| Message ID | 20210112223847.1915615-4-sdf@google.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | bpf: misc performance improvements for cgroup hooks | expand |
On Wed, Jan 13, 2021 at 11:03 AM Martin KaFai Lau <kafai@fb.com> wrote: > > On Tue, Jan 12, 2021 at 02:38:46PM -0800, Stanislav Fomichev wrote: > > When we attach a bpf program to cgroup/getsockopt any other getsockopt() > > syscall starts incurring kzalloc/kfree cost. > > > > Let add a small buffer on the stack and use it for small (majority) > > {s,g}etsockopt values. The buffer is small enough to fit into > > the cache line and cover the majority of simple options (most > > of them are 4 byte ints). > > > > It seems natural to do the same for setsockopt, but it's a bit more > > involved when the BPF program modifies the data (where we have to > > kmalloc). The assumption is that for the majority of setsockopt > > calls (which are doing pure BPF options or apply policy) this > > will bring some benefit as well. > > > > Without this patch (we remove about 1% __kmalloc): > > 3.38% 0.07% tcp_mmap [kernel.kallsyms] [k] __cgroup_bpf_run_filter_getsockopt > > | > > --3.30%--__cgroup_bpf_run_filter_getsockopt > > | > > --0.81%--__kmalloc > > > > Signed-off-by: Stanislav Fomichev <sdf@google.com> > > Cc: Martin KaFai Lau <kafai@fb.com> > > Cc: Song Liu <songliubraving@fb.com> > > --- > > include/linux/filter.h | 5 ++++ > > kernel/bpf/cgroup.c | 52 ++++++++++++++++++++++++++++++++++++------ > > 2 files changed, 50 insertions(+), 7 deletions(-) > > > > diff --git a/include/linux/filter.h b/include/linux/filter.h > > index 29c27656165b..8739f1d4cac4 100644 > > --- a/include/linux/filter.h > > +++ b/include/linux/filter.h > > @@ -1281,6 +1281,11 @@ struct bpf_sysctl_kern { > > u64 tmp_reg; > > }; > > > > +#define BPF_SOCKOPT_KERN_BUF_SIZE 32 > > +struct bpf_sockopt_buf { > > + u8 data[BPF_SOCKOPT_KERN_BUF_SIZE]; > > +}; > > + > > struct bpf_sockopt_kern { > > struct sock *sk; > > u8 *optval; > > diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c > > index 416e7738981b..dbeef7afbbf9 100644 > > --- a/kernel/bpf/cgroup.c > > +++ b/kernel/bpf/cgroup.c > > @@ -1298,7 +1298,8 @@ static bool __cgroup_bpf_prog_array_is_empty(struct cgroup *cgrp, > > return empty; > > } > > > > -static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen) > > +static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen, > > + struct bpf_sockopt_buf *buf) > > { > > if (unlikely(max_optlen < 0)) > > return -EINVAL; > > @@ -1310,6 +1311,15 @@ static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen) > > max_optlen = PAGE_SIZE; > > } > > > > + if (max_optlen <= sizeof(buf->data)) { > > + /* When the optval fits into BPF_SOCKOPT_KERN_BUF_SIZE > > + * bytes avoid the cost of kzalloc. > > + */ > > + ctx->optval = buf->data; > > + ctx->optval_end = ctx->optval + max_optlen; > > + return max_optlen; > > + } > > + > > ctx->optval = kzalloc(max_optlen, GFP_USER); > > if (!ctx->optval) > > return -ENOMEM; > > @@ -1319,16 +1329,26 @@ static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen) > > return max_optlen; > > } > > > > -static void sockopt_free_buf(struct bpf_sockopt_kern *ctx) > > +static void sockopt_free_buf(struct bpf_sockopt_kern *ctx, > > + struct bpf_sockopt_buf *buf) > > { > > + if (ctx->optval == buf->data) > > + return; > > kfree(ctx->optval); > > } > > > > +static bool sockopt_buf_allocated(struct bpf_sockopt_kern *ctx, > > + struct bpf_sockopt_buf *buf) > > +{ > > + return ctx->optval != buf->data; > > +} > > + > > int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level, > > int *optname, char __user *optval, > > int *optlen, char **kernel_optval) > > { > > struct cgroup *cgrp = sock_cgroup_ptr(&sk->sk_cgrp_data); > > + struct bpf_sockopt_buf buf = {}; > > struct bpf_sockopt_kern ctx = { > > .sk = sk, > > .level = *level, > > @@ -1350,7 +1370,7 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level, > > */ > > max_optlen = max_t(int, 16, *optlen); > > > > - max_optlen = sockopt_alloc_buf(&ctx, max_optlen); > > + max_optlen = sockopt_alloc_buf(&ctx, max_optlen, &buf); > > if (max_optlen < 0) > > return max_optlen; > > > > @@ -1390,14 +1410,31 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level, > > */ > > if (ctx.optlen != 0) { > > *optlen = ctx.optlen; > > - *kernel_optval = ctx.optval; > > + /* We've used bpf_sockopt_kern->buf as an intermediary > > + * storage, but the BPF program indicates that we need > > + * to pass this data to the kernel setsockopt handler. > > + * No way to export on-stack buf, have to allocate a > > + * new buffer. > > + */ > > + if (!sockopt_buf_allocated(&ctx, &buf)) { > > + void *p = kzalloc(ctx.optlen, GFP_USER); > nit. zero-ing is unnecessary when memcpy() will be done later. SG, will switch to kmalloc, thanks! > Acked-by: Martin KaFai Lau <kafai@fb.com>
diff --git a/include/linux/filter.h b/include/linux/filter.h index 29c27656165b..8739f1d4cac4 100644 --- a/include/linux/filter.h +++ b/include/linux/filter.h @@ -1281,6 +1281,11 @@ struct bpf_sysctl_kern { u64 tmp_reg; }; +#define BPF_SOCKOPT_KERN_BUF_SIZE 32 +struct bpf_sockopt_buf { + u8 data[BPF_SOCKOPT_KERN_BUF_SIZE]; +}; + struct bpf_sockopt_kern { struct sock *sk; u8 *optval; diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c index 416e7738981b..dbeef7afbbf9 100644 --- a/kernel/bpf/cgroup.c +++ b/kernel/bpf/cgroup.c @@ -1298,7 +1298,8 @@ static bool __cgroup_bpf_prog_array_is_empty(struct cgroup *cgrp, return empty; } -static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen) +static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen, + struct bpf_sockopt_buf *buf) { if (unlikely(max_optlen < 0)) return -EINVAL; @@ -1310,6 +1311,15 @@ static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen) max_optlen = PAGE_SIZE; } + if (max_optlen <= sizeof(buf->data)) { + /* When the optval fits into BPF_SOCKOPT_KERN_BUF_SIZE + * bytes avoid the cost of kzalloc. + */ + ctx->optval = buf->data; + ctx->optval_end = ctx->optval + max_optlen; + return max_optlen; + } + ctx->optval = kzalloc(max_optlen, GFP_USER); if (!ctx->optval) return -ENOMEM; @@ -1319,16 +1329,26 @@ static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen) return max_optlen; } -static void sockopt_free_buf(struct bpf_sockopt_kern *ctx) +static void sockopt_free_buf(struct bpf_sockopt_kern *ctx, + struct bpf_sockopt_buf *buf) { + if (ctx->optval == buf->data) + return; kfree(ctx->optval); } +static bool sockopt_buf_allocated(struct bpf_sockopt_kern *ctx, + struct bpf_sockopt_buf *buf) +{ + return ctx->optval != buf->data; +} + int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level, int *optname, char __user *optval, int *optlen, char **kernel_optval) { struct cgroup *cgrp = sock_cgroup_ptr(&sk->sk_cgrp_data); + struct bpf_sockopt_buf buf = {}; struct bpf_sockopt_kern ctx = { .sk = sk, .level = *level, @@ -1350,7 +1370,7 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level, */ max_optlen = max_t(int, 16, *optlen); - max_optlen = sockopt_alloc_buf(&ctx, max_optlen); + max_optlen = sockopt_alloc_buf(&ctx, max_optlen, &buf); if (max_optlen < 0) return max_optlen; @@ -1390,14 +1410,31 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level, */ if (ctx.optlen != 0) { *optlen = ctx.optlen; - *kernel_optval = ctx.optval; + /* We've used bpf_sockopt_kern->buf as an intermediary + * storage, but the BPF program indicates that we need + * to pass this data to the kernel setsockopt handler. + * No way to export on-stack buf, have to allocate a + * new buffer. + */ + if (!sockopt_buf_allocated(&ctx, &buf)) { + void *p = kzalloc(ctx.optlen, GFP_USER); + + if (!p) { + ret = -ENOMEM; + goto out; + } + memcpy(p, ctx.optval, ctx.optlen); + *kernel_optval = p; + } else { + *kernel_optval = ctx.optval; + } /* export and don't free sockopt buf */ return 0; } } out: - sockopt_free_buf(&ctx); + sockopt_free_buf(&ctx, &buf); return ret; } @@ -1407,6 +1444,7 @@ int __cgroup_bpf_run_filter_getsockopt(struct sock *sk, int level, int retval) { struct cgroup *cgrp = sock_cgroup_ptr(&sk->sk_cgrp_data); + struct bpf_sockopt_buf buf = {}; struct bpf_sockopt_kern ctx = { .sk = sk, .level = level, @@ -1425,7 +1463,7 @@ int __cgroup_bpf_run_filter_getsockopt(struct sock *sk, int level, ctx.optlen = max_optlen; - max_optlen = sockopt_alloc_buf(&ctx, max_optlen); + max_optlen = sockopt_alloc_buf(&ctx, max_optlen, &buf); if (max_optlen < 0) return max_optlen; @@ -1483,7 +1521,7 @@ int __cgroup_bpf_run_filter_getsockopt(struct sock *sk, int level, ret = ctx.retval; out: - sockopt_free_buf(&ctx); + sockopt_free_buf(&ctx, &buf); return ret; }
When we attach a bpf program to cgroup/getsockopt any other getsockopt() syscall starts incurring kzalloc/kfree cost. Let add a small buffer on the stack and use it for small (majority) {s,g}etsockopt values. The buffer is small enough to fit into the cache line and cover the majority of simple options (most of them are 4 byte ints). It seems natural to do the same for setsockopt, but it's a bit more involved when the BPF program modifies the data (where we have to kmalloc). The assumption is that for the majority of setsockopt calls (which are doing pure BPF options or apply policy) this will bring some benefit as well. Without this patch (we remove about 1% __kmalloc): 3.38% 0.07% tcp_mmap [kernel.kallsyms] [k] __cgroup_bpf_run_filter_getsockopt | --3.30%--__cgroup_bpf_run_filter_getsockopt | --0.81%--__kmalloc Signed-off-by: Stanislav Fomichev <sdf@google.com> Cc: Martin KaFai Lau <kafai@fb.com> Cc: Song Liu <songliubraving@fb.com> --- include/linux/filter.h | 5 ++++ kernel/bpf/cgroup.c | 52 ++++++++++++++++++++++++++++++++++++------ 2 files changed, 50 insertions(+), 7 deletions(-)