Skip to contentSkip to conversation

bpf: support input xdp_md context in BPF_PROG_TEST_RUN

PATCH Superseded Zvi Effron 3 participants 10 messages Jun 17, 2021 – Jun 21, 2021

Subsystems (6)

v6 of 8 Superseded Latest is v8 →

lore
Share & tools
lore

Zvi Effron

author
#0 lore raw

v6 cover bpf: support input xdp_md context in BPF_PROG_TEST_RUN

This patchset adds support for passing an xdp_md via ctx_in/ctx_out in bpf_attr for BPF_PROG_TEST_RUN of XDP programs.

Patch 1 adds a function to validate XDP meta data lengths.

Patch 2 adds initial support for passing XDP meta data in addition to packet data.

Patch 3 adds support for also specifying the ingress interface and rx queue.

Patch 4 adds selftests to ensure functionality is correct.

Changelog:
----------
v5->v6
v5: https://lore.kernel.org/bpf/[email protected]/

* Correct commit messages in patches 1 and 3
* Add Acked-by to commit message in patch 4
* Use gotos instead of returns to correctly free resources in bpf_prog_test_run_xdp
* Rename xdp_metalen_valid to xdp_metalen_invalid
* Improve the function signature for xdp_metalen_invalid
* Merged declaration of ingress_ifindex and rx_queue_index into one line

v4->v5
v4: https://lore.kernel.org/bpf/[email protected]/

* Add new patch to introduce xdp_metalen_valid inline function to avoid duplicated code from net/core/filter.c
* Correct size of bad_ctx in selftests
* Make all declarations reverse Christmas tree
* Move data check from xdp_convert_md_to_buff to bpf_prog_test_run_xdp
* Merge xdp_convert_buff_to_md into bpf_prog_test_run_xdp
* Fix line too long
* Extracted common checks in selftests to a helper function
* Removed redundant assignment in selftests
* Reordered test cases in selftests
* Check data against 0 instead of data_meta in selftests
* Made selftests use EINVAL instead of hardcoded 22
* Dropped "_" from XDP function name
* Changed casts in XDP program from unsigned long to long
* Added a comment explaining the use of the loopback interface in selftests
* Change parameter order in xdp_convert_md_to_buff to be input first
* Assigned xdp->ingress_ifindex and xdp->rx_queue_index to local variables in xdp_convert_md_to_buff
* Made use of "meta data" versus "metadata" consistent in comments and commit messages

v3->v4
v3: https://lore.kernel.org/bpf/[email protected]/

* Clean up nits
* Validate xdp_md->data_end in bpf_prog_test_run_xdp
* Remove intermediate metalen variables

v2 -> v3
v2: https://lore.kernel.org/bpf/[email protected]/

* Check errno first in selftests
* Use DECLARE_LIBBPF_OPTS
* Rename tattr to opts in selftests
* Remove extra new line
* Rename convert_xdpmd_to_xdpb to xdp_convert_md_to_buff
* Rename convert_xdpb_to_xdpmd to xdp_convert_buff_to_md
* Move declaration of device and rxqueue in xdp_convert_md_to_buff to patch 2
* Reorder the kfree calls in bpf_prog_test_run_xdp

v1 -> v2
v1: https://lore.kernel.org/bpf/[email protected]

* Fix null pointer dereference with no context
* Use the BPF skeleton and replace CHECK with ASSERT macros

Zvi Effron (4):
bpf: add function for XDP meta data length check
bpf: support input xdp_md context in BPF_PROG_TEST_RUN
bpf: support specifying ingress via xdp_md context in
BPF_PROG_TEST_RUN
selftests/bpf: Add test for xdp_md context in BPF_PROG_TEST_RUN

 include/net/xdp.h                             |   5 +
 include/uapi/linux/bpf.h                      |   3 -
 net/bpf/test_run.c                            |  87 +++++++++++++--
 net/core/filter.c                             |   4 +-
 .../bpf/prog_tests/xdp_context_test_run.c     | 105 ++++++++++++++++++
 .../bpf/progs/test_xdp_context_test_run.c     |  20 ++++
 6 files changed, 211 insertions(+), 13 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/xdp_context_test_run.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_xdp_context_test_run.c

base-commit: 8fe088bd4fd12f4c8899b51d5bc3daad98767d49

Signature · 1 line
2.31.1
author
#1 lore raw

PATCH v6 1/4 bpf: add function for XDP meta data length check

This commit prepares to use the XDP meta data length check in multiple places by making it into a static inline function instead of a literal.

---
 include/net/xdp.h | 5 +++++
 net/core/filter.c | 4 ++--
 2 files changed, 7 insertions(+), 2 deletions(-)
include/net/xdp.h +5 -0
index 5533f0ab2afc..ad5b02dcb6f4 100644
@@ -276,6 +276,11 @@ xdp_data_meta_unsupported(const struct xdp_buff *xdp)
 	return unlikely(xdp->data_meta > xdp->data);
 }
 
+static inline bool xdp_metalen_invalid(unsigned long metalen)
+{
+	return (metalen & (sizeof(__u32) - 1)) || (metalen > 32);
+}
+
 struct xdp_attachment_info {
 	struct bpf_prog *prog;
 	u32 flags;
net/core/filter.c +2 -2
index 0b13d8157a8f..118158d6e883 100644
@@ -77,6 +77,7 @@
 #include <net/transp_v6.h>
 #include <linux/btf_ids.h>
 #include <net/tls.h>
+#include <net/xdp.h>
 
 static const struct bpf_func_proto *
 bpf_sk_base_func_proto(enum bpf_func_id func_id);
@@ -3906,8 +3907,7 @@ BPF_CALL_2(bpf_xdp_adjust_meta, struct xdp_buff *, xdp, int, offset)
 	if (unlikely(meta < xdp_frame_end ||
 		     meta > xdp->data))
 		return -EINVAL;
-	if (unlikely((metalen & (sizeof(__u32) - 1)) ||
-		     (metalen > 32)))
+	if (unlikely(xdp_metalen_invalid(metalen)))
 		return -EACCES;
 
 	xdp->data_meta = meta;
Signature · 1 line
2.31.1
Acked-by on v6 1/4
#2 lore raw

Re: [PATCH bpf-next v6 1/4] bpf: add function for XDP meta data length check

Quoted 8 lines · Zvi Effron

This commit prepares to use the XDP meta data length check in multiple places by making it into a static inline function instead of a literal.

author
#3 lore raw

PATCH v6 2/4 bpf: support input xdp_md context in BPF_PROG_TEST_RUN

In reply to Zvi Effron

Support passing a xdp_md via ctx_in/ctx_out in bpf_attr for BPF_PROG_TEST_RUN.

The intended use case is to pass some XDP meta data to the test runs of XDP programs that are used as tail calls.

For programs that use bpf_prog_test_run_xdp, support xdp_md input and output. Unlike with an actual xdp_md during a non-test run, data_meta must be 0 because it must point to the start of the provided user data. From the initial xdp_md, use data and data_end to adjust the pointers in the generated xdp_buff. All other non-zero fields are prohibited (with EINVAL). If the user has set ctx_out/ctx_size_out, copy the (potentially different) xdp_md back to the userspace.

We require all fields of input xdp_md except the ones we explicitly support to be set to zero. The expectation is that in the future we might add support for more fields and we want to fail explicitly if the user runs the program on the kernel where we don't yet support them.

---
 include/uapi/linux/bpf.h |  3 --
 net/bpf/test_run.c       | 67 +++++++++++++++++++++++++++++++++++-----
 2 files changed, 59 insertions(+), 11 deletions(-)
include/uapi/linux/bpf.h +0 -3
index bf9252c7381e..b46a383e8db7 100644
@@ -324,9 +324,6 @@ union bpf_iter_link_info {
  *		**BPF_PROG_TYPE_SK_LOOKUP**
  *			*data_in* and *data_out* must be NULL.
  *
- *		**BPF_PROG_TYPE_XDP**
- *			*ctx_in* and *ctx_out* must be NULL.
- *
  *		**BPF_PROG_TYPE_RAW_TRACEPOINT**,
  *		**BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
  *
net/bpf/test_run.c +59 -8

106 lines not on this page · raw message

Signature · 1 line
2.31.1
Acked-by on v6 2/4
#4 lore raw
Quoted 19 lines · Zvi Effron

Support passing a xdp_md via ctx_in/ctx_out in bpf_attr for BPF_PROG_TEST_RUN.

The intended use case is to pass some XDP meta data to the test runs of XDP programs that are used as tail calls.

For programs that use bpf_prog_test_run_xdp, support xdp_md input and output. Unlike with an actual xdp_md during a non-test run, data_meta must be 0 because it must point to the start of the provided user data. From the initial xdp_md, use data and data_end to adjust the pointers in the generated xdp_buff. All other non-zero fields are prohibited (with EINVAL). If the user has set ctx_out/ctx_size_out, copy the (potentially different) xdp_md back to the userspace.

We require all fields of input xdp_md except the ones we explicitly support to be set to zero. The expectation is that in the future we might add support for more fields and we want to fail explicitly if the user runs the program on the kernel where we don't yet support them.

Quoted 5 lines
author
#5 lore raw

PATCH v6 3/4 bpf: support specifying ingress via xdp_md context in BPF_PROG_TEST_RUN

In reply to Zvi Effron

Support specifying the ingress_ifindex and rx_queue_index of xdp_md contexts for BPF_PROG_TEST_RUN.

The intended use case is to allow testing XDP programs that make decisions based on the ingress interface or RX queue.

If ingress_ifindex is specified, look up the device by the provided index in the current namespace and use its xdp_rxq for the xdp_buff. If the rx_queue_index is out of range, or is non-zero when the ingress_ifindex is 0, return -EINVAL.

---
 net/bpf/test_run.c | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)
net/bpf/test_run.c +21 -1
index 229c5deb813c..1ba15c741517 100644
@@ -690,15 +690,35 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 
 static int xdp_convert_md_to_buff(struct xdp_md *xdp_md, struct xdp_buff *xdp)
 {
+	unsigned int ingress_ifindex, rx_queue_index;
+	struct netdev_rx_queue *rxqueue;
+	struct net_device *device;
+
 	if (!xdp_md)
 		return 0;
 
 	if (xdp_md->egress_ifindex != 0)
 		return -EINVAL;
 
-	if (xdp_md->ingress_ifindex != 0 || xdp_md->rx_queue_index != 0)
+	ingress_ifindex = xdp_md->ingress_ifindex;
+	rx_queue_index = xdp_md->rx_queue_index;
+
+	if (!ingress_ifindex && rx_queue_index)
 		return -EINVAL;
 
+	if (ingress_ifindex) {
+		device = dev_get_by_index(current->nsproxy->net_ns,
+					  ingress_ifindex);
+		if (!device)
+			return -EINVAL;
+
+		if (rx_queue_index >= device->real_num_rx_queues)
+			return -EINVAL;
+
+		rxqueue = __netif_get_rx_queue(device, rx_queue_index);
+		xdp->rxq = &rxqueue->xdp_rxq;
+	}
+
 	xdp->data = xdp->data_meta + xdp_md->data;
 
 	return 0;
Signature · 1 line
2.31.1
Acked-by on v6 3/4
#6 lore raw

Re: [PATCH bpf-next v6 3/4] bpf: support specifying ingress via xdp_md context in BPF_PROG_TEST_RUN

Quoted 11 lines · Zvi Effron

Support specifying the ingress_ifindex and rx_queue_index of xdp_md contexts for BPF_PROG_TEST_RUN.

The intended use case is to allow testing XDP programs that make decisions based on the ingress interface or RX queue.

If ingress_ifindex is specified, look up the device by the provided index in the current namespace and use its xdp_rxq for the xdp_buff. If the rx_queue_index is out of range, or is non-zero when the ingress_ifindex is 0, return -EINVAL.

Quoted 5 lines
maintainer
#7 lore raw

Re: [PATCH bpf-next v6 3/4] bpf: support specifying ingress via xdp_md context in BPF_PROG_TEST_RUN

In reply to Zvi Effron
Quoted 38 lines · Zvi Effron

Support specifying the ingress_ifindex and rx_queue_index of xdp_md contexts for BPF_PROG_TEST_RUN.

The intended use case is to allow testing XDP programs that make decisions based on the ingress interface or RX queue.

If ingress_ifindex is specified, look up the device by the provided index in the current namespace and use its xdp_rxq for the xdp_buff. If the rx_queue_index is out of range, or is non-zero when the ingress_ifindex is 0, return -EINVAL.

---
  net/bpf/test_run.c | 22 +++++++++++++++++++++-
  1 file changed, 21 insertions(+), 1 deletion(-)
net/bpf/test_run.c +4 -0
index 229c5deb813c..1ba15c741517 100644
@@ -690,15 +690,35 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 
  static int xdp_convert_md_to_buff(struct xdp_md *xdp_md, struct xdp_buff *xdp)
  {
+	unsigned int ingress_ifindex, rx_queue_index;
+	struct netdev_rx_queue *rxqueue;
+	struct net_device *device;
+
  	if (!xdp_md)
  		return 0;
 
  	if (xdp_md->egress_ifindex != 0)
  		return -EINVAL;
 

net/bpf/test_run.c int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, · line 711

Quoted 10 lines
net/bpf/test_run.c +7 -1
-	if (xdp_md->ingress_ifindex != 0 || xdp_md->rx_queue_index != 0)
+	ingress_ifindex = xdp_md->ingress_ifindex;
+	rx_queue_index = xdp_md->rx_queue_index;
+
+	if (!ingress_ifindex && rx_queue_index)
  		return -EINVAL;
 
+	if (ingress_ifindex) {
+		device = dev_get_by_index(current->nsproxy->net_ns,
+					  ingress_ifindex);

This takes a reference on the device, which seems to be leaked here?

Quoted 14 lines
net/bpf/test_run.c +10 -0
+		if (!device)
+			return -EINVAL;
+
+		if (rx_queue_index >= device->real_num_rx_queues)
+			return -EINVAL;
+
+		rxqueue = __netif_get_rx_queue(device, rx_queue_index);
+		xdp->rxq = &rxqueue->xdp_rxq;
+	}
+
  	xdp->data = xdp->data_meta + xdp_md->data;
 
  	return 0;
author
#8 lore raw

Re: [PATCH bpf-next v6 3/4] bpf: support specifying ingress via xdp_md context in BPF_PROG_TEST_RUN

Quoted 47 lines · Daniel Borkmann
Quoted 45 lines · Zvi Effron

Support specifying the ingress_ifindex and rx_queue_index of xdp_md contexts for BPF_PROG_TEST_RUN.

The intended use case is to allow testing XDP programs that make decisions based on the ingress interface or RX queue.

If ingress_ifindex is specified, look up the device by the provided index in the current namespace and use its xdp_rxq for the xdp_buff. If the rx_queue_index is out of range, or is non-zero when the ingress_ifindex is 0, return -EINVAL.

---
  net/bpf/test_run.c | 22 +++++++++++++++++++++-
  1 file changed, 21 insertions(+), 1 deletion(-)
net/bpf/test_run.c +8 -1
index 229c5deb813c..1ba15c741517 100644
@@ -690,15 +690,35 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 
  static int xdp_convert_md_to_buff(struct xdp_md *xdp_md, struct xdp_buff *xdp)
  {
+     unsigned int ingress_ifindex, rx_queue_index;
+     struct netdev_rx_queue *rxqueue;
+     struct net_device *device;
+
      if (!xdp_md)
              return 0;
 
      if (xdp_md->egress_ifindex != 0)
              return -EINVAL;
 
-     if (xdp_md->ingress_ifindex != 0 || xdp_md->rx_queue_index != 0)
+     ingress_ifindex = xdp_md->ingress_ifindex;
+     rx_queue_index = xdp_md->rx_queue_index;
+
+     if (!ingress_ifindex && rx_queue_index)
              return -EINVAL;
 
Quoted 5 lines

net/bpf/test_run.c int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, · line 711

Quoted 3 lines
net/bpf/test_run.c +3 -0
+     if (ingress_ifindex) {
+             device = dev_get_by_index(current->nsproxy->net_ns,
+                                       ingress_ifindex);

This takes a reference on the device, which seems to be leaked here?

That would be an accurate assessment. Good catch. We'll fix that!

Quoted 16 lines
Quoted 15 lines
net/bpf/test_run.c +10 -0
+             if (!device)
+                     return -EINVAL;
+
+             if (rx_queue_index >= device->real_num_rx_queues)
+                     return -EINVAL;
+
+             rxqueue = __netif_get_rx_queue(device, rx_queue_index);
+             xdp->rxq = &rxqueue->xdp_rxq;
+     }
+
      xdp->data = xdp->data_meta + xdp_md->data;
 
      return 0;
author
#9 lore raw

PATCH v6 4/4 selftests/bpf: Add test for xdp_md context in BPF_PROG_TEST_RUN

In reply to Zvi Effron

Add a test for using xdp_md as a context to BPF_PROG_TEST_RUN for XDP programs.

The test uses a BPF program that takes in a return value from XDP meta data, then reduces the size of the XDP meta data by 4 bytes.

Test cases validate the possible failure cases for passing in invalid xdp_md contexts, that the return value is successfully passed in, and that the adjusted meta data is successfully copied out.

---
 .../bpf/prog_tests/xdp_context_test_run.c     | 105 ++++++++++++++++++
 .../bpf/progs/test_xdp_context_test_run.c     |  20 ++++
 2 files changed, 125 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/xdp_context_test_run.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_xdp_context_test_run.c
tools/testing/selftests/bpf/prog_tests/xdp_context_test_run.c new file +105 -0
new file mode 100644
index 000000000000..4fdb991482cb
@@ -0,0 +1,105 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <test_progs.h>
+#include <network_helpers.h>
+#include "test_xdp_context_test_run.skel.h"
+
+void test_xdp_context_error(int prog_fd, struct bpf_test_run_opts opts,
+			    __u32 data_meta, __u32 data, __u32 data_end,
+			    __u32 ingress_ifindex, __u32 rx_queue_index,
+			    __u32 egress_ifindex)
+{
+	struct xdp_md ctx = {
+		.data = data,
+		.data_end = data_end,
+		.data_meta = data_meta,
+		.ingress_ifindex = ingress_ifindex,
+		.rx_queue_index = rx_queue_index,
+		.egress_ifindex = egress_ifindex,
+	};
+	int err;
+
+	opts.ctx_in = &ctx;
+	opts.ctx_size_in = sizeof(ctx);
+	err = bpf_prog_test_run_opts(prog_fd, &opts);
+	ASSERT_EQ(errno, EINVAL, "errno-EINVAL");
+	ASSERT_ERR(err, "bpf_prog_test_run");
+}
+
+void test_xdp_context_test_run(void)
+{
+	struct test_xdp_context_test_run *skel = NULL;
+	char data[sizeof(pkt_v4) + sizeof(__u32)];
+	char bad_ctx[sizeof(struct xdp_md) + 1];
+	struct xdp_md ctx_in, ctx_out;
+	DECLARE_LIBBPF_OPTS(bpf_test_run_opts, opts,
+			    .data_in = &data,
+			    .data_size_in = sizeof(data),
+			    .ctx_out = &ctx_out,
+			    .ctx_size_out = sizeof(ctx_out),
+			    .repeat = 1,
+		);
+	int err, prog_fd;
+
+	skel = test_xdp_context_test_run__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "skel"))
+		return;
+	prog_fd = bpf_program__fd(skel->progs.xdp_context);
+
+	/* Data past the end of the kernel's struct xdp_md must be 0 */
+	bad_ctx[sizeof(bad_ctx) - 1] = 1;
+	opts.ctx_in = bad_ctx;
+	opts.ctx_size_in = sizeof(bad_ctx);
+	err = bpf_prog_test_run_opts(prog_fd, &opts);
+	ASSERT_EQ(errno, E2BIG, "extradata-errno");
+	ASSERT_ERR(err, "bpf_prog_test_run(extradata)");
+
+	*(__u32 *)data = XDP_PASS;
+	*(struct ipv4_packet *)(data + sizeof(__u32)) = pkt_v4;
+	opts.ctx_in = &ctx_in;
+	opts.ctx_size_in = sizeof(ctx_in);
+	memset(&ctx_in, 0, sizeof(ctx_in));
+	ctx_in.data_meta = 0;
+	ctx_in.data = sizeof(__u32);
+	ctx_in.data_end = ctx_in.data + sizeof(pkt_v4);
+	err = bpf_prog_test_run_opts(prog_fd, &opts);
+	ASSERT_OK(err, "bpf_prog_test_run(valid)");
+	ASSERT_EQ(opts.retval, XDP_PASS, "valid-retval");
+	ASSERT_EQ(opts.data_size_out, sizeof(pkt_v4), "valid-datasize");
+	ASSERT_EQ(opts.ctx_size_out, opts.ctx_size_in, "valid-ctxsize");
+	ASSERT_EQ(ctx_out.data_meta, 0, "valid-datameta");
+	ASSERT_EQ(ctx_out.data, 0, "valid-data");
+	ASSERT_EQ(ctx_out.data_end, sizeof(pkt_v4), "valid-dataend");
+
+	/* Meta data's size must be a multiple of 4 */
+	test_xdp_context_error(prog_fd, opts, 0, 1, sizeof(data), 0, 0, 0);
+
+	/* data_meta must reference the start of data */
+	test_xdp_context_error(prog_fd, opts, 4, sizeof(__u32), sizeof(data),
+			       0, 0, 0);
+
+	/* Meta data must be 32 bytes or smaller */
+	test_xdp_context_error(prog_fd, opts, 0, 36, 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);
+	test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32),
+			       sizeof(data) + 1, 0, 0, 0);
+
+	/* RX queue cannot be specified without specifying an ingress */
+	test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), sizeof(data),
+			       0, 1, 0);
+
+	/* Interface 1 is always the loopback interface which always has only
+	 * one RX queue (index 0). This makes index 1 an invalid index for
+	 * interface 1.
+	 */
+	test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), sizeof(data),
+			       1, 1, 0);
+
+	/* The egress cannot be specified */
+	test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), sizeof(data),
+			       0, 0, 1);
+
+	test_xdp_context_test_run__destroy(skel);
+}
tools/testing/selftests/bpf/progs/test_xdp_context_test_run.c new file +20 -0
new file mode 100644
index 000000000000..d7b88cd05afd
@@ -0,0 +1,20 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+
+SEC("xdp")
+int xdp_context(struct xdp_md *xdp)
+{
+	void *data = (void *)(long)xdp->data;
+	__u32 *metadata = (void *)(long)xdp->data_meta;
+	__u32 ret;
+
+	if (metadata + 1 > data)
+		return XDP_ABORTED;
+	ret = *metadata;
+	if (bpf_xdp_adjust_meta(xdp, 4))
+		return XDP_ABORTED;
+	return ret;
+}
+
+char _license[] SEC("license") = "GPL";
Signature · 1 line
2.31.1