About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / bpf / bpf_devel_QA.txt




Custom Search

Based on kernel version 4.16.1. Page generated on 2018-04-09 11:52 EST.

1	This document provides information for the BPF subsystem about various
2	workflows related to reporting bugs, submitting patches, and queueing
3	patches for stable kernels.
4	
5	For general information about submitting patches, please refer to
6	Documentation/process/. This document only describes additional specifics
7	related to BPF.
8	
9	Reporting bugs:
10	---------------
11	
12	Q: How do I report bugs for BPF kernel code?
13	
14	A: Since all BPF kernel development as well as bpftool and iproute2 BPF
15	   loader development happens through the netdev kernel mailing list,
16	   please report any found issues around BPF to the following mailing
17	   list:
18	
19	     netdev@vger.kernel.org
20	
21	   This may also include issues related to XDP, BPF tracing, etc.
22	
23	   Given netdev has a high volume of traffic, please also add the BPF
24	   maintainers to Cc (from kernel MAINTAINERS file):
25	
26	     Alexei Starovoitov <ast@kernel.org>
27	     Daniel Borkmann <daniel@iogearbox.net>
28	
29	   In case a buggy commit has already been identified, make sure to keep
30	   the actual commit authors in Cc as well for the report. They can
31	   typically be identified through the kernel's git tree.
32	
33	   Please do *not* report BPF issues to bugzilla.kernel.org since it
34	   is a guarantee that the reported issue will be overlooked.
35	
36	Submitting patches:
37	-------------------
38	
39	Q: To which mailing list do I need to submit my BPF patches?
40	
41	A: Please submit your BPF patches to the netdev kernel mailing list:
42	
43	     netdev@vger.kernel.org
44	
45	   Historically, BPF came out of networking and has always been maintained
46	   by the kernel networking community. Although these days BPF touches
47	   many other subsystems as well, the patches are still routed mainly
48	   through the networking community.
49	
50	   In case your patch has changes in various different subsystems (e.g.
51	   tracing, security, etc), make sure to Cc the related kernel mailing
52	   lists and maintainers from there as well, so they are able to review
53	   the changes and provide their Acked-by's to the patches.
54	
55	Q: Where can I find patches currently under discussion for BPF subsystem?
56	
57	A: All patches that are Cc'ed to netdev are queued for review under netdev
58	   patchwork project:
59	
60	     http://patchwork.ozlabs.org/project/netdev/list/
61	
62	   Those patches which target BPF, are assigned to a 'bpf' delegate for
63	   further processing from BPF maintainers. The current queue with
64	   patches under review can be found at:
65	
66	     https://patchwork.ozlabs.org/project/netdev/list/?delegate=77147
67	
68	   Once the patches have been reviewed by the BPF community as a whole
69	   and approved by the BPF maintainers, their status in patchwork will be
70	   changed to 'Accepted' and the submitter will be notified by mail. This
71	   means that the patches look good from a BPF perspective and have been
72	   applied to one of the two BPF kernel trees.
73	
74	   In case feedback from the community requires a respin of the patches,
75	   their status in patchwork will be set to 'Changes Requested', and purged
76	   from the current review queue. Likewise for cases where patches would
77	   get rejected or are not applicable to the BPF trees (but assigned to
78	   the 'bpf' delegate).
79	
80	Q: How do the changes make their way into Linux?
81	
82	A: There are two BPF kernel trees (git repositories). Once patches have
83	   been accepted by the BPF maintainers, they will be applied to one
84	   of the two BPF trees:
85	
86	     https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git/
87	     https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/
88	
89	   The bpf tree itself is for fixes only, whereas bpf-next for features,
90	   cleanups or other kind of improvements ("next-like" content). This is
91	   analogous to net and net-next trees for networking. Both bpf and
92	   bpf-next will only have a master branch in order to simplify against
93	   which branch patches should get rebased to.
94	
95	   Accumulated BPF patches in the bpf tree will regularly get pulled
96	   into the net kernel tree. Likewise, accumulated BPF patches accepted
97	   into the bpf-next tree will make their way into net-next tree. net and
98	   net-next are both run by David S. Miller. From there, they will go
99	   into the kernel mainline tree run by Linus Torvalds. To read up on the
100	   process of net and net-next being merged into the mainline tree, see
101	   the netdev FAQ under:
102	
103	     Documentation/networking/netdev-FAQ.txt
104	
105	   Occasionally, to prevent merge conflicts, we might send pull requests
106	   to other trees (e.g. tracing) with a small subset of the patches, but
107	   net and net-next are always the main trees targeted for integration.
108	
109	   The pull requests will contain a high-level summary of the accumulated
110	   patches and can be searched on netdev kernel mailing list through the
111	   following subject lines (yyyy-mm-dd is the date of the pull request):
112	
113	     pull-request: bpf yyyy-mm-dd
114	     pull-request: bpf-next yyyy-mm-dd
115	
116	Q: How do I indicate which tree (bpf vs. bpf-next) my patch should be
117	   applied to?
118	
119	A: The process is the very same as described in the netdev FAQ, so
120	   please read up on it. The subject line must indicate whether the
121	   patch is a fix or rather "next-like" content in order to let the
122	   maintainers know whether it is targeted at bpf or bpf-next.
123	
124	   For fixes eventually landing in bpf -> net tree, the subject must
125	   look like:
126	
127	     git format-patch --subject-prefix='PATCH bpf' start..finish
128	
129	   For features/improvements/etc that should eventually land in
130	   bpf-next -> net-next, the subject must look like:
131	
132	     git format-patch --subject-prefix='PATCH bpf-next' start..finish
133	
134	   If unsure whether the patch or patch series should go into bpf
135	   or net directly, or bpf-next or net-next directly, it is not a
136	   problem either if the subject line says net or net-next as target.
137	   It is eventually up to the maintainers to do the delegation of
138	   the patches.
139	
140	   If it is clear that patches should go into bpf or bpf-next tree,
141	   please make sure to rebase the patches against those trees in
142	   order to reduce potential conflicts.
143	
144	   In case the patch or patch series has to be reworked and sent out
145	   again in a second or later revision, it is also required to add a
146	   version number (v2, v3, ...) into the subject prefix:
147	
148	     git format-patch --subject-prefix='PATCH net-next v2' start..finish
149	
150	   When changes have been requested to the patch series, always send the
151	   whole patch series again with the feedback incorporated (never send
152	   individual diffs on top of the old series).
153	
154	Q: What does it mean when a patch gets applied to bpf or bpf-next tree?
155	
156	A: It means that the patch looks good for mainline inclusion from
157	   a BPF point of view.
158	
159	   Be aware that this is not a final verdict that the patch will
160	   automatically get accepted into net or net-next trees eventually:
161	
162	   On the netdev kernel mailing list reviews can come in at any point
163	   in time. If discussions around a patch conclude that they cannot
164	   get included as-is, we will either apply a follow-up fix or drop
165	   them from the trees entirely. Therefore, we also reserve to rebase
166	   the trees when deemed necessary. After all, the purpose of the tree
167	   is to i) accumulate and stage BPF patches for integration into trees
168	   like net and net-next, and ii) run extensive BPF test suite and
169	   workloads on the patches before they make their way any further.
170	
171	   Once the BPF pull request was accepted by David S. Miller, then
172	   the patches end up in net or net-next tree, respectively, and
173	   make their way from there further into mainline. Again, see the
174	   netdev FAQ for additional information e.g. on how often they are
175	   merged to mainline.
176	
177	Q: How long do I need to wait for feedback on my BPF patches?
178	
179	A: We try to keep the latency low. The usual time to feedback will
180	   be around 2 or 3 business days. It may vary depending on the
181	   complexity of changes and current patch load.
182	
183	Q: How often do you send pull requests to major kernel trees like
184	   net or net-next?
185	
186	A: Pull requests will be sent out rather often in order to not
187	   accumulate too many patches in bpf or bpf-next.
188	
189	   As a rule of thumb, expect pull requests for each tree regularly
190	   at the end of the week. In some cases pull requests could additionally
191	   come also in the middle of the week depending on the current patch
192	   load or urgency.
193	
194	Q: Are patches applied to bpf-next when the merge window is open?
195	
196	A: For the time when the merge window is open, bpf-next will not be
197	   processed. This is roughly analogous to net-next patch processing,
198	   so feel free to read up on the netdev FAQ about further details.
199	
200	   During those two weeks of merge window, we might ask you to resend
201	   your patch series once bpf-next is open again. Once Linus released
202	   a v*-rc1 after the merge window, we continue processing of bpf-next.
203	
204	   For non-subscribers to kernel mailing lists, there is also a status
205	   page run by David S. Miller on net-next that provides guidance:
206	
207	     http://vger.kernel.org/~davem/net-next.html
208	
209	Q: I made a BPF verifier change, do I need to add test cases for
210	   BPF kernel selftests?
211	
212	A: If the patch has changes to the behavior of the verifier, then yes,
213	   it is absolutely necessary to add test cases to the BPF kernel
214	   selftests suite. If they are not present and we think they are
215	   needed, then we might ask for them before accepting any changes.
216	
217	   In particular, test_verifier.c is tracking a high number of BPF test
218	   cases, including a lot of corner cases that LLVM BPF back end may
219	   generate out of the restricted C code. Thus, adding test cases is
220	   absolutely crucial to make sure future changes do not accidentally
221	   affect prior use-cases. Thus, treat those test cases as: verifier
222	   behavior that is not tracked in test_verifier.c could potentially
223	   be subject to change.
224	
225	Q: When should I add code to samples/bpf/ and when to BPF kernel
226	   selftests?
227	
228	A: In general, we prefer additions to BPF kernel selftests rather than
229	   samples/bpf/. The rationale is very simple: kernel selftests are
230	   regularly run by various bots to test for kernel regressions.
231	
232	   The more test cases we add to BPF selftests, the better the coverage
233	   and the less likely it is that those could accidentally break. It is
234	   not that BPF kernel selftests cannot demo how a specific feature can
235	   be used.
236	
237	   That said, samples/bpf/ may be a good place for people to get started,
238	   so it might be advisable that simple demos of features could go into
239	   samples/bpf/, but advanced functional and corner-case testing rather
240	   into kernel selftests.
241	
242	   If your sample looks like a test case, then go for BPF kernel selftests
243	   instead!
244	
245	Q: When should I add code to the bpftool?
246	
247	A: The main purpose of bpftool (under tools/bpf/bpftool/) is to provide
248	   a central user space tool for debugging and introspection of BPF programs
249	   and maps that are active in the kernel. If UAPI changes related to BPF
250	   enable for dumping additional information of programs or maps, then
251	   bpftool should be extended as well to support dumping them.
252	
253	Q: When should I add code to iproute2's BPF loader?
254	
255	A: For UAPI changes related to the XDP or tc layer (e.g. cls_bpf), the
256	   convention is that those control-path related changes are added to
257	   iproute2's BPF loader as well from user space side. This is not only
258	   useful to have UAPI changes properly designed to be usable, but also
259	   to make those changes available to a wider user base of major
260	   downstream distributions.
261	
262	Q: Do you accept patches as well for iproute2's BPF loader?
263	
264	A: Patches for the iproute2's BPF loader have to be sent to:
265	
266	     netdev@vger.kernel.org
267	
268	   While those patches are not processed by the BPF kernel maintainers,
269	   please keep them in Cc as well, so they can be reviewed.
270	
271	   The official git repository for iproute2 is run by Stephen Hemminger
272	   and can be found at:
273	
274	     https://git.kernel.org/pub/scm/linux/kernel/git/shemminger/iproute2.git/
275	
276	   The patches need to have a subject prefix of '[PATCH iproute2 master]'
277	   or '[PATCH iproute2 net-next]'. 'master' or 'net-next' describes the
278	   target branch where the patch should be applied to. Meaning, if kernel
279	   changes went into the net-next kernel tree, then the related iproute2
280	   changes need to go into the iproute2 net-next branch, otherwise they
281	   can be targeted at master branch. The iproute2 net-next branch will get
282	   merged into the master branch after the current iproute2 version from
283	   master has been released.
284	
285	   Like BPF, the patches end up in patchwork under the netdev project and
286	   are delegated to 'shemminger' for further processing:
287	
288	     http://patchwork.ozlabs.org/project/netdev/list/?delegate=389
289	
290	Q: What is the minimum requirement before I submit my BPF patches?
291	
292	A: When submitting patches, always take the time and properly test your
293	   patches *prior* to submission. Never rush them! If maintainers find
294	   that your patches have not been properly tested, it is a good way to
295	   get them grumpy. Testing patch submissions is a hard requirement!
296	
297	   Note, fixes that go to bpf tree *must* have a Fixes: tag included. The
298	   same applies to fixes that target bpf-next, where the affected commit
299	   is in net-next (or in some cases bpf-next). The Fixes: tag is crucial
300	   in order to identify follow-up commits and tremendously helps for people
301	   having to do backporting, so it is a must have!
302	
303	   We also don't accept patches with an empty commit message. Take your
304	   time and properly write up a high quality commit message, it is
305	   essential!
306	
307	   Think about it this way: other developers looking at your code a month
308	   from now need to understand *why* a certain change has been done that
309	   way, and whether there have been flaws in the analysis or assumptions
310	   that the original author did. Thus providing a proper rationale and
311	   describing the use-case for the changes is a must.
312	
313	   Patch submissions with >1 patch must have a cover letter which includes
314	   a high level description of the series. This high level summary will
315	   then be placed into the merge commit by the BPF maintainers such that
316	   it is also accessible from the git log for future reference.
317	
318	Q: What do I need to consider when adding a new instruction or feature
319	   that would require BPF JIT and/or LLVM integration as well?
320	
321	A: We try hard to keep all BPF JITs up to date such that the same user
322	   experience can be guaranteed when running BPF programs on different
323	   architectures without having the program punt to the less efficient
324	   interpreter in case the in-kernel BPF JIT is enabled.
325	
326	   If you are unable to implement or test the required JIT changes for
327	   certain architectures, please work together with the related BPF JIT
328	   developers in order to get the feature implemented in a timely manner.
329	   Please refer to the git log (arch/*/net/) to locate the necessary
330	   people for helping out.
331	
332	   Also always make sure to add BPF test cases (e.g. test_bpf.c and
333	   test_verifier.c) for new instructions, so that they can receive
334	   broad test coverage and help run-time testing the various BPF JITs.
335	
336	   In case of new BPF instructions, once the changes have been accepted
337	   into the Linux kernel, please implement support into LLVM's BPF back
338	   end. See LLVM section below for further information.
339	
340	Stable submission:
341	------------------
342	
343	Q: I need a specific BPF commit in stable kernels. What should I do?
344	
345	A: In case you need a specific fix in stable kernels, first check whether
346	   the commit has already been applied in the related linux-*.y branches:
347	
348	     https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/
349	
350	   If not the case, then drop an email to the BPF maintainers with the
351	   netdev kernel mailing list in Cc and ask for the fix to be queued up:
352	
353	     netdev@vger.kernel.org
354	
355	   The process in general is the same as on netdev itself, see also the
356	   netdev FAQ document.
357	
358	Q: Do you also backport to kernels not currently maintained as stable?
359	
360	A: No. If you need a specific BPF commit in kernels that are currently not
361	   maintained by the stable maintainers, then you are on your own.
362	
363	   The current stable and longterm stable kernels are all listed here:
364	
365	     https://www.kernel.org/
366	
367	Q: The BPF patch I am about to submit needs to go to stable as well. What
368	   should I do?
369	
370	A: The same rules apply as with netdev patch submissions in general, see
371	   netdev FAQ under:
372	
373	     Documentation/networking/netdev-FAQ.txt
374	
375	   Never add "Cc: stable@vger.kernel.org" to the patch description, but
376	   ask the BPF maintainers to queue the patches instead. This can be done
377	   with a note, for example, under the "---" part of the patch which does
378	   not go into the git log. Alternatively, this can be done as a simple
379	   request by mail instead.
380	
381	Q: Where do I find currently queued BPF patches that will be submitted
382	   to stable?
383	
384	A: Once patches that fix critical bugs got applied into the bpf tree, they
385	   are queued up for stable submission under:
386	
387	     http://patchwork.ozlabs.org/bundle/bpf/stable/?state=*
388	
389	   They will be on hold there at minimum until the related commit made its
390	   way into the mainline kernel tree.
391	
392	   After having been under broader exposure, the queued patches will be
393	   submitted by the BPF maintainers to the stable maintainers.
394	
395	Testing patches:
396	----------------
397	
398	Q: Which BPF kernel selftests version should I run my kernel against?
399	
400	A: If you run a kernel xyz, then always run the BPF kernel selftests from
401	   that kernel xyz as well. Do not expect that the BPF selftest from the
402	   latest mainline tree will pass all the time.
403	
404	   In particular, test_bpf.c and test_verifier.c have a large number of
405	   test cases and are constantly updated with new BPF test sequences, or
406	   existing ones are adapted to verifier changes e.g. due to verifier
407	   becoming smarter and being able to better track certain things.
408	
409	LLVM:
410	-----
411	
412	Q: Where do I find LLVM with BPF support?
413	
414	A: The BPF back end for LLVM is upstream in LLVM since version 3.7.1.
415	
416	   All major distributions these days ship LLVM with BPF back end enabled,
417	   so for the majority of use-cases it is not required to compile LLVM by
418	   hand anymore, just install the distribution provided package.
419	
420	   LLVM's static compiler lists the supported targets through 'llc --version',
421	   make sure BPF targets are listed. Example:
422	
423	     $ llc --version
424	     LLVM (http://llvm.org/):
425	       LLVM version 6.0.0svn
426	       Optimized build.
427	       Default target: x86_64-unknown-linux-gnu
428	       Host CPU: skylake
429	
430	       Registered Targets:
431	         bpf    - BPF (host endian)
432	         bpfeb  - BPF (big endian)
433	         bpfel  - BPF (little endian)
434	         x86    - 32-bit X86: Pentium-Pro and above
435	         x86-64 - 64-bit X86: EM64T and AMD64
436	
437	   For developers in order to utilize the latest features added to LLVM's
438	   BPF back end, it is advisable to run the latest LLVM releases. Support
439	   for new BPF kernel features such as additions to the BPF instruction
440	   set are often developed together.
441	
442	   All LLVM releases can be found at: http://releases.llvm.org/
443	
444	Q: Got it, so how do I build LLVM manually anyway?
445	
446	A: You need cmake and gcc-c++ as build requisites for LLVM. Once you have
447	   that set up, proceed with building the latest LLVM and clang version
448	   from the git repositories:
449	
450	     $ git clone http://llvm.org/git/llvm.git
451	     $ cd llvm/tools
452	     $ git clone --depth 1 http://llvm.org/git/clang.git
453	     $ cd ..; mkdir build; cd build
454	     $ cmake .. -DLLVM_TARGETS_TO_BUILD="BPF;X86" \
455	                -DBUILD_SHARED_LIBS=OFF           \
456	                -DCMAKE_BUILD_TYPE=Release        \
457	                -DLLVM_BUILD_RUNTIME=OFF
458	     $ make -j $(getconf _NPROCESSORS_ONLN)
459	
460	   The built binaries can then be found in the build/bin/ directory, where
461	   you can point the PATH variable to.
462	
463	Q: Should I notify BPF kernel maintainers about issues in LLVM's BPF code
464	   generation back end or about LLVM generated code that the verifier
465	   refuses to accept?
466	
467	A: Yes, please do! LLVM's BPF back end is a key piece of the whole BPF
468	   infrastructure and it ties deeply into verification of programs from the
469	   kernel side. Therefore, any issues on either side need to be investigated
470	   and fixed whenever necessary.
471	
472	   Therefore, please make sure to bring them up at netdev kernel mailing
473	   list and Cc BPF maintainers for LLVM and kernel bits:
474	
475	     Yonghong Song <yhs@fb.com>
476	     Alexei Starovoitov <ast@kernel.org>
477	     Daniel Borkmann <daniel@iogearbox.net>
478	
479	   LLVM also has an issue tracker where BPF related bugs can be found:
480	
481	     https://bugs.llvm.org/buglist.cgi?quicksearch=bpf
482	
483	   However, it is better to reach out through mailing lists with having
484	   maintainers in Cc.
485	
486	Q: I have added a new BPF instruction to the kernel, how can I integrate
487	   it into LLVM?
488	
489	A: LLVM has a -mcpu selector for the BPF back end in order to allow the
490	   selection of BPF instruction set extensions. By default the 'generic'
491	   processor target is used, which is the base instruction set (v1) of BPF.
492	
493	   LLVM has an option to select -mcpu=probe where it will probe the host
494	   kernel for supported BPF instruction set extensions and selects the
495	   optimal set automatically.
496	
497	   For cross-compilation, a specific version can be select manually as well.
498	
499	     $ llc -march bpf -mcpu=help
500	     Available CPUs for this target:
501	
502	       generic - Select the generic processor.
503	       probe   - Select the probe processor.
504	       v1      - Select the v1 processor.
505	       v2      - Select the v2 processor.
506	     [...]
507	
508	   Newly added BPF instructions to the Linux kernel need to follow the same
509	   scheme, bump the instruction set version and implement probing for the
510	   extensions such that -mcpu=probe users can benefit from the optimization
511	   transparently when upgrading their kernels.
512	
513	   If you are unable to implement support for the newly added BPF instruction
514	   please reach out to BPF developers for help.
515	
516	   By the way, the BPF kernel selftests run with -mcpu=probe for better
517	   test coverage.
518	
519	Q: In some cases clang flag "-target bpf" is used but in other cases the
520	   default clang target, which matches the underlying architecture, is used.
521	   What is the difference and when I should use which?
522	
523	A: Although LLVM IR generation and optimization try to stay architecture
524	   independent, "-target <arch>" still has some impact on generated code:
525	
526	     - BPF program may recursively include header file(s) with file scope
527	       inline assembly codes. The default target can handle this well,
528	       while bpf target may fail if bpf backend assembler does not
529	       understand these assembly codes, which is true in most cases.
530	
531	     - When compiled without -g, additional elf sections, e.g.,
532	       .eh_frame and .rela.eh_frame, may be present in the object file
533	       with default target, but not with bpf target.
534	
535	     - The default target may turn a C switch statement into a switch table
536	       lookup and jump operation. Since the switch table is placed
537	       in the global readonly section, the bpf program will fail to load.
538	       The bpf target does not support switch table optimization.
539	       The clang option "-fno-jump-tables" can be used to disable
540	       switch table generation.
541	
542	   You should use default target when:
543	
544	     - Your program includes a header file, e.g., ptrace.h, which eventually
545	       pulls in some header files containing file scope host assembly codes.
546	     - You can add "-fno-jump-tables" to work around the switch table issue.
547	
548	   Otherwise, you can use bpf target.
549	
550	Happy BPF hacking!
Hide Line Numbers
About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Information is copyright its respective author. All material is available from the Linux Kernel Source distributed under a GPL License. This page is provided as a free service by mjmwired.net.