Commit | Line | Data |
---|---|---|
54222838 JDB |
1 | ================================= |
2 | HOWTO interact with BPF subsystem | |
3 | ================================= | |
4 | ||
34f15bf3 DB |
5 | This document provides information for the BPF subsystem about various |
6 | workflows related to reporting bugs, submitting patches, and queueing | |
7 | patches for stable kernels. | |
8 | ||
9 | For general information about submitting patches, please refer to | |
54222838 | 10 | `Documentation/process/`_. This document only describes additional specifics |
34f15bf3 DB |
11 | related to BPF. |
12 | ||
54222838 JDB |
13 | .. contents:: |
14 | :local: | |
15 | :depth: 2 | |
34f15bf3 | 16 | |
54222838 JDB |
17 | Reporting bugs |
18 | ============== | |
34f15bf3 | 19 | |
54222838 JDB |
20 | Q: How do I report bugs for BPF kernel code? |
21 | -------------------------------------------- | |
34f15bf3 | 22 | A: Since all BPF kernel development as well as bpftool and iproute2 BPF |
54222838 JDB |
23 | loader development happens through the netdev kernel mailing list, |
24 | please report any found issues around BPF to the following mailing | |
25 | list: | |
34f15bf3 | 26 | |
54222838 | 27 | netdev@vger.kernel.org |
34f15bf3 | 28 | |
54222838 | 29 | This may also include issues related to XDP, BPF tracing, etc. |
34f15bf3 | 30 | |
54222838 JDB |
31 | Given netdev has a high volume of traffic, please also add the BPF |
32 | maintainers to Cc (from kernel MAINTAINERS_ file): | |
34f15bf3 | 33 | |
54222838 JDB |
34 | * Alexei Starovoitov <ast@kernel.org> |
35 | * Daniel Borkmann <daniel@iogearbox.net> | |
34f15bf3 | 36 | |
54222838 JDB |
37 | In case a buggy commit has already been identified, make sure to keep |
38 | the actual commit authors in Cc as well for the report. They can | |
39 | typically be identified through the kernel's git tree. | |
34f15bf3 | 40 | |
54222838 JDB |
41 | **Please do NOT report BPF issues to bugzilla.kernel.org since it |
42 | is a guarantee that the reported issue will be overlooked.** | |
34f15bf3 | 43 | |
54222838 JDB |
44 | Submitting patches |
45 | ================== | |
34f15bf3 DB |
46 | |
47 | Q: To which mailing list do I need to submit my BPF patches? | |
54222838 | 48 | ------------------------------------------------------------ |
34f15bf3 DB |
49 | A: Please submit your BPF patches to the netdev kernel mailing list: |
50 | ||
54222838 | 51 | netdev@vger.kernel.org |
34f15bf3 | 52 | |
54222838 JDB |
53 | Historically, BPF came out of networking and has always been maintained |
54 | by the kernel networking community. Although these days BPF touches | |
55 | many other subsystems as well, the patches are still routed mainly | |
56 | through the networking community. | |
34f15bf3 | 57 | |
54222838 JDB |
58 | In case your patch has changes in various different subsystems (e.g. |
59 | tracing, security, etc), make sure to Cc the related kernel mailing | |
60 | lists and maintainers from there as well, so they are able to review | |
61 | the changes and provide their Acked-by's to the patches. | |
34f15bf3 DB |
62 | |
63 | Q: Where can I find patches currently under discussion for BPF subsystem? | |
54222838 | 64 | ------------------------------------------------------------------------- |
34f15bf3 | 65 | A: All patches that are Cc'ed to netdev are queued for review under netdev |
54222838 | 66 | patchwork project: |
34f15bf3 | 67 | |
54222838 | 68 | http://patchwork.ozlabs.org/project/netdev/list/ |
34f15bf3 | 69 | |
54222838 JDB |
70 | Those patches which target BPF, are assigned to a 'bpf' delegate for |
71 | further processing from BPF maintainers. The current queue with | |
72 | patches under review can be found at: | |
34f15bf3 | 73 | |
54222838 | 74 | https://patchwork.ozlabs.org/project/netdev/list/?delegate=77147 |
34f15bf3 | 75 | |
54222838 JDB |
76 | Once the patches have been reviewed by the BPF community as a whole |
77 | and approved by the BPF maintainers, their status in patchwork will be | |
78 | changed to 'Accepted' and the submitter will be notified by mail. This | |
79 | means that the patches look good from a BPF perspective and have been | |
80 | applied to one of the two BPF kernel trees. | |
34f15bf3 | 81 | |
54222838 JDB |
82 | In case feedback from the community requires a respin of the patches, |
83 | their status in patchwork will be set to 'Changes Requested', and purged | |
84 | from the current review queue. Likewise for cases where patches would | |
85 | get rejected or are not applicable to the BPF trees (but assigned to | |
86 | the 'bpf' delegate). | |
34f15bf3 DB |
87 | |
88 | Q: How do the changes make their way into Linux? | |
54222838 | 89 | ------------------------------------------------ |
34f15bf3 | 90 | A: There are two BPF kernel trees (git repositories). Once patches have |
54222838 JDB |
91 | been accepted by the BPF maintainers, they will be applied to one |
92 | of the two BPF trees: | |
34f15bf3 | 93 | |
54222838 JDB |
94 | * https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git/ |
95 | * https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/ | |
34f15bf3 | 96 | |
54222838 JDB |
97 | The bpf tree itself is for fixes only, whereas bpf-next for features, |
98 | cleanups or other kind of improvements ("next-like" content). This is | |
99 | analogous to net and net-next trees for networking. Both bpf and | |
100 | bpf-next will only have a master branch in order to simplify against | |
101 | which branch patches should get rebased to. | |
34f15bf3 | 102 | |
54222838 JDB |
103 | Accumulated BPF patches in the bpf tree will regularly get pulled |
104 | into the net kernel tree. Likewise, accumulated BPF patches accepted | |
105 | into the bpf-next tree will make their way into net-next tree. net and | |
106 | net-next are both run by David S. Miller. From there, they will go | |
107 | into the kernel mainline tree run by Linus Torvalds. To read up on the | |
108 | process of net and net-next being merged into the mainline tree, see | |
287f4fa9 TH |
109 | the :ref:`netdev-FAQ` |
110 | ||
34f15bf3 | 111 | |
34f15bf3 | 112 | |
54222838 JDB |
113 | Occasionally, to prevent merge conflicts, we might send pull requests |
114 | to other trees (e.g. tracing) with a small subset of the patches, but | |
115 | net and net-next are always the main trees targeted for integration. | |
34f15bf3 | 116 | |
54222838 JDB |
117 | The pull requests will contain a high-level summary of the accumulated |
118 | patches and can be searched on netdev kernel mailing list through the | |
119 | following subject lines (``yyyy-mm-dd`` is the date of the pull | |
120 | request):: | |
34f15bf3 | 121 | |
54222838 JDB |
122 | pull-request: bpf yyyy-mm-dd |
123 | pull-request: bpf-next yyyy-mm-dd | |
34f15bf3 | 124 | |
54222838 JDB |
125 | Q: How do I indicate which tree (bpf vs. bpf-next) my patch should be applied to? |
126 | --------------------------------------------------------------------------------- | |
34f15bf3 | 127 | |
287f4fa9 TH |
128 | A: The process is the very same as described in the :ref:`netdev-FAQ`, |
129 | so please read up on it. The subject line must indicate whether the | |
54222838 JDB |
130 | patch is a fix or rather "next-like" content in order to let the |
131 | maintainers know whether it is targeted at bpf or bpf-next. | |
34f15bf3 | 132 | |
54222838 JDB |
133 | For fixes eventually landing in bpf -> net tree, the subject must |
134 | look like:: | |
34f15bf3 | 135 | |
54222838 | 136 | git format-patch --subject-prefix='PATCH bpf' start..finish |
34f15bf3 | 137 | |
54222838 JDB |
138 | For features/improvements/etc that should eventually land in |
139 | bpf-next -> net-next, the subject must look like:: | |
34f15bf3 | 140 | |
54222838 | 141 | git format-patch --subject-prefix='PATCH bpf-next' start..finish |
34f15bf3 | 142 | |
54222838 JDB |
143 | If unsure whether the patch or patch series should go into bpf |
144 | or net directly, or bpf-next or net-next directly, it is not a | |
145 | problem either if the subject line says net or net-next as target. | |
146 | It is eventually up to the maintainers to do the delegation of | |
147 | the patches. | |
34f15bf3 | 148 | |
54222838 JDB |
149 | If it is clear that patches should go into bpf or bpf-next tree, |
150 | please make sure to rebase the patches against those trees in | |
151 | order to reduce potential conflicts. | |
34f15bf3 | 152 | |
54222838 JDB |
153 | In case the patch or patch series has to be reworked and sent out |
154 | again in a second or later revision, it is also required to add a | |
155 | version number (``v2``, ``v3``, ...) into the subject prefix:: | |
34f15bf3 | 156 | |
54222838 | 157 | git format-patch --subject-prefix='PATCH net-next v2' start..finish |
34f15bf3 | 158 | |
54222838 JDB |
159 | When changes have been requested to the patch series, always send the |
160 | whole patch series again with the feedback incorporated (never send | |
161 | individual diffs on top of the old series). | |
34f15bf3 DB |
162 | |
163 | Q: What does it mean when a patch gets applied to bpf or bpf-next tree? | |
54222838 | 164 | ----------------------------------------------------------------------- |
34f15bf3 | 165 | A: It means that the patch looks good for mainline inclusion from |
54222838 | 166 | a BPF point of view. |
34f15bf3 | 167 | |
54222838 JDB |
168 | Be aware that this is not a final verdict that the patch will |
169 | automatically get accepted into net or net-next trees eventually: | |
170 | ||
171 | On the netdev kernel mailing list reviews can come in at any point | |
172 | in time. If discussions around a patch conclude that they cannot | |
173 | get included as-is, we will either apply a follow-up fix or drop | |
174 | them from the trees entirely. Therefore, we also reserve to rebase | |
175 | the trees when deemed necessary. After all, the purpose of the tree | |
176 | is to: | |
177 | ||
178 | i) accumulate and stage BPF patches for integration into trees | |
179 | like net and net-next, and | |
34f15bf3 | 180 | |
54222838 JDB |
181 | ii) run extensive BPF test suite and |
182 | workloads on the patches before they make their way any further. | |
183 | ||
184 | Once the BPF pull request was accepted by David S. Miller, then | |
185 | the patches end up in net or net-next tree, respectively, and | |
186 | make their way from there further into mainline. Again, see the | |
287f4fa9 | 187 | :ref:`netdev-FAQ` for additional information e.g. on how often they are |
54222838 JDB |
188 | merged to mainline. |
189 | ||
190 | Q: How long do I need to wait for feedback on my BPF patches? | |
191 | ------------------------------------------------------------- | |
34f15bf3 | 192 | A: We try to keep the latency low. The usual time to feedback will |
54222838 JDB |
193 | be around 2 or 3 business days. It may vary depending on the |
194 | complexity of changes and current patch load. | |
34f15bf3 | 195 | |
54222838 JDB |
196 | Q: How often do you send pull requests to major kernel trees like net or net-next? |
197 | ---------------------------------------------------------------------------------- | |
34f15bf3 DB |
198 | |
199 | A: Pull requests will be sent out rather often in order to not | |
54222838 | 200 | accumulate too many patches in bpf or bpf-next. |
34f15bf3 | 201 | |
54222838 JDB |
202 | As a rule of thumb, expect pull requests for each tree regularly |
203 | at the end of the week. In some cases pull requests could additionally | |
204 | come also in the middle of the week depending on the current patch | |
205 | load or urgency. | |
34f15bf3 DB |
206 | |
207 | Q: Are patches applied to bpf-next when the merge window is open? | |
54222838 | 208 | ----------------------------------------------------------------- |
34f15bf3 | 209 | A: For the time when the merge window is open, bpf-next will not be |
54222838 | 210 | processed. This is roughly analogous to net-next patch processing, |
287f4fa9 | 211 | so feel free to read up on the :ref:`netdev-FAQ` about further details. |
34f15bf3 | 212 | |
54222838 JDB |
213 | During those two weeks of merge window, we might ask you to resend |
214 | your patch series once bpf-next is open again. Once Linus released | |
215 | a ``v*-rc1`` after the merge window, we continue processing of bpf-next. | |
34f15bf3 | 216 | |
54222838 JDB |
217 | For non-subscribers to kernel mailing lists, there is also a status |
218 | page run by David S. Miller on net-next that provides guidance: | |
34f15bf3 | 219 | |
54222838 | 220 | http://vger.kernel.org/~davem/net-next.html |
34f15bf3 | 221 | |
54222838 JDB |
222 | Q: Verifier changes and test cases |
223 | ---------------------------------- | |
34f15bf3 | 224 | Q: I made a BPF verifier change, do I need to add test cases for |
54222838 | 225 | BPF kernel selftests_? |
34f15bf3 DB |
226 | |
227 | A: If the patch has changes to the behavior of the verifier, then yes, | |
54222838 JDB |
228 | it is absolutely necessary to add test cases to the BPF kernel |
229 | selftests_ suite. If they are not present and we think they are | |
230 | needed, then we might ask for them before accepting any changes. | |
231 | ||
232 | In particular, test_verifier.c is tracking a high number of BPF test | |
233 | cases, including a lot of corner cases that LLVM BPF back end may | |
234 | generate out of the restricted C code. Thus, adding test cases is | |
235 | absolutely crucial to make sure future changes do not accidentally | |
236 | affect prior use-cases. Thus, treat those test cases as: verifier | |
237 | behavior that is not tracked in test_verifier.c could potentially | |
238 | be subject to change. | |
239 | ||
240 | Q: samples/bpf preference vs selftests? | |
241 | --------------------------------------- | |
242 | Q: When should I add code to `samples/bpf/`_ and when to BPF kernel | |
243 | selftests_ ? | |
244 | ||
245 | A: In general, we prefer additions to BPF kernel selftests_ rather than | |
246 | `samples/bpf/`_. The rationale is very simple: kernel selftests are | |
247 | regularly run by various bots to test for kernel regressions. | |
248 | ||
249 | The more test cases we add to BPF selftests, the better the coverage | |
250 | and the less likely it is that those could accidentally break. It is | |
251 | not that BPF kernel selftests cannot demo how a specific feature can | |
252 | be used. | |
253 | ||
254 | That said, `samples/bpf/`_ may be a good place for people to get started, | |
255 | so it might be advisable that simple demos of features could go into | |
256 | `samples/bpf/`_, but advanced functional and corner-case testing rather | |
257 | into kernel selftests. | |
258 | ||
259 | If your sample looks like a test case, then go for BPF kernel selftests | |
260 | instead! | |
34f15bf3 DB |
261 | |
262 | Q: When should I add code to the bpftool? | |
54222838 | 263 | ----------------------------------------- |
34f15bf3 | 264 | A: The main purpose of bpftool (under tools/bpf/bpftool/) is to provide |
54222838 JDB |
265 | a central user space tool for debugging and introspection of BPF programs |
266 | and maps that are active in the kernel. If UAPI changes related to BPF | |
267 | enable for dumping additional information of programs or maps, then | |
268 | bpftool should be extended as well to support dumping them. | |
34f15bf3 DB |
269 | |
270 | Q: When should I add code to iproute2's BPF loader? | |
54222838 JDB |
271 | --------------------------------------------------- |
272 | A: For UAPI changes related to the XDP or tc layer (e.g. ``cls_bpf``), | |
273 | the convention is that those control-path related changes are added to | |
274 | iproute2's BPF loader as well from user space side. This is not only | |
275 | useful to have UAPI changes properly designed to be usable, but also | |
276 | to make those changes available to a wider user base of major | |
277 | downstream distributions. | |
34f15bf3 DB |
278 | |
279 | Q: Do you accept patches as well for iproute2's BPF loader? | |
54222838 | 280 | ----------------------------------------------------------- |
34f15bf3 DB |
281 | A: Patches for the iproute2's BPF loader have to be sent to: |
282 | ||
54222838 | 283 | netdev@vger.kernel.org |
34f15bf3 | 284 | |
54222838 JDB |
285 | While those patches are not processed by the BPF kernel maintainers, |
286 | please keep them in Cc as well, so they can be reviewed. | |
34f15bf3 | 287 | |
54222838 JDB |
288 | The official git repository for iproute2 is run by Stephen Hemminger |
289 | and can be found at: | |
34f15bf3 | 290 | |
54222838 | 291 | https://git.kernel.org/pub/scm/linux/kernel/git/shemminger/iproute2.git/ |
34f15bf3 | 292 | |
54222838 JDB |
293 | The patches need to have a subject prefix of '``[PATCH iproute2 |
294 | master]``' or '``[PATCH iproute2 net-next]``'. '``master``' or | |
295 | '``net-next``' describes the target branch where the patch should be | |
296 | applied to. Meaning, if kernel changes went into the net-next kernel | |
297 | tree, then the related iproute2 changes need to go into the iproute2 | |
298 | net-next branch, otherwise they can be targeted at master branch. The | |
299 | iproute2 net-next branch will get merged into the master branch after | |
300 | the current iproute2 version from master has been released. | |
34f15bf3 | 301 | |
54222838 JDB |
302 | Like BPF, the patches end up in patchwork under the netdev project and |
303 | are delegated to 'shemminger' for further processing: | |
34f15bf3 | 304 | |
54222838 | 305 | http://patchwork.ozlabs.org/project/netdev/list/?delegate=389 |
34f15bf3 DB |
306 | |
307 | Q: What is the minimum requirement before I submit my BPF patches? | |
54222838 | 308 | ------------------------------------------------------------------ |
34f15bf3 | 309 | A: When submitting patches, always take the time and properly test your |
54222838 JDB |
310 | patches *prior* to submission. Never rush them! If maintainers find |
311 | that your patches have not been properly tested, it is a good way to | |
312 | get them grumpy. Testing patch submissions is a hard requirement! | |
313 | ||
314 | Note, fixes that go to bpf tree *must* have a ``Fixes:`` tag included. | |
315 | The same applies to fixes that target bpf-next, where the affected | |
316 | commit is in net-next (or in some cases bpf-next). The ``Fixes:`` tag is | |
317 | crucial in order to identify follow-up commits and tremendously helps | |
318 | for people having to do backporting, so it is a must have! | |
319 | ||
320 | We also don't accept patches with an empty commit message. Take your | |
321 | time and properly write up a high quality commit message, it is | |
322 | essential! | |
323 | ||
324 | Think about it this way: other developers looking at your code a month | |
325 | from now need to understand *why* a certain change has been done that | |
326 | way, and whether there have been flaws in the analysis or assumptions | |
327 | that the original author did. Thus providing a proper rationale and | |
328 | describing the use-case for the changes is a must. | |
329 | ||
330 | Patch submissions with >1 patch must have a cover letter which includes | |
331 | a high level description of the series. This high level summary will | |
332 | then be placed into the merge commit by the BPF maintainers such that | |
333 | it is also accessible from the git log for future reference. | |
334 | ||
335 | Q: Features changing BPF JIT and/or LLVM | |
336 | ---------------------------------------- | |
34f15bf3 | 337 | Q: What do I need to consider when adding a new instruction or feature |
54222838 | 338 | that would require BPF JIT and/or LLVM integration as well? |
34f15bf3 DB |
339 | |
340 | A: We try hard to keep all BPF JITs up to date such that the same user | |
54222838 JDB |
341 | experience can be guaranteed when running BPF programs on different |
342 | architectures without having the program punt to the less efficient | |
343 | interpreter in case the in-kernel BPF JIT is enabled. | |
34f15bf3 | 344 | |
54222838 JDB |
345 | If you are unable to implement or test the required JIT changes for |
346 | certain architectures, please work together with the related BPF JIT | |
347 | developers in order to get the feature implemented in a timely manner. | |
348 | Please refer to the git log (``arch/*/net/``) to locate the necessary | |
349 | people for helping out. | |
34f15bf3 | 350 | |
54222838 JDB |
351 | Also always make sure to add BPF test cases (e.g. test_bpf.c and |
352 | test_verifier.c) for new instructions, so that they can receive | |
353 | broad test coverage and help run-time testing the various BPF JITs. | |
34f15bf3 | 354 | |
54222838 JDB |
355 | In case of new BPF instructions, once the changes have been accepted |
356 | into the Linux kernel, please implement support into LLVM's BPF back | |
357 | end. See LLVM_ section below for further information. | |
34f15bf3 | 358 | |
54222838 JDB |
359 | Stable submission |
360 | ================= | |
34f15bf3 DB |
361 | |
362 | Q: I need a specific BPF commit in stable kernels. What should I do? | |
54222838 | 363 | -------------------------------------------------------------------- |
34f15bf3 | 364 | A: In case you need a specific fix in stable kernels, first check whether |
54222838 | 365 | the commit has already been applied in the related ``linux-*.y`` branches: |
34f15bf3 | 366 | |
54222838 | 367 | https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/ |
34f15bf3 | 368 | |
54222838 JDB |
369 | If not the case, then drop an email to the BPF maintainers with the |
370 | netdev kernel mailing list in Cc and ask for the fix to be queued up: | |
34f15bf3 | 371 | |
54222838 | 372 | netdev@vger.kernel.org |
34f15bf3 | 373 | |
54222838 | 374 | The process in general is the same as on netdev itself, see also the |
287f4fa9 | 375 | :ref:`netdev-FAQ`. |
34f15bf3 DB |
376 | |
377 | Q: Do you also backport to kernels not currently maintained as stable? | |
54222838 | 378 | ---------------------------------------------------------------------- |
34f15bf3 | 379 | A: No. If you need a specific BPF commit in kernels that are currently not |
54222838 | 380 | maintained by the stable maintainers, then you are on your own. |
34f15bf3 | 381 | |
54222838 | 382 | The current stable and longterm stable kernels are all listed here: |
34f15bf3 | 383 | |
54222838 | 384 | https://www.kernel.org/ |
34f15bf3 | 385 | |
54222838 JDB |
386 | Q: The BPF patch I am about to submit needs to go to stable as well |
387 | ------------------------------------------------------------------- | |
388 | What should I do? | |
34f15bf3 DB |
389 | |
390 | A: The same rules apply as with netdev patch submissions in general, see | |
287f4fa9 | 391 | the :ref:`netdev-FAQ`. |
34f15bf3 | 392 | |
54222838 JDB |
393 | Never add "``Cc: stable@vger.kernel.org``" to the patch description, but |
394 | ask the BPF maintainers to queue the patches instead. This can be done | |
395 | with a note, for example, under the ``---`` part of the patch which does | |
396 | not go into the git log. Alternatively, this can be done as a simple | |
397 | request by mail instead. | |
34f15bf3 | 398 | |
54222838 JDB |
399 | Q: Queue stable patches |
400 | ----------------------- | |
34f15bf3 | 401 | Q: Where do I find currently queued BPF patches that will be submitted |
54222838 | 402 | to stable? |
34f15bf3 DB |
403 | |
404 | A: Once patches that fix critical bugs got applied into the bpf tree, they | |
54222838 | 405 | are queued up for stable submission under: |
34f15bf3 | 406 | |
54222838 | 407 | http://patchwork.ozlabs.org/bundle/bpf/stable/?state=* |
34f15bf3 | 408 | |
54222838 JDB |
409 | They will be on hold there at minimum until the related commit made its |
410 | way into the mainline kernel tree. | |
34f15bf3 | 411 | |
54222838 JDB |
412 | After having been under broader exposure, the queued patches will be |
413 | submitted by the BPF maintainers to the stable maintainers. | |
34f15bf3 | 414 | |
54222838 JDB |
415 | Testing patches |
416 | =============== | |
34f15bf3 | 417 | |
b7a27c3a JDB |
418 | Q: How to run BPF selftests |
419 | --------------------------- | |
420 | A: After you have booted into the newly compiled kernel, navigate to | |
421 | the BPF selftests_ suite in order to test BPF functionality (current | |
422 | working directory points to the root of the cloned git tree):: | |
423 | ||
424 | $ cd tools/testing/selftests/bpf/ | |
425 | $ make | |
426 | ||
427 | To run the verifier tests:: | |
428 | ||
429 | $ sudo ./test_verifier | |
430 | ||
431 | The verifier tests print out all the current checks being | |
432 | performed. The summary at the end of running all tests will dump | |
433 | information of test successes and failures:: | |
434 | ||
435 | Summary: 418 PASSED, 0 FAILED | |
436 | ||
437 | In order to run through all BPF selftests, the following command is | |
438 | needed:: | |
439 | ||
440 | $ sudo make run_tests | |
441 | ||
442 | See the kernels selftest `Documentation/dev-tools/kselftest.rst`_ | |
443 | document for further documentation. | |
444 | ||
34f15bf3 | 445 | Q: Which BPF kernel selftests version should I run my kernel against? |
54222838 JDB |
446 | --------------------------------------------------------------------- |
447 | A: If you run a kernel ``xyz``, then always run the BPF kernel selftests | |
448 | from that kernel ``xyz`` as well. Do not expect that the BPF selftest | |
449 | from the latest mainline tree will pass all the time. | |
34f15bf3 | 450 | |
54222838 JDB |
451 | In particular, test_bpf.c and test_verifier.c have a large number of |
452 | test cases and are constantly updated with new BPF test sequences, or | |
453 | existing ones are adapted to verifier changes e.g. due to verifier | |
454 | becoming smarter and being able to better track certain things. | |
34f15bf3 | 455 | |
54222838 JDB |
456 | LLVM |
457 | ==== | |
34f15bf3 DB |
458 | |
459 | Q: Where do I find LLVM with BPF support? | |
54222838 | 460 | ----------------------------------------- |
34f15bf3 DB |
461 | A: The BPF back end for LLVM is upstream in LLVM since version 3.7.1. |
462 | ||
54222838 JDB |
463 | All major distributions these days ship LLVM with BPF back end enabled, |
464 | so for the majority of use-cases it is not required to compile LLVM by | |
465 | hand anymore, just install the distribution provided package. | |
34f15bf3 | 466 | |
54222838 JDB |
467 | LLVM's static compiler lists the supported targets through |
468 | ``llc --version``, make sure BPF targets are listed. Example:: | |
34f15bf3 DB |
469 | |
470 | $ llc --version | |
471 | LLVM (http://llvm.org/): | |
472 | LLVM version 6.0.0svn | |
473 | Optimized build. | |
474 | Default target: x86_64-unknown-linux-gnu | |
475 | Host CPU: skylake | |
476 | ||
477 | Registered Targets: | |
478 | bpf - BPF (host endian) | |
479 | bpfeb - BPF (big endian) | |
480 | bpfel - BPF (little endian) | |
481 | x86 - 32-bit X86: Pentium-Pro and above | |
482 | x86-64 - 64-bit X86: EM64T and AMD64 | |
483 | ||
54222838 JDB |
484 | For developers in order to utilize the latest features added to LLVM's |
485 | BPF back end, it is advisable to run the latest LLVM releases. Support | |
486 | for new BPF kernel features such as additions to the BPF instruction | |
487 | set are often developed together. | |
34f15bf3 | 488 | |
54222838 | 489 | All LLVM releases can be found at: http://releases.llvm.org/ |
34f15bf3 DB |
490 | |
491 | Q: Got it, so how do I build LLVM manually anyway? | |
54222838 | 492 | -------------------------------------------------- |
34f15bf3 | 493 | A: You need cmake and gcc-c++ as build requisites for LLVM. Once you have |
54222838 JDB |
494 | that set up, proceed with building the latest LLVM and clang version |
495 | from the git repositories:: | |
34f15bf3 DB |
496 | |
497 | $ git clone http://llvm.org/git/llvm.git | |
498 | $ cd llvm/tools | |
499 | $ git clone --depth 1 http://llvm.org/git/clang.git | |
500 | $ cd ..; mkdir build; cd build | |
501 | $ cmake .. -DLLVM_TARGETS_TO_BUILD="BPF;X86" \ | |
502 | -DBUILD_SHARED_LIBS=OFF \ | |
503 | -DCMAKE_BUILD_TYPE=Release \ | |
504 | -DLLVM_BUILD_RUNTIME=OFF | |
505 | $ make -j $(getconf _NPROCESSORS_ONLN) | |
506 | ||
54222838 JDB |
507 | The built binaries can then be found in the build/bin/ directory, where |
508 | you can point the PATH variable to. | |
34f15bf3 | 509 | |
54222838 JDB |
510 | Q: Reporting LLVM BPF issues |
511 | ---------------------------- | |
34f15bf3 | 512 | Q: Should I notify BPF kernel maintainers about issues in LLVM's BPF code |
54222838 JDB |
513 | generation back end or about LLVM generated code that the verifier |
514 | refuses to accept? | |
515 | ||
516 | A: Yes, please do! | |
34f15bf3 | 517 | |
54222838 JDB |
518 | LLVM's BPF back end is a key piece of the whole BPF |
519 | infrastructure and it ties deeply into verification of programs from the | |
520 | kernel side. Therefore, any issues on either side need to be investigated | |
521 | and fixed whenever necessary. | |
34f15bf3 | 522 | |
54222838 JDB |
523 | Therefore, please make sure to bring them up at netdev kernel mailing |
524 | list and Cc BPF maintainers for LLVM and kernel bits: | |
34f15bf3 | 525 | |
54222838 JDB |
526 | * Yonghong Song <yhs@fb.com> |
527 | * Alexei Starovoitov <ast@kernel.org> | |
528 | * Daniel Borkmann <daniel@iogearbox.net> | |
34f15bf3 | 529 | |
54222838 | 530 | LLVM also has an issue tracker where BPF related bugs can be found: |
34f15bf3 | 531 | |
54222838 | 532 | https://bugs.llvm.org/buglist.cgi?quicksearch=bpf |
34f15bf3 | 533 | |
54222838 JDB |
534 | However, it is better to reach out through mailing lists with having |
535 | maintainers in Cc. | |
34f15bf3 | 536 | |
54222838 JDB |
537 | Q: New BPF instruction for kernel and LLVM |
538 | ------------------------------------------ | |
34f15bf3 | 539 | Q: I have added a new BPF instruction to the kernel, how can I integrate |
54222838 | 540 | it into LLVM? |
34f15bf3 | 541 | |
54222838 JDB |
542 | A: LLVM has a ``-mcpu`` selector for the BPF back end in order to allow |
543 | the selection of BPF instruction set extensions. By default the | |
544 | ``generic`` processor target is used, which is the base instruction set | |
545 | (v1) of BPF. | |
34f15bf3 | 546 | |
54222838 JDB |
547 | LLVM has an option to select ``-mcpu=probe`` where it will probe the host |
548 | kernel for supported BPF instruction set extensions and selects the | |
549 | optimal set automatically. | |
34f15bf3 | 550 | |
54222838 | 551 | For cross-compilation, a specific version can be select manually as well :: |
34f15bf3 DB |
552 | |
553 | $ llc -march bpf -mcpu=help | |
554 | Available CPUs for this target: | |
555 | ||
556 | generic - Select the generic processor. | |
557 | probe - Select the probe processor. | |
558 | v1 - Select the v1 processor. | |
559 | v2 - Select the v2 processor. | |
560 | [...] | |
561 | ||
54222838 JDB |
562 | Newly added BPF instructions to the Linux kernel need to follow the same |
563 | scheme, bump the instruction set version and implement probing for the | |
564 | extensions such that ``-mcpu=probe`` users can benefit from the | |
565 | optimization transparently when upgrading their kernels. | |
34f15bf3 | 566 | |
54222838 JDB |
567 | If you are unable to implement support for the newly added BPF instruction |
568 | please reach out to BPF developers for help. | |
34f15bf3 | 569 | |
54222838 JDB |
570 | By the way, the BPF kernel selftests run with ``-mcpu=probe`` for better |
571 | test coverage. | |
34f15bf3 | 572 | |
54222838 JDB |
573 | Q: clang flag for target bpf? |
574 | ----------------------------- | |
575 | Q: In some cases clang flag ``-target bpf`` is used but in other cases the | |
576 | default clang target, which matches the underlying architecture, is used. | |
577 | What is the difference and when I should use which? | |
6215ea6b YS |
578 | |
579 | A: Although LLVM IR generation and optimization try to stay architecture | |
54222838 JDB |
580 | independent, ``-target <arch>`` still has some impact on generated code: |
581 | ||
582 | - BPF program may recursively include header file(s) with file scope | |
583 | inline assembly codes. The default target can handle this well, | |
584 | while ``bpf`` target may fail if bpf backend assembler does not | |
585 | understand these assembly codes, which is true in most cases. | |
586 | ||
587 | - When compiled without ``-g``, additional elf sections, e.g., | |
588 | .eh_frame and .rela.eh_frame, may be present in the object file | |
589 | with default target, but not with ``bpf`` target. | |
590 | ||
591 | - The default target may turn a C switch statement into a switch table | |
592 | lookup and jump operation. Since the switch table is placed | |
593 | in the global readonly section, the bpf program will fail to load. | |
594 | The bpf target does not support switch table optimization. | |
595 | The clang option ``-fno-jump-tables`` can be used to disable | |
596 | switch table generation. | |
597 | ||
598 | - For clang ``-target bpf``, it is guaranteed that pointer or long / | |
599 | unsigned long types will always have a width of 64 bit, no matter | |
600 | whether underlying clang binary or default target (or kernel) is | |
601 | 32 bit. However, when native clang target is used, then it will | |
602 | compile these types based on the underlying architecture's conventions, | |
603 | meaning in case of 32 bit architecture, pointer or long / unsigned | |
604 | long types e.g. in BPF context structure will have width of 32 bit | |
605 | while the BPF LLVM back end still operates in 64 bit. The native | |
606 | target is mostly needed in tracing for the case of walking ``pt_regs`` | |
607 | or other kernel structures where CPU's register width matters. | |
608 | Otherwise, ``clang -target bpf`` is generally recommended. | |
609 | ||
610 | You should use default target when: | |
611 | ||
612 | - Your program includes a header file, e.g., ptrace.h, which eventually | |
613 | pulls in some header files containing file scope host assembly codes. | |
614 | ||
615 | - You can add ``-fno-jump-tables`` to work around the switch table issue. | |
616 | ||
617 | Otherwise, you can use ``bpf`` target. Additionally, you *must* use bpf target | |
618 | when: | |
619 | ||
620 | - Your program uses data structures with pointer or long / unsigned long | |
621 | types that interface with BPF helpers or context data structures. Access | |
622 | into these structures is verified by the BPF verifier and may result | |
623 | in verification failures if the native architecture is not aligned with | |
624 | the BPF architecture, e.g. 64-bit. An example of this is | |
625 | BPF_PROG_TYPE_SK_MSG require ``-target bpf`` | |
626 | ||
627 | ||
628 | .. Links | |
629 | .. _Documentation/process/: https://www.kernel.org/doc/html/latest/process/ | |
630 | .. _MAINTAINERS: ../../MAINTAINERS | |
287f4fa9 | 631 | .. _netdev-FAQ: ../networking/netdev-FAQ.rst |
54222838 JDB |
632 | .. _samples/bpf/: ../../samples/bpf/ |
633 | .. _selftests: ../../tools/testing/selftests/bpf/ | |
b7a27c3a JDB |
634 | .. _Documentation/dev-tools/kselftest.rst: |
635 | https://www.kernel.org/doc/html/latest/dev-tools/kselftest.html | |
6215ea6b | 636 | |
34f15bf3 | 637 | Happy BPF hacking! |