summaryrefslogtreecommitdiffstatshomepage
path: root/src (follow)
Commit message (Collapse)AuthorAgeFilesLines
* curve25519: use precomp implementation instead of sandy2xJason A. Donenfeld2018-03-094-3438/+2071
| | | | It's faster and doesn't use the FPU.
* compat: support OpenSUSE 15Jason A. Donenfeld2018-03-091-3/+6
|
* compat: silence warning on frankenkernelsJason A. Donenfeld2018-03-091-0/+3
| | | | Some Android 3.18 devices backport this macro.
* compat: stable kernels are now receiving b87b619Jason A. Donenfeld2018-03-071-1/+1
|
* wg-quick: hide errors on saveJason A. Donenfeld2018-03-041-1/+1
|
* version: bump snapshot0.0.20180304Jason A. Donenfeld2018-03-042-2/+2
|
* wg-quick: if resolvconf/run/iface exists, use itJason A. Donenfeld2018-03-041-1/+1
| | | | | Some older broken resolvconfs don't support resolvconf -l, but do have a file in a standard location, so use it.
* wg-quick: if resolvconf/interface-order exists, use itJason A. Donenfeld2018-03-041-3/+12
| | | | | | | Some older broken resolvconf implementations ignore -m, but do have an interface-order list. It's better to use this list dynamically, in case it changes, or in case it's not used by the OS's resolvconf implementation, such as in the case of systemd or openresolv.
* noise: align static_identity keysJason A. Donenfeld2018-03-021-1/+1
|
* compat: use correct -include pathJason A. Donenfeld2018-03-021-5/+6
|
* noise: fix function prototypeJason A. Donenfeld2018-03-021-1/+1
|
* global: in gnu code, use un-underscored asmJason A. Donenfeld2018-03-023-3/+3
|
* messages: MESSAGE_TOTAL is unusedJason A. Donenfeld2018-03-021-2/+1
|
* crypto: read only after initJason A. Donenfeld2018-03-024-10/+11
|
* Kconfig: require DST_CACHE explicitlyJason A. Donenfeld2018-03-021-0/+1
|
* compat: workaround netlink refcount bugJason A. Donenfeld2018-03-021-1/+1
| | | | | | | For now it looks like only 4.16 has this, but we'll keep track in case others add it too. Upstream-fix: b87b6194be63 ("netlink: put module reference if dump start fails")
* allowedips: fix comment styleJason A. Donenfeld2018-02-211-1/+2
|
* queueing: skb_reset: mark as xnetJason A. Donenfeld2018-02-201-1/+1
| | | | | | | | This was avoided for a long time, because I wanted the packet to be charged to the original socket for as long as possible. However, this broke net_cls, which looks at skb->sk for additional late-stage routing decisions. So we had no choice but to ensure that skb->sk is NULL by the time of xmit, and this means calling the skb destructor.
* version: bump snapshot0.0.20180218Jason A. Donenfeld2018-02-182-2/+2
|
* tools: fixup errno handlingJason A. Donenfeld2018-02-172-8/+8
|
* tools: FreeBSD doesn't have EAI_NODATAJason A. Donenfeld2018-02-171-2/+15
|
* tools: do not collide types with libc clashesJason A. Donenfeld2018-02-171-4/+10
|
* wg(8): clarify phrasingJason A. Donenfeld2018-02-171-3/+3
|
* tools: allow in-line commentsJason A. Donenfeld2018-02-173-11/+17
|
* tools: normalize strncpy/snprintf usageJason A. Donenfeld2018-02-144-18/+22
|
* blake2s: use union instead of castingJason A. Donenfeld2018-02-142-19/+17
| | | | | This deals with alignment more easily and also helps squelch a clang-analyzer warning.
* allowedips: indicate to clang-analyzer that trie is non-nullJason A. Donenfeld2018-02-141-0/+1
| | | | | We check it in the block just above the only call to node_placement, so we're certain this is the case.
* socket: free skb if there isn't an endpointJason A. Donenfeld2018-02-121-0/+2
| | | | | | | | | | | | | | Because the packet transmission interface in Linux, in general, has the characteristic of "consuming" skbs, our own socket_send family too must always consume skbs. This means that when a caller passes an skb to a socket_send function, he looses ownership and expects for the skb to be freed by the socket_send function, not by himself. Therefore, it's important that we actually carry out this contract under all circumstances, even when there's no endpoint available. This condition could be hit by setting up a peer with a persistent keepalive but no endpoint, or by removing an endpoint once packets were queued up to be sent already.
* qemu: more granular memleak detectionJason A. Donenfeld2018-02-121-2/+15
|
* netns: replace n0 ip with ip0, per customJason A. Donenfeld2018-02-121-2/+2
|
* poly1305: fix up selftest counterJason A. Donenfeld2018-02-081-1/+1
|
* qemu: add support for powerpcJason A. Donenfeld2018-02-072-2/+21
|
* wg-quick: match from beginning rather than shift rightJason A. Donenfeld2018-02-061-2/+2
| | | | | Before, this meant that it simply took the last 15 characters, instead of erroring out when there's more than 15 chars.
* tools: endian.h is not portableJason A. Donenfeld2018-02-051-4/+9
|
* version: bump snapshot0.0.20180202Jason A. Donenfeld2018-02-022-2/+2
|
* qemu: update base versionsJason A. Donenfeld2018-02-021-2/+2
|
* curve25519: break more things with more test casesJason A. Donenfeld2018-02-011-0/+12
|
* curve25519: replace fiat64 with faster hacl64Jason A. Donenfeld2018-02-015-474/+887
| | | | | This reverts commit da4ff396cc5d5e0ff21f9ecbc2f951c048c63fff and adds some optimizations to hacl64.
* curve25519: replace hacl64 with fiat64Jason A. Donenfeld2018-02-015-880/+474
| | | | | | | | | | For now, it's faster: hacl64: 109782 cycles per call fiat64: 108984 cycles per call It's quite possible this commit will be reverted with nice changes from INRIA, though.
* curve25519: verify that specialized basepoint implementations are correctJason A. Donenfeld2018-02-011-3/+17
|
* tools: dedup secret normalizationJason A. Donenfeld2018-01-311-7/+1
|
* chacha20poly1305: better buffer alignmentJason A. Donenfeld2018-01-301-9/+8
|
* chacha20poly1305: use existing rol32 functionJason A. Donenfeld2018-01-301-9/+4
|
* tools: fread doesn't change errnoJason A. Donenfeld2018-01-301-4/+0
| | | | | Thus we might be responding to an old errno, which could cause this to unnecessarily fail.
* device: let udev know what kind of device we areJason A. Donenfeld2018-01-301-0/+6
|
* qemu: disable AVX-512 in userlandJason A. Donenfeld2018-01-231-0/+2
| | | | | | | We're trying to test kernel code, so the userland tools we use for doing that testing don't really matter to us. We turn off AVX512, because WireGuard supports kernels that do not support AVX512 userlands. It's easier to just blanket-disable it all, since it doesn't matter anyway.
* qemu: disable PIE for compilationJason A. Donenfeld2018-01-231-1/+3
| | | | | | Some old kernels never backported this fix to the build system, and it's required if we want to build those old kernels with PIE-by-default compilers.
* tools: share curve25519 implementations with kernelJason A. Donenfeld2018-01-232-1541/+32
|
* poly1305: add poly-specific self-testsJason A. Donenfeld2018-01-194-1/+1569
|
* curve25519-fiat32: uninline certain functionsJason A. Donenfeld2018-01-182-8/+8
| | | | | | | | | | | While this has a negative performance impact on x86_64, it has a positive performance impact on smaller machines, which is where we're actually using this code. For example, an A53: Before: fiat32: 228605 cycles per call After: fiat32: 188307 cycles per call
Copyright © 1996 – 2023 Jason A. Donenfeld. All Rights Reverse Engineered.