From 49f1048f3d89f9986a5f0c1cc8678f49f020e956 Mon Sep 17 00:00:00 2001 From: Amery Hung Date: Fri, 19 Sep 2025 16:09:46 -0700 Subject: [PATCH 1/7] bpf: Clear pfmemalloc flag when freeing all fragments It is possible for bpf_xdp_adjust_tail() to free all fragments. The kfunc currently clears the XDP_FLAGS_HAS_FRAGS bit, but not XDP_FLAGS_FRAGS_PF_MEMALLOC. So far, this has not caused a issue when building sk_buff from xdp_buff since all readers of xdp_buff->flags use the flag only when there are fragments. Clear the XDP_FLAGS_FRAGS_PF_MEMALLOC bit as well to make the flags correct. Signed-off-by: Amery Hung Reviewed-by: Maciej Fijalkowski --- include/net/xdp.h | 5 +++++ net/core/filter.c | 1 + 2 files changed, 6 insertions(+) diff --git a/include/net/xdp.h b/include/net/xdp.h index b40f1f96cb117..f288c348a6c13 100644 --- a/include/net/xdp.h +++ b/include/net/xdp.h @@ -115,6 +115,11 @@ static __always_inline void xdp_buff_set_frag_pfmemalloc(struct xdp_buff *xdp) xdp->flags |= XDP_FLAGS_FRAGS_PF_MEMALLOC; } +static __always_inline void xdp_buff_clear_frag_pfmemalloc(struct xdp_buff *xdp) +{ + xdp->flags &= ~XDP_FLAGS_FRAGS_PF_MEMALLOC; +} + static __always_inline void xdp_init_buff(struct xdp_buff *xdp, u32 frame_sz, struct xdp_rxq_info *rxq) { diff --git a/net/core/filter.c b/net/core/filter.c index 8342f810ad856..a3902ec01355c 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -4210,6 +4210,7 @@ static int bpf_xdp_frags_shrink_tail(struct xdp_buff *xdp, int offset) if (unlikely(!sinfo->nr_frags)) { xdp_buff_clear_frags_flag(xdp); + xdp_buff_clear_frag_pfmemalloc(xdp); xdp->data_end -= offset; } From f792139fc1e9429bb3fcc49720cdc521a5f4157b Mon Sep 17 00:00:00 2001 From: Amery Hung Date: Fri, 19 Sep 2025 16:09:47 -0700 Subject: [PATCH 2/7] bpf: Allow bpf_xdp_shrink_data to shrink a frag from head and tail Move skb_frag_t adjustment into bpf_xdp_shrink_data() and extend its functionality to be able to shrink an xdp fragment from both head and tail. In a later patch, bpf_xdp_pull_data() will reuse it to shrink an xdp fragment from head. Additionally, in bpf_xdp_frags_shrink_tail(), breaking the loop when bpf_xdp_shrink_data() returns false (i.e., not releasing the current fragment) is not necessary as the loop condition, offset > 0, has the same effect. Remove the else branch to simplify the code. Signed-off-by: Amery Hung Reviewed-by: Maciej Fijalkowski --- include/net/xdp_sock_drv.h | 21 ++++++++++++++++--- net/core/filter.c | 41 ++++++++++++++++++++++---------------- 2 files changed, 42 insertions(+), 20 deletions(-) diff --git a/include/net/xdp_sock_drv.h b/include/net/xdp_sock_drv.h index 513c8e9704f65..4f2d3268a6769 100644 --- a/include/net/xdp_sock_drv.h +++ b/include/net/xdp_sock_drv.h @@ -160,13 +160,23 @@ static inline struct xdp_buff *xsk_buff_get_frag(const struct xdp_buff *first) return ret; } -static inline void xsk_buff_del_tail(struct xdp_buff *tail) +static inline void xsk_buff_del_frag(struct xdp_buff *xdp) { - struct xdp_buff_xsk *xskb = container_of(tail, struct xdp_buff_xsk, xdp); + struct xdp_buff_xsk *xskb = container_of(xdp, struct xdp_buff_xsk, xdp); list_del(&xskb->list_node); } +static inline struct xdp_buff *xsk_buff_get_head(struct xdp_buff *first) +{ + struct xdp_buff_xsk *xskb = container_of(first, struct xdp_buff_xsk, xdp); + struct xdp_buff_xsk *frag; + + frag = list_first_entry(&xskb->pool->xskb_list, struct xdp_buff_xsk, + list_node); + return &frag->xdp; +} + static inline struct xdp_buff *xsk_buff_get_tail(struct xdp_buff *first) { struct xdp_buff_xsk *xskb = container_of(first, struct xdp_buff_xsk, xdp); @@ -389,8 +399,13 @@ static inline struct xdp_buff *xsk_buff_get_frag(const struct xdp_buff *first) return NULL; } -static inline void xsk_buff_del_tail(struct xdp_buff *tail) +static inline void xsk_buff_del_frag(struct xdp_buff *xdp) +{ +} + +static inline struct xdp_buff *xsk_buff_get_head(struct xdp_buff *first) { + return NULL; } static inline struct xdp_buff *xsk_buff_get_tail(struct xdp_buff *first) diff --git a/net/core/filter.c b/net/core/filter.c index a3902ec01355c..bd31a5a83ef75 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -4153,34 +4153,45 @@ static int bpf_xdp_frags_increase_tail(struct xdp_buff *xdp, int offset) return 0; } -static void bpf_xdp_shrink_data_zc(struct xdp_buff *xdp, int shrink, - enum xdp_mem_type mem_type, bool release) +static struct xdp_buff *bpf_xdp_shrink_data_zc(struct xdp_buff *xdp, int shrink, + bool tail, bool release) { - struct xdp_buff *zc_frag = xsk_buff_get_tail(xdp); + struct xdp_buff *zc_frag = tail ? xsk_buff_get_tail(xdp) : + xsk_buff_get_head(xdp); if (release) { - xsk_buff_del_tail(zc_frag); - __xdp_return(0, mem_type, false, zc_frag); + xsk_buff_del_frag(zc_frag); } else { - zc_frag->data_end -= shrink; + if (tail) + zc_frag->data_end -= shrink; + else + zc_frag->data += shrink; } + + return zc_frag; } static bool bpf_xdp_shrink_data(struct xdp_buff *xdp, skb_frag_t *frag, - int shrink) + int shrink, bool tail) { enum xdp_mem_type mem_type = xdp->rxq->mem.type; bool release = skb_frag_size(frag) == shrink; + netmem_ref netmem = skb_frag_netmem(frag); + struct xdp_buff *zc_frag = NULL; if (mem_type == MEM_TYPE_XSK_BUFF_POOL) { - bpf_xdp_shrink_data_zc(xdp, shrink, mem_type, release); - goto out; + netmem = 0; + zc_frag = bpf_xdp_shrink_data_zc(xdp, shrink, tail, release); } - if (release) - __xdp_return(skb_frag_netmem(frag), mem_type, false, NULL); + if (release) { + __xdp_return(netmem, mem_type, false, zc_frag); + } else { + if (!tail) + skb_frag_off_add(frag, shrink); + skb_frag_size_sub(frag, shrink); + } -out: return release; } @@ -4198,12 +4209,8 @@ static int bpf_xdp_frags_shrink_tail(struct xdp_buff *xdp, int offset) len_free += shrink; offset -= shrink; - if (bpf_xdp_shrink_data(xdp, frag, shrink)) { + if (bpf_xdp_shrink_data(xdp, frag, shrink, true)) n_frags_free++; - } else { - skb_frag_size_sub(frag, shrink); - break; - } } sinfo->nr_frags -= n_frags_free; sinfo->xdp_frags_size -= len_free; From 1d25bf54d2e63d5d90b5516face1d852d5d6bcb0 Mon Sep 17 00:00:00 2001 From: Amery Hung Date: Fri, 19 Sep 2025 16:09:48 -0700 Subject: [PATCH 3/7] bpf: Support pulling non-linear xdp data Add kfunc, bpf_xdp_pull_data(), to support pulling data from xdp fragments. Similar to bpf_skb_pull_data(), bpf_xdp_pull_data() makes the first len bytes of data directly readable and writable in bpf programs. If the "len" argument is larger than the linear data size, data in fragments will be copied to the linear data area when there is enough room. Specifically, the kfunc will try to use the tailroom first. When the tailroom is not enough, metadata and data will be shifted down to make room for pulling data. A use case of the kfunc is to decapsulate headers residing in xdp fragments. It is possible for a NIC driver to place headers in xdp fragments. To keep using direct packet access for parsing and decapsulating headers, users can pull headers into the linear data area by calling bpf_xdp_pull_data() and then pop the header with bpf_xdp_adjust_head(). Reviewed-by: Jakub Kicinski Signed-off-by: Amery Hung --- net/core/filter.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 93 insertions(+) diff --git a/net/core/filter.c b/net/core/filter.c index bd31a5a83ef75..b20d59bb19b8c 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -12218,6 +12218,98 @@ __bpf_kfunc int bpf_sock_ops_enable_tx_tstamp(struct bpf_sock_ops_kern *skops, return 0; } +/** + * bpf_xdp_pull_data() - Pull in non-linear xdp data. + * @x: &xdp_md associated with the XDP buffer + * @len: length of data to be made directly accessible in the linear part + * + * Pull in data in case the XDP buffer associated with @x is non-linear and + * not all @len are in the linear data area. + * + * Direct packet access allows reading and writing linear XDP data through + * packet pointers (i.e., &xdp_md->data + offsets). The amount of data which + * ends up in the linear part of the xdp_buff depends on the NIC and its + * configuration. When a frag-capable XDP program wants to directly access + * headers that may be in the non-linear area, call this kfunc to make sure + * the data is available in the linear area. Alternatively, use dynptr or + * bpf_xdp_{load,store}_bytes() to access data without pulling. + * + * This kfunc can also be used with bpf_xdp_adjust_head() to decapsulate + * headers in the non-linear data area. + * + * A call to this kfunc may reduce headroom. If there is not enough tailroom + * in the linear data area, metadata and data will be shifted down. + * + * A call to this kfunc is susceptible to change the buffer geometry. + * Therefore, at load time, all checks on pointers previously done by the + * verifier are invalidated and must be performed again, if the kfunc is used + * in combination with direct packet access. + * + * Return: + * * %0 - success + * * %-EINVAL - invalid len + */ +__bpf_kfunc int bpf_xdp_pull_data(struct xdp_md *x, u32 len) +{ + struct xdp_buff *xdp = (struct xdp_buff *)x; + struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); + int i, delta, shift, headroom, tailroom, n_frags_free = 0; + void *data_hard_end = xdp_data_hard_end(xdp); + int data_len = xdp->data_end - xdp->data; + void *start; + + if (len <= data_len) + return 0; + + if (unlikely(len > xdp_get_buff_len(xdp))) + return -EINVAL; + + start = xdp_data_meta_unsupported(xdp) ? xdp->data : xdp->data_meta; + + headroom = start - xdp->data_hard_start - sizeof(struct xdp_frame); + tailroom = data_hard_end - xdp->data_end; + + delta = len - data_len; + if (unlikely(delta > tailroom + headroom)) + return -EINVAL; + + shift = delta - tailroom; + if (shift > 0) { + memmove(start - shift, start, xdp->data_end - start); + + xdp->data_meta -= shift; + xdp->data -= shift; + xdp->data_end -= shift; + } + + for (i = 0; i < sinfo->nr_frags && delta; i++) { + skb_frag_t *frag = &sinfo->frags[i]; + u32 shrink = min_t(u32, delta, skb_frag_size(frag)); + + memcpy(xdp->data_end, skb_frag_address(frag), shrink); + + xdp->data_end += shrink; + sinfo->xdp_frags_size -= shrink; + delta -= shrink; + if (bpf_xdp_shrink_data(xdp, frag, shrink, false)) + n_frags_free++; + } + + if (unlikely(n_frags_free)) { + memmove(sinfo->frags, sinfo->frags + n_frags_free, + (sinfo->nr_frags - n_frags_free) * sizeof(skb_frag_t)); + + sinfo->nr_frags -= n_frags_free; + + if (!sinfo->nr_frags) { + xdp_buff_clear_frags_flag(xdp); + xdp_buff_clear_frag_pfmemalloc(xdp); + } + } + + return 0; +} + __bpf_kfunc_end_defs(); int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags, @@ -12245,6 +12337,7 @@ BTF_KFUNCS_END(bpf_kfunc_check_set_skb_meta) BTF_KFUNCS_START(bpf_kfunc_check_set_xdp) BTF_ID_FLAGS(func, bpf_dynptr_from_xdp) +BTF_ID_FLAGS(func, bpf_xdp_pull_data) BTF_KFUNCS_END(bpf_kfunc_check_set_xdp) BTF_KFUNCS_START(bpf_kfunc_check_set_sock_addr) From 3d3dd196b1668aab5b65ccd4536dd8d6045af9f8 Mon Sep 17 00:00:00 2001 From: Amery Hung Date: Fri, 19 Sep 2025 16:09:49 -0700 Subject: [PATCH 4/7] bpf: Clear packet pointers after changing packet data in kfuncs bpf_xdp_pull_data() may change packet data and therefore packet pointers need to be invalidated. Add bpf_xdp_pull_data() to the special kfunc list instead of introducing a new KF_ flag until there are more kfuncs changing packet data. Signed-off-by: Amery Hung --- kernel/bpf/verifier.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 1d4183bc3cd1e..0f0867f109b6c 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -12145,6 +12145,7 @@ enum special_kfunc_type { KF_bpf_dynptr_from_skb, KF_bpf_dynptr_from_xdp, KF_bpf_dynptr_from_skb_meta, + KF_bpf_xdp_pull_data, KF_bpf_dynptr_slice, KF_bpf_dynptr_slice_rdwr, KF_bpf_dynptr_clone, @@ -12195,10 +12196,12 @@ BTF_ID(func, bpf_rbtree_right) BTF_ID(func, bpf_dynptr_from_skb) BTF_ID(func, bpf_dynptr_from_xdp) BTF_ID(func, bpf_dynptr_from_skb_meta) +BTF_ID(func, bpf_xdp_pull_data) #else BTF_ID_UNUSED BTF_ID_UNUSED BTF_ID_UNUSED +BTF_ID_UNUSED #endif BTF_ID(func, bpf_dynptr_slice) BTF_ID(func, bpf_dynptr_slice_rdwr) @@ -12268,6 +12271,11 @@ static bool is_kfunc_bpf_preempt_enable(struct bpf_kfunc_call_arg_meta *meta) return meta->func_id == special_kfunc_list[KF_bpf_preempt_enable]; } +static bool is_kfunc_pkt_changing(struct bpf_kfunc_call_arg_meta *meta) +{ + return meta->func_id == special_kfunc_list[KF_bpf_xdp_pull_data]; +} + static enum kfunc_ptr_arg_type get_kfunc_ptr_arg_type(struct bpf_verifier_env *env, struct bpf_kfunc_call_arg_meta *meta, @@ -13997,6 +14005,9 @@ static int check_kfunc_call(struct bpf_verifier_env *env, struct bpf_insn *insn, } } + if (is_kfunc_pkt_changing(&meta)) + clear_all_pkt_pointers(env); + nargs = btf_type_vlen(meta.func_proto); args = (const struct btf_param *)(meta.func_proto + 1); for (i = 0; i < nargs; i++) { @@ -17714,6 +17725,8 @@ static int visit_insn(int t, struct bpf_verifier_env *env) */ if (ret == 0 && is_kfunc_sleepable(&meta)) mark_subprog_might_sleep(env, t); + if (ret == 0 && is_kfunc_pkt_changing(&meta)) + mark_subprog_changes_pkt_data(env, t); } return visit_func_call_insn(t, insns, env, insn->src_reg == BPF_PSEUDO_CALL); From 17a413f5400ac7ac0094aece51dee5cf4770e311 Mon Sep 17 00:00:00 2001 From: Amery Hung Date: Fri, 19 Sep 2025 16:09:50 -0700 Subject: [PATCH 5/7] bpf: Support specifying linear xdp packet data size for BPF_PROG_TEST_RUN To test bpf_xdp_pull_data(), an xdp packet containing fragments as well as free linear data area after xdp->data_end needs to be created. However, bpf_prog_test_run_xdp() always fills the linear area with data_in before creating fragments, leaving no space to pull data. This patch will allow users to specify the linear data size through ctx->data_end. Currently, ctx_in->data_end must match data_size_in and will not be the final ctx->data_end seen by xdp programs. This is because ctx->data_end is populated according to the xdp_buff passed to test_run. The linear data area available in an xdp_buff, max_data_sz, is alawys filled up before copying data_in into fragments. This patch will allow users to specify the size of data that goes into the linear area. When ctx_in->data_end is different from data_size_in, only ctx_in->data_end bytes of data will be put into the linear area when creating the xdp_buff. While ctx_in->data_end will be allowed to be different from data_size_in, it cannot be larger than the data_size_in as there will be no data to copy from user space. If it is larger than the maximum linear data area size, the layout suggested by the user will not be honored. Data beyond max_data_sz bytes will still be copied into fragments. Finally, since it is possible for a NIC to produce a xdp_buff with empty linear data area, allow it when calling bpf_test_init() from bpf_prog_test_run_xdp() so that we can test XDP kfuncs with such xdp_buff. This is done by moving lower-bound check to callers as most of them already do except bpf_prog_test_run_skb(). Signed-off-by: Amery Hung --- net/bpf/test_run.c | 9 +++++++-- .../selftests/bpf/prog_tests/xdp_context_test_run.c | 4 +--- 2 files changed, 8 insertions(+), 5 deletions(-) diff --git a/net/bpf/test_run.c b/net/bpf/test_run.c index 4a862d6053861..0cbd3b898c458 100644 --- a/net/bpf/test_run.c +++ b/net/bpf/test_run.c @@ -665,7 +665,7 @@ static void *bpf_test_init(const union bpf_attr *kattr, u32 user_size, void __user *data_in = u64_to_user_ptr(kattr->test.data_in); void *data; - if (user_size < ETH_HLEN || user_size > PAGE_SIZE - headroom - tailroom) + if (user_size > PAGE_SIZE - headroom - tailroom) return ERR_PTR(-EINVAL); size = SKB_DATA_ALIGN(size); @@ -1001,6 +1001,9 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, kattr->test.cpu || kattr->test.batch_size) return -EINVAL; + if (size < ETH_HLEN) + return -EINVAL; + data = bpf_test_init(kattr, kattr->test.data_size_in, size, NET_SKB_PAD + NET_IP_ALIGN, SKB_DATA_ALIGN(sizeof(struct skb_shared_info))); @@ -1246,13 +1249,15 @@ int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, if (ctx) { /* There can't be user provided data before the meta data */ - if (ctx->data_meta || ctx->data_end != size || + if (ctx->data_meta || ctx->data_end > size || ctx->data > ctx->data_end || unlikely(xdp_metalen_invalid(ctx->data)) || (do_live && (kattr->test.data_out || kattr->test.ctx_out))) goto free_ctx; /* Meta data is allocated from the headroom */ headroom -= ctx->data; + + size = ctx->data_end; } max_data_sz = PAGE_SIZE - headroom - tailroom; diff --git a/tools/testing/selftests/bpf/prog_tests/xdp_context_test_run.c b/tools/testing/selftests/bpf/prog_tests/xdp_context_test_run.c index 46e0730174ed6..178292d1251a1 100644 --- a/tools/testing/selftests/bpf/prog_tests/xdp_context_test_run.c +++ b/tools/testing/selftests/bpf/prog_tests/xdp_context_test_run.c @@ -97,9 +97,7 @@ void test_xdp_context_test_run(void) /* Meta data must be 255 bytes or smaller */ test_xdp_context_error(prog_fd, opts, 0, 256, sizeof(data), 0, 0, 0); - /* Total size of data must match data_end - data_meta */ - test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), - sizeof(data) - 1, 0, 0, 0); + /* Total size of data must be data_end - data_meta or larger */ test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), sizeof(data) + 1, 0, 0, 0); From 3aecf83a317e0bdc331f90c8766b9e0a62a20663 Mon Sep 17 00:00:00 2001 From: Amery Hung Date: Fri, 19 Sep 2025 16:09:51 -0700 Subject: [PATCH 6/7] selftests/bpf: Test bpf_xdp_pull_data Test bpf_xdp_pull_data() with xdp packets with different layouts. The xdp bpf program first checks if the layout is as expected. Then, it calls bpf_xdp_pull_data(). Finally, it checks the 0xbb marker at offset 1024 using directly packet access. Signed-off-by: Amery Hung --- .../selftests/bpf/prog_tests/xdp_pull_data.c | 179 ++++++++++++++++++ .../selftests/bpf/progs/test_xdp_pull_data.c | 48 +++++ 2 files changed, 227 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c create mode 100644 tools/testing/selftests/bpf/progs/test_xdp_pull_data.c diff --git a/tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c b/tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c new file mode 100644 index 0000000000000..efa350d04ec5f --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/xdp_pull_data.c @@ -0,0 +1,179 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include +#include +#include "test_xdp_pull_data.skel.h" + +#define PULL_MAX (1 << 31) +#define PULL_PLUS_ONE (1 << 30) + +#define XDP_PACKET_HEADROOM 256 + +/* Find headroom and tailroom occupied by struct xdp_frame and struct + * skb_shared_info so that we can calculate the maximum pull lengths for + * test cases. They might not be the real size of the structures due to + * cache alignment. + */ +static int find_xdp_sizes(struct test_xdp_pull_data *skel, int frame_sz) +{ + LIBBPF_OPTS(bpf_test_run_opts, topts); + struct xdp_md ctx = {}; + int prog_fd, err; + __u8 *buf; + + buf = calloc(frame_sz, sizeof(__u8)); + if (!ASSERT_OK_PTR(buf, "calloc buf")) + return -ENOMEM; + + topts.data_in = buf; + topts.data_out = buf; + topts.data_size_in = frame_sz; + topts.data_size_out = frame_sz; + /* Pass a data_end larger than the linear space available to make sure + * bpf_prog_test_run_xdp() will fill the linear data area so that + * xdp_find_sizes can infer the size of struct skb_shared_info + */ + ctx.data_end = frame_sz; + topts.ctx_in = &ctx; + topts.ctx_out = &ctx; + topts.ctx_size_in = sizeof(ctx); + topts.ctx_size_out = sizeof(ctx); + + prog_fd = bpf_program__fd(skel->progs.xdp_find_sizes); + err = bpf_prog_test_run_opts(prog_fd, &topts); + ASSERT_OK(err, "bpf_prog_test_run_opts"); + + free(buf); + + return err; +} + +/* xdp_pull_data_prog will directly read a marker 0xbb stored at buf[1024] + * so caller expecting XDP_PASS should always pass pull_len no less than 1024 + */ +static void run_test(struct test_xdp_pull_data *skel, int retval, + int frame_sz, int buff_len, int meta_len, int data_len, + int pull_len) +{ + LIBBPF_OPTS(bpf_test_run_opts, topts); + struct xdp_md ctx = {}; + int prog_fd, err; + __u8 *buf; + + buf = calloc(buff_len, sizeof(__u8)); + if (!ASSERT_OK_PTR(buf, "calloc buf")) + return; + + buf[meta_len + 1023] = 0xaa; + buf[meta_len + 1024] = 0xbb; + buf[meta_len + 1025] = 0xcc; + + topts.data_in = buf; + topts.data_out = buf; + topts.data_size_in = buff_len; + topts.data_size_out = buff_len; + ctx.data = meta_len; + ctx.data_end = meta_len + data_len; + topts.ctx_in = &ctx; + topts.ctx_out = &ctx; + topts.ctx_size_in = sizeof(ctx); + topts.ctx_size_out = sizeof(ctx); + + skel->bss->data_len = data_len; + if (pull_len & PULL_MAX) { + int headroom = XDP_PACKET_HEADROOM - meta_len - skel->bss->xdpf_sz; + int tailroom = frame_sz - XDP_PACKET_HEADROOM - + data_len - skel->bss->sinfo_sz; + + pull_len = pull_len & PULL_PLUS_ONE ? 1 : 0; + pull_len += headroom + tailroom + data_len; + } + skel->bss->pull_len = pull_len; + + prog_fd = bpf_program__fd(skel->progs.xdp_pull_data_prog); + err = bpf_prog_test_run_opts(prog_fd, &topts); + ASSERT_OK(err, "bpf_prog_test_run_opts"); + ASSERT_EQ(topts.retval, retval, "xdp_pull_data_prog retval"); + + if (retval == XDP_DROP) + goto out; + + ASSERT_EQ(ctx.data_end, meta_len + pull_len, "linear data size"); + ASSERT_EQ(topts.data_size_out, buff_len, "linear + non-linear data size"); + /* Make sure data around xdp->data_end was not messed up by + * bpf_xdp_pull_data() + */ + ASSERT_EQ(buf[meta_len + 1023], 0xaa, "data[1023]"); + ASSERT_EQ(buf[meta_len + 1024], 0xbb, "data[1024]"); + ASSERT_EQ(buf[meta_len + 1025], 0xcc, "data[1025]"); +out: + free(buf); +} + +static void test_xdp_pull_data_basic(void) +{ + u32 pg_sz, max_meta_len, max_data_len; + struct test_xdp_pull_data *skel; + + skel = test_xdp_pull_data__open_and_load(); + if (!ASSERT_OK_PTR(skel, "test_xdp_pull_data__open_and_load")) + return; + + pg_sz = sysconf(_SC_PAGE_SIZE); + + if (find_xdp_sizes(skel, pg_sz)) + goto out; + + max_meta_len = XDP_PACKET_HEADROOM - skel->bss->xdpf_sz; + max_data_len = pg_sz - XDP_PACKET_HEADROOM - skel->bss->sinfo_sz; + + /* linear xdp pkt, pull 0 byte */ + run_test(skel, XDP_PASS, pg_sz, 2048, 0, 2048, 2048); + + /* multi-buf pkt, pull results in linear xdp pkt */ + run_test(skel, XDP_PASS, pg_sz, 2048, 0, 1024, 2048); + + /* multi-buf pkt, pull 1 byte to linear data area */ + run_test(skel, XDP_PASS, pg_sz, 9000, 0, 1024, 1025); + + /* multi-buf pkt, pull 0 byte to linear data area */ + run_test(skel, XDP_PASS, pg_sz, 9000, 0, 1025, 1025); + + /* multi-buf pkt, empty linear data area, pull requires memmove */ + run_test(skel, XDP_PASS, pg_sz, 9000, 0, 0, PULL_MAX); + + /* multi-buf pkt, no headroom */ + run_test(skel, XDP_PASS, pg_sz, 9000, max_meta_len, 1024, PULL_MAX); + + /* multi-buf pkt, no tailroom, pull requires memmove */ + run_test(skel, XDP_PASS, pg_sz, 9000, 0, max_data_len, PULL_MAX); + + /* Test cases with invalid pull length */ + + /* linear xdp pkt, pull more than total data len */ + run_test(skel, XDP_DROP, pg_sz, 2048, 0, 2048, 2049); + + /* multi-buf pkt with no space left in linear data area */ + run_test(skel, XDP_DROP, pg_sz, 9000, max_meta_len, max_data_len, + PULL_MAX | PULL_PLUS_ONE); + + /* multi-buf pkt, empty linear data area */ + run_test(skel, XDP_DROP, pg_sz, 9000, 0, 0, PULL_MAX | PULL_PLUS_ONE); + + /* multi-buf pkt, no headroom */ + run_test(skel, XDP_DROP, pg_sz, 9000, max_meta_len, 1024, + PULL_MAX | PULL_PLUS_ONE); + + /* multi-buf pkt, no tailroom */ + run_test(skel, XDP_DROP, pg_sz, 9000, 0, max_data_len, + PULL_MAX | PULL_PLUS_ONE); + +out: + test_xdp_pull_data__destroy(skel); +} + +void test_xdp_pull_data(void) +{ + if (test__start_subtest("xdp_pull_data")) + test_xdp_pull_data_basic(); +} diff --git a/tools/testing/selftests/bpf/progs/test_xdp_pull_data.c b/tools/testing/selftests/bpf/progs/test_xdp_pull_data.c new file mode 100644 index 0000000000000..c41a21413eaa2 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_xdp_pull_data.c @@ -0,0 +1,48 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include "vmlinux.h" +#include + +int xdpf_sz; +int sinfo_sz; +int data_len; +int pull_len; + +#define XDP_PACKET_HEADROOM 256 + +SEC("xdp.frags") +int xdp_find_sizes(struct xdp_md *ctx) +{ + xdpf_sz = sizeof(struct xdp_frame); + sinfo_sz = __PAGE_SIZE - XDP_PACKET_HEADROOM - + (ctx->data_end - ctx->data); + + return XDP_PASS; +} + +SEC("xdp.frags") +int xdp_pull_data_prog(struct xdp_md *ctx) +{ + __u8 *data_end = (void *)(long)ctx->data_end; + __u8 *data = (void *)(long)ctx->data; + __u8 *val_p; + int err; + + if (data_len != data_end - data) + return XDP_DROP; + + err = bpf_xdp_pull_data(ctx, pull_len); + if (err) + return XDP_DROP; + + val_p = (void *)(long)ctx->data + 1024; + if (val_p + 1 > (void *)(long)ctx->data_end) + return XDP_DROP; + + if (*val_p != 0xbb) + return XDP_DROP; + + return XDP_PASS; +} + +char _license[] SEC("license") = "GPL"; From 7ce8e389c6fd0ffd3d3e3c5850690493b8c76971 Mon Sep 17 00:00:00 2001 From: Amery Hung Date: Fri, 19 Sep 2025 16:09:52 -0700 Subject: [PATCH 7/7] selftests: drv-net: Pull data before parsing headers It is possible for drivers to generate xdp packets with data residing entirely in fragments. To keep parsing headers using direcy packet access, call bpf_xdp_pull_data() to pull headers into the linear data area. Signed-off-by: Amery Hung --- .../selftests/net/lib/xdp_native.bpf.c | 89 +++++++++++++++---- 1 file changed, 74 insertions(+), 15 deletions(-) diff --git a/tools/testing/selftests/net/lib/xdp_native.bpf.c b/tools/testing/selftests/net/lib/xdp_native.bpf.c index 521ba38f2ddda..df4eea5c192b3 100644 --- a/tools/testing/selftests/net/lib/xdp_native.bpf.c +++ b/tools/testing/selftests/net/lib/xdp_native.bpf.c @@ -14,6 +14,8 @@ #define MAX_PAYLOAD_LEN 5000 #define MAX_HDR_LEN 64 +extern int bpf_xdp_pull_data(struct xdp_md *xdp, __u32 len) __ksym __weak; + enum { XDP_MODE = 0, XDP_PORT = 1, @@ -68,30 +70,57 @@ static void record_stats(struct xdp_md *ctx, __u32 stat_type) static struct udphdr *filter_udphdr(struct xdp_md *ctx, __u16 port) { - void *data_end = (void *)(long)ctx->data_end; - void *data = (void *)(long)ctx->data; struct udphdr *udph = NULL; - struct ethhdr *eth = data; + void *data, *data_end; + struct ethhdr *eth; + int err; + + err = bpf_xdp_pull_data(ctx, sizeof(*eth)); + if (err) + return NULL; + + data_end = (void *)(long)ctx->data_end; + data = eth = (void *)(long)ctx->data; if (data + sizeof(*eth) > data_end) return NULL; if (eth->h_proto == bpf_htons(ETH_P_IP)) { - struct iphdr *iph = data + sizeof(*eth); + struct iphdr *iph; + + err = bpf_xdp_pull_data(ctx, sizeof(*eth) + sizeof(*iph) + + sizeof(*udph)); + if (err) + return NULL; + + data_end = (void *)(long)ctx->data_end; + data = (void *)(long)ctx->data; + + iph = data + sizeof(*eth); if (iph + 1 > (struct iphdr *)data_end || iph->protocol != IPPROTO_UDP) return NULL; - udph = (void *)eth + sizeof(*iph) + sizeof(*eth); - } else if (eth->h_proto == bpf_htons(ETH_P_IPV6)) { - struct ipv6hdr *ipv6h = data + sizeof(*eth); + udph = data + sizeof(*iph) + sizeof(*eth); + } else if (eth->h_proto == bpf_htons(ETH_P_IPV6)) { + struct ipv6hdr *ipv6h; + + err = bpf_xdp_pull_data(ctx, sizeof(*eth) + sizeof(*ipv6h) + + sizeof(*udph)); + if (err) + return NULL; + + data_end = (void *)(long)ctx->data_end; + data = (void *)(long)ctx->data; + + ipv6h = data + sizeof(*eth); if (ipv6h + 1 > (struct ipv6hdr *)data_end || ipv6h->nexthdr != IPPROTO_UDP) return NULL; - udph = (void *)eth + sizeof(*ipv6h) + sizeof(*eth); + udph = data + sizeof(*ipv6h) + sizeof(*eth); } else { return NULL; } @@ -145,17 +174,34 @@ static void swap_machdr(void *data) static int xdp_mode_tx_handler(struct xdp_md *ctx, __u16 port) { - void *data_end = (void *)(long)ctx->data_end; - void *data = (void *)(long)ctx->data; struct udphdr *udph = NULL; - struct ethhdr *eth = data; + void *data, *data_end; + struct ethhdr *eth; + int err; + + err = bpf_xdp_pull_data(ctx, sizeof(*eth)); + if (err) + return XDP_PASS; + + data_end = (void *)(long)ctx->data_end; + data = eth = (void *)(long)ctx->data; if (data + sizeof(*eth) > data_end) return XDP_PASS; if (eth->h_proto == bpf_htons(ETH_P_IP)) { - struct iphdr *iph = data + sizeof(*eth); - __be32 tmp_ip = iph->saddr; + struct iphdr *iph; + __be32 tmp_ip; + + err = bpf_xdp_pull_data(ctx, sizeof(*eth) + sizeof(*iph) + + sizeof(*udph)); + if (err) + return XDP_PASS; + + data_end = (void *)(long)ctx->data_end; + data = (void *)(long)ctx->data; + + iph = data + sizeof(*eth); if (iph + 1 > (struct iphdr *)data_end || iph->protocol != IPPROTO_UDP) @@ -169,8 +215,10 @@ static int xdp_mode_tx_handler(struct xdp_md *ctx, __u16 port) return XDP_PASS; record_stats(ctx, STATS_RX); + eth = data; swap_machdr((void *)eth); + tmp_ip = iph->saddr; iph->saddr = iph->daddr; iph->daddr = tmp_ip; @@ -178,9 +226,19 @@ static int xdp_mode_tx_handler(struct xdp_md *ctx, __u16 port) return XDP_TX; - } else if (eth->h_proto == bpf_htons(ETH_P_IPV6)) { - struct ipv6hdr *ipv6h = data + sizeof(*eth); + } else if (eth->h_proto == bpf_htons(ETH_P_IPV6)) { struct in6_addr tmp_ipv6; + struct ipv6hdr *ipv6h; + + err = bpf_xdp_pull_data(ctx, sizeof(*eth) + sizeof(*ipv6h) + + sizeof(*udph)); + if (err) + return XDP_PASS; + + data_end = (void *)(long)ctx->data_end; + data = (void *)(long)ctx->data; + + ipv6h = data + sizeof(*eth); if (ipv6h + 1 > (struct ipv6hdr *)data_end || ipv6h->nexthdr != IPPROTO_UDP) @@ -194,6 +252,7 @@ static int xdp_mode_tx_handler(struct xdp_md *ctx, __u16 port) return XDP_PASS; record_stats(ctx, STATS_RX); + eth = data; swap_machdr((void *)eth); __builtin_memcpy(&tmp_ipv6, &ipv6h->saddr, sizeof(tmp_ipv6));