[PATCH v7 0/2] libbpf: fix USDT SIB argument handling causing unrecognized register error
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
The current USDT implementation in libbpf cannot parse these two formats, causing `bpf_program__attach_usdt()` to fail with -ENOENT (unrecognized register).
This patch series adds support for SIB addressing mode in USDT probes. The main changes include: - add correct handling logic for SIB-addressed arguments in `parse_usdt_arg`. - add an usdt_o2 test case to cover SIB addressing mode.
Testing shows that the SIB probe correctly generates 8@(%rcx,%rax,8) argument spec and passes all validation checks.
The modification history of this patch series: Change since v1: - refactor the code to make it more readable - modify the commit message to explain why and how
Change since v2: - fix the `scale` uninitialized error
Change since v3: - force -O2 optimization for usdt.test.o to generate SIB addressing usdt and pass all test cases.
Change since v4: - split the patch into two parts, one for the fix and the other for the test
Change since v5: - Only enable optimization for x86 architecture to generate SIB addressing usdt argument spec.
Change since v6: - Add an usdt_o2 test case to cover SIB addressing mode. - Reinstate the usdt.c test case.
Do we need to add support for PC-relative USDT argument spec handling in libbpf? I have some interest in this question, but currently have no ideas. Getting offsets based on symbols requires dependency on the symbol table. However, once the binary file is stripped, the symtab will also be removed, which will cause this approach to fail. Does anyone have any thoughts on this?
Jiawei Zhao (2): libbpf: fix USDT SIB argument handling causing unrecognized register error selftests/bpf: Force -O2 for USDT selftests to cover SIB handling logic
tools/lib/bpf/usdt.bpf.h | 33 ++++++++- tools/lib/bpf/usdt.c | 43 +++++++++-- tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 5 files changed, 185 insertions(+), 7 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
On x86-64, USDT arguments can be specified using Scale-Index-Base (SIB) addressing, e.g. "1@-96(%rbp,%rax,8)". The current USDT implementation in libbpf cannot parse this format, causing `bpf_program__attach_usdt()` to fail with -ENOENT (unrecognized register).
This patch fixes this by implementing the necessary changes: - add correct handling for SIB-addressed arguments in `bpf_usdt_arg`. - add adaptive support to `__bpf_usdt_arg_type` and `__bpf_usdt_arg_spec` to represent SIB addressing parameters.
Signed-off-by: Jiawei Zhao phoenix500526@163.com --- tools/lib/bpf/usdt.bpf.h | 33 +++++++++++++++++++++++++++++- tools/lib/bpf/usdt.c | 43 ++++++++++++++++++++++++++++++++++------ 2 files changed, 69 insertions(+), 7 deletions(-)
diff --git a/tools/lib/bpf/usdt.bpf.h b/tools/lib/bpf/usdt.bpf.h index 2a7865c8e3fe..246513088c3a 100644 --- a/tools/lib/bpf/usdt.bpf.h +++ b/tools/lib/bpf/usdt.bpf.h @@ -34,6 +34,7 @@ enum __bpf_usdt_arg_type { BPF_USDT_ARG_CONST, BPF_USDT_ARG_REG, BPF_USDT_ARG_REG_DEREF, + BPF_USDT_ARG_SIB, };
struct __bpf_usdt_arg_spec { @@ -43,6 +44,10 @@ struct __bpf_usdt_arg_spec { enum __bpf_usdt_arg_type arg_type; /* offset of referenced register within struct pt_regs */ short reg_off; + /* offset of index register in pt_regs, only used in SIB mode */ + short idx_reg_off; + /* scale factor for index register, only used in SIB mode */ + short scale; /* whether arg should be interpreted as signed value */ bool arg_signed; /* number of bits that need to be cleared and, optionally, @@ -149,7 +154,7 @@ int bpf_usdt_arg(struct pt_regs *ctx, __u64 arg_num, long *res) { struct __bpf_usdt_spec *spec; struct __bpf_usdt_arg_spec *arg_spec; - unsigned long val; + unsigned long val, idx; int err, spec_id;
*res = 0; @@ -202,6 +207,32 @@ int bpf_usdt_arg(struct pt_regs *ctx, __u64 arg_num, long *res) return err; #if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ val >>= arg_spec->arg_bitshift; +#endif + break; + case BPF_USDT_ARG_SIB: + /* Arg is in memory addressed by SIB (Scale-Index-Base) mode + * (e.g., "-1@-96(%rbp,%rax,8)" in USDT arg spec). Register + * is identified like with BPF_USDT_ARG_SIB case, the offset + * is in arg_spec->val_off, the scale factor is in arg_spec->scale. + * Firstly, we fetch the base register contents and the index + * register contents from pt_regs. Secondly, we multiply the + * index register contents by the scale factor, then add the + * base address and the offset to get the final address. Finally, + * we do another user-space probe read to fetch argument value + * itself. + */ + err = bpf_probe_read_kernel(&val, sizeof(val), (void *)ctx + arg_spec->reg_off); + if (err) + return err; + err = bpf_probe_read_kernel(&idx, sizeof(idx), (void *)ctx + arg_spec->idx_reg_off); + if (err) + return err; + err = bpf_probe_read_user(&val, sizeof(val), + (void *)val + idx * arg_spec->scale + arg_spec->val_off); + if (err) + return err; +#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ + val >>= arg_spec->arg_bitshift; #endif break; default: diff --git a/tools/lib/bpf/usdt.c b/tools/lib/bpf/usdt.c index 4e4a52742b01..1f8b9e1c9819 100644 --- a/tools/lib/bpf/usdt.c +++ b/tools/lib/bpf/usdt.c @@ -200,6 +200,7 @@ enum usdt_arg_type { USDT_ARG_CONST, USDT_ARG_REG, USDT_ARG_REG_DEREF, + USDT_ARG_SIB, };
/* should match exactly struct __bpf_usdt_arg_spec from usdt.bpf.h */ @@ -207,6 +208,8 @@ struct usdt_arg_spec { __u64 val_off; enum usdt_arg_type arg_type; short reg_off; + short idx_reg_off; + short scale; bool arg_signed; char arg_bitshift; }; @@ -1283,11 +1286,39 @@ static int calc_pt_regs_off(const char *reg_name)
static int parse_usdt_arg(const char *arg_str, int arg_num, struct usdt_arg_spec *arg, int *arg_sz) { - char reg_name[16]; - int len, reg_off; - long off; + char reg_name[16] = {0}, idx_reg_name[16] = {0}; + int len, reg_off, idx_reg_off, scale = 1; + long off = 0; + + if (sscanf(arg_str, " %d @ %ld ( %%%15[^,] , %%%15[^,] , %d ) %n", + arg_sz, &off, reg_name, idx_reg_name, &scale, &len) == 5 || + sscanf(arg_str, " %d @ ( %%%15[^,] , %%%15[^,] , %d ) %n", + arg_sz, reg_name, idx_reg_name, &scale, &len) == 4 || + sscanf(arg_str, " %d @ %ld ( %%%15[^,] , %%%15[^)] ) %n", + arg_sz, &off, reg_name, idx_reg_name, &len) == 4 || + sscanf(arg_str, " %d @ ( %%%15[^,] , %%%15[^)] ) %n", + arg_sz, reg_name, idx_reg_name, &len) == 3 + ) { + /* Scale Index Base case, e.g., 1@-96(%rbp,%rax,8) + * 1@(%rbp,%rax,8) + * 1@-96(%rbp,%rax) + * 1@(%rbp,%rax) + */ + arg->arg_type = USDT_ARG_SIB; + arg->val_off = off; + arg->scale = scale; + + reg_off = calc_pt_regs_off(reg_name); + if (reg_off < 0) + return reg_off; + arg->reg_off = reg_off;
- if (sscanf(arg_str, " %d @ %ld ( %%%15[^)] ) %n", arg_sz, &off, reg_name, &len) == 3) { + idx_reg_off = calc_pt_regs_off(idx_reg_name); + if (idx_reg_off < 0) + return idx_reg_off; + arg->idx_reg_off = idx_reg_off; + } else if (sscanf(arg_str, " %d @ %ld ( %%%15[^)] ) %n", + arg_sz, &off, reg_name, &len) == 3) { /* Memory dereference case, e.g., -4@-20(%rbp) */ arg->arg_type = USDT_ARG_REG_DEREF; arg->val_off = off; @@ -1298,7 +1329,7 @@ static int parse_usdt_arg(const char *arg_str, int arg_num, struct usdt_arg_spec } else if (sscanf(arg_str, " %d @ ( %%%15[^)] ) %n", arg_sz, reg_name, &len) == 2) { /* Memory dereference case without offset, e.g., 8@(%rsp) */ arg->arg_type = USDT_ARG_REG_DEREF; - arg->val_off = 0; + arg->val_off = off; reg_off = calc_pt_regs_off(reg_name); if (reg_off < 0) return reg_off; @@ -1306,7 +1337,7 @@ static int parse_usdt_arg(const char *arg_str, int arg_num, struct usdt_arg_spec } else if (sscanf(arg_str, " %d @ %%%15s %n", arg_sz, reg_name, &len) == 2) { /* Register read case, e.g., -4@%eax */ arg->arg_type = USDT_ARG_REG; - arg->val_off = 0; + arg->val_off = off;
reg_off = calc_pt_regs_off(reg_name); if (reg_off < 0)
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
In this patch: - add usdt_o2 test case to cover SIB addressing usdt argument spec handling logic
Signed-off-by: Jiawei Zhao phoenix500526@163.com --- tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 3 files changed, 116 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..68cf6a9cf05f 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -759,6 +759,14 @@ TRUNNER_BPF_BUILD_RULE := $$(error no BPF objects should be built) TRUNNER_BPF_CFLAGS := $(eval $(call DEFINE_TEST_RUNNER,test_maps))
+# Use -O2 optimization to generate SIB addressing usdt argument spec +# Only apply on x86 architecture where SIB addressing is relevant +ifeq ($(ARCH), x86) +$(OUTPUT)/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/cpuv4/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/no_alu32/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +endif + # Define test_verifier test runner. # It is much simpler than test_maps/test_progs and sufficiently different from # them (e.g., test.h is using completely pattern), that it's worth just diff --git a/tools/testing/selftests/bpf/prog_tests/usdt_o2.c b/tools/testing/selftests/bpf/prog_tests/usdt_o2.c new file mode 100644 index 000000000000..f04b756b3640 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/usdt_o2.c @@ -0,0 +1,71 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2025 Jiawei Zhao phoenix500526@163.com. */ +#include <test_progs.h> + +#define _SDT_HAS_SEMAPHORES 1 +#include "../sdt.h" +#include "test_usdt_o2.skel.h" + +int lets_test_this(int); + +#define test_value 0xFEDCBA9876543210ULL +#define SEC(name) __attribute__((section(name), used)) + + +static volatile __u64 array[1] = {test_value}; +unsigned short test_usdt1_semaphore SEC(".probes"); + +static __always_inline void trigger_func(void) +{ + /* Base address + offset + (index * scale) */ + if (test_usdt1_semaphore) { + for (volatile int i = 0; i <= 0; i++) + STAP_PROBE1(test, usdt1, array[i]); + } +} + +static void basic_sib_usdt(void) +{ + LIBBPF_OPTS(bpf_usdt_opts, opts); + struct test_usdt_o2 *skel; + struct test_usdt_o2__bss *bss; + int err; + + skel = test_usdt_o2__open_and_load(); + if (!ASSERT_OK_PTR(skel, "skel_open")) + return; + + bss = skel->bss; + bss->my_pid = getpid(); + + err = test_usdt_o2__attach(skel); + if (!ASSERT_OK(err, "skel_attach")) + goto cleanup; + + /* usdt1 won't be auto-attached */ + opts.usdt_cookie = 0xcafedeadbeeffeed; + skel->links.usdt1 = bpf_program__attach_usdt(skel->progs.usdt1, + 0 /*self*/, "/proc/self/exe", + "test", "usdt1", &opts); + if (!ASSERT_OK_PTR(skel->links.usdt1, "usdt1_link")) + goto cleanup; + + trigger_func(); + + ASSERT_EQ(bss->usdt1_called, 1, "usdt1_called"); + ASSERT_EQ(bss->usdt1_cookie, 0xcafedeadbeeffeed, "usdt1_cookie"); + ASSERT_EQ(bss->usdt1_arg_cnt, 1, "usdt1_arg_cnt"); + ASSERT_EQ(bss->usdt1_arg, test_value, "usdt1_arg"); + ASSERT_EQ(bss->usdt1_arg_ret, 0, "usdt1_arg_ret"); + ASSERT_EQ(bss->usdt1_arg_size, sizeof(array[0]), "usdt1_arg_size"); + +cleanup: + test_usdt_o2__destroy(skel); +} + + + +void test_usdt_o2(void) +{ + basic_sib_usdt(); +} diff --git a/tools/testing/selftests/bpf/progs/test_usdt_o2.c b/tools/testing/selftests/bpf/progs/test_usdt_o2.c new file mode 100644 index 000000000000..14602aa54578 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_usdt_o2.c @@ -0,0 +1,37 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2022 Meta Platforms, Inc. and affiliates. */ + +#include "vmlinux.h" +#include <bpf/bpf_helpers.h> +#include <bpf/usdt.bpf.h> + +int my_pid; + +int usdt1_called; +u64 usdt1_cookie; +int usdt1_arg_cnt; +int usdt1_arg_ret; +u64 usdt1_arg; +int usdt1_arg_size; + +SEC("usdt") +int usdt1(struct pt_regs *ctx) +{ + long tmp; + + if (my_pid != (bpf_get_current_pid_tgid() >> 32)) + return 0; + + __sync_fetch_and_add(&usdt1_called, 1); + + usdt1_cookie = bpf_usdt_cookie(ctx); + usdt1_arg_cnt = bpf_usdt_arg_cnt(ctx); + + usdt1_arg_ret = bpf_usdt_arg(ctx, 0, &tmp); + usdt1_arg = (u64)tmp; + usdt1_arg_size = bpf_usdt_arg_size(ctx, 0); + + return 0; +} + +char _license[] SEC("license") = "GPL";
On 8/6/25 2:24 AM, Jiawei Zhao wrote:
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
In this patch:
- add usdt_o2 test case to cover SIB addressing usdt argument spec handling logic
Signed-off-by: Jiawei Zhao phoenix500526@163.com
tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 3 files changed, 116 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..68cf6a9cf05f 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -759,6 +759,14 @@ TRUNNER_BPF_BUILD_RULE := $$(error no BPF objects should be built) TRUNNER_BPF_CFLAGS := $(eval $(call DEFINE_TEST_RUNNER,test_maps)) +# Use -O2 optimization to generate SIB addressing usdt argument spec +# Only apply on x86 architecture where SIB addressing is relevant +ifeq ($(ARCH), x86) +$(OUTPUT)/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/cpuv4/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/no_alu32/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +endif
I tried your selftest with gcc14 and llvm20 in my environment. See below:
llvm20: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x0000002f NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x00000000000003ac, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@-64(%rbp)
gcc14: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000034 NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x0000000000000334, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@array(,%rax,8)
llvm20 and gcc14 generate different usdt patterns. '8@-64(%rbp)' already supports so with SIB support, the test should pass CI, I think.
- # Define test_verifier test runner. # It is much simpler than test_maps/test_progs and sufficiently different from # them (e.g., test.h is using completely pattern), that it's worth just
diff --git a/tools/testing/selftests/bpf/prog_tests/usdt_o2.c b/tools/testing/selftests/bpf/prog_tests/usdt_o2.c new file mode 100644 index 000000000000..f04b756b3640 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/usdt_o2.c @@ -0,0 +1,71 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2025 Jiawei Zhao phoenix500526@163.com. */ +#include <test_progs.h>
+#define _SDT_HAS_SEMAPHORES 1 +#include "../sdt.h" +#include "test_usdt_o2.skel.h"
+int lets_test_this(int);
+#define test_value 0xFEDCBA9876543210ULL +#define SEC(name) __attribute__((section(name), used))
+static volatile __u64 array[1] = {test_value}; +unsigned short test_usdt1_semaphore SEC(".probes");
+static __always_inline void trigger_func(void) +{
- /* Base address + offset + (index * scale) */
- if (test_usdt1_semaphore) {
for (volatile int i = 0; i <= 0; i++)STAP_PROBE1(test, usdt1, array[i]);- }
+}
+static void basic_sib_usdt(void) +{
- LIBBPF_OPTS(bpf_usdt_opts, opts);
- struct test_usdt_o2 *skel;
- struct test_usdt_o2__bss *bss;
- int err;
- skel = test_usdt_o2__open_and_load();
- if (!ASSERT_OK_PTR(skel, "skel_open"))
return;- bss = skel->bss;
- bss->my_pid = getpid();
- err = test_usdt_o2__attach(skel);
- if (!ASSERT_OK(err, "skel_attach"))
goto cleanup;- /* usdt1 won't be auto-attached */
- opts.usdt_cookie = 0xcafedeadbeeffeed;
- skel->links.usdt1 = bpf_program__attach_usdt(skel->progs.usdt1,
0 /*self*/, "/proc/self/exe","test", "usdt1", &opts);- if (!ASSERT_OK_PTR(skel->links.usdt1, "usdt1_link"))
goto cleanup;- trigger_func();
- ASSERT_EQ(bss->usdt1_called, 1, "usdt1_called");
- ASSERT_EQ(bss->usdt1_cookie, 0xcafedeadbeeffeed, "usdt1_cookie");
- ASSERT_EQ(bss->usdt1_arg_cnt, 1, "usdt1_arg_cnt");
- ASSERT_EQ(bss->usdt1_arg, test_value, "usdt1_arg");
- ASSERT_EQ(bss->usdt1_arg_ret, 0, "usdt1_arg_ret");
- ASSERT_EQ(bss->usdt1_arg_size, sizeof(array[0]), "usdt1_arg_size");
+cleanup:
- test_usdt_o2__destroy(skel);
+}
+void test_usdt_o2(void) +{
- basic_sib_usdt();
+} diff --git a/tools/testing/selftests/bpf/progs/test_usdt_o2.c b/tools/testing/selftests/bpf/progs/test_usdt_o2.c new file mode 100644 index 000000000000..14602aa54578 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_usdt_o2.c @@ -0,0 +1,37 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2022 Meta Platforms, Inc. and affiliates. */
+#include "vmlinux.h" +#include <bpf/bpf_helpers.h> +#include <bpf/usdt.bpf.h>
+int my_pid;
+int usdt1_called; +u64 usdt1_cookie; +int usdt1_arg_cnt; +int usdt1_arg_ret; +u64 usdt1_arg; +int usdt1_arg_size;
+SEC("usdt") +int usdt1(struct pt_regs *ctx) +{
- long tmp;
- if (my_pid != (bpf_get_current_pid_tgid() >> 32))
return 0;- __sync_fetch_and_add(&usdt1_called, 1);
- usdt1_cookie = bpf_usdt_cookie(ctx);
- usdt1_arg_cnt = bpf_usdt_arg_cnt(ctx);
- usdt1_arg_ret = bpf_usdt_arg(ctx, 0, &tmp);
- usdt1_arg = (u64)tmp;
- usdt1_arg_size = bpf_usdt_arg_size(ctx, 0);
- return 0;
+}
+char _license[] SEC("license") = "GPL";
Hi Yonghong,
I noticed that the USDT argument specification generated by GCC 14 is '8@array(,%rax,8)'. This pattern is currently not handled correctly. I'm exploring whether I can use DWARF information to calculate the address of this variable. This approach seems to work. However, since I can't reproduce the same issue on my machine, I plan to implement this approach for the PC-relative issue in a separate patch. Would that affect the merging of this patch?
At 2025-08-07 02:17:34, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/6/25 2:24 AM, Jiawei Zhao wrote:
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
In this patch:
- add usdt_o2 test case to cover SIB addressing usdt argument spec handling logic
Signed-off-by: Jiawei Zhao phoenix500526@163.com
tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 3 files changed, 116 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..68cf6a9cf05f 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -759,6 +759,14 @@ TRUNNER_BPF_BUILD_RULE := $$(error no BPF objects should be built) TRUNNER_BPF_CFLAGS := $(eval $(call DEFINE_TEST_RUNNER,test_maps)) +# Use -O2 optimization to generate SIB addressing usdt argument spec +# Only apply on x86 architecture where SIB addressing is relevant +ifeq ($(ARCH), x86) +$(OUTPUT)/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/cpuv4/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/no_alu32/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +endif
I tried your selftest with gcc14 and llvm20 in my environment. See below:
llvm20: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x0000002f NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x00000000000003ac, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@-64(%rbp)
gcc14: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000034 NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x0000000000000334, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@array(,%rax,8)
llvm20 and gcc14 generate different usdt patterns. '8@-64(%rbp)' already supports so with SIB support, the test should pass CI, I think.
- # Define test_verifier test runner. # It is much simpler than test_maps/test_progs and sufficiently different from # them (e.g., test.h is using completely pattern), that it's worth just
diff --git a/tools/testing/selftests/bpf/prog_tests/usdt_o2.c b/tools/testing/selftests/bpf/prog_tests/usdt_o2.c new file mode 100644 index 000000000000..f04b756b3640 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/usdt_o2.c @@ -0,0 +1,71 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2025 Jiawei Zhao phoenix500526@163.com. */ +#include <test_progs.h>
+#define _SDT_HAS_SEMAPHORES 1 +#include "../sdt.h" +#include "test_usdt_o2.skel.h"
+int lets_test_this(int);
+#define test_value 0xFEDCBA9876543210ULL +#define SEC(name) __attribute__((section(name), used))
+static volatile __u64 array[1] = {test_value}; +unsigned short test_usdt1_semaphore SEC(".probes");
+static __always_inline void trigger_func(void) +{
- /* Base address + offset + (index * scale) */
- if (test_usdt1_semaphore) {
for (volatile int i = 0; i <= 0; i++)STAP_PROBE1(test, usdt1, array[i]);- }
+}
+static void basic_sib_usdt(void) +{
- LIBBPF_OPTS(bpf_usdt_opts, opts);
- struct test_usdt_o2 *skel;
- struct test_usdt_o2__bss *bss;
- int err;
- skel = test_usdt_o2__open_and_load();
- if (!ASSERT_OK_PTR(skel, "skel_open"))
return;- bss = skel->bss;
- bss->my_pid = getpid();
- err = test_usdt_o2__attach(skel);
- if (!ASSERT_OK(err, "skel_attach"))
goto cleanup;- /* usdt1 won't be auto-attached */
- opts.usdt_cookie = 0xcafedeadbeeffeed;
- skel->links.usdt1 = bpf_program__attach_usdt(skel->progs.usdt1,
0 /*self*/, "/proc/self/exe","test", "usdt1", &opts);- if (!ASSERT_OK_PTR(skel->links.usdt1, "usdt1_link"))
goto cleanup;- trigger_func();
- ASSERT_EQ(bss->usdt1_called, 1, "usdt1_called");
- ASSERT_EQ(bss->usdt1_cookie, 0xcafedeadbeeffeed, "usdt1_cookie");
- ASSERT_EQ(bss->usdt1_arg_cnt, 1, "usdt1_arg_cnt");
- ASSERT_EQ(bss->usdt1_arg, test_value, "usdt1_arg");
- ASSERT_EQ(bss->usdt1_arg_ret, 0, "usdt1_arg_ret");
- ASSERT_EQ(bss->usdt1_arg_size, sizeof(array[0]), "usdt1_arg_size");
+cleanup:
- test_usdt_o2__destroy(skel);
+}
+void test_usdt_o2(void) +{
- basic_sib_usdt();
+} diff --git a/tools/testing/selftests/bpf/progs/test_usdt_o2.c b/tools/testing/selftests/bpf/progs/test_usdt_o2.c new file mode 100644 index 000000000000..14602aa54578 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_usdt_o2.c @@ -0,0 +1,37 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2022 Meta Platforms, Inc. and affiliates. */
+#include "vmlinux.h" +#include <bpf/bpf_helpers.h> +#include <bpf/usdt.bpf.h>
+int my_pid;
+int usdt1_called; +u64 usdt1_cookie; +int usdt1_arg_cnt; +int usdt1_arg_ret; +u64 usdt1_arg; +int usdt1_arg_size;
+SEC("usdt") +int usdt1(struct pt_regs *ctx) +{
- long tmp;
- if (my_pid != (bpf_get_current_pid_tgid() >> 32))
return 0;- __sync_fetch_and_add(&usdt1_called, 1);
- usdt1_cookie = bpf_usdt_cookie(ctx);
- usdt1_arg_cnt = bpf_usdt_arg_cnt(ctx);
- usdt1_arg_ret = bpf_usdt_arg(ctx, 0, &tmp);
- usdt1_arg = (u64)tmp;
- usdt1_arg_size = bpf_usdt_arg_size(ctx, 0);
- return 0;
+}
+char _license[] SEC("license") = "GPL";
On 8/6/25 7:57 PM, 赵佳炜 wrote:
Hi Yonghong,
I noticed that the USDT argument specification generated by GCC 14 is '8@array(,%rax,8)'. This pattern is currently not handled correctly. I'm exploring whether I can use DWARF information to calculate the address of this variable. This approach seems to work. However, since I can't
I think 'array' should be in symbol table, so there is no need to check dwarf in my opinion.
reproduce the same issue on my machine, I plan to implement this approach for the PC-relative issue in a separate patch. Would that affect the merging of this patch?
Let us handle this since '8@array(,%rax,8)' may appear in CI environment.
At 2025-08-07 02:17:34, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/6/25 2:24 AM, Jiawei Zhao wrote:
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
In this patch:
- add usdt_o2 test case to cover SIB addressing usdt argument spec handling logic
Signed-off-by: Jiawei Zhao phoenix500526@163.com
tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 3 files changed, 116 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..68cf6a9cf05f 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -759,6 +759,14 @@ TRUNNER_BPF_BUILD_RULE := $$(error no BPF objects should be built) TRUNNER_BPF_CFLAGS := $(eval $(call DEFINE_TEST_RUNNER,test_maps)) +# Use -O2 optimization to generate SIB addressing usdt argument spec +# Only apply on x86 architecture where SIB addressing is relevant +ifeq ($(ARCH), x86) +$(OUTPUT)/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/cpuv4/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/no_alu32/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +endif
I tried your selftest with gcc14 and llvm20 in my environment. See below:
llvm20: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x0000002f NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x00000000000003ac, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@-64(%rbp)
gcc14: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000034 NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x0000000000000334, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@array(,%rax,8)
llvm20 and gcc14 generate different usdt patterns. '8@-64(%rbp)' already supports so with SIB support, the test should pass CI, I think.
[...]
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Thanks, Jiawei Zhao
At 2025-08-08 02:01:08, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/6/25 7:57 PM, 赵佳炜 wrote:
Hi Yonghong,
I noticed that the USDT argument specification generated by GCC 14 is '8@array(,%rax,8)'. This pattern is currently not handled correctly. I'm exploring whether I can use DWARF information to calculate the address of this variable. This approach seems to work. However, since I can't
I think 'array' should be in symbol table, so there is no need to check dwarf in my opinion.
reproduce the same issue on my machine, I plan to implement this approach for the PC-relative issue in a separate patch. Would that affect the merging of this patch?
Let us handle this since '8@array(,%rax,8)' may appear in CI environment.
At 2025-08-07 02:17:34, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/6/25 2:24 AM, Jiawei Zhao wrote:
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
In this patch:
- add usdt_o2 test case to cover SIB addressing usdt argument spec handling logic
Signed-off-by: Jiawei Zhao phoenix500526@163.com
tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 3 files changed, 116 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..68cf6a9cf05f 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -759,6 +759,14 @@ TRUNNER_BPF_BUILD_RULE := $$(error no BPF objects should be built) TRUNNER_BPF_CFLAGS := $(eval $(call DEFINE_TEST_RUNNER,test_maps)) +# Use -O2 optimization to generate SIB addressing usdt argument spec +# Only apply on x86 architecture where SIB addressing is relevant +ifeq ($(ARCH), x86) +$(OUTPUT)/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/cpuv4/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/no_alu32/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +endif
I tried your selftest with gcc14 and llvm20 in my environment. See below:
llvm20: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x0000002f NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x00000000000003ac, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@-64(%rbp)
gcc14: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000034 NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x0000000000000334, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@array(,%rax,8)
llvm20 and gcc14 generate different usdt patterns. '8@-64(%rbp)' already supports so with SIB support, the test should pass CI, I think.
[...]
Or maybe we can just throw an error when LOCAL symbols are duplicated. WDYT?
At 2025-08-10 16:55:38, "赵佳炜" phoenix500526@163.com wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Thanks, Jiawei Zhao
At 2025-08-08 02:01:08, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/6/25 7:57 PM, 赵佳炜 wrote:
Hi Yonghong,
I noticed that the USDT argument specification generated by GCC 14 is '8@array(,%rax,8)'. This pattern is currently not handled correctly. I'm exploring whether I can use DWARF information to calculate the address of this variable. This approach seems to work. However, since I can't
I think 'array' should be in symbol table, so there is no need to check dwarf in my opinion.
reproduce the same issue on my machine, I plan to implement this approach for the PC-relative issue in a separate patch. Would that affect the merging of this patch?
Let us handle this since '8@array(,%rax,8)' may appear in CI environment.
At 2025-08-07 02:17:34, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/6/25 2:24 AM, Jiawei Zhao wrote:
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
In this patch:
- add usdt_o2 test case to cover SIB addressing usdt argument spec handling logic
Signed-off-by: Jiawei Zhao phoenix500526@163.com
tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 3 files changed, 116 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..68cf6a9cf05f 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -759,6 +759,14 @@ TRUNNER_BPF_BUILD_RULE := $$(error no BPF objects should be built) TRUNNER_BPF_CFLAGS := $(eval $(call DEFINE_TEST_RUNNER,test_maps)) +# Use -O2 optimization to generate SIB addressing usdt argument spec +# Only apply on x86 architecture where SIB addressing is relevant +ifeq ($(ARCH), x86) +$(OUTPUT)/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/cpuv4/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/no_alu32/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +endif
I tried your selftest with gcc14 and llvm20 in my environment. See below:
llvm20: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x0000002f NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x00000000000003ac, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@-64(%rbp)
gcc14: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000034 NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x0000000000000334, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@array(,%rax,8)
llvm20 and gcc14 generate different usdt patterns. '8@-64(%rbp)' already supports so with SIB support, the test should pass CI, I think.
[...]
On 8/10/25 1:55 AM, 赵佳炜 wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Did you check relocations? Relocaitons should be able to point exact which symbol.
Thanks, Jiawei Zhao
[...]
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/10/25 1:55 AM, 赵佳炜 wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Did you check relocations? Relocaitons should be able to point exact which symbol.
Thanks, Jiawei Zhao
[...]
On 8/12/25 12:02 AM, 赵佳炜 wrote:
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
Could you share the complete source codes and compiler options which reproduce the above result?
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/10/25 1:55 AM, 赵佳炜 wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Did you check relocations? Relocaitons should be able to point exact which symbol.
Thanks, Jiawei Zhao
[...]
Sure.
The usdt_rip.c source code:
```C // the usdt_rip.c file #include <stddef.h> #include <stdint.h> #include "sdt.h"
static volatile char ti = 0;
int add(int a, int b) { return a + b; } int (*add_ptr)(int, int) = add;
struct st{ int a; char b; };
volatile struct st t1 = {.a = 1, .b = 'a'};
static void __always_inline trigger_func() { STAP_PROBE4(usdt_rip, rip_global_var, ti, add_ptr, t1.b, ti); }
int main() { trigger_func(); return 0; } ```
The test/usdt_rip.c source code:
```C // The test/usdt_rip.c file static volatile char ti = 0; ```
The compiler option is `gcc -O2 -g usdt_rip.c test/usdt_rip.c -o usdt_rip`
My compiler version: ```bash $ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr --with-gcc-major-version-only --program-suffix=-13 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib --enable-libphobos-checking=release --with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch --disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr --enable-offload-defaulted --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu --with-build-config=bootstrap-lto-lean --enable-link-serialization=2 Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04) ```
At 2025-08-13 00:11:45, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 12:02 AM, 赵佳炜 wrote:
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
Could you share the complete source codes and compiler options which reproduce the above result?
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/10/25 1:55 AM, 赵佳炜 wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Did you check relocations? Relocaitons should be able to point exact which symbol.
Thanks, Jiawei Zhao
[...]
On 8/12/25 7:27 PM, 赵佳炜 wrote:
Sure.
The usdt_rip.c source code:
// the usdt_rip.c file #include <stddef.h> #include <stdint.h> #include "sdt.h" static volatile char ti = 0; int add(int a, int b) { return a + b; } int (*add_ptr)(int, int) = add; struct st{ int a; char b; }; volatile struct st t1 = {.a = 1, .b = 'a'}; static void __always_inline trigger_func() { STAP_PROBE4(usdt_rip, rip_global_var, ti, add_ptr, t1.b, ti); } int main() { trigger_func(); return 0; }The test/usdt_rip.c source code:
// The test/usdt_rip.c file static volatile char ti = 0;The compiler option is `gcc -O2 -g usdt_rip.c test/usdt_rip.c -o usdt_rip`
My compiler version:
$ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr --with-gcc-major-version-only --program-suffix=-13 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib --enable-libphobos-checking=release --with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch --disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr --enable-offload-defaulted --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu --with-build-config=bootstrap-lto-lean --enable-link-serialization=2 Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04)
Thanks for the code. I compiled with gcc14 compiler and the below is the asm code for 'main' function:
0000000000401020 <main>: 401020: 90 nop 401021: 31 c0 xor %eax,%eax 401023: c3 ret 401024: 66 2e 0f 1f 84 00 00 cs nopw 0x0(%rax,%rax,1) 40102b: 00 00 00 40102e: 66 90 xchg %ax,%ax
stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000401020, Base: 0x0000000000402010, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@t1+4(%rip) -1@ti(%rip)
$ readelf -s usdt_rip | grep ti 4: 0000000000404019 1 OBJECT LOCAL DEFAULT 24 ti 14: 000000000040401a 1 OBJECT LOCAL DEFAULT 24 ti
So yet, it would be hard to find which 'ti' should be used to resolve usdt arguments.
I then tried with clang21. The 'main' function:
0000000000001140 <main>: 1140: 0f b6 05 f2 2e 00 00 movzbl 0x2ef2(%rip),%eax # 4039 <ti> 1147: 48 8b 0d da 2e 00 00 mov 0x2eda(%rip),%rcx # 4028 <add_ptr> 114e: 0f b6 15 df 2e 00 00 movzbl 0x2edf(%rip),%edx # 4034 <t1+0x4> 1155: 0f b6 35 dd 2e 00 00 movzbl 0x2edd(%rip),%esi # 4039 <ti> 115c: 88 44 24 f7 mov %al,-0x9(%rsp) 1160: 48 89 4c 24 f8 mov %rcx,-0x8(%rsp) 1165: 88 54 24 f6 mov %dl,-0xa(%rsp) 1169: 40 88 74 24 f5 mov %sil,-0xb(%rsp) 116e: 90 nop 116f: 31 c0 xor %eax,%eax 1171: c3 ret 1172: 66 90 xchg %ax,%ax
For this case, usdt should work properly. The following is related symbol table:
74: 0000000000004028 8 OBJECT GLOBAL DEFAULT 26 add_ptr 64: 0000000000004030 8 OBJECT GLOBAL DEFAULT 26 t1 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti <=== for '# 4039 <ti>'.
And for llvm21, only one 'ti' in the symbol table.
$ readelf -s usdt_rip | grep ti 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti
The stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000061 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x000000000000116e, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp)
It looks like clang optimized away the following: // The test/usdt_rip.c file static volatile char ti = 0; but gcc does not. But even if the above 'ti' is preserved for clang, clang generated code should still be fine since the argument is -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp).
I think you can file a bug for gcc community.
At 2025-08-13 00:11:45, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 12:02 AM, 赵佳炜 wrote:
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
Could you share the complete source codes and compiler options which reproduce the above result?
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/10/25 1:55 AM, 赵佳炜 wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Did you check relocations? Relocaitons should be able to point exact which symbol.
Thanks, Jiawei Zhao
[...]
Thanks a lot! I'll file a bug for gcc community.
At 2025-08-13 13:24:39, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 7:27 PM, 赵佳炜 wrote:
Sure.
The usdt_rip.c source code:
// the usdt_rip.c file #include <stddef.h> #include <stdint.h> #include "sdt.h" static volatile char ti = 0; int add(int a, int b) { return a + b; } int (*add_ptr)(int, int) = add; struct st{ int a; char b; }; volatile struct st t1 = {.a = 1, .b = 'a'}; static void __always_inline trigger_func() { STAP_PROBE4(usdt_rip, rip_global_var, ti, add_ptr, t1.b, ti); } int main() { trigger_func(); return 0; }The test/usdt_rip.c source code:
// The test/usdt_rip.c file static volatile char ti = 0;The compiler option is `gcc -O2 -g usdt_rip.c test/usdt_rip.c -o usdt_rip`
My compiler version:
$ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr --with-gcc-major-version-only --program-suffix=-13 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib --enable-libphobos-checking=release --with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch --disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr --enable-offload-defaulted --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu --with-build-config=bootstrap-lto-lean --enable-link-serialization=2 Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04)Thanks for the code. I compiled with gcc14 compiler and the below is the asm code for 'main' function:
0000000000401020 <main>: 401020: 90 nop 401021: 31 c0 xor %eax,%eax 401023: c3 ret 401024: 66 2e 0f 1f 84 00 00 cs nopw 0x0(%rax,%rax,1) 40102b: 00 00 00 40102e: 66 90 xchg %ax,%ax
stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000401020, Base: 0x0000000000402010, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@t1+4(%rip) -1@ti(%rip)
$ readelf -s usdt_rip | grep ti 4: 0000000000404019 1 OBJECT LOCAL DEFAULT 24 ti 14: 000000000040401a 1 OBJECT LOCAL DEFAULT 24 ti
So yet, it would be hard to find which 'ti' should be used to resolve usdt arguments.
I then tried with clang21. The 'main' function:
0000000000001140 <main>: 1140: 0f b6 05 f2 2e 00 00 movzbl 0x2ef2(%rip),%eax # 4039 <ti> 1147: 48 8b 0d da 2e 00 00 mov 0x2eda(%rip),%rcx # 4028 <add_ptr> 114e: 0f b6 15 df 2e 00 00 movzbl 0x2edf(%rip),%edx # 4034 <t1+0x4> 1155: 0f b6 35 dd 2e 00 00 movzbl 0x2edd(%rip),%esi # 4039 <ti> 115c: 88 44 24 f7 mov %al,-0x9(%rsp) 1160: 48 89 4c 24 f8 mov %rcx,-0x8(%rsp) 1165: 88 54 24 f6 mov %dl,-0xa(%rsp) 1169: 40 88 74 24 f5 mov %sil,-0xb(%rsp) 116e: 90 nop 116f: 31 c0 xor %eax,%eax 1171: c3 ret 1172: 66 90 xchg %ax,%ax
For this case, usdt should work properly. The following is related symbol table:
74: 0000000000004028 8 OBJECT GLOBAL DEFAULT 26 add_ptr 64: 0000000000004030 8 OBJECT GLOBAL DEFAULT 26 t1 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti <=== for '# 4039 <ti>'.And for llvm21, only one 'ti' in the symbol table.
$ readelf -s usdt_rip | grep ti 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti
The stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000061 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x000000000000116e, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp)
It looks like clang optimized away the following: // The test/usdt_rip.c file static volatile char ti = 0; but gcc does not. But even if the above 'ti' is preserved for clang, clang generated code should still be fine since the argument is -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp).
I think you can file a bug for gcc community.
At 2025-08-13 00:11:45, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 12:02 AM, 赵佳炜 wrote:
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
Could you share the complete source codes and compiler options which reproduce the above result?
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/10/25 1:55 AM, 赵佳炜 wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Did you check relocations? Relocaitons should be able to point exact which symbol.
Thanks, Jiawei Zhao
[...]
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
At 2025-08-13 13:24:39, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 7:27 PM, 赵佳炜 wrote:
Sure.
The usdt_rip.c source code:
// the usdt_rip.c file #include <stddef.h> #include <stdint.h> #include "sdt.h" static volatile char ti = 0; int add(int a, int b) { return a + b; } int (*add_ptr)(int, int) = add; struct st{ int a; char b; }; volatile struct st t1 = {.a = 1, .b = 'a'}; static void __always_inline trigger_func() { STAP_PROBE4(usdt_rip, rip_global_var, ti, add_ptr, t1.b, ti); } int main() { trigger_func(); return 0; }The test/usdt_rip.c source code:
// The test/usdt_rip.c file static volatile char ti = 0;The compiler option is `gcc -O2 -g usdt_rip.c test/usdt_rip.c -o usdt_rip`
My compiler version:
$ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr --with-gcc-major-version-only --program-suffix=-13 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib --enable-libphobos-checking=release --with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch --disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr --enable-offload-defaulted --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu --with-build-config=bootstrap-lto-lean --enable-link-serialization=2 Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04)Thanks for the code. I compiled with gcc14 compiler and the below is the asm code for 'main' function:
0000000000401020 <main>: 401020: 90 nop 401021: 31 c0 xor %eax,%eax 401023: c3 ret 401024: 66 2e 0f 1f 84 00 00 cs nopw 0x0(%rax,%rax,1) 40102b: 00 00 00 40102e: 66 90 xchg %ax,%ax
stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000401020, Base: 0x0000000000402010, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@t1+4(%rip) -1@ti(%rip)
$ readelf -s usdt_rip | grep ti 4: 0000000000404019 1 OBJECT LOCAL DEFAULT 24 ti 14: 000000000040401a 1 OBJECT LOCAL DEFAULT 24 ti
So yet, it would be hard to find which 'ti' should be used to resolve usdt arguments.
I then tried with clang21. The 'main' function:
0000000000001140 <main>: 1140: 0f b6 05 f2 2e 00 00 movzbl 0x2ef2(%rip),%eax # 4039 <ti> 1147: 48 8b 0d da 2e 00 00 mov 0x2eda(%rip),%rcx # 4028 <add_ptr> 114e: 0f b6 15 df 2e 00 00 movzbl 0x2edf(%rip),%edx # 4034 <t1+0x4> 1155: 0f b6 35 dd 2e 00 00 movzbl 0x2edd(%rip),%esi # 4039 <ti> 115c: 88 44 24 f7 mov %al,-0x9(%rsp) 1160: 48 89 4c 24 f8 mov %rcx,-0x8(%rsp) 1165: 88 54 24 f6 mov %dl,-0xa(%rsp) 1169: 40 88 74 24 f5 mov %sil,-0xb(%rsp) 116e: 90 nop 116f: 31 c0 xor %eax,%eax 1171: c3 ret 1172: 66 90 xchg %ax,%ax
For this case, usdt should work properly. The following is related symbol table:
74: 0000000000004028 8 OBJECT GLOBAL DEFAULT 26 add_ptr 64: 0000000000004030 8 OBJECT GLOBAL DEFAULT 26 t1 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti <=== for '# 4039 <ti>'.And for llvm21, only one 'ti' in the symbol table.
$ readelf -s usdt_rip | grep ti 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti
The stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000061 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x000000000000116e, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp)
It looks like clang optimized away the following: // The test/usdt_rip.c file static volatile char ti = 0; but gcc does not. But even if the above 'ti' is preserved for clang, clang generated code should still be fine since the argument is -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp).
I think you can file a bug for gcc community.
At 2025-08-13 00:11:45, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 12:02 AM, 赵佳炜 wrote:
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
Could you share the complete source codes and compiler options which reproduce the above result?
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/10/25 1:55 AM, 赵佳炜 wrote:
Hi Yonghong,
I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart.
For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`.
After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries:
$ readelf -s usdt_rip
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__
As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish.
I’m unsure how to handle this situation. Do you have any suggestions?
Did you check relocations? Relocaitons should be able to point exact which symbol.
Thanks, Jiawei Zhao
[...]
On 8/17/25 6:43 AM, 赵佳炜 wrote:
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
Thanks for filing an issue on gcc and getting some feedback/suggestions from gcc community.
Currently, libbpf/usdt does not suport format like '-1@ti(%rip)'. If we do intend to implement this. libbpf/usdt can reject that if 'ti' is a static variable. libbpf can provide some hints about how to make it work (see above [1] and [2]). Then, it would be user's reponsibility to change code so libbpf can support it.
At 2025-08-13 13:24:39, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 7:27 PM, 赵佳炜 wrote:
Sure.
The usdt_rip.c source code:
// the usdt_rip.c file #include <stddef.h> #include <stdint.h> #include "sdt.h" static volatile char ti = 0; int add(int a, int b) { return a + b; } int (*add_ptr)(int, int) = add; struct st{ int a; char b; }; volatile struct st t1 = {.a = 1, .b = 'a'}; static void __always_inline trigger_func() { STAP_PROBE4(usdt_rip, rip_global_var, ti, add_ptr, t1.b, ti); } int main() { trigger_func(); return 0; }The test/usdt_rip.c source code:
// The test/usdt_rip.c file static volatile char ti = 0;The compiler option is `gcc -O2 -g usdt_rip.c test/usdt_rip.c -o usdt_rip`
My compiler version:
$ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr --with-gcc-major-version-only --program-suffix=-13 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib --enable-libphobos-checking=release --with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch --disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr --enable-offload-defaulted --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu --with-build-config=bootstrap-lto-lean --enable-link-serialization=2 Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04)Thanks for the code. I compiled with gcc14 compiler and the below is the asm code for 'main' function:
0000000000401020 <main>: 401020: 90 nop 401021: 31 c0 xor %eax,%eax 401023: c3 ret 401024: 66 2e 0f 1f 84 00 00 cs nopw 0x0(%rax,%rax,1) 40102b: 00 00 00 40102e: 66 90 xchg %ax,%ax
stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000401020, Base: 0x0000000000402010, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@t1+4(%rip) -1@ti(%rip)
$ readelf -s usdt_rip | grep ti 4: 0000000000404019 1 OBJECT LOCAL DEFAULT 24 ti 14: 000000000040401a 1 OBJECT LOCAL DEFAULT 24 ti
So yet, it would be hard to find which 'ti' should be used to resolve usdt arguments.
I then tried with clang21. The 'main' function:
0000000000001140 <main>: 1140: 0f b6 05 f2 2e 00 00 movzbl 0x2ef2(%rip),%eax # 4039 <ti> 1147: 48 8b 0d da 2e 00 00 mov 0x2eda(%rip),%rcx # 4028 <add_ptr> 114e: 0f b6 15 df 2e 00 00 movzbl 0x2edf(%rip),%edx # 4034 <t1+0x4> 1155: 0f b6 35 dd 2e 00 00 movzbl 0x2edd(%rip),%esi # 4039 <ti> 115c: 88 44 24 f7 mov %al,-0x9(%rsp) 1160: 48 89 4c 24 f8 mov %rcx,-0x8(%rsp) 1165: 88 54 24 f6 mov %dl,-0xa(%rsp) 1169: 40 88 74 24 f5 mov %sil,-0xb(%rsp) 116e: 90 nop 116f: 31 c0 xor %eax,%eax 1171: c3 ret 1172: 66 90 xchg %ax,%ax
For this case, usdt should work properly. The following is related symbol table:
74: 0000000000004028 8 OBJECT GLOBAL DEFAULT 26 add_ptr 64: 0000000000004030 8 OBJECT GLOBAL DEFAULT 26 t1 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti <=== for '# 4039 <ti>'.And for llvm21, only one 'ti' in the symbol table.
$ readelf -s usdt_rip | grep ti 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti
The stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000061 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x000000000000116e, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp)
It looks like clang optimized away the following: // The test/usdt_rip.c file static volatile char ti = 0; but gcc does not. But even if the above 'ti' is preserved for clang, clang generated code should still be fine since the argument is -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp).
I think you can file a bug for gcc community.
At 2025-08-13 00:11:45, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 12:02 AM, 赵佳炜 wrote:
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
Could you share the complete source codes and compiler options which reproduce the above result?
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/10/25 1:55 AM, 赵佳炜 wrote: > Hi Yonghong, > > I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart. > > For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`. > > After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries: > > $ readelf -s usdt_rip > > Symbol table '.symtab' contains 42 entries: > Num: Value Size Type Bind Vis Ndx Name > 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND > 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o > 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag > 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c > 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti > 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c > 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones > 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones > 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux > 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 > 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] > 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy > 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] > 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c > 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti > 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c > 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ > > > As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish. > > I’m unsure how to handle this situation. Do you have any suggestions? Did you check relocations? Relocaitons should be able to point exact which symbol.
> Thanks, > Jiawei Zhao [...]
libbpf/usdt can reject that if 'ti' is a static variable.
But how can we do that? It seems we cannot determine whether a variable is static during the preprocessing phase.
At 2025-08-19 01:35:04, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/17/25 6:43 AM, 赵佳炜 wrote:
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
Thanks for filing an issue on gcc and getting some feedback/suggestions from gcc community.
Currently, libbpf/usdt does not suport format like '-1@ti(%rip)'. If we do intend to implement this. libbpf/usdt can reject that if 'ti' is a static variable. libbpf can provide some hints about how to make it work (see above [1] and [2]). Then, it would be user's reponsibility to change code so libbpf can support it.
At 2025-08-13 13:24:39, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 7:27 PM, 赵佳炜 wrote:
Sure.
The usdt_rip.c source code:
// the usdt_rip.c file #include <stddef.h> #include <stdint.h> #include "sdt.h" static volatile char ti = 0; int add(int a, int b) { return a + b; } int (*add_ptr)(int, int) = add; struct st{ int a; char b; }; volatile struct st t1 = {.a = 1, .b = 'a'}; static void __always_inline trigger_func() { STAP_PROBE4(usdt_rip, rip_global_var, ti, add_ptr, t1.b, ti); } int main() { trigger_func(); return 0; }The test/usdt_rip.c source code:
// The test/usdt_rip.c file static volatile char ti = 0;The compiler option is `gcc -O2 -g usdt_rip.c test/usdt_rip.c -o usdt_rip`
My compiler version:
$ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr --with-gcc-major-version-only --program-suffix=-13 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib --enable-libphobos-checking=release --with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch --disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr --enable-offload-defaulted --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu --with-build-config=bootstrap-lto-lean --enable-link-serialization=2 Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04)Thanks for the code. I compiled with gcc14 compiler and the below is the asm code for 'main' function:
0000000000401020 <main>: 401020: 90 nop 401021: 31 c0 xor %eax,%eax 401023: c3 ret 401024: 66 2e 0f 1f 84 00 00 cs nopw 0x0(%rax,%rax,1) 40102b: 00 00 00 40102e: 66 90 xchg %ax,%ax
stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000401020, Base: 0x0000000000402010, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@t1+4(%rip) -1@ti(%rip)
$ readelf -s usdt_rip | grep ti 4: 0000000000404019 1 OBJECT LOCAL DEFAULT 24 ti 14: 000000000040401a 1 OBJECT LOCAL DEFAULT 24 ti
So yet, it would be hard to find which 'ti' should be used to resolve usdt arguments.
I then tried with clang21. The 'main' function:
0000000000001140 <main>: 1140: 0f b6 05 f2 2e 00 00 movzbl 0x2ef2(%rip),%eax # 4039 <ti> 1147: 48 8b 0d da 2e 00 00 mov 0x2eda(%rip),%rcx # 4028 <add_ptr> 114e: 0f b6 15 df 2e 00 00 movzbl 0x2edf(%rip),%edx # 4034 <t1+0x4> 1155: 0f b6 35 dd 2e 00 00 movzbl 0x2edd(%rip),%esi # 4039 <ti> 115c: 88 44 24 f7 mov %al,-0x9(%rsp) 1160: 48 89 4c 24 f8 mov %rcx,-0x8(%rsp) 1165: 88 54 24 f6 mov %dl,-0xa(%rsp) 1169: 40 88 74 24 f5 mov %sil,-0xb(%rsp) 116e: 90 nop 116f: 31 c0 xor %eax,%eax 1171: c3 ret 1172: 66 90 xchg %ax,%ax
For this case, usdt should work properly. The following is related symbol table:
74: 0000000000004028 8 OBJECT GLOBAL DEFAULT 26 add_ptr 64: 0000000000004030 8 OBJECT GLOBAL DEFAULT 26 t1 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti <=== for '# 4039 <ti>'.And for llvm21, only one 'ti' in the symbol table.
$ readelf -s usdt_rip | grep ti 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti
The stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000061 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x000000000000116e, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp)
It looks like clang optimized away the following: // The test/usdt_rip.c file static volatile char ti = 0; but gcc does not. But even if the above 'ti' is preserved for clang, clang generated code should still be fine since the argument is -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp).
I think you can file a bug for gcc community.
At 2025-08-13 00:11:45, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 12:02 AM, 赵佳炜 wrote:
Yes, I've tried that but it didn't help. FYI:
$ readelf -nsr usdt_rip
Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: Offset Info Type Sym. Value Sym. Name + Addend 000000003df0 000000000008 R_X86_64_RELATIVE 1150 000000003df8 000000000008 R_X86_64_RELATIVE 1110 000000004008 000000000008 R_X86_64_RELATIVE 4008 000000004018 000000000008 R_X86_64_RELATIVE 1160 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0
Symbol table '.dynsym' contains 6 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3)
Symbol table '.symtab' contains 42 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init
Displaying notes found in: .note.gnu.property Owner Data size Description GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 Properties: x86 feature: IBT, SHSTK x86 ISA needed: x86-64-baseline
Displaying notes found in: .note.gnu.build-id Owner Data size Description GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) Build ID: eb615daa575687cc44edc1d339b27890c12c27f1
Displaying notes found in: .note.ABI-tag Owner Data size Description GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) OS: Linux, ABI: 3.2.0
Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip)
Could you share the complete source codes and compiler options which reproduce the above result?
At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote: > On 8/10/25 1:55 AM, 赵佳炜 wrote: >> Hi Yonghong, >> >> I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart. >> >> For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`. >> >> After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries: >> >> $ readelf -s usdt_rip >> >> Symbol table '.symtab' contains 42 entries: >> Num: Value Size Type Bind Vis Ndx Name >> 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND >> 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o >> 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag >> 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c >> 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti >> 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c >> 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones >> 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones >> 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux >> 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 >> 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] >> 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy >> 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] >> 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c >> 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti >> 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c >> 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ >> >> >> As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish. >> >> I’m unsure how to handle this situation. Do you have any suggestions? > Did you check relocations? Relocaitons should be able to point exact which symbol. > >> Thanks, >> Jiawei Zhao > [...]
On 8/19/25 2:50 AM, 赵佳炜 wrote:
libbpf/usdt can reject that if 'ti' is a static variable.
But how can we do that? It seems we cannot determine whether a variable is static during the preprocessing phase.
What I mean is after compilation and during libbpf processing usdt requests. At that time, libbpf can inspect the application binary and all usdt's. My above statement libbpf/usdt can reject that if 'ti' is a static variable. is not not precise, it should be libbpf/usdt can reject if there are ambiguity and provides some hints.
If the preprocessing phase means compilation, yes, we cannot do anything.
At 2025-08-19 01:35:04, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/17/25 6:43 AM, 赵佳炜 wrote:
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
Thanks for filing an issue on gcc and getting some feedback/suggestions from gcc community. Currently, libbpf/usdt does not suport format like '-1@ti(%rip)'. If we do intend to implement this. libbpf/usdt can reject that if 'ti' is a static variable. libbpf can provide some hints about how to make it work (see above [1] and [2]). Then, it would be user's reponsibility to change code so libbpf can support it.
At 2025-08-13 13:24:39, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 7:27 PM, 赵佳炜 wrote:
Sure.
The usdt_rip.c source code:
// the usdt_rip.c file #include <stddef.h> #include <stdint.h> #include "sdt.h" static volatile char ti = 0; int add(int a, int b) { return a + b; } int (*add_ptr)(int, int) = add; struct st{ int a; char b; }; volatile struct st t1 = {.a = 1, .b = 'a'}; static void __always_inline trigger_func() { STAP_PROBE4(usdt_rip, rip_global_var, ti, add_ptr, t1.b, ti); } int main() { trigger_func(); return 0; }The test/usdt_rip.c source code:
// The test/usdt_rip.c file static volatile char ti = 0;The compiler option is `gcc -O2 -g usdt_rip.c test/usdt_rip.c -o usdt_rip`
My compiler version:
$ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr --with-gcc-major-version-only --program-suffix=-13 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace --enable-gnu-unique-object --disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib --enable-libphobos-checking=release --with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch --disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr --enable-offload-defaulted --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu --with-build-config=bootstrap-lto-lean --enable-link-serialization=2 Thread model: posix Supported LTO compression algorithms: zlib zstd gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04)Thanks for the code. I compiled with gcc14 compiler and the below is the asm code for 'main' function:
0000000000401020 <main>: 401020: 90 nop 401021: 31 c0 xor %eax,%eax 401023: c3 ret 401024: 66 2e 0f 1f 84 00 00 cs nopw 0x0(%rax,%rax,1) 40102b: 00 00 00 40102e: 66 90 xchg %ax,%ax
stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x0000000000401020, Base: 0x0000000000402010, Semaphore: 0x0000000000000000 Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@t1+4(%rip) -1@ti(%rip)
$ readelf -s usdt_rip | grep ti 4: 0000000000404019 1 OBJECT LOCAL DEFAULT 24 ti 14: 000000000040401a 1 OBJECT LOCAL DEFAULT 24 ti
So yet, it would be hard to find which 'ti' should be used to resolve usdt arguments.
I then tried with clang21. The 'main' function:
0000000000001140 <main>: 1140: 0f b6 05 f2 2e 00 00 movzbl 0x2ef2(%rip),%eax # 4039 <ti> 1147: 48 8b 0d da 2e 00 00 mov 0x2eda(%rip),%rcx # 4028 <add_ptr> 114e: 0f b6 15 df 2e 00 00 movzbl 0x2edf(%rip),%edx # 4034 <t1+0x4> 1155: 0f b6 35 dd 2e 00 00 movzbl 0x2edd(%rip),%esi # 4039 <ti> 115c: 88 44 24 f7 mov %al,-0x9(%rsp) 1160: 48 89 4c 24 f8 mov %rcx,-0x8(%rsp) 1165: 88 54 24 f6 mov %dl,-0xa(%rsp) 1169: 40 88 74 24 f5 mov %sil,-0xb(%rsp) 116e: 90 nop 116f: 31 c0 xor %eax,%eax 1171: c3 ret 1172: 66 90 xchg %ax,%ax
For this case, usdt should work properly. The following is related symbol table:
74: 0000000000004028 8 OBJECT GLOBAL DEFAULT 26 add_ptr 64: 0000000000004030 8 OBJECT GLOBAL DEFAULT 26 t1 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti <=== for '# 4039 <ti>'.And for llvm21, only one 'ti' in the symbol table.
$ readelf -s usdt_rip | grep ti 50: 0000000000004039 1 OBJECT LOCAL DEFAULT 27 ti
The stapsdt note: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000061 NT_STAPSDT (SystemTap probe descriptors) Provider: usdt_rip Name: rip_global_var Location: 0x000000000000116e, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 Arguments: -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp)
It looks like clang optimized away the following: // The test/usdt_rip.c file static volatile char ti = 0; but gcc does not. But even if the above 'ti' is preserved for clang, clang generated code should still be fine since the argument is -1@-9(%rsp) 8@-8(%rsp) -1@-10(%rsp) -1@-11(%rsp).
I think you can file a bug for gcc community.
At 2025-08-13 00:11:45, "Yonghong Song" yonghong.song@linux.dev wrote:
On 8/12/25 12:02 AM, 赵佳炜 wrote: > Yes, I've tried that but it didn't help. FYI: > > $ readelf -nsr usdt_rip > > > Relocation section '.rela.dyn' at offset 0x530 contains 9 entries: > Offset Info Type Sym. Value Sym. Name + Addend > 000000003df0 000000000008 R_X86_64_RELATIVE 1150 > 000000003df8 000000000008 R_X86_64_RELATIVE 1110 > 000000004008 000000000008 R_X86_64_RELATIVE 4008 > 000000004018 000000000008 R_X86_64_RELATIVE 1160 > 000000003fd8 000100000006 R_X86_64_GLOB_DAT 0000000000000000 __libc_start_main@GLIBC_2.34 + 0 > 000000003fe0 000200000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_deregisterTM[...] + 0 > 000000003fe8 000300000006 R_X86_64_GLOB_DAT 0000000000000000 __gmon_start__ + 0 > 000000003ff0 000400000006 R_X86_64_GLOB_DAT 0000000000000000 _ITM_registerTMCl[...] + 0 > 000000003ff8 000500000006 R_X86_64_GLOB_DAT 0000000000000000 __cxa_finalize@GLIBC_2.2.5 + 0 > > > Symbol table '.dynsym' contains 6 entries: > Num: Value Size Type Bind Vis Ndx Name > 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND > 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND _[...]@GLIBC_2.34 (2) > 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] > 3: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ > 4: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] > 5: 0000000000000000 0 FUNC WEAK DEFAULT UND [...]@GLIBC_2.2.5 (3) > > > Symbol table '.symtab' contains 42 entries: > Num: Value Size Type Bind Vis Ndx Name > 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND > 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o > 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag > 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c > 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti > 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c > 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones > 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones > 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux > 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 > 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] > 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy > 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] > 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c > 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti > 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c > 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ > 17: 0000000000000000 0 FILE LOCAL DEFAULT ABS > 18: 0000000000003e00 0 OBJECT LOCAL DEFAULT 22 _DYNAMIC > 19: 0000000000002008 0 NOTYPE LOCAL DEFAULT 18 __GNU_EH_FRAME_HDR > 20: 0000000000003fc0 0 OBJECT LOCAL DEFAULT 23 _GLOBAL_OFFSET_TABLE_ > 21: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_mai[...] > 22: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_deregisterT[...] > 23: 0000000000004000 0 NOTYPE WEAK DEFAULT 24 data_start > 24: 0000000000001160 8 FUNC GLOBAL DEFAULT 14 add > 25: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 24 _edata > 26: 0000000000002004 1 NOTYPE WEAK HIDDEN 17 _.stapsdt.base > 27: 0000000000004010 8 OBJECT GLOBAL DEFAULT 24 t1 > 28: 0000000000001168 0 FUNC GLOBAL HIDDEN 15 _fini > 29: 0000000000004000 0 NOTYPE GLOBAL DEFAULT 24 __data_start > 30: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ > 31: 0000000000004008 0 OBJECT GLOBAL HIDDEN 24 __dso_handle > 32: 0000000000002000 4 OBJECT GLOBAL DEFAULT 16 _IO_stdin_used > 33: 0000000000004028 0 NOTYPE GLOBAL DEFAULT 25 _end > 34: 0000000000001070 38 FUNC GLOBAL DEFAULT 14 _start > 35: 0000000000004020 0 NOTYPE GLOBAL DEFAULT 25 __bss_start > 36: 0000000000001040 48 FUNC GLOBAL DEFAULT 14 main > 37: 0000000000004018 8 OBJECT GLOBAL DEFAULT 24 add_ptr > 38: 0000000000004020 0 OBJECT GLOBAL HIDDEN 24 __TMC_END__ > 39: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _ITM_registerTMC[...] > 40: 0000000000000000 0 FUNC WEAK DEFAULT UND __cxa_finalize@G[...] > 41: 0000000000001000 0 FUNC GLOBAL HIDDEN 11 _init > > > Displaying notes found in: .note.gnu.property > Owner Data size Description > GNU 0x00000020 NT_GNU_PROPERTY_TYPE_0 > Properties: x86 feature: IBT, SHSTK > x86 ISA needed: x86-64-baseline > > > Displaying notes found in: .note.gnu.build-id > Owner Data size Description > GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring) > Build ID: eb615daa575687cc44edc1d339b27890c12c27f1 > > > Displaying notes found in: .note.ABI-tag > Owner Data size Description > GNU 0x00000010 NT_GNU_ABI_TAG (ABI version tag) > OS: Linux, ABI: 3.2.0 > > > Displaying notes found in: .note.stapsdt > Owner Data size Description > stapsdt 0x00000066 NT_STAPSDT (SystemTap probe descriptors) > Provider: usdt_rip > Name: rip_global_var > Location: 0x0000000000001058, Base: 0x0000000000002004, Semaphore: 0x0000000000000000 > Arguments: -1@ti(%rip) 8@add_ptr(%rip) -1@4+t1(%rip) -1@ti(%rip) Could you share the complete source codes and compiler options which reproduce the above result?
> At 2025-08-12 13:06:40, "Yonghong Song" yonghong.song@linux.dev wrote: >> On 8/10/25 1:55 AM, 赵佳炜 wrote: >>> Hi Yonghong, >>> >>> I found another issue where symbols can be duplicated, and I’m not sure how to tell them apart. >>> >>> For example, I created two C files named usdt_rip.c and hello.c. Both define their own static ti variables, like:`static volatile char ti = 0;`. >>> >>> After compiling, I obtained an ELF file usdt_rip whose .symtab contains the following entries: >>> >>> $ readelf -s usdt_rip >>> >>> Symbol table '.symtab' contains 42 entries: >>> Num: Value Size Type Bind Vis Ndx Name >>> 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND >>> 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS Scrt1.o >>> 2: 000000000000038c 32 OBJECT LOCAL DEFAULT 4 __abi_tag >>> 3: 0000000000000000 0 FILE LOCAL DEFAULT ABS usdt_rip.c >>> 4: 0000000000004021 1 OBJECT LOCAL DEFAULT 25 ti >>> 5: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c >>> 6: 00000000000010a0 0 FUNC LOCAL DEFAULT 14 deregister_tm_clones >>> 7: 00000000000010d0 0 FUNC LOCAL DEFAULT 14 register_tm_clones >>> 8: 0000000000001110 0 FUNC LOCAL DEFAULT 14 __do_global_dtors_aux >>> 9: 0000000000004020 1 OBJECT LOCAL DEFAULT 25 completed.0 >>> 10: 0000000000003df8 0 OBJECT LOCAL DEFAULT 21 __do_global_dtor[...] >>> 11: 0000000000001150 0 FUNC LOCAL DEFAULT 14 frame_dummy >>> 12: 0000000000003df0 0 OBJECT LOCAL DEFAULT 20 __frame_dummy_in[...] >>> 13: 0000000000000000 0 FILE LOCAL DEFAULT ABS damo.c >>> 14: 0000000000004022 1 OBJECT LOCAL DEFAULT 25 ti >>> 15: 0000000000000000 0 FILE LOCAL DEFAULT ABS crtstuff.c >>> 16: 00000000000020d8 0 OBJECT LOCAL DEFAULT 19 __FRAME_END__ >>> >>> >>> As you can see, there are two ti variables in the .symtab section. Their values are very close, making them hard to distinguish. >>> >>> I’m unsure how to handle this situation. Do you have any suggestions? >> Did you check relocations? Relocaitons should be able to point exact which symbol. >> >>> Thanks, >>> Jiawei Zhao >> [...]
On Mon, Aug 18, 2025 at 10:35 AM Yonghong Song yonghong.song@linux.dev wrote:
On 8/17/25 6:43 AM, 赵佳炜 wrote:
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
Thanks for filing an issue on gcc and getting some feedback/suggestions from gcc community.
Currently, libbpf/usdt does not suport format like '-1@ti(%rip)'. If we do
Exactly, it doesn't. I haven't yet ran into a case where real-world applications would use such an argument format, so there was no incentive in trying to support it.
Was this issue discovered as part of testing on some real world application, or it's mostly through testing on synthetic cases?
intend to implement this. libbpf/usdt can reject that if 'ti' is a static variable. libbpf can provide some hints about how to make it work (see above [1] and [2]). Then, it would be user's reponsibility to change code so libbpf can support it.
[...]
In the previous discussion with Yonghong Song, we found that some compiler would generate such an arguement format. Although I have never encounter such an issue, I found that the global volatile variable could trigger the compiler to generate this argument spec. So I tried to solve this problem. I guess this would not be a problem since we have already used STAP_PROBE_ASM to reliably generate SIB argument spec.
BTW, I have another issue to discuss.
Now, bcc framework is not a recommendation for writing bpf program, so bpftrace is now migrating from bcc framework to libbpf. Bcc framework provides some relevant APIs for get usdt probe info[1]. And I found that there is not similar APIs in libbpf, therefore I have to parse elf file manually.
Could we add some relevant APIs, maybe like `bpf_program__usdt_probe_list`, in libbpf? I can make a patch to implement it. WDYT?
[1]. https://github.com/bpftrace/bpftrace/blob/1cd4bbdd4a13dd55880f2cc638dde641fb...
At 2025-08-21 07:00:35, "Andrii Nakryiko" andrii.nakryiko@gmail.com wrote:
On Mon, Aug 18, 2025 at 10:35 AM Yonghong Song yonghong.song@linux.dev wrote:
On 8/17/25 6:43 AM, 赵佳炜 wrote:
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
Thanks for filing an issue on gcc and getting some feedback/suggestions from gcc community.
Currently, libbpf/usdt does not suport format like '-1@ti(%rip)'. If we do
Exactly, it doesn't. I haven't yet ran into a case where real-world applications would use such an argument format, so there was no incentive in trying to support it.
Was this issue discovered as part of testing on some real world application, or it's mostly through testing on synthetic cases?
intend to implement this. libbpf/usdt can reject that if 'ti' is a static variable. libbpf can provide some hints about how to make it work (see above [1] and [2]). Then, it would be user's reponsibility to change code so libbpf can support it.
[...]
On Thu, Aug 21, 2025 at 8:38 AM 赵佳炜 phoenix500526@163.com wrote:
In the previous discussion with Yonghong Song, we found that some compiler would generate such an arguement format. Although I have never encounter such an issue, I found that the global volatile variable could trigger the compiler to generate this argument spec. So I tried to solve this problem. I guess this would not be a problem since we have already used STAP_PROBE_ASM to reliably generate SIB argument spec.
Yep, let's hold off on implementing this, as it doesn't seem to be really necessary in practice.
BTW, I have another issue to discuss.
Now, bcc framework is not a recommendation for writing bpf program, so bpftrace is now migrating from bcc framework to libbpf. Bcc framework provides some relevant APIs for get usdt probe info[1]. And I found that there is not similar APIs in libbpf, therefore I have to parse elf file manually.
Could we add some relevant APIs, maybe like `bpf_program__usdt_probe_list`, in libbpf? I can make a patch to implement it. WDYT?
I'm not yet convinced this belongs in libbpf, tbh. The process of discovering USDTs actually involved two separate tasks: discovering binaries (executable and shared libs) that constitute a process, and then for each binary discovering which USDTs belong to it. The first part is pretty clearly out of scope for libbpf. Second one in principle can be added, but as I said I'm a bit hesitant at this point, as we don't even have "USDT manager" exposed through public API.
So not yet.
[1]. https://github.com/bpftrace/bpftrace/blob/1cd4bbdd4a13dd55880f2cc638dde641fb...
At 2025-08-21 07:00:35, "Andrii Nakryiko" andrii.nakryiko@gmail.com wrote:
On Mon, Aug 18, 2025 at 10:35 AM Yonghong Song yonghong.song@linux.dev wrote:
On 8/17/25 6:43 AM, 赵佳炜 wrote:
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
Thanks for filing an issue on gcc and getting some feedback/suggestions from gcc community.
Currently, libbpf/usdt does not suport format like '-1@ti(%rip)'. If we do
Exactly, it doesn't. I haven't yet ran into a case where real-world applications would use such an argument format, so there was no incentive in trying to support it.
Was this issue discovered as part of testing on some real world application, or it's mostly through testing on synthetic cases?
intend to implement this. libbpf/usdt can reject that if 'ti' is a static variable. libbpf can provide some hints about how to make it work (see above [1] and [2]). Then, it would be user's reponsibility to change code so libbpf can support it.
[...]
Ok, I see. I file an issue in libbpf repo to track this discussion.
FYI: https://github.com/libbpf/libbpf/issues/918
At 2025-08-22 02:28:31, "Andrii Nakryiko" andrii.nakryiko@gmail.com wrote:
On Thu, Aug 21, 2025 at 8:38 AM 赵佳炜 phoenix500526@163.com wrote:
In the previous discussion with Yonghong Song, we found that some compiler would generate such an arguement format. Although I have never encounter such an issue, I found that the global volatile variable could trigger the compiler to generate this argument spec. So I tried to solve this problem. I guess this would not be a problem since we have already used STAP_PROBE_ASM to reliably generate SIB argument spec.
Yep, let's hold off on implementing this, as it doesn't seem to be really necessary in practice.
BTW, I have another issue to discuss.
Now, bcc framework is not a recommendation for writing bpf program, so bpftrace is now migrating from bcc framework to libbpf. Bcc framework provides some relevant APIs for get usdt probe info[1]. And I found that there is not similar APIs in libbpf, therefore I have to parse elf file manually.
Could we add some relevant APIs, maybe like `bpf_program__usdt_probe_list`, in libbpf? I can make a patch to implement it. WDYT?
I'm not yet convinced this belongs in libbpf, tbh. The process of discovering USDTs actually involved two separate tasks: discovering binaries (executable and shared libs) that constitute a process, and then for each binary discovering which USDTs belong to it. The first part is pretty clearly out of scope for libbpf. Second one in principle can be added, but as I said I'm a bit hesitant at this point, as we don't even have "USDT manager" exposed through public API.
So not yet.
[1]. https://github.com/bpftrace/bpftrace/blob/1cd4bbdd4a13dd55880f2cc638dde641fb...
At 2025-08-21 07:00:35, "Andrii Nakryiko" andrii.nakryiko@gmail.com wrote:
On Mon, Aug 18, 2025 at 10:35 AM Yonghong Song yonghong.song@linux.dev wrote:
On 8/17/25 6:43 AM, 赵佳炜 wrote:
Hi, Yonghong. I've already filed an issue[1] in GCC community.
Accroding to the discussion, it's not a gcc bug but may be a systemtap bug. I don't know how to report this bug to systemtap, but I found that the libbpf/usdt have the same problem. I've filed an issue in libbpf/usdt repo[2].
I also have some ideas about it. I wrote it down in the issue[2] comment. May be we can discuss there.
[1]. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121569 [2]. https://github.com/libbpf/usdt/issues/13
Thanks for filing an issue on gcc and getting some feedback/suggestions from gcc community.
Currently, libbpf/usdt does not suport format like '-1@ti(%rip)'. If we do
Exactly, it doesn't. I haven't yet ran into a case where real-world applications would use such an argument format, so there was no incentive in trying to support it.
Was this issue discovered as part of testing on some real world application, or it's mostly through testing on synthetic cases?
intend to implement this. libbpf/usdt can reject that if 'ti' is a static variable. libbpf can provide some hints about how to make it work (see above [1] and [2]). Then, it would be user's reponsibility to change code so libbpf can support it.
[...]
On Wed, Aug 06, 2025 at 11:17:34AM -0700, Yonghong Song wrote:
SNIP
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index 910d8d6402ef..68cf6a9cf05f 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -759,6 +759,14 @@ TRUNNER_BPF_BUILD_RULE := $$(error no BPF objects should be built) TRUNNER_BPF_CFLAGS := $(eval $(call DEFINE_TEST_RUNNER,test_maps)) +# Use -O2 optimization to generate SIB addressing usdt argument spec +# Only apply on x86 architecture where SIB addressing is relevant +ifeq ($(ARCH), x86) +$(OUTPUT)/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/cpuv4/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +$(OUTPUT)/no_alu32/usdt_o2.test.o: CFLAGS:=$(subst O0,O2,$(CFLAGS)) +endif
I tried your selftest with gcc14 and llvm20 in my environment. See below:
llvm20: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x0000002f NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x00000000000003ac, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@-64(%rbp)
gcc14: Displaying notes found in: .note.stapsdt Owner Data size Description stapsdt 0x00000034 NT_STAPSDT (SystemTap probe descriptors) Provider: test Name: usdt1 Location: 0x0000000000000334, Base: 0x0000000000000000, Semaphore: 0x0000000000000000 Arguments: 8@array(,%rax,8)
llvm20 and gcc14 generate different usdt patterns. '8@-64(%rbp)' already supports so with SIB support, the test should pass CI, I think.
I see the same with gcc 15 and clang 20
Arguments: 8@array(,%rax,8)
jirka
On 8/6/25 2:24 AM, Jiawei Zhao wrote:
When using GCC on x86-64 to compile an usdt prog with -O1 or higher optimization, the compiler will generate SIB addressing mode for global array and PC-relative addressing mode for global variable, e.g. "1@-96(%rbp,%rax,8)" and "-1@4+t1(%rip)".
The current USDT implementation in libbpf cannot parse these two formats, causing `bpf_program__attach_usdt()` to fail with -ENOENT (unrecognized register).
This patch series adds support for SIB addressing mode in USDT probes. The main changes include:
- add correct handling logic for SIB-addressed arguments in `parse_usdt_arg`.
- add an usdt_o2 test case to cover SIB addressing mode.
Testing shows that the SIB probe correctly generates 8@(%rcx,%rax,8) argument spec and passes all validation checks.
The modification history of this patch series: Change since v1:
- refactor the code to make it more readable
- modify the commit message to explain why and how
Change since v2:
- fix the `scale` uninitialized error
Change since v3:
- force -O2 optimization for usdt.test.o to generate SIB addressing usdt and pass all test cases.
Change since v4:
- split the patch into two parts, one for the fix and the other for the test
Change since v5:
- Only enable optimization for x86 architecture to generate SIB addressing usdt argument spec.
Change since v6:
- Add an usdt_o2 test case to cover SIB addressing mode.
- Reinstate the usdt.c test case.
Do we need to add support for PC-relative USDT argument spec handling in libbpf? I have some interest in this question, but currently have no ideas. Getting offsets based on symbols requires dependency on the symbol table. However, once the binary file is stripped, the symtab will also be removed, which will cause this approach to fail. Does anyone have any thoughts on this?
Jiawei Zhao (2): libbpf: fix USDT SIB argument handling causing unrecognized register error selftests/bpf: Force -O2 for USDT selftests to cover SIB handling logic
tools/lib/bpf/usdt.bpf.h | 33 ++++++++- tools/lib/bpf/usdt.c | 43 +++++++++-- tools/testing/selftests/bpf/Makefile | 8 +++ .../selftests/bpf/prog_tests/usdt_o2.c | 71 +++++++++++++++++++ .../selftests/bpf/progs/test_usdt_o2.c | 37 ++++++++++ 5 files changed, 185 insertions(+), 7 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/usdt_o2.c create mode 100644 tools/testing/selftests/bpf/progs/test_usdt_o2.c
Please add proper tag like [PATCH bpf-next v7 ...] so CI can test it properly.
linux-kselftest-mirror@lists.linaro.org
-
Andrii Nakryiko -
Jiawei Zhao -
Jiri Olsa -
Yonghong Song -
赵佳炜