Skip to contentSkip to conversation

uprobe: Fix uretprobe syscall wiring

PATCH In review Jiri Olsa 6 participants 9 messages Jul 12, 2024 – Jul 24, 2024

Subsystems (7)
x86 Kselftest BPF Tracing +3 more subsystems
lore
Share & tools
lore

Jiri Olsa

author reviewer
#0 lore raw

cover uprobe: Fix uretprobe syscall wiring

hi,
the uretprobe syscall clashed in linux-next tree with xattrat syscalls, so after discussing with Arnd, changing the syscall number to 467.

I'm also changing the ABI to 'common' which will ease up the global scripts/syscall.tbl management.

I split the change into syscall_64.tbl and selftest change, but it could be merged together if needed.

The patches are based on:
https://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace.git
probes/for-next

thanks,
jirka

---
Jiri Olsa (2):
uprobe: Change uretprobe syscall scope and number selftests/bpf: Change uretprobe syscall number in uprobe_syscall test

 arch/x86/entry/syscalls/syscall_64.tbl                  | 2 +-
 tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
author reviewer
#1 lore raw

PATCH 1/2 uprobe: Change uretprobe syscall scope and number

After discussing with Arnd [1] it's preferable to change uretprobe syscall number to 467 to omit the merge conflict with xattrat syscalls.

Also changing the ABI to 'common' which will ease up the global scripts/syscall.tbl management. One consequence is we generate uretprobe syscall numbers for ABIs that do not support uretprobe syscall, but the syscall still returns -ENOSYS when called in that ABI.

[1] https://lore.kernel.org/lkml/[email protected]/

Fixes 190fec72df4a ("uprobe: Wire up uretprobe system call")
---
 arch/x86/entry/syscalls/syscall_64.tbl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/entry/syscalls/syscall_64.tbl +1 -1
index 6452c2ec469a..dabf1982de6d 100644
@@ -384,7 +384,7 @@
 460	common	lsm_set_self_attr	sys_lsm_set_self_attr
 461	common	lsm_list_modules	sys_lsm_list_modules
 462 	common  mseal			sys_mseal
-463	64	uretprobe		sys_uretprobe
+467	common	uretprobe		sys_uretprobe
 
 #
 # Due to a historical design error, certain syscalls are numbered differently
Signature · 1 line
2.45.2
Acked-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] uprobe: Change uretprobe syscall scope and number

Quoted 9 lines · Jiri Olsa

After discussing with Arnd [1] it's preferable to change uretprobe syscall number to 467 to omit the merge conflict with xattrat syscalls.

Also changing the ABI to 'common' which will ease up the global scripts/syscall.tbl management. One consequence is we generate uretprobe syscall numbers for ABIs that do not support uretprobe syscall, but the syscall still returns -ENOSYS when called in that ABI.

Quoted 9 lines

[1] https://lore.kernel.org/lkml/[email protected]/

Fixes 190fec72df4a ("uprobe: Wire up uretprobe system call")
---
 arch/x86/entry/syscalls/syscall_64.tbl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

sure, why not

Quoted 16 lines
arch/x86/entry/syscalls/syscall_64.tbl +1 -2
index 6452c2ec469a..dabf1982de6d 100644
@@ -384,7 +384,7 @@
 460    common  lsm_set_self_attr       sys_lsm_set_self_attr
 461    common  lsm_list_modules        sys_lsm_list_modules
 462    common  mseal                   sys_mseal
-463    64      uretprobe               sys_uretprobe
+467    common  uretprobe               sys_uretprobe
 
 #
 # Due to a historical design error, certain syscalls are numbered differently
--

2.45.2

#3 lore raw

Re: [PATCH 1/2] uprobe: Change uretprobe syscall scope and number

In reply to Jiri Olsa
Quoted 22 lines · Jiri Olsa

After discussing with Arnd [1] it's preferable to change uretprobe syscall number to 467 to omit the merge conflict with xattrat syscalls.

Also changing the ABI to 'common' which will ease up the global scripts/syscall.tbl management. One consequence is we generate uretprobe syscall numbers for ABIs that do not support uretprobe syscall, but the syscall still returns -ENOSYS when called in that ABI.

[1] https://lore.kernel.org/lkml/[email protected]/

Fixes 190fec72df4a ("uprobe: Wire up uretprobe system call")
---
 arch/x86/entry/syscalls/syscall_64.tbl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/entry/syscalls/syscall_64.tbl +0 -0
index 6452c2ec469a..dabf1982de6d 100644

arch/x86/entry/syscalls/syscall_64.tbl line 390

Quoted 8 lines
arch/x86/entry/syscalls/syscall_64.tbl +1 -1
@@ -384,7 +384,7 @@
 460	common	lsm_set_self_attr	sys_lsm_set_self_attr
 461	common	lsm_list_modules	sys_lsm_list_modules
 462 	common  mseal			sys_mseal
-463	64	uretprobe		sys_uretprobe
+467	common	uretprobe		sys_uretprobe
 
 #
 # Due to a historical design error, certain syscalls are numbered differently

Isn't include/uapi/asm-generic/unistd.h expected to be updated as well? As of mainline commit v6.10-12246-g786c8248dbd3, it still contains

#define __NR_uretprobe 463

Signature · 1 line
ldv
#4 lore raw

Re: [PATCH 1/2] uprobe: Change uretprobe syscall scope and number

Quoted 12 lines · Dmitry V. Levin
Quoted 11 lines · Jiri Olsa
arch/x86/entry/syscalls/syscall_64.tbl +1 -1
index 6452c2ec469a..dabf1982de6d 100644
@@ -384,7 +384,7 @@
 460	common	lsm_set_self_attr	sys_lsm_set_self_attr
 461	common	lsm_list_modules	sys_lsm_list_modules
 462 	common  mseal			sys_mseal
-463	64	uretprobe		sys_uretprobe
+467	common	uretprobe		sys_uretprobe
 
Quoted 7 lines

arch/x86/entry/syscalls/syscall_64.tbl line 390

Quoted 2 lines
arch/x86/entry/syscalls/syscall_64.tbl +0 -0
 #
 # Due to a historical design error, certain syscalls are numbered differently

Isn't include/uapi/asm-generic/unistd.h expected to be updated as well? As of mainline commit v6.10-12246-g786c8248dbd3, it still contains

#define __NR_uretprobe 463

The file is currently unused and replaced with scripts/syscall.tbl, my plan was to remove the old file in the 6.12 syscall cleanups.

The number in scripts/syscall.tbl is now 467, so its users (arc, arm64, csky, hegagon, loongarch, nios2 openrisc and riscv) have the same number as on x86.
However, the corresponding change did not make it into the other syscall.tbl files (alpha, arm, m68k, microblaze, parisc, powerpc, s390, sh, sparc and xtensa), which is rather inconsistent.

I think we should definitely make all non-x86 architectures behave the same way, either with or without an entry for uretprobe. There are three ways do do this:

a) remove it from both include/uapi/asm/unistd.h and scripts/syscall.tbl, and change the x86-64 system call to a private number such as 335

b) remove it from both include/uapi/asm/unistd.h and scripts/syscall.tbl, but leave the number at 467

c) add the syscall to all other architectures for consistency, but continue to have it return -ENOSYS.

From Linus' earlier comments, I would guess that a) would be the least bad of those. I'm also unsure about the status of the xattrat patches, which were the reason for changing uretprobe from 463 to 467. Those patches are still not merged either, and disappeared from linux-next between Friday and Monday.

Arnd

author reviewer
#5 lore raw

PATCH 2/2 selftests/bpf: Change uretprobe syscall number in uprobe_syscall test

In reply to Jiri Olsa

Fixing the syscall number value.

Fixes 9e7f74e64ae5 ("selftests/bpf: Add uretprobe syscall call from user space test")
---
 tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c +1 -1
index c8517c8f5313..bd8c75b620c2 100644
@@ -216,7 +216,7 @@ static void test_uretprobe_regs_change(void)
 }
 
 #ifndef __NR_uretprobe
-#define __NR_uretprobe 463
+#define __NR_uretprobe 467
 #endif
 
 __naked unsigned long uretprobe_syscall_call_1(void)
Signature · 1 line
2.45.2
Acked-by on v1 2/2
#6 lore raw

Re: [PATCH 2/2] selftests/bpf: Change uretprobe syscall number in uprobe_syscall test

Quoted 9 lines · Jiri Olsa

Fixing the syscall number value.

Fixes 9e7f74e64ae5 ("selftests/bpf: Add uretprobe syscall call from user space test")
---
 tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

is this selftest in probes/for-next already? If yes, I'd combine these two patches to avoid any bisection problems

but either way

Quoted 16 lines
tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c +1 -2
index c8517c8f5313..bd8c75b620c2 100644
@@ -216,7 +216,7 @@ static void test_uretprobe_regs_change(void)
 }
 
 #ifndef __NR_uretprobe
-#define __NR_uretprobe 463
+#define __NR_uretprobe 467
 #endif
 
 __naked unsigned long uretprobe_syscall_call_1(void)
--

2.45.2

#7 lore raw

Re: [PATCH 2/2] selftests/bpf: Change uretprobe syscall number in uprobe_syscall test

Quoted 11 lines · On Fri, 12 Jul 2024 11:27:30 -0700 Andrii Nakryiko <[email protected]>…
Quoted 9 lines · Jiri Olsa

Fixing the syscall number value.

Fixes 9e7f74e64ae5 ("selftests/bpf: Add uretprobe syscall call from user space test")
---
 tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 6 lines

is this selftest in probes/for-next already? If yes, I'd combine these two patches to avoid any bisection problems

but either way

Thanks, let me pick it to for-next branch.

Thank you,

Quoted 18 lines
Quoted 18 lines
tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c +1 -2
index c8517c8f5313..bd8c75b620c2 100644
@@ -216,7 +216,7 @@ static void test_uretprobe_regs_change(void)
 }
 
 #ifndef __NR_uretprobe
-#define __NR_uretprobe 463
+#define __NR_uretprobe 467
 #endif
 
 __naked unsigned long uretprobe_syscall_call_1(void)
--

2.45.2

Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
#8 lore raw

Re: [PATCH 2/2] selftests/bpf: Change uretprobe syscall number in uprobe_syscall test

Quoted 14 lines · Masami Hiramatsu
Quoted 11 lines · On Fri, 12 Jul 2024 11:27:30 -0700 Andrii Nakryiko <[email protected]>…
Quoted 9 lines · Jiri Olsa

Fixing the syscall number value.

Fixes 9e7f74e64ae5 ("selftests/bpf: Add uretprobe syscall call from user space test")
---
 tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 2 lines
Quoted 2 lines

is this selftest in probes/for-next already? If yes, I'd combine these two patches to avoid any bisection problems

yes it's all there.. I don't mind squashing it, I just did not want to combine kernel and user space parts.. up to Masami I guess

Quoted 6 lines
Quoted 4 lines

but either way

Thanks, let me pick it to for-next branch.

thanks,
jirka

Quoted 25 lines

Thank you,

Quoted 18 lines
Quoted 18 lines
tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c +1 -2
index c8517c8f5313..bd8c75b620c2 100644
@@ -216,7 +216,7 @@ static void test_uretprobe_regs_change(void)
 }
 
 #ifndef __NR_uretprobe
-#define __NR_uretprobe 463
+#define __NR_uretprobe 467
 #endif
 
 __naked unsigned long uretprobe_syscall_call_1(void)
--

2.45.2

Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>