Skip to contentSkip to conversation

bpf: support input __sk_buff context in BPF_PROG_TEST_RUN

PATCH Superseded Stanislav Fomichev 3 participants 6 messages Apr 8, 2019 – Apr 9, 2019

Subsystems (7)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/3 bpf: support input __sk_buff context in BPF_PROG_TEST_RUN

Add new set of arguments to bpf_attr for BPF_PROG_TEST_RUN:
* ctx_in/ctx_size_in - input context
* ctx_out/ctx_size_out - output context

The intended use case is to pass some meta data to the test runs that operate on skb (this has being brought up on recent LPC).

For programs that use bpf_prog_test_run_skb, support __sk_buff input and output. Initially, from input __sk_buff, copy _only_ cb and priority into skb, all other non-zero fields are prohibited (with EINVAL). If the user has set ctx_out/ctx_size_out, copy the potentially modified __sk_buff back to the userspace.

We require all fields of input __sk_buff 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.

The API is intentionally vague (i.e. we don't explicitly add __sk_buff to bpf_attr, but ctx_in) to potentially let other test_run types use this interface in the future (this can be xdp_md for xdp types for example).

v2:
* Addressed comments from Martin Lau

---
 include/uapi/linux/bpf.h |   7 ++
 kernel/bpf/syscall.c     |  10 ++-
 net/bpf/test_run.c       | 135 ++++++++++++++++++++++++++++++++++++---
 3 files changed, 143 insertions(+), 9 deletions(-)
include/uapi/linux/bpf.h +7 -0
index 837024512baf..8e96f99cebf8 100644
@@ -396,6 +396,13 @@ union bpf_attr {
 		__aligned_u64	data_out;
 		__u32		repeat;
 		__u32		duration;
+		__u32		ctx_size_in;	/* input: len of ctx_in */
+		__u32		ctx_size_out;	/* input/output: len of ctx_out
+						 *   returns ENOSPC if ctx_out
+						 *   is too small.
+						 */
+		__aligned_u64	ctx_in;
+		__aligned_u64	ctx_out;
 	} test;
 
 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
kernel/bpf/syscall.c +9 -1
index 1d65e56594db..5bb963e8f9b0 100644
@@ -1949,7 +1949,7 @@ static int bpf_prog_query(const union bpf_attr *attr,
 	return cgroup_bpf_prog_query(attr, uattr);
 }
 
-#define BPF_PROG_TEST_RUN_LAST_FIELD test.duration
+#define BPF_PROG_TEST_RUN_LAST_FIELD test.ctx_out
 
 static int bpf_prog_test_run(const union bpf_attr *attr,
 			     union bpf_attr __user *uattr)
@@ -1962,6 +1962,14 @@ static int bpf_prog_test_run(const union bpf_attr *attr,
 	if (CHECK_ATTR(BPF_PROG_TEST_RUN))
 		return -EINVAL;
 
+	if ((attr->test.ctx_size_in && !attr->test.ctx_in) ||
+	    (!attr->test.ctx_size_in && attr->test.ctx_in))
+		return -EINVAL;
+
+	if ((attr->test.ctx_size_out && !attr->test.ctx_out) ||
+	    (!attr->test.ctx_size_out && attr->test.ctx_out))
+		return -EINVAL;
+
 	prog = bpf_prog_get(attr->test.prog_fd);
 	if (IS_ERR(prog))
 		return PTR_ERR(prog);
net/bpf/test_run.c +127 -8

189 lines not on this page · raw message

Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#1 lore raw

PATCH v2 2/3 libbpf: add support for ctx_{size,}_{in,out} in BPF_PROG_TEST_RUN

Support recently introduced input/output context for test runs. We extend only bpf_prog_test_run_xattr. bpf_prog_test_run is unextendable and left as is.

---
 tools/include/uapi/linux/bpf.h | 7 +++++++
 tools/lib/bpf/bpf.c            | 5 +++++
 tools/lib/bpf/bpf.h            | 5 +++++
 3 files changed, 17 insertions(+)
tools/include/uapi/linux/bpf.h +7 -0
index 837024512baf..8e96f99cebf8 100644
@@ -396,6 +396,13 @@ union bpf_attr {
 		__aligned_u64	data_out;
 		__u32		repeat;
 		__u32		duration;
+		__u32		ctx_size_in;	/* input: len of ctx_in */
+		__u32		ctx_size_out;	/* input/output: len of ctx_out
+						 *   returns ENOSPC if ctx_out
+						 *   is too small.
+						 */
+		__aligned_u64	ctx_in;
+		__aligned_u64	ctx_out;
 	} test;
 
 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
tools/lib/bpf/bpf.c +5 -0
index a1db869a6fda..d740da7ae1d3 100644
@@ -545,10 +545,15 @@ int bpf_prog_test_run_xattr(struct bpf_prog_test_run_attr *test_attr)
 	attr.test.data_out = ptr_to_u64(test_attr->data_out);
 	attr.test.data_size_in = test_attr->data_size_in;
 	attr.test.data_size_out = test_attr->data_size_out;
+	attr.test.ctx_in = ptr_to_u64(test_attr->ctx_in);
+	attr.test.ctx_out = ptr_to_u64(test_attr->ctx_out);
+	attr.test.ctx_size_in = test_attr->ctx_size_in;
+	attr.test.ctx_size_out = test_attr->ctx_size_out;
 	attr.test.repeat = test_attr->repeat;
 
 	ret = sys_bpf(BPF_PROG_TEST_RUN, &attr, sizeof(attr));
 	test_attr->data_size_out = attr.test.data_size_out;
+	test_attr->ctx_size_out = attr.test.ctx_size_out;
 	test_attr->retval = attr.test.retval;
 	test_attr->duration = attr.test.duration;
 	return ret;
tools/lib/bpf/bpf.h +5 -0
index e2c0df7b831f..385418a6d54b 100644
@@ -135,6 +135,11 @@ struct bpf_prog_test_run_attr {
 			      * out: length of data_out */
 	__u32 retval;        /* out: return code of the BPF program */
 	__u32 duration;      /* out: average per repetition in ns */
+	const void *ctx_in; /* optional */
+	__u32 ctx_size_in;
+	void *ctx_out;      /* optional */
+	__u32 ctx_size_out; /* in: max length of ctx_out
+			     * out: length of cxt_out */
 };
 
 LIBBPF_API int bpf_prog_test_run_xattr(struct bpf_prog_test_run_attr *test_attr);
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#2 lore raw

PATCH v2 3/3 selftests: bpf: add selftest for __sk_buff context in BPF_PROG_TEST_RUN

In reply to Stanislav Fomichev

Simple test that sets cb to {1,2,3,4,5} and priority to 6, runs bpf program that fails if cb is not what we expect and increments cb[i] and priority. When the test finishes, we check that cb is now {2,3,4,5,6} and priority is 7.

We also test the sanity checks:
* ctx_in is provided, but ctx_size_in is zero (same for ctx_out/ctx_size_out)
* unexpected non-zero fields in __sk_buff return EINVAL

---
 .../selftests/bpf/prog_tests/skb_ctx.c        | 89 +++++++++++++++++++
 .../selftests/bpf/progs/test_skb_ctx.c        | 21 +++++
 2 files changed, 110 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/skb_ctx.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_skb_ctx.c
tools/testing/selftests/bpf/prog_tests/skb_ctx.c new file +89 -0
new file mode 100644
index 000000000000..e95baa32e277
@@ -0,0 +1,89 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <test_progs.h>
+
+void test_skb_ctx(void)
+{
+	struct __sk_buff skb = {
+		.cb[0] = 1,
+		.cb[1] = 2,
+		.cb[2] = 3,
+		.cb[3] = 4,
+		.cb[4] = 5,
+		.priority = 6,
+	};
+	struct bpf_prog_test_run_attr tattr = {
+		.data_in = &pkt_v4,
+		.data_size_in = sizeof(pkt_v4),
+		.ctx_in = &skb,
+		.ctx_size_in = sizeof(skb),
+		.ctx_out = &skb,
+		.ctx_size_out = sizeof(skb),
+	};
+	struct bpf_object *obj;
+	int err;
+	int i;
+
+	err = bpf_prog_load("./test_skb_ctx.o", BPF_PROG_TYPE_SCHED_CLS, &obj,
+			    &tattr.prog_fd);
+	if (CHECK_ATTR(err, "load", "err %d errno %d\n", err, errno))
+		return;
+
+	/* ctx_in != NULL, ctx_size_in == 0 */
+
+	tattr.ctx_size_in = 0;
+	err = bpf_prog_test_run_xattr(&tattr);
+	CHECK_ATTR(err == 0, "ctx_size_in", "err %d errno %d\n", err, errno);
+	tattr.ctx_size_in = sizeof(skb);
+
+	/* ctx_out != NULL, ctx_size_out == 0 */
+
+	tattr.ctx_size_out = 0;
+	err = bpf_prog_test_run_xattr(&tattr);
+	CHECK_ATTR(err == 0, "ctx_size_out", "err %d errno %d\n", err, errno);
+	tattr.ctx_size_out = sizeof(skb);
+
+	/* non-zero [len, tc_index] fields should be rejected*/
+
+	skb.len = 1;
+	err = bpf_prog_test_run_xattr(&tattr);
+	CHECK_ATTR(err == 0, "len", "err %d errno %d\n", err, errno);
+	skb.len = 0;
+
+	skb.tc_index = 1;
+	err = bpf_prog_test_run_xattr(&tattr);
+	CHECK_ATTR(err == 0, "tc_index", "err %d errno %d\n", err, errno);
+	skb.tc_index = 0;
+
+	/* non-zero [hash, sk] fields should be rejected */
+
+	skb.hash = 1;
+	err = bpf_prog_test_run_xattr(&tattr);
+	CHECK_ATTR(err == 0, "hash", "err %d errno %d\n", err, errno);
+	skb.hash = 0;
+
+	skb.sk = (struct bpf_sock *)1;
+	err = bpf_prog_test_run_xattr(&tattr);
+	CHECK_ATTR(err == 0, "sk", "err %d errno %d\n", err, errno);
+	skb.sk = 0;
+
+	err = bpf_prog_test_run_xattr(&tattr);
+	CHECK_ATTR(err != 0 || tattr.retval,
+		   "run",
+		   "err %d errno %d retval %d\n",
+		   err, errno, tattr.retval);
+
+	CHECK_ATTR(tattr.ctx_size_out != sizeof(skb),
+		   "ctx_size_out",
+		   "incorrect output size, want %lu have %u\n",
+		   sizeof(skb), tattr.ctx_size_out);
+
+	for (i = 0; i < 5; i++)
+		CHECK_ATTR(skb.cb[i] != i + 2,
+			   "ctx_out_cb",
+			   "skb->cb[i] == %d, expected %d\n",
+			   skb.cb[i], i + 2);
+	CHECK_ATTR(skb.priority != 7,
+		   "ctx_out_priority",
+		   "skb->priority == %d, expected %d\n",
+		   skb.priority, 7);
+}
tools/testing/selftests/bpf/progs/test_skb_ctx.c new file +21 -0
new file mode 100644
index 000000000000..7a80960d7df1
@@ -0,0 +1,21 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/bpf.h>
+#include "bpf_helpers.h"
+
+int _version SEC("version") = 1;
+char _license[] SEC("license") = "GPL";
+
+SEC("skb_ctx")
+int process(struct __sk_buff *skb)
+{
+	#pragma clang loop unroll(full)
+	for (int i = 0; i < 5; i++) {
+		if (skb->cb[i] != i + 1)
+			return 1;
+		skb->cb[i]++;
+	}
+	skb->priority++;
+
+	return 0;
+}
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#3 lore raw
In reply to Stanislav Fomichev
Quoted 86 lines · Stanislav Fomichev

Add new set of arguments to bpf_attr for BPF_PROG_TEST_RUN:
* ctx_in/ctx_size_in - input context
* ctx_out/ctx_size_out - output context

The intended use case is to pass some meta data to the test runs that operate on skb (this has being brought up on recent LPC).

For programs that use bpf_prog_test_run_skb, support __sk_buff input and output. Initially, from input __sk_buff, copy _only_ cb and priority into skb, all other non-zero fields are prohibited (with EINVAL). If the user has set ctx_out/ctx_size_out, copy the potentially modified __sk_buff back to the userspace.

We require all fields of input __sk_buff 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.

The API is intentionally vague (i.e. we don't explicitly add __sk_buff to bpf_attr, but ctx_in) to potentially let other test_run types use this interface in the future (this can be xdp_md for xdp types for example).

v2:
* Addressed comments from Martin Lau

---
 include/uapi/linux/bpf.h |   7 ++
 kernel/bpf/syscall.c     |  10 ++-
 net/bpf/test_run.c       | 135 ++++++++++++++++++++++++++++++++++++---
 3 files changed, 143 insertions(+), 9 deletions(-)
include/uapi/linux/bpf.h +7 -0
index 837024512baf..8e96f99cebf8 100644
@@ -396,6 +396,13 @@ union bpf_attr {
 		__aligned_u64	data_out;
 		__u32		repeat;
 		__u32		duration;
+		__u32		ctx_size_in;	/* input: len of ctx_in */
+		__u32		ctx_size_out;	/* input/output: len of ctx_out
+						 *   returns ENOSPC if ctx_out
+						 *   is too small.
+						 */
+		__aligned_u64	ctx_in;
+		__aligned_u64	ctx_out;
 	} test;
 
 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
kernel/bpf/syscall.c +9 -1
index 1d65e56594db..5bb963e8f9b0 100644
@@ -1949,7 +1949,7 @@ static int bpf_prog_query(const union bpf_attr *attr,
 	return cgroup_bpf_prog_query(attr, uattr);
 }
 
-#define BPF_PROG_TEST_RUN_LAST_FIELD test.duration
+#define BPF_PROG_TEST_RUN_LAST_FIELD test.ctx_out
 
 static int bpf_prog_test_run(const union bpf_attr *attr,
 			     union bpf_attr __user *uattr)
@@ -1962,6 +1962,14 @@ static int bpf_prog_test_run(const union bpf_attr *attr,
 	if (CHECK_ATTR(BPF_PROG_TEST_RUN))
 		return -EINVAL;
 
+	if ((attr->test.ctx_size_in && !attr->test.ctx_in) ||
+	    (!attr->test.ctx_size_in && attr->test.ctx_in))
+		return -EINVAL;
+
+	if ((attr->test.ctx_size_out && !attr->test.ctx_out) ||
+	    (!attr->test.ctx_size_out && attr->test.ctx_out))
+		return -EINVAL;
+
 	prog = bpf_prog_get(attr->test.prog_fd);
 	if (IS_ERR(prog))
 		return PTR_ERR(prog);
net/bpf/test_run.c +0 -0
index fab142b796ef..3c6e3855c886 100644
@@ -123,12 +123,118 @@ static void *bpf_test_init(const union bpf_attr *kattr, u32 size,
 	return data;

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 133

Quoted 10 lines
net/bpf/test_run.c +8 -0
 }
 
+static void *bpf_ctx_init(const union bpf_attr *kattr, u32 max_size)
+{
+	void __user *data_in = u64_to_user_ptr(kattr->test.ctx_in);
+	u32 size = kattr->test.ctx_size_in;
+	void *data;
+
+	if (!size)
+		return NULL;

A note for a later comment in bpf_prog_test_run_skb(). This function may return NULL.

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 135

Quoted 2 lines
net/bpf/test_run.c +2 -0
+
+	if (size > max_size)

If should be fine if the kernel unsupported bytes are zeros (i.e. tailing zeros in data_in).
Please refer to the existing bpf_check_uarg_tail_zero() usages or also CHECK_ATTR() in syscall.c. Something like this if bpf_check_uarg_tail_zero() is used,

	err = bpf_check_uarg_tail_zero(data_in, max_size, size));
	if (err)
		return ERR_PTR(err);
Quoted 11 lines
net/bpf/test_run.c +11 -0
+		return ERR_PTR(-E2BIG);
+
+	data = kzalloc(max_size, GFP_USER);
+	if (!data)
+		return ERR_PTR(-ENOMEM);
+
+	if (copy_from_user(data, data_in, size)) {
+		kfree(data);
+		return ERR_PTR(-EFAULT);
+	}
+	return data;

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 156

Quoted 10 lines
net/bpf/test_run.c +10 -0
+}
+
+static int bpf_ctx_finish(const union bpf_attr *kattr,
+			  union bpf_attr __user *uattr, const void *data,
+			  u32 size)
+{
+	void __user *data_out = u64_to_user_ptr(kattr->test.ctx_out);
+	int err = -EFAULT;
+	u32 copy_size = size;
+

data is not checked for NULL here.

Quoted 17 lines
net/bpf/test_run.c +17 -0
+	if (!kattr->test.ctx_size_out)
+		return 0;
+
+	if (copy_size > kattr->test.ctx_size_out) {
+		copy_size = kattr->test.ctx_size_out;
+		err = -ENOSPC;
+	}
+
+	if (copy_to_user(data_out, data, copy_size))
+		goto out;
+	if (copy_to_user(&uattr->test.ctx_size_out, &size, sizeof(size)))
+		goto out;
+	if (err != -ENOSPC)
+		err = 0;
+out:
+	return err;
+}

[ ... ]

Quoted 67 lines

int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,

net/bpf/test_run.c +16 -8
 			  union bpf_attr __user *uattr)
 {
 	bool is_l2 = false, is_direct_pkt_access = false;
 	u32 size = kattr->test.data_size_in;
 	u32 repeat = kattr->test.repeat;
+	struct __sk_buff *ctx = NULL;
 	u32 retval, duration;
 	int hh_len = ETH_HLEN;
 	struct sk_buff *skb;
@@ -141,6 +247,12 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	if (IS_ERR(data))
 		return PTR_ERR(data);
 
+	ctx = bpf_ctx_init(kattr, sizeof(struct __sk_buff));
+	if (IS_ERR(ctx)) {
+		kfree(data);
+		return PTR_ERR(ctx);
+	}
+
 	switch (prog->type) {
 	case BPF_PROG_TYPE_SCHED_CLS:
 	case BPF_PROG_TYPE_SCHED_ACT:
@@ -158,6 +270,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	sk = kzalloc(sizeof(struct sock), GFP_USER);
 	if (!sk) {
 		kfree(data);
+		kfree(ctx);
 		return -ENOMEM;
 	}
 	sock_net_set(sk, current->nsproxy->net_ns);
@@ -166,6 +279,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	skb = build_skb(data, 0);
 	if (!skb) {
 		kfree(data);
+		kfree(ctx);
 		kfree(sk);
 		return -ENOMEM;
 	}
@@ -180,32 +294,37 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 		__skb_push(skb, hh_len);
 	if (is_direct_pkt_access)
 		bpf_compute_data_pointers(skb);
+	ret = convert___skb_to_skb(skb, ctx);
+	if (ret)
+		goto out;
 	ret = bpf_test_run(prog, skb, repeat, &retval, &duration);
-	if (ret) {
-		kfree_skb(skb);
-		kfree(sk);
-		return ret;
-	}
+	if (ret)
+		goto out;
 	if (!is_l2) {
 		if (skb_headroom(skb) < hh_len) {
 			int nhead = HH_DATA_ALIGN(hh_len - skb_headroom(skb));
 
 			if (pskb_expand_head(skb, nhead, 0, GFP_USER)) {
-				kfree_skb(skb);
-				kfree(sk);
-				return -ENOMEM;
+				ret = -ENOMEM;
+				goto out;
 			}
 		}
 		memset(__skb_push(skb, hh_len), 0, hh_len);

net/bpf/test_run.c int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, · line 322

Quoted 10 lines
net/bpf/test_run.c +3 -0
 	}
+	convert_skb_to___skb(skb, ctx);
 
 	size = skb->len;
 	/* bpf program can never convert linear skb to non-linear */
 	if (WARN_ON_ONCE(skb_is_nonlinear(skb)))
 		size = skb_headlen(skb);
 	ret = bpf_test_finish(kattr, uattr, skb->data, size, retval, duration);
+	if (!ret)
+		ret = bpf_ctx_finish(kattr, uattr, ctx,

ctx is not checked for NULL in here either.

Quoted 11 lines
net/bpf/test_run.c +3 -0
+				     sizeof(struct __sk_buff));
+out:
 	kfree_skb(skb);
 	kfree(sk);
+	kfree(ctx);
 	return ret;
 }
 
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#4 lore raw
Quoted 108 lines · Martin Lau
Quoted 86 lines · Stanislav Fomichev

Add new set of arguments to bpf_attr for BPF_PROG_TEST_RUN:
* ctx_in/ctx_size_in - input context
* ctx_out/ctx_size_out - output context

The intended use case is to pass some meta data to the test runs that operate on skb (this has being brought up on recent LPC).

For programs that use bpf_prog_test_run_skb, support __sk_buff input and output. Initially, from input __sk_buff, copy _only_ cb and priority into skb, all other non-zero fields are prohibited (with EINVAL). If the user has set ctx_out/ctx_size_out, copy the potentially modified __sk_buff back to the userspace.

We require all fields of input __sk_buff 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.

The API is intentionally vague (i.e. we don't explicitly add __sk_buff to bpf_attr, but ctx_in) to potentially let other test_run types use this interface in the future (this can be xdp_md for xdp types for example).

v2:
* Addressed comments from Martin Lau

---
 include/uapi/linux/bpf.h |   7 ++
 kernel/bpf/syscall.c     |  10 ++-
 net/bpf/test_run.c       | 135 ++++++++++++++++++++++++++++++++++++---
 3 files changed, 143 insertions(+), 9 deletions(-)
include/uapi/linux/bpf.h +7 -0
index 837024512baf..8e96f99cebf8 100644
@@ -396,6 +396,13 @@ union bpf_attr {
 		__aligned_u64	data_out;
 		__u32		repeat;
 		__u32		duration;
+		__u32		ctx_size_in;	/* input: len of ctx_in */
+		__u32		ctx_size_out;	/* input/output: len of ctx_out
+						 *   returns ENOSPC if ctx_out
+						 *   is too small.
+						 */
+		__aligned_u64	ctx_in;
+		__aligned_u64	ctx_out;
 	} test;
 
 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
kernel/bpf/syscall.c +9 -1
index 1d65e56594db..5bb963e8f9b0 100644
@@ -1949,7 +1949,7 @@ static int bpf_prog_query(const union bpf_attr *attr,
 	return cgroup_bpf_prog_query(attr, uattr);
 }
 
-#define BPF_PROG_TEST_RUN_LAST_FIELD test.duration
+#define BPF_PROG_TEST_RUN_LAST_FIELD test.ctx_out
 
 static int bpf_prog_test_run(const union bpf_attr *attr,
 			     union bpf_attr __user *uattr)
@@ -1962,6 +1962,14 @@ static int bpf_prog_test_run(const union bpf_attr *attr,
 	if (CHECK_ATTR(BPF_PROG_TEST_RUN))
 		return -EINVAL;
 
+	if ((attr->test.ctx_size_in && !attr->test.ctx_in) ||
+	    (!attr->test.ctx_size_in && attr->test.ctx_in))
+		return -EINVAL;
+
+	if ((attr->test.ctx_size_out && !attr->test.ctx_out) ||
+	    (!attr->test.ctx_size_out && attr->test.ctx_out))
+		return -EINVAL;
+
 	prog = bpf_prog_get(attr->test.prog_fd);
 	if (IS_ERR(prog))
 		return PTR_ERR(prog);
net/bpf/test_run.c +0 -0
index fab142b796ef..3c6e3855c886 100644
@@ -123,12 +123,118 @@ static void *bpf_test_init(const union bpf_attr *kattr, u32 size,
 	return data;

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 133

Quoted 10 lines
net/bpf/test_run.c +8 -0
 }
 
+static void *bpf_ctx_init(const union bpf_attr *kattr, u32 max_size)
+{
+	void __user *data_in = u64_to_user_ptr(kattr->test.ctx_in);
+	u32 size = kattr->test.ctx_size_in;
+	void *data;
+
+	if (!size)
+		return NULL;

A note for a later comment in bpf_prog_test_run_skb(). This function may return NULL.

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 135

Quoted 2 lines
net/bpf/test_run.c +2 -0
+
+	if (size > max_size)

If should be fine if the kernel unsupported bytes are zeros (i.e. tailing zeros in data_in).
Please refer to the existing bpf_check_uarg_tail_zero() usages or also CHECK_ATTR() in syscall.c. Something like this if bpf_check_uarg_tail_zero() is used,

Quoted 4 lines
	err = bpf_check_uarg_tail_zero(data_in, max_size, size));
	if (err)
		return ERR_PTR(err);

Nice! Will use that instead.

Quoted 14 lines
Quoted 14 lines
net/bpf/test_run.c +14 -0
+		return ERR_PTR(-E2BIG);
+
+	data = kzalloc(max_size, GFP_USER);
+	if (!data)
+		return ERR_PTR(-ENOMEM);
+
+	if (copy_from_user(data, data_in, size)) {
+		kfree(data);
+		return ERR_PTR(-EFAULT);
+	}
+	return data;
+}
+
+static int bpf_ctx_finish(const union bpf_attr *kattr,
Quoted 8 lines

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 156

Quoted 7 lines
net/bpf/test_run.c +7 -0
+			  union bpf_attr __user *uattr, const void *data,
+			  u32 size)
+{
+	void __user *data_out = u64_to_user_ptr(kattr->test.ctx_out);
+	int err = -EFAULT;
+	u32 copy_size = size;
+

data is not checked for NULL here.

I do check kattr->test.ctx_size_out on the next line. And I've added checks to the syscall.c that make sure that ctx_size_out and ctx_out are consistent (i.e. if ctx_size_out is 0, ctx_out is NULL).

Let me instead check kattr->test.ctx_out, that should be more clear.

Quoted 90 lines
Quoted 17 lines
net/bpf/test_run.c +17 -0
+	if (!kattr->test.ctx_size_out)
+		return 0;
+
+	if (copy_size > kattr->test.ctx_size_out) {
+		copy_size = kattr->test.ctx_size_out;
+		err = -ENOSPC;
+	}
+
+	if (copy_to_user(data_out, data, copy_size))
+		goto out;
+	if (copy_to_user(&uattr->test.ctx_size_out, &size, sizeof(size)))
+		goto out;
+	if (err != -ENOSPC)
+		err = 0;
+out:
+	return err;
+}

[ ... ]

Quoted 70 lines

int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,

net/bpf/test_run.c +17 -8
 			  union bpf_attr __user *uattr)
 {
 	bool is_l2 = false, is_direct_pkt_access = false;
 	u32 size = kattr->test.data_size_in;
 	u32 repeat = kattr->test.repeat;
+	struct __sk_buff *ctx = NULL;
 	u32 retval, duration;
 	int hh_len = ETH_HLEN;
 	struct sk_buff *skb;
@@ -141,6 +247,12 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	if (IS_ERR(data))
 		return PTR_ERR(data);
 
+	ctx = bpf_ctx_init(kattr, sizeof(struct __sk_buff));
+	if (IS_ERR(ctx)) {
+		kfree(data);
+		return PTR_ERR(ctx);
+	}
+
 	switch (prog->type) {
 	case BPF_PROG_TYPE_SCHED_CLS:
 	case BPF_PROG_TYPE_SCHED_ACT:
@@ -158,6 +270,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	sk = kzalloc(sizeof(struct sock), GFP_USER);
 	if (!sk) {
 		kfree(data);
+		kfree(ctx);
 		return -ENOMEM;
 	}
 	sock_net_set(sk, current->nsproxy->net_ns);
@@ -166,6 +279,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	skb = build_skb(data, 0);
 	if (!skb) {
 		kfree(data);
+		kfree(ctx);
 		kfree(sk);
 		return -ENOMEM;
 	}
@@ -180,32 +294,37 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 		__skb_push(skb, hh_len);
 	if (is_direct_pkt_access)
 		bpf_compute_data_pointers(skb);
+	ret = convert___skb_to_skb(skb, ctx);
+	if (ret)
+		goto out;
 	ret = bpf_test_run(prog, skb, repeat, &retval, &duration);
-	if (ret) {
-		kfree_skb(skb);
-		kfree(sk);
-		return ret;
-	}
+	if (ret)
+		goto out;
 	if (!is_l2) {
 		if (skb_headroom(skb) < hh_len) {
 			int nhead = HH_DATA_ALIGN(hh_len - skb_headroom(skb));
 
 			if (pskb_expand_head(skb, nhead, 0, GFP_USER)) {
-				kfree_skb(skb);
-				kfree(sk);
-				return -ENOMEM;
+				ret = -ENOMEM;
+				goto out;
 			}
 		}
 		memset(__skb_push(skb, hh_len), 0, hh_len);
 	}
+	convert_skb_to___skb(skb, ctx);
 
Quoted 8 lines

net/bpf/test_run.c int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, · line 322

Quoted 7 lines
net/bpf/test_run.c +2 -0
 	size = skb->len;
 	/* bpf program can never convert linear skb to non-linear */
 	if (WARN_ON_ONCE(skb_is_nonlinear(skb)))
 		size = skb_headlen(skb);
 	ret = bpf_test_finish(kattr, uattr, skb->data, size, retval, duration);
+	if (!ret)
+		ret = bpf_ctx_finish(kattr, uattr, ctx,

ctx is not checked for NULL in here either.

Good catch, I didn't think about the case where ctx_in is not set, but ctx_out is set. Let me handle that in the v3.

Thank you for a review!

Quoted 12 lines
Quoted 12 lines
net/bpf/test_run.c +3 -0
+				     sizeof(struct __sk_buff));
+out:
 	kfree_skb(skb);
 	kfree(sk);
+	kfree(ctx);
 	return ret;
 }
 
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#5 lore raw
Quoted 137 lines · Stanislav Fomichev
Quoted 108 lines · Martin Lau
Quoted 86 lines · Stanislav Fomichev

Add new set of arguments to bpf_attr for BPF_PROG_TEST_RUN:
* ctx_in/ctx_size_in - input context
* ctx_out/ctx_size_out - output context

The intended use case is to pass some meta data to the test runs that operate on skb (this has being brought up on recent LPC).

For programs that use bpf_prog_test_run_skb, support __sk_buff input and output. Initially, from input __sk_buff, copy _only_ cb and priority into skb, all other non-zero fields are prohibited (with EINVAL). If the user has set ctx_out/ctx_size_out, copy the potentially modified __sk_buff back to the userspace.

We require all fields of input __sk_buff 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.

The API is intentionally vague (i.e. we don't explicitly add __sk_buff to bpf_attr, but ctx_in) to potentially let other test_run types use this interface in the future (this can be xdp_md for xdp types for example).

v2:
* Addressed comments from Martin Lau

---
 include/uapi/linux/bpf.h |   7 ++
 kernel/bpf/syscall.c     |  10 ++-
 net/bpf/test_run.c       | 135 ++++++++++++++++++++++++++++++++++++---
 3 files changed, 143 insertions(+), 9 deletions(-)
include/uapi/linux/bpf.h +7 -0
index 837024512baf..8e96f99cebf8 100644
@@ -396,6 +396,13 @@ union bpf_attr {
 		__aligned_u64	data_out;
 		__u32		repeat;
 		__u32		duration;
+		__u32		ctx_size_in;	/* input: len of ctx_in */
+		__u32		ctx_size_out;	/* input/output: len of ctx_out
+						 *   returns ENOSPC if ctx_out
+						 *   is too small.
+						 */
+		__aligned_u64	ctx_in;
+		__aligned_u64	ctx_out;
 	} test;
 
 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
kernel/bpf/syscall.c +9 -1
index 1d65e56594db..5bb963e8f9b0 100644
@@ -1949,7 +1949,7 @@ static int bpf_prog_query(const union bpf_attr *attr,
 	return cgroup_bpf_prog_query(attr, uattr);
 }
 
-#define BPF_PROG_TEST_RUN_LAST_FIELD test.duration
+#define BPF_PROG_TEST_RUN_LAST_FIELD test.ctx_out
 
 static int bpf_prog_test_run(const union bpf_attr *attr,
 			     union bpf_attr __user *uattr)
@@ -1962,6 +1962,14 @@ static int bpf_prog_test_run(const union bpf_attr *attr,
 	if (CHECK_ATTR(BPF_PROG_TEST_RUN))
 		return -EINVAL;
 
+	if ((attr->test.ctx_size_in && !attr->test.ctx_in) ||
+	    (!attr->test.ctx_size_in && attr->test.ctx_in))
+		return -EINVAL;
+
+	if ((attr->test.ctx_size_out && !attr->test.ctx_out) ||
+	    (!attr->test.ctx_size_out && attr->test.ctx_out))
+		return -EINVAL;
+
 	prog = bpf_prog_get(attr->test.prog_fd);
 	if (IS_ERR(prog))
 		return PTR_ERR(prog);
net/bpf/test_run.c +0 -0
index fab142b796ef..3c6e3855c886 100644
@@ -123,12 +123,118 @@ static void *bpf_test_init(const union bpf_attr *kattr, u32 size,
 	return data;

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 133

Quoted 10 lines
net/bpf/test_run.c +8 -0
 }
 
+static void *bpf_ctx_init(const union bpf_attr *kattr, u32 max_size)
+{
+	void __user *data_in = u64_to_user_ptr(kattr->test.ctx_in);
+	u32 size = kattr->test.ctx_size_in;
+	void *data;
+
+	if (!size)
+		return NULL;

A note for a later comment in bpf_prog_test_run_skb(). This function may return NULL.

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 135

Quoted 2 lines
net/bpf/test_run.c +2 -0
+
+	if (size > max_size)

If should be fine if the kernel unsupported bytes are zeros (i.e. tailing zeros in data_in).
Please refer to the existing bpf_check_uarg_tail_zero() usages or also CHECK_ATTR() in syscall.c. Something like this if bpf_check_uarg_tail_zero() is used,

Quoted 4 lines
	err = bpf_check_uarg_tail_zero(data_in, max_size, size));
	if (err)
		return ERR_PTR(err);

Nice! Will use that instead.

Quoted 14 lines
Quoted 14 lines
net/bpf/test_run.c +14 -0
+		return ERR_PTR(-E2BIG);
+
+	data = kzalloc(max_size, GFP_USER);
+	if (!data)
+		return ERR_PTR(-ENOMEM);
+
+	if (copy_from_user(data, data_in, size)) {
+		kfree(data);
+		return ERR_PTR(-EFAULT);
+	}
+	return data;
+}
+
+static int bpf_ctx_finish(const union bpf_attr *kattr,
Quoted 8 lines

net/bpf/test_run.c static void *bpf_test_init(const union bpf_attr *kattr, u32 size, · line 156

Quoted 7 lines
net/bpf/test_run.c +7 -0
+			  union bpf_attr __user *uattr, const void *data,
+			  u32 size)
+{
+	void __user *data_out = u64_to_user_ptr(kattr->test.ctx_out);
+	int err = -EFAULT;
+	u32 copy_size = size;
+

data is not checked for NULL here.

Quoted 5 lines

I do check kattr->test.ctx_size_out on the next line. And I've added checks to the syscall.c that make sure that ctx_size_out and ctx_out are consistent (i.e. if ctx_size_out is 0, ctx_out is NULL).

Let me instead check kattr->test.ctx_out, that should be more clear.

Note that I meant "data" instead of "data_out". The caller "bpf_prog_test_run_skb()" passed in "ctx" (input) as "data" here.

I think !kattr->test.ctx_size_out and !data_out are equally good. Your call ;)

Thanks!

[ ... ]

Quoted 94 lines
Quoted 70 lines
Quoted 70 lines
net/bpf/test_run.c +17 -8
 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 			  union bpf_attr __user *uattr)
 {
 	bool is_l2 = false, is_direct_pkt_access = false;
 	u32 size = kattr->test.data_size_in;
 	u32 repeat = kattr->test.repeat;
+	struct __sk_buff *ctx = NULL;
 	u32 retval, duration;
 	int hh_len = ETH_HLEN;
 	struct sk_buff *skb;
@@ -141,6 +247,12 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	if (IS_ERR(data))
 		return PTR_ERR(data);
 
+	ctx = bpf_ctx_init(kattr, sizeof(struct __sk_buff));
+	if (IS_ERR(ctx)) {
+		kfree(data);
+		return PTR_ERR(ctx);
+	}
+
 	switch (prog->type) {
 	case BPF_PROG_TYPE_SCHED_CLS:
 	case BPF_PROG_TYPE_SCHED_ACT:
@@ -158,6 +270,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	sk = kzalloc(sizeof(struct sock), GFP_USER);
 	if (!sk) {
 		kfree(data);
+		kfree(ctx);
 		return -ENOMEM;
 	}
 	sock_net_set(sk, current->nsproxy->net_ns);
@@ -166,6 +279,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 	skb = build_skb(data, 0);
 	if (!skb) {
 		kfree(data);
+		kfree(ctx);
 		kfree(sk);
 		return -ENOMEM;
 	}
@@ -180,32 +294,37 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 		__skb_push(skb, hh_len);
 	if (is_direct_pkt_access)
 		bpf_compute_data_pointers(skb);
+	ret = convert___skb_to_skb(skb, ctx);
+	if (ret)
+		goto out;
 	ret = bpf_test_run(prog, skb, repeat, &retval, &duration);
-	if (ret) {
-		kfree_skb(skb);
-		kfree(sk);
-		return ret;
-	}
+	if (ret)
+		goto out;
 	if (!is_l2) {
 		if (skb_headroom(skb) < hh_len) {
 			int nhead = HH_DATA_ALIGN(hh_len - skb_headroom(skb));
 
 			if (pskb_expand_head(skb, nhead, 0, GFP_USER)) {
-				kfree_skb(skb);
-				kfree(sk);
-				return -ENOMEM;
+				ret = -ENOMEM;
+				goto out;
 			}
 		}
 		memset(__skb_push(skb, hh_len), 0, hh_len);
 	}
+	convert_skb_to___skb(skb, ctx);
 
Quoted 8 lines

net/bpf/test_run.c int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, · line 322

Quoted 7 lines
net/bpf/test_run.c +2 -0
 	size = skb->len;
 	/* bpf program can never convert linear skb to non-linear */
 	if (WARN_ON_ONCE(skb_is_nonlinear(skb)))
 		size = skb_headlen(skb);
 	ret = bpf_test_finish(kattr, uattr, skb->data, size, retval, duration);
+	if (!ret)
+		ret = bpf_ctx_finish(kattr, uattr, ctx,

ctx is not checked for NULL in here either.

Good catch, I didn't think about the case where ctx_in is not set, but ctx_out is set. Let me handle that in the v3.

Thank you for a review!

Quoted 12 lines
Quoted 12 lines
net/bpf/test_run.c +3 -0
+				     sizeof(struct __sk_buff));
+out:
 	kfree_skb(skb);
 	kfree(sk);
+	kfree(ctx);
 	return ret;
 }
 
Signature · 1 line
2.21.0.392.gf8f6787159e-goog