Operators want to be able to ensure enough tracepoints exist on the system for kernel components as well as for user components. Since there are only up to 64K events, by default allow up to half to be used by user events. Add a kernel sysctl parameter (kernel.user_events_max) to set a global limit that is honored among all groups on the system. This ensures hard limits can be setup to prevent user processes from consuming all event IDs on the system. Link: https://lkml.kernel.org/r/20230328235219.203-12-beaub@linux.microsoft.com
Unified · Split
| @@ -20,6 +20,7 @@ · 19 unmodified lines | ||
| 20 | 20 | #include <linux/types.h> |
| 21 | 21 | #include <linux/uaccess.h> |
| 22 | 22 | #include <linux/highmem.h> |
| 23 | +#include <linux/init.h> | |
| 23 | 24 | #include <linux/user_events.h> |
| 24 | 25 | #include "trace.h" |
| 25 | 26 | #include "trace_dynevent.h" |
| @@ -61,6 +62,12 @@ struct user_event_group { · 35 unmodified lines | ||
| 61 | 62 | /* Group for init_user_ns mapping, top-most group */ |
| 62 | 63 | static struct user_event_group *init_group; |
| 63 | 64 | |
| 65 | +/* Max allowed events for the whole system */ | |
| 66 | +static unsigned int max_user_events = 32768; | |
| 67 | + | |
| 68 | +/* Current number of events on the whole system */ | |
| 69 | +static unsigned int current_user_events; | |
| 70 | + | |
| 64 | 71 | /* |
| 65 | 72 | * Stores per-event properties, as users register events |
| 66 | 73 | * within a file a user_event might be created if it does not |
| @@ -1241,6 +1248,8 @@ static int destroy_user_event(struct user_event *user) · 1174 unmodified lines | ||
| 1241 | 1248 | { |
| 1242 | 1249 | int ret = 0; |
| 1243 | 1250 | |
| 1251 | + lockdep_assert_held(&event_mutex); | |
| 1252 | + | |
| 1244 | 1253 | /* Must destroy fields before call removal */ |
| 1245 | 1254 | user_event_destroy_fields(user); |
| 1246 | 1255 | |
| @@ -1257,6 +1266,11 @@ static int destroy_user_event(struct user_event *user) · 10 unmodified lines | ||
| 1257 | 1266 | kfree(EVENT_NAME(user)); |
| 1258 | 1267 | kfree(user); |
| 1259 | 1268 | |
| 1269 | + if (current_user_events > 0) | |
| 1270 | + current_user_events--; | |
| 1271 | + else | |
| 1272 | + pr_alert("BUG: Bad current_user_events\n"); | |
| 1273 | + | |
| 1260 | 1274 | return ret; |
| 1261 | 1275 | } |
| 1262 | 1276 | |
| @@ -1744,6 +1758,11 @@ static int user_event_parse(struct user_event_group *group, char *name, · 481 unmodified lines | ||
| 1744 | 1758 | |
| 1745 | 1759 | mutex_lock(&event_mutex); |
| 1746 | 1760 | |
| 1761 | + if (current_user_events >= max_user_events) { | |
| 1762 | + ret = -EMFILE; | |
| 1763 | + goto put_user_lock; | |
| 1764 | + } | |
| 1765 | + | |
| 1747 | 1766 | ret = user_event_trace_register(user); |
| 1748 | 1767 | |
| 1749 | 1768 | if (ret) |
| @@ -1755,6 +1774,7 @@ static int user_event_parse(struct user_event_group *group, char *name, · 5 unmodified lines | ||
| 1755 | 1774 | dyn_event_init(&user->devent, &user_event_dops); |
| 1756 | 1775 | dyn_event_add(&user->devent, &user->call); |
| 1757 | 1776 | hash_add(group->register_table, &user->node, key); |
| 1777 | + current_user_events++; | |
| 1758 | 1778 | |
| 1759 | 1779 | mutex_unlock(&event_mutex); |
| 1760 | 1780 | |
| @@ -2390,6 +2410,31 @@ err: · 629 unmodified lines | ||
| 2390 | 2410 | return -ENODEV; |
| 2391 | 2411 | } |
| 2392 | 2412 | |
| 2413 | +static int set_max_user_events_sysctl(struct ctl_table *table, int write, | |
| 2414 | + void *buffer, size_t *lenp, loff_t *ppos) | |
| 2415 | +{ | |
| 2416 | + int ret; | |
| 2417 | + | |
| 2418 | + mutex_lock(&event_mutex); | |
| 2419 | + | |
| 2420 | + ret = proc_douintvec(table, write, buffer, lenp, ppos); | |
| 2421 | + | |
| 2422 | + mutex_unlock(&event_mutex); | |
| 2423 | + | |
| 2424 | + return ret; | |
| 2425 | +} | |
| 2426 | + | |
| 2427 | +static struct ctl_table user_event_sysctls[] = { | |
| 2428 | + { | |
| 2429 | + .procname = "user_events_max", | |
| 2430 | + .data = &max_user_events, | |
| 2431 | + .maxlen = sizeof(unsigned int), | |
| 2432 | + .mode = 0644, | |
| 2433 | + .proc_handler = set_max_user_events_sysctl, | |
| 2434 | + }, | |
| 2435 | + {} | |
| 2436 | +}; | |
| 2437 | + | |
| 2393 | 2438 | static int __init trace_events_user_init(void) |
| 2394 | 2439 | { |
| 2395 | 2440 | int ret; |
| @@ -2419,6 +2464,8 @@ static int __init trace_events_user_init(void) · 23 unmodified lines | ||
| 2419 | 2464 | if (dyn_event_register(&user_event_dops)) |
| 2420 | 2465 | pr_warn("user_events could not register with dyn_events\n"); |
| 2421 | 2466 | |
| 2467 | + register_sysctl_init("kernel", user_event_sysctls); | |
| 2468 | + | |
| 2422 | 2469 | return 0; |
| 2423 | 2470 | } |
| 2424 | 2471 | |