[PATCH v2 bpf-next 0/4] bpf: Make the verifier recognize llvm register allocation patterns.

[PATCH v2 bpf-next 0/4] bpf: Make the verifier recognize llvm register allocation patterns.

[Alexei Starovoitov]

From: Alexei Starovoitov <ast@kernel.org>

v1->v2:
- fixed 32-bit mov issue spotted by John.
- allowed r2=r1; r3=r2; sequence as suggested by John.
- added comments, acks, more tests.

Make two verifier improvements:
- The llvm register allocator may use two different registers representing the
  same virtual register. Teach the verifier to recognize that.
- Track bounded scalar spill/fill.

The profiler[123] test in patch 3 will fail to load without patches 1 and 2.
The profiler[23] test may fail to load on older llvm due to speculative
code motion nd instruction combining optimizations that are fixed in
https://reviews.llvm.org/D85570

Alexei Starovoitov (3):
  bpf: Propagate scalar ranges through register assignments.
  selftests/bpf: Add profiler test
  selftests/bpf: Asm tests for the verifier regalloc tracking.

Yonghong Song (1):
  bpf: Track spill/fill of bounded scalars.

 kernel/bpf/verifier.c                         |  66 +-
 tools/testing/selftests/bpf/README.rst        |  38 +
 .../testing/selftests/bpf/prog_tests/align.c  |  16 +-
 .../selftests/bpf/prog_tests/test_profiler.c  |  72 ++
 tools/testing/selftests/bpf/progs/profiler.h  | 177 ++++
 .../selftests/bpf/progs/profiler.inc.h        | 969 ++++++++++++++++++
 tools/testing/selftests/bpf/progs/profiler1.c |   6 +
 tools/testing/selftests/bpf/progs/profiler2.c |   6 +
 tools/testing/selftests/bpf/progs/profiler3.c |   6 +
 .../bpf/verifier/direct_packet_access.c       |   2 +-
 .../testing/selftests/bpf/verifier/regalloc.c | 243 +++++
 11 files changed, 1591 insertions(+), 10 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/test_profiler.c
 create mode 100644 tools/testing/selftests/bpf/progs/profiler.h
 create mode 100644 tools/testing/selftests/bpf/progs/profiler.inc.h
 create mode 100644 tools/testing/selftests/bpf/progs/profiler1.c
 create mode 100644 tools/testing/selftests/bpf/progs/profiler2.c
 create mode 100644 tools/testing/selftests/bpf/progs/profiler3.c
 create mode 100644 tools/testing/selftests/bpf/verifier/regalloc.c

-- 
2.23.0

[PATCH v2 bpf-next 1/4] bpf: Propagate scalar ranges through register assignments.

[Alexei Starovoitov]

From: Alexei Starovoitov <ast@kernel.org>

The llvm register allocator may use two different registers representing the
same virtual register. In such case the following pattern can be observed:
1047: (bf) r9 = r6
1048: (a5) if r6 < 0x1000 goto pc+1
1050: ...
1051: (a5) if r9 < 0x2 goto pc+66
1052: ...
1053: (bf) r2 = r9 /* r2 needs to have upper and lower bounds */

This is normal behavior of greedy register allocator.
The slides 137+ explain why regalloc introduces such register copy:
http://llvm.org/devmtg/2018-04/slides/Yatsina-LLVM%20Greedy%20Register%20Allocator.pdf
There is no way to tell llvm 'not to do this'.
Hence the verifier has to recognize such patterns.

In order to track this information without backtracking allocate ID
for scalars in a similar way as it's done for find_good_pkt_pointers().

When the verifier encounters r9 = r6 assignment it will assign the same ID
to both registers. Later if either register range is narrowed via conditional
jump propagate the register state into the other register.

Clear register ID in adjust_reg_min_max_vals() for any alu instruction. The
register ID is ignored for scalars in regsafe() and doesn't affect state
pruning. mark_reg_unknown() clears the ID. It's used to process call, endian
and other instructions. Hence ID is explicitly cleared only in
adjust_reg_min_max_vals() and in 32-bit mov.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
---
 kernel/bpf/verifier.c                         | 50 +++++++++++++++++++
 .../testing/selftests/bpf/prog_tests/align.c  | 16 +++---
 .../bpf/verifier/direct_packet_access.c       |  2 +-
 3 files changed, 59 insertions(+), 9 deletions(-)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 62b804651a48..ba96f7e9bbc0 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -6436,6 +6436,11 @@ static int adjust_reg_min_max_vals(struct bpf_verifier_env *env,
 	src_reg = NULL;
 	if (dst_reg->type != SCALAR_VALUE)
 		ptr_reg = dst_reg;
+	else
+		/* Make sure ID is cleared otherwise dst_reg min/max could be
+		 * incorrectly propagated into other registers by find_equal_scalars()
+		 */
+		dst_reg->id = 0;
 	if (BPF_SRC(insn->code) == BPF_X) {
 		src_reg = &regs[insn->src_reg];
 		if (src_reg->type != SCALAR_VALUE) {
@@ -6569,6 +6574,12 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn)
 				/* case: R1 = R2
 				 * copy register state to dest reg
 				 */
+				if (src_reg->type == SCALAR_VALUE && !src_reg->id)
+					/* Assign src and dst registers the same ID
+					 * that will be used by find_equal_scalars()
+					 * to propagate min/max range.
+					 */
+					src_reg->id = ++env->id_gen;
 				*dst_reg = *src_reg;
 				dst_reg->live |= REG_LIVE_WRITTEN;
 				dst_reg->subreg_def = DEF_NOT_SUBREG;
@@ -6581,6 +6592,11 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn)
 					return -EACCES;
 				} else if (src_reg->type == SCALAR_VALUE) {
 					*dst_reg = *src_reg;
+					/* Make sure ID is cleared otherwise
+					 * dst_reg min/max could be incorrectly
+					 * propagated into src_reg by find_equal_scalars()
+					 */
+					dst_reg->id = 0;
 					dst_reg->live |= REG_LIVE_WRITTEN;
 					dst_reg->subreg_def = env->insn_idx + 1;
 				} else {
@@ -7369,6 +7385,30 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn,
 	return true;
 }
 
+static void find_equal_scalars(struct bpf_verifier_state *vstate,
+			       struct bpf_reg_state *known_reg)
+{
+	struct bpf_func_state *state;
+	struct bpf_reg_state *reg;
+	int i, j;
+
+	for (i = 0; i <= vstate->curframe; i++) {
+		state = vstate->frame[i];
+		for (j = 0; j < MAX_BPF_REG; j++) {
+			reg = &state->regs[j];
+			if (reg->type == SCALAR_VALUE && reg->id == known_reg->id)
+				*reg = *known_reg;
+		}
+
+		bpf_for_each_spilled_reg(j, state, reg) {
+			if (!reg)
+				continue;
+			if (reg->type == SCALAR_VALUE && reg->id == known_reg->id)
+				*reg = *known_reg;
+		}
+	}
+}
+
 static int check_cond_jmp_op(struct bpf_verifier_env *env,
 			     struct bpf_insn *insn, int *insn_idx)
 {
@@ -7497,6 +7537,11 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env,
 				reg_combine_min_max(&other_branch_regs[insn->src_reg],
 						    &other_branch_regs[insn->dst_reg],
 						    src_reg, dst_reg, opcode);
+			if (src_reg->id) {
+				find_equal_scalars(this_branch, src_reg);
+				find_equal_scalars(other_branch, &other_branch_regs[insn->src_reg]);
+			}
+
 		}
 	} else if (dst_reg->type == SCALAR_VALUE) {
 		reg_set_min_max(&other_branch_regs[insn->dst_reg],
@@ -7504,6 +7549,11 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env,
 					opcode, is_jmp32);
 	}
 
+	if (dst_reg->type == SCALAR_VALUE && dst_reg->id) {
+		find_equal_scalars(this_branch, dst_reg);
+		find_equal_scalars(other_branch, &other_branch_regs[insn->dst_reg]);
+	}
+
 	/* detect if R == 0 where R is returned from bpf_map_lookup_elem().
 	 * NOTE: these optimizations below are related with pointer comparison
 	 *       which will never be JMP32.
diff --git a/tools/testing/selftests/bpf/prog_tests/align.c b/tools/testing/selftests/bpf/prog_tests/align.c
index c548aded6585..52414058a627 100644
--- a/tools/testing/selftests/bpf/prog_tests/align.c
+++ b/tools/testing/selftests/bpf/prog_tests/align.c
@@ -195,13 +195,13 @@ static struct bpf_align_test tests[] = {
 		.prog_type = BPF_PROG_TYPE_SCHED_CLS,
 		.matches = {
 			{7, "R3_w=inv(id=0,umax_value=255,var_off=(0x0; 0xff))"},
-			{8, "R4_w=inv(id=0,umax_value=255,var_off=(0x0; 0xff))"},
+			{8, "R4_w=inv(id=1,umax_value=255,var_off=(0x0; 0xff))"},
 			{9, "R4_w=inv(id=0,umax_value=255,var_off=(0x0; 0xff))"},
-			{10, "R4_w=inv(id=0,umax_value=255,var_off=(0x0; 0xff))"},
+			{10, "R4_w=inv(id=1,umax_value=255,var_off=(0x0; 0xff))"},
 			{11, "R4_w=inv(id=0,umax_value=510,var_off=(0x0; 0x1fe))"},
-			{12, "R4_w=inv(id=0,umax_value=255,var_off=(0x0; 0xff))"},
+			{12, "R4_w=inv(id=1,umax_value=255,var_off=(0x0; 0xff))"},
 			{13, "R4_w=inv(id=0,umax_value=1020,var_off=(0x0; 0x3fc))"},
-			{14, "R4_w=inv(id=0,umax_value=255,var_off=(0x0; 0xff))"},
+			{14, "R4_w=inv(id=1,umax_value=255,var_off=(0x0; 0xff))"},
 			{15, "R4_w=inv(id=0,umax_value=2040,var_off=(0x0; 0x7f8))"},
 			{16, "R4_w=inv(id=0,umax_value=4080,var_off=(0x0; 0xff0))"},
 		},
@@ -518,7 +518,7 @@ static struct bpf_align_test tests[] = {
 			 * the total offset is 4-byte aligned and meets the
 			 * load's requirements.
 			 */
-			{20, "R5=pkt(id=1,off=0,r=4,umin_value=2,umax_value=1034,var_off=(0x2; 0x7fc)"},
+			{20, "R5=pkt(id=2,off=0,r=4,umin_value=2,umax_value=1034,var_off=(0x2; 0x7fc)"},
 
 		},
 	},
@@ -561,18 +561,18 @@ static struct bpf_align_test tests[] = {
 			/* Adding 14 makes R6 be (4n+2) */
 			{11, "R6_w=inv(id=0,umin_value=14,umax_value=74,var_off=(0x2; 0x7c))"},
 			/* Subtracting from packet pointer overflows ubounds */
-			{13, "R5_w=pkt(id=1,off=0,r=8,umin_value=18446744073709551542,umax_value=18446744073709551602,var_off=(0xffffffffffffff82; 0x7c)"},
+			{13, "R5_w=pkt(id=2,off=0,r=8,umin_value=18446744073709551542,umax_value=18446744073709551602,var_off=(0xffffffffffffff82; 0x7c)"},
 			/* New unknown value in R7 is (4n), >= 76 */
 			{15, "R7_w=inv(id=0,umin_value=76,umax_value=1096,var_off=(0x0; 0x7fc))"},
 			/* Adding it to packet pointer gives nice bounds again */
-			{16, "R5_w=pkt(id=2,off=0,r=0,umin_value=2,umax_value=1082,var_off=(0x2; 0xfffffffc)"},
+			{16, "R5_w=pkt(id=3,off=0,r=0,umin_value=2,umax_value=1082,var_off=(0x2; 0xfffffffc)"},
 			/* At the time the word size load is performed from R5,
 			 * its total fixed offset is NET_IP_ALIGN + reg->off (0)
 			 * which is 2.  Then the variable offset is (4n+2), so
 			 * the total offset is 4-byte aligned and meets the
 			 * load's requirements.
 			 */
-			{20, "R5=pkt(id=2,off=0,r=4,umin_value=2,umax_value=1082,var_off=(0x2; 0xfffffffc)"},
+			{20, "R5=pkt(id=3,off=0,r=4,umin_value=2,umax_value=1082,var_off=(0x2; 0xfffffffc)"},
 		},
 	},
 };
diff --git a/tools/testing/selftests/bpf/verifier/direct_packet_access.c b/tools/testing/selftests/bpf/verifier/direct_packet_access.c
index 2c5fbe7bcd27..ae72536603fe 100644
--- a/tools/testing/selftests/bpf/verifier/direct_packet_access.c
+++ b/tools/testing/selftests/bpf/verifier/direct_packet_access.c
@@ -529,7 +529,7 @@
 	},
 	.prog_type = BPF_PROG_TYPE_SCHED_CLS,
 	.result = REJECT,
-	.errstr = "invalid access to packet, off=0 size=8, R5(id=1,off=0,r=0)",
+	.errstr = "invalid access to packet, off=0 size=8, R5(id=2,off=0,r=0)",
 	.flags = F_NEEDS_EFFICIENT_UNALIGNED_ACCESS,
 },
 {
-- 
2.23.0

[PATCH v2 bpf-next 2/4] bpf: Track spill/fill of bounded scalars.

[Alexei Starovoitov]

From: Yonghong Song <yhs@fb.com>

Under register pressure the llvm may spill registers with bounds into the stack.
The verifier has to track them through spill/fill otherwise many kinds of bound
errors will be seen. The spill/fill of induction variables was already
happening. This patch extends this logic from tracking spill/fill of a constant
into any bounded register. There is no need to track spill/fill of unbounded,
since no new information will be retrieved from the stack during register fill.

Though extra stack difference could cause state pruning to be less effective, no
adverse affects were seen from this patch on selftests and on cilium programs.

Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
---
 kernel/bpf/verifier.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index ba96f7e9bbc0..f3e36eade3d4 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -2227,6 +2227,20 @@ static bool register_is_const(struct bpf_reg_state *reg)
 	return reg->type == SCALAR_VALUE && tnum_is_const(reg->var_off);
 }
 
+static bool __is_scalar_unbounded(struct bpf_reg_state *reg)
+{
+	return tnum_is_unknown(reg->var_off) &&
+	       reg->smin_value == S64_MIN && reg->smax_value == S64_MAX &&
+	       reg->umin_value == 0 && reg->umax_value == U64_MAX &&
+	       reg->s32_min_value == S32_MIN && reg->s32_max_value == S32_MAX &&
+	       reg->u32_min_value == 0 && reg->u32_max_value == U32_MAX;
+}
+
+static bool register_is_bounded(struct bpf_reg_state *reg)
+{
+	return reg->type == SCALAR_VALUE && !__is_scalar_unbounded(reg);
+}
+
 static bool __is_pointer_value(bool allow_ptr_leaks,
 			       const struct bpf_reg_state *reg)
 {
@@ -2278,7 +2292,7 @@ static int check_stack_write(struct bpf_verifier_env *env,
 	if (value_regno >= 0)
 		reg = &cur->regs[value_regno];
 
-	if (reg && size == BPF_REG_SIZE && register_is_const(reg) &&
+	if (reg && size == BPF_REG_SIZE && register_is_bounded(reg) &&
 	    !register_is_null(reg) && env->bpf_capable) {
 		if (dst_reg != BPF_REG_FP) {
 			/* The backtracking logic can only recognize explicit
-- 
2.23.0

[PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Alexei Starovoitov]

From: Alexei Starovoitov <ast@kernel.org>

The main purpose of the profiler test to check different llvm generation
patterns to make sure the verifier can load these large programs.

Note that profiler.inc.h test doesn't follow strict kernel coding style.
The code was formatted in the kernel style, but variable declarations are
kept as-is to preserve original llvm IR pattern.

profiler1.c should pass with older and newer llvm

profiler[23].c may fail on older llvm that don't have:
https://reviews.llvm.org/D85570
because llvm may do speculative code motion optimization that
will generate code like this:

// r9 is a pointer to map_value
// r7 is a scalar
17:       bf 96 00 00 00 00 00 00 r6 = r9
18:       0f 76 00 00 00 00 00 00 r6 += r7
19:       a5 07 01 00 01 01 00 00 if r7 < 257 goto +1
20:       bf 96 00 00 00 00 00 00 r6 = r9
// r6 is used here

The verifier will reject such code with the error:
"math between map_value pointer and register with unbounded min value is not allowed"
At insn 18 the r7 is indeed unbounded. The later insn 19 checks the bounds and
the insn 20 undoes map_value addition. It is currently impossible for the
verifier to understand such speculative pointer arithmetic. Hence llvm D85570
addresses it on the compiler side.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
---
 tools/testing/selftests/bpf/README.rst        |  38 +
 .../selftests/bpf/prog_tests/test_profiler.c  |  72 ++
 tools/testing/selftests/bpf/progs/profiler.h  | 177 ++++
 .../selftests/bpf/progs/profiler.inc.h        | 969 ++++++++++++++++++
 tools/testing/selftests/bpf/progs/profiler1.c |   6 +
 tools/testing/selftests/bpf/progs/profiler2.c |   6 +
 tools/testing/selftests/bpf/progs/profiler3.c |   6 +
 7 files changed, 1274 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/test_profiler.c
 create mode 100644 tools/testing/selftests/bpf/progs/profiler.h
 create mode 100644 tools/testing/selftests/bpf/progs/profiler.inc.h
 create mode 100644 tools/testing/selftests/bpf/progs/profiler1.c
 create mode 100644 tools/testing/selftests/bpf/progs/profiler2.c
 create mode 100644 tools/testing/selftests/bpf/progs/profiler3.c

diff --git a/tools/testing/selftests/bpf/README.rst b/tools/testing/selftests/bpf/README.rst
index 66acfcf15ff2..ac9eda830187 100644
--- a/tools/testing/selftests/bpf/README.rst
+++ b/tools/testing/selftests/bpf/README.rst
@@ -7,6 +7,44 @@ General instructions on running selftests can be found in
 Additional information about selftest failures are
 documented here.
 
+profiler[23] test failures with clang/llvm <12.0.0
+==================================================
+
+With clang/llvm <12.0.0, the profiler[23] test may fail.
+The symptom looks like
+
+.. code-block:: c
+
+  // r9 is a pointer to map_value
+  // r7 is a scalar
+  17:       bf 96 00 00 00 00 00 00 r6 = r9
+  18:       0f 76 00 00 00 00 00 00 r6 += r7
+  math between map_value pointer and register with unbounded min value is not allowed
+
+  // the instructions below will not be seen in the verifier log
+  19:       a5 07 01 00 01 01 00 00 if r7 < 257 goto +1
+  20:       bf 96 00 00 00 00 00 00 r6 = r9
+  // r6 is used here
+
+The verifier will reject such code with above error.
+At insn 18 the r7 is indeed unbounded. The later insn 19 checks the bounds and
+the insn 20 undoes map_value addition. It is currently impossible for the
+verifier to understand such speculative pointer arithmetic.
+Hence
+    https://reviews.llvm.org/D85570
+addresses it on the compiler side. It was committed on llvm 12.
+
+The corresponding C code
+.. code-block:: c
+
+  for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {
+          filepart_length = bpf_probe_read_str(payload, ...);
+          if (filepart_length <= MAX_PATH) {
+                  barrier_var(filepart_length); // workaround
+                  payload += filepart_length;
+          }
+  }
+
 bpf_iter test failures with clang/llvm 10.0.0
 =============================================
 
diff --git a/tools/testing/selftests/bpf/prog_tests/test_profiler.c b/tools/testing/selftests/bpf/prog_tests/test_profiler.c
new file mode 100644
index 000000000000..4ca275101ee0
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/test_profiler.c
@@ -0,0 +1,72 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#include <test_progs.h>
+#include "progs/profiler.h"
+#include "profiler1.skel.h"
+#include "profiler2.skel.h"
+#include "profiler3.skel.h"
+
+static int sanity_run(struct bpf_program *prog)
+{
+	struct bpf_prog_test_run_attr test_attr = {};
+	__u64 args[] = {1, 2, 3};
+	__u32 duration = 0;
+	int err, prog_fd;
+
+	prog_fd = bpf_program__fd(prog);
+	test_attr.prog_fd = prog_fd;
+	test_attr.ctx_in = args;
+	test_attr.ctx_size_in = sizeof(args);
+	err = bpf_prog_test_run_xattr(&test_attr);
+	if (CHECK(err || test_attr.retval, "test_run",
+		  "err %d errno %d retval %d duration %d\n",
+		  err, errno, test_attr.retval, duration))
+		return -1;
+	return 0;
+}
+
+void test_test_profiler(void)
+{
+	struct profiler1 *profiler1_skel = NULL;
+	struct profiler2 *profiler2_skel = NULL;
+	struct profiler3 *profiler3_skel = NULL;
+	__u32 duration = 0;
+	int err;
+
+	profiler1_skel = profiler1__open_and_load();
+	if (CHECK(!profiler1_skel, "profiler1_skel_load", "profiler1 skeleton failed\n"))
+		goto cleanup;
+
+	err = profiler1__attach(profiler1_skel);
+	if (CHECK(err, "profiler1_attach", "profiler1 attach failed: %d\n", err))
+		goto cleanup;
+
+	if (sanity_run(profiler1_skel->progs.raw_tracepoint__sched_process_exec))
+		goto cleanup;
+
+	profiler2_skel = profiler2__open_and_load();
+	if (CHECK(!profiler2_skel, "profiler2_skel_load", "profiler2 skeleton failed\n"))
+		goto cleanup;
+
+	err = profiler2__attach(profiler2_skel);
+	if (CHECK(err, "profiler2_attach", "profiler2 attach failed: %d\n", err))
+		goto cleanup;
+
+	if (sanity_run(profiler2_skel->progs.raw_tracepoint__sched_process_exec))
+		goto cleanup;
+
+	profiler3_skel = profiler3__open_and_load();
+	if (CHECK(!profiler3_skel, "profiler3_skel_load", "profiler3 skeleton failed\n"))
+		goto cleanup;
+
+	err = profiler3__attach(profiler3_skel);
+	if (CHECK(err, "profiler3_attach", "profiler3 attach failed: %d\n", err))
+		goto cleanup;
+
+	if (sanity_run(profiler3_skel->progs.raw_tracepoint__sched_process_exec))
+		goto cleanup;
+cleanup:
+	profiler1__destroy(profiler1_skel);
+	profiler2__destroy(profiler2_skel);
+	profiler3__destroy(profiler3_skel);
+}
diff --git a/tools/testing/selftests/bpf/progs/profiler.h b/tools/testing/selftests/bpf/progs/profiler.h
new file mode 100644
index 000000000000..3bac4fdd4bdf
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/profiler.h
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#pragma once
+
+#define TASK_COMM_LEN 16
+#define MAX_ANCESTORS 4
+#define MAX_PATH 256
+#define KILL_TARGET_LEN 64
+#define CTL_MAXNAME 10
+#define MAX_ARGS_LEN 4096
+#define MAX_FILENAME_LEN 512
+#define MAX_ENVIRON_LEN 8192
+#define MAX_PATH_DEPTH 32
+#define MAX_FILEPATH_LENGTH (MAX_PATH_DEPTH * MAX_PATH)
+#define MAX_CGROUPS_PATH_DEPTH 8
+
+#define MAX_METADATA_PAYLOAD_LEN TASK_COMM_LEN
+
+#define MAX_CGROUP_PAYLOAD_LEN \
+	(MAX_PATH * 2 + (MAX_PATH * MAX_CGROUPS_PATH_DEPTH))
+
+#define MAX_CAP_PAYLOAD_LEN (MAX_METADATA_PAYLOAD_LEN + MAX_CGROUP_PAYLOAD_LEN)
+
+#define MAX_SYSCTL_PAYLOAD_LEN \
+	(MAX_METADATA_PAYLOAD_LEN + MAX_CGROUP_PAYLOAD_LEN + CTL_MAXNAME + MAX_PATH)
+
+#define MAX_KILL_PAYLOAD_LEN \
+	(MAX_METADATA_PAYLOAD_LEN + MAX_CGROUP_PAYLOAD_LEN + TASK_COMM_LEN + \
+	 KILL_TARGET_LEN)
+
+#define MAX_EXEC_PAYLOAD_LEN \
+	(MAX_METADATA_PAYLOAD_LEN + MAX_CGROUP_PAYLOAD_LEN + MAX_FILENAME_LEN + \
+	 MAX_ARGS_LEN + MAX_ENVIRON_LEN)
+
+#define MAX_FILEMOD_PAYLOAD_LEN \
+	(MAX_METADATA_PAYLOAD_LEN + MAX_CGROUP_PAYLOAD_LEN + MAX_FILEPATH_LENGTH + \
+	 MAX_FILEPATH_LENGTH)
+
+enum data_type {
+	INVALID_EVENT,
+	EXEC_EVENT,
+	FORK_EVENT,
+	KILL_EVENT,
+	SYSCTL_EVENT,
+	FILEMOD_EVENT,
+	MAX_DATA_TYPE_EVENT
+};
+
+enum filemod_type {
+	FMOD_OPEN,
+	FMOD_LINK,
+	FMOD_SYMLINK,
+};
+
+struct ancestors_data_t {
+	pid_t ancestor_pids[MAX_ANCESTORS];
+	uint32_t ancestor_exec_ids[MAX_ANCESTORS];
+	uint64_t ancestor_start_times[MAX_ANCESTORS];
+	uint32_t num_ancestors;
+};
+
+struct var_metadata_t {
+	enum data_type type;
+	pid_t pid;
+	uint32_t exec_id;
+	uid_t uid;
+	gid_t gid;
+	uint64_t start_time;
+	uint32_t cpu_id;
+	uint64_t bpf_stats_num_perf_events;
+	uint64_t bpf_stats_start_ktime_ns;
+	uint8_t comm_length;
+};
+
+struct cgroup_data_t {
+	ino_t cgroup_root_inode;
+	ino_t cgroup_proc_inode;
+	uint64_t cgroup_root_mtime;
+	uint64_t cgroup_proc_mtime;
+	uint16_t cgroup_root_length;
+	uint16_t cgroup_proc_length;
+	uint16_t cgroup_full_length;
+	int cgroup_full_path_root_pos;
+};
+
+struct var_sysctl_data_t {
+	struct var_metadata_t meta;
+	struct cgroup_data_t cgroup_data;
+	struct ancestors_data_t ancestors_info;
+	uint8_t sysctl_val_length;
+	uint16_t sysctl_path_length;
+	char payload[MAX_SYSCTL_PAYLOAD_LEN];
+};
+
+struct var_kill_data_t {
+	struct var_metadata_t meta;
+	struct cgroup_data_t cgroup_data;
+	struct ancestors_data_t ancestors_info;
+	pid_t kill_target_pid;
+	int kill_sig;
+	uint32_t kill_count;
+	uint64_t last_kill_time;
+	uint8_t kill_target_name_length;
+	uint8_t kill_target_cgroup_proc_length;
+	char payload[MAX_KILL_PAYLOAD_LEN];
+	size_t payload_length;
+};
+
+struct var_exec_data_t {
+	struct var_metadata_t meta;
+	struct cgroup_data_t cgroup_data;
+	pid_t parent_pid;
+	uint32_t parent_exec_id;
+	uid_t parent_uid;
+	uint64_t parent_start_time;
+	uint16_t bin_path_length;
+	uint16_t cmdline_length;
+	uint16_t environment_length;
+	char payload[MAX_EXEC_PAYLOAD_LEN];
+};
+
+struct var_fork_data_t {
+	struct var_metadata_t meta;
+	pid_t parent_pid;
+	uint32_t parent_exec_id;
+	uint64_t parent_start_time;
+	char payload[MAX_METADATA_PAYLOAD_LEN];
+};
+
+struct var_filemod_data_t {
+	struct var_metadata_t meta;
+	struct cgroup_data_t cgroup_data;
+	enum filemod_type fmod_type;
+	unsigned int dst_flags;
+	uint32_t src_device_id;
+	uint32_t dst_device_id;
+	ino_t src_inode;
+	ino_t dst_inode;
+	uint16_t src_filepath_length;
+	uint16_t dst_filepath_length;
+	char payload[MAX_FILEMOD_PAYLOAD_LEN];
+};
+
+struct profiler_config_struct {
+	bool fetch_cgroups_from_bpf;
+	ino_t cgroup_fs_inode;
+	ino_t cgroup_login_session_inode;
+	uint64_t kill_signals_mask;
+	ino_t inode_filter;
+	uint32_t stale_info_secs;
+	bool use_variable_buffers;
+	bool read_environ_from_exec;
+	bool enable_cgroup_v1_resolver;
+};
+
+struct bpf_func_stats_data {
+	uint64_t time_elapsed_ns;
+	uint64_t num_executions;
+	uint64_t num_perf_events;
+};
+
+struct bpf_func_stats_ctx {
+	uint64_t start_time_ns;
+	struct bpf_func_stats_data* bpf_func_stats_data_val;
+};
+
+enum bpf_function_id {
+	profiler_bpf_proc_sys_write,
+	profiler_bpf_sched_process_exec,
+	profiler_bpf_sched_process_exit,
+	profiler_bpf_sys_enter_kill,
+	profiler_bpf_do_filp_open_ret,
+	profiler_bpf_sched_process_fork,
+	profiler_bpf_vfs_link,
+	profiler_bpf_vfs_symlink,
+	profiler_bpf_max_function_id
+};
diff --git a/tools/testing/selftests/bpf/progs/profiler.inc.h b/tools/testing/selftests/bpf/progs/profiler.inc.h
new file mode 100644
index 000000000000..00578311a423
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/profiler.inc.h
@@ -0,0 +1,969 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#include <vmlinux.h>
+#include <bpf/bpf_core_read.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+#include "profiler.h"
+
+#ifndef NULL
+#define NULL 0
+#endif
+
+#define O_WRONLY 00000001
+#define O_RDWR 00000002
+#define O_DIRECTORY 00200000
+#define __O_TMPFILE 020000000
+#define O_TMPFILE (__O_TMPFILE | O_DIRECTORY)
+#define MAX_ERRNO 4095
+#define S_IFMT 00170000
+#define S_IFSOCK 0140000
+#define S_IFLNK 0120000
+#define S_IFREG 0100000
+#define S_IFBLK 0060000
+#define S_IFDIR 0040000
+#define S_IFCHR 0020000
+#define S_IFIFO 0010000
+#define S_ISUID 0004000
+#define S_ISGID 0002000
+#define S_ISVTX 0001000
+#define S_ISLNK(m) (((m)&S_IFMT) == S_IFLNK)
+#define S_ISDIR(m) (((m)&S_IFMT) == S_IFDIR)
+#define S_ISCHR(m) (((m)&S_IFMT) == S_IFCHR)
+#define S_ISBLK(m) (((m)&S_IFMT) == S_IFBLK)
+#define S_ISFIFO(m) (((m)&S_IFMT) == S_IFIFO)
+#define S_ISSOCK(m) (((m)&S_IFMT) == S_IFSOCK)
+#define IS_ERR_VALUE(x) (unsigned long)(void*)(x) >= (unsigned long)-MAX_ERRNO
+
+#define KILL_DATA_ARRAY_SIZE 8
+
+struct var_kill_data_arr_t {
+	struct var_kill_data_t array[KILL_DATA_ARRAY_SIZE];
+};
+
+union any_profiler_data_t {
+	struct var_exec_data_t var_exec;
+	struct var_kill_data_t var_kill;
+	struct var_sysctl_data_t var_sysctl;
+	struct var_filemod_data_t var_filemod;
+	struct var_fork_data_t var_fork;
+	struct var_kill_data_arr_t var_kill_data_arr;
+};
+
+volatile struct profiler_config_struct bpf_config = {};
+
+#define FETCH_CGROUPS_FROM_BPF (bpf_config.fetch_cgroups_from_bpf)
+#define CGROUP_FS_INODE (bpf_config.cgroup_fs_inode)
+#define CGROUP_LOGIN_SESSION_INODE \
+	(bpf_config.cgroup_login_session_inode)
+#define KILL_SIGNALS (bpf_config.kill_signals_mask)
+#define STALE_INFO (bpf_config.stale_info_secs)
+#define INODE_FILTER (bpf_config.inode_filter)
+#define READ_ENVIRON_FROM_EXEC (bpf_config.read_environ_from_exec)
+#define ENABLE_CGROUP_V1_RESOLVER (bpf_config.enable_cgroup_v1_resolver)
+
+struct kernfs_iattrs___52 {
+	struct iattr ia_iattr;
+};
+
+struct kernfs_node___52 {
+	union /* kernfs_node_id */ {
+		struct {
+			u32 ino;
+			u32 generation;
+		};
+		u64 id;
+	} id;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__uint(max_entries, 1);
+	__type(key, u32);
+	__type(value, union any_profiler_data_t);
+} data_heap SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERF_EVENT_ARRAY);
+	__uint(key_size, sizeof(int));
+	__uint(value_size, sizeof(int));
+} events SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(max_entries, KILL_DATA_ARRAY_SIZE);
+	__type(key, u32);
+	__type(value, struct var_kill_data_arr_t);
+} var_tpid_to_data SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
+	__uint(max_entries, profiler_bpf_max_function_id);
+	__type(key, u32);
+	__type(value, struct bpf_func_stats_data);
+} bpf_func_stats SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, u32);
+	__type(value, bool);
+	__uint(max_entries, 16);
+} allowed_devices SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, u64);
+	__type(value, bool);
+	__uint(max_entries, 1024);
+} allowed_file_inodes SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, u64);
+	__type(value, bool);
+	__uint(max_entries, 1024);
+} allowed_directory_inodes SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, u32);
+	__type(value, bool);
+	__uint(max_entries, 16);
+} disallowed_exec_inodes SEC(".maps");
+
+#ifndef ARRAY_SIZE
+#define ARRAY_SIZE(arr) (sizeof(arr) / sizeof(arr[0]))
+#endif
+
+static INLINE bool IS_ERR(const void* ptr)
+{
+	return IS_ERR_VALUE((unsigned long)ptr);
+}
+
+static INLINE u32 get_userspace_pid()
+{
+	return bpf_get_current_pid_tgid() >> 32;
+}
+
+static INLINE bool is_init_process(u32 tgid)
+{
+	return tgid == 1 || tgid == 0;
+}
+
+static INLINE unsigned long
+probe_read_lim(void* dst, void* src, unsigned long len, unsigned long max)
+{
+	len = len < max ? len : max;
+	if (len > 1) {
+		if (bpf_probe_read(dst, len, src))
+			return 0;
+	} else if (len == 1) {
+		if (bpf_probe_read(dst, 1, src))
+			return 0;
+	}
+	return len;
+}
+
+static INLINE int get_var_spid_index(struct var_kill_data_arr_t* arr_struct,
+				     int spid)
+{
+#ifdef UNROLL
+#pragma unroll
+#endif
+	for (int i = 0; i < ARRAY_SIZE(arr_struct->array); i++)
+		if (arr_struct->array[i].meta.pid == spid)
+			return i;
+	return -1;
+}
+
+static INLINE void populate_ancestors(struct task_struct* task,
+				      struct ancestors_data_t* ancestors_data)
+{
+	struct task_struct* parent = task;
+	u32 num_ancestors, ppid;
+
+	ancestors_data->num_ancestors = 0;
+#ifdef UNROLL
+#pragma unroll
+#endif
+	for (num_ancestors = 0; num_ancestors < MAX_ANCESTORS; num_ancestors++) {
+		parent = BPF_CORE_READ(parent, real_parent);
+		if (parent == NULL)
+			break;
+		ppid = BPF_CORE_READ(parent, tgid);
+		if (is_init_process(ppid))
+			break;
+		ancestors_data->ancestor_pids[num_ancestors] = ppid;
+		ancestors_data->ancestor_exec_ids[num_ancestors] =
+			BPF_CORE_READ(parent, self_exec_id);
+		ancestors_data->ancestor_start_times[num_ancestors] =
+			BPF_CORE_READ(parent, start_time);
+		ancestors_data->num_ancestors = num_ancestors;
+	}
+}
+
+static INLINE void* read_full_cgroup_path(struct kernfs_node* cgroup_node,
+					  struct kernfs_node* cgroup_root_node,
+					  void* payload,
+					  int* root_pos)
+{
+	void* payload_start = payload;
+	size_t filepart_length;
+
+#ifdef UNROLL
+#pragma unroll
+#endif
+	for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {
+		filepart_length =
+			bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(cgroup_node, name));
+		if (!cgroup_node)
+			return payload;
+		if (cgroup_node == cgroup_root_node)
+			*root_pos = payload - payload_start;
+		if (filepart_length <= MAX_PATH) {
+			barrier_var(filepart_length);
+			payload += filepart_length;
+		}
+		cgroup_node = BPF_CORE_READ(cgroup_node, parent);
+	}
+	return payload;
+}
+
+static ino_t get_inode_from_kernfs(struct kernfs_node* node)
+{
+	struct kernfs_node___52* node52 = (void*)node;
+
+	if (bpf_core_field_exists(node52->id.ino)) {
+		barrier_var(node52);
+		return BPF_CORE_READ(node52, id.ino);
+	} else {
+		barrier_var(node);
+		return (u64)BPF_CORE_READ(node, id);
+	}
+}
+
+int pids_cgrp_id = 1;
+
+static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
+					 struct task_struct* task,
+					 void* payload)
+{
+	struct kernfs_node* root_kernfs =
+		BPF_CORE_READ(task, nsproxy, cgroup_ns, root_cset, dfl_cgrp, kn);
+	struct kernfs_node* proc_kernfs = BPF_CORE_READ(task, cgroups, dfl_cgrp, kn);
+
+	if (ENABLE_CGROUP_V1_RESOLVER) {
+#ifdef UNROLL
+#pragma unroll
+#endif
+		for (int i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
+			struct cgroup_subsys_state* subsys =
+				BPF_CORE_READ(task, cgroups, subsys[i]);
+			if (subsys != NULL) {
+				int subsys_id = BPF_CORE_READ(subsys, ss, id);
+				if (subsys_id == pids_cgrp_id) {
+					proc_kernfs = BPF_CORE_READ(subsys, cgroup, kn);
+					root_kernfs = BPF_CORE_READ(subsys, ss, root, kf_root, kn);
+					break;
+				}
+			}
+		}
+	}
+
+	cgroup_data->cgroup_root_inode = get_inode_from_kernfs(root_kernfs);
+	cgroup_data->cgroup_proc_inode = get_inode_from_kernfs(proc_kernfs);
+
+	if (bpf_core_field_exists(root_kernfs->iattr->ia_mtime)) {
+		cgroup_data->cgroup_root_mtime =
+			BPF_CORE_READ(root_kernfs, iattr, ia_mtime.tv_nsec);
+		cgroup_data->cgroup_proc_mtime =
+			BPF_CORE_READ(proc_kernfs, iattr, ia_mtime.tv_nsec);
+	} else {
+		struct kernfs_iattrs___52* root_iattr =
+			(struct kernfs_iattrs___52*)BPF_CORE_READ(root_kernfs, iattr);
+		cgroup_data->cgroup_root_mtime =
+			BPF_CORE_READ(root_iattr, ia_iattr.ia_mtime.tv_nsec);
+
+		struct kernfs_iattrs___52* proc_iattr =
+			(struct kernfs_iattrs___52*)BPF_CORE_READ(proc_kernfs, iattr);
+		cgroup_data->cgroup_proc_mtime =
+			BPF_CORE_READ(proc_iattr, ia_iattr.ia_mtime.tv_nsec);
+	}
+
+	cgroup_data->cgroup_root_length = 0;
+	cgroup_data->cgroup_proc_length = 0;
+	cgroup_data->cgroup_full_length = 0;
+
+	size_t cgroup_root_length =
+		bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(root_kernfs, name));
+	barrier_var(cgroup_root_length);
+	if (cgroup_root_length <= MAX_PATH) {
+		barrier_var(cgroup_root_length);
+		cgroup_data->cgroup_root_length = cgroup_root_length;
+		payload += cgroup_root_length;
+	}
+
+	size_t cgroup_proc_length =
+		bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(proc_kernfs, name));
+	barrier_var(cgroup_proc_length);
+	if (cgroup_proc_length <= MAX_PATH) {
+		barrier_var(cgroup_proc_length);
+		cgroup_data->cgroup_proc_length = cgroup_proc_length;
+		payload += cgroup_proc_length;
+	}
+
+	if (FETCH_CGROUPS_FROM_BPF) {
+		cgroup_data->cgroup_full_path_root_pos = -1;
+		void* payload_end_pos = read_full_cgroup_path(proc_kernfs, root_kernfs, payload,
+							      &cgroup_data->cgroup_full_path_root_pos);
+		cgroup_data->cgroup_full_length = payload_end_pos - payload;
+		payload = payload_end_pos;
+	}
+
+	return (void*)payload;
+}
+
+static INLINE void* populate_var_metadata(struct var_metadata_t* metadata,
+					  struct task_struct* task,
+					  u32 pid, void* payload)
+{
+	u64 uid_gid = bpf_get_current_uid_gid();
+
+	metadata->uid = (u32)uid_gid;
+	metadata->gid = uid_gid >> 32;
+	metadata->pid = pid;
+	metadata->exec_id = BPF_CORE_READ(task, self_exec_id);
+	metadata->start_time = BPF_CORE_READ(task, start_time);
+	metadata->comm_length = 0;
+
+	size_t comm_length = bpf_core_read_str(payload, TASK_COMM_LEN, &task->comm);
+	barrier_var(comm_length);
+	if (comm_length <= TASK_COMM_LEN) {
+		barrier_var(comm_length);
+		metadata->comm_length = comm_length;
+		payload += comm_length;
+	}
+
+	return (void*)payload;
+}
+
+static INLINE struct var_kill_data_t*
+get_var_kill_data(struct pt_regs* ctx, int spid, int tpid, int sig)
+{
+	int zero = 0;
+	struct var_kill_data_t* kill_data = bpf_map_lookup_elem(&data_heap, &zero);
+
+	if (kill_data == NULL)
+		return NULL;
+	struct task_struct* task = (struct task_struct*)bpf_get_current_task();
+
+	void* payload = populate_var_metadata(&kill_data->meta, task, spid, kill_data->payload);
+	payload = populate_cgroup_info(&kill_data->cgroup_data, task, payload);
+	size_t payload_length = payload - (void*)kill_data->payload;
+	kill_data->payload_length = payload_length;
+	populate_ancestors(task, &kill_data->ancestors_info);
+	kill_data->meta.type = KILL_EVENT;
+	kill_data->kill_target_pid = tpid;
+	kill_data->kill_sig = sig;
+	kill_data->kill_count = 1;
+	kill_data->last_kill_time = bpf_ktime_get_ns();
+	return kill_data;
+}
+
+static INLINE int trace_var_sys_kill(void* ctx, int tpid, int sig)
+{
+	if ((KILL_SIGNALS & (1ULL << sig)) == 0)
+		return 0;
+
+	u32 spid = get_userspace_pid();
+	struct var_kill_data_arr_t* arr_struct = bpf_map_lookup_elem(&var_tpid_to_data, &tpid);
+
+	if (arr_struct == NULL) {
+		struct var_kill_data_t* kill_data = get_var_kill_data(ctx, spid, tpid, sig);
+		int zero = 0;
+
+		if (kill_data == NULL)
+			return 0;
+		arr_struct = bpf_map_lookup_elem(&data_heap, &zero);
+		if (arr_struct == NULL)
+			return 0;
+		bpf_probe_read(&arr_struct->array[0], sizeof(arr_struct->array[0]), kill_data);
+	} else {
+		int index = get_var_spid_index(arr_struct, spid);
+
+		if (index == -1) {
+			struct var_kill_data_t* kill_data =
+				get_var_kill_data(ctx, spid, tpid, sig);
+			if (kill_data == NULL)
+				return 0;
+#ifdef UNROLL
+#pragma unroll
+#endif
+			for (int i = 0; i < ARRAY_SIZE(arr_struct->array); i++)
+				if (arr_struct->array[i].meta.pid == 0) {
+					bpf_probe_read(&arr_struct->array[i],
+						       sizeof(arr_struct->array[i]), kill_data);
+					bpf_map_update_elem(&var_tpid_to_data, &tpid,
+							    arr_struct, 0);
+
+					return 0;
+				}
+			return 0;
+		}
+
+		struct var_kill_data_t* kill_data = &arr_struct->array[index];
+
+		u64 delta_sec =
+			(bpf_ktime_get_ns() - kill_data->last_kill_time) / 1000000000;
+
+		if (delta_sec < STALE_INFO) {
+			kill_data->kill_count++;
+			kill_data->last_kill_time = bpf_ktime_get_ns();
+			bpf_probe_read(&arr_struct->array[index],
+				       sizeof(arr_struct->array[index]),
+				       kill_data);
+		} else {
+			struct var_kill_data_t* kill_data =
+				get_var_kill_data(ctx, spid, tpid, sig);
+			if (kill_data == NULL)
+				return 0;
+			bpf_probe_read(&arr_struct->array[index],
+				       sizeof(arr_struct->array[index]),
+				       kill_data);
+		}
+	}
+	bpf_map_update_elem(&var_tpid_to_data, &tpid, arr_struct, 0);
+	return 0;
+}
+
+static INLINE void bpf_stats_enter(struct bpf_func_stats_ctx* bpf_stat_ctx,
+				   enum bpf_function_id func_id)
+{
+	int func_id_key = func_id;
+
+	bpf_stat_ctx->start_time_ns = bpf_ktime_get_ns();
+	bpf_stat_ctx->bpf_func_stats_data_val =
+		bpf_map_lookup_elem(&bpf_func_stats, &func_id_key);
+	if (bpf_stat_ctx->bpf_func_stats_data_val)
+		bpf_stat_ctx->bpf_func_stats_data_val->num_executions++;
+}
+
+static INLINE void bpf_stats_exit(struct bpf_func_stats_ctx* bpf_stat_ctx)
+{
+	if (bpf_stat_ctx->bpf_func_stats_data_val)
+		bpf_stat_ctx->bpf_func_stats_data_val->time_elapsed_ns +=
+			bpf_ktime_get_ns() - bpf_stat_ctx->start_time_ns;
+}
+
+static INLINE void
+bpf_stats_pre_submit_var_perf_event(struct bpf_func_stats_ctx* bpf_stat_ctx,
+				    struct var_metadata_t* meta)
+{
+	if (bpf_stat_ctx->bpf_func_stats_data_val) {
+		bpf_stat_ctx->bpf_func_stats_data_val->num_perf_events++;
+		meta->bpf_stats_num_perf_events =
+			bpf_stat_ctx->bpf_func_stats_data_val->num_perf_events;
+	}
+	meta->bpf_stats_start_ktime_ns = bpf_stat_ctx->start_time_ns;
+	meta->cpu_id = bpf_get_smp_processor_id();
+}
+
+static INLINE size_t
+read_absolute_file_path_from_dentry(struct dentry* filp_dentry, void* payload)
+{
+	size_t length = 0;
+	size_t filepart_length;
+	struct dentry* parent_dentry;
+
+#ifdef UNROLL
+#pragma unroll
+#endif
+	for (int i = 0; i < MAX_PATH_DEPTH; i++) {
+		filepart_length = bpf_probe_read_str(payload, MAX_PATH,
+						     BPF_CORE_READ(filp_dentry, d_name.name));
+		barrier_var(filepart_length);
+		if (filepart_length > MAX_PATH)
+			break;
+		barrier_var(filepart_length);
+		payload += filepart_length;
+		length += filepart_length;
+
+		parent_dentry = BPF_CORE_READ(filp_dentry, d_parent);
+		if (filp_dentry == parent_dentry)
+			break;
+		filp_dentry = parent_dentry;
+	}
+
+	return length;
+}
+
+static INLINE bool
+is_ancestor_in_allowed_inodes(struct dentry* filp_dentry)
+{
+	struct dentry* parent_dentry;
+#ifdef UNROLL
+#pragma unroll
+#endif
+	for (int i = 0; i < MAX_PATH_DEPTH; i++) {
+		u64 dir_ino = BPF_CORE_READ(filp_dentry, d_inode, i_ino);
+		bool* allowed_dir = bpf_map_lookup_elem(&allowed_directory_inodes, &dir_ino);
+
+		if (allowed_dir != NULL)
+			return true;
+		parent_dentry = BPF_CORE_READ(filp_dentry, d_parent);
+		if (filp_dentry == parent_dentry)
+			break;
+		filp_dentry = parent_dentry;
+	}
+	return false;
+}
+
+static INLINE bool is_dentry_allowed_for_filemod(struct dentry* file_dentry,
+						 u32* device_id,
+						 u64* file_ino)
+{
+	u32 dev_id = BPF_CORE_READ(file_dentry, d_sb, s_dev);
+	*device_id = dev_id;
+	bool* allowed_device = bpf_map_lookup_elem(&allowed_devices, &dev_id);
+
+	if (allowed_device == NULL)
+		return false;
+
+	u64 ino = BPF_CORE_READ(file_dentry, d_inode, i_ino);
+	*file_ino = ino;
+	bool* allowed_file = bpf_map_lookup_elem(&allowed_file_inodes, &ino);
+
+	if (allowed_file == NULL)
+		if (!is_ancestor_in_allowed_inodes(BPF_CORE_READ(file_dentry, d_parent)))
+			return false;
+	return true;
+}
+
+SEC("kprobe/proc_sys_write")
+ssize_t BPF_KPROBE(kprobe__proc_sys_write,
+		   struct file* filp, const char* buf,
+		   size_t count, loff_t* ppos)
+{
+	struct bpf_func_stats_ctx stats_ctx;
+	bpf_stats_enter(&stats_ctx, profiler_bpf_proc_sys_write);
+
+	u32 pid = get_userspace_pid();
+	int zero = 0;
+	struct var_sysctl_data_t* sysctl_data =
+		bpf_map_lookup_elem(&data_heap, &zero);
+	if (!sysctl_data)
+		goto out;
+
+	struct task_struct* task = (struct task_struct*)bpf_get_current_task();
+	sysctl_data->meta.type = SYSCTL_EVENT;
+	void* payload = populate_var_metadata(&sysctl_data->meta, task, pid, sysctl_data->payload);
+	payload = populate_cgroup_info(&sysctl_data->cgroup_data, task, payload);
+
+	populate_ancestors(task, &sysctl_data->ancestors_info);
+
+	sysctl_data->sysctl_val_length = 0;
+	sysctl_data->sysctl_path_length = 0;
+
+	size_t sysctl_val_length = bpf_probe_read_str(payload, CTL_MAXNAME, buf);
+	barrier_var(sysctl_val_length);
+	if (sysctl_val_length <= CTL_MAXNAME) {
+		barrier_var(sysctl_val_length);
+		sysctl_data->sysctl_val_length = sysctl_val_length;
+		payload += sysctl_val_length;
+	}
+
+	size_t sysctl_path_length = bpf_probe_read_str(payload, MAX_PATH,
+						       BPF_CORE_READ(filp, f_path.dentry, d_name.name));
+	barrier_var(sysctl_path_length);
+	if (sysctl_path_length <= MAX_PATH) {
+		barrier_var(sysctl_path_length);
+		sysctl_data->sysctl_path_length = sysctl_path_length;
+		payload += sysctl_path_length;
+	}
+
+	bpf_stats_pre_submit_var_perf_event(&stats_ctx, &sysctl_data->meta);
+	unsigned long data_len = payload - (void*)sysctl_data;
+	data_len = data_len > sizeof(struct var_sysctl_data_t)
+		? sizeof(struct var_sysctl_data_t)
+		: data_len;
+	bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, sysctl_data, data_len);
+out:
+	bpf_stats_exit(&stats_ctx);
+	return 0;
+}
+
+SEC("tracepoint/syscalls/sys_enter_kill")
+int tracepoint__syscalls__sys_enter_kill(struct trace_event_raw_sys_enter* ctx)
+{
+	struct bpf_func_stats_ctx stats_ctx;
+
+	bpf_stats_enter(&stats_ctx, profiler_bpf_sys_enter_kill);
+	int pid = ctx->args[0];
+	int sig = ctx->args[1];
+	int ret = trace_var_sys_kill(ctx, pid, sig);
+	bpf_stats_exit(&stats_ctx);
+	return ret;
+};
+
+SEC("raw_tracepoint/sched_process_exit")
+int raw_tracepoint__sched_process_exit(void* ctx)
+{
+	int zero = 0;
+	struct bpf_func_stats_ctx stats_ctx;
+	bpf_stats_enter(&stats_ctx, profiler_bpf_sched_process_exit);
+
+	u32 tpid = get_userspace_pid();
+
+	struct var_kill_data_arr_t* arr_struct = bpf_map_lookup_elem(&var_tpid_to_data, &tpid);
+	struct var_kill_data_t* kill_data = bpf_map_lookup_elem(&data_heap, &zero);
+
+	if (arr_struct == NULL || kill_data == NULL)
+		goto out;
+
+	struct task_struct* task = (struct task_struct*)bpf_get_current_task();
+	struct kernfs_node* proc_kernfs = BPF_CORE_READ(task, cgroups, dfl_cgrp, kn);
+
+#ifdef UNROLL
+#pragma unroll
+#endif
+	for (int i = 0; i < ARRAY_SIZE(arr_struct->array); i++) {
+		struct var_kill_data_t* past_kill_data = &arr_struct->array[i];
+
+		if (past_kill_data != NULL && past_kill_data->kill_target_pid == tpid) {
+			bpf_probe_read(kill_data, sizeof(*past_kill_data), past_kill_data);
+			void* payload = kill_data->payload;
+			size_t offset = kill_data->payload_length;
+			if (offset >= MAX_METADATA_PAYLOAD_LEN + MAX_CGROUP_PAYLOAD_LEN)
+				return 0;
+			payload += offset;
+
+			kill_data->kill_target_name_length = 0;
+			kill_data->kill_target_cgroup_proc_length = 0;
+
+			size_t comm_length = bpf_core_read_str(payload, TASK_COMM_LEN, &task->comm);
+			barrier_var(comm_length);
+			if (comm_length <= TASK_COMM_LEN) {
+				barrier_var(comm_length);
+				kill_data->kill_target_name_length = comm_length;
+				payload += comm_length;
+			}
+
+			size_t cgroup_proc_length = bpf_probe_read_str(payload, KILL_TARGET_LEN,
+								       BPF_CORE_READ(proc_kernfs, name));
+			barrier_var(cgroup_proc_length);
+			if (cgroup_proc_length <= KILL_TARGET_LEN) {
+				barrier_var(cgroup_proc_length);
+				kill_data->kill_target_cgroup_proc_length = cgroup_proc_length;
+				payload += cgroup_proc_length;
+			}
+
+			bpf_stats_pre_submit_var_perf_event(&stats_ctx, &kill_data->meta);
+			unsigned long data_len = (void*)payload - (void*)kill_data;
+			data_len = data_len > sizeof(struct var_kill_data_t)
+				? sizeof(struct var_kill_data_t)
+				: data_len;
+			bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, kill_data, data_len);
+		}
+	}
+	bpf_map_delete_elem(&var_tpid_to_data, &tpid);
+out:
+	bpf_stats_exit(&stats_ctx);
+	return 0;
+}
+
+SEC("raw_tracepoint/sched_process_exec")
+int raw_tracepoint__sched_process_exec(struct bpf_raw_tracepoint_args* ctx)
+{
+	struct bpf_func_stats_ctx stats_ctx;
+	bpf_stats_enter(&stats_ctx, profiler_bpf_sched_process_exec);
+
+	struct linux_binprm* bprm = (struct linux_binprm*)ctx->args[2];
+	u64 inode = BPF_CORE_READ(bprm, file, f_inode, i_ino);
+
+	bool* should_filter_binprm = bpf_map_lookup_elem(&disallowed_exec_inodes, &inode);
+	if (should_filter_binprm != NULL)
+		goto out;
+
+	int zero = 0;
+	struct var_exec_data_t* proc_exec_data = bpf_map_lookup_elem(&data_heap, &zero);
+	if (!proc_exec_data)
+		goto out;
+
+	if (INODE_FILTER && inode != INODE_FILTER)
+		return 0;
+
+	u32 pid = get_userspace_pid();
+	struct task_struct* task = (struct task_struct*)bpf_get_current_task();
+
+	proc_exec_data->meta.type = EXEC_EVENT;
+	proc_exec_data->bin_path_length = 0;
+	proc_exec_data->cmdline_length = 0;
+	proc_exec_data->environment_length = 0;
+	void* payload = populate_var_metadata(&proc_exec_data->meta, task, pid,
+					      proc_exec_data->payload);
+	payload = populate_cgroup_info(&proc_exec_data->cgroup_data, task, payload);
+
+	struct task_struct* parent_task = BPF_CORE_READ(task, real_parent);
+	proc_exec_data->parent_pid = BPF_CORE_READ(parent_task, tgid);
+	proc_exec_data->parent_uid = BPF_CORE_READ(parent_task, real_cred, uid.val);
+	proc_exec_data->parent_exec_id = BPF_CORE_READ(parent_task, self_exec_id);
+	proc_exec_data->parent_start_time = BPF_CORE_READ(parent_task, start_time);
+
+	const char* filename = BPF_CORE_READ(bprm, filename);
+	size_t bin_path_length = bpf_probe_read_str(payload, MAX_FILENAME_LEN, filename);
+	barrier_var(bin_path_length);
+	if (bin_path_length <= MAX_FILENAME_LEN) {
+		barrier_var(bin_path_length);
+		proc_exec_data->bin_path_length = bin_path_length;
+		payload += bin_path_length;
+	}
+
+	void* arg_start = (void*)BPF_CORE_READ(task, mm, arg_start);
+	void* arg_end = (void*)BPF_CORE_READ(task, mm, arg_end);
+	unsigned int cmdline_length = probe_read_lim(payload, arg_start,
+						     arg_end - arg_start, MAX_ARGS_LEN);
+
+	if (cmdline_length <= MAX_ARGS_LEN) {
+		barrier_var(cmdline_length);
+		proc_exec_data->cmdline_length = cmdline_length;
+		payload += cmdline_length;
+	}
+
+	if (READ_ENVIRON_FROM_EXEC) {
+		void* env_start = (void*)BPF_CORE_READ(task, mm, env_start);
+		void* env_end = (void*)BPF_CORE_READ(task, mm, env_end);
+		unsigned long env_len = probe_read_lim(payload, env_start,
+						       env_end - env_start, MAX_ENVIRON_LEN);
+		if (cmdline_length <= MAX_ENVIRON_LEN) {
+			proc_exec_data->environment_length = env_len;
+			payload += env_len;
+		}
+	}
+
+	bpf_stats_pre_submit_var_perf_event(&stats_ctx, &proc_exec_data->meta);
+	unsigned long data_len = payload - (void*)proc_exec_data;
+	data_len = data_len > sizeof(struct var_exec_data_t)
+		? sizeof(struct var_exec_data_t)
+		: data_len;
+	bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, proc_exec_data, data_len);
+out:
+	bpf_stats_exit(&stats_ctx);
+	return 0;
+}
+
+SEC("kretprobe/do_filp_open")
+int kprobe_ret__do_filp_open(struct pt_regs* ctx)
+{
+	struct bpf_func_stats_ctx stats_ctx;
+	bpf_stats_enter(&stats_ctx, profiler_bpf_do_filp_open_ret);
+
+	struct file* filp = (struct file*)PT_REGS_RC_CORE(ctx);
+
+	if (filp == NULL || IS_ERR(filp))
+		goto out;
+	unsigned int flags = BPF_CORE_READ(filp, f_flags);
+	if ((flags & (O_RDWR | O_WRONLY)) == 0)
+		goto out;
+	if ((flags & O_TMPFILE) > 0)
+		goto out;
+	struct inode* file_inode = BPF_CORE_READ(filp, f_inode);
+	umode_t mode = BPF_CORE_READ(file_inode, i_mode);
+	if (S_ISDIR(mode) || S_ISCHR(mode) || S_ISBLK(mode) || S_ISFIFO(mode) ||
+	    S_ISSOCK(mode))
+		goto out;
+
+	struct dentry* filp_dentry = BPF_CORE_READ(filp, f_path.dentry);
+	u32 device_id = 0;
+	u64 file_ino = 0;
+	if (!is_dentry_allowed_for_filemod(filp_dentry, &device_id, &file_ino))
+		goto out;
+
+	int zero = 0;
+	struct var_filemod_data_t* filemod_data = bpf_map_lookup_elem(&data_heap, &zero);
+	if (!filemod_data)
+		goto out;
+
+	u32 pid = get_userspace_pid();
+	struct task_struct* task = (struct task_struct*)bpf_get_current_task();
+
+	filemod_data->meta.type = FILEMOD_EVENT;
+	filemod_data->fmod_type = FMOD_OPEN;
+	filemod_data->dst_flags = flags;
+	filemod_data->src_inode = 0;
+	filemod_data->dst_inode = file_ino;
+	filemod_data->src_device_id = 0;
+	filemod_data->dst_device_id = device_id;
+	filemod_data->src_filepath_length = 0;
+	filemod_data->dst_filepath_length = 0;
+
+	void* payload = populate_var_metadata(&filemod_data->meta, task, pid,
+					      filemod_data->payload);
+	payload = populate_cgroup_info(&filemod_data->cgroup_data, task, payload);
+
+	size_t len = read_absolute_file_path_from_dentry(filp_dentry, payload);
+	barrier_var(len);
+	if (len <= MAX_FILEPATH_LENGTH) {
+		barrier_var(len);
+		payload += len;
+		filemod_data->dst_filepath_length = len;
+	}
+	bpf_stats_pre_submit_var_perf_event(&stats_ctx, &filemod_data->meta);
+	unsigned long data_len = payload - (void*)filemod_data;
+	data_len = data_len > sizeof(*filemod_data) ? sizeof(*filemod_data) : data_len;
+	bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, filemod_data, data_len);
+out:
+	bpf_stats_exit(&stats_ctx);
+	return 0;
+}
+
+SEC("kprobe/vfs_link")
+int BPF_KPROBE(kprobe__vfs_link,
+	       struct dentry* old_dentry, struct inode* dir,
+	       struct dentry* new_dentry, struct inode** delegated_inode)
+{
+	struct bpf_func_stats_ctx stats_ctx;
+	bpf_stats_enter(&stats_ctx, profiler_bpf_vfs_link);
+
+	u32 src_device_id = 0;
+	u64 src_file_ino = 0;
+	u32 dst_device_id = 0;
+	u64 dst_file_ino = 0;
+	if (!is_dentry_allowed_for_filemod(old_dentry, &src_device_id, &src_file_ino) &&
+	    !is_dentry_allowed_for_filemod(new_dentry, &dst_device_id, &dst_file_ino))
+		goto out;
+
+	int zero = 0;
+	struct var_filemod_data_t* filemod_data = bpf_map_lookup_elem(&data_heap, &zero);
+	if (!filemod_data)
+		goto out;
+
+	u32 pid = get_userspace_pid();
+	struct task_struct* task = (struct task_struct*)bpf_get_current_task();
+
+	filemod_data->meta.type = FILEMOD_EVENT;
+	filemod_data->fmod_type = FMOD_LINK;
+	filemod_data->dst_flags = 0;
+	filemod_data->src_inode = src_file_ino;
+	filemod_data->dst_inode = dst_file_ino;
+	filemod_data->src_device_id = src_device_id;
+	filemod_data->dst_device_id = dst_device_id;
+	filemod_data->src_filepath_length = 0;
+	filemod_data->dst_filepath_length = 0;
+
+	void* payload = populate_var_metadata(&filemod_data->meta, task, pid,
+					      filemod_data->payload);
+	payload = populate_cgroup_info(&filemod_data->cgroup_data, task, payload);
+
+	size_t len = read_absolute_file_path_from_dentry(old_dentry, payload);
+	barrier_var(len);
+	if (len <= MAX_FILEPATH_LENGTH) {
+		barrier_var(len);
+		payload += len;
+		filemod_data->src_filepath_length = len;
+	}
+
+	len = read_absolute_file_path_from_dentry(new_dentry, payload);
+	barrier_var(len);
+	if (len <= MAX_FILEPATH_LENGTH) {
+		barrier_var(len);
+		payload += len;
+		filemod_data->dst_filepath_length = len;
+	}
+
+	bpf_stats_pre_submit_var_perf_event(&stats_ctx, &filemod_data->meta);
+	unsigned long data_len = payload - (void*)filemod_data;
+	data_len = data_len > sizeof(*filemod_data) ? sizeof(*filemod_data) : data_len;
+	bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, filemod_data, data_len);
+out:
+	bpf_stats_exit(&stats_ctx);
+	return 0;
+}
+
+SEC("kprobe/vfs_symlink")
+int BPF_KPROBE(kprobe__vfs_symlink, struct inode* dir, struct dentry* dentry,
+	       const char* oldname)
+{
+	struct bpf_func_stats_ctx stats_ctx;
+	bpf_stats_enter(&stats_ctx, profiler_bpf_vfs_symlink);
+
+	u32 dst_device_id = 0;
+	u64 dst_file_ino = 0;
+	if (!is_dentry_allowed_for_filemod(dentry, &dst_device_id, &dst_file_ino))
+		goto out;
+
+	int zero = 0;
+	struct var_filemod_data_t* filemod_data = bpf_map_lookup_elem(&data_heap, &zero);
+	if (!filemod_data)
+		goto out;
+
+	u32 pid = get_userspace_pid();
+	struct task_struct* task = (struct task_struct*)bpf_get_current_task();
+
+	filemod_data->meta.type = FILEMOD_EVENT;
+	filemod_data->fmod_type = FMOD_SYMLINK;
+	filemod_data->dst_flags = 0;
+	filemod_data->src_inode = 0;
+	filemod_data->dst_inode = dst_file_ino;
+	filemod_data->src_device_id = 0;
+	filemod_data->dst_device_id = dst_device_id;
+	filemod_data->src_filepath_length = 0;
+	filemod_data->dst_filepath_length = 0;
+
+	void* payload = populate_var_metadata(&filemod_data->meta, task, pid,
+					      filemod_data->payload);
+	payload = populate_cgroup_info(&filemod_data->cgroup_data, task, payload);
+
+	size_t len = bpf_probe_read_str(payload, MAX_FILEPATH_LENGTH, oldname);
+	barrier_var(len);
+	if (len <= MAX_FILEPATH_LENGTH) {
+		barrier_var(len);
+		payload += len;
+		filemod_data->src_filepath_length = len;
+	}
+	len = read_absolute_file_path_from_dentry(dentry, payload);
+	barrier_var(len);
+	if (len <= MAX_FILEPATH_LENGTH) {
+		barrier_var(len);
+		payload += len;
+		filemod_data->dst_filepath_length = len;
+	}
+	bpf_stats_pre_submit_var_perf_event(&stats_ctx, &filemod_data->meta);
+	unsigned long data_len = payload - (void*)filemod_data;
+	data_len = data_len > sizeof(*filemod_data) ? sizeof(*filemod_data) : data_len;
+	bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, filemod_data, data_len);
+out:
+	bpf_stats_exit(&stats_ctx);
+	return 0;
+}
+
+SEC("raw_tracepoint/sched_process_fork")
+int raw_tracepoint__sched_process_fork(struct bpf_raw_tracepoint_args* ctx)
+{
+	struct bpf_func_stats_ctx stats_ctx;
+	bpf_stats_enter(&stats_ctx, profiler_bpf_sched_process_fork);
+
+	int zero = 0;
+	struct var_fork_data_t* fork_data = bpf_map_lookup_elem(&data_heap, &zero);
+	if (!fork_data)
+		goto out;
+
+	struct task_struct* parent = (struct task_struct*)ctx->args[0];
+	struct task_struct* child = (struct task_struct*)ctx->args[1];
+	fork_data->meta.type = FORK_EVENT;
+
+	void* payload = populate_var_metadata(&fork_data->meta, child,
+					      BPF_CORE_READ(child, pid), fork_data->payload);
+	fork_data->parent_pid = BPF_CORE_READ(parent, pid);
+	fork_data->parent_exec_id = BPF_CORE_READ(parent, self_exec_id);
+	fork_data->parent_start_time = BPF_CORE_READ(parent, start_time);
+	bpf_stats_pre_submit_var_perf_event(&stats_ctx, &fork_data->meta);
+
+	unsigned long data_len = payload - (void*)fork_data;
+	data_len = data_len > sizeof(*fork_data) ? sizeof(*fork_data) : data_len;
+	bpf_perf_event_output(ctx, &events, BPF_F_CURRENT_CPU, fork_data, data_len);
+out:
+	bpf_stats_exit(&stats_ctx);
+	return 0;
+}
+char _license[] SEC("license") = "GPL";
diff --git a/tools/testing/selftests/bpf/progs/profiler1.c b/tools/testing/selftests/bpf/progs/profiler1.c
new file mode 100644
index 000000000000..4df9088bfc00
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/profiler1.c
@@ -0,0 +1,6 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#define barrier_var(var) asm volatile("" : "=r"(var) : "0"(var))
+#define UNROLL
+#define INLINE __always_inline
+#include "profiler.inc.h"
diff --git a/tools/testing/selftests/bpf/progs/profiler2.c b/tools/testing/selftests/bpf/progs/profiler2.c
new file mode 100644
index 000000000000..0f32a3cbf556
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/profiler2.c
@@ -0,0 +1,6 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#define barrier_var(var) /**/
+/* undef #define UNROLL */
+#define INLINE /**/
+#include "profiler.inc.h"
diff --git a/tools/testing/selftests/bpf/progs/profiler3.c b/tools/testing/selftests/bpf/progs/profiler3.c
new file mode 100644
index 000000000000..6249fc31ccb0
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/profiler3.c
@@ -0,0 +1,6 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#define barrier_var(var) /**/
+#define UNROLL
+#define INLINE __noinline
+#include "profiler.inc.h"
-- 
2.23.0

[PATCH v2 bpf-next 4/4] selftests/bpf: Asm tests for the verifier regalloc tracking.

[Alexei Starovoitov]

From: Alexei Starovoitov <ast@kernel.org>

Add asm tests for register allocator tracking logic.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
---
 .../testing/selftests/bpf/verifier/regalloc.c | 243 ++++++++++++++++++
 1 file changed, 243 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/verifier/regalloc.c

diff --git a/tools/testing/selftests/bpf/verifier/regalloc.c b/tools/testing/selftests/bpf/verifier/regalloc.c
new file mode 100644
index 000000000000..ac71b824f97a
--- /dev/null
+++ b/tools/testing/selftests/bpf/verifier/regalloc.c
@@ -0,0 +1,243 @@
+{
+	"regalloc basic",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 8),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_0, 20, 4),
+	BPF_JMP_IMM(BPF_JSLT, BPF_REG_2, 0, 3),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_0),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_2),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = ACCEPT,
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc negative",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 8),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_0, 24, 4),
+	BPF_JMP_IMM(BPF_JSLT, BPF_REG_2, 0, 3),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_0),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_2),
+	BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = REJECT,
+	.errstr = "invalid access to map value, value_size=48 off=48 size=1",
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc src_reg mark",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 9),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_0, 20, 5),
+	BPF_MOV64_IMM(BPF_REG_3, 0),
+	BPF_JMP_REG(BPF_JSGE, BPF_REG_3, BPF_REG_2, 3),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_0),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_2),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = ACCEPT,
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc src_reg negative",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 9),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_0, 22, 5),
+	BPF_MOV64_IMM(BPF_REG_3, 0),
+	BPF_JMP_REG(BPF_JSGE, BPF_REG_3, BPF_REG_2, 3),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_0),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_2),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = REJECT,
+	.errstr = "invalid access to map value, value_size=48 off=44 size=8",
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc and spill",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 11),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_0, 20, 7),
+	/* r0 has upper bound that should propagate into r2 */
+	BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_2, -8), /* spill r2 */
+	BPF_MOV64_IMM(BPF_REG_0, 0),
+	BPF_MOV64_IMM(BPF_REG_2, 0), /* clear r0 and r2 */
+	BPF_LDX_MEM(BPF_DW, BPF_REG_3, BPF_REG_10, -8), /* fill r3 */
+	BPF_JMP_REG(BPF_JSGE, BPF_REG_0, BPF_REG_3, 2),
+	/* r3 has lower and upper bounds */
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_3),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = ACCEPT,
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc and spill negative",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 11),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_0, 48, 7),
+	/* r0 has upper bound that should propagate into r2 */
+	BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_2, -8), /* spill r2 */
+	BPF_MOV64_IMM(BPF_REG_0, 0),
+	BPF_MOV64_IMM(BPF_REG_2, 0), /* clear r0 and r2 */
+	BPF_LDX_MEM(BPF_DW, BPF_REG_3, BPF_REG_10, -8), /* fill r3 */
+	BPF_JMP_REG(BPF_JSGE, BPF_REG_0, BPF_REG_3, 2),
+	/* r3 has lower and upper bounds */
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_3),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = REJECT,
+	.errstr = "invalid access to map value, value_size=48 off=48 size=8",
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc three regs",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 10),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_MOV64_REG(BPF_REG_4, BPF_REG_2),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_0, 12, 5),
+	BPF_JMP_IMM(BPF_JSLT, BPF_REG_2, 0, 4),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_0),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_2),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_4),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = ACCEPT,
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc after call",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 10),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_8, BPF_REG_0),
+	BPF_MOV64_REG(BPF_REG_9, BPF_REG_0),
+	BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 1, 0, 6),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_8, 20, 4),
+	BPF_JMP_IMM(BPF_JSLT, BPF_REG_9, 0, 3),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_8),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_7, BPF_REG_9),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_7, 0),
+	BPF_EXIT_INSN(),
+	BPF_MOV64_IMM(BPF_REG_0, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = ACCEPT,
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
+{
+	"regalloc in callee",
+	.insns = {
+	BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+	BPF_LD_MAP_FD(BPF_REG_1, 0),
+	BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
+	BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 6),
+	BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
+	BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
+	BPF_MOV64_REG(BPF_REG_3, BPF_REG_7),
+	BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 1, 0, 1),
+	BPF_EXIT_INSN(),
+	BPF_JMP_IMM(BPF_JSGT, BPF_REG_1, 20, 5),
+	BPF_JMP_IMM(BPF_JSLT, BPF_REG_2, 0, 4),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_3, BPF_REG_1),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_3, BPF_REG_2),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_3, 0),
+	BPF_EXIT_INSN(),
+	BPF_MOV64_IMM(BPF_REG_0, 0),
+	BPF_EXIT_INSN(),
+	},
+	.fixup_map_hash_48b = { 4 },
+	.result = ACCEPT,
+	.prog_type = BPF_PROG_TYPE_TRACEPOINT,
+},
-- 
2.23.0

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Yonghong Song]

On 10/8/20 6:12 PM, Alexei Starovoitov wrote:
> From: Alexei Starovoitov <ast@kernel.org>
> 
> The main purpose of the profiler test to check different llvm generation
> patterns to make sure the verifier can load these large programs.
> 
> Note that profiler.inc.h test doesn't follow strict kernel coding style.
> The code was formatted in the kernel style, but variable declarations are
> kept as-is to preserve original llvm IR pattern.
> 
> profiler1.c should pass with older and newer llvm
> 
> profiler[23].c may fail on older llvm that don't have:
> https://reviews.llvm.org/D85570

Not sure but the below equivalent URL may be more intuitive:
   https://reviews.llvm.org/D85570

> because llvm may do speculative code motion optimization that
> will generate code like this:
> 
> // r9 is a pointer to map_value
> // r7 is a scalar
> 17:       bf 96 00 00 00 00 00 00 r6 = r9
> 18:       0f 76 00 00 00 00 00 00 r6 += r7
> 19:       a5 07 01 00 01 01 00 00 if r7 < 257 goto +1
> 20:       bf 96 00 00 00 00 00 00 r6 = r9
> // r6 is used here
> 
> The verifier will reject such code with the error:
> "math between map_value pointer and register with unbounded min value is not allowed"
> At insn 18 the r7 is indeed unbounded. The later insn 19 checks the bounds and
> the insn 20 undoes map_value addition. It is currently impossible for the
> verifier to understand such speculative pointer arithmetic. Hence llvm D85570
> addresses it on the compiler side.
> 
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Acked-by: Yonghong Song <yhs@fb.com>

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Alexei Starovoitov]

On Thu, Oct 8, 2020 at 11:49 PM Yonghong Song <yhs@fb.com> wrote:
> >
> > profiler[23].c may fail on older llvm that don't have:
> > https://reviews.llvm.org/D85570
>
> Not sure but the below equivalent URL may be more intuitive:
>    https://reviews.llvm.org/D85570

It's the same URL :)
See it unmangled here:
https://patchwork.kernel.org/patch/11824813/#23675319

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Yonghong Song]

On 10/9/20 8:08 AM, Alexei Starovoitov wrote:
> On Thu, Oct 8, 2020 at 11:49 PM Yonghong Song <yhs@fb.com> wrote:
>>>
>>> profiler[23].c may fail on older llvm that don't have:
>>> https://reviews.llvm.org/D85570
>>
>> Not sure but the below equivalent URL may be more intuitive:
>>     https://reviews.llvm.org/D85570
> 
> It's the same URL :)
> See it unmangled here:
> https://patchwork.kernel.org/patch/11824813/#23675319

WoW, I never pay attention to this. Thanks for pointing it out.

RE: [PATCH v2 bpf-next 1/4] bpf: Propagate scalar ranges through register assignments.

[John Fastabend]

Alexei Starovoitov wrote:
> From: Alexei Starovoitov <ast@kernel.org>
> 
> The llvm register allocator may use two different registers representing the
> same virtual register. In such case the following pattern can be observed:
> 1047: (bf) r9 = r6
> 1048: (a5) if r6 < 0x1000 goto pc+1
> 1050: ...
> 1051: (a5) if r9 < 0x2 goto pc+66
> 1052: ...
> 1053: (bf) r2 = r9 /* r2 needs to have upper and lower bounds */
> 
> This is normal behavior of greedy register allocator.
> The slides 137+ explain why regalloc introduces such register copy:
> http://llvm.org/devmtg/2018-04/slides/Yatsina-LLVM%20Greedy%20Register%20Allocator.pdf
> There is no way to tell llvm 'not to do this'.
> Hence the verifier has to recognize such patterns.
> 
> In order to track this information without backtracking allocate ID
> for scalars in a similar way as it's done for find_good_pkt_pointers().
> 
> When the verifier encounters r9 = r6 assignment it will assign the same ID
> to both registers. Later if either register range is narrowed via conditional
> jump propagate the register state into the other register.
> 
> Clear register ID in adjust_reg_min_max_vals() for any alu instruction. The
> register ID is ignored for scalars in regsafe() and doesn't affect state
> pruning. mark_reg_unknown() clears the ID. It's used to process call, endian
> and other instructions. Hence ID is explicitly cleared only in
> adjust_reg_min_max_vals() and in 32-bit mov.
> 
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
> Acked-by: Andrii Nakryiko <andrii@kernel.org>
> ---
>  kernel/bpf/verifier.c                         | 50 +++++++++++++++++++
>  .../testing/selftests/bpf/prog_tests/align.c  | 16 +++---
>  .../bpf/verifier/direct_packet_access.c       |  2 +-
>  3 files changed, 59 insertions(+), 9 deletions(-)
> 

I also walked through the stack read case and I don't see any issues
there either. We will run check_mem_access on the ld with src_reg->type set
to PTR_TO_STACK. So we run through check_stack_read() with a value_regno
set to the dst_reg (>=0). The case I was thinking of was if we had a
size != BPF_REG_SIZE. But, this case is handled by marking the dst reg
unknown and never copying over the stack[spi]. If size == BPF_REG_SIZE
its a full read and we also ensure all stype[] are STACK_SPILL. Then
everything looks safe to do state->regs[value_reno] = *reg e.g. assign
register state from stack complete with reg.id. 

Looking the other way at stack writes. We only allow SCALAR type bounds to
be saved on the stack if they are constant and size = BPF_REG_SIZE so no
problems there.

LGTM!

Acked-by: John Fastabend <john.fastabend@gmail.com>

RE: [PATCH v2 bpf-next 2/4] bpf: Track spill/fill of bounded scalars.

[John Fastabend]

Alexei Starovoitov wrote:
> From: Yonghong Song <yhs@fb.com>
> 
> Under register pressure the llvm may spill registers with bounds into the stack.
> The verifier has to track them through spill/fill otherwise many kinds of bound
> errors will be seen. The spill/fill of induction variables was already
> happening. This patch extends this logic from tracking spill/fill of a constant
> into any bounded register. There is no need to track spill/fill of unbounded,
> since no new information will be retrieved from the stack during register fill.
> 
> Though extra stack difference could cause state pruning to be less effective, no
> adverse affects were seen from this patch on selftests and on cilium programs.
> 
> Signed-off-by: Yonghong Song <yhs@fb.com>
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
> Acked-by: Andrii Nakryiko <andrii@kernel.org>
> ---
>  kernel/bpf/verifier.c | 16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)

LGTM and will be useful.

Acked-by: John Fastabend <john.fastabend@gmail.com>

RE: [PATCH v2 bpf-next 4/4] selftests/bpf: Asm tests for the verifier regalloc tracking.

[John Fastabend]

Alexei Starovoitov wrote:
> From: Alexei Starovoitov <ast@kernel.org>
> 
> Add asm tests for register allocator tracking logic.
> 
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
> Acked-by: Andrii Nakryiko <andrii@kernel.org>
> ---
>  .../testing/selftests/bpf/verifier/regalloc.c | 243 ++++++++++++++++++
>  1 file changed, 243 insertions(+)
>  create mode 100644 tools/testing/selftests/bpf/verifier/regalloc.c
> 

I'm writing some extra tests now for a few things so I'll probably also add
some to track the bounds for non-const through stack spill today or
monday.

Acked-by: John Fastabend <john.fastabend@gmail.com>

Re: [PATCH v2 bpf-next 0/4] bpf: Make the verifier recognize llvm register allocation patterns.

[patchwork-bot+netdevbpf]

Hello:

This series was applied to bpf/bpf-next.git (refs/heads/master):

On Thu,  8 Oct 2020 18:12:36 -0700 you wrote:
> From: Alexei Starovoitov <ast@kernel.org>
> 
> v1->v2:
> - fixed 32-bit mov issue spotted by John.
> - allowed r2=r1; r3=r2; sequence as suggested by John.
> - added comments, acks, more tests.
> 
> [...]

Here is the summary with links:
  - [v2,bpf-next,1/4] bpf: Propagate scalar ranges through register assignments.
    https://git.kernel.org/bpf/bpf-next/c/75748837b7e5
  - [v2,bpf-next,2/4] bpf: Track spill/fill of bounded scalars.
    https://git.kernel.org/bpf/bpf-next/c/5689d49b71ad
  - [v2,bpf-next,3/4] selftests/bpf: Add profiler test
    https://git.kernel.org/bpf/bpf-next/c/03d4d13fab3f
  - [v2,bpf-next,4/4] selftests/bpf: Asm tests for the verifier regalloc tracking.
    https://git.kernel.org/bpf/bpf-next/c/54fada41e8a1

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Jiri Olsa]

On Thu, Oct 08, 2020 at 06:12:39PM -0700, Alexei Starovoitov wrote:

SNIP

> +
> +#ifdef UNROLL
> +#pragma unroll
> +#endif
> +	for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {
> +		filepart_length =
> +			bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(cgroup_node, name));
> +		if (!cgroup_node)
> +			return payload;
> +		if (cgroup_node == cgroup_root_node)
> +			*root_pos = payload - payload_start;
> +		if (filepart_length <= MAX_PATH) {
> +			barrier_var(filepart_length);
> +			payload += filepart_length;
> +		}
> +		cgroup_node = BPF_CORE_READ(cgroup_node, parent);
> +	}
> +	return payload;
> +}
> +
> +static ino_t get_inode_from_kernfs(struct kernfs_node* node)
> +{
> +	struct kernfs_node___52* node52 = (void*)node;
> +
> +	if (bpf_core_field_exists(node52->id.ino)) {
> +		barrier_var(node52);
> +		return BPF_CORE_READ(node52, id.ino);
> +	} else {
> +		barrier_var(node);
> +		return (u64)BPF_CORE_READ(node, id);
> +	}
> +}
> +
> +int pids_cgrp_id = 1;


hi,
I'm getting compilation failure with this:

	  CLNG-LLC [test_maps] profiler2.o
	In file included from progs/profiler2.c:6:
	progs/profiler.inc.h:246:5: error: redefinition of 'pids_cgrp_id' as different kind of symbol
	int pids_cgrp_id = 1;
	    ^
	/home/jolsa/linux-qemu/tools/testing/selftests/bpf/tools/include/vmlinux.h:14531:2: note: previous definition is here
		pids_cgrp_id = 11,
		^
	1 error generated.
	/opt/clang/bin/llc: error: /opt/clang/bin/llc: <stdin>:1:1: error: expected top-level entity
	BPF obj compilation failed
	^
	make: *** [Makefile:396: /home/jolsa/linux-qemu/tools/testing/selftests/bpf/profiler2.o] Error 1

jirka

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Alexei Starovoitov]

On Tue, Oct 13, 2020 at 12:56 PM Jiri Olsa <jolsa@redhat.com> wrote:
>
> On Thu, Oct 08, 2020 at 06:12:39PM -0700, Alexei Starovoitov wrote:
>
> SNIP
>
> > +
> > +#ifdef UNROLL
> > +#pragma unroll
> > +#endif
> > +     for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {
> > +             filepart_length =
> > +                     bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(cgroup_node, name));
> > +             if (!cgroup_node)
> > +                     return payload;
> > +             if (cgroup_node == cgroup_root_node)
> > +                     *root_pos = payload - payload_start;
> > +             if (filepart_length <= MAX_PATH) {
> > +                     barrier_var(filepart_length);
> > +                     payload += filepart_length;
> > +             }
> > +             cgroup_node = BPF_CORE_READ(cgroup_node, parent);
> > +     }
> > +     return payload;
> > +}
> > +
> > +static ino_t get_inode_from_kernfs(struct kernfs_node* node)
> > +{
> > +     struct kernfs_node___52* node52 = (void*)node;
> > +
> > +     if (bpf_core_field_exists(node52->id.ino)) {
> > +             barrier_var(node52);
> > +             return BPF_CORE_READ(node52, id.ino);
> > +     } else {
> > +             barrier_var(node);
> > +             return (u64)BPF_CORE_READ(node, id);
> > +     }
> > +}
> > +
> > +int pids_cgrp_id = 1;
>
>
> hi,
> I'm getting compilation failure with this:
>
>           CLNG-LLC [test_maps] profiler2.o
>         In file included from progs/profiler2.c:6:
>         progs/profiler.inc.h:246:5: error: redefinition of 'pids_cgrp_id' as different kind of symbol
>         int pids_cgrp_id = 1;
>             ^
>         /home/jolsa/linux-qemu/tools/testing/selftests/bpf/tools/include/vmlinux.h:14531:2: note: previous definition is here
>                 pids_cgrp_id = 11,

Interesting.
You probably have CONFIG_CGROUP_PIDS in your .config?
I don't and bpf CI doesn't have it either, so this issue wasn't spotted earlier.

I can hard code 11, of course, but
but maybe Andrii has a cool way to use co-re to deal with this?
I think
"extern bool CONFIG_CGROUP_PIDS __kconfig"
won't work.
A good opportunity to try to use bpf_core_enum_value_exists() ?

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Andrii Nakryiko]

On Tue, Oct 13, 2020 at 2:03 PM Alexei Starovoitov
<alexei.starovoitov@gmail.com> wrote:
>
> On Tue, Oct 13, 2020 at 12:56 PM Jiri Olsa <jolsa@redhat.com> wrote:
> >
> > On Thu, Oct 08, 2020 at 06:12:39PM -0700, Alexei Starovoitov wrote:
> >
> > SNIP
> >
> > > +
> > > +#ifdef UNROLL
> > > +#pragma unroll
> > > +#endif
> > > +     for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {
> > > +             filepart_length =
> > > +                     bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(cgroup_node, name));
> > > +             if (!cgroup_node)
> > > +                     return payload;
> > > +             if (cgroup_node == cgroup_root_node)
> > > +                     *root_pos = payload - payload_start;
> > > +             if (filepart_length <= MAX_PATH) {
> > > +                     barrier_var(filepart_length);
> > > +                     payload += filepart_length;
> > > +             }
> > > +             cgroup_node = BPF_CORE_READ(cgroup_node, parent);
> > > +     }
> > > +     return payload;
> > > +}
> > > +
> > > +static ino_t get_inode_from_kernfs(struct kernfs_node* node)
> > > +{
> > > +     struct kernfs_node___52* node52 = (void*)node;
> > > +
> > > +     if (bpf_core_field_exists(node52->id.ino)) {
> > > +             barrier_var(node52);
> > > +             return BPF_CORE_READ(node52, id.ino);
> > > +     } else {
> > > +             barrier_var(node);
> > > +             return (u64)BPF_CORE_READ(node, id);
> > > +     }
> > > +}
> > > +
> > > +int pids_cgrp_id = 1;
> >
> >
> > hi,
> > I'm getting compilation failure with this:
> >
> >           CLNG-LLC [test_maps] profiler2.o
> >         In file included from progs/profiler2.c:6:
> >         progs/profiler.inc.h:246:5: error: redefinition of 'pids_cgrp_id' as different kind of symbol
> >         int pids_cgrp_id = 1;
> >             ^
> >         /home/jolsa/linux-qemu/tools/testing/selftests/bpf/tools/include/vmlinux.h:14531:2: note: previous definition is here
> >                 pids_cgrp_id = 11,
>
> Interesting.
> You probably have CONFIG_CGROUP_PIDS in your .config?
> I don't and bpf CI doesn't have it either, so this issue wasn't spotted earlier.
>
> I can hard code 11, of course, but
> but maybe Andrii has a cool way to use co-re to deal with this?

Cool or not, but we do have a way to deal with it. :)

> I think
> "extern bool CONFIG_CGROUP_PIDS __kconfig"
> won't work.
> A good opportunity to try to use bpf_core_enum_value_exists() ?

There are fews parts here.

First, we can't rely that vmlinux.h has pids_cgrp_id enum defined, so
we need to define our own. But in such a way that it doesn't collide
with enum cgroup_subsys_id in vmlinux.h. ___suffix for the rescue
here:

enum cgroup_subsys_id___local {
    pids_cgrp_id___local = 1, /* anything but zero */
};

If value is zero, built-in will complain. Need to check with Yonghong
on why that is. For all things CO-RE, ___local suffix is going to be
ignored.

Second, detecting if the kernel even has this pids_cgrp_id defined.
That could be done with __kconfig as you mentioned, actually (but need
__weak as well). But one can also use bpf_core_enum_value_exists as
well. So it's either:

extern bool CONFIG_CGROUP_PIDS __kconfig __weak;

...

if (ENABLE_CGROUP_V1_RESOLVER && CONFIG_CGROUP_PIDS) {
   ...
}


Or a bit more verbose way with relos:

if (ENABLE_CGROUP_V1_RESOLVER &&
    bpf_core_enum_value_exists(enum cgroup_subsys_id___local,
pids_cgrp_id___local)) {
   ...
}

Third, actually getting the value of enum. This one would be
impossible without CO-RE reloc, but that's exactly what
bpf_core_enum_value() exists for:


int cgrp_id = bpf_core_enum_value(enum cgroup_subsys_id___local,
pids_cgrp_id___local);


I'd go with Kconfig + bpf_core_enum_value(), as it's shorter and
nicer. This compiles and works with my Kconfig, but I haven't checked
with CONFIG_CGROUP_PIDS defined.


diff --git a/tools/testing/selftests/bpf/progs/profiler.inc.h
b/tools/testing/selftests/bpf/progs/profiler.inc.h
index 00578311a423..79b8d2860a5c 100644
--- a/tools/testing/selftests/bpf/progs/profiler.inc.h
+++ b/tools/testing/selftests/bpf/progs/profiler.inc.h
@@ -243,7 +243,11 @@ static ino_t get_inode_from_kernfs(struct
kernfs_node* node)
        }
 }

-int pids_cgrp_id = 1;
+extern bool CONFIG_CGROUP_PIDS __kconfig __weak;
+
+enum cgroup_subsys_id___local {
+       pids_cgrp_id___local = 1, /* anything but zero */
+};

 static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
                                         struct task_struct* task,
@@ -253,7 +257,9 @@ static INLINE void* populate_cgroup_info(struct
cgroup_data_t* cgroup_data,
                BPF_CORE_READ(task, nsproxy, cgroup_ns, root_cset,
dfl_cgrp, kn);
        struct kernfs_node* proc_kernfs = BPF_CORE_READ(task, cgroups,
dfl_cgrp, kn);

-       if (ENABLE_CGROUP_V1_RESOLVER) {
+       if (ENABLE_CGROUP_V1_RESOLVER && CONFIG_CGROUP_PIDS) {
+               int cgrp_id = bpf_core_enum_value(enum
cgroup_subsys_id___local, pids_cgrp_id___local);
+
 #ifdef UNROLL
 #pragma unroll
 #endif
@@ -262,7 +268,7 @@ static INLINE void* populate_cgroup_info(struct
cgroup_data_t* cgroup_data,
                                BPF_CORE_READ(task, cgroups, subsys[i]);
                        if (subsys != NULL) {
                                int subsys_id = BPF_CORE_READ(subsys, ss, id);
-                               if (subsys_id == pids_cgrp_id) {
+                               if (subsys_id == cgrp_id) {
                                        proc_kernfs =
BPF_CORE_READ(subsys, cgroup, kn);
                                        root_kernfs =
BPF_CORE_READ(subsys, ss, root, kf_root, kn);
                                        break;

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Jiri Olsa]

On Tue, Oct 13, 2020 at 02:03:33PM -0700, Alexei Starovoitov wrote:
> On Tue, Oct 13, 2020 at 12:56 PM Jiri Olsa <jolsa@redhat.com> wrote:
> >
> > On Thu, Oct 08, 2020 at 06:12:39PM -0700, Alexei Starovoitov wrote:
> >
> > SNIP
> >
> > > +
> > > +#ifdef UNROLL
> > > +#pragma unroll
> > > +#endif
> > > +     for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {
> > > +             filepart_length =
> > > +                     bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(cgroup_node, name));
> > > +             if (!cgroup_node)
> > > +                     return payload;
> > > +             if (cgroup_node == cgroup_root_node)
> > > +                     *root_pos = payload - payload_start;
> > > +             if (filepart_length <= MAX_PATH) {
> > > +                     barrier_var(filepart_length);
> > > +                     payload += filepart_length;
> > > +             }
> > > +             cgroup_node = BPF_CORE_READ(cgroup_node, parent);
> > > +     }
> > > +     return payload;
> > > +}
> > > +
> > > +static ino_t get_inode_from_kernfs(struct kernfs_node* node)
> > > +{
> > > +     struct kernfs_node___52* node52 = (void*)node;
> > > +
> > > +     if (bpf_core_field_exists(node52->id.ino)) {
> > > +             barrier_var(node52);
> > > +             return BPF_CORE_READ(node52, id.ino);
> > > +     } else {
> > > +             barrier_var(node);
> > > +             return (u64)BPF_CORE_READ(node, id);
> > > +     }
> > > +}
> > > +
> > > +int pids_cgrp_id = 1;
> >
> >
> > hi,
> > I'm getting compilation failure with this:
> >
> >           CLNG-LLC [test_maps] profiler2.o
> >         In file included from progs/profiler2.c:6:
> >         progs/profiler.inc.h:246:5: error: redefinition of 'pids_cgrp_id' as different kind of symbol
> >         int pids_cgrp_id = 1;
> >             ^
> >         /home/jolsa/linux-qemu/tools/testing/selftests/bpf/tools/include/vmlinux.h:14531:2: note: previous definition is here
> >                 pids_cgrp_id = 11,
> 
> Interesting.
> You probably have CONFIG_CGROUP_PIDS in your .config?

yes

jirka


> I don't and bpf CI doesn't have it either, so this issue wasn't spotted earlier.
> 
> I can hard code 11, of course, but
> but maybe Andrii has a cool way to use co-re to deal with this?
> I think
> "extern bool CONFIG_CGROUP_PIDS __kconfig"
> won't work.
> A good opportunity to try to use bpf_core_enum_value_exists() ?
> 

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Song Liu]

On Tue, Oct 13, 2020 at 12:57 PM Jiri Olsa <jolsa@redhat.com> wrote:
>
> On Thu, Oct 08, 2020 at 06:12:39PM -0700, Alexei Starovoitov wrote:
>
> SNIP
>
> > +
> > +#ifdef UNROLL
> > +#pragma unroll
> > +#endif
> > +     for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {
> > +             filepart_length =
> > +                     bpf_probe_read_str(payload, MAX_PATH, BPF_CORE_READ(cgroup_node, name));
> > +             if (!cgroup_node)
> > +                     return payload;
> > +             if (cgroup_node == cgroup_root_node)
> > +                     *root_pos = payload - payload_start;
> > +             if (filepart_length <= MAX_PATH) {
> > +                     barrier_var(filepart_length);
> > +                     payload += filepart_length;
> > +             }
> > +             cgroup_node = BPF_CORE_READ(cgroup_node, parent);
> > +     }
> > +     return payload;
> > +}
> > +
> > +static ino_t get_inode_from_kernfs(struct kernfs_node* node)
> > +{
> > +     struct kernfs_node___52* node52 = (void*)node;
> > +
> > +     if (bpf_core_field_exists(node52->id.ino)) {
> > +             barrier_var(node52);
> > +             return BPF_CORE_READ(node52, id.ino);
> > +     } else {
> > +             barrier_var(node);
> > +             return (u64)BPF_CORE_READ(node, id);
> > +     }
> > +}
> > +
> > +int pids_cgrp_id = 1;
>
>
> hi,
> I'm getting compilation failure with this:
>
>           CLNG-LLC [test_maps] profiler2.o
>         In file included from progs/profiler2.c:6:
>         progs/profiler.inc.h:246:5: error: redefinition of 'pids_cgrp_id' as different kind of symbol
>         int pids_cgrp_id = 1;
>             ^
>         /home/jolsa/linux-qemu/tools/testing/selftests/bpf/tools/include/vmlinux.h:14531:2: note: previous definition is here
>                 pids_cgrp_id = 11,
>                 ^
>         1 error generated.
>         /opt/clang/bin/llc: error: /opt/clang/bin/llc: <stdin>:1:1: error: expected top-level entity
>         BPF obj compilation failed
>         ^
>         make: *** [Makefile:396: /home/jolsa/linux-qemu/tools/testing/selftests/bpf/profiler2.o] Error 1

I got the same error with some .config. We should fix it with something below.

I will send official patch soon.

Thanks,
Song

diff --git i/tools/testing/selftests/bpf/progs/profiler.inc.h
w/tools/testing/selftests/bpf/progs/profiler.inc.h
index 00578311a4233..54f594efccbca 100644
--- i/tools/testing/selftests/bpf/progs/profiler.inc.h
+++ w/tools/testing/selftests/bpf/progs/profiler.inc.h
@@ -243,7 +243,7 @@ static ino_t get_inode_from_kernfs(struct kernfs_node* node)
        }
 }

-int pids_cgrp_id = 1;
+int pids_cgrp_id_ = 1;

 static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
                                         struct task_struct* task,
@@ -262,7 +262,7 @@ static INLINE void* populate_cgroup_info(struct
cgroup_data_t* cgroup_data,
                                BPF_CORE_READ(task, cgroups, subsys[i]);
                        if (subsys != NULL) {
                                int subsys_id = BPF_CORE_READ(subsys, ss, id);
-                               if (subsys_id == pids_cgrp_id) {
+                               if (subsys_id == pids_cgrp_id_) {
                                        proc_kernfs =
BPF_CORE_READ(subsys, cgroup, kn);
                                        root_kernfs =
BPF_CORE_READ(subsys, ss, root, kf_root, kn);
                                        break;

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Song Liu]

> On Oct 13, 2020, at 2:56 PM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:

[...]

> 
> I'd go with Kconfig + bpf_core_enum_value(), as it's shorter and
> nicer. This compiles and works with my Kconfig, but I haven't checked
> with CONFIG_CGROUP_PIDS defined.

Tested with CONFIG_CGROUP_PIDS, it looks good. 

Tested-by: Song Liu <songliubraving@fb.com>

> 
> 
> diff --git a/tools/testing/selftests/bpf/progs/profiler.inc.h
> b/tools/testing/selftests/bpf/progs/profiler.inc.h
> index 00578311a423..79b8d2860a5c 100644
> --- a/tools/testing/selftests/bpf/progs/profiler.inc.h
> +++ b/tools/testing/selftests/bpf/progs/profiler.inc.h
> @@ -243,7 +243,11 @@ static ino_t get_inode_from_kernfs(struct
> kernfs_node* node)
>        }
> }
> 
> -int pids_cgrp_id = 1;
> +extern bool CONFIG_CGROUP_PIDS __kconfig __weak;
> +
> +enum cgroup_subsys_id___local {
> +       pids_cgrp_id___local = 1, /* anything but zero */
> +};
> 
> static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
>                                         struct task_struct* task,
> @@ -253,7 +257,9 @@ static INLINE void* populate_cgroup_info(struct
> cgroup_data_t* cgroup_data,
>                BPF_CORE_READ(task, nsproxy, cgroup_ns, root_cset,
> dfl_cgrp, kn);
>        struct kernfs_node* proc_kernfs = BPF_CORE_READ(task, cgroups,
> dfl_cgrp, kn);
> 
> -       if (ENABLE_CGROUP_V1_RESOLVER) {
> +       if (ENABLE_CGROUP_V1_RESOLVER && CONFIG_CGROUP_PIDS) {
> +               int cgrp_id = bpf_core_enum_value(enum
> cgroup_subsys_id___local, pids_cgrp_id___local);
> +
> #ifdef UNROLL
> #pragma unroll
> #endif
> @@ -262,7 +268,7 @@ static INLINE void* populate_cgroup_info(struct
> cgroup_data_t* cgroup_data,
>                                BPF_CORE_READ(task, cgroups, subsys[i]);
>                        if (subsys != NULL) {
>                                int subsys_id = BPF_CORE_READ(subsys, ss, id);
> -                               if (subsys_id == pids_cgrp_id) {
> +                               if (subsys_id == cgrp_id) {
>                                        proc_kernfs =
> BPF_CORE_READ(subsys, cgroup, kn);
>                                        root_kernfs =
> BPF_CORE_READ(subsys, ss, root, kf_root, kn);
>                                        break;

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Jiri Olsa]

On Thu, Oct 15, 2020 at 06:09:14AM +0000, Song Liu wrote:
> 
> 
> > On Oct 13, 2020, at 2:56 PM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> 
> [...]
> 
> > 
> > I'd go with Kconfig + bpf_core_enum_value(), as it's shorter and
> > nicer. This compiles and works with my Kconfig, but I haven't checked
> > with CONFIG_CGROUP_PIDS defined.
> 
> Tested with CONFIG_CGROUP_PIDS, it looks good. 
> 
> Tested-by: Song Liu <songliubraving@fb.com>

hi,
I still need to apply my workaround to compile tests,
so I wonder this fell through cracks

thanks,
jirka

> 
> > 
> > 
> > diff --git a/tools/testing/selftests/bpf/progs/profiler.inc.h
> > b/tools/testing/selftests/bpf/progs/profiler.inc.h
> > index 00578311a423..79b8d2860a5c 100644
> > --- a/tools/testing/selftests/bpf/progs/profiler.inc.h
> > +++ b/tools/testing/selftests/bpf/progs/profiler.inc.h
> > @@ -243,7 +243,11 @@ static ino_t get_inode_from_kernfs(struct
> > kernfs_node* node)
> >        }
> > }
> > 
> > -int pids_cgrp_id = 1;
> > +extern bool CONFIG_CGROUP_PIDS __kconfig __weak;
> > +
> > +enum cgroup_subsys_id___local {
> > +       pids_cgrp_id___local = 1, /* anything but zero */
> > +};
> > 
> > static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
> >                                         struct task_struct* task,
> > @@ -253,7 +257,9 @@ static INLINE void* populate_cgroup_info(struct
> > cgroup_data_t* cgroup_data,
> >                BPF_CORE_READ(task, nsproxy, cgroup_ns, root_cset,
> > dfl_cgrp, kn);
> >        struct kernfs_node* proc_kernfs = BPF_CORE_READ(task, cgroups,
> > dfl_cgrp, kn);
> > 
> > -       if (ENABLE_CGROUP_V1_RESOLVER) {
> > +       if (ENABLE_CGROUP_V1_RESOLVER && CONFIG_CGROUP_PIDS) {
> > +               int cgrp_id = bpf_core_enum_value(enum
> > cgroup_subsys_id___local, pids_cgrp_id___local);
> > +
> > #ifdef UNROLL
> > #pragma unroll
> > #endif
> > @@ -262,7 +268,7 @@ static INLINE void* populate_cgroup_info(struct
> > cgroup_data_t* cgroup_data,
> >                                BPF_CORE_READ(task, cgroups, subsys[i]);
> >                        if (subsys != NULL) {
> >                                int subsys_id = BPF_CORE_READ(subsys, ss, id);
> > -                               if (subsys_id == pids_cgrp_id) {
> > +                               if (subsys_id == cgrp_id) {
> >                                        proc_kernfs =
> > BPF_CORE_READ(subsys, cgroup, kn);
> >                                        root_kernfs =
> > BPF_CORE_READ(subsys, ss, root, kf_root, kn);
> >                                        break;
> 

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Andrii Nakryiko]

On Wed, Nov 4, 2020 at 8:46 AM Jiri Olsa <jolsa@redhat.com> wrote:
>
> On Thu, Oct 15, 2020 at 06:09:14AM +0000, Song Liu wrote:
> >
> >
> > > On Oct 13, 2020, at 2:56 PM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> >
> > [...]
> >
> > >
> > > I'd go with Kconfig + bpf_core_enum_value(), as it's shorter and
> > > nicer. This compiles and works with my Kconfig, but I haven't checked
> > > with CONFIG_CGROUP_PIDS defined.
> >
> > Tested with CONFIG_CGROUP_PIDS, it looks good.
> >
> > Tested-by: Song Liu <songliubraving@fb.com>
>
> hi,
> I still need to apply my workaround to compile tests,
> so I wonder this fell through cracks


The fix was already applied ([0]). Do you still see issues?

  [0] https://patchwork.kernel.org/project/netdevbpf/patch/20201022202739.3667367-1-andrii@kernel.org/

>
> thanks,
> jirka
>
> >
> > >
> > >
> > > diff --git a/tools/testing/selftests/bpf/progs/profiler.inc.h
> > > b/tools/testing/selftests/bpf/progs/profiler.inc.h
> > > index 00578311a423..79b8d2860a5c 100644
> > > --- a/tools/testing/selftests/bpf/progs/profiler.inc.h
> > > +++ b/tools/testing/selftests/bpf/progs/profiler.inc.h
> > > @@ -243,7 +243,11 @@ static ino_t get_inode_from_kernfs(struct
> > > kernfs_node* node)
> > >        }
> > > }
> > >
> > > -int pids_cgrp_id = 1;
> > > +extern bool CONFIG_CGROUP_PIDS __kconfig __weak;
> > > +
> > > +enum cgroup_subsys_id___local {
> > > +       pids_cgrp_id___local = 1, /* anything but zero */
> > > +};
> > >
> > > static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
> > >                                         struct task_struct* task,
> > > @@ -253,7 +257,9 @@ static INLINE void* populate_cgroup_info(struct
> > > cgroup_data_t* cgroup_data,
> > >                BPF_CORE_READ(task, nsproxy, cgroup_ns, root_cset,
> > > dfl_cgrp, kn);
> > >        struct kernfs_node* proc_kernfs = BPF_CORE_READ(task, cgroups,
> > > dfl_cgrp, kn);
> > >
> > > -       if (ENABLE_CGROUP_V1_RESOLVER) {
> > > +       if (ENABLE_CGROUP_V1_RESOLVER && CONFIG_CGROUP_PIDS) {
> > > +               int cgrp_id = bpf_core_enum_value(enum
> > > cgroup_subsys_id___local, pids_cgrp_id___local);
> > > +
> > > #ifdef UNROLL
> > > #pragma unroll
> > > #endif
> > > @@ -262,7 +268,7 @@ static INLINE void* populate_cgroup_info(struct
> > > cgroup_data_t* cgroup_data,
> > >                                BPF_CORE_READ(task, cgroups, subsys[i]);
> > >                        if (subsys != NULL) {
> > >                                int subsys_id = BPF_CORE_READ(subsys, ss, id);
> > > -                               if (subsys_id == pids_cgrp_id) {
> > > +                               if (subsys_id == cgrp_id) {
> > >                                        proc_kernfs =
> > > BPF_CORE_READ(subsys, cgroup, kn);
> > >                                        root_kernfs =
> > > BPF_CORE_READ(subsys, ss, root, kf_root, kn);
> > >                                        break;
> >
>

Re: [PATCH v2 bpf-next 3/4] selftests/bpf: Add profiler test

[Jiri Olsa]

On Wed, Nov 04, 2020 at 12:50:25PM -0800, Andrii Nakryiko wrote:
> On Wed, Nov 4, 2020 at 8:46 AM Jiri Olsa <jolsa@redhat.com> wrote:
> >
> > On Thu, Oct 15, 2020 at 06:09:14AM +0000, Song Liu wrote:
> > >
> > >
> > > > On Oct 13, 2020, at 2:56 PM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> > >
> > > [...]
> > >
> > > >
> > > > I'd go with Kconfig + bpf_core_enum_value(), as it's shorter and
> > > > nicer. This compiles and works with my Kconfig, but I haven't checked
> > > > with CONFIG_CGROUP_PIDS defined.
> > >
> > > Tested with CONFIG_CGROUP_PIDS, it looks good.
> > >
> > > Tested-by: Song Liu <songliubraving@fb.com>
> >
> > hi,
> > I still need to apply my workaround to compile tests,
> > so I wonder this fell through cracks
> 
> 
> The fix was already applied ([0]). Do you still see issues?
> 
>   [0] https://patchwork.kernel.org/project/netdevbpf/patch/20201022202739.3667367-1-andrii@kernel.org/

nope, it's working now, I was on the wrong branch, sry for noise

thanks,
jirka