This was during the early days of the program I was implementing, I
later found out I needed PTRACE_O_TRACESYSGOOD anyways as the manual
suggests. Still it's something that should be documented! Caused a lot
of frustration.
Regarding your suggestion, it's great but I have something to point out.
In the case of PTRACE_EVENT_SECCOMP the union is filled normally,
without any need for the PTRACE_O_TRACESYSGOOD option to be set. So
unless I am missing something, please consider including that!
On 24/2/23 20:53, Dmitry V. Levin wrote:
On Fri, Feb 24, 2023 at 06:31:42PM +0200, Fotios Valasiadis wrote:
From: Fotios Valasiadis <fvalasiad@xxxxxxxxxxx>
Fixed incomplete doc. PTRACE_GET_SYSCALL_INFO's description wouldn't inform the user that they need to enable PTRACE_O_TRACESYSGOOD to get detailed information from said operation.
Came upon this bug after writing a test program using PTRACE_O_TRACESYSGOOD. After failing to find what's wrong I posted a stackoverflow question which you can find right here: https://stackoverflow.com/questions/72410182/ptrace-get-syscall-info-always-returns-info-op-as-ptrace-syscall-info-none
Nate Eldredge found out what's wrong by looking into the kernel's source code, here is a link to the relevant part https://github.com/torvalds/linux/blob/8291eaafed36f575f23951f3ce18407f480e9ecf/kernel/ptrace.c#L1018
In the code it can be seen that the union is filled if and only if the signal matches "SIGTRAP | 0x80", a signal which is only sent if the PTRACE_O_TRACESYSGOOD option is set. You can read about that in the PTRACE_O_TRACESYSGOOD section of ptrace.2 's manual.
I am unaware if Nate wants to be included as a co-author of this commit, nor am I aware of any way to contact him.
Cc'ed Nate just in case.
Signed-off-by: Fotios Valasiadis <fvalasiad@xxxxxxxxx>
---
man2/ptrace.2 | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/man2/ptrace.2 b/man2/ptrace.2
index 55d9fd36d..b437b4f44 100644
--- a/man2/ptrace.2
+++ b/man2/ptrace.2
@@ -1033,6 +1033,11 @@ If the size of the data to be written by the kernel exceeds the size
specified by the
.I addr
argument, the output data is truncated.
+To obtain complete information during system call entry or exit stops,
+you should enable the
+.B PTRACE_O_TRACESYSGOOD
+option using
+.B PTRACE_SETOPTIONS.
.IP
The
.I ptrace_syscall_info
I suggest the following wording:
diff --git a/man2/ptrace.2 b/man2/ptrace.2
index 55d9fd36d..496d8fae0 100644
--- a/man2/ptrace.2
+++ b/man2/ptrace.2
@@ -1111,6 +1111,15 @@ stop.
.B PTRACE_SYSCALL_INFO_NONE
No component of the union contains relevant information.
.RE
+.IP
+Note that the data returned by
+.B PTRACE_GET_SYSCALL_INFO
+is limited to type
+.B
+PTRACE_SYSCALL_INFO_NONE
+unless
+.B PTRACE_O_TRACESYSGOOD
+option is set before the corresponding ptrace stop has occurred.
.\"
.SS Death under ptrace
When a (possibly multithreaded) process receives a killing signal
As written in the manual page, using the PTRACE_O_TRACESYSGOOD option is
the recommended method to distinguish syscall-stops from other kinds of
ptrace-stops, since it is reliable and does not incur a performance
penalty. In other words, if you use ptrace(2) at all, please consider
using PTRACE_O_TRACESYSGOOD option.
