- 24 Apr, 2024 17 commits
-
-
Alexei Starovoitov authored
Benjamin Tissoires says: ==================== Introduce bpf_wq This is a followup of sleepable bpf_timer[0]. When discussing sleepable bpf_timer, it was thought that we should give a try to bpf_wq, as the 2 APIs are similar but distinct enough to justify a new one. So here it is. I tried to keep as much as possible common code in kernel/bpf/helpers.c but I couldn't get away with code duplication in kernel/bpf/verifier.c. This series introduces a basic bpf_wq support: - creation is supported - assignment is supported - running a simple bpf_wq is also supported. We will probably need to extend the API further with: - a full delayed_work API (can be piggy backed on top with a correct flag) - bpf_wq_cancel() <- apparently not, this is shooting ourself in the foot - bpf_wq_cancel_sync() (for sleepable programs) - documentation --- For reference, the use cases I have in mind: --- Basically, I need to be able to defer a HID-BPF program for the following reasons (from the aforementioned patch): 1. defer an event: Sometimes we receive an out of proximity event, but the device can not be trusted enough, and we need to ensure that we won't receive another one in the following n milliseconds. So we need to wait those n milliseconds, and eventually re-inject that event in the stack. 2. inject new events in reaction to one given event: We might want to transform one given event into several. This is the case for macro keys where a single key press is supposed to send a sequence of key presses. But this could also be used to patch a faulty behavior, if a device forgets to send a release event. 3. communicate with the device in reaction to one event: We might want to communicate back to the device after a given event. For example a device might send us an event saying that it came back from sleeping state and needs to be re-initialized. Currently we can achieve that by keeping a userspace program around, raise a bpf event, and let that userspace program inject the events and commands. However, we are just keeping that program alive as a daemon for just scheduling commands. There is no logic in it, so it doesn't really justify an actual userspace wakeup. So a kernel workqueue seems simpler to handle. bpf_timers are currently running in a soft IRQ context, this patch series implements a sleppable context for them. Cheers, Benjamin [0] https://lore.kernel.org/all/20240408-hid-bpf-sleepable-v6-0-0499ddd91b94@kernel.org/ Changes in v2: - took previous review into account - mainly dropped BPF_F_WQ_SLEEPABLE - Link to v1: https://lore.kernel.org/r/20240416-bpf_wq-v1-0-c9e66092f842@kernel.org ==================== Link: https://lore.kernel.org/r/20240420-bpf_wq-v2-0-6c986a5a741f@kernel.orgSigned-off-by:
Alexei Starovoitov <ast@kernel.org>
-
Benjamin Tissoires authored
Allows to test if allocation/free works Signed-off-by:
Benjamin Tissoires <bentiss@kernel.org> Link: https://lore.kernel.org/r/20240420-bpf_wq-v2-16-6c986a5a741f@kernel.orgSigned-off-by:
-
Benjamin Tissoires authored
again, copy/paste from bpf_timer_start(). Signed-off-by:
-
Benjamin Tissoires authored
We assign the callback and set everything up. The actual tests of these callbacks will be done when bpf_wq_start() is available. Signed-off-by:
-
Benjamin Tissoires authored
To support sleepable async callbacks, we need to tell push_async_cb() whether the cb is sleepable or not. The verifier now detects that we are in bpf_wq_set_callback_impl and can allow a sleepable callback to happen. Signed-off-by:
-
Benjamin Tissoires authored
Allows to test if allocation/free works Signed-off-by:
-
Benjamin Tissoires authored
We need to teach the verifier about the second argument which is declared as void * but which is of type KF_ARG_PTR_TO_MAP. We could have dropped this extra case if we declared the second argument as struct bpf_map *, but that means users will have to do extra casting to have their program compile. We also need to duplicate the timer code for the checking if the map argument is matching the provided workqueue. Signed-off-by:
-
Benjamin Tissoires authored
We simply try in all supported map types if we can store/load a bpf_wq. Signed-off-by:
-
Benjamin Tissoires authored
Currently bpf_wq_cancel_and_free() is just a placeholder as there is no memory allocation for bpf_wq just yet. Again, duplication of the bpf_timer approach Signed-off-by:
-
Benjamin Tissoires authored
Introduce support for KF_ARG_PTR_TO_WORKQUEUE. The kfuncs will use bpf_wq as argument and that will be recognized as workqueue argument by verifier. bpf_wq_kern casting can happen inside kfunc, but using bpf_wq in argument makes life easier for users who work with non-kern type in BPF progs. Duplicate process_timer_func into process_wq_func. meta argument is only needed to ensure bpf_wq_init's workqueue and map arguments are coming from the same map (map_uid logic is necessary for correct inner-map handling), so also amend check_kfunc_args() to match what helpers functions check is doing. Signed-off-by:
-
Benjamin Tissoires authored
When a kfunc is declared with a KF_ARG_PTR_TO_MAP, we should have reg->map_ptr set to a non NULL value, otherwise, that means that the underlying type is not a map. Signed-off-by:
-
Benjamin Tissoires authored
cp include/uapi/linux/bpf.h tools/include/uapi/linux/bpf.h Signed-off-by:
-
Benjamin Tissoires authored
Mostly a copy/paste from the bpf_timer API, without the initialization and free, as they will be done in a separate patch. Signed-off-by:
-
Benjamin Tissoires authored
Same reason than most bpf_timer* functions, we need almost the same for workqueues. So extract the generic part out of it so bpf_wq_cancel_and_free can reuse it. Signed-off-by:
-
Benjamin Tissoires authored
In the same way we have a generic __bpf_async_init(), we also need to share code between timer and workqueue for the set_callback call. We just add an unused flags parameter, as it will be used for workqueues. Signed-off-by:
-
Benjamin Tissoires authored
No code change except for the new flags argument being stored in the local data struct. Signed-off-by:
-
Benjamin Tissoires authored
To be able to add workqueues and reuse most of the timer code, we need to make bpf_hrtimer more generic. There is no code change except that the new struct gets a new u64 flags attribute. We are still below 2 cache lines, so this shouldn't impact the current running codes. The ordering is also changed. Everything related to async callback is now on top of bpf_hrtimer. Signed-off-by:
-
- 22 Apr, 2024 2 commits
-
-
Rafael Passos authored
Found the following typos in comments, and fixed them: s/unpriviledged/unprivileged/ s/reponsible/responsible/ s/possiblities/possibilities/ s/Divison/Division/ s/precsion/precision/ s/havea/have a/ s/reponsible/responsible/ s/responsibile/responsible/ s/tigher/tighter/ s/respecitve/respective/ Signed-off-by:
Rafael Passos <rafael@rcpassos.me> Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Link: https://lore.kernel.org/bpf/6af7deb4-bb24-49e8-b3f1-8dd410597337@smtp-relay.sendinblue.com
-
Rafael Passos authored
I found this typo in the save_aux_ptr_type function. s/allow_trust_missmatch/allow_trust_mismatch/ I did not find this anywhere else in the codebase. Signed-off-by:
-
- 21 Apr, 2024 2 commits
-
-
Dave Thaler authored
Other places that had pseudocode were prefixed with :: so as to appear in a literal block, but one place was inconsistent. This patch fixes that inconsistency. Signed-off-by:
Dave Thaler <dthaler1968@googlemail.com> Acked-by:
David Vernet <void@manifault.com> Link: https://lore.kernel.org/r/20240419213826.7301-1-dthaler1968@gmail.comSigned-off-by:
-
Dave Thaler authored
Per IETF 119 meeting discussion and mailing list discussion at https://mailarchive.ietf.org/arch/msg/bpf/2JwWQwFdOeMGv0VTbD0CKWwAOEA/ the following changes are made. First, say call by "static ID" rather than call by "address" Second, change "pointer" to "address" Signed-off-by:
Dave Thaler <dthaler1968@gmail.com> Acked-by:
-
- 20 Apr, 2024 7 commits
-
-
Martin KaFai Lau authored
Geliang Tang says: ==================== v5: - address Martin's comments for v4. (thanks) - drop start_server_addr_opts, add opts as a argument of start_server_addr. - add opts argument for connect_to_addr too. - move some patches out of this set, stay with start_server_addr() and connect_to_addr() only in it. v4: - add more patches using make_sockaddr and get_socket_local_port helpers. v3: - address comments of Martin and Eduard in v2. (thanks) - move "int type" to the first argument of start_server_addr and connect_to_addr. - add start_server_addr_opts. - using "sockaddr_storage" instead of "sockaddr". - move start_server_setsockopt patches out of this series. v2: - update patch 6 only, fix errors reported by CI. This patchset uses public helpers start_server_* and connect_to_* defined in network_helpers.c to drop duplicate code. ==================== Signed-off-by:Martin KaFai Lau <martin.lau@kernel.org>
-
Geliang Tang authored
This patch uses public helper connect_to_addr() exported in network_helpers.h instead of the local defined function connect_to_server() in prog_tests/sk_assign.c. This can avoid duplicate code. The code that sets SO_SNDTIMEO timeout as timeo_sec (3s) can be dropped, since connect_to_addr() sets default timeout as 3s. Signed-off-by:
Geliang Tang <tanggeliang@kylinos.cn> Link: https://lore.kernel.org/r/98fdd384872bda10b2adb052e900a2212c9047b9.1713427236.git.tanggeliang@kylinos.cnSigned-off-by:
-
Geliang Tang authored
This patch uses public helper connect_to_addr() exported in network_helpers.h instead of the local defined function connect_to_server() in prog_tests/cls_redirect.c. This can avoid duplicate code. Signed-off-by:
-
Geliang Tang authored
Move the third argument "int type" of connect_to_addr() to the first one which is closer to how the socket syscall is doing it. And add a network_helper_opts argument as the fourth one. Then change its usages in sock_addr.c too. Signed-off-by:
-
Geliang Tang authored
Include network_helpers.h in prog_tests/sk_assign.c, use the newly added public helper start_server_addr() instead of the local defined function start_server(). This can avoid duplicate code. The code that sets SO_RCVTIMEO timeout as timeo_sec (3s) can be dropped, since start_server_addr() sets default timeout as 3s. Signed-off-by:
-
Geliang Tang authored
Include network_helpers.h in prog_tests/cls_redirect.c, use the newly added public helper start_server_addr() instead of the local defined function start_server(). This can avoid duplicate code. Signed-off-by:
-
Geliang Tang authored
In order to pair up with connect_to_addr(), this patch adds a new helper start_server_addr(), which is a wrapper of __start_server(). It accepts an argument 'addr' of 'struct sockaddr_storage' type instead of a string type argument like start_server(), and a network_helper_opts argument as the last one. Signed-off-by:
-
- 18 Apr, 2024 1 commit
-
-
Alexei Starovoitov authored
The codegen for is_mov_percpu_addr instruction works for rax/r8 registers only. Fix it to generate proper x86 byte code for other registers. Fixes: 7bdbf744 ("bpf: add special internal-only MOV instruction to resolve per-CPU addrs") Signed-off-by:
Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/bpf/20240417214406.15788-1-alexei.starovoitov@gmail.com
-
- 17 Apr, 2024 2 commits
-
-
Quentin Deslandes authored
When dumping a character array, libbpf will watch for a '\0' and set is_array_terminated=true if found. This prevents libbpf from printing the remaining characters of the array, treating it as a nul-terminated string. However, once this flag is set, it's never reset, leading to subsequent characters array not being printed properly: .str_multi = (__u8[2][16])[ [ 'H', 'e', 'l', ], ], This patch saves the is_array_terminated flag and restores its default (false) value before looping over the elements of an array, then restores it afterward. This way, libbpf's behavior is unchanged when dumping the characters of an array, but subsequent arrays are printed properly: .str_multi = (__u8[2][16])[ [ 'H', 'e', 'l', ], [ 'l', 'o', ], ], Signed-off-by:Quentin Deslandes <qde@naccy.de> Signed-off-by:
-
Quentin Deslandes authored
In btf_dump_array_data(), libbpf will call btf_dump_dump_type_data() for each element. For an array of characters, each element will be processed the following way: - btf_dump_dump_type_data() is called to print the character - btf_dump_data_pfx() prefixes the current line with the proper number of indentations - btf_dump_int_data() is called to print the character - After the last character is printed, btf_dump_dump_type_data() calls btf_dump_data_pfx() before writing the closing bracket However, for an array containing characters, btf_dump_int_data() won't print any '\0' and subsequent characters. This leads to situations where the line prefix is written, no character is added, then the prefix is written again before adding the closing bracket: (struct sk_metadata){ .str_array = (__u8[14])[ 'H', 'e', 'l', 'l', 'o', ], This change solves this issue by printing the '\0' character, which has two benefits: - The bracket closing the array is properly aligned - It's clear from a user point of view that libbpf uses '\0' as a terminator for arrays of characters. Signed-off-by:
-
- 16 Apr, 2024 5 commits
-
-
Quentin Monnet authored
This commit contains a series of clean-ups and fixes for bpftool's bash completion file: - Make sure all local variables are declared as such. - Make sure variables are initialised before being read. - Update ELF section ("maps" -> ".maps") for looking up map names in object files. - Fix call to _init_completion. - Move definition for MAP_TYPE and PROG_TYPE higher up in the scope to avoid defining them multiple times, reuse MAP_TYPE where relevant. - Simplify completion for "duration" keyword in "bpftool prog profile". - Fix completion for "bpftool struct_ops register" and "bpftool link (pin|detach)" where we would repeatedly suggest file names instead of suggesting just one name. - Fix completion for "bpftool iter pin ... map MAP" to account for the "map" keyword. - Add missing "detach" suggestion for "bpftool link". Signed-off-by:Quentin Monnet <qmo@kernel.org> Signed-off-by:
-
Quentin Monnet authored
When using references to BPF programs, bpftool supports passing programs by name on the command line. The manual pages for "bpftool prog" and "bpftool map" (for prog_array updates) mention it, but we have a few additional subcommands that support referencing programs by name but do not mention it in their documentation. Let's update the pages for subcommands "btf", "cgroup", and "net". Similarly, we can reference maps by name when passing them to "bpftool prog load", so we update the page for "bpftool prog" as well. Signed-off-by:
-
Harishankar Vishwanathan authored
This patch addresses a latent unsoundness issue in the scalar(32)_min_max_and/or/xor functions. While it is not a bugfix, it ensures that the functions produce sound outputs for all inputs. The issue occurs in these functions when setting signed bounds. The following example illustrates the issue for scalar_min_max_and(), but it applies to the other functions. In scalar_min_max_and() the following clause is executed when ANDing positive numbers: /* ANDing two positives gives a positive, so safe to * cast result into s64. */ dst_reg->smin_value = dst_reg->umin_value; dst_reg->smax_value = dst_reg->umax_value; However, if umin_value and umax_value of dst_reg cross the sign boundary (i.e., if (s64)dst_reg->umin_value > (s64)dst_reg->umax_value), then we will end up with smin_value > smax_value, which is unsound. Previous works [1, 2] have discovered and reported this issue. Our tool Agni [2, 3] consideres it a false positive. This is because, during the verification of the abstract operator scalar_min_max_and(), Agni restricts its inputs to those passing through reg_bounds_sync(). This mimics real-world verifier behavior, as reg_bounds_sync() is invariably executed at the tail of every abstract operator. Therefore, such behavior is unlikely in an actual verifier execution. However, it is still unsound for an abstract operator to set signed bounds such that smin_value > smax_value. This patch fixes it, making the abstract operator sound for all (well-formed) inputs. It is worth noting that while the previous code updated the signed bounds (using the output unsigned bounds) only when the *input signed* bounds were positive, the new code updates them whenever the *output unsigned* bounds do not cross the sign boundary. An alternative approach to fix this latent unsoundness would be to unconditionally set the signed bounds to unbounded [S64_MIN, S64_MAX], and let reg_bounds_sync() refine the signed bounds using the unsigned bounds and the tnum. We found that our approach produces more precise (tighter) bounds. For example, consider these inputs to BPF_AND: /* dst_reg */ var_off.value: 8608032320201083347 var_off.mask: 615339716653692460 smin_value: 8070450532247928832 smax_value: 8070450532247928832 umin_value: 13206380674380886586 umax_value: 13206380674380886586 s32_min_value: -2110561598 s32_max_value: -133438816 u32_min_value: 4135055354 u32_max_value: 4135055354 /* src_reg */ var_off.value: 8584102546103074815 var_off.mask: 9862641527606476800 smin_value: 2920655011908158522 smax_value: 7495731535348625717 umin_value: 7001104867969363969 umax_value: 8584102543730304042 s32_min_value: -2097116671 s32_max_value: 71704632 u32_min_value: 1047457619 u32_max_value: 4268683090 After going through tnum_and() -> scalar32_min_max_and() -> scalar_min_max_and() -> reg_bounds_sync(), our patch produces the following bounds for s32: s32_min_value: -1263875629 s32_max_value: -159911942 Whereas, setting the signed bounds to unbounded in scalar_min_max_and() produces: s32_min_value: -1263875629 s32_max_value: -1 As observed, our patch produces a tighter s32 bound. We also confirmed using Agni and SMT verification that our patch always produces signed bounds that are equal to or more precise than setting the signed bounds to unbounded in scalar_min_max_and(). [1] https://sanjit-bhat.github.io/assets/pdf/ebpf-verifier-range-analysis22.pdf [2] https://link.springer.com/chapter/10.1007/978-3-031-37709-9_12 [3] https://github.com/bpfverif/agniCo-developed-by:
Matan Shachnai <m.shachnai@rutgers.edu> Signed-off-by:
Srinivas Narayana <srinivas.narayana@rutgers.edu> Signed-off-by:
Santosh Nagarakatte <santosh.nagarakatte@rutgers.edu> Signed-off-by:
Harishankar Vishwanathan <harishankar.vishwanathan@gmail.com> Signed-off-by:
-
Chen Pei authored
Currently, there are two comments with same name "64-bit ATOMIC magnitudes", the second one should be "32-bit ATOMIC magnitudes" based on the context. Signed-off-by:
Chen Pei <cp0613@linux.alibaba.com> Signed-off-by:
Yonghong Song <yonghong.song@linux.dev> Link: https://lore.kernel.org/bpf/20240415081928.17440-1-cp0613@linux.alibaba.com
-
Ard Biesheuvel authored
If the BTF code is enabled in the build configuration, the start/stop BTF markers are guaranteed to exist. Only when CONFIG_DEBUG_INFO_BTF=n, the references in btf_parse_vmlinux() will remain unsatisfied, relying on the weak linkage of the external references to avoid breaking the build. Avoid GOT based relocations to these markers in the final executable by dropping the weak attribute and instead, make btf_parse_vmlinux() return ERR_PTR(-ENOENT) directly if CONFIG_DEBUG_INFO_BTF is not enabled to begin with. The compiler will drop any subsequent references to __start_BTF and __stop_BTF in that case, allowing the link to succeed. Note that Clang will notice that taking the address of __start_BTF can no longer yield NULL, so testing for that condition becomes unnecessary. Signed-off-by:
Ard Biesheuvel <ardb@kernel.org> Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Acked-by:
Jiri Olsa <jolsa@kernel.org> Link: https://lore.kernel.org/bpf/20240415162041.2491523-8-ardb+git@google.com
-
- 12 Apr, 2024 2 commits
-
-
Jiri Olsa authored
We have two printk tests reading trace_pipe in non blocking way, with the very same code. Moving that in new read_trace_pipe_iter function. Current read_trace_pipe is used from samples/bpf and needs to do blocking read and printf of the trace_pipe data, using new read_trace_pipe_iter to implement that. Both printk tests do early checks for the number of found messages and can bail earlier, but I did not find any speed difference w/o that condition, so I did not complicate the change more for that. Some of the samples/bpf programs use read_trace_pipe function, so I kept that interface untouched. I did not see any issues with affected samples/bpf programs other than there's slight change in read_trace_pipe output. The current code uses puts that adds new line after the printed string, so we would occasionally see extra new line. With this patch we read output per lines, so there's no need to use puts and we can use just printf instead without extra new line. Signed-off-by:
-
Thorsten Blum authored
s/at at/at a/ Signed-off-by:
Thorsten Blum <thorsten.blum@toblux.com> Signed-off-by:
-
- 11 Apr, 2024 2 commits
-
-
Martin KaFai Lau authored
Geliang Tang says: ==================== v5: - address Martin's comments for v4 (thanks). - update patch 2, use 'return err' instead of 'return -1/0'. - drop patch 3 in v4. v4: - fix a bug in v3, it should be 'if (err)', not 'if (!err)'. - move "selftests/bpf: Use log_err in network_helpers" out of this series. v3: - add two more patches. - use log_err instead of ASSERT in v3. - let send_recv_data return int as Martin suggested. v2: Address Martin's comments for v1 (thanks.) - drop patch 1, "export send_byte helper". - drop "WRITE_ONCE(arg.stop, 0)". - rebased. send_recv_data will be re-used in MPTCP bpf tests, but not included in this set because it depends on other patches that have not been in the bpf-next yet. It will be sent as another set soon. ==================== Signed-off-by: -
Geliang Tang authored
This patch extracts the code to send and receive data into a new helper named send_recv_data() in network_helpers.c and export it in network_helpers.h. This helper will be used for MPTCP BPF selftests. Signed-off-by:
-
