diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2023-02-06 16:11:30 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2023-02-06 16:11:30 +0000 |
commit | aa2fe8ccbfcb117efa207d10229eeeac5d0f97c7 (patch) | |
tree | 941cbdd387b41c1a81587c20a6df9f0e5e0ff7ab /collectors/ebpf.plugin | |
parent | Adding upstream version 1.37.1. (diff) | |
download | netdata-aa2fe8ccbfcb117efa207d10229eeeac5d0f97c7.tar.xz netdata-aa2fe8ccbfcb117efa207d10229eeeac5d0f97c7.zip |
Adding upstream version 1.38.0.upstream/1.38.0
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
34 files changed, 474 insertions, 1209 deletions
diff --git a/collectors/ebpf.plugin/README.md b/collectors/ebpf.plugin/README.md index 7762ed34f..deedf4d79 100644 --- a/collectors/ebpf.plugin/README.md +++ b/collectors/ebpf.plugin/README.md @@ -1,9 +1,11 @@ <!-- -title: "eBPF monitoring with Netdata" -description: "Use Netdata's extended Berkeley Packet Filter (eBPF) collector to monitor kernel-level metrics about your -complex applications with per-second granularity." -custom_edit_url: https://github.com/netdata/netdata/edit/master/collectors/ebpf.plugin/README.md -sidebar_label: "eBPF" +title: "Kernel traces/metrics (eBPF) monitoring with Netdata" +description: "Use Netdata's extended Berkeley Packet Filter (eBPF) collector to monitor kernel-level metrics about yourcomplex applications with per-second granularity." +custom_edit_url: "https://github.com/netdata/netdata/edit/master/collectors/ebpf.plugin/README.md" +sidebar_label: "Kernel traces/metrics (eBPF)" +learn_status: "Published" +learn_topic_type: "References" +learn_rel_path: "References/Collectors references/System metrics" --> # eBPF monitoring with Netdata @@ -13,7 +15,7 @@ The Netdata Agent provides many [eBPF](https://ebpf.io/what-is-ebpf/) programs t > ❗ eBPF monitoring only works on Linux systems and with specific Linux kernels, including all kernels newer than `4.11.0`, and all kernels on CentOS 7.6 or later. For kernels older than `4.11.0`, improved support is in active development. This document provides comprehensive details about the `ebpf.plugin`. -For hands-on configuration and troubleshooting tips see our [tutorial on troubleshooting apps with eBPF metrics](/docs/guides/troubleshoot/monitor-debug-applications-ebpf.md). +For hands-on configuration and troubleshooting tips see our [tutorial on troubleshooting apps with eBPF metrics](https://github.com/netdata/netdata/blob/master/docs/guides/troubleshoot/monitor-debug-applications-ebpf.md). <figure> <img src="https://user-images.githubusercontent.com/1153921/74746434-ad6a1e00-5222-11ea-858a-a7882617ae02.png" alt="An example of VFS charts, made possible by the eBPF collector plugin" /> @@ -42,12 +44,12 @@ If your Agent is v1.22 or older, you may to enable the collector yourself. To enable or disable the entire eBPF collector: -1. Navigate to the [Netdata config directory](/docs/configure/nodes.md#the-netdata-config-directory). +1. Navigate to the [Netdata config directory](https://github.com/netdata/netdata/blob/master/docs/configure/nodes.md#the-netdata-config-directory). ```bash cd /etc/netdata ``` -2. Use the [`edit-config`](/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit `netdata.conf`. +2. Use the [`edit-config`](https://github.com/netdata/netdata/blob/master/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit `netdata.conf`. ```bash ./edit-config netdata.conf @@ -67,11 +69,11 @@ You can configure the eBPF collector's behavior to fine-tune which metrics you r To edit the `ebpf.d.conf`: -1. Navigate to the [Netdata config directory](/docs/configure/nodes.md#the-netdata-config-directory). +1. Navigate to the [Netdata config directory](https://github.com/netdata/netdata/blob/master/docs/configure/nodes.md#the-netdata-config-directory). ```bash cd /etc/netdata ``` -2. Use the [`edit-config`](/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit [`ebpf.d.conf`](https://github.com/netdata/netdata/blob/master/collectors/ebpf.plugin/ebpf.d.conf). +2. Use the [`edit-config`](https://github.com/netdata/netdata/blob/master/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit [`ebpf.d.conf`](https://github.com/netdata/netdata/blob/master/collectors/ebpf.plugin/ebpf.d.conf). ```bash ./edit-config ebpf.d.conf @@ -103,11 +105,10 @@ accepts the following values: #### Integration with `apps.plugin` The eBPF collector also creates charts for each running application through an integration with the -[`apps.plugin`](/collectors/apps.plugin/README.md). This integration helps you understand how specific applications +[`apps.plugin`](https://github.com/netdata/netdata/blob/master/collectors/apps.plugin/README.md). This integration helps you understand how specific applications interact with the Linux kernel. -If you want to _disable_ the integration with `apps.plugin` along with the above charts, change the setting `apps` to -`no`. +If you want to enable `apps.plugin` integration, change the "apps" setting to "yes". ```conf [global] @@ -122,7 +123,7 @@ it runs. #### Integration with `cgroups.plugin` The eBPF collector also creates charts for each cgroup through an integration with the -[`cgroups.plugin`](/collectors/cgroups.plugin/README.md). This integration helps you understand how a specific cgroup +[`cgroups.plugin`](https://github.com/netdata/netdata/blob/master/collectors/cgroups.plugin/README.md). This integration helps you understand how a specific cgroup interacts with the Linux kernel. The integration with `cgroups.plugin` is disabled by default to avoid creating overhead on your system. If you want to @@ -244,7 +245,7 @@ The eBPF collector enables and runs the following eBPF programs by default: You can also enable the following eBPF programs: - `cachestat`: Netdata's eBPF data collector creates charts about the memory page cache. When the integration with - [`apps.plugin`](/collectors/apps.plugin/README.md) is enabled, this collector creates charts for the whole host _and_ + [`apps.plugin`](https://github.com/netdata/netdata/blob/master/collectors/apps.plugin/README.md) is enabled, this collector creates charts for the whole host _and_ for each application. - `dcstat` : This eBPF program creates charts that show information about file access using directory cache. It appends `kprobes` for `lookup_fast()` and `d_lookup()` to identify if files are inside directory cache, outside and files are @@ -261,11 +262,11 @@ You can configure each thread of the eBPF data collector. This allows you to ove To configure an eBPF thread: -1. Navigate to the [Netdata config directory](/docs/configure/nodes.md#the-netdata-config-directory). +1. Navigate to the [Netdata config directory](https://github.com/netdata/netdata/blob/master/docs/configure/nodes.md#the-netdata-config-directory). ```bash cd /etc/netdata ``` -2. Use the [`edit-config`](/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit a thread configuration file. The following configuration files are available: +2. Use the [`edit-config`](https://github.com/netdata/netdata/blob/master/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit a thread configuration file. The following configuration files are available: - `network.conf`: Configuration for the [`network` thread](#network-configuration). This config file overwrites the global options and also lets you specify which network the eBPF collector monitors. @@ -304,7 +305,7 @@ You can configure the information shown on `outbound` and `inbound` charts with When you define a `ports` setting, Netdata will collect network metrics for that specific port. For example, if you write `ports = 19999`, Netdata will collect only connections for itself. The `hostnames` setting accepts -[simple patterns](/libnetdata/simple_pattern/README.md). The `ports`, and `ips` settings accept negation (`!`) to deny +[simple patterns](https://github.com/netdata/netdata/blob/master/libnetdata/simple_pattern/README.md). The `ports`, and `ips` settings accept negation (`!`) to deny specific values or asterisk alone to define all values. In the above example, Netdata will collect metrics for all ports between 1 and 443, with the exception of 53 (domain) @@ -881,7 +882,7 @@ significantly increases kernel memory usage by several hundred MB. If your node is experiencing high memory usage and there is no obvious culprit to be found in the `apps.mem` chart, consider testing for high kernel memory usage by [disabling eBPF monitoring](#configuring-ebpfplugin). Next, -[restart Netdata](/docs/configure/start-stop-restart.md) with `sudo systemctl restart netdata` to see if system memory +[restart Netdata](https://github.com/netdata/netdata/blob/master/docs/configure/start-stop-restart.md) with `sudo systemctl restart netdata` to see if system memory usage (see the `system.ram` chart) has dropped significantly. Beginning with `v1.31`, kernel memory usage is configurable via the [`pid table size` setting](#ebpf-load-mode) diff --git a/collectors/ebpf.plugin/ebpf.c b/collectors/ebpf.plugin/ebpf.c index 00b53a57d..67fe477c2 100644 --- a/collectors/ebpf.plugin/ebpf.c +++ b/collectors/ebpf.plugin/ebpf.c @@ -483,6 +483,16 @@ static void ebpf_exit() if (unlink(filename)) error("Cannot remove PID file %s", filename); +#ifdef NETDATA_INTERNAL_CHECKS + error("Good bye world! I was PID %d", main_thread_id); +#endif + printf("DISABLE\n"); + + if (shm_ebpf_cgroup.header) { + munmap(shm_ebpf_cgroup.header, shm_ebpf_cgroup.header->body_length); + shm_unlink(NETDATA_SHARED_MEMORY_EBPF_CGROUP_NAME); + } + exit(0); } @@ -534,7 +544,7 @@ static void ebpf_stop_threads(int sig) pthread_mutex_unlock(&ebpf_exit_cleanup); ebpf_exit_plugin = 1; - usec_t max = 3 * USEC_PER_SEC, step = 100000; + usec_t max = USEC_PER_SEC, step = 100000; while (i && max) { max -= step; sleep_usec(step); @@ -548,32 +558,35 @@ static void ebpf_stop_threads(int sig) pthread_mutex_unlock(&ebpf_exit_cleanup); } - //Unload threads(except sync and filesystem) - pthread_mutex_lock(&ebpf_exit_cleanup); - for (i = 0; ebpf_threads[i].name != NULL; i++) { - if (ebpf_threads[i].enabled == NETDATA_THREAD_EBPF_STOPPED && i != EBPF_MODULE_FILESYSTEM_IDX && - i != EBPF_MODULE_SYNC_IDX) - ebpf_unload_legacy_code(ebpf_modules[i].objects, ebpf_modules[i].probe_links); - } - pthread_mutex_unlock(&ebpf_exit_cleanup); + if (!i) { + //Unload threads(except sync and filesystem) + pthread_mutex_lock(&ebpf_exit_cleanup); + for (i = 0; ebpf_threads[i].name != NULL; i++) { + if (ebpf_threads[i].enabled == NETDATA_THREAD_EBPF_STOPPED && i != EBPF_MODULE_FILESYSTEM_IDX && + i != EBPF_MODULE_SYNC_IDX) + ebpf_unload_legacy_code(ebpf_modules[i].objects, ebpf_modules[i].probe_links); + } + pthread_mutex_unlock(&ebpf_exit_cleanup); - //Unload filesystem - pthread_mutex_lock(&ebpf_exit_cleanup); - if (ebpf_threads[EBPF_MODULE_FILESYSTEM_IDX].enabled == NETDATA_THREAD_EBPF_STOPPED) { - for (i = 0; localfs[i].filesystem != NULL; i++) { - ebpf_unload_legacy_code(localfs[i].objects, localfs[i].probe_links); + //Unload filesystem + pthread_mutex_lock(&ebpf_exit_cleanup); + if (ebpf_threads[EBPF_MODULE_FILESYSTEM_IDX].enabled == NETDATA_THREAD_EBPF_STOPPED) { + for (i = 0; localfs[i].filesystem != NULL; i++) { + ebpf_unload_legacy_code(localfs[i].objects, localfs[i].probe_links); + } } - } - pthread_mutex_unlock(&ebpf_exit_cleanup); + pthread_mutex_unlock(&ebpf_exit_cleanup); - //Unload Sync - pthread_mutex_lock(&ebpf_exit_cleanup); - if (ebpf_threads[EBPF_MODULE_SYNC_IDX].enabled == NETDATA_THREAD_EBPF_STOPPED) { - for (i = 0; local_syscalls[i].syscall != NULL; i++) { - ebpf_unload_legacy_code(local_syscalls[i].objects, local_syscalls[i].probe_links); + //Unload Sync + pthread_mutex_lock(&ebpf_exit_cleanup); + if (ebpf_threads[EBPF_MODULE_SYNC_IDX].enabled == NETDATA_THREAD_EBPF_STOPPED) { + for (i = 0; local_syscalls[i].syscall != NULL; i++) { + ebpf_unload_legacy_code(local_syscalls[i].objects, local_syscalls[i].probe_links); + } } + pthread_mutex_unlock(&ebpf_exit_cleanup); + } - pthread_mutex_unlock(&ebpf_exit_cleanup); ebpf_exit(); } @@ -1317,7 +1330,7 @@ static void read_local_addresses() } } - fill_ip_list((family == AF_INET)?&network_viewer_opt.ipv4_local_ip:&network_viewer_opt.ipv6_local_ip, + ebpf_fill_ip_list((family == AF_INET)?&network_viewer_opt.ipv4_local_ip:&network_viewer_opt.ipv6_local_ip, w, "selector"); } @@ -1520,13 +1533,8 @@ static void read_collector_values(int *disable_apps, int *disable_cgroups, enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, ebpf_modules[EBPF_MODULE_SOCKET_IDX].config_name, CONFIG_BOOLEAN_NO); - if (enabled) { ebpf_enable_chart(EBPF_MODULE_SOCKET_IDX, *disable_apps, *disable_cgroups); - // Read network viewer section if network viewer is enabled - // This is kept here to keep backward compatibility - parse_network_viewer_section(&collector_config); - parse_service_name_section(&collector_config); started++; } @@ -1536,7 +1544,17 @@ static void read_collector_values(int *disable_apps, int *disable_cgroups, if (!enabled) enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, "network connections", CONFIG_BOOLEAN_NO); - ebpf_modules[EBPF_MODULE_SOCKET_IDX].optional = (int)enabled; + network_viewer_opt.enabled = enabled; + if (enabled) { + if (!ebpf_modules[EBPF_MODULE_SOCKET_IDX].enabled) + ebpf_enable_chart(EBPF_MODULE_SOCKET_IDX, *disable_apps, *disable_cgroups); + + // Read network viewer section if network viewer is enabled + // This is kept here to keep backward compatibility + parse_network_viewer_section(&collector_config); + parse_service_name_section(&collector_config); + started++; + } enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, "cachestat", CONFIG_BOOLEAN_NO); @@ -1642,8 +1660,10 @@ static void read_collector_values(int *disable_apps, int *disable_cgroups, ebpf_enable_all_charts(*disable_apps, *disable_cgroups); // Read network viewer section // This is kept here to keep backward compatibility - parse_network_viewer_section(&collector_config); - parse_service_name_section(&collector_config); + if (network_viewer_opt.enabled) { + parse_network_viewer_section(&collector_config); + parse_service_name_section(&collector_config); + } } } @@ -2158,6 +2178,7 @@ static void ebpf_manage_pid(pid_t pid) */ int main(int argc, char **argv) { + stderror = stderr; clocks_init(); main_thread_id = gettid(); @@ -2237,13 +2258,26 @@ int main(int argc, char **argv) } } - usec_t step = EBPF_DEFAULT_UPDATE_EVERY * USEC_PER_SEC; + usec_t step = USEC_PER_SEC; + int counter = NETDATA_EBPF_CGROUP_UPDATE - 1; heartbeat_t hb; heartbeat_init(&hb); //Plugin will be killed when it receives a signal while (!ebpf_exit_plugin) { (void)heartbeat_next(&hb, step); + + // We are using a small heartbeat time to wake up thread, + // but we should not update so frequently the shared memory data + if (++counter >= NETDATA_EBPF_CGROUP_UPDATE) { + counter = 0; + if (!shm_ebpf_cgroup.header) + ebpf_map_cgroup_shared_memory(); + + ebpf_parse_cgroup_shm_data(); + } } + ebpf_stop_threads(0); + return 0; } diff --git a/collectors/ebpf.plugin/ebpf.d.conf b/collectors/ebpf.plugin/ebpf.d.conf index cf5c740fc..112df275d 100644 --- a/collectors/ebpf.plugin/ebpf.d.conf +++ b/collectors/ebpf.plugin/ebpf.d.conf @@ -17,7 +17,7 @@ # [global] ebpf load mode = entry - apps = yes + apps = no cgroups = no update every = 5 pid table size = 32768 @@ -50,7 +50,7 @@ # When plugin detects that system has support to BTF, it enables integration with apps.plugin. # [ebpf programs] - cachestat = no + cachestat = yes dcstat = no disk = no fd = yes @@ -60,10 +60,10 @@ mount = yes oomkill = yes process = yes - shm = no - socket = yes + shm = yes + socket = no softirq = yes sync = yes - swap = no - vfs = yes + swap = yes + vfs = no network connections = no diff --git a/collectors/ebpf.plugin/ebpf.h b/collectors/ebpf.plugin/ebpf.h index 28b04ce48..16e62498c 100644 --- a/collectors/ebpf.plugin/ebpf.h +++ b/collectors/ebpf.plugin/ebpf.h @@ -123,6 +123,9 @@ enum ebpf_threads_status { #endif #endif +// Messages +#define NETDATA_EBPF_DEFAULT_FNT_NOT_FOUND "Cannot find the necessary functions to monitor" + // Chart definitions #define NETDATA_EBPF_FAMILY "ebpf" #define NETDATA_EBPF_IP_FAMILY "ip" diff --git a/collectors/ebpf.plugin/ebpf_cachestat.c b/collectors/ebpf.plugin/ebpf_cachestat.c index 4c410647d..b21cc6103 100644 --- a/collectors/ebpf.plugin/ebpf_cachestat.c +++ b/collectors/ebpf.plugin/ebpf_cachestat.c @@ -15,15 +15,6 @@ netdata_cachestat_pid_t *cachestat_vector = NULL; static netdata_idx_t cachestat_hash_values[NETDATA_CACHESTAT_END]; static netdata_idx_t *cachestat_values = NULL; -struct netdata_static_thread cachestat_threads = {.name = "CACHESTAT KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL}; - ebpf_local_maps_t cachestat_maps[] = {{.name = "cstat_global", .internal_input = NETDATA_CACHESTAT_END, .user_input = 0, .type = NETDATA_EBPF_MAP_STATIC, .map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED}, @@ -51,6 +42,9 @@ netdata_ebpf_targets_t cachestat_targets[] = { {.name = "add_to_page_cache_lru", {.name = "mark_buffer_dirty", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = NULL, .mode = EBPF_LOAD_TRAMPOLINE}}; +static char *account_page[NETDATA_CACHESTAT_ACCOUNT_DIRTY_END] ={ "account_page_dirtied", + "__set_page_dirty", "__folio_mark_dirty" }; + #ifdef LIBBPF_MAJOR_VERSION #include "includes/cachestat.skel.h" // BTF code @@ -83,10 +77,12 @@ static void ebpf_cachestat_disable_probe(struct cachestat_bpf *obj) */ static void ebpf_cachestat_disable_specific_probe(struct cachestat_bpf *obj) { - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_16) { + if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_FOLIO_DIRTY])) { bpf_program__set_autoload(obj->progs.netdata_account_page_dirtied_kprobe, false); bpf_program__set_autoload(obj->progs.netdata_set_page_dirty_kprobe, false); - } else if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_15) { + } else if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_SET_PAGE_DIRTY])) { bpf_program__set_autoload(obj->progs.netdata_folio_mark_dirty_kprobe, false); bpf_program__set_autoload(obj->progs.netdata_account_page_dirtied_kprobe, false); } else { @@ -122,10 +118,12 @@ static void ebpf_cachestat_disable_trampoline(struct cachestat_bpf *obj) */ static void ebpf_cachestat_disable_specific_trampoline(struct cachestat_bpf *obj) { - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_16) { + if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_FOLIO_DIRTY])) { bpf_program__set_autoload(obj->progs.netdata_account_page_dirtied_fentry, false); bpf_program__set_autoload(obj->progs.netdata_set_page_dirty_fentry, false); - } else if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_15) { + } else if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_SET_PAGE_DIRTY])) { bpf_program__set_autoload(obj->progs.netdata_folio_mark_dirty_fentry, false); bpf_program__set_autoload(obj->progs.netdata_account_page_dirtied_fentry, false); } else { @@ -149,10 +147,12 @@ static inline void netdata_set_trampoline_target(struct cachestat_bpf *obj) bpf_program__set_attach_target(obj->progs.netdata_mark_page_accessed_fentry, 0, cachestat_targets[NETDATA_KEY_CALLS_MARK_PAGE_ACCESSED].name); - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_16) { + if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_FOLIO_DIRTY])) { bpf_program__set_attach_target(obj->progs.netdata_folio_mark_dirty_fentry, 0, cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name); - } else if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_15) { + } else if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_SET_PAGE_DIRTY])) { bpf_program__set_attach_target(obj->progs.netdata_set_page_dirty_fentry, 0, cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name); } else { @@ -192,12 +192,14 @@ static int ebpf_cachestat_attach_probe(struct cachestat_bpf *obj) if (ret) return -1; - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_16) { + if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_FOLIO_DIRTY])) { obj->links.netdata_folio_mark_dirty_kprobe = bpf_program__attach_kprobe(obj->progs.netdata_folio_mark_dirty_kprobe, false, cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name); ret = libbpf_get_error(obj->links.netdata_folio_mark_dirty_kprobe); - } else if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_15) { + } else if (!strcmp(cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name, + account_page[NETDATA_CACHESTAT_SET_PAGE_DIRTY])) { obj->links.netdata_set_page_dirty_kprobe = bpf_program__attach_kprobe(obj->progs.netdata_set_page_dirty_kprobe, false, cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name); @@ -278,7 +280,7 @@ static void ebpf_cachestat_disable_release_task(struct cachestat_bpf *obj) * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_cachestat_load_and_attach(struct cachestat_bpf *obj, ebpf_module_t *em) { @@ -331,18 +333,13 @@ static inline int ebpf_cachestat_load_and_attach(struct cachestat_bpf *obj, ebpf static void ebpf_cachestat_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); ebpf_cleanup_publish_syscall(cachestat_counter_publish_aggregated); freez(cachestat_vector); freez(cachestat_values); - freez(cachestat_threads.thread); #ifdef LIBBPF_MAJOR_VERSION if (bpf_obj) @@ -363,20 +360,7 @@ static void ebpf_cachestat_free(ebpf_module_t *em) static void ebpf_cachestat_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*cachestat_threads.thread); - ebpf_cachestat_free(em); -} -/** - * Cachestat cleanup - * - * Clean up allocated addresses. - * - * @param ptr thread data. - */ -static void ebpf_cachestat_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_cachestat_free(em); } @@ -656,7 +640,7 @@ void ebpf_cachestat_create_apps_charts(struct ebpf_module *em, void *ptr) * * Read the table with number of calls for all functions */ -static void read_global_table() +static void ebpf_cachestat_read_global_table() { uint32_t idx; netdata_idx_t *val = cachestat_hash_values; @@ -677,35 +661,6 @@ static void read_global_table() } /** - * Socket read hash - * - * This is the thread callback. - * This thread is necessary, because we cannot freeze the whole plugin to read the data on very busy socket. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_cachestat_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_cachestat_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_LATENCY_CACHESTAT_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Send global * * Send global charts to Netdata @@ -1106,26 +1061,23 @@ void ebpf_cachestat_send_cgroup_data(int update_every) */ static void cachestat_collector(ebpf_module_t *em) { - cachestat_threads.thread = callocz(1, sizeof(netdata_thread_t)); - cachestat_threads.start_routine = ebpf_cachestat_read_hash; - - netdata_thread_create(cachestat_threads.thread, cachestat_threads.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_cachestat_read_hash, em); - netdata_publish_cachestat_t publish; memset(&publish, 0, sizeof(publish)); int cgroups = em->cgroup_charts; int update_every = em->update_every; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = update_every * USEC_PER_SEC; + int counter = update_every - 1; //This will be cancelled by its parent while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; netdata_apps_integration_flags_t apps = em->apps_charts; + ebpf_cachestat_read_global_table(); pthread_mutex_lock(&collect_data_mutex); if (apps) read_apps_table(); @@ -1237,16 +1189,28 @@ static void ebpf_cachestat_allocate_global_vectors(int apps) * Update Internal value * * Update values used during runtime. + * + * @return It returns 0 when one of the functions is present and -1 otherwise. */ -static void ebpf_cachestat_set_internal_value() +static int ebpf_cachestat_set_internal_value() { - static char *account_page[] = { "account_page_dirtied", "__set_page_dirty", "__folio_mark_dirty" }; - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_16) - cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name = account_page[NETDATA_CACHESTAT_FOLIO_DIRTY]; - else if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_15) - cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name = account_page[NETDATA_CACHESTAT_SET_PAGE_DIRTY]; - else - cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name = account_page[NETDATA_CACHESTAT_ACCOUNT_PAGE_DIRTY]; + ebpf_addresses_t address = {.function = NULL, .hash = 0, .addr = 0}; + int i; + for (i = 0; i < NETDATA_CACHESTAT_ACCOUNT_DIRTY_END ; i++) { + address.function = account_page[i]; + ebpf_load_addresses(&address, -1); + if (address.addr) + break; + } + + if (!address.addr) { + error("%s cachestat.", NETDATA_EBPF_DEFAULT_FNT_NOT_FOUND); + return -1; + } + + cachestat_targets[NETDATA_KEY_CALLS_ACCOUNT_PAGE_DIRTIED].name = address.function; + + return 0; } /* @@ -1300,7 +1264,10 @@ void *ebpf_cachestat_thread(void *ptr) ebpf_update_pid_table(&cachestat_maps[NETDATA_CACHESTAT_PID_STATS], em); - ebpf_cachestat_set_internal_value(); + if (ebpf_cachestat_set_internal_value()) { + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; + goto endcachestat; + } #ifdef LIBBPF_MAJOR_VERSION ebpf_adjust_thread_load(em, default_btf); diff --git a/collectors/ebpf.plugin/ebpf_cachestat.h b/collectors/ebpf.plugin/ebpf_cachestat.h index 07f0745d4..15b06511e 100644 --- a/collectors/ebpf.plugin/ebpf_cachestat.h +++ b/collectors/ebpf.plugin/ebpf_cachestat.h @@ -19,8 +19,6 @@ #define EBPF_CACHESTAT_DIMENSION_HITS "hits/s" #define EBPF_CACHESTAT_DIMENSION_MISSES "misses/s" -#define NETDATA_LATENCY_CACHESTAT_SLEEP_MS 600000ULL - // configuration file #define NETDATA_CACHESTAT_CONFIG_FILE "cachestat.conf" @@ -48,7 +46,9 @@ enum cachestat_counters { enum cachestat_account_dirty_pages { NETDATA_CACHESTAT_ACCOUNT_PAGE_DIRTY, NETDATA_CACHESTAT_SET_PAGE_DIRTY, - NETDATA_CACHESTAT_FOLIO_DIRTY + NETDATA_CACHESTAT_FOLIO_DIRTY, + + NETDATA_CACHESTAT_ACCOUNT_DIRTY_END }; enum cachestat_indexes { diff --git a/collectors/ebpf.plugin/ebpf_dcstat.c b/collectors/ebpf.plugin/ebpf_dcstat.c index 71169e153..75e83214a 100644 --- a/collectors/ebpf.plugin/ebpf_dcstat.c +++ b/collectors/ebpf.plugin/ebpf_dcstat.c @@ -19,15 +19,6 @@ struct config dcstat_config = { .first_section = NULL, .index = { .avl_tree = { .root = NULL, .compar = appconfig_section_compare }, .rwlock = AVL_LOCK_INITIALIZER } }; -struct netdata_static_thread dcstat_threads = {"DCSTAT KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL}; - ebpf_local_maps_t dcstat_maps[] = {{.name = "dcstat_global", .internal_input = NETDATA_DIRECTORY_CACHE_END, .user_input = 0, .type = NETDATA_EBPF_MAP_STATIC, .map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED}, @@ -216,7 +207,7 @@ static void ebpf_dc_disable_release_task(struct dc_bpf *obj) * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_dc_load_and_attach(struct dc_bpf *obj, ebpf_module_t *em) { @@ -303,16 +294,11 @@ void ebpf_dcstat_clean_names() static void ebpf_dcstat_free(ebpf_module_t *em ) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); freez(dcstat_vector); freez(dcstat_values); - freez(dcstat_threads.thread); ebpf_cleanup_publish_syscall(dcstat_counter_publish_aggregated); @@ -338,18 +324,6 @@ static void ebpf_dcstat_free(ebpf_module_t *em ) static void ebpf_dcstat_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*dcstat_threads.thread); - ebpf_dcstat_free(em); -} - -/** - * Clean up the main thread. - * - * @param ptr thread data. - */ -static void ebpf_dcstat_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_dcstat_free(em); } @@ -538,7 +512,7 @@ static void ebpf_update_dc_cgroup() * * Read the table with number of calls for all functions */ -static void read_global_table() +static void ebpf_dc_read_global_table() { uint32_t idx; netdata_idx_t *val = dcstat_hash_values; @@ -559,35 +533,6 @@ static void read_global_table() } /** - * DCstat read hash - * - * This is the thread callback. - * This thread is necessary, because we cannot freeze the whole plugin to read the data. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_dcstat_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_dcstat_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_LATENCY_DCSTAT_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Cachestat sum PIDs * * Sum values for all PIDs associated to a group @@ -1034,25 +979,22 @@ void ebpf_dc_send_cgroup_data(int update_every) */ static void dcstat_collector(ebpf_module_t *em) { - dcstat_threads.thread = mallocz(sizeof(netdata_thread_t)); - dcstat_threads.start_routine = ebpf_dcstat_read_hash; - - netdata_thread_create(dcstat_threads.thread, dcstat_threads.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_dcstat_read_hash, em); - netdata_publish_dcstat_t publish; memset(&publish, 0, sizeof(publish)); int cgroups = em->cgroup_charts; int update_every = em->update_every; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = update_every * USEC_PER_SEC; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; netdata_apps_integration_flags_t apps = em->apps_charts; + ebpf_dc_read_global_table(); pthread_mutex_lock(&collect_data_mutex); if (apps) read_apps_table(); diff --git a/collectors/ebpf.plugin/ebpf_dcstat.h b/collectors/ebpf.plugin/ebpf_dcstat.h index d8687f968..201fc8a02 100644 --- a/collectors/ebpf.plugin/ebpf_dcstat.h +++ b/collectors/ebpf.plugin/ebpf_dcstat.h @@ -28,8 +28,6 @@ #define NETDATA_SYSTEMD_DC_NOT_CACHE_CONTEXT "services.dc_not_cache" #define NETDATA_SYSTEMD_DC_NOT_FOUND_CONTEXT "services.dc_not_found" -#define NETDATA_LATENCY_DCSTAT_SLEEP_MS 700000ULL - enum directory_cache_indexes { NETDATA_DCSTAT_IDX_RATIO, NETDATA_DCSTAT_IDX_REFERENCE, diff --git a/collectors/ebpf.plugin/ebpf_disk.c b/collectors/ebpf.plugin/ebpf_disk.c index a27bd81e3..5e7e2599d 100644 --- a/collectors/ebpf.plugin/ebpf_disk.c +++ b/collectors/ebpf.plugin/ebpf_disk.c @@ -33,16 +33,6 @@ static netdata_syscall_stat_t disk_aggregated_data[NETDATA_EBPF_HIST_MAX_BINS]; static netdata_publish_syscall_t disk_publish_aggregated[NETDATA_EBPF_HIST_MAX_BINS]; static netdata_idx_t *disk_hash_values = NULL; -static struct netdata_static_thread disk_threads = { - .name = "DISK KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; ebpf_publish_disk_t *plot_disks = NULL; pthread_mutex_t plot_mutex; @@ -439,11 +429,7 @@ static void ebpf_cleanup_disk_list() static void ebpf_disk_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); ebpf_disk_disable_tracepoints(); @@ -452,7 +438,6 @@ static void ebpf_disk_free(ebpf_module_t *em) ebpf_histogram_dimension_cleanup(dimensions, NETDATA_EBPF_HIST_MAX_BINS); freez(disk_hash_values); - freez(disk_threads.thread); pthread_mutex_destroy(&plot_mutex); ebpf_cleanup_plot_disks(); @@ -473,20 +458,6 @@ static void ebpf_disk_free(ebpf_module_t *em) static void ebpf_disk_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*disk_threads.thread); - ebpf_disk_free(em); -} - -/** - * Disk Cleanup - * - * Clean up allocated memory. - * - * @param ptr thread data. - */ -static void ebpf_disk_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_disk_free(em); } @@ -592,35 +563,6 @@ static void read_hard_disk_tables(int table) } /** - * Disk read hash - * - * This is the thread callback. - * This thread is necessary, because we cannot freeze the whole plugin to read the data on very busy socket. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_disk_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_disk_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_LATENCY_DISK_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_hard_disk_tables(disk_maps[NETDATA_DISK_READ].map_fd); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Obsolete Hard Disk charts * * Make Hard disk charts and fill chart name @@ -743,21 +685,19 @@ static void ebpf_latency_send_hd_data(int update_every) static void disk_collector(ebpf_module_t *em) { disk_hash_values = callocz(ebpf_nprocs, sizeof(netdata_idx_t)); - disk_threads.thread = mallocz(sizeof(netdata_thread_t)); - disk_threads.start_routine = ebpf_disk_read_hash; - - netdata_thread_create(disk_threads.thread, disk_threads.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_disk_read_hash, em); int update_every = em->update_every; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = update_every * USEC_PER_SEC; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; + read_hard_disk_tables(disk_maps[NETDATA_DISK_READ].map_fd); pthread_mutex_lock(&lock); ebpf_remove_pointer_from_plot_disk(em); ebpf_latency_send_hd_data(update_every); diff --git a/collectors/ebpf.plugin/ebpf_disk.h b/collectors/ebpf.plugin/ebpf_disk.h index c14b887f8..c606d6594 100644 --- a/collectors/ebpf.plugin/ebpf_disk.h +++ b/collectors/ebpf.plugin/ebpf_disk.h @@ -11,8 +11,6 @@ #define NETDATA_EBPF_PROC_PARTITIONS "/proc/partitions" -#define NETDATA_LATENCY_DISK_SLEEP_MS 650000ULL - // Process configuration name #define NETDATA_DISK_CONFIG_FILE "disk.conf" diff --git a/collectors/ebpf.plugin/ebpf_fd.c b/collectors/ebpf.plugin/ebpf_fd.c index 30b7f22ce..79537066c 100644 --- a/collectors/ebpf.plugin/ebpf_fd.c +++ b/collectors/ebpf.plugin/ebpf_fd.c @@ -6,6 +6,9 @@ static char *fd_dimension_names[NETDATA_FD_SYSCALL_END] = { "open", "close" }; static char *fd_id_names[NETDATA_FD_SYSCALL_END] = { "do_sys_open", "__close_fd" }; +static char *close_targets[NETDATA_EBPF_MAX_FD_TARGETS] = {"close_fd", "__close_fd"}; +static char *open_targets[NETDATA_EBPF_MAX_FD_TARGETS] = {"do_sys_openat2", "do_sys_open"}; + static netdata_syscall_stat_t fd_aggregated_data[NETDATA_FD_SYSCALL_END]; static netdata_publish_syscall_t fd_publish_aggregated[NETDATA_FD_SYSCALL_END]; @@ -29,15 +32,6 @@ struct config fd_config = { .first_section = NULL, .last_section = NULL, .mutex .index = {.avl_tree = { .root = NULL, .compar = appconfig_section_compare }, .rwlock = AVL_LOCK_INITIALIZER } }; -struct netdata_static_thread fd_thread = {"FD KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL}; - static netdata_idx_t fd_hash_values[NETDATA_FD_COUNTER]; static netdata_idx_t *fd_values = NULL; @@ -65,7 +59,7 @@ static inline void ebpf_fd_disable_probes(struct fd_bpf *obj) bpf_program__set_autoload(obj->progs.netdata_sys_open_kprobe, false); bpf_program__set_autoload(obj->progs.netdata_sys_open_kretprobe, false); bpf_program__set_autoload(obj->progs.netdata_release_task_fd_kprobe, false); - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_11) { + if (!strcmp(fd_targets[NETDATA_FD_SYSCALL_CLOSE].name, close_targets[NETDATA_FD_CLOSE_FD])) { bpf_program__set_autoload(obj->progs.netdata___close_fd_kretprobe, false); bpf_program__set_autoload(obj->progs.netdata___close_fd_kprobe, false); bpf_program__set_autoload(obj->progs.netdata_close_fd_kprobe, false); @@ -85,7 +79,7 @@ static inline void ebpf_fd_disable_probes(struct fd_bpf *obj) */ static inline void ebpf_disable_specific_probes(struct fd_bpf *obj) { - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_11) { + if (!strcmp(fd_targets[NETDATA_FD_SYSCALL_CLOSE].name, close_targets[NETDATA_FD_CLOSE_FD])) { bpf_program__set_autoload(obj->progs.netdata___close_fd_kretprobe, false); bpf_program__set_autoload(obj->progs.netdata___close_fd_kprobe, false); } else { @@ -121,7 +115,7 @@ static inline void ebpf_disable_trampoline(struct fd_bpf *obj) */ static inline void ebpf_disable_specific_trampoline(struct fd_bpf *obj) { - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_11) { + if (!strcmp(fd_targets[NETDATA_FD_SYSCALL_CLOSE].name, close_targets[NETDATA_FD_CLOSE_FD])) { bpf_program__set_autoload(obj->progs.netdata___close_fd_fentry, false); bpf_program__set_autoload(obj->progs.netdata___close_fd_fexit, false); } else { @@ -143,7 +137,7 @@ static void ebpf_set_trampoline_target(struct fd_bpf *obj) bpf_program__set_attach_target(obj->progs.netdata_sys_open_fexit, 0, fd_targets[NETDATA_FD_SYSCALL_OPEN].name); bpf_program__set_attach_target(obj->progs.netdata_release_task_fd_fentry, 0, EBPF_COMMON_FNCT_CLEAN_UP); - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_11) { + if (!strcmp(fd_targets[NETDATA_FD_SYSCALL_CLOSE].name, close_targets[NETDATA_FD_CLOSE_FD])) { bpf_program__set_attach_target( obj->progs.netdata_close_fd_fentry, 0, fd_targets[NETDATA_FD_SYSCALL_CLOSE].name); bpf_program__set_attach_target(obj->progs.netdata_close_fd_fexit, 0, fd_targets[NETDATA_FD_SYSCALL_CLOSE].name); @@ -185,7 +179,7 @@ static int ebpf_fd_attach_probe(struct fd_bpf *obj) if (ret) return -1; - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_11) { + if (!strcmp(fd_targets[NETDATA_FD_SYSCALL_CLOSE].name, close_targets[NETDATA_FD_CLOSE_FD])) { obj->links.netdata_close_fd_kretprobe = bpf_program__attach_kprobe(obj->progs.netdata_close_fd_kretprobe, true, fd_targets[NETDATA_FD_SYSCALL_CLOSE].name); ret = libbpf_get_error(obj->links.netdata_close_fd_kretprobe); @@ -217,24 +211,50 @@ static int ebpf_fd_attach_probe(struct fd_bpf *obj) } /** - * Set target values + * FD Fill Address * - * Set pointers used to laod data. + * Fill address value used to load probes/trampoline. */ -static void ebpf_fd_set_target_values() +static inline void ebpf_fd_fill_address(ebpf_addresses_t *address, char **targets) { - static char *close_targets[] = {"close_fd", "__close_fd"}; - static char *open_targets[] = {"do_sys_openat2", "do_sys_open"}; - if (running_on_kernel >= NETDATA_EBPF_KERNEL_5_11) { - fd_targets[NETDATA_FD_SYSCALL_OPEN].name = open_targets[0]; - fd_targets[NETDATA_FD_SYSCALL_CLOSE].name = close_targets[0]; - } else { - fd_targets[NETDATA_FD_SYSCALL_OPEN].name = open_targets[1]; - fd_targets[NETDATA_FD_SYSCALL_CLOSE].name = close_targets[1]; + int i; + for (i = 0; i < NETDATA_EBPF_MAX_FD_TARGETS; i++) { + address->function = targets[i]; + ebpf_load_addresses(address, -1); + if (address->addr) + break; } } /** + * Set target values + * + * Set pointers used to load data. + * + * @return It returns 0 on success and -1 otherwise. + */ +static int ebpf_fd_set_target_values() +{ + ebpf_addresses_t address = {.function = NULL, .hash = 0, .addr = 0}; + ebpf_fd_fill_address(&address, close_targets); + + if (!address.addr) + return -1; + + fd_targets[NETDATA_FD_SYSCALL_CLOSE].name = address.function; + + address.addr = 0; + ebpf_fd_fill_address(&address, open_targets); + + if (!address.addr) + return -1; + + fd_targets[NETDATA_FD_SYSCALL_OPEN].name = address.function; + + return 0; +} + +/** * Set hash tables * * Set the values for maps according the value given by kernel. @@ -283,14 +303,18 @@ static void ebpf_fd_disable_release_task(struct fd_bpf *obj) * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_fd_load_and_attach(struct fd_bpf *obj, ebpf_module_t *em) { netdata_ebpf_targets_t *mt = em->targets; netdata_ebpf_program_loaded_t test = mt[NETDATA_FD_SYSCALL_OPEN].mode; - ebpf_fd_set_target_values(); + if (ebpf_fd_set_target_values()) { + error("%s file descriptor.", NETDATA_EBPF_DEFAULT_FNT_NOT_FOUND); + return -1; + } + if (test == EBPF_LOAD_TRAMPOLINE) { ebpf_fd_disable_probes(obj); ebpf_disable_specific_trampoline(obj); @@ -340,15 +364,10 @@ static inline int ebpf_fd_load_and_attach(struct fd_bpf *obj, ebpf_module_t *em) static void ebpf_fd_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); ebpf_cleanup_publish_syscall(fd_publish_aggregated); - freez(fd_thread.thread); freez(fd_values); freez(fd_vector); @@ -372,18 +391,6 @@ static void ebpf_fd_free(ebpf_module_t *em) static void ebpf_fd_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*fd_thread.thread); - ebpf_fd_free(em); -} - -/** - * Clean up the main thread. - * - * @param ptr thread data. - */ -static void ebpf_fd_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_fd_free(em); } @@ -420,7 +427,7 @@ static void ebpf_fd_send_data(ebpf_module_t *em) * * Read the table with number of calls for all functions */ -static void read_global_table() +static void ebpf_fd_read_global_table() { uint32_t idx; netdata_idx_t *val = fd_hash_values; @@ -441,34 +448,6 @@ static void read_global_table() } /** - * File descriptor read hash - * - * This is the thread callback. - * This thread is necessary, because we cannot freeze the whole plugin to read the data. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_fd_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_fd_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - usec_t step = NETDATA_FD_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Apps Accumulator * * Sum all values read from kernel and store in the first address. @@ -942,22 +921,20 @@ static void ebpf_fd_send_cgroup_data(ebpf_module_t *em) */ static void fd_collector(ebpf_module_t *em) { - fd_thread.thread = mallocz(sizeof(netdata_thread_t)); - fd_thread.start_routine = ebpf_fd_read_hash; - - netdata_thread_create(fd_thread.thread, fd_thread.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_fd_read_hash, em); - int cgroups = em->cgroup_charts; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = em->update_every * USEC_PER_SEC; + int update_every = em->update_every; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; netdata_apps_integration_flags_t apps = em->apps_charts; + ebpf_fd_read_global_table(); pthread_mutex_lock(&collect_data_mutex); if (apps) read_apps_table(); diff --git a/collectors/ebpf.plugin/ebpf_fd.h b/collectors/ebpf.plugin/ebpf_fd.h index 914a34b98..e6545d79c 100644 --- a/collectors/ebpf.plugin/ebpf_fd.h +++ b/collectors/ebpf.plugin/ebpf_fd.h @@ -6,8 +6,6 @@ // Module name #define NETDATA_EBPF_MODULE_NAME_FD "filedescriptor" -#define NETDATA_FD_SLEEP_MS 850000ULL - // Menu group #define NETDATA_FILE_GROUP "file_access" @@ -36,9 +34,6 @@ #define NETDATA_SYSTEMD_FD_CLOSE_ERR_CONTEXT "services.fd_close_error" typedef struct netdata_fd_stat { - uint64_t pid_tgid; // Unique identifier - uint32_t pid; // Process ID - uint32_t open_call; // Open syscalls (open and openat) uint32_t close_call; // Close syscall (close) @@ -74,6 +69,14 @@ enum fd_syscalls { NETDATA_FD_SYSCALL_END }; +enum fd_close_syscall { + NETDATA_FD_CLOSE_FD, + NETDATA_FD___CLOSE_FD, + + NETDATA_FD_CLOSE_END +}; + +#define NETDATA_EBPF_MAX_FD_TARGETS 2 void *ebpf_fd_thread(void *ptr); void ebpf_fd_create_apps_charts(struct ebpf_module *em, void *ptr); diff --git a/collectors/ebpf.plugin/ebpf_filesystem.c b/collectors/ebpf.plugin/ebpf_filesystem.c index 7dbec7410..5250ed8af 100644 --- a/collectors/ebpf.plugin/ebpf_filesystem.c +++ b/collectors/ebpf.plugin/ebpf_filesystem.c @@ -30,17 +30,6 @@ static ebpf_local_maps_t fs_maps[] = {{.name = "tbl_ext4", .internal_input = NET .type = NETDATA_EBPF_MAP_CONTROLLER, .map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED}}; -struct netdata_static_thread filesystem_threads = { - .name = "EBPF FS READ", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - static netdata_syscall_stat_t filesystem_aggregated_data[NETDATA_EBPF_HIST_MAX_BINS]; static netdata_publish_syscall_t filesystem_publish_aggregated[NETDATA_EBPF_HIST_MAX_BINS]; @@ -337,14 +326,9 @@ void ebpf_filesystem_cleanup_ebpf_data() static void ebpf_filesystem_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); - freez(filesystem_threads.thread); ebpf_cleanup_publish_syscall(filesystem_publish_aggregated); ebpf_filesystem_cleanup_ebpf_data(); @@ -367,20 +351,6 @@ static void ebpf_filesystem_free(ebpf_module_t *em) static void ebpf_filesystem_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*filesystem_threads.thread); - ebpf_filesystem_free(em); -} - -/** - * File system cleanup - * - * Clean up allocated thread. - * - * @param ptr thread data. - */ -static void ebpf_filesystem_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_filesystem_free(em); } @@ -483,30 +453,16 @@ static void read_filesystem_tables() * * @return It always returns NULL. */ -void *ebpf_filesystem_read_hash(void *ptr) +void ebpf_filesystem_read_hash(ebpf_module_t *em) { - netdata_thread_cleanup_push(ebpf_filesystem_cleanup, ptr); - ebpf_module_t *em = (ebpf_module_t *)ptr; - - heartbeat_t hb; - heartbeat_init(&hb); - usec_t step = NETDATA_FILESYSTEM_READ_SLEEP_MS * em->update_every; - int update_every = em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); + ebpf_obsolete_fs_charts(em->update_every); - (void) ebpf_update_partitions(em); - ebpf_obsolete_fs_charts(update_every); + (void) ebpf_update_partitions(em); - // No more partitions, it is not necessary to read tables - if (em->optional) - continue; - - read_filesystem_tables(); - } + if (em->optional) + return; - netdata_thread_cleanup_pop(1); - return NULL; + read_filesystem_tables(); } /** @@ -543,21 +499,18 @@ static void ebpf_histogram_send_data() */ static void filesystem_collector(ebpf_module_t *em) { - filesystem_threads.thread = mallocz(sizeof(netdata_thread_t)); - filesystem_threads.start_routine = ebpf_filesystem_read_hash; - - netdata_thread_create(filesystem_threads.thread, filesystem_threads.name, - NETDATA_THREAD_OPTION_DEFAULT, ebpf_filesystem_read_hash, em); - int update_every = em->update_every; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = update_every * USEC_PER_SEC; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; + ebpf_filesystem_read_hash(em); pthread_mutex_lock(&lock); ebpf_create_fs_charts(update_every); diff --git a/collectors/ebpf.plugin/ebpf_filesystem.h b/collectors/ebpf.plugin/ebpf_filesystem.h index 0d558df7d..cf19b253e 100644 --- a/collectors/ebpf.plugin/ebpf_filesystem.h +++ b/collectors/ebpf.plugin/ebpf_filesystem.h @@ -11,7 +11,6 @@ #define NETDATA_FS_MAX_DIST_NAME 64UL #define NETDATA_FILESYSTEM_CONFIG_NAME "filesystem" -#define NETDATA_FILESYSTEM_READ_SLEEP_MS 600000ULL // Process configuration name #define NETDATA_FILESYSTEM_CONFIG_FILE "filesystem.conf" diff --git a/collectors/ebpf.plugin/ebpf_hardirq.c b/collectors/ebpf.plugin/ebpf_hardirq.c index b07dd24ca..20c4b9d05 100644 --- a/collectors/ebpf.plugin/ebpf_hardirq.c +++ b/collectors/ebpf.plugin/ebpf_hardirq.c @@ -135,17 +135,6 @@ static hardirq_ebpf_val_t *hardirq_ebpf_vals = NULL; // tmp store for static hard IRQ values we get from a per-CPU eBPF map. static hardirq_ebpf_static_val_t *hardirq_ebpf_static_vals = NULL; -static struct netdata_static_thread hardirq_threads = { - .name = "HARDIRQ KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - /** * Hardirq Free * @@ -156,21 +145,18 @@ static struct netdata_static_thread hardirq_threads = { static void ebpf_hardirq_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); - freez(hardirq_threads.thread); for (int i = 0; hardirq_tracepoints[i].class != NULL; i++) { ebpf_disable_tracepoint(&hardirq_tracepoints[i]); } freez(hardirq_ebpf_vals); freez(hardirq_ebpf_static_vals); + pthread_mutex_lock(&ebpf_exit_cleanup); em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; + pthread_mutex_unlock(&ebpf_exit_cleanup); } /** @@ -183,20 +169,6 @@ static void ebpf_hardirq_free(ebpf_module_t *em) static void hardirq_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*hardirq_threads.thread); - ebpf_hardirq_free(em); -} - -/** - * Hardirq clean up - * - * Clean up allocated memory. - * - * @param ptr thread data. - */ -static void hardirq_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_hardirq_free(em); } @@ -331,24 +303,10 @@ static void hardirq_read_latency_static_map(int mapfd) /** * Read eBPF maps for hard IRQ. */ -static void *hardirq_reader(void *ptr) +static void hardirq_reader() { - netdata_thread_cleanup_push(hardirq_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_HARDIRQ_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - hardirq_read_latency_map(hardirq_maps[HARDIRQ_MAP_LATENCY].map_fd); - hardirq_read_latency_static_map(hardirq_maps[HARDIRQ_MAP_LATENCY_STATIC].map_fd); - } - - netdata_thread_cleanup_pop(1); - return NULL; + hardirq_read_latency_map(hardirq_maps[HARDIRQ_MAP_LATENCY].map_fd); + hardirq_read_latency_static_map(hardirq_maps[HARDIRQ_MAP_LATENCY_STATIC].map_fd); } static void hardirq_create_charts(int update_every) @@ -428,17 +386,6 @@ static void hardirq_collector(ebpf_module_t *em) avl_init_lock(&hardirq_pub, hardirq_val_cmp); - // create reader thread. - hardirq_threads.thread = mallocz(sizeof(netdata_thread_t)); - hardirq_threads.start_routine = hardirq_reader; - netdata_thread_create( - hardirq_threads.thread, - hardirq_threads.name, - NETDATA_THREAD_OPTION_DEFAULT, - hardirq_reader, - em - ); - // create chart and static dims. pthread_mutex_lock(&lock); hardirq_create_charts(em->update_every); @@ -449,13 +396,17 @@ static void hardirq_collector(ebpf_module_t *em) // loop and read from published data until ebpf plugin is closed. heartbeat_t hb; heartbeat_init(&hb); - usec_t step = em->update_every * USEC_PER_SEC; + int update_every = em->update_every; + int counter = update_every - 1; //This will be cancelled by its parent while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; + hardirq_reader(); pthread_mutex_lock(&lock); // write dims now for all hitherto discovered IRQs. diff --git a/collectors/ebpf.plugin/ebpf_hardirq.h b/collectors/ebpf.plugin/ebpf_hardirq.h index 381da57d8..fe38b1bb1 100644 --- a/collectors/ebpf.plugin/ebpf_hardirq.h +++ b/collectors/ebpf.plugin/ebpf_hardirq.h @@ -47,7 +47,6 @@ typedef struct hardirq_ebpf_static_val { *****************************************************************/ #define NETDATA_EBPF_MODULE_NAME_HARDIRQ "hardirq" -#define NETDATA_HARDIRQ_SLEEP_MS 650000ULL #define NETDATA_HARDIRQ_CONFIG_FILE "hardirq.conf" typedef struct hardirq_val { diff --git a/collectors/ebpf.plugin/ebpf_mdflush.c b/collectors/ebpf.plugin/ebpf_mdflush.c index dc805da23..1a5a7731e 100644 --- a/collectors/ebpf.plugin/ebpf_mdflush.c +++ b/collectors/ebpf.plugin/ebpf_mdflush.c @@ -35,17 +35,6 @@ static avl_tree_lock mdflush_pub; // tmp store for mdflush values we get from a per-CPU eBPF map. static mdflush_ebpf_val_t *mdflush_ebpf_vals = NULL; -static struct netdata_static_thread mdflush_threads = { - .name = "MDFLUSH KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - /** * MDflush Free * @@ -55,18 +44,10 @@ static struct netdata_static_thread mdflush_threads = { */ static void ebpf_mdflush_free(ebpf_module_t *em) { - pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } - pthread_mutex_unlock(&ebpf_exit_cleanup); - freez(mdflush_ebpf_vals); - freez(mdflush_threads.thread); - + pthread_mutex_lock(&ebpf_exit_cleanup); em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; + pthread_mutex_unlock(&ebpf_exit_cleanup); } /** @@ -83,20 +64,6 @@ static void mdflush_exit(void *ptr) } /** - * CLeanup - * - * Clean allocated memory. - * - * @param ptr thread data. - */ -static void mdflush_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*mdflush_threads.thread); - ebpf_mdflush_free(em); -} - -/** * Compare mdflush values. * * @param a `netdata_mdflush_t *`. @@ -188,28 +155,6 @@ static void mdflush_read_count_map() } } -/** - * Read eBPF maps for mdflush. - */ -static void *mdflush_reader(void *ptr) -{ - netdata_thread_cleanup_push(mdflush_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_MDFLUSH_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - mdflush_read_count_map(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - static void mdflush_create_charts(int update_every) { ebpf_create_chart( @@ -256,34 +201,27 @@ static void mdflush_collector(ebpf_module_t *em) { mdflush_ebpf_vals = callocz(ebpf_nprocs, sizeof(mdflush_ebpf_val_t)); + int update_every = em->update_every; avl_init_lock(&mdflush_pub, mdflush_val_cmp); - // create reader thread. - mdflush_threads.thread = mallocz(sizeof(netdata_thread_t)); - mdflush_threads.start_routine = mdflush_reader; - netdata_thread_create( - mdflush_threads.thread, - mdflush_threads.name, - NETDATA_THREAD_OPTION_DEFAULT, - mdflush_reader, - em - ); - // create chart and static dims. pthread_mutex_lock(&lock); - mdflush_create_charts(em->update_every); + mdflush_create_charts(update_every); ebpf_update_stats(&plugin_statistics, em); pthread_mutex_unlock(&lock); // loop and read from published data until ebpf plugin is closed. heartbeat_t hb; heartbeat_init(&hb); - usec_t step = em->update_every * USEC_PER_SEC; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + + counter = 0; + mdflush_read_count_map(); // write dims now for all hitherto discovered devices. write_begin_chart("mdstat", "mdstat_flush"); avl_traverse_lock(&mdflush_pub, mdflush_write_dims, NULL); diff --git a/collectors/ebpf.plugin/ebpf_mdflush.h b/collectors/ebpf.plugin/ebpf_mdflush.h index b04eefd28..4913ad019 100644 --- a/collectors/ebpf.plugin/ebpf_mdflush.h +++ b/collectors/ebpf.plugin/ebpf_mdflush.h @@ -6,8 +6,6 @@ // Module name #define NETDATA_EBPF_MODULE_NAME_MDFLUSH "mdflush" -#define NETDATA_MDFLUSH_SLEEP_MS 850000ULL - // charts #define NETDATA_MDFLUSH_GLOBAL_CHART "mdflush" diff --git a/collectors/ebpf.plugin/ebpf_mount.c b/collectors/ebpf.plugin/ebpf_mount.c index ec1f07a65..e06010b5b 100644 --- a/collectors/ebpf.plugin/ebpf_mount.c +++ b/collectors/ebpf.plugin/ebpf_mount.c @@ -22,17 +22,6 @@ static netdata_idx_t *mount_values = NULL; static netdata_idx_t mount_hash_values[NETDATA_MOUNT_END]; -struct netdata_static_thread mount_thread = { - .name = "MOUNT KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - netdata_ebpf_targets_t mount_targets[] = { {.name = "mount", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = "umount", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = NULL, .mode = EBPF_LOAD_TRAMPOLINE}}; @@ -187,7 +176,7 @@ static void ebpf_mount_set_hash_tables(struct mount_bpf *obj) * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_mount_load_and_attach(struct mount_bpf *obj, ebpf_module_t *em) { @@ -239,14 +228,9 @@ static inline int ebpf_mount_load_and_attach(struct mount_bpf *obj, ebpf_module_ static void ebpf_mount_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); - freez(mount_thread.thread); freez(mount_values); #ifdef LIBBPF_MAJOR_VERSION @@ -269,20 +253,6 @@ static void ebpf_mount_free(ebpf_module_t *em) static void ebpf_mount_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*mount_thread.thread); - ebpf_mount_free(em); -} - -/** - * Mount cleanup - * - * Clean up allocated memory. - * - * @param ptr thread data. - */ -static void ebpf_mount_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_mount_free(em); } @@ -297,7 +267,7 @@ static void ebpf_mount_cleanup(void *ptr) * * Read the table with number of calls for all functions */ -static void read_global_table() +static void ebpf_mount_read_global_table() { uint32_t idx; netdata_idx_t *val = mount_hash_values; @@ -318,36 +288,6 @@ static void read_global_table() } /** - * Mount read hash - * - * This is the thread callback. - * This thread is necessary, because we cannot freeze the whole plugin to read the data. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_mount_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_mount_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_LATENCY_MOUNT_SLEEP_MS * em->update_every; - //This will be cancelled by its parent - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Send data to Netdata calling auxiliary functions. */ static void ebpf_mount_send_data() @@ -371,23 +311,20 @@ static void ebpf_mount_send_data() */ static void mount_collector(ebpf_module_t *em) { - mount_thread.thread = mallocz(sizeof(netdata_thread_t)); - mount_thread.start_routine = ebpf_mount_read_hash; - memset(mount_hash_values, 0, sizeof(mount_hash_values)); - mount_values = callocz((size_t)ebpf_nprocs, sizeof(netdata_idx_t)); - - netdata_thread_create(mount_thread.thread, mount_thread.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_mount_read_hash, em); + memset(mount_hash_values, 0, sizeof(mount_hash_values)); heartbeat_t hb; heartbeat_init(&hb); - usec_t step = em->update_every * USEC_PER_SEC; + int update_every = em->update_every; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; + ebpf_mount_read_global_table(); pthread_mutex_lock(&lock); ebpf_mount_send_data(); diff --git a/collectors/ebpf.plugin/ebpf_mount.h b/collectors/ebpf.plugin/ebpf_mount.h index 5a8d11a59..11b21f832 100644 --- a/collectors/ebpf.plugin/ebpf_mount.h +++ b/collectors/ebpf.plugin/ebpf_mount.h @@ -8,8 +8,6 @@ #define NETDATA_EBPF_MOUNT_SYSCALL 2 -#define NETDATA_LATENCY_MOUNT_SLEEP_MS 700000ULL - #define NETDATA_EBPF_MOUNT_CALLS "call" #define NETDATA_EBPF_MOUNT_ERRORS "error" #define NETDATA_EBPF_MOUNT_FAMILY "mount (eBPF)" diff --git a/collectors/ebpf.plugin/ebpf_oomkill.c b/collectors/ebpf.plugin/ebpf_oomkill.c index d93e4159e..82420d54e 100644 --- a/collectors/ebpf.plugin/ebpf_oomkill.c +++ b/collectors/ebpf.plugin/ebpf_oomkill.c @@ -46,7 +46,9 @@ static netdata_publish_syscall_t oomkill_publish_aggregated = {.name = "oomkill" static void oomkill_cleanup(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; + pthread_mutex_lock(&ebpf_exit_cleanup); em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; + pthread_mutex_unlock(&ebpf_exit_cleanup); } static void oomkill_write_data(int32_t *keys, uint32_t total) @@ -294,12 +296,13 @@ static void oomkill_collector(ebpf_module_t *em) // loop and read until ebpf plugin is closed. heartbeat_t hb; heartbeat_init(&hb); - usec_t step = update_every * USEC_PER_SEC; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (!ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; pthread_mutex_lock(&collect_data_mutex); pthread_mutex_lock(&lock); diff --git a/collectors/ebpf.plugin/ebpf_oomkill.h b/collectors/ebpf.plugin/ebpf_oomkill.h index 786086384..f921f9d87 100644 --- a/collectors/ebpf.plugin/ebpf_oomkill.h +++ b/collectors/ebpf.plugin/ebpf_oomkill.h @@ -17,7 +17,6 @@ typedef uint8_t oomkill_ebpf_val_t; *****************************************************************/ #define NETDATA_EBPF_MODULE_NAME_OOMKILL "oomkill" -#define NETDATA_OOMKILL_SLEEP_MS 650000ULL #define NETDATA_OOMKILL_CONFIG_FILE "oomkill.conf" #define NETDATA_OOMKILL_CHART "oomkills" diff --git a/collectors/ebpf.plugin/ebpf_process.c b/collectors/ebpf.plugin/ebpf_process.c index 682577da7..9a191d391 100644 --- a/collectors/ebpf.plugin/ebpf_process.c +++ b/collectors/ebpf.plugin/ebpf_process.c @@ -57,17 +57,6 @@ struct config process_config = { .first_section = NULL, static char *threads_stat[NETDATA_EBPF_THREAD_STAT_END] = {"total", "running"}; static char *load_event_stat[NETDATA_EBPF_LOAD_STAT_END] = {"legacy", "co-re"}; -static struct netdata_static_thread cgroup_thread = { - .name = "EBPF CGROUP", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - /***************************************************************** * * PROCESS DATA AND SEND TO NETDATA @@ -327,55 +316,6 @@ static void ebpf_process_update_apps_data() } /** - * Cgroup Exit - * - * Function used with netdata_thread_clean_push - * - * @param ptr unused argument - */ -static void ebpf_cgroup_exit(void *ptr) -{ - UNUSED(ptr); -} - -/** - * Cgroup update shm - * - * This is the thread callback. - * This thread is necessary, because we cannot freeze the whole plugin to read the data from shared memory. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_cgroup_update_shm(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_cgroup_exit, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - usec_t step = 3 * USEC_PER_SEC; - int counter = NETDATA_EBPF_CGROUP_UPDATE - 1; - //This will be cancelled by its parent - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - // We are using a small heartbeat time to wake up thread, - // but we should not update so frequently the shared memory data - if (++counter >= NETDATA_EBPF_CGROUP_UPDATE) { - counter = 0; - if (!shm_ebpf_cgroup.header) - ebpf_map_cgroup_shared_memory(); - - ebpf_parse_cgroup_shm_data(); - } - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Update cgroup * * Update cgroup data based in @@ -745,7 +685,6 @@ static void ebpf_process_exit(void *ptr) pthread_mutex_lock(&ebpf_exit_cleanup); em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; pthread_mutex_unlock(&ebpf_exit_cleanup); - pthread_cancel(*cgroup_thread.thread); } /***************************************************************** @@ -1104,13 +1043,6 @@ void ebpf_send_statistic_data() */ static void process_collector(ebpf_module_t *em) { - // Start cgroup integration before other threads - cgroup_thread.thread = mallocz(sizeof(netdata_thread_t)); - cgroup_thread.start_routine = ebpf_cgroup_update_shm; - - netdata_thread_create(cgroup_thread.thread, cgroup_thread.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_cgroup_update_shm, NULL); - heartbeat_t hb; heartbeat_init(&hb); int publish_global = em->global_charts; @@ -1152,7 +1084,7 @@ static void process_collector(ebpf_module_t *em) ebpf_process_update_apps_data(); } - if (cgroups) { + if (cgroups && shm_ebpf_cgroup.header) { ebpf_update_process_cgroup(); } } @@ -1169,7 +1101,7 @@ static void process_collector(ebpf_module_t *em) ebpf_process_send_apps_data(apps_groups_root_target, em); } - if (cgroups) { + if (cgroups && shm_ebpf_cgroup.header) { ebpf_process_send_cgroup_data(em); } } diff --git a/collectors/ebpf.plugin/ebpf_process.h b/collectors/ebpf.plugin/ebpf_process.h index 43df34d48..6fded16fc 100644 --- a/collectors/ebpf.plugin/ebpf_process.h +++ b/collectors/ebpf.plugin/ebpf_process.h @@ -39,7 +39,7 @@ #define NETDATA_SYSTEMD_PROCESS_EXIT_CONTEXT "services.task_exit" #define NETDATA_SYSTEMD_PROCESS_ERROR_CONTEXT "services.task_error" -#define NETDATA_EBPF_CGROUP_UPDATE 10 +#define NETDATA_EBPF_CGROUP_UPDATE 30 // Statistical information enum netdata_ebpf_thread_stats{ diff --git a/collectors/ebpf.plugin/ebpf_shm.c b/collectors/ebpf.plugin/ebpf_shm.c index f81287d82..4057eff7f 100644 --- a/collectors/ebpf.plugin/ebpf_shm.c +++ b/collectors/ebpf.plugin/ebpf_shm.c @@ -34,17 +34,6 @@ static ebpf_local_maps_t shm_maps[] = {{.name = "tbl_pid_shm", .internal_input = .map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED}, {.name = NULL, .internal_input = 0, .user_input = 0}}; -struct netdata_static_thread shm_threads = { - .name = "SHM KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - netdata_ebpf_targets_t shm_targets[] = { {.name = "shmget", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = "shmat", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = "shmdt", .mode = EBPF_LOAD_TRAMPOLINE}, @@ -246,7 +235,7 @@ static void ebpf_shm_adjust_map_size(struct shm_bpf *obj, ebpf_module_t *em) * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_shm_load_and_attach(struct shm_bpf *obj, ebpf_module_t *em) { @@ -299,11 +288,7 @@ static inline int ebpf_shm_load_and_attach(struct shm_bpf *obj, ebpf_module_t *e static void ebpf_shm_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); ebpf_cleanup_publish_syscall(shm_publish_aggregated); @@ -316,7 +301,9 @@ static void ebpf_shm_free(ebpf_module_t *em) shm_bpf__destroy(bpf_obj); #endif + pthread_mutex_lock(&ebpf_exit_cleanup); em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; + pthread_mutex_unlock(&ebpf_exit_cleanup); } /** @@ -329,20 +316,6 @@ static void ebpf_shm_free(ebpf_module_t *em) static void ebpf_shm_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*shm_threads.thread); - ebpf_shm_free(em); -} - -/** - * SHM Cleanup - * - * Clean up allocated memory. - * - * @param ptr thread data. - */ -static void ebpf_shm_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_shm_free(em); } @@ -491,7 +464,7 @@ static void shm_send_global() * * Read the table with number of calls for all functions */ -static void read_global_table() +static void ebpf_shm_read_global_table() { netdata_idx_t *stored = shm_values; netdata_idx_t *val = shm_hash_values; @@ -512,30 +485,6 @@ static void read_global_table() } /** - * Shared memory reader thread. - * - * @param ptr It is a NULL value for this thread. - * @return It always returns NULL. - */ -void *ebpf_shm_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_shm_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - usec_t step = NETDATA_SHM_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Sum values for all targets. */ static void ebpf_shm_sum_pids(netdata_publish_shm_t *shm, struct pid_on_target *root) @@ -894,28 +843,19 @@ void ebpf_shm_send_cgroup_data(int update_every) */ static void shm_collector(ebpf_module_t *em) { - shm_threads.thread = mallocz(sizeof(netdata_thread_t)); - shm_threads.start_routine = ebpf_shm_read_hash; - - netdata_thread_create( - shm_threads.thread, - shm_threads.name, - NETDATA_THREAD_OPTION_DEFAULT, - ebpf_shm_read_hash, - em - ); - int cgroups = em->cgroup_charts; int update_every = em->update_every; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = update_every * USEC_PER_SEC; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; netdata_apps_integration_flags_t apps = em->apps_charts; + ebpf_shm_read_global_table(); pthread_mutex_lock(&collect_data_mutex); if (apps) { read_apps_table(); diff --git a/collectors/ebpf.plugin/ebpf_shm.h b/collectors/ebpf.plugin/ebpf_shm.h index 4e068819b..b06a4a5d1 100644 --- a/collectors/ebpf.plugin/ebpf_shm.h +++ b/collectors/ebpf.plugin/ebpf_shm.h @@ -6,8 +6,6 @@ // Module name #define NETDATA_EBPF_MODULE_NAME_SHM "shm" -#define NETDATA_SHM_SLEEP_MS 850000ULL - // charts #define NETDATA_SHM_GLOBAL_CHART "shared_memory_calls" #define NETDATA_SHMGET_CHART "shmget_call" diff --git a/collectors/ebpf.plugin/ebpf_socket.c b/collectors/ebpf.plugin/ebpf_socket.c index 3a023e4a4..1954be714 100644 --- a/collectors/ebpf.plugin/ebpf_socket.c +++ b/collectors/ebpf.plugin/ebpf_socket.c @@ -62,8 +62,6 @@ ebpf_socket_publish_apps_t **socket_bandwidth_curr = NULL; static ebpf_bandwidth_t *bandwidth_vector = NULL; pthread_mutex_t nv_mutex; -int wait_to_plot = 0; - netdata_vector_plot_t inbound_vectors = { .plot = NULL, .next = 0, .last = 0 }; netdata_vector_plot_t outbound_vectors = { .plot = NULL, .next = 0, .last = 0 }; netdata_socket_t *socket_values; @@ -389,7 +387,7 @@ static void ebpf_socket_adjust_map_size(struct socket_bpf *obj, ebpf_module_t *e * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_socket_load_and_attach(struct socket_bpf *obj, ebpf_module_t *em) { @@ -459,6 +457,9 @@ static inline void clean_internal_socket_plot(netdata_socket_plot_t *ptr) */ static void clean_allocated_socket_plot() { + if (!network_viewer_opt.enabled) + return; + uint32_t i; uint32_t end = inbound_vectors.last; netdata_socket_plot_t *plot = inbound_vectors.plot; @@ -647,7 +648,8 @@ static void ebpf_socket_free(ebpf_module_t *em ) static void ebpf_socket_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*socket_threads.thread); + if (socket_threads.thread) + netdata_thread_cancel(*socket_threads.thread); ebpf_socket_free(em); } @@ -724,7 +726,7 @@ static void ebpf_update_global_publish( */ static inline void update_nv_plot_data(netdata_plot_values_t *plot, netdata_socket_t *sock) { - if (sock->ct > plot->last_time) { + if (sock->ct != plot->last_time) { plot->last_time = sock->ct; plot->plot_recv_packets = sock->recv_packets; plot->plot_sent_packets = sock->sent_packets; @@ -747,6 +749,7 @@ static inline void update_nv_plot_data(netdata_plot_values_t *plot, netdata_sock */ static inline void calculate_nv_plot() { + pthread_mutex_lock(&nv_mutex); uint32_t i; uint32_t end = inbound_vectors.next; for (i = 0; i < end; i++) { @@ -764,9 +767,12 @@ static inline void calculate_nv_plot() } outbound_vectors.max_plot = end; + /* // The 'Other' dimension is always calculated for the chart to have at least one dimension update_nv_plot_data(&outbound_vectors.plot[outbound_vectors.last].plot, &outbound_vectors.plot[outbound_vectors.last].sock); + */ + pthread_mutex_unlock(&nv_mutex); } /** @@ -1441,17 +1447,17 @@ static void ebpf_socket_create_nv_charts(netdata_vector_plot_t *ptr, int update_ * * @return It returns 1 if the IP is inside the range and 0 otherwise */ -static int is_specific_ip_inside_range(union netdata_ip_t *cmp, int family) +static int ebpf_is_specific_ip_inside_range(union netdata_ip_t *cmp, int family) { if (!network_viewer_opt.excluded_ips && !network_viewer_opt.included_ips) return 1; - uint32_t ipv4_test = ntohl(cmp->addr32[0]); + uint32_t ipv4_test = htonl(cmp->addr32[0]); ebpf_network_viewer_ip_list_t *move = network_viewer_opt.excluded_ips; while (move) { if (family == AF_INET) { - if (ntohl(move->first.addr32[0]) <= ipv4_test && - ipv4_test <= ntohl(move->last.addr32[0]) ) + if (move->first.addr32[0] <= ipv4_test && + ipv4_test <= move->last.addr32[0]) return 0; } else { if (memcmp(move->first.addr8, cmp->addr8, sizeof(union netdata_ip_t)) <= 0 && @@ -1464,12 +1470,13 @@ static int is_specific_ip_inside_range(union netdata_ip_t *cmp, int family) move = network_viewer_opt.included_ips; while (move) { - if (family == AF_INET) { - if (ntohl(move->first.addr32[0]) <= ipv4_test && - ntohl(move->last.addr32[0]) >= ipv4_test) + if (family == AF_INET && move->ver == AF_INET) { + if (move->first.addr32[0] <= ipv4_test && + move->last.addr32[0] >= ipv4_test) return 1; } else { - if (memcmp(move->first.addr8, cmp->addr8, sizeof(union netdata_ip_t)) <= 0 && + if (move->ver == AF_INET6 && + memcmp(move->first.addr8, cmp->addr8, sizeof(union netdata_ip_t)) <= 0 && memcmp(move->last.addr8, cmp->addr8, sizeof(union netdata_ip_t)) >= 0) { return 1; } @@ -1565,7 +1572,7 @@ int is_socket_allowed(netdata_socket_idx_t *key, int family) if (!is_port_inside_range(key->dport)) return 0; - return is_specific_ip_inside_range(&key->daddr, family); + return ebpf_is_specific_ip_inside_range(&key->daddr, family); } /** @@ -1580,38 +1587,26 @@ int is_socket_allowed(netdata_socket_idx_t *key, int family) * * @return It returns 0 case the values are equal, 1 case a is bigger than b and -1 case a is smaller than b. */ -static int compare_sockets(void *a, void *b) +static int ebpf_compare_sockets(void *a, void *b) { struct netdata_socket_plot *val1 = a; struct netdata_socket_plot *val2 = b; - int cmp; + int cmp = 0; // We do not need to compare val2 family, because data inside hash table is always from the same family if (val1->family == AF_INET) { //IPV4 - if (val1->flags & NETDATA_INBOUND_DIRECTION) { - if (val1->index.sport == val2->index.sport) - cmp = 0; - else { - cmp = (val1->index.sport > val2->index.sport)?1:-1; - } - } else { + if (network_viewer_opt.included_port || network_viewer_opt.excluded_port) cmp = memcmp(&val1->index.dport, &val2->index.dport, sizeof(uint16_t)); - if (!cmp) { - cmp = memcmp(&val1->index.daddr.addr32[0], &val2->index.daddr.addr32[0], sizeof(uint32_t)); - } + + if (!cmp) { + cmp = memcmp(&val1->index.daddr.addr32[0], &val2->index.daddr.addr32[0], sizeof(uint32_t)); } } else { - if (val1->flags & NETDATA_INBOUND_DIRECTION) { - if (val1->index.sport == val2->index.sport) - cmp = 0; - else { - cmp = (val1->index.sport > val2->index.sport)?1:-1; - } - } else { + if (network_viewer_opt.included_port || network_viewer_opt.excluded_port) cmp = memcmp(&val1->index.dport, &val2->index.dport, sizeof(uint16_t)); - if (!cmp) { - cmp = memcmp(&val1->index.daddr.addr32, &val2->index.daddr.addr32, 4*sizeof(uint32_t)); - } + + if (!cmp) { + cmp = memcmp(&val1->index.daddr.addr32, &val2->index.daddr.addr32, 4*sizeof(uint32_t)); } } @@ -1631,12 +1626,15 @@ static int compare_sockets(void *a, void *b) * * @return it returns the size of the data copied on success and -1 otherwise. */ -static inline int build_outbound_dimension_name(char *dimname, char *hostname, char *service_name, - char *proto, int family) +static inline int ebpf_build_outbound_dimension_name(char *dimname, char *hostname, char *service_name, + char *proto, int family) { - return snprintf(dimname, CONFIG_MAX_NAME - 7, (family == AF_INET)?"%s:%s:%s_":"%s:%s:[%s]_", - service_name, proto, - hostname); + if (network_viewer_opt.included_port || network_viewer_opt.excluded_port) + return snprintf(dimname, CONFIG_MAX_NAME - 7, (family == AF_INET)?"%s:%s:%s_":"%s:%s:[%s]_", + service_name, proto, hostname); + + return snprintf(dimname, CONFIG_MAX_NAME - 7, (family == AF_INET)?"%s:%s_":"%s:[%s]_", + proto, hostname); } /** @@ -1692,7 +1690,7 @@ static inline void fill_resolved_name(netdata_socket_plot_t *ptr, char *hostname } if (is_outbound) - size = build_outbound_dimension_name(dimname, hostname, service_name, protocol, ptr->family); + size = ebpf_build_outbound_dimension_name(dimname, hostname, service_name, protocol, ptr->family); else size = build_inbound_dimension_name(dimname,service_name, protocol); @@ -1850,14 +1848,12 @@ static void fill_last_nv_dimension(netdata_socket_plot_t *ptr, int is_outbound) */ static inline void update_socket_data(netdata_socket_t *sock, netdata_socket_t *lvalues) { - sock->recv_packets += lvalues->recv_packets; - sock->sent_packets += lvalues->sent_packets; - sock->recv_bytes += lvalues->recv_bytes; - sock->sent_bytes += lvalues->sent_bytes; - sock->retransmit += lvalues->retransmit; - - if (lvalues->ct > sock->ct) - sock->ct = lvalues->ct; + sock->recv_packets = lvalues->recv_packets; + sock->sent_packets = lvalues->sent_packets; + sock->recv_bytes = lvalues->recv_bytes; + sock->sent_bytes = lvalues->sent_bytes; + sock->retransmit = lvalues->retransmit; + sock->ct = lvalues->ct; } /** @@ -1881,7 +1877,7 @@ static void store_socket_inside_avl(netdata_vector_plot_t *out, netdata_socket_t ret = (netdata_socket_plot_t *) avl_search_lock(&out->tree, (avl_t *)&test); if (ret) { - if (lvalues->ct > ret->plot.last_time) { + if (lvalues->ct != ret->plot.last_time) { update_socket_data(&ret->sock, lvalues); } } else { @@ -1892,7 +1888,7 @@ static void store_socket_inside_avl(netdata_vector_plot_t *out, netdata_socket_t int resolved; if (curr == last) { - if (lvalues->ct > w->plot.last_time) { + if (lvalues->ct != w->plot.last_time) { update_socket_data(&w->sock, lvalues); } return; @@ -1977,6 +1973,9 @@ netdata_vector_plot_t * select_vector_to_store(uint32_t *direction, netdata_sock */ static void hash_accumulator(netdata_socket_t *values, netdata_socket_idx_t *key, int family, int end) { + if (!network_viewer_opt.enabled || !is_socket_allowed(key, family)) + return; + uint64_t bsent = 0, brecv = 0, psent = 0, precv = 0; uint16_t retransmit = 0; int i; @@ -1994,7 +1993,7 @@ static void hash_accumulator(netdata_socket_t *values, netdata_socket_idx_t *key if (!protocol) protocol = w->protocol; - if (w->ct > ct) + if (w->ct != ct) ct = w->ct; } @@ -2006,11 +2005,9 @@ static void hash_accumulator(netdata_socket_t *values, netdata_socket_idx_t *key values[0].protocol = (!protocol)?IPPROTO_TCP:protocol; values[0].ct = ct; - if (is_socket_allowed(key, family)) { - uint32_t dir; - netdata_vector_plot_t *table = select_vector_to_store(&dir, key, protocol); - store_socket_inside_avl(table, &values[0], key, family, dir); - } + uint32_t dir; + netdata_vector_plot_t *table = select_vector_to_store(&dir, key, protocol); + store_socket_inside_avl(table, &values[0], key, family, dir); } /** @@ -2018,16 +2015,13 @@ static void hash_accumulator(netdata_socket_t *values, netdata_socket_idx_t *key * * Read data from hash tables created on kernel ring. * - * @param fd the hash table with data. - * @param family the family associated to the hash table + * @param fd the hash table with data. + * @param family the family associated to the hash table * * @return it returns 0 on success and -1 otherwise. */ -static void read_socket_hash_table(int fd, int family, int network_connection) +static void ebpf_read_socket_hash_table(int fd, int family) { - if (wait_to_plot) - return; - netdata_socket_idx_t key = {}; netdata_socket_idx_t next_key = {}; @@ -2046,9 +2040,7 @@ static void read_socket_hash_table(int fd, int family, int network_connection) continue; } - if (network_connection) { - hash_accumulator(values, &key, family, end); - } + hash_accumulator(values, &key, family, end); key = next_key; } @@ -2057,12 +2049,12 @@ static void read_socket_hash_table(int fd, int family, int network_connection) /** * Fill Network Viewer Port list * - * Fill the strcture with values read from /proc or hash table. + * Fill the structure with values read from /proc or hash table. * * @param out the structure where we will store data. * @param value the ports we are listen to. * @param proto the protocol used for this connection. - * @param in the strcuture with values read form different sources. + * @param in the structure with values read form different sources. */ static inline void fill_nv_port_list(ebpf_network_viewer_port_list_t *out, uint16_t value, uint16_t proto, netdata_passive_connection_t *in) @@ -2081,7 +2073,7 @@ static inline void fill_nv_port_list(ebpf_network_viewer_port_list_t *out, uint1 * * @param value the ports we are listen to. * @param proto the protocol used with port connection. - * @param in the strcuture with values read form different sources. + * @param in the structure with values read form different sources. */ void update_listen_table(uint16_t value, uint16_t proto, netdata_passive_connection_t *in) { @@ -2159,22 +2151,19 @@ static void read_listen_table() void *ebpf_socket_read_hash(void *ptr) { netdata_thread_cleanup_push(ebpf_socket_cleanup, ptr); - ebpf_module_t *em = (ebpf_module_t *)ptr; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = NETDATA_SOCKET_READ_SLEEP_MS * em->update_every; int fd_ipv4 = socket_maps[NETDATA_SOCKET_TABLE_IPV4].map_fd; int fd_ipv6 = socket_maps[NETDATA_SOCKET_TABLE_IPV6].map_fd; - int network_connection = em->optional; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin) + continue; pthread_mutex_lock(&nv_mutex); - read_listen_table(); - read_socket_hash_table(fd_ipv4, AF_INET, network_connection); - read_socket_hash_table(fd_ipv6, AF_INET6, network_connection); - wait_to_plot = 1; + ebpf_read_socket_hash_table(fd_ipv4, AF_INET); + ebpf_read_socket_hash_table(fd_ipv6, AF_INET6); pthread_mutex_unlock(&nv_mutex); } @@ -2863,44 +2852,50 @@ static void ebpf_socket_send_cgroup_data(int update_every) /** * Main loop for this collector. * - * @param step the number of microseconds used with heart beat * @param em the structure with thread information */ -static void socket_collector(usec_t step, ebpf_module_t *em) +static void socket_collector(ebpf_module_t *em) { heartbeat_t hb; heartbeat_init(&hb); + uint32_t network_connection = network_viewer_opt.enabled; - socket_threads.thread = mallocz(sizeof(netdata_thread_t)); - socket_threads.start_routine = ebpf_socket_read_hash; + if (network_connection) { + socket_threads.thread = mallocz(sizeof(netdata_thread_t)); + socket_threads.start_routine = ebpf_socket_read_hash; - netdata_thread_create(socket_threads.thread, socket_threads.name, - NETDATA_THREAD_OPTION_DEFAULT, ebpf_socket_read_hash, em); + netdata_thread_create(socket_threads.thread, socket_threads.name, + NETDATA_THREAD_OPTION_DEFAULT, ebpf_socket_read_hash, em); + } int cgroups = em->cgroup_charts; if (cgroups) ebpf_socket_update_cgroup_algorithm(); int socket_global_enabled = em->global_charts; - int network_connection = em->optional; int update_every = em->update_every; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; netdata_apps_integration_flags_t socket_apps_enabled = em->apps_charts; - pthread_mutex_lock(&collect_data_mutex); - if (socket_global_enabled) + if (socket_global_enabled) { + read_listen_table(); read_hash_global_tables(); + } + pthread_mutex_lock(&collect_data_mutex); if (socket_apps_enabled) ebpf_socket_update_apps_data(); if (cgroups) ebpf_update_socket_cgroup(); - calculate_nv_plot(); + if (network_connection) + calculate_nv_plot(); pthread_mutex_lock(&lock); if (socket_global_enabled) @@ -2925,7 +2920,6 @@ static void socket_collector(usec_t step, ebpf_module_t *em) ebpf_socket_create_nv_charts(&outbound_vectors, update_every); fflush(stdout); ebpf_socket_send_nv_data(&outbound_vectors); - wait_to_plot = 0; pthread_mutex_unlock(&nv_mutex); } @@ -2959,8 +2953,10 @@ static void ebpf_socket_allocate_global_vectors(int apps) bandwidth_vector = callocz((size_t)ebpf_nprocs, sizeof(ebpf_bandwidth_t)); socket_values = callocz((size_t)ebpf_nprocs, sizeof(netdata_socket_t)); - inbound_vectors.plot = callocz(network_viewer_opt.max_dim, sizeof(netdata_socket_plot_t)); - outbound_vectors.plot = callocz(network_viewer_opt.max_dim, sizeof(netdata_socket_plot_t)); + if (network_viewer_opt.enabled) { + inbound_vectors.plot = callocz(network_viewer_opt.max_dim, sizeof(netdata_socket_plot_t)); + outbound_vectors.plot = callocz(network_viewer_opt.max_dim, sizeof(netdata_socket_plot_t)); + } } /** @@ -3219,12 +3215,11 @@ static void get_ipv6_first_addr(union netdata_ip_t *out, union netdata_ip_t *in, * * @return It returns 1 if the IP is inside the range and 0 otherwise */ -static int is_ip_inside_range(union netdata_ip_t *rfirst, union netdata_ip_t *rlast, - union netdata_ip_t *cmpfirst, union netdata_ip_t *cmplast, int family) +static int ebpf_is_ip_inside_range(union netdata_ip_t *rfirst, union netdata_ip_t *rlast, + union netdata_ip_t *cmpfirst, union netdata_ip_t *cmplast, int family) { if (family == AF_INET) { - if (ntohl(rfirst->addr32[0]) <= ntohl(cmpfirst->addr32[0]) && - ntohl(rlast->addr32[0]) >= ntohl(cmplast->addr32[0])) + if ((rfirst->addr32[0] <= cmpfirst->addr32[0]) && (rlast->addr32[0] >= cmplast->addr32[0])) return 1; } else { if (memcmp(rfirst->addr8, cmpfirst->addr8, sizeof(union netdata_ip_t)) <= 0 && @@ -3241,16 +3236,22 @@ static int is_ip_inside_range(union netdata_ip_t *rfirst, union netdata_ip_t *rl * * @param out a pointer to the link list. * @param in the structure that will be linked. + * @param table the modified table. */ -void fill_ip_list(ebpf_network_viewer_ip_list_t **out, ebpf_network_viewer_ip_list_t *in, char *table) +void ebpf_fill_ip_list(ebpf_network_viewer_ip_list_t **out, ebpf_network_viewer_ip_list_t *in, char *table) { #ifndef NETDATA_INTERNAL_CHECKS UNUSED(table); #endif + if (in->ver == AF_INET) { // It is simpler to compare using host order + in->first.addr32[0] = ntohl(in->first.addr32[0]); + in->last.addr32[0] = ntohl(in->last.addr32[0]); + } if (likely(*out)) { ebpf_network_viewer_ip_list_t *move = *out, *store = *out; while (move) { - if (in->ver == move->ver && is_ip_inside_range(&move->first, &move->last, &in->first, &in->last, in->ver)) { + if (in->ver == move->ver && + ebpf_is_ip_inside_range(&move->first, &move->last, &in->first, &in->last, in->ver)) { info("The range/value (%s) is inside the range/value (%s) already inserted, it will be ignored.", in->value, move->value); freez(in->value); @@ -3267,14 +3268,12 @@ void fill_ip_list(ebpf_network_viewer_ip_list_t **out, ebpf_network_viewer_ip_li } #ifdef NETDATA_INTERNAL_CHECKS - char first[512], last[512]; + char first[256], last[512]; if (in->ver == AF_INET) { - if (inet_ntop(AF_INET, in->first.addr8, first, INET_ADDRSTRLEN) && - inet_ntop(AF_INET, in->last.addr8, last, INET_ADDRSTRLEN)) - info("Adding values %s - %s to %s IP list \"%s\" used on network viewer", - first, last, - (*out == network_viewer_opt.included_ips)?"included":"excluded", - table); + info("Adding values %s: (%u - %u) to %s IP list \"%s\" used on network viewer", + in->value, in->first.addr32[0], in->last.addr32[0], + (*out == network_viewer_opt.included_ips)?"included":"excluded", + table); } else { if (inet_ntop(AF_INET6, in->first.addr8, first, INET6_ADDRSTRLEN) && inet_ntop(AF_INET6, in->last.addr8, last, INET6_ADDRSTRLEN)) @@ -3294,7 +3293,7 @@ void fill_ip_list(ebpf_network_viewer_ip_list_t **out, ebpf_network_viewer_ip_li * @param out a pointer to store the link list * @param ip the value given as parameter */ -static void parse_ip_list(void **out, char *ip) +static void ebpf_parse_ip_list(void **out, char *ip) { ebpf_network_viewer_ip_list_t **list = (ebpf_network_viewer_ip_list_t **)out; @@ -3442,7 +3441,7 @@ static void parse_ip_list(void **out, char *ip) ebpf_network_viewer_ip_list_t *store; - storethisip: +storethisip: store = callocz(1, sizeof(ebpf_network_viewer_ip_list_t)); store->value = ipdup; store->hash = simple_hash(ipdup); @@ -3450,7 +3449,7 @@ static void parse_ip_list(void **out, char *ip) memcpy(store->first.addr8, first.addr8, sizeof(first.addr8)); memcpy(store->last.addr8, last.addr8, sizeof(last.addr8)); - fill_ip_list(list, store, "socket"); + ebpf_fill_ip_list(list, store, "socket"); return; cleanipdup: @@ -3464,7 +3463,7 @@ cleanipdup: * * @param ptr is a pointer with the text to parse. */ -static void parse_ips(char *ptr) +static void ebpf_parse_ips(char *ptr) { // No value if (unlikely(!ptr)) @@ -3491,8 +3490,9 @@ static void parse_ips(char *ptr) } if (isascii(*ptr)) { // Parse port - parse_ip_list((!neg)?(void **)&network_viewer_opt.included_ips:(void **)&network_viewer_opt.excluded_ips, - ptr); + ebpf_parse_ip_list((!neg)?(void **)&network_viewer_opt.included_ips: + (void **)&network_viewer_opt.excluded_ips, + ptr); } ptr = end; @@ -3762,7 +3762,7 @@ void parse_network_viewer_section(struct config *cfg) value = appconfig_get(cfg, EBPF_NETWORK_VIEWER_SECTION, "ips", "!127.0.0.1/8 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 fc00::/7 !::1/128"); - parse_ips(value); + ebpf_parse_ips(value); } /** @@ -3916,16 +3916,9 @@ void *ebpf_socket_thread(void *ptr) { netdata_thread_cleanup_push(ebpf_socket_exit, ptr); - memset(&inbound_vectors.tree, 0, sizeof(avl_tree_lock)); - memset(&outbound_vectors.tree, 0, sizeof(avl_tree_lock)); - avl_init_lock(&inbound_vectors.tree, compare_sockets); - avl_init_lock(&outbound_vectors.tree, compare_sockets); - ebpf_module_t *em = (ebpf_module_t *)ptr; em->maps = socket_maps; - parse_network_viewer_section(&socket_config); - parse_service_name_section(&socket_config); parse_table_size_options(&socket_config); if (pthread_mutex_init(&nv_mutex, NULL)) { @@ -3935,7 +3928,15 @@ void *ebpf_socket_thread(void *ptr) } ebpf_socket_allocate_global_vectors(em->apps_charts); - initialize_inbound_outbound(); + + if (network_viewer_opt.enabled) { + memset(&inbound_vectors.tree, 0, sizeof(avl_tree_lock)); + memset(&outbound_vectors.tree, 0, sizeof(avl_tree_lock)); + avl_init_lock(&inbound_vectors.tree, ebpf_compare_sockets); + avl_init_lock(&outbound_vectors.tree, ebpf_compare_sockets); + + initialize_inbound_outbound(); + } if (running_on_kernel < NETDATA_EBPF_KERNEL_5_0) em->mode = MODE_ENTRY; @@ -3966,7 +3967,7 @@ void *ebpf_socket_thread(void *ptr) pthread_mutex_unlock(&lock); - socket_collector((usec_t)(em->update_every * USEC_PER_SEC), em); + socket_collector(em); endsocket: ebpf_update_disabled_plugin_stats(em); diff --git a/collectors/ebpf.plugin/ebpf_socket.h b/collectors/ebpf.plugin/ebpf_socket.h index ca6b193f0..63b1e107b 100644 --- a/collectors/ebpf.plugin/ebpf_socket.h +++ b/collectors/ebpf.plugin/ebpf_socket.h @@ -244,6 +244,7 @@ typedef struct ebpf_network_viewer_hostname_list { #define NETDATA_NV_CAP_VALUE 50L typedef struct ebpf_network_viewer_options { + uint32_t enabled; uint32_t max_dim; // Store value read from 'maximum dimensions' uint32_t hostname_resolution_enabled; @@ -360,7 +361,7 @@ void clean_port_structure(ebpf_network_viewer_port_list_t **clean); extern ebpf_network_viewer_port_list_t *listen_ports; void update_listen_table(uint16_t value, uint16_t proto, netdata_passive_connection_t *values); void parse_network_viewer_section(struct config *cfg); -void fill_ip_list(ebpf_network_viewer_ip_list_t **out, ebpf_network_viewer_ip_list_t *in, char *table); +void ebpf_fill_ip_list(ebpf_network_viewer_ip_list_t **out, ebpf_network_viewer_ip_list_t *in, char *table); void parse_service_name_section(struct config *cfg); extern ebpf_socket_publish_apps_t **socket_bandwidth_curr; diff --git a/collectors/ebpf.plugin/ebpf_softirq.c b/collectors/ebpf.plugin/ebpf_softirq.c index 3b5d15921..49e9c3051 100644 --- a/collectors/ebpf.plugin/ebpf_softirq.c +++ b/collectors/ebpf.plugin/ebpf_softirq.c @@ -54,17 +54,6 @@ static softirq_val_t softirq_vals[] = { // tmp store for soft IRQ values we get from a per-CPU eBPF map. static softirq_ebpf_val_t *softirq_ebpf_vals = NULL; -static struct netdata_static_thread softirq_threads = { - .name = "SOFTIRQ KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - /** * Cachestat Free * @@ -75,40 +64,20 @@ static struct netdata_static_thread softirq_threads = { static void ebpf_softirq_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); - freez(softirq_threads.thread); - for (int i = 0; softirq_tracepoints[i].class != NULL; i++) { ebpf_disable_tracepoint(&softirq_tracepoints[i]); } freez(softirq_ebpf_vals); pthread_mutex_lock(&ebpf_exit_cleanup); - em->thread->enabled = NETDATA_MAIN_THREAD_EXITED; + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; pthread_mutex_unlock(&ebpf_exit_cleanup); } /** - * Exit - * - * Cancel thread. - * - * @param ptr thread data. - */ -static void softirq_exit(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*softirq_threads.thread); - ebpf_softirq_free(em); -} - -/** * Cleanup * * Clean up allocated memory. @@ -146,28 +115,6 @@ static void softirq_read_latency_map() } } -/** - * Read eBPF maps for soft IRQ. - */ -static void *softirq_reader(void *ptr) -{ - netdata_thread_cleanup_push(softirq_exit, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_SOFTIRQ_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - softirq_read_latency_map(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - static void softirq_create_charts(int update_every) { ebpf_create_chart( @@ -212,17 +159,6 @@ static void softirq_collector(ebpf_module_t *em) { softirq_ebpf_vals = callocz(ebpf_nprocs, sizeof(softirq_ebpf_val_t)); - // create reader thread. - softirq_threads.thread = mallocz(sizeof(netdata_thread_t)); - softirq_threads.start_routine = softirq_reader; - netdata_thread_create( - softirq_threads.thread, - softirq_threads.name, - NETDATA_THREAD_OPTION_DEFAULT, - softirq_reader, - em - ); - // create chart and static dims. pthread_mutex_lock(&lock); softirq_create_charts(em->update_every); @@ -233,13 +169,16 @@ static void softirq_collector(ebpf_module_t *em) // loop and read from published data until ebpf plugin is closed. heartbeat_t hb; heartbeat_init(&hb); - usec_t step = em->update_every * USEC_PER_SEC; + int update_every = em->update_every; + int counter = update_every - 1; //This will be cancelled by its parent while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; + softirq_read_latency_map(); pthread_mutex_lock(&lock); // write dims now for all hitherto discovered IRQs. diff --git a/collectors/ebpf.plugin/ebpf_softirq.h b/collectors/ebpf.plugin/ebpf_softirq.h index 7dcddbb49..eea2a1841 100644 --- a/collectors/ebpf.plugin/ebpf_softirq.h +++ b/collectors/ebpf.plugin/ebpf_softirq.h @@ -20,7 +20,6 @@ typedef struct softirq_ebpf_val { *****************************************************************/ #define NETDATA_EBPF_MODULE_NAME_SOFTIRQ "softirq" -#define NETDATA_SOFTIRQ_SLEEP_MS 650000ULL #define NETDATA_SOFTIRQ_CONFIG_FILE "softirq.conf" typedef struct sofirq_val { diff --git a/collectors/ebpf.plugin/ebpf_swap.c b/collectors/ebpf.plugin/ebpf_swap.c index 8199573a9..059efb63b 100644 --- a/collectors/ebpf.plugin/ebpf_swap.c +++ b/collectors/ebpf.plugin/ebpf_swap.c @@ -34,17 +34,6 @@ static ebpf_local_maps_t swap_maps[] = {{.name = "tbl_pid_swap", .internal_input .map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED}, {.name = NULL, .internal_input = 0, .user_input = 0}}; -struct netdata_static_thread swap_threads = { - .name = "SWAP KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - netdata_ebpf_targets_t swap_targets[] = { {.name = "swap_readpage", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = "swap_writepage", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = NULL, .mode = EBPF_LOAD_TRAMPOLINE}}; @@ -184,7 +173,7 @@ static void ebpf_swap_disable_release_task(struct swap_bpf *obj) * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_swap_load_and_attach(struct swap_bpf *obj, ebpf_module_t *em) { @@ -236,18 +225,13 @@ static inline int ebpf_swap_load_and_attach(struct swap_bpf *obj, ebpf_module_t static void ebpf_swap_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); ebpf_cleanup_publish_syscall(swap_publish_aggregated); freez(swap_vector); freez(swap_values); - freez(swap_threads.thread); #ifdef LIBBPF_MAJOR_VERSION if (bpf_obj) @@ -268,20 +252,6 @@ static void ebpf_swap_free(ebpf_module_t *em) static void ebpf_swap_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*swap_threads.thread); - ebpf_swap_free(em); -} - -/** - * Swap cleanup - * - * Clean up allocated memory. - * - * @param ptr thread data. - */ -static void ebpf_swap_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_swap_free(em); } @@ -412,7 +382,7 @@ static void swap_send_global() * * Read the table with number of calls to all functions */ -static void read_global_table() +static void ebpf_swap_read_global_table() { netdata_idx_t *stored = swap_values; netdata_idx_t *val = swap_hash_values; @@ -433,33 +403,6 @@ static void read_global_table() } /** - * Swap read hash - * - * This is the thread callback. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_swap_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_swap_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - usec_t step = NETDATA_SWAP_SLEEP_MS * em->update_every; - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Sum PIDs * * Sum values for all targets. @@ -714,23 +657,19 @@ void ebpf_swap_send_cgroup_data(int update_every) */ static void swap_collector(ebpf_module_t *em) { - swap_threads.thread = mallocz(sizeof(netdata_thread_t)); - swap_threads.start_routine = ebpf_swap_read_hash; - - netdata_thread_create(swap_threads.thread, swap_threads.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_swap_read_hash, em); - int cgroup = em->cgroup_charts; int update_every = em->update_every; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = update_every * USEC_PER_SEC; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; netdata_apps_integration_flags_t apps = em->apps_charts; + ebpf_swap_read_global_table(); pthread_mutex_lock(&collect_data_mutex); if (apps) read_apps_table(); diff --git a/collectors/ebpf.plugin/ebpf_sync.c b/collectors/ebpf.plugin/ebpf_sync.c index 840497533..7c81c1df3 100644 --- a/collectors/ebpf.plugin/ebpf_sync.c +++ b/collectors/ebpf.plugin/ebpf_sync.c @@ -10,17 +10,6 @@ static netdata_publish_syscall_t sync_counter_publish_aggregated[NETDATA_SYNC_ID static netdata_idx_t sync_hash_values[NETDATA_SYNC_IDX_END]; -struct netdata_static_thread sync_threads = { - .name = "SYNC KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - static ebpf_local_maps_t sync_maps[] = {{.name = "tbl_sync", .internal_input = NETDATA_SYNC_END, .user_input = 0, .type = NETDATA_EBPF_MAP_STATIC, .map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED}, @@ -77,7 +66,7 @@ static inline void ebpf_sync_disable_probe(struct sync_bpf *obj) } /** - * Disable tramppoline + * Disable trampoline * * Disable trampoline to use another method. * @@ -140,7 +129,7 @@ static void ebpf_sync_set_hash_tables(struct sync_bpf *obj, sync_syscalls_index_ * @param target the syscall that we are attaching a tracer. * @param idx the index for the main structure * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_sync_load_and_attach(struct sync_bpf *obj, ebpf_module_t *em, char *target, sync_syscalls_index_t idx) @@ -216,17 +205,12 @@ void ebpf_sync_cleanup_objects() static void ebpf_sync_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); #ifdef LIBBPF_MAJOR_VERSION ebpf_sync_cleanup_objects(); #endif - freez(sync_threads.thread); pthread_mutex_lock(&ebpf_exit_cleanup); em->thread->enabled = NETDATA_THREAD_EBPF_STOPPED; @@ -243,18 +227,6 @@ static void ebpf_sync_free(ebpf_module_t *em) static void ebpf_sync_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*sync_threads.thread); - ebpf_sync_free(em); -} - -/** - * Clean up the main thread. - * - * @param ptr thread data. - */ -static void ebpf_sync_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_sync_free(em); } @@ -350,7 +322,7 @@ static int ebpf_sync_initialize_syscall(ebpf_module_t *em) * * Read the table with number of calls for all functions */ -static void read_global_table() +static void ebpf_sync_read_global_table() { netdata_idx_t stored; uint32_t idx = NETDATA_SYNC_CALL; @@ -366,34 +338,6 @@ static void read_global_table() } /** - * Sync read hash - * - * This is the thread callback. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_sync_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_sync_cleanup, ptr); - ebpf_module_t *em = (ebpf_module_t *)ptr; - - heartbeat_t hb; - heartbeat_init(&hb); - usec_t step = NETDATA_EBPF_SYNC_SLEEP_MS * em->update_every; - - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Create Sync charts * * Create charts and dimensions according user input. @@ -452,20 +396,17 @@ static void sync_send_data() */ static void sync_collector(ebpf_module_t *em) { - sync_threads.thread = mallocz(sizeof(netdata_thread_t)); - sync_threads.start_routine = ebpf_sync_read_hash; - - netdata_thread_create(sync_threads.thread, sync_threads.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_sync_read_hash, em); - heartbeat_t hb; heartbeat_init(&hb); - usec_t step = em->update_every * USEC_PER_SEC; + int update_every = em->update_every; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; + ebpf_sync_read_global_table(); pthread_mutex_lock(&lock); sync_send_data(); diff --git a/collectors/ebpf.plugin/ebpf_vfs.c b/collectors/ebpf.plugin/ebpf_vfs.c index ad6de4a07..b3c0ba45d 100644 --- a/collectors/ebpf.plugin/ebpf_vfs.c +++ b/collectors/ebpf.plugin/ebpf_vfs.c @@ -34,17 +34,6 @@ struct config vfs_config = { .first_section = NULL, .index = { .avl_tree = { .root = NULL, .compar = appconfig_section_compare }, .rwlock = AVL_LOCK_INITIALIZER } }; -struct netdata_static_thread vfs_threads = { - .name = "VFS KERNEL", - .config_section = NULL, - .config_name = NULL, - .env_name = NULL, - .enabled = 1, - .thread = NULL, - .init_routine = NULL, - .start_routine = NULL -}; - netdata_ebpf_targets_t vfs_targets[] = { {.name = "vfs_write", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = "vfs_writev", .mode = EBPF_LOAD_TRAMPOLINE}, {.name = "vfs_read", .mode = EBPF_LOAD_TRAMPOLINE}, @@ -357,7 +346,7 @@ static void ebpf_vfs_disable_release_task(struct vfs_bpf *obj) * @param obj is the main structure for bpf objects. * @param em structure with configuration * - * @return it returns 0 on succes and -1 otherwise + * @return it returns 0 on success and -1 otherwise */ static inline int ebpf_vfs_load_and_attach(struct vfs_bpf *obj, ebpf_module_t *em) { @@ -409,16 +398,11 @@ static inline int ebpf_vfs_load_and_attach(struct vfs_bpf *obj, ebpf_module_t *e static void ebpf_vfs_free(ebpf_module_t *em) { pthread_mutex_lock(&ebpf_exit_cleanup); - if (em->thread->enabled == NETDATA_THREAD_EBPF_RUNNING) { - em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; - pthread_mutex_unlock(&ebpf_exit_cleanup); - return; - } + em->thread->enabled = NETDATA_THREAD_EBPF_STOPPING; pthread_mutex_unlock(&ebpf_exit_cleanup); freez(vfs_hash_values); freez(vfs_vector); - freez(vfs_threads.thread); #ifdef LIBBPF_MAJOR_VERSION if (bpf_obj) @@ -440,18 +424,6 @@ static void ebpf_vfs_free(ebpf_module_t *em) static void ebpf_vfs_exit(void *ptr) { ebpf_module_t *em = (ebpf_module_t *)ptr; - netdata_thread_cancel(*vfs_threads.thread); - ebpf_vfs_free(em); -} - -/** -* Clean up the main thread. -* -* @param ptr thread data. -**/ -static void ebpf_vfs_cleanup(void *ptr) -{ - ebpf_module_t *em = (ebpf_module_t *)ptr; ebpf_vfs_free(em); } @@ -518,7 +490,7 @@ static void ebpf_vfs_send_data(ebpf_module_t *em) /** * Read the hash table and store data to allocated vectors. */ -static void read_global_table() +static void ebpf_vfs_read_global_table() { uint64_t idx; netdata_idx_t res[NETDATA_VFS_COUNTER]; @@ -874,36 +846,6 @@ static void read_update_vfs_cgroup() } /** - * VFS read hash - * - * This is the thread callback. - * This thread is necessary, because we cannot freeze the whole plugin to read the data. - * - * @param ptr It is a NULL value for this thread. - * - * @return It always returns NULL. - */ -void *ebpf_vfs_read_hash(void *ptr) -{ - netdata_thread_cleanup_push(ebpf_vfs_cleanup, ptr); - heartbeat_t hb; - heartbeat_init(&hb); - - ebpf_module_t *em = (ebpf_module_t *)ptr; - - usec_t step = NETDATA_LATENCY_VFS_SLEEP_MS * em->update_every; - //This will be cancelled by its parent - while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - - read_global_table(); - } - - netdata_thread_cleanup_pop(1); - return NULL; -} - -/** * Sum PIDs * * Sum values for all targets. @@ -1525,22 +1467,19 @@ static void ebpf_vfs_send_cgroup_data(ebpf_module_t *em) */ static void vfs_collector(ebpf_module_t *em) { - vfs_threads.thread = mallocz(sizeof(netdata_thread_t)); - vfs_threads.start_routine = ebpf_vfs_read_hash; - - netdata_thread_create(vfs_threads.thread, vfs_threads.name, NETDATA_THREAD_OPTION_DEFAULT, - ebpf_vfs_read_hash, em); - int cgroups = em->cgroup_charts; heartbeat_t hb; heartbeat_init(&hb); - usec_t step = em->update_every * USEC_PER_SEC; + int update_every = em->update_every; + int counter = update_every - 1; while (!ebpf_exit_plugin) { - (void)heartbeat_next(&hb, step); - if (ebpf_exit_plugin) - break; + (void)heartbeat_next(&hb, USEC_PER_SEC); + if (ebpf_exit_plugin || ++counter != update_every) + continue; + counter = 0; netdata_apps_integration_flags_t apps = em->apps_charts; + ebpf_vfs_read_global_table(); pthread_mutex_lock(&collect_data_mutex); if (apps) ebpf_vfs_read_apps(); diff --git a/collectors/ebpf.plugin/ebpf_vfs.h b/collectors/ebpf.plugin/ebpf_vfs.h index 2e3c7cc29..d7fc2672f 100644 --- a/collectors/ebpf.plugin/ebpf_vfs.h +++ b/collectors/ebpf.plugin/ebpf_vfs.h @@ -8,8 +8,6 @@ #define NETDATA_DIRECTORY_VFS_CONFIG_FILE "vfs.conf" -#define NETDATA_LATENCY_VFS_SLEEP_MS 750000ULL - // Global chart name #define NETDATA_VFS_FILE_CLEAN_COUNT "vfs_deleted_objects" #define NETDATA_VFS_FILE_IO_COUNT "vfs_io" |