Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | data: simplify no-keypair failure case | Samuel Holland | 2017-07-20 | 1 | -5/+2 |
| | | | | Signed-off-by: Samuel Holland <samuel@sholland.org> | ||||
* | data: use KMEM_CACHE macro | Jason A. Donenfeld | 2017-07-20 | 1 | -2/+2 |
| | | | | Suggested-by: Samuel Holland <samuel@sholland.org> | ||||
* | global: use pointer to net_device | Jason A. Donenfeld | 2017-07-20 | 1 | -1/+1 |
| | | | | | | DaveM prefers it to be this way per [1]. [1] http://www.spinics.net/lists/netdev/msg443992.html | ||||
* | counter: use correct unit for indices | Jason A. Donenfeld | 2017-07-06 | 1 | -3/+3 |
| | | | | | Even though redundant bits == bits per long, we're indexing into something that uses longs as its unit, so this is correct. | ||||
* | main: annotate init/exit functions to save memory | Jason A. Donenfeld | 2017-06-24 | 1 | -1/+1 |
| | |||||
* | receive: trim incoming packets to IP header length | Jason A. Donenfeld | 2017-06-01 | 1 | -1/+1 |
| | |||||
* | debug: print interface name in dmesg | Jason A. Donenfeld | 2017-05-31 | 1 | -1/+1 |
| | |||||
* | data: enable BH during parallel crypto on ARM/NEON | Jason A. Donenfeld | 2017-05-31 | 1 | -0/+12 |
| | | | | | | padata disables it, but in order to use SIMD on ARM, we can't be in an interrupt. We only do this on ARM since it adds jitter to the performance. | ||||
* | style: spaces after for loops | Jason A. Donenfeld | 2017-05-30 | 1 | -2/+2 |
| | |||||
* | data: alloca is actually as dangerous as they say | Jason A. Donenfeld | 2017-04-09 | 1 | -6/+4 |
| | | | | | | | | | | | | | It turns out that calling alloca from an inline function means that the memory isn't ever deallocated until the caller function exits, which means we were using tons of stack space for every iteration of the call. So, we hard code the sg array. While 128 seems like a reasonable number, we actually wind up using "MAX_SKB_FRAGS * 2 + 1". An skb has its data segment, so that's 1. Then it has its frags, which are MAX_SKB_FRAGS at max. Then it has its frag list, which, so far as I can tell, are potentially unbounded. So we just hope it's no more than MAX_SKB_FRAGS, and so we plan for at most two of those. | ||||
* | data: cleanup parallel workqueue and use two max_active | Jason A. Donenfeld | 2017-04-08 | 1 | -2/+2 |
| | |||||
* | data: simplify flow | Jason A. Donenfeld | 2017-04-04 | 1 | -81/+36 |
| | |||||
* | chacha20poly1305: check return values of sgops | Jason A. Donenfeld | 2017-04-04 | 1 | -5/+5 |
| | |||||
* | locking: always use _bh | Jason A. Donenfeld | 2017-04-04 | 1 | -6/+6 |
| | | | | | All locks are potentially between user context and softirq, which means we need to take the _bh variant. | ||||
* | data: big refactoring | Jason A. Donenfeld | 2017-03-20 | 1 | -96/+71 |
| | |||||
* | data: transition to skb_reset_tc for 4.11 | Jason A. Donenfeld | 2017-02-27 | 1 | -3/+1 |
| | |||||
* | Use __read_mostly attribute when possible | Jason A. Donenfeld | 2017-01-13 | 1 | -2/+2 |
| | |||||
* | Update copyright | Jason A. Donenfeld | 2017-01-10 | 1 | -1/+1 |
| | |||||
* | messages: increase header by 3 bytes for alignment | Jason A. Donenfeld | 2016-12-16 | 1 | -1/+1 |
| | |||||
* | types: enforce consistency | Jason A. Donenfeld | 2016-12-16 | 1 | -5/+5 |
| | |||||
* | global: move to consistent use of uN instead of uintN_t for kernel code | Jason A. Donenfeld | 2016-12-11 | 1 | -8/+8 |
| | |||||
* | data: reset tc when resetting skb | Jason A. Donenfeld | 2016-12-09 | 1 | -0/+6 |
| | |||||
* | headers: cleanup notices | Jason A. Donenfeld | 2016-11-21 | 1 | -1/+1 |
| | |||||
* | various: nits from willy | Jason A. Donenfeld | 2016-11-15 | 1 | -4/+2 |
| | |||||
* | socket: keep track of src address in sending packets | Jason A. Donenfeld | 2016-11-15 | 1 | -8/+8 |
| | |||||
* | data: only uses kmem_cache for parallism | Jason A. Donenfeld | 2016-11-06 | 1 | -0/+2 |
| | |||||
* | data: squelch compiler warning on PARALLEL=n | Jason A. Donenfeld | 2016-11-06 | 1 | -1/+1 |
| | |||||
* | data: we care about per-peer, not per-device, inflight encryptions | Jason A. Donenfeld | 2016-11-06 | 1 | -1/+4 |
| | |||||
* | data: use a memory cache for parallel ctx | Jason A. Donenfeld | 2016-11-04 | 1 | -60/+81 |
| | |||||
* | data: keep FPU on when possible | Jason A. Donenfeld | 2016-11-04 | 1 | -7/+5 |
| | |||||
* | send: queue bundles on same CPU | Jason A. Donenfeld | 2016-11-04 | 1 | -54/+95 |
| | |||||
* | data: use smaller types | Jason A. Donenfeld | 2016-11-04 | 1 | -6/+6 |
| | |||||
* | data: take reference to peer | Jason A. Donenfeld | 2016-11-04 | 1 | -1/+8 |
| | |||||
* | data: do not allow usage of keypair just before hash removal | Jason A. Donenfeld | 2016-11-01 | 1 | -4/+2 |
| | |||||
* | data: reset all packet fields like tun.c | Jason A. Donenfeld | 2016-10-23 | 1 | -2/+2 |
| | |||||
* | send: only avoid parallel path when there aren't inflight jobs | Jason A. Donenfeld | 2016-10-02 | 1 | -1/+1 |
| | | | | Otherwise we get packet reordering. | ||||
* | Rework headers and includes | Jason A. Donenfeld | 2016-09-29 | 1 | -3/+5 |
| | |||||
* | selftest: move to subfolder | Jason A. Donenfeld | 2016-08-02 | 1 | -88/+1 |
| | |||||
* | tests: make fatal | Jason A. Donenfeld | 2016-06-25 | 1 | -1/+2 |
| | |||||
* | nonce: switch to RFC6479 to better support packet reordering | Jason A. Donenfeld | 2016-06-25 | 1 | -58/+86 |
| | | | | | With packets hitting multiple cores, a 64bit backtrack was too small. This algorithm increases our backtrack to 1984bits. | ||||
* | Initial commit | Jason A. Donenfeld | 2016-06-25 | 1 | -0/+477 |