Skip to contentSkip to conversation

bpf, arm64: enable kfunc call

PATCH In review Hou Tao 3 participants 6 messages Jan 30, 2022 – Feb 4, 2022

Subsystems (8)
lore
Share & tools
lore

Hou Tao

author
#0 lore raw

v3 cover bpf, arm64: enable kfunc call

Hi,

The simple patchset tries to enable kfunc call for arm64. Patch #1 just overrides bpf_jit_supports_kfunc_call() to enable kfunc call, patch #2 unexports the subtests in ksyms_module.c to fix the confusion in test output and patch #3 add a test in ksyms_module.c to ensure s32 is sufficient for kfunc offset.

Change Log:
v3:
* patch #2: newly-addded to unexport unnecessary subtests
* patch #3: use kallsyms_find() instead of reimplementing it.
* patch #3: ensure kfunc call is supported before checking whether s32 will be overflowed or not.

v2: https://lore.kernel.org/bpf/[email protected]/
* add a test to check whether imm will be overflowed for kfunc call

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

Hou Tao (3):
bpf, arm64: enable kfunc call
selftests/bpf: do not export subtest as standalone test selftests/bpf: check whether s32 is sufficient for kfunc offset

 arch/arm64/net/bpf_jit_comp.c                 |  5 ++
 .../selftests/bpf/prog_tests/ksyms_module.c   | 46 ++++++++++++++++++-
 .../bpf/prog_tests/xdp_adjust_frags.c         |  6 ---
 .../bpf/prog_tests/xdp_adjust_tail.c          |  4 +-
 .../bpf/prog_tests/xdp_cpumap_attach.c        |  4 +-
 .../bpf/prog_tests/xdp_devmap_attach.c        |  2 +-
 6 files changed, 54 insertions(+), 13 deletions(-)
Signature · 1 line
2.20.1
author
#1 lore raw

PATCH v3 1/3 bpf, arm64: enable kfunc call

From: Hou Tao <[email protected]>

Since commit b2eed9b58811 ("arm64/kernel: kaslr: reduce module randomization range to 2 GB"), for arm64 whether KASLR is enabled or not, the module is placed within 2GB of the kernel region, so s32 in bpf_kfunc_desc is sufficient to represente the offset of module function relative to __bpf_call_base. The only thing needed is to override bpf_jit_supports_kfunc_call().

---
 arch/arm64/net/bpf_jit_comp.c | 5 +++++
 1 file changed, 5 insertions(+)
arch/arm64/net/bpf_jit_comp.c +5 -0
index e96d4d87291f..74f9a9b6a053 100644
@@ -1143,6 +1143,11 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 	return prog;
 }
 
+bool bpf_jit_supports_kfunc_call(void)
+{
+	return true;
+}
+
 u64 bpf_jit_alloc_exec_limit(void)
 {
 	return VMALLOC_END - VMALLOC_START;
Signature · 1 line
2.20.1
maintainer
#2 lore raw
Quoted 10 lines · Hou Tao

From: Hou Tao <[email protected]>

Since commit b2eed9b58811 ("arm64/kernel: kaslr: reduce module randomization range to 2 GB"), for arm64 whether KASLR is enabled or not, the module is placed within 2GB of the kernel region, so s32 in bpf_kfunc_desc is sufficient to represente the offset of module function relative to __bpf_call_base. The only thing needed is to override bpf_jit_supports_kfunc_call().

Could you rebase patch 2 & 3 and resend as they don't apply to bpf-next right now. Meanwhile, applied this one, thanks a lot!

author
#3 lore raw

PATCH v3 2/3 selftests/bpf: do not export subtest as standalone test

In reply to Hou Tao

Two subtests in ksyms_module.c are not qualified as static, so these subtests are exported as standalone tests in tests.h and lead to confusion for the output of "./test_progs -t ksyms_module".

By using the following command:

  grep "^void \(serial_\)\?test_[a-zA-Z0-9_]\+(\(void\)\?)" \
      tools/testing/selftests/bpf/prog_tests/*.c | \
	awk -F : '{print $1}' | sort | uniq -c | awk '$1 != 1'

Find out that other tests also have the similar problem, so fix these tests by marking subtests in these tests as static. For xdp_adjust_frags.c, there is just one subtest, so just export the subtest directly.

---
 tools/testing/selftests/bpf/prog_tests/ksyms_module.c      | 4 ++--
 tools/testing/selftests/bpf/prog_tests/xdp_adjust_frags.c  | 6 ------
 tools/testing/selftests/bpf/prog_tests/xdp_adjust_tail.c   | 4 ++--
 tools/testing/selftests/bpf/prog_tests/xdp_cpumap_attach.c | 4 ++--
 tools/testing/selftests/bpf/prog_tests/xdp_devmap_attach.c | 2 +-
 5 files changed, 7 insertions(+), 13 deletions(-)
tools/testing/selftests/bpf/prog_tests/ksyms_module.c +2 -2
index d490ad80eccb..997aa90dea64 100644
@@ -6,7 +6,7 @@
 #include "test_ksyms_module.lskel.h"
 #include "test_ksyms_module.skel.h"
 
-void test_ksyms_module_lskel(void)
+static void test_ksyms_module_lskel(void)
 {
 	struct test_ksyms_module_lskel *skel;
 	int retval;
@@ -30,7 +30,7 @@ void test_ksyms_module_lskel(void)
 	test_ksyms_module_lskel__destroy(skel);
 }
 
-void test_ksyms_module_libbpf(void)
+static void test_ksyms_module_libbpf(void)
 {
 	struct test_ksyms_module *skel;
 	int retval, err;
tools/testing/selftests/bpf/prog_tests/xdp_adjust_frags.c +0 -6
index 31c188666e81..6ca3582bb056 100644
@@ -96,9 +96,3 @@ void test_xdp_update_frags(void)
 out:
 	bpf_object__close(obj);
 }
-
-void test_xdp_adjust_frags(void)
-{
-	if (test__start_subtest("xdp_adjust_frags"))
-		test_xdp_update_frags();
-}
tools/testing/selftests/bpf/prog_tests/xdp_adjust_tail.c +2 -2
index ccc9e63254a8..a254c54c3ada 100644
@@ -118,7 +118,7 @@ static void test_xdp_adjust_tail_grow2(void)
 	bpf_object__close(obj);
 }
 
-void test_xdp_adjust_frags_tail_shrink(void)
+static void test_xdp_adjust_frags_tail_shrink(void)
 {
 	const char *file = "./test_xdp_adjust_tail_shrink.o";
 	__u32 duration, retval, size, exp_size;
@@ -180,7 +180,7 @@ void test_xdp_adjust_frags_tail_shrink(void)
 	bpf_object__close(obj);
 }
 
-void test_xdp_adjust_frags_tail_grow(void)
+static void test_xdp_adjust_frags_tail_grow(void)
 {
 	const char *file = "./test_xdp_adjust_tail_grow.o";
 	__u32 duration, retval, size, exp_size;
tools/testing/selftests/bpf/prog_tests/xdp_cpumap_attach.c +2 -2
index 13aabb3b6cf2..15e48098c8b8 100644
@@ -8,7 +8,7 @@
 
 #define IFINDEX_LO	1
 
-void test_xdp_with_cpumap_helpers(void)
+static void test_xdp_with_cpumap_helpers(void)
 {
 	struct test_xdp_with_cpumap_helpers *skel;
 	struct bpf_prog_info info = {};
@@ -68,7 +68,7 @@ void test_xdp_with_cpumap_helpers(void)
 	test_xdp_with_cpumap_helpers__destroy(skel);
 }
 
-void test_xdp_with_cpumap_frags_helpers(void)
+static void test_xdp_with_cpumap_frags_helpers(void)
 {
 	struct test_xdp_with_cpumap_frags_helpers *skel;
 	struct bpf_prog_info info = {};
tools/testing/selftests/bpf/prog_tests/xdp_devmap_attach.c +1 -1
index 2a784ccd3136..08d074ec471d 100644
@@ -81,7 +81,7 @@ static void test_neg_xdp_devmap_helpers(void)
 	}
 }
 
-void test_xdp_with_devmap_frags_helpers(void)
+static void test_xdp_with_devmap_frags_helpers(void)
 {
 	struct test_xdp_with_devmap_frags_helpers *skel;
 	struct bpf_prog_info info = {};
Signature · 1 line
2.20.1
author
#4 lore raw

PATCH v3 3/3 selftests/bpf: check whether s32 is sufficient for kfunc offset

In reply to Hou Tao

From: Hou Tao <[email protected]>

In add_kfunc_call(), bpf_kfunc_desc->imm with type s32 is used to represent the offset of called kfunc from __bpf_call_base, so add a test to ensure that the offset will not be overflowed.

---
 .../selftests/bpf/prog_tests/ksyms_module.c   | 42 +++++++++++++++++++
 1 file changed, 42 insertions(+)
tools/testing/selftests/bpf/prog_tests/ksyms_module.c +42 -0
index 997aa90dea64..955a9ede4756 100644
@@ -3,9 +3,49 @@
 
 #include <test_progs.h>
 #include <network_helpers.h>
+#include <trace_helpers.h>
 #include "test_ksyms_module.lskel.h"
 #include "test_ksyms_module.skel.h"
 
+/*
+ * Check whether or not s32 in bpf_kfunc_desc is sufficient
+ * to represent the offset between bpf_testmod_test_mod_kfunc
+ * and __bpf_call_base.
+ */
+static void test_ksyms_module_valid_offset(void)
+{
+	struct test_ksyms_module *skel;
+	unsigned long long kfunc_addr;
+	unsigned long long base_addr;
+	long long actual_offset;
+	int used_offset;
+	int err;
+
+	if (!env.has_testmod) {
+		test__skip();
+		return;
+	}
+
+	/* Ensure kfunc call is supported */
+	skel = test_ksyms_module__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "test_ksyms_module__open"))
+		return;
+
+	err = kallsyms_find("bpf_testmod_test_mod_kfunc", &kfunc_addr);
+	if (!ASSERT_OK(err, "find kfunc addr"))
+		goto cleanup;
+
+	err = kallsyms_find("__bpf_call_base", &base_addr);
+	if (!ASSERT_OK(err, "find base addr"))
+		goto cleanup;
+
+	used_offset = kfunc_addr - base_addr;
+	actual_offset = kfunc_addr - base_addr;
+	ASSERT_EQ((long long)used_offset, actual_offset, "kfunc offset overflowed");
+cleanup:
+	test_ksyms_module__destroy(skel);
+}
+
 static void test_ksyms_module_lskel(void)
 {
 	struct test_ksyms_module_lskel *skel;
@@ -55,6 +95,8 @@ static void test_ksyms_module_libbpf(void)
 
 void test_ksyms_module(void)
 {
+	if (test__start_subtest("valid_offset"))
+		test_ksyms_module_valid_offset();
 	if (test__start_subtest("lskel"))
 		test_ksyms_module_lskel();
 	if (test__start_subtest("libbpf"))
Signature · 1 line
2.20.1
#5 lore raw
In reply to Hou Tao
This series was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:

Hello:

This series was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:

Quoted 9 lines · On Sun, 30 Jan 2022 17:29:14 +0800 you wrote:

Hi,

The simple patchset tries to enable kfunc call for arm64. Patch #1 just overrides bpf_jit_supports_kfunc_call() to enable kfunc call, patch #2 unexports the subtests in ksyms_module.c to fix the confusion in test output and patch #3 add a test in ksyms_module.c to ensure s32 is sufficient for kfunc offset.

[...]

Here is the summary with links:
- [bpf-next,v3,1/3] bpf, arm64: enable kfunc call

    https://git.kernel.org/bpf/bpf-next/c/b5e975d256db

- [bpf-next,v3,2/3] selftests/bpf: do not export subtest as standalone test (no matching commit)
- [bpf-next,v3,3/3] selftests/bpf: check whether s32 is sufficient for kfunc offset (no matching commit)

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html