diff options
| author | Mateusz Guzik <mjg@FreeBSD.org> | 2018-06-08 21:40:03 +0000 |
|---|---|---|
| committer | Mateusz Guzik <mjg@FreeBSD.org> | 2018-06-08 21:40:03 +0000 |
| commit | 4e180881ae4a325cfaeef336aa1a16b0d5bbae67 (patch) | |
| tree | c5a4fb99800ba3d30a7f2b689c21dc1e6ff74ceb /sys/netipsec | |
| parent | 4f63fbc955b7de6332b3c5507f57389dbd6e75f8 (diff) | |
| download | src-4e180881ae4a325cfaeef336aa1a16b0d5bbae67.tar.gz src-4e180881ae4a325cfaeef336aa1a16b0d5bbae67.zip | |
uma: implement provisional api for per-cpu zones
Per-cpu zone allocations are very rarely done compared to regular zones.
The intent is to avoid pessimizing the latter case with per-cpu specific
code.
In particular contrary to the claim in r334824, M_ZERO is sometimes being
used for such zones. But the zeroing method is completely different and
braching on it in the fast path for regular zones is a waste of time.
Notes
Notes:
svn path=/head/; revision=334858
Diffstat (limited to 'sys/netipsec')
| -rw-r--r-- | sys/netipsec/key.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/sys/netipsec/key.c b/sys/netipsec/key.c index 96ef209518d1..0b90d46d4d0b 100644 --- a/sys/netipsec/key.c +++ b/sys/netipsec/key.c @@ -2957,7 +2957,7 @@ key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx, goto done; } mtx_init(sav->lock, "ipsec association", NULL, MTX_DEF); - sav->lft_c = uma_zalloc(V_key_lft_zone, M_NOWAIT); + sav->lft_c = uma_zalloc_pcpu(V_key_lft_zone, M_NOWAIT); if (sav->lft_c == NULL) { *errp = ENOBUFS; goto done; @@ -3049,7 +3049,7 @@ done: free(sav->lock, M_IPSEC_MISC); } if (sav->lft_c != NULL) - uma_zfree(V_key_lft_zone, sav->lft_c); + uma_zfree_pcpu(V_key_lft_zone, sav->lft_c); free(sav, M_IPSEC_SA), sav = NULL; } if (sah != NULL) |