You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@nuttx.apache.org by "patacongo (via GitHub)" <gi...@apache.org> on 2023/07/28 13:58:42 UTC

[GitHub] [nuttx] patacongo commented on a diff in pull request #9923: ltp: sigprocmask fix

patacongo commented on code in PR #9923:
URL: https://github.com/apache/nuttx/pull/9923#discussion_r1277560626


##########
sched/signal/sig_procmask.c:
##########
@@ -187,8 +187,27 @@ int nxsig_procmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 
 int sigprocmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 {
+  sigset_t nset;
   int ret;
 
+  /* SIGKILL and SIGSTOP should not be added to signal mask */
+
+  if (set != NULL)
+    {
+      nset = *set;
+      if (nxsig_ismember(&nset, SIGKILL))
+        {
+          nxsig_delset(&nset, SIGKILL);
+        }
+
+      if (nxsig_ismember(&nset, SIGSTOP))
+        {
+          nxsig_delset(&nset, SIGSTOP);
+        }
+

Review Comment:
   The nxsig_delset() is rather expensive; nxsig_member not so much.  The code is fine here, but you might want to consider using a single nxsig_nandset.  I am not sure if that is an improvment or not.



##########
sched/signal/sig_procmask.c:
##########
@@ -187,8 +187,27 @@ int nxsig_procmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 
 int sigprocmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 {
+  sigset_t nset;
   int ret;
 
+  /* SIGKILL and SIGSTOP should not be added to signal mask */
+
+  if (set != NULL)
+    {
+      nset = *set;
+      if (nxsig_ismember(&nset, SIGKILL))
+        {
+          nxsig_delset(&nset, SIGKILL);
+        }
+
+      if (nxsig_ismember(&nset, SIGSTOP))
+        {
+          nxsig_delset(&nset, SIGSTOP);
+        }
+

Review Comment:
   In the event of an error in some other part of the signal handling logic, if SIGKILL or SIGSTOP ever get set in the sigprocmask, they may never be cleared.  They can be cleared with SIG_SETMASK but that is not often used after initialization.
   
   Wouldn't it be safer and more robust to mask out SIGKILL and SIGSTOP _after_ the new sigprocmaks (or pthread sigprocmask) is set.



##########
sched/signal/sig_procmask.c:
##########
@@ -187,8 +187,27 @@ int nxsig_procmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 
 int sigprocmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 {
+  sigset_t nset;
   int ret;
 
+  /* SIGKILL and SIGSTOP should not be added to signal mask */
+
+  if (set != NULL)
+    {
+      nset = *set;
+      if (nxsig_ismember(&nset, SIGKILL))
+        {
+          nxsig_delset(&nset, SIGKILL);
+        }
+
+      if (nxsig_ismember(&nset, SIGSTOP))
+        {
+          nxsig_delset(&nset, SIGSTOP);
+        }
+
+      set = &nset;
+    }
+
   /* Let nxsig_procmask do all of the work */
 
   ret = nxsig_procmask(how, set, oset);

Review Comment:
   NOTE: While a signal handler is running, all signals are blocked (as I recall), including SIGKILL and SIGSTOP) but are restored when the signal handler returns.  This is a case where it is necessary to set even these signals to avoid unhandled re-entrancy.



##########
sched/signal/sig_procmask.c:
##########
@@ -187,8 +187,27 @@ int nxsig_procmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 
 int sigprocmask(int how, FAR const sigset_t *set, FAR sigset_t *oset)
 {
+  sigset_t nset;
   int ret;
 
+  /* SIGKILL and SIGSTOP should not be added to signal mask */
+
+  if (set != NULL)
+    {
+      nset = *set;
+      if (nxsig_ismember(&nset, SIGKILL))
+        {
+          nxsig_delset(&nset, SIGKILL);
+        }
+
+      if (nxsig_ismember(&nset, SIGSTOP))
+        {
+          nxsig_delset(&nset, SIGSTOP);
+        }
+
+      set = &nset;
+    }
+
   /* Let nxsig_procmask do all of the work */
 
   ret = nxsig_procmask(how, set, oset);

Review Comment:
   Note that there are two different instances of sigprocmask:  One common sigprocmask in the task group structure, and one in each TCB.  The latter are used by pthread_kill(), pthread_sigprocmask(), etc.
   
   Shouldn't the logic that removes SIGKILL and SIGSTOP also be in pthread_sigprocmask()?  The effect of sending SIGKILL (or SIGSTOP) via kill(SIGKILL, pid) and pthread_kill(thread, SIGKILL) must be identical.  SIGKILL and SIGSTOP cannot be masked in either and both effect the whole process/task group.
   



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@nuttx.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org