| Message ID | 51aabb2e4680d830b5ba7b7623e0c7697bcb6338.1573675056.git.jstancek@redhat.com |
|---|---|
| State | Accepted, archived |
| Headers | show |
| Series | mbind04: terminate child processes | expand |
Hi! > verify_policy() forks a child, which continues to run > pass end of function and into same loop in verify_mbind() > as parent, effectively creating fork bomb. With 2 nodes, > there's 45 forked children. With 4 nodes, 765 processes. > > Terminate child process at the end of verify_policy(); Good catch, thanks for fixing this. > Signed-off-by: Jan Stancek <jstancek@redhat.com> > --- > testcases/kernel/syscalls/mbind/mbind04.c | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/testcases/kernel/syscalls/mbind/mbind04.c b/testcases/kernel/syscalls/mbind/mbind04.c > index 6a7a2a2321a8..aaad8a988bad 100644 > --- a/testcases/kernel/syscalls/mbind/mbind04.c > +++ b/testcases/kernel/syscalls/mbind/mbind04.c > @@ -47,6 +47,7 @@ static void verify_policy(unsigned int node, int mode, unsigned flag) > struct bitmask *bm = numa_allocate_nodemask(); > unsigned int i; > void *ptr; > + pid_t child; > unsigned long size = PAGES_ALLOCATED * page_size; > > numa_bitmask_setbit(bm, node); > @@ -69,7 +70,8 @@ static void verify_policy(unsigned int node, int mode, unsigned flag) > > const char *prefix = "child: "; > > - if (SAFE_FORK()) { > + child = SAFE_FORK(); > + if (child) { > prefix = "parent: "; > tst_reap_children(); > } > @@ -104,6 +106,9 @@ static void verify_policy(unsigned int node, int mode, unsigned flag) > > if (fail) > tst_nodemap_print_counters(nodes); > + > + if (!child) > + exit(0); > } I find the name for the variable quite confusing here, because it suggests that the parent exits here, while the opposite is true. Can we please just use pid or something less confusing for the variable name? Other than that it's fine. Reviewed-by: Cyril Hrubis <chrubis@suse.cz>
> I find the name for the variable quite confusing here, because it > suggests that the parent exits here, while the opposite is true. Can we > please just use pid or something less confusing for the variable name? Pushed with renamed variable.
diff --git a/testcases/kernel/syscalls/mbind/mbind04.c b/testcases/kernel/syscalls/mbind/mbind04.c index 6a7a2a2321a8..aaad8a988bad 100644 --- a/testcases/kernel/syscalls/mbind/mbind04.c +++ b/testcases/kernel/syscalls/mbind/mbind04.c @@ -47,6 +47,7 @@ static void verify_policy(unsigned int node, int mode, unsigned flag) struct bitmask *bm = numa_allocate_nodemask(); unsigned int i; void *ptr; + pid_t child; unsigned long size = PAGES_ALLOCATED * page_size; numa_bitmask_setbit(bm, node); @@ -69,7 +70,8 @@ static void verify_policy(unsigned int node, int mode, unsigned flag) const char *prefix = "child: "; - if (SAFE_FORK()) { + child = SAFE_FORK(); + if (child) { prefix = "parent: "; tst_reap_children(); } @@ -104,6 +106,9 @@ static void verify_policy(unsigned int node, int mode, unsigned flag) if (fail) tst_nodemap_print_counters(nodes); + + if (!child) + exit(0); } static const int modes[] = {
verify_policy() forks a child, which continues to run pass end of function and into same loop in verify_mbind() as parent, effectively creating fork bomb. With 2 nodes, there's 45 forked children. With 4 nodes, 765 processes. Terminate child process at the end of verify_policy(); Signed-off-by: Jan Stancek <jstancek@redhat.com> --- testcases/kernel/syscalls/mbind/mbind04.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)