2023-05-19 08:52:01

by sunliming

Subject: [PATCH RESEND] tracing/user_events: Prevent same name event per process

User processes register name_args for events. If the same name are registered
multiple times in the same process, it can cause undefined behavior. Because
the same name may be used for a diffrent event. If this event has the same
format as the original event, it is impossible to distinguish the trace output
of these two events. If the event has a different format from the original event,
the trace output of the new event is incorrect.

Return EADDRINUSE back to the user process if the same event has being registered
in the same process.

Signed-off-by: sunliming <[email protected]>
---
kernel/trace/trace_events_user.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/trace/trace_events_user.c b/kernel/trace/trace_events_user.c
index b1ecd7677642..4ef6bdb5c07c 100644
--- a/kernel/trace/trace_events_user.c
+++ b/kernel/trace/trace_events_user.c
@@ -1996,7 +1996,7 @@ static int user_events_ref_add(struct user_event_file_info *info,

for (i = 0; i < count; ++i)
if (refs->events[i] == user)
- return i;
+ return -EADDRINUSE;
}

size = struct_size(refs, events, count + 1);
--
2.25.1

2023-05-19 23:54:38

by Beau Belgrave

Subject: Re: [PATCH RESEND] tracing/user_events: Prevent same name event per process

> User processes register name_args for events. If the same name are registered
> multiple times in the same process, it can cause undefined behavior. Because
> the same name may be used for a diffrent event. If this event has the same
> format as the original event, it is impossible to distinguish the trace output
> of these two events. If the event has a different format from the original event,
> the trace output of the new event is incorrect.
>
> Return EADDRINUSE back to the user process if the same event has being registered
> in the same process.
>
> Signed-off-by: sunliming <[email protected]>
> ---

Thank you for the patch, a few comments.

> kernel/trace/trace_events_user.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/trace/trace_events_user.c b/kernel/trace/trace_events_user.c
> index b1ecd7677642..4ef6bdb5c07c 100644
> --- a/kernel/trace/trace_events_user.c
> +++ b/kernel/trace/trace_events_user.c
> @@ -1996,7 +1996,7 @@ static int user_events_ref_add(struct user_event_file_info *info,
>
> for (i = 0; i < count; ++i)
> if (refs->events[i] == user)
> - return i;
> + return -EADDRINUSE;

The ABI currently allows this to allow for shared libraries to share a common write
FD. Do you feel strongly that this should not be allowed?

If you and others feel this should not be allowed, I'd kindly ask you to also update
the self-tests at tools/testing/selftests/user_events/ in the tree and ensure they
pass fully.

For example, see ftrace_test.c at the above tree location:
...

TEST_F(user, register_events) {

...


/* Multiple registers to the same addr + bit should fail */
ASSERT_EQ(-1, ioctl(self->data_fd, DIAG_IOCSREG, &reg));
ASSERT_EQ(EADDRINUSE, errno);

/* Multiple registers to same name should result in same index */
reg.enable_bit = 30;
ASSERT_EQ(0, ioctl(self->data_fd, DIAG_IOCSREG, &reg));
ASSERT_EQ(0, reg.write_index);

...

Thanks,
-Beau

> }
>
> size = struct_size(refs, events, count + 1);
> --
> 2.25.1