- 30 Jul, 2023 1 commit
-
-
Yonghong Song authored
Two issues are fixed: 1. Malformed table due to newly-introduced BPF_MOVSX 2. Missing reference link for ``Sign-extension load operations`` Fixes: 245d4c40 ("docs/bpf: Add documentation for new instructions") Cc: bpf@ietf.org Reported-by:
kernel test robot <lkp@intel.com> Closes: https://lore.kernel.org/oe-kbuild-all/202307291840.Cqhj7uox-lkp@intel.com/Signed-off-by:
Yonghong Song <yonghong.song@linux.dev> Link: https://lore.kernel.org/r/20230730004251.381307-1-yonghong.song@linux.devSigned-off-by:
Alexei Starovoitov <ast@kernel.org>
-
- 28 Jul, 2023 29 commits
-
-
Alexei Starovoitov authored
Daniel Xu says: ==================== Support defragmenting IPv(4|6) packets in BPF === Context === In the context of a middlebox, fragmented packets are tricky to handle. The full 5-tuple of a packet is often only available in the first fragment which makes enforcing consistent policy difficult. There are really only two stateless options, neither of which are very nice: 1. Enforce policy on first fragment and accept all subsequent fragments. This works but may let in certain attacks or allow data exfiltration. 2. Enforce policy on first fragment and drop all subsequent fragments. This does not really work b/c some protocols may rely on fragmentation. For example, DNS may rely on oversized UDP packets for large responses. So stateful tracking is the only sane option. RFC 8900 [0] calls this out as well in section 6.3: Middleboxes [...] should process IP fragments in a manner that is consistent with [RFC0791] and [RFC8200]. In many cases, middleboxes must maintain state in order to achieve this goal. === BPF related bits === Policy has traditionally been enforced from XDP/TC hooks. Both hooks run before kernel reassembly facilities. However, with the new BPF_PROG_TYPE_NETFILTER, we can rather easily hook into existing netfilter reassembly infra. The basic idea is we bump a refcnt on the netfilter defrag module and then run the bpf prog after the defrag module runs. This allows bpf progs to transparently see full, reassembled packets. The nice thing about this is that progs don't have to carry around logic to detect fragments. === Changelog === Changes from v5: * Fix defrag disable codepaths Changes from v4: * Refactor module handling code to not sleep in rcu_read_lock() * Also unify the v4 and v6 hook structs so they can share codepaths * Fixed some checkpatch.pl formatting warnings Changes from v3: * Correctly initialize `addrlen` stack var for recvmsg() Changes from v2: * module_put() if ->enable() fails * Fix CI build errors Changes from v1: * Drop bpf_program__attach_netfilter() patches * static -> static const where appropriate * Fix callback assignment order during registration * Only request_module() if callbacks are missing * Fix retval when modprobe fails in userspace * Fix v6 defrag module name (nf_defrag_ipv6_hooks -> nf_defrag_ipv6) * Simplify priority checking code * Add warning if module doesn't assign callbacks in the future * Take refcnt on module while defrag link is active [0]: https://datatracker.ietf.org/doc/html/rfc8900 ==================== Link: https://lore.kernel.org/r/cover.1689970773.git.dxu@dxuuu.xyzSigned-off-by: -
Daniel Xu authored
These selftests tests 2 major scenarios: the BPF based defragmentation can successfully be done and that packet pointers are invalidated after calls to the kfunc. The logic is similar for both ipv4 and ipv6. In the first scenario, we create a UDP client and UDP echo server. The the server side is fairly straightforward: we attach the prog and simply echo back the message. The on the client side, we send fragmented packets to and expect the reassembled message back from the server. Signed-off-by:
Daniel Xu <dxu@dxuuu.xyz> Link: https://lore.kernel.org/r/33e40fdfddf43be93f2cb259303f132f46750953.1689970773.git.dxu@dxuuu.xyzSigned-off-by:
-
Daniel Xu authored
Extend connect_to_fd_opts() to take optional type and protocol parameters for the client socket. These parameters are useful when opening a raw socket to send IP fragments. Signed-off-by:
-
Daniel Xu authored
For connectionless protocols or raw sockets we do not want to actually connect() to the server. Signed-off-by:
-
Daniel Xu authored
This commit adds support for enabling IP defrag using pre-existing netfilter defrag support. Basically all the flag does is bump a refcnt while the link the active. Checks are also added to ensure the prog requesting defrag support is run _after_ netfilter defrag hooks. We also take care to avoid any issues w.r.t. module unloading -- while defrag is active on a link, the module is prevented from unloading. Signed-off-by:
Florian Westphal <fw@strlen.de> Link: https://lore.kernel.org/r/5cff26f97e55161b7d56b09ddcf5f8888a5add1d.1689970773.git.dxu@dxuuu.xyzSigned-off-by:
-
Daniel Xu authored
We want to be able to enable/disable IP packet defrag from core bpf/netfilter code. In other words, execute code from core that could possibly be built as a module. To help avoid symbol resolution errors, use glue hooks that the modules will register callbacks with during module init. Signed-off-by:
-
Yonghong Song authored
Improve documentation for cpu=v4 instructions based on David's suggestions. Cc: bpf@ietf.org Suggested-by:
David Vernet <void@manifault.com> Acked-by:
-
YiFei Zhu authored
In internal testing of test_maps, we sometimes observed failures like: test_maps: test_maps.c:173: void test_hashmap_percpu(unsigned int, void *): Assertion `bpf_map_update_elem(fd, &key, value, BPF_ANY) == 0' failed. where the errno is ENOMEM. After some troubleshooting and enabling the warnings, we saw: [ 91.304708] percpu: allocation failed, size=8 align=8 atomic=1, atomic alloc failed, no space left [ 91.304716] CPU: 51 PID: 24145 Comm: test_maps Kdump: loaded Tainted: G N 6.1.38-smp-DEV #7 [ 91.304719] Hardware name: Google Astoria/astoria, BIOS 0.20230627.0-0 06/27/2023 [ 91.304721] Call Trace: [ 91.304724] <TASK> [ 91.304730] [<ffffffffa7ef83b9>] dump_stack_lvl+0x59/0x88 [ 91.304737] [<ffffffffa7ef83f8>] dump_stack+0x10/0x18 [ 91.304738] [<ffffffffa75caa0c>] pcpu_alloc+0x6fc/0x870 [ 91.304741] [<ffffffffa75ca302>] __alloc_percpu_gfp+0x12/0x20 [ 91.304743] [<ffffffffa756785e>] alloc_bulk+0xde/0x1e0 [ 91.304746] [<ffffffffa7566c02>] bpf_mem_alloc_init+0xd2/0x2f0 [ 91.304747] [<ffffffffa7547c69>] htab_map_alloc+0x479/0x650 [ 91.304750] [<ffffffffa751d6e0>] map_create+0x140/0x2e0 [ 91.304752] [<ffffffffa751d413>] __sys_bpf+0x5a3/0x6c0 [ 91.304753] [<ffffffffa751c3ec>] __x64_sys_bpf+0x1c/0x30 [ 91.304754] [<ffffffffa7ef847a>] do_syscall_64+0x5a/0x80 [ 91.304756] [<ffffffffa800009b>] entry_SYSCALL_64_after_hwframe+0x63/0xcd This makes sense, because in atomic context, percpu allocation would not create new chunks; it would only create in non-atomic contexts. And if during prefill all precpu chunks are full, -ENOMEM would happen immediately upon next unit_alloc. Prefill phase does not actually run in atomic context, so we can use this fact to allocate non-atomically with GFP_KERNEL instead of GFP_NOWAIT. This avoids the immediate -ENOMEM. GFP_NOWAIT has to be used in unit_alloc when bpf program runs in atomic context. Even if bpf program runs in non-atomic context, in most cases, rcu read lock is enabled for the program so GFP_NOWAIT is still needed. This is often also the case for BPF_MAP_UPDATE_ELEM syscalls. Signed-off-by:YiFei Zhu <zhuyifei@google.com> Acked-by:
Hou Tao <houtao1@huawei.com> Link: https://lore.kernel.org/r/20230728043359.3324347-1-zhuyifei@google.comSigned-off-by:
-
Yonghong Song authored
Currently, test_progs-cpuv4 is generated with clang build kernel when bpf cpu=v4 is supported by the clang compiler. Let us enable test_progs-cpuv4 for gcc build kernel as well. Signed-off-by:
-
Yonghong Song authored
The kernel test robot reported compilation warnings when -Wparentheses is added to KBUILD_CFLAGS with gcc compiler. The following is the error message: .../bpf-next/kernel/bpf/verifier.c: In function ‘coerce_reg_to_size_sx’: .../bpf-next/kernel/bpf/verifier.c:5901:14: error: suggest parentheses around comparison in operand of ‘==’ [-Werror=parentheses] if (s64_max >= 0 == s64_min >= 0) { ~~~~~~~~^~~~ .../bpf-next/kernel/bpf/verifier.c: In function ‘coerce_subreg_to_size_sx’: .../bpf-next/kernel/bpf/verifier.c:5965:14: error: suggest parentheses around comparison in operand of ‘==’ [-Werror=parentheses] if (s32_min >= 0 == s32_max >= 0) { ~~~~~~~~^~~~ To fix the issue, add proper parentheses for the above '>=' condition to silence the warning/error. I tried a few clang compilers like clang16 and clang18 and they do not emit such warnings with -Wparentheses. Reported-by:Jiri Olsa <jolsa@kernel.org> Link: https://lore.kernel.org/r/20230728055740.2284534-1-yonghong.song@linux.devSigned-off-by:
-
Alexei Starovoitov authored
Yonghong Song says: ==================== bpf: Support new insns from cpu v4 In previous discussion ([1]), it is agreed that we should introduce cpu version 4 (llvm flag -mcpu=v4) which contains some instructions which can simplify code, make code easier to understand, fix the existing problem, or simply for feature completeness. More specifically, the following new insns are proposed: . sign extended load . sign extended mov . bswap . signed div/mod . ja with 32-bit offset This patch set added kernel support for insns proposed in [1] except BPF_ST which already has full kernel support. Beside the above proposed insns, LLVM will generate BPF_ST insn as well under -mcpu=v4. The llvm patch ([2]) has been merged into llvm-project 'main' branch. The patchset implements interpreter, jit and verifier support for these new insns. For this patch set, I tested cpu v2/v3/v4 and the selftests are all passed. I also tested selftests introduced in this patch set with additional changes beside normal jit testing (bpf_jit_enable = 1 and bpf_jit_harden = 0) - bpf_jit_enable = 0 - bpf_jit_enable = 1 and bpf_jit_harden = 1 and both testing passed. [1] https://lore.kernel.org/bpf/4bfe98be-5333-1c7e-2f6d-42486c8ec039@meta.com/ [2] https://reviews.llvm.org/D144829 Changelogs: v4 -> v5: . for v4, patch 8/17 missed in mailing list and patchwork, so resend. . rebase on top of master v3 -> v4: . some minor asm syntax adjustment based on llvm change. . add clang version and target arch guard for new tests so they can still compile with old llvm compilers. . some changes to the bpf doc. v2 -> v3: . add missed disasm change from v2. . handle signed load of ctx fields properly. . fix some interpreter sdiv/smod error when bpf_jit_enable = 0. . fix some verifier range bounding errors. . add more C tests. RFCv1 -> v2: . add more verifier supports for signed extend load and mov insns. . rename some insn names to be more consistent with intel practice. . add cpuv4 test runner for test progs. . add more unit and C tests. . add documentation. ==================== Link: https://lore.kernel.org/r/20230728011143.3710005-1-yonghong.song@linux.devSigned-off-by:
-
Yonghong Song authored
Add documentation in instruction-set.rst for new instruction encoding and their corresponding operations. Also removed the question related to 'no BPF_SDIV' in bpf_design_QA.rst since we have BPF_SDIV insn now. Cc: bpf@ietf.org Signed-off-by:
-
Yonghong Song authored
The following ldsx cases are tested: - signed readonly map value - read/write map value - probed memory - not-narrowed ctx field access - narrowed ctx field access. Without previous proper verifier/git handling, the test will fail. If cpuv4 is not supported either by compiler or by jit, the test will be skipped. # ./test_progs -t ldsx_insn #113/1 ldsx_insn/map_val and probed_memory:SKIP #113/2 ldsx_insn/ctx_member_sign_ext:SKIP #113/3 ldsx_insn/ctx_member_narrow_sign_ext:SKIP #113 ldsx_insn:SKIP Summary: 1/0 PASSED, 3 SKIPPED, 0 FAILED Signed-off-by:
-
Yonghong Song authored
Add unit tests for gotol insn. Signed-off-by:
-
Yonghong Song authored
Add unit tests for sdiv/smod insns. Signed-off-by:
-
Yonghong Song authored
Add unit tests for bswap insns. Signed-off-by:
-
Yonghong Song authored
Add unit tests for movsx insns. Signed-off-by:
-
Yonghong Song authored
Add unit tests for new ldsx insns. The test includes sign-extension with a single value or with a value range. If cpuv4 is not supported due to (1) older compiler, e.g., less than clang version 18, or (2) test runner test_progs and test_progs-no_alu32 which tests cpu v2 and v3, or (3) non-x86_64 arch not supporting new insns in jit yet, a dummy program is added with below output: #318/1 verifier_ldsx/cpuv4 is not supported by compiler or jit, use a dummy test:OK #318 verifier_ldsx:OK to indicate the test passed with a dummy test instead of actually testing cpuv4. I am using a dummy prog to avoid changing the verifier testing infrastructure. Once clang 18 is widely available and other architectures support cpuv4, at least for CI run, the dummy program can be removed. Signed-off-by: -
Yonghong Song authored
Similar to no-alu32 runner, if clang compiler supports -mcpu=v4, a cpuv4 runner is created to test bpf programs compiled with -mcpu=v4. The following are some num-of-insn statistics for each newer instructions based on existing selftests, excluding subsequent cpuv4 insn specific tests. insn pattern # of instructions reg = (s8)reg 4 reg = (s16)reg 4 reg = (s32)reg 144 reg = *(s8 *)(reg + off) 13 reg = *(s16 *)(reg + off) 14 reg = *(s32 *)(reg + off) 15215 reg = bswap16 reg 142 reg = bswap32 reg 38 reg = bswap64 reg 14 reg s/= reg 0 reg s%= reg 0 gotol <offset> 58 Note that in llvm -mcpu=v4 implementation, the compiler is a little bit conservative about generating 'gotol' insn (32-bit branch offset) as it didn't precise count the number of insns (e.g., some insns are debug insns, etc.). Compared to old 'goto' insn, newer 'gotol' insn should have comparable verification states to 'goto' insn. With current patch set, all selftests passed with -mcpu=v4 when running test_progs-cpuv4 binary. The -mcpu=v3 and -mcpu=v2 run are also successful. Signed-off-by:
-
Yonghong Song authored
The following test_verifier subtest failed due to new encoding for BSWAP. $ ./test_verifier ... #99/u invalid 64-bit BPF_END FAIL Unexpected success to load! verification time 215 usec stack depth 0 processed 3 insns (limit 1000000) max_states_per_insn 0 total_states 0 peak_states 0 mark_read 0 #99/p invalid 64-bit BPF_END FAIL Unexpected success to load! verification time 198 usec stack depth 0 processed 3 insns (limit 1000000) max_states_per_insn 0 total_states 0 peak_states 0 mark_read 0 Tighten the test so it still reports a failure. Acked-by:
Eduard Zingerman <eddyz87@gmail.com> Signed-off-by:
-
Yonghong Song authored
Add asm support for new instructions so kernel verifier and bpftool xlated insn dumps can have proper asm syntax for new instructions. Acked-by:
Quentin Monnet <quentin@isovalent.com> Signed-off-by:
-
Yonghong Song authored
Add interpreter/jit/verifier support for 32bit offset jmp instruction. If a conditional jmp instruction needs more than 16bit offset, it can be simulated with a conditional jmp + a 32bit jmp insn. Acked-by:
-
Yonghong Song authored
Handle new insns properly in bpf_jit_blind_insn() function. Acked-by:
-
Yonghong Song authored
Add interpreter/jit support for new signed div/mod insns. The new signed div/mod instructions are encoded with unsigned div/mod instructions plus insn->off == 1. Also add basic verifier support to ensure new insns get accepted. Acked-by:
-
Yonghong Song authored
The existing 'be' and 'le' insns will do conditional bswap depends on host endianness. This patch implements unconditional bswap insns. Acked-by:
-
Yonghong Song authored
Currently, if user accesses a ctx member with signed types, the compiler will generate an unsigned load followed by necessary left and right shifts. With the introduction of sign-extension load, compiler may just emit a ldsx insn instead. Let us do a final movsx sign extension to the final unsigned ctx load result to satisfy original sign extension requirement. Acked-by:
-
Yonghong Song authored
Add interpreter/jit support for new sign-extension mov insns. The original 'MOV' insn is extended to support reg-to-reg signed version for both ALU and ALU64 operations. For ALU mode, the insn->off value of 8 or 16 indicates sign-extension from 8- or 16-bit value to 32-bit value. For ALU64 mode, the insn->off value of 8/16/32 indicates sign-extension from 8-, 16- or 32-bit value to 64-bit value. Acked-by:
-
Yonghong Song authored
Add interpreter/jit support for new sign-extension load insns which adds a new mode (BPF_MEMSX). Also add verifier support to recognize these insns and to do proper verification with new insns. In verifier, besides to deduce proper bounds for the dst_reg, probed memory access is also properly handled. Acked-by:
-
Jose E. Marchesi authored
This patch fixes the documentation of the BPF_NEG instruction to denote that it does not use the source register operand. Signed-off-by:
Jose E. Marchesi <jose.marchesi@oracle.com> Acked-by:
Dave Thaler <dthaler@microsoft.com> Acked-by:
-
- 26 Jul, 2023 1 commit
-
-
Arnd Bergmann authored
Splitting these out into separate helper functions means that we actually pass an uninitialized variable into another function call if dec_active() happens to not be inlined, and CONFIG_PREEMPT_RT is disabled: kernel/bpf/memalloc.c: In function 'add_obj_to_free_list': kernel/bpf/memalloc.c:200:9: error: 'flags' is used uninitialized [-Werror=uninitialized] 200 | dec_active(c, flags); Avoid this by passing the flags by reference, so they either get initialized and dereferenced through a pointer, or the pointer never gets accessed at all. Fixes: 18e027b1 ("bpf: Factor out inc/dec of active flag into helpers.") Suggested-by:
Alexei Starovoitov <alexei.starovoitov@gmail.com> Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Link: https://lore.kernel.org/r/20230725202653.2905259-1-arnd@kernel.orgSigned-off-by:
-
- 25 Jul, 2023 9 commits
-
-
Colin Ian King authored
There is a spelling mistake in an error message. Fix it. Signed-off-by:
Colin Ian King <colin.i.king@gmail.com> Acked-by:
Björn Töpel <bjorn@kernel.org> Link: https://lore.kernel.org/r/20230720104815.123146-1-colin.i.king@gmail.comSigned-off-by:
Martin KaFai Lau <martin.lau@kernel.org>
-
Martin KaFai Lau authored
Lorenz Bauer says: ==================== We want to replace iptables TPROXY with a BPF program at TC ingress. To make this work in all cases we need to assign a SO_REUSEPORT socket to an skb, which is currently prohibited. This series adds support for such sockets to bpf_sk_assing. I did some refactoring to cut down on the amount of duplicate code. The key to this is to use INDIRECT_CALL in the reuseport helpers. To show that this approach is not just beneficial to TC sk_assign I removed duplicate code for bpf_sk_lookup as well. Joint work with Daniel Borkmann. Signed-off-by:
Lorenz Bauer <lmb@isovalent.com> --- Changes in v6: - Reject unhashed UDP sockets in bpf_sk_assign to avoid ref leak - Link to v5: https://lore.kernel.org/r/20230613-so-reuseport-v5-0-f6686a0dbce0@isovalent.com Changes in v5: - Drop reuse_sk == sk check in inet[6]_steal_stock (Kuniyuki) - Link to v4: https://lore.kernel.org/r/20230613-so-reuseport-v4-0-4ece76708bba@isovalent.com Changes in v4: - WARN_ON_ONCE if reuseport socket is refcounted (Kuniyuki) - Use inet[6]_ehashfn_t to shorten function declarations (Kuniyuki) - Shuffle documentation patch around (Kuniyuki) - Update commit message to explain why IPv6 needs EXPORT_SYMBOL - Link to v3: https://lore.kernel.org/r/20230613-so-reuseport-v3-0-907b4cbb7b99@isovalent.com Changes in v3: - Fix warning re udp_ehashfn and udp6_ehashfn (Simon) - Return higher scoring connected UDP reuseport sockets (Kuniyuki) - Fix ipv6 module builds - Link to v2: https://lore.kernel.org/r/20230613-so-reuseport-v2-0-b7c69a342613@isovalent.com Changes in v2: - Correct commit abbrev length (Kuniyuki) - Reduce duplication (Kuniyuki) - Add checks on sk_state (Martin) - Split exporting inet[6]_lookup_reuseport into separate patch (Eric) --- Daniel Borkmann (1): selftests/bpf: Test that SO_REUSEPORT can be used with sk_assign helper ==================== Signed-off-by:
-
Daniel Borkmann authored
We use two programs to check that the new reuseport logic is executed appropriately. The first is a TC clsact program which bpf_sk_assigns the skb to a UDP or TCP socket created by user space. Since the test communicates via lo we see both directions of packets in the eBPF. Traffic ingressing to the reuseport socket is identified by looking at the destination port. For TCP, we additionally need to make sure that we only assign the initial SYN packets towards our listening socket. The network stack then creates a request socket which transitions to ESTABLISHED after the 3WHS. The second is a reuseport program which shares the fact that it has been executed with user space. This tells us that the delayed lookup mechanism is working. Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Co-developed-by:
-
Lorenz Bauer authored
Currently the bpf_sk_assign helper in tc BPF context refuses SO_REUSEPORT sockets. This means we can't use the helper to steer traffic to Envoy, which configures SO_REUSEPORT on its sockets. In turn, we're blocked from removing TPROXY from our setup. The reason that bpf_sk_assign refuses such sockets is that the bpf_sk_lookup helpers don't execute SK_REUSEPORT programs. Instead, one of the reuseport sockets is selected by hash. This could cause dispatch to the "wrong" socket: sk = bpf_sk_lookup_tcp(...) // select SO_REUSEPORT by hash bpf_sk_assign(skb, sk) // SK_REUSEPORT wasn't executed Fixing this isn't as simple as invoking SK_REUSEPORT from the lookup helpers unfortunately. In the tc context, L2 headers are at the start of the skb, while SK_REUSEPORT expects L3 headers instead. Instead, we execute the SK_REUSEPORT program when the assigned socket is pulled out of the skb, further up the stack. This creates some trickiness with regards to refcounting as bpf_sk_assign will put both refcounted and RCU freed sockets in skb->sk. reuseport sockets are RCU freed. We can infer that the sk_assigned socket is RCU freed if the reuseport lookup succeeds, but convincing yourself of this fact isn't straight forward. Therefore we defensively check refcounting on the sk_assign sock even though it's probably not required in practice. Fixes: 8e368dc7 ("bpf: Fix use of sk->sk_reuseport from sk_assign") Fixes: cf7fbe66 ("bpf: Add socket assign support") Co-developed-by:Kuniyuki Iwashima <kuniyu@amazon.com> Signed-off-by:
-
Lorenz Bauer authored
Now that inet[6]_lookup_reuseport are parameterised on the ehashfn we can remove two sk_lookup helpers. Reviewed-by:
-
Lorenz Bauer authored
The current implementation was extracted from inet[6]_lhash2_lookup in commit 80b373f7 ("inet: Extract helper for selecting socket from reuseport group") and commit 5df65312 ("inet6: Extract helper for selecting socket from reuseport group"). In the original context, sk is always in TCP_LISTEN state and so did not have a separate check. Add documentation that specifies which sk_state are valid to pass to the function. Reviewed-by:
-
Lorenz Bauer authored
There are currently four copies of reuseport_lookup: one each for (TCP, UDP)x(IPv4, IPv6). This forces us to duplicate all callers of those functions as well. This is already the case for sk_lookup helpers (inet,inet6,udp4,udp6)_lookup_run_bpf. There are two differences between the reuseport_lookup helpers: 1. They call different hash functions depending on protocol 2. UDP reuseport_lookup checks that sk_state != TCP_ESTABLISHED Move the check for sk_state into the caller and use the INDIRECT_CALL infrastructure to cut down the helpers to one per IP version. Reviewed-by:
-
Lorenz Bauer authored
Rename the existing reuseport helpers for IPv4 and IPv6 so that they can be invoked in the follow up commit. Export them so that building DCCP and IPv6 as a module works. No change in functionality. Reviewed-by:
-
Lorenz Bauer authored
The semantics for bpf_sk_assign are as follows: sk = some_lookup_func() bpf_sk_assign(skb, sk) bpf_sk_release(sk) That is, the sk is not consumed by bpf_sk_assign. The function therefore needs to make sure that sk lives long enough to be consumed from __inet_lookup_skb. The path through the stack for a TCPv4 packet is roughly: netif_receive_skb_core: takes RCU read lock __netif_receive_skb_core: sch_handle_ingress: tcf_classify: bpf_sk_assign() deliver_ptype_list_skb: deliver_skb: ip_packet_type->func == ip_rcv: ip_rcv_core: ip_rcv_finish_core: dst_input: ip_local_deliver: ip_local_deliver_finish: ip_protocol_deliver_rcu: tcp_v4_rcv: __inet_lookup_skb: skb_steal_sock The existing helper takes advantage of the fact that everything happens in the same RCU critical section: for sockets with SOCK_RCU_FREE set bpf_sk_assign never takes a reference. skb_steal_sock then checks SOCK_RCU_FREE again and does sock_put if necessary. This approach assumes that SOCK_RCU_FREE is never set on a sk between bpf_sk_assign and skb_steal_sock, but this invariant is violated by unhashed UDP sockets. A new UDP socket is created in TCP_CLOSE state but without SOCK_RCU_FREE set. That flag is only added in udp_lib_get_port() which happens when a socket is bound. When bpf_sk_assign was added it wasn't possible to access unhashed UDP sockets from BPF, so this wasn't a problem. This changed in commit 0c48eefa ("sock_map: Lift socket state restriction for datagram sockets"), but the helper wasn't adjusted accordingly. The following sequence of events will therefore lead to a refcount leak: 1. Add socket(AF_INET, SOCK_DGRAM) to a sockmap. 2. Pull socket out of sockmap and bpf_sk_assign it. Since SOCK_RCU_FREE is not set we increment the refcount. 3. bind() or connect() the socket, setting SOCK_RCU_FREE. 4. skb_steal_sock will now set refcounted = false due to SOCK_RCU_FREE. 5. tcp_v4_rcv() skips sock_put(). Fix the problem by rejecting unhashed sockets in bpf_sk_assign(). This matches the behaviour of __inet_lookup_skb which is ultimately the goal of bpf_sk_assign(). Fixes: cf7fbe66 ("bpf: Add socket assign support") Cc: Joe Stringer <joe@cilium.io> Signed-off-by:
-
