seccomp: Add find_notification helper
[ Kees, why am I getting tons and tons of these warnings? Are we not
going to initialize things manually any more? ]
Hello Sargun Dhillon,
The patch 186f03857c48: "seccomp: Add find_notification helper" from Jun 1, 2020, leads to the following static checker warning:
kernel/seccomp.c:1124 seccomp_notify_recv() error: uninitialized symbol 'knotif'.
kernel/seccomp.c
1091 static long seccomp_notify_recv(struct seccomp_filter *filter,
1092 void __user *buf)
1093 {
1094 struct seccomp_knotif *knotif, *cur;
^^^^^^This used to be initialized to NULL here.
1095 struct seccomp_notif unotif;
1096 ssize_t ret;
1097
1098 /* Verify that we're not given garbage to keep struct extensible. */
1099 ret = check_zeroed_user(buf, sizeof(unotif));
1100 if (ret < 0)
1101 return ret;
1102 if (!ret)
1103 return -EINVAL;
1104
1105 memset(&unotif, 0, sizeof(unotif));
1106
1107 ret = down_interruptible(&filter->notif->request);
1108 if (ret < 0)
1109 return ret;
1110
1111 mutex_lock(&filter->notify_lock);
1112 list_for_each_entry(cur, &filter->notif->notifications, list) {
1113 if (cur->state == SECCOMP_NOTIFY_INIT) {
1114 knotif = cur;
^^^^^^^^^^^^ 1115 break;
1116 }
1117 }
1118
1119 /*
1120 * If we didn't find a notification, it could be that the task was
1121 * interrupted by a fatal signal between the time we were woken and
1122 * when we were able to acquire the rw lock.
1123 */
1124 if (!knotif) {
^^^^^^But now it's uninitialized.
1125 ret = -ENOENT; 1126 goto out; 1127 } 1128 1129 unotif.id = knotif->id;
regards,
dan carpenter
Quoted 3 lines · Dan Carpenter
[ Kees, why am I getting tons and tons of these warnings? Are we not
going to initialize things manually any more? ]
We are, yes. This is "just" a bug.
Quoted 8 lines
Hello Sargun Dhillon,
The patch
186f03857c48: "seccomp: Addfind_notificationhelper" from Jun 1, 2020, leads to the following static checker warning:
kernel/seccomp.c:1124seccomp_notify_recv()
error: uninitialized symbol 'knotif'.
Thanks for the heads-up! This was also reported by the ClangBuiltLinux project, and I've since fixed it. It should be visible in my for-next/seccomp tree now.
-Kees
Quoted 52 lines
kernel/seccomp.c
1091 static long seccomp_notify_recv(structseccomp_filter*filter,
1092 void__user*buf)
1093 {1094 struct seccomp_knotif *knotif, *cur; ^^^^^^This used to be initialized to NULL here.
1095 struct seccomp_notif unotif; 1096 ssize_t ret; 1097 1098 /* Verify that we're not given garbage to keep struct extensible. */ 1099 ret = check_zeroed_user(buf, sizeof(unotif)); 1100 if (ret < 0) 1101 return ret; 1102 if (!ret) 1103 return -EINVAL; 1104 1105 memset(&unotif, 0, sizeof(unotif)); 1106 1107 ret = down_interruptible(&filter->notif->request); 1108 if (ret < 0) 1109 return ret; 1110 1111 mutex_lock(&filter->notify_lock); 1112 list_for_each_entry(cur, &filter->notif->notifications, list) { 1113 if (cur->state == SECCOMP_NOTIFY_INIT) { 1114 knotif = cur; ^^^^^^^^^^^^1115 break;
1116 }
1117 }
1118
1119 /*1120 * If we didn't find a notification, it could be that the task was 1121 * interrupted by a fatal signal between the time we were woken and 1122 * when we were able to acquire the rw lock. 1123 */ 1124 if (!knotif) { ^^^^^^But now it's uninitialized.
1125 ret = -ENOENT; 1126 goto out; 1127 } 1128 1129 unotif.id = knotif->id;regards,
dan carpenter
Signature · 4 lines
Kees Cook -- Kees Cook
Quoted 6 lines · Kees Cook
Quoted 3 lines · Dan Carpenter
[ Kees, why am I getting tons and tons of these warnings? Are we not
going to initialize things manually any more? ]We are, yes. This is "just" a bug.
GCC has been spotty for some years because of the optimization bug, but it really feels like something changed recent and I'm getting lots more warnings now.
regards,
dan carpenter