utils: Add helpers for runtime dynamic patching #1705
Conversation
clementguidi
force-pushed
the
utils-runtime-dynamic-patching
branch
from
3aff835
to
b748861
Compare
| continue; | ||
| } | ||
|
|
||
| if (!oldact.sa_handler) |
There was a problem hiding this comment.
Also this can be racy if other thread installs a new sighandler at the same time. But there's not much we can do..
There was a problem hiding this comment.
True, and furthermore, there is no guarantee that we're going to keep this signal to ourselves even if we don't get into a race condition here. The application can change our signal's disposition at any other point without notice. However, this function is only triggered upon dynamic runtime patching and unpatching via (do_dynamic_update) which we expect to be 1) infrequent 2) not happening at any particular time, e.g., application startup. Hence, the likelihood of this condition is low.
There was a problem hiding this comment.
Right, then we need to take care of releasing the rt signal after use.
Glibc < 2.30 doen't provide wrappers for 'gettid()' and 'tgkill()' so we define them. Signed-off-by: Clément Guidi <[email protected]>
Functions to setup real-time signals and broadcast signals to all threads in an application. Useful for runtime synchronization mechanisms. Co-authored-by: Gabriel-Andrew Pollo-Guilbert <[email protected]> Signed-off-by: Clément Guidi <[email protected]>
clementguidi
force-pushed
the
utils-runtime-dynamic-patching
branch
from
b748861
to
0398ceb
Compare
| if (errno != ESRCH) { | ||
| pr_dbg2("cannot signal thread %u: %s\n", task, strerror(errno)); | ||
| errno = 0; | ||
| } |
There was a problem hiding this comment.
It seems we should not take ESRCH as an error. Something like this?
if (tgkill(pid, task, sig) < 0) {
if (errno != ESRCH)
pr_dbg2(...);
errno = 0;
}| continue; | ||
| } | ||
|
|
||
| if (!oldact.sa_handler) |
There was a problem hiding this comment.
Right, then we need to take care of releasing the rt signal after use.
This is the fourth PR in a series of patches intended to bring runtime dynamic tracing on x86_64 to uftrace.
We add syscall wrappers and signal-related functions that will be useful for runtime dynamic instrumentation.
Related: #1698