change, but let all involved parties know about the risk. Hence, = make +sure your patch description makes this aspect obvious. Once the change is +merged, tell the Linux kernel's regression tracker and the regressions mai= ling +list about the risk, so everyone has the change on the radar in case repor= ts +trickle in. Depending on the risk, you also might want to ask the subsystem +maintainer to mention the issue in his mainline pull request. + + +Everything developers need to know about regression tracking +------------------------------------------------------------ + +Do I have to use regzbot? +~~~~~~~~~~~~~~~~~~~~~~~~~ + +It's in the interest of everyone if you do, as kernel maintainers like Lin= us +Torvalds partly rely on regzbot's tracking in their work -- for example wh= en +deciding to release a new version or extend the development phase. For thi= s they +need to be aware of all unfixed regression; to do that, Linus is known to = look +into the weekly reports sent by regzbot. + +Do I have to tell regzbot about every regression I stumble upon? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Ideally yes: we are all humans and easily forget problems when something m= ore +important unexpectedly comes up -- for example a bigger problem in the Lin= ux +kernel or something in real life that's keeping us away from keyboards for= a +while. Hence, it's best to tell regzbot about every regression, except whe= n you +immediately write a fix and commit it to a tree regularly merged to the af= fected +kernel series. + +Why does the Linux kernel need a regression tracker, and why does he utili= ze regzbot? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~= ~~~~~~~~~~~ + +Rules like 'no regressions' need someone to enforce them, otherwise they a= re +broken either accidentally or on purpose. History has shown that this is t= rue +for the Linux kernel as well. That's why Thorsten volunteered to keep an e= ye on +things. + +Tracking regressions completely manually has proven to be exhausting and +demotivating, which is why earlier attempts to establish it failed after a +while. To prevent this from happening again, Thorsten developed regzbot to +facilitate the work, with the long term goal to automate regression tracki= ng as +much as possible for everyone involved. + +How does regression tracking work with regzbot? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The bot keeps track of all the reports and monitors their fixing progress.= It +tries to do that with as little overhead as possible for both reporters and +developers. + +In fact, only reporters or someone helping them are burdened with an extra= duty: +they need to tell regzbot about the regression report using one of the +``#regzbot introduced`` commands outlined above. + +For developers there normally is no extra work involved, they just need to= do +something that's expected from them already: add 'Link:' tags to the patch +description pointing to all reports about the issue fixed. + +Thanks to these tags regzbot can associate regression reports with patches= to +fix the issue, whenever they are posted for review or applied to a git tre= e. The +bot additionally watches out for replies to the report. All this data comb= ined +provides a good impression about the current status of the fixing process. + +How to see which regressions regzbot tracks currently? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Check `regzbot's web-interface `_ +for the latest info; alternatively, `search for the latest regression repo= rt +`_, +which regzbot normally sends out once a week on Sunday evening (UTC), whic= h is a +few hours before Linus usually publishes new (pre-)releases. + +What places is regzbot monitoring? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Regzbot is watching the most important Linux mailing lists as well as the +linux-next, mainline and stable/longterm git repositories. + +How to interact with regzbot? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Everyone can interact with the bot using mails containing 'regzbot command= s', +which need to be in their own paragraph (IOW: they need to be separated fr= om the +rest of the mail using blank lines). One such command is ``#regzbot introd= uced +``, which adds a report to the tracking, as already des= cribed +above; ``#regzbot ^introduced `` is another such comman= d, +which makes regzbot consider the parent mail as a report for a regression = which +it starts to track. + +Once one of those two commands has been utilized, other regzbot commands c= an be +used. You can write them below one of the `introduced` commands or in repl= ies to +the mail that used one of them or itself is a reply to that mail: + + * Set or update the title:: + + #regzbot title: foo + + * Link to a related discussion (for example the posting of a patch to fix= the + issue) and monitor it:: + + #regzbot monitor: https://lore.kernel.org/all/30th.anniversary.repo= st@klaava.Helsinki.FI/ + + Monitoring only works for lore.kernel.org; regzbot will consider all me= ssages + in that thread as related to the fixing process. + + * Point to a place with further details, like a bug tracker or a related + mailing list post:: + + #regzbot link: https://bugzilla.kernel.org/show_bug.cgi?id=3D123456= 789 + + * Mark a regression as fixed by a commit that is heading upstream or alre= ady + landed:: + + #regzbot fixed-by: 1f2e3d4c5d + + * Mark a regression as a duplicate of another one already tracked by regz= bot:: + + #regzbot dup-of: https://lore.kernel.org/all/30th.anniversary.repos= t@klaava.Helsinki.FI/ + + * Mark a regression as invalid:: + + #regzbot invalid: wasn't a regression, problem has always existed + +Is there more to tell about regzbot and its commands? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +More detailed and up-to-date information about the Linux +kernel's regression tracking bot can be found on its +`project page `_, which among others +contains a `getting started guide `_ +and `reference documentation `_ +which both cover more details than above section. + + +Quotes from Linus about regression +---------------------------------- + +Find below a few real life examples of how Linus Torvalds expects regressi= ons to +be handled: + + * From `2017-10-26 (1/2) + `_:: + + If you break existing user space setups THAT IS A REGRESSION. + + It's not ok to say "but we'll fix the user space setup". + + Really. NOT OK. + + [...] + + The first rule is: + + - we don't cause regressions + + and the corollary is that when regressions *do* occur, we admit to + them and fix them, instead of blaming user space. + + The fact that you have apparently been denying the regression now f= or + three weeks means that I will revert, and I will stop pulling appar= mor + requests until the people involved understand how kernel development + is done. + + * From `2017-10-26 (2/2) + `_:: + + People should basically always feel like they can update their kern= el + and simply not have to worry about it. + + I refuse to introduce "you can only update the kernel if you also + update that other program" kind of limitations. If the kernel used = to + work for you, the rule is that it continues to work for you. + + There have been exceptions, but they are few and far between, and t= hey + generally have some major and fundamental reasons for having happen= ed, + that were basically entirely unavoidable, and people _tried_hard_ to + avoid them. Maybe we can't practically support the hardware any more + after it is decades old and nobody uses it with modern kernels any + more. Maybe there's a serious security issue with how we did things, + and people actually depended on that fundamentally broken model. Ma= ybe + there was some fundamental other breakage that just _had_ to have a + flag day for very core and fundamental reasons. + + And notice that this is very much about *breaking* peoples environm= ents. + + Behavioral changes happen, and maybe we don't even support some + feature any more. There's a number of fields in /proc//stat th= at + are printed out as zeroes, simply because they don't even *exist* in + the kernel any more, or because showing them was a mistake (typical= ly + an information leak). But the numbers got replaced by zeroes, so th= at + the code that used to parse the fields still works. The user might = not + see everything they used to see, and so behavior is clearly differe= nt, + but things still _work_, even if they might no longer show sensitive + (or no longer relevant) information. + + But if something actually breaks, then the change must get fixed or + reverted. And it gets fixed in the *kernel*. Not by saying "well, f= ix + your user space then". It was a kernel change that exposed the + problem, it needs to be the kernel that corrects for it, because we + have a "upgrade in place" model. We don't have a "upgrade with new + user space". + + And I seriously will refuse to take code from people who do not + understand and honor this very simple rule. + + This rule is also not going to change. + + And yes, I realize that the kernel is "special" in this respect. I'm + proud of it. + + I have seen, and can point to, lots of projects that go "We need to + break that use case in order to make progress" or "you relied on + undocumented behavior, it sucks to be you" or "there's a better way= to + do what you want to do, and you have to change to that new better + way", and I simply don't think that's acceptable outside of very ea= rly + alpha releases that have experimental users that know what they sig= ned + up for. The kernel hasn't been in that situation for the last two + decades. + + We do API breakage _inside_ the kernel all the time. We will fix + internal problems by saying "you now need to do XYZ", but then it's + about internal kernel API's, and the people who do that then also + obviously have to fix up all the in-kernel users of that API. Nobody + can say "I now broke the API you used, and now _you_ need to fix it + up". Whoever broke something gets to fix it too. + + And we simply do not break user space. + + * From `2020-05-21 + `_:: + + The rules about regressions have never been about any kind of + documented behavior, or where the code lives. + + The rules about regressions are always about "breaks user workflow". + + Users are literally the _only_ thing that matters. + + No amount of "you shouldn't have used this" or "that behavior was + undefined, it's your own fault your app broke" or "that used to work + simply because of a kernel bug" is at all relevant. + + Now, reality is never entirely black-and-white. So we've had things + like "serious security issue" etc that just forces us to make chang= es + that may break user space. But even then the rule is that we don't + really have other options that would allow things to continue. + + And obviously, if users take years to even notice that something + broke, or if we have sane ways to work around the breakage that + doesn't make for too much trouble for users (ie "ok, there are a + handful of users, and they can use a kernel command line to work + around it" kind of things) we've also been a bit less strict. + + But no, "that was documented to be broken" (whether it's because the + code was in staging or because the man-page said something else) is + irrelevant. If staging code is so useful that people end up using i= t, + that means that it's basically regular kernel code with a flag sayi= ng + "please clean this up". + + The other side of the coin is that people who talk about "API + stability" are entirely wrong. API's don't matter either. You can m= ake + any changes to an API you like - as long as nobody notices. + + Again, the regression rule is not about documentation, not about + API's, and not about the phase of the moon. + + It's entirely about "we caused problems for user space that used to= work". + + * From `2012-07-06 `_:: + + > Now this got me wondering if Debian _unstable_ actually qualifies= as a + > standard distro userspace. + + Oh, if the kernel breaks some standard user space, that counts. Tons + of people run Debian unstable (and from my limited interactions with + it, for damn good reasons: -stable tends to run so old versions of + everything that you have to sometimes deal with cuneiform writing w= hen + using it) + + * From `2017-11-05 + `_:: + + And our regression rule has never been "behavior doesn't change". + That would mean that we could never make any changes at all. + + For example, we do things like add new error handling etc all the + time, which we then sometimes even add tests for in our kselftest + directory. + + So clearly behavior changes all the time and we don't consider that= a + regression per se. + + The rule for a regression for the kernel is that some real user + workflow breaks. Not some test. Not a "look, I used to be able to do + X, now I can't". + + * From `2018-08-03 + `_:: + + YOU ARE MISSING THE #1 KERNEL RULE. + + We do not regress, and we do not regress exactly because your are 1= 00% wrong. + + And the reason you state for your opinion is in fact exactly *WHY* = you + are wrong. + + Your "good reasons" are pure and utter garbage. + + The whole point of "we do not regress" is so that people can upgrade + the kernel and never have to worry about it. + + > Kernel had a bug which has been fixed + + That is *ENTIRELY* immaterial. + + Guys, whether something was buggy or not DOES NOT MATTER. + + Why? + + Bugs happen. That's a fact of life. Arguing that "we had to break + something because we were fixing a bug" is completely insane. We fix + tens of bugs every single day, thinking that "fixing a bug" means t= hat + we can break something is simply NOT TRUE. + + So bugs simply aren't even relevant to the discussion. They happen, + they get found, they get fixed, and it has nothing to do with "we + break users". + + Because the only thing that matters IS THE USER. + + How hard is that to understand? + + Anybody who uses "but it was buggy" as an argument is entirely miss= ing + the point. As far as the USER was concerned, it wasn't buggy - it + worked for him/her. + + Maybe it worked *because* the user had taken the bug into account, + maybe it worked because the user didn't notice - again, it doesn't + matter. It worked for the user. + + Breaking a user workflow for a "bug" is absolutely the WORST reason + for breakage you can imagine. + + It's basically saying "I took something that worked, and I broke it, + but now it's better". Do you not see how f*cking insane that statem= ent + is? + + And without users, your program is not a program, it's a pointless + piece of code that you might as well throw away. + + Seriously. This is *why* the #1 rule for kernel development is "we + don't break users". Because "I fixed a bug" is absolutely NOT AN + ARGUMENT if that bug fix broke a user setup. You actually introduce= d a + MUCH BIGGER bug by "fixing" something that the user clearly didn't + even care about. + + And dammit, we upgrade the kernel ALL THE TIME without upgrading any + other programs at all. It is absolutely required, because flag-days + and dependencies are horribly bad. + + And it is also required simply because I as a kernel developer do n= ot + upgrade random other tools that I don't even care about as I develop + the kernel, and I want any of my users to feel safe doing the same + time. + + So no. Your rule is COMPLETELY wrong. If you cannot upgrade a kernel + without upgrading some other random binary, then we have a problem. + + * From `2021-06-05 + `_:: + + THERE ARE NO VALID ARGUMENTS FOR REGRESSIONS. + + Honestly, security people need to understand that "not working" is = not + a success case of security. It's a failure case. + + Yes, "not working" may be secure. But security in that case is *poi= ntless*. + + * From `2021-07-30 + `_:: + + But we have the policy that regressions aren't about documentation = or + even sane behavior. + + Regressions are about whether a user application broke in a noticea= ble way. + + * From `2011-05-06 (1/3) + `_:: + + Binary compatibility is more important. + + And if binaries don't use the interface to parse the format (or just + parse it wrongly - see the fairly recent example of adding uuid's to + /proc/self/mountinfo), then it's a regression. + + And regressions get reverted, unless there are security issues or + similar that makes us go "Oh Gods, we really have to break things". + + I don't understand why this simple logic is so hard for some kernel + developers to understand. Reality matters. Your personal wishes mat= ter + NOT AT ALL. + + If you made an interface that can be used without parsing the + interface description, then we're stuck with the interface. Theory + simply doesn't matter. + + You could help fix the tools, and try to avoid the compatibility + issues that way. There aren't that many of them. + + * From `2011-05-06 (2/3) + `_:: + + it's clearly NOT an internal tracepoint. By definition. It's being + used by powertop. + + * From `2011-05-06 (3/3) + `_:: + + We have programs that use that ABI and thus it's a regression if th= ey break. + + * From `2006-02-21 + `_:: + + The fact is, if changing the kernel breaks user-space, it's a regre= ssion. + IT DOES NOT MATTER WHETHER IT'S IN /sbin/hotplug OR ANYTHING ELSE. = If it + was installed by a distribution, it's user-space. If it got install= ed by + "vmlinux", it's the kernel. + + The only piece of user-space code we ship with the kernel is the sy= stem + call trampoline etc that the kernel sets up. THOSE interfaces we can + really change, because it changes with the kernel. + + * From `2019-09-15 + `_:: + + One _particularly_ last-minute revert is the top-most commit (ignor= ing + the version change itself) done just before the release, and while + it's very annoying, it's perhaps also instructive. + + What's instructive about it is that I reverted a commit that wasn't + actually buggy. In fact, it was doing exactly what it set out to do, + and did it very well. In fact it did it _so_ well that the much + improved IO patterns it caused then ended up revealing a user-visib= le + regression due to a real bug in a completely unrelated area. + + The actual details of that regression are not the reason I point th= at + revert out as instructive, though. It's more that it's an instructi= ve + example of what counts as a regression, and what the whole "no + regressions" kernel rule means. The reverted commit didn't change a= ny + API's, and it didn't introduce any new bugs. But it ended up exposi= ng + another problem, and as such caused a kernel upgrade to fail for a + user. So it got reverted. + + The point here being that we revert based on user-reported _behavio= r_, + not based on some "it changes the ABI" or "it caused a bug" concept. + The problem was really pre-existing, and it just didn't happen to + trigger before. The better IO patterns introduced by the change just + happened to expose an old bug, and people had grown to depend on the + previously benign behavior of that old issue. + + And never fear, we'll re-introduce the fix that improved on the IO + patterns once we've decided just how to handle the fact that we had= a + bad interaction with an interface that people had then just happened + to rely on incidental behavior for before. It's just that we'll have + to hash through how to do that (there are no less than three differ= ent + patches by three different developers being discussed, and there mi= ght + be more coming...). In the meantime, I reverted the thing that expo= sed + the problem to users for this release, even if I hope it will be + re-introduced (perhaps even backported as a stable patch) once we h= ave + consensus about the issue it exposed. + + Take-away from the whole thing: it's not about whether you change t= he + kernel-userspace ABI, or fix a bug, or about whether the old code + "should never have worked in the first place". It's about whether + something breaks existing users' workflow. + + Anyway, that was my little aside on the whole regression thing. Si= nce + it's that "first rule of kernel programming", I felt it is perhaps + worth just bringing it up every once in a while. diff --git a/MAINTAINERS b/MAINTAINERS index ea3e6c914384..03bb629302cb 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -10438,6 +10438,7 @@ KERNEL REGRESSIONS M: Thorsten Leemhuis L: regressions@lists.linux.dev S: Supported +F: Documentation/admin-guide/regressions.rst =20 KERNEL SELFTEST FRAMEWORK M: Shuah Khan --=20 2.31.1 From nobody Tue Jun 30 04:30:21 2026 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 84DD3C433F5 for ; Tue, 25 Jan 2022 11:55:31 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1380687AbiAYLyu (ORCPT ); Tue, 25 Jan 2022 06:54:50 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:56394 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1383256AbiAYLpL (ORCPT ); Tue, 25 Jan 2022 06:45:11 -0500 Received: from wp530.webpack.hosteurope.de (wp530.webpack.hosteurope.de [IPv6:2a01:488:42:1000:50ed:8234::]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id CDE31C06175F; Tue, 25 Jan 2022 03:44:17 -0800 (PST) Received: from ip4d173d02.dynamic.kabel-deutschland.de ([77.23.61.2] helo=truhe.fritz.box); authenticated by wp530.webpack.hosteurope.de running ExIM with esmtpsa (TLS1.3:ECDHE_RSA_AES_256_GCM_SHA384:256) id 1nCKFD-0001tA-UU; Tue, 25 Jan 2022 12:44:16 +0100 From: Thorsten Leemhuis To: linux-doc@vger.kernel.org, Linus Torvalds Cc: workflows@vger.kernel.org, Linux Kernel Mailing List , Jonathan Corbet , Randy Dunlap , regressions@lists.linux.dev, Greg Kroah-Hartman , Lukas Bulwahn Subject: [PATCH v3 2/2] docs: regressions.rst: rules of thumb for handling regressions Date: Tue, 25 Jan 2022 12:44:14 +0100 Message-Id: <9eb100798e963e6f47c28cba81668d4068843876.1643110442.git.linux@leemhuis.info> X-Mailer: git-send-email 2.31.1 In-Reply-To: References: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable X-bounce-key: webpack.hosteurope.de;linux@leemhuis.info;1643111057;57976a3c; X-HE-SMSGID: 1nCKFD-0001tA-UU Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Type: text/plain; charset="utf-8" Add a section with a few rules of thumb about how quickly regressions should be fixed. They are written after studying the quotes from Linus found in the modified document and especially influenced by statements like "Users are literally the _only_ thing that matters" and "without users, your program is not a program, it's a pointless piece of code that you might as well throw away". The author interpreted those in perspective to how the various Linux kernel series are maintained and what those practices might mean for users running into a regression when updating. That for example lead to the paragraph starting with "Aim to get fixes for regressions mainlined within one week after identifying the culprit, if the regression was introduced in a stable/longterm release or the devel cycle for the latest mainline release". This is a pretty high bar, but on the other hand needed to not leave users out in the cold for to long. This can quickly happen, as the previous stable series is normally stamped "End of Life" about three or four weeks after a new mainline release, which makes a lot of users switch during this timeframe. Any of them risk running into regressions not promptly fixed; even worse, once the previous stable series is EOLed for real, users that face a regression might be left with only three options: (1) continue running an outdated and thus potentially insecure kernel version from an abandoned stable series (2) run the kernel with the regression (3) downgrade to an earlier longterm series still supported This is better avoided, as (1) puts users and their data in danger, (2) will only be possible if it's a minor regression that doesn't interfere with booting or serious usage, and (3) might be regression itself or impossible on the particular machine, as some users will require drivers or features only introduced after the latest longterm series took of. In the end this lead to the "Aim to fix regression within one week" part. It also is the reason for the "Try to resolve any regressions introduced in the current development cycle before its end.". Signed-off-by: Thorsten Leemhuis CC: Linus Torvalds Acked-by: Greg Kroah-Hartman Reviewed-by: Lukas Bulwahn --- Hi! A lot of developers are doing a good job in fixing regressions, but I noticed it sometimes takes many weeks to get even simple fixes for regressions merged. Most of the time this is due to one of these factors: * it takes a long time to get the fix ready, as some developers apparently don't prioritize work on fixing regressions * fully developed fixes linger in git trees of maintainers for weeks, sometimes even without the fix being in linux-next This afaics is especially a problem for regressions introduced in mainline, but only found after a recent release or a stable kernel series derived from it. Sometimes fixes for these regressions are even left lying around for weeks until the next merge window, which contributes to a huge pile of fixes getting backported to stable and longterm releases during or shortly after merge windows. Asking developers to speed things up rarely helped, as people have different options on how fast regression fixes need to be developed and merged upstream. That's why it would be a great help to my work as regression tracker if we had some rough written down guidelines for handling regressions, as proposed by the patch below. I'm well aware that the text sets a pretty high bar. That's because I approached the problem primarily from the point of a user, as can be seen by the patch description. The text added by this patch likely will lead to some discussions, that's why I submit it separately from the rest of the new document on regressions, which is found in patch 1/2; I also CCed Linus on this patch and hope they state his opinion or ACK is. In the end I can easily tone this down or write something totally different: that's totally fine for me, I'm mainly interested in having some expectations roughly documented to get everyone on the same page. --- Documentation/admin-guide/regressions.rst | 81 +++++++++++++++++++++++ 1 file changed, 81 insertions(+) diff --git a/Documentation/admin-guide/regressions.rst b/Documentation/admi= n-guide/regressions.rst index 837b1658d149..903f3cb69b75 100644 --- a/Documentation/admin-guide/regressions.rst +++ b/Documentation/admin-guide/regressions.rst @@ -46,6 +46,10 @@ ensuring all regressions are acted upon in a timely mann= er. The important bits for people fixing regressions =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D =20 +Try to fix regressions quickly once the culprit has been identified; fixes= for +most regressions should be merged within two weeks, but some need to be +addressed within two or three days. + When submitting fixes for regressions, add "Link:" tags pointing to all pl= aces where the issue was reported, as tools like the Linux kernel regression bot 'regzbot' heavily rely on them. These pointers are also of great value when @@ -257,6 +261,83 @@ In case you are having trouble, simply forward the rep= ort or a pointer to it without further ado to Thorsten Leemhuis (regressions@leemhuis.info), he t= hen will handle things. =20 +How quickly should regressions be fixed? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Developers should fix any reported regression as quickly as possible, to p= rovide +affected users with a solution in a timely manner and prevent more users f= rom +running into the issue; nevertheless developers need to take enough time a= nd +care to ensure regression fixes do not cause additional damage. + +In the end though, developers should give their best to prevent users from +running into situations where a regression leaves them only three options:= 'run +a kernel with a regression that seriously impacts usage', 'continue runnin= g an +outdated and thus potentially insecure kernel version for more than two we= eks +after a regression's culprit was identified', and 'downgrade to a still +supported kernel series that's missing required features'. + +How to realize this depends a lot on the situation. Here are a few rules of +thumb for developers, in order or importance: + + * Prioritize work on handling regression reports and fixing regression ov= er all + other Linux kernel work, unless the latter concerns acute security issu= es or + bugs causing data loss or damage. + + * Always consider reverting the culprit commits and reapplying them later + together with necessary fixes, as this might be the least dangerous and + quickest way to fix a regression. + + * Try to resolve any regressions introduced in the current development be= fore + its end. If you fear a fix might be too risky to apply only days before= a new + mainline release, let Linus decide: submit the fix separately to him as= soon + as possible with the explanation of the situation. He then can make a c= all + and postpone the release if necessary, for example if multiple such cha= nges + show up in his inbox. + + * Address regressions in stable, longterm, or proper mainline releases wi= th + more urgency than regressions in mainline pre-releases. That changes af= ter + the release of the fifth pre-release, aka '-rc5': mainline then becomes= as + important, to ensure all the improvements and fixes are ideally tested + together for at least one week before Linus releases a new mainline ver= sion. + + * Fix regressions within two or three days, if they are critical for some + reason -- for example, if the issue is likely to affect many users of t= he + kernel series in question on all or certain architectures. This thus in= cludes + fixes for compile errors in mainline, as they otherwise might prevent + testers and continuous integration systems from doing their work. + + * Aim to merge regression fixes into mainline within one week after the c= ulprit + was identified, if the regression was introduced in a stable/longterm r= elease + or the development cycle for the latest mainline release (say v5.14). If + possible, try to address the issue even quicker, if the previous stable + series (v5.13.y) will be abandoned soon or already was stamped 'End-of-= Life' + (EOL) -- this usually happens about three to four weeks after a new mai= nline + release. + + * Try to fix all other regressions within two weeks after the culprit was + found. Two or three additional weeks are acceptable for performance + regressions and other issues which are annoying, but don't prevent anyo= ne + from running Linux (unless it's an issue in the current development cyc= le, + as those should ideally be addressed before the release). A few weeks in + total are acceptable if a regression can only be fixed with a risky cha= nge + and at the same time is affecting only a few users; as much time is + also okay if the regression is already present in the second newest lon= gterm + kernel series. + +Note: The aforementioned time frames for resolving regressions are meant to +include getting the fix tested, reviewed, and merged into mainline, ideall= y with +the fix being in linux-next for two days. + +Developers thus need to get their fixes finished earlier, as each of these= steps +takes additional time. + +Subsystem maintainers are expected to assist in reaching those periods by = doing +timely reviews and quick handling of accepted patches. They thus might hav= e to +send git-pull requests earlier or more often than usual; depending on the = fix, +it might even be acceptable to skip testing in linux-next. Especially fixe= s for +regressions in stable and longterm kernels need to be handled quickly, as = fixes +need to be merged in mainline before they can be backported to older serie= s. + Are really all regressions fixed? ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ =20 --=20 2.31.1