diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-27 11:08:07 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-27 11:08:07 +0000 |
commit | c69cb8cc094cc916adbc516b09e944cd3d137c01 (patch) | |
tree | f2878ec41fb6d0e3613906c6722fc02b934eeb80 /collectors/ebpf.plugin | |
parent | Initial commit. (diff) | |
download | netdata-upstream.tar.xz netdata-upstream.zip |
Adding upstream version 1.29.3.upstream/1.29.3upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'collectors/ebpf.plugin')
-rw-r--r-- | collectors/ebpf.plugin/Makefile.am | 25 | ||||
-rw-r--r-- | collectors/ebpf.plugin/README.md | 400 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf.c | 1977 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf.conf | 45 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf.h | 207 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf_apps.c | 1082 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf_apps.h | 430 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf_kernel_reject_list.txt | 1 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf_process.c | 1057 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf_process.h | 139 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf_socket.c | 1920 | ||||
-rw-r--r-- | collectors/ebpf.plugin/ebpf_socket.h | 275 | ||||
-rw-r--r-- | collectors/ebpf.plugin/reset_netdata_trace.sh.in | 9 |
13 files changed, 7567 insertions, 0 deletions
diff --git a/collectors/ebpf.plugin/Makefile.am b/collectors/ebpf.plugin/Makefile.am new file mode 100644 index 0000000..1327d47 --- /dev/null +++ b/collectors/ebpf.plugin/Makefile.am @@ -0,0 +1,25 @@ +# SPDX-License-Identifier: GPL-3.0-or-later + +AUTOMAKE_OPTIONS = subdir-objects +MAINTAINERCLEANFILES = $(srcdir)/Makefile.in + +CLEANFILES = \ + reset_netdata_trace.sh \ + $(NULL) + +include $(top_srcdir)/build/subst.inc +SUFFIXES = .in + +dist_plugins_SCRIPTS = \ + reset_netdata_trace.sh \ + $(NULL) + +dist_noinst_DATA = \ + reset_netdata_trace.sh.in \ + README.md \ + $(NULL) + +dist_libconfig_DATA = \ + ebpf.conf \ + ebpf_kernel_reject_list.txt \ + $(NULL) diff --git a/collectors/ebpf.plugin/README.md b/collectors/ebpf.plugin/README.md new file mode 100644 index 0000000..5ea3b49 --- /dev/null +++ b/collectors/ebpf.plugin/README.md @@ -0,0 +1,400 @@ +<!-- +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" +--> + +# eBPF monitoring with Netdata + +Netdata's extended Berkeley Packet Filter (eBPF) collector monitors kernel-level metrics for file descriptors, virtual +filesystem IO, and process management on Linux systems. You can use our eBPF collector to analyze how and when a process +accesses files, when it makes system calls, whether it leaks memory or creating zombie processes, and more. + +Netdata's eBPF monitoring toolkit uses two custom eBPF programs. The default, called `entry`, monitors calls to a +variety of kernel functions, such as `do_sys_open`, `__close_fd`, `vfs_read`, `vfs_write`, `_do_fork`, and more. The +`return` program also monitors the return of each kernel functions to deliver more granular metrics about how your +system and its applications interact with the Linux kernel. + +eBPF monitoring can help you troubleshoot and debug how applications interact with the Linux kernel. See our [guide on +troubleshooting apps with eBPF metrics](/docs/guides/troubleshoot/monitor-debug-applications-ebpf.md) for configuration +and troubleshooting tips. + +<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" /> + <figcaption>An example of VFS charts made possible by the eBPF collector plugin.</figcaption> +</figure> + +## Enable the collector on Linux + +**The eBPF collector is installed and enabled by default on most new installations of the Agent**. The eBPF collector +does not currently work with [static build installations](/packaging/installer/methods/kickstart-64.md), but improved +support is in active development. + +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. + +If your Agent is v1.22 or older, you may to enable the collector yourself. See the [configuration](#configuration) +section for details. + +## Charts + +The eBPF collector creates an **eBPF** menu in the Agent's dashboard along with three sub-menus: **File**, **VFS**, and +**Process**. All the charts in this section update every second. The collector stores the actual value inside of its +process, but charts only show the difference between the values collected in the previous and current seconds. + +### File + +This group has two charts demonstrating how software interacts with the Linux kernel to open and close file descriptors. + +#### File descriptor + +This chart contains two dimensions that show the number of calls to the functions `do_sys_open` and `__close_fd`. Most +software do not commonly call these functions directly, but they are behind the system calls `open(2)`, `openat(2)`, +and `close(2)`. + +#### File error + +This chart shows the number of times some software tried and failed to open or close a file descriptor. + +### VFS + +A [virtual file system](https://en.wikipedia.org/wiki/Virtual_file_system) (VFS) is a layer on top of regular +filesystems. The functions present inside this API are used for all filesystems, so it's possible the charts in this +group won't show _all_ the actions that occurred on your system. + +#### Deleted objects + +This chart monitors calls for `vfs_unlink`. This function is responsible for removing objects from the file system. + +#### IO + +This chart shows the number of calls to the functions `vfs_read` and `vfs_write`. + +#### IO bytes + +This chart also monitors `vfs_read` and `vfs_write`, but instead shows the total of bytes read and written with these +functions. + +The Agent displays the number of bytes written as negative because they are moving down to disk. + +#### IO errors + +The Agent counts and shows the number of instances where a running program experiences a read or write error. + +### Process + +For this group, the eBPF collector monitors process/thread creation and process end, and then displays any errors in the +following charts. + +#### Process thread + +Internally, the Linux kernel treats both processes and threads as `tasks`. To create a thread, the kernel offers a few +system calls: `fork(2)`, `vfork(2)` and `clone(2)`. In turn, each of these system calls use the function `_do_fork`. To +generate this chart, the eBPF collector monitors `_do_fork` to populate the `process` dimension, and monitors +`sys_clone` to identify threads. + +#### Exit + +Ending a task requires two steps. The first is a call to the internal function `do_exit`, which notifies the operating +system that the task is finishing its work. The second step is to release the kernel information with the internal +function `release_task`. The difference between the two dimensions can help you discover [zombie +processes](https://en.wikipedia.org/wiki/Zombie_process). + +#### Task error + +The functions responsible for ending tasks do not return values, so this chart contains information about failures on +process and thread creation. + +## Configuration + +Enable or disable the entire eBPF collector by editing `netdata.conf`. + +```bash +cd /etc/netdata/ # Replace with your Netdata configuration directory, if not /etc/netdata/ +./edit-config netdata.conf +``` + +To enable the collector, scroll down to the `[plugins]` section ensure the relevant line references `ebpf` (not +`ebpf_process`), is uncommented, and is set to `yes`. + +```conf +[plugins] + ebpf = yes +``` + +You can also configure the eBPF collector's behavior by editing `ebpf.conf`. + +```bash +cd /etc/netdata/ # Replace with your Netdata configuration directory, if not /etc/netdata/ +./edit-config ebpf.conf +``` + +### `[global]` + +The `[global]` section defines settings for the whole eBPF collector. + +#### ebpf load mode + +The collector has two different eBPF programs. These programs monitor the same functions inside the kernel, but they +monitor, process, and display different kinds of information. + +By default, this plugin uses the `entry` mode. Changing this mode can create significant overhead on your operating +system, but also offer valuable information if you are developing or debugging software. The `ebpf load mode` option +accepts the following values: + +- `entry`: This is the default mode. In this mode, the eBPF collector only monitors calls for the functions described + in the sections above, and does not show charts related to errors. +- `return`: In the `return` mode, the eBPF collector monitors the same kernel functions as `entry`, but also creates + new charts for the return of these functions, such as errors. Monitoring function returns can help in debugging + software, such as failing to close file descriptors or creating zombie processes. + +#### 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 +interact with the Linux kernel. + +When the integration is enabled, your dashboard will also show the following charts using low-level Linux metrics: + +- eBPF file + - Number of calls to open files. (`apps.file_open`) + - Number of files closed. (`apps.file_closed`) + - Number of calls to open files that returned errors. + - Number of calls to close files that returned errors. +- eBPF syscall + - Number of calls to delete files. (`apps.file_deleted`) + - Number of calls to `vfs_write`. (`apps.vfs_write_call`) + - Number of calls to `vfs_read`. (`apps.vfs_read_call`) + - Number of bytes written with `vfs_write`. (`apps.vfs_write_bytes`) + - Number of bytes read with `vfs_read`. (`apps.vfs_read_bytes`) + - Number of calls to write a file that returned errors. + - Number of calls to read a file that returned errors. +- eBPF process + - Number of process created with `do_fork`. (`apps.process_create`) + - Number of threads created with `do_fork` or `__x86_64_sys_clone`, depending on your system's kernel version. (`apps.thread_create`) + - Number of times that a process called `do_exit`. (`apps.task_close`) +- eBPF net + - Number of bytes sent. (`apps.bandwidth_sent`) + - Number of bytes received. (`apps.bandwidth_recv`) + +If you want to _disable_ the integration with `apps.plugin` along with the above charts, change the setting `apps` to +`no`. + +```conf +[global] + apps = yes +``` + +### `[ebpf programs]` + +The eBPF collector enables and runs the following eBPF programs by default: + +- `process`: This eBPF program creates charts that show information about process creation, VFS IO, and files removed. + When in `return` mode, it also creates charts showing errors when these operations are executed. +- `network viewer`: This eBPF program creates charts with information about `TCP` and `UDP` functions, including the + bandwidth consumed by each. + +### `[network connections]` + +You can configure the information shown on `outbound` and `inbound` charts with the settings in this section. + +```conf +[network connections] + maximum dimensions = 500 + resolve hostname ips = no + ports = 1-1024 !145 !domain + hostnames = !example.com + ips = !127.0.0.1/8 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 fc00::/7 +``` + +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 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) +and 145. + +The following options are available: + +- `ports`: Define the destination ports for Netdata to monitor. +- `hostnames`: The list of hostnames that can be resolved to an IP address. +- `ips`: The IP or range of IPs that you want to monitor. You can use IPv4 or IPv6 addresses, use dashes to define a + range of IPs, or use CIDR values. The default behavior is to only collect data for private IP addresses, but this + can be changed with the `ips` setting. + +By default, Netdata displays up to 500 dimensions on network connection charts. If there are more possible dimensions, +they will be bundled into the `other` dimension. You can increase the number of shown dimensions by changing the `maximum +dimensions` setting. + +The dimensions for the traffic charts are created using the destination IPs of the sockets by default. This can be +changed setting `resolve hostname ips = yes` and restarting Netdata, after this Netdata will create dimensions using +the `hostnames` every time that is possible to resolve IPs to their hostnames. + +### `[service name]` + +Netdata uses the list of services in `/etc/services` to plot network connection charts. If this file does not contain the +name for a particular service you use in your infrastructure, you will need to add it to the `[service name]` section. + +For example, Netdata's default port (`19999`) is not listed in `/etc/services`. To associate that port with the Netdata +service in network connection charts, and thus see the name of the service instead of its port, define it: + +```conf +[service name] + 19999 = Netdata +``` + +## Troubleshooting + +If the eBPF collector does not work, you can troubleshoot it by running the `ebpf.plugin` command and investigating its +output. + +```bash +cd /usr/libexec/netdata/plugins.d/ +sudo su -s /bin/bash netdata +./ebpf.plugin +``` + +You can also use `grep` to search the Agent's `error.log` for messages related to eBPF monitoring. + +```bash +grep -i ebpf /var/log/netdata/error.log +``` + +### Confirm kernel compatibility + +The eBPF collector only works on Linux systems and with specific Linux kernels. We support all kernels more recent than +`4.11.0`, and all kernels on CentOS 7.6 or later. + +You can run our helper script to determine whether your system can support eBPF monitoring. + +```bash +curl -sSL https://raw.githubusercontent.com/netdata/kernel-collector/master/tools/check-kernel-config.sh | sudo bash +``` + +If this script returns no output, your system is ready to compile and run the eBPF collector. + +If you see a warning about a missing kernel configuration (`KPROBES KPROBES_ON_FTRACE HAVE_KPROBES BPF BPF_SYSCALL +BPF_JIT`), you will need to recompile your kernel to support this configuration. The process of recompiling Linux +kernels varies based on your distribution and version. Read the documentation for your system's distribution to learn +more about the specific workflow for recompiling the kernel, ensuring that you set all the necessary + +- [Ubuntu](https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel) +- [Debian](https://kernel-team.pages.debian.net/kernel-handbook/ch-common-tasks.html#s-common-official) +- [Fedora](https://fedoraproject.org/wiki/Building_a_custom_kernel) +- [CentOS](https://wiki.centos.org/HowTos/Custom_Kernel) +- [Arch Linux](https://wiki.archlinux.org/index.php/Kernel/Traditional_compilation) +- [Slackware](https://docs.slackware.com/howtos:slackware_admin:kernelbuilding) + +### Mount `debugfs` and `tracefs` + +The eBPF collector also requires both the `tracefs` and `debugfs` filesystems. Try mounting the `tracefs` and `debugfs` +filesystems using the commands below: + +```bash +sudo mount -t debugfs nodev /sys/kernel/debug +sudo mount -t tracefs nodev /sys/kernel/tracing +``` + +If they are already mounted, you will see an error. You can also configure your system's `/etc/fstab` configuration to +mount these filesystems on startup. More information can be found in the [ftrace documentation](https://www.kernel.org/doc/Documentation/trace/ftrace.txt). + +## Performance + +Because eBPF monitoring is complex, we are evaluating the performance of this new collector in various real-world +conditions, across various system loads, and when monitoring complex applications. + +Our [initial testing](https://github.com/netdata/netdata/issues/8195) shows the performance of the eBPF collector is +nearly identical to our [apps.plugin collector](/collectors/apps.plugin/README.md), despite collecting and displaying +much more sophisticated metrics. You can now use the eBPF to gather deeper insights without affecting the performance of +your complex applications at any load. + +## SELinux + +When [SELinux](https://www.redhat.com/en/topics/linux/what-is-selinux) is enabled, it may prevent `ebpf.plugin` from +starting correctly. Check the Agent's `error.log` file for errors like the ones below: + +```bash +2020-06-14 15:32:08: ebpf.plugin ERROR : EBPF PROCESS : Cannot load program: /usr/libexec/netdata/plugins.d/pnetdata_ebpf_process.3.10.0.o (errno 13, Permission denied) +2020-06-14 15:32:19: netdata ERROR : PLUGINSD[ebpf] : read failed: end of file (errno 9, Bad file descriptor) +``` + +You can also check for errors related to `ebpf.plugin` inside `/var/log/audit/audit.log`: + +```bash +type=AVC msg=audit(1586260134.952:97): avc: denied { map_create } for pid=1387 comm="ebpf.pl" scontext=system_u:system_r:unconfined_service_t:s0 tcontext=system_u:system_r:unconfined_service_t:s0 tclass=bpf permissive=0 +type=SYSCALL msg=audit(1586260134.952:97): arch=c000003e syscall=321 success=no exit=-13 a0=0 a1=7ffe6b36f000 a2=70 a3=0 items=0 ppid=1135 pid=1387 auid=4294967295 uid=994 gid=990 euid=0 suid=0 fsuid=0 egid=990 sgid=990 fsgid=990 tty=(none) ses=4294967295 comm="ebpf_proc +ess.pl" exe="/usr/libexec/netdata/plugins.d/ebpf.plugin" subj=system_u:system_r:unconfined_service_t:s0 key=(null) +``` + +If you see similar errors, you will have to adjust SELinux's policies to enable the eBPF collector. + +### Creation of bpf policies + +To enable `ebpf.plugin` to run on a distribution with SELinux enabled, it will be necessary to take the following +actions. + +First, stop the Netdata Agent. + +```bash +# systemctl stop netdata +``` + +Next, create a policy with the `audit.log` file you examined earlier. + +```bash +# grep ebpf.plugin /var/log/audit/audit.log | audit2allow -M netdata_ebpf +``` + +This will create two new files: `netdata_ebpf.te` and `netdata_ebpf.mod`. + +Edit the `netdata_ebpf.te` file to change the options `class` and `allow`. You should have the following at the end of +the `netdata_ebpf.te` file. + +```conf +module netdata_ebpf 1.0; +require { + type unconfined_service_t; + class bpf { map_create map_read map_write prog_load prog_run }; +} +#============= unconfined_service_t ============== +allow unconfined_service_t self:bpf { map_create map_read map_write prog_load prog_run }; +``` + +Then compile your `netdata_ebpf.te` file with the following commands to create a binary that loads the new policies: + +```bash +# checkmodule -M -m -o netdata_ebpf.mod netdata_ebpf.te +# semodule_package -o netdata_ebpf.pp -m netdata_ebpf.mod +``` + +Finally, you can load the new policy and start the Netdata agent again: + +```bash +# semodule -i netdata_ebpf.pp +# systemctl start netdata +``` + +## Lockdown + +Beginning with [version 5.4](https://www.zdnet.com/article/linux-to-get-kernel-lockdown-feature/), the Linux kernel has +a feature called "lockdown," which may affect `ebpf.plugin` depending how the kernel was compiled. The following table +shows how the lockdown module impacts `ebpf.plugin` based on the selected options: + +| Enforcing kernel lockdown | Enable lockdown LSM early in init | Default lockdown mode | Can `ebpf.plugin` run with this? | +|:------------------------- |:--------------------------------- |:--------------------- |:-------------------------------- | +| YES | NO | NO | YES | +| YES | Yes | None | YES | +| YES | Yes | Integrity | YES | +| YES | Yes | Confidentiality | NO | + +If you or your distribution compiled the kernel with the last combination, your system cannot load shared libraries +required to run `ebpf.plugin`. + +## Cleaning `kprobe_events` +The eBPF collector adds entries to the file `/sys/kernel/debug/tracing/kprobe_events`, and cleans them on exit, unless +another process prevents it. If you need to clean the eBPF entries safely, you can manually run the script +`/usr/libexec/netdata/plugins.d/reset_netdata_trace.sh`. + +[![analytics](https://www.google-analytics.com/collect?v=1&aip=1&t=pageview&_s=1&ds=github&dr=https%3A%2F%2Fgithub.com%2Fnetdata%2Fnetdata&dl=https%3A%2F%2Fmy-netdata.io%2Fgithub%2Fcollectors%2Febpf.plugin%2FREADME&_u=MAC~&cid=5792dfd7-8dc4-476b-af31-da2fdb9f93d2&tid=UA-64295674-3)](<>) diff --git a/collectors/ebpf.plugin/ebpf.c b/collectors/ebpf.plugin/ebpf.c new file mode 100644 index 0000000..26bcfcf --- /dev/null +++ b/collectors/ebpf.plugin/ebpf.c @@ -0,0 +1,1977 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#include <sys/time.h> +#include <sys/resource.h> +#include <ifaddrs.h> + +#include "ebpf.h" +#include "ebpf_socket.h" + +/***************************************************************** + * + * FUNCTIONS USED BY NETDATA + * + *****************************************************************/ + +// callback required by eval() +int health_variable_lookup(const char *variable, uint32_t hash, struct rrdcalc *rc, calculated_number *result) +{ + UNUSED(variable); + UNUSED(hash); + UNUSED(rc); + UNUSED(result); + return 0; +}; + +void send_statistics(const char *action, const char *action_result, const char *action_data) +{ + UNUSED(action); + UNUSED(action_result); + UNUSED(action_data); +} + +// callbacks required by popen() +void signals_block(void){}; +void signals_unblock(void){}; +void signals_reset(void){}; + +// required by get_system_cpus() +char *netdata_configured_host_prefix = ""; + +// callback required by fatal() +void netdata_cleanup_and_exit(int ret) +{ + exit(ret); +} + +// ---------------------------------------------------------------------- +/***************************************************************** + * + * GLOBAL VARIABLES + * + *****************************************************************/ + +char *ebpf_plugin_dir = PLUGINS_DIR; +char *ebpf_user_config_dir = CONFIG_DIR; +char *ebpf_stock_config_dir = LIBCONFIG_DIR; +static char *ebpf_configured_log_dir = LOG_DIR; + +char *ebpf_algorithms[] = {"absolute", "incremental"}; +int update_every = 1; +static int thread_finished = 0; +int close_ebpf_plugin = 0; +struct config collector_config = { .first_section = NULL, + .last_section = NULL, + .mutex = NETDATA_MUTEX_INITIALIZER, + .index = { .avl_tree = { .root = NULL, .compar = appconfig_section_compare }, + .rwlock = AVL_LOCK_INITIALIZER } }; + +int running_on_kernel = 0; +char kernel_string[64]; +int ebpf_nprocs; +static int isrh; +uint32_t finalized_threads = 1; + +pthread_mutex_t lock; +pthread_mutex_t collect_data_mutex; +pthread_cond_t collect_data_cond_var; + +ebpf_module_t ebpf_modules[] = { + { .thread_name = "process", .config_name = "process", .enabled = 0, .start_routine = ebpf_process_thread, + .update_time = 1, .global_charts = 1, .apps_charts = 1, .mode = MODE_ENTRY, + .optional = 0 }, + { .thread_name = "socket", .config_name = "socket", .enabled = 0, .start_routine = ebpf_socket_thread, + .update_time = 1, .global_charts = 1, .apps_charts = 1, .mode = MODE_ENTRY, + .optional = 0 }, + { .thread_name = NULL, .enabled = 0, .start_routine = NULL, .update_time = 1, + .global_charts = 0, .apps_charts = 1, .mode = MODE_ENTRY, + .optional = 0 }, +}; + +// Link with apps.plugin +ebpf_process_stat_t *global_process_stat = NULL; + +//Network viewer +ebpf_network_viewer_options_t network_viewer_opt; + +/***************************************************************** + * + * FUNCTIONS USED TO CLEAN MEMORY AND OPERATE SYSTEM FILES + * + *****************************************************************/ + +/** + * Cleanup publish syscall + * + * @param nps list of structures to clean + */ +void ebpf_cleanup_publish_syscall(netdata_publish_syscall_t *nps) +{ + while (nps) { + freez(nps->algorithm); + nps = nps->next; + } +} + +/** + * Clean port Structure + * + * Clean the allocated list. + * + * @param clean the list that will be cleaned + */ +void clean_port_structure(ebpf_network_viewer_port_list_t **clean) +{ + ebpf_network_viewer_port_list_t *move = *clean; + while (move) { + ebpf_network_viewer_port_list_t *next = move->next; + freez(move->value); + freez(move); + + move = next; + } + *clean = NULL; +} + +/** + * Clean IP structure + * + * Clean the allocated list. + * + * @param clean the list that will be cleaned + */ +static void clean_ip_structure(ebpf_network_viewer_ip_list_t **clean) +{ + ebpf_network_viewer_ip_list_t *move = *clean; + while (move) { + ebpf_network_viewer_ip_list_t *next = move->next; + freez(move); + + move = next; + } + *clean = NULL; +} + +/** + * Clean Loaded Events + * + * This function cleans the events previous loaded on Linux. +void clean_loaded_events() +{ + int event_pid; + for (event_pid = 0; ebpf_modules[event_pid].probes; event_pid++) + clean_kprobe_events(NULL, (int)ebpf_modules[event_pid].thread_id, ebpf_modules[event_pid].probes); +} + */ + +/** + * Close the collector gracefully + * + * @param sig is the signal number used to close the collector + */ +static void ebpf_exit(int sig) +{ + close_ebpf_plugin = 1; + + // When both threads were not finished case I try to go in front this address, the collector will crash + if (!thread_finished) { + return; + } + + freez(global_process_stat); + + /* + int ret = fork(); + if (ret < 0) // error + error("Cannot fork(), so I won't be able to clean %skprobe_events", NETDATA_DEBUGFS); + else if (!ret) { // child + int i; + for (i = getdtablesize(); i >= 0; --i) + close(i); + + int fd = open("/dev/null", O_RDWR, 0); + if (fd != -1) { + dup2(fd, STDIN_FILENO); + dup2(fd, STDOUT_FILENO); + dup2(fd, STDERR_FILENO); + } + + if (fd > 2) + close(fd); + + int sid = setsid(); + if (sid >= 0) { + debug(D_EXIT, "Wait for father %d die", getpid()); + sleep_usec(200000); // Sleep 200 miliseconds to father dies. + clean_loaded_events(); + } else { + error("Cannot become session id leader, so I won't try to clean kprobe_events.\n"); + } + } else { // parent + exit(0); + } + */ + + exit(sig); +} + +/***************************************************************** + * + * FUNCTIONS TO CREATE CHARTS + * + *****************************************************************/ + +/** + * Get a value from a structure. + * + * @param basis it is the first address of the structure + * @param offset it is the offset of the data you want to access. + * @return + */ +collected_number get_value_from_structure(char *basis, size_t offset) +{ + collected_number *value = (collected_number *)(basis + offset); + + collected_number ret = (collected_number)llabs(*value); + // this reset is necessary to avoid keep a constant value while processing is not executing a task + *value = 0; + + return ret; +} + +/** + * Write begin command on standard output + * + * @param family the chart family name + * @param name the chart name + */ +void write_begin_chart(char *family, char *name) +{ + printf("BEGIN %s.%s\n", family, name); +} + +/** + * Write END command on stdout. + */ +inline void write_end_chart() +{ + printf("END\n"); +} + +/** + * Write set command on standard output + * + * @param dim the dimension name + * @param value the value for the dimension + */ +void write_chart_dimension(char *dim, long long value) +{ + int ret = printf("SET %s = %lld\n", dim, value); + UNUSED(ret); +} + +/** + * Call the necessary functions to create a chart. + * + * @param name the chart name + * @param family the chart family + * @param move the pointer with the values that will be published + * @param end the number of values that will be written on standard output + * + * @return It returns a variable tha maps the charts that did not have zero values. + */ +void write_count_chart(char *name, char *family, netdata_publish_syscall_t *move, uint32_t end) +{ + write_begin_chart(family, name); + + uint32_t i = 0; + while (move && i < end) { + write_chart_dimension(move->name, move->ncall); + + move = move->next; + i++; + } + + write_end_chart(); +} + +/** + * Call the necessary functions to create a chart. + * + * @param name the chart name + * @param family the chart family + * @param move the pointer with the values that will be published + * @param end the number of values that will be written on standard output + */ +void write_err_chart(char *name, char *family, netdata_publish_syscall_t *move, int end) +{ + write_begin_chart(family, name); + + int i = 0; + while (move && i < end) { + write_chart_dimension(move->name, move->nerr); + + move = move->next; + i++; + } + + write_end_chart(); +} + +/** + * Call the necessary functions to create a chart. + * + * @param chart the chart name + * @param family the chart family + * @param dwrite the dimension name + * @param vwrite the value for previous dimension + * @param dread the dimension name + * @param vread the value for previous dimension + * + * @return It returns a variable tha maps the charts that did not have zero values. + */ +void write_io_chart(char *chart, char *family, char *dwrite, long long vwrite, char *dread, long long vread) +{ + write_begin_chart(family, chart); + + write_chart_dimension(dwrite, vwrite); + write_chart_dimension(dread, vread); + + write_end_chart(); +} + +/** + * Write chart cmd on standard output + * + * @param type the chart type + * @param id the chart id + * @param title the chart title + * @param units the units label + * @param family the group name used to attach the chart on dashaboard + * @param charttype the chart type + * @param order the chart order + */ +void ebpf_write_chart_cmd(char *type, char *id, char *title, char *units, char *family, char *charttype, int order) +{ + printf("CHART %s.%s '' '%s' '%s' '%s' '' %s %d %d\n", + type, + id, + title, + units, + family, + charttype, + order, + update_every); +} + +/** + * Write the dimension command on standard output + * + * @param name the dimension name + * @param id the dimension id + * @param algo the dimension algorithm + */ +void ebpf_write_global_dimension(char *name, char *id, char *algorithm) +{ + printf("DIMENSION %s %s %s 1 1\n", name, id, algorithm); +} + +/** + * Call ebpf_write_global_dimension to create the dimensions for a specific chart + * + * @param ptr a pointer to a structure of the type netdata_publish_syscall_t + * @param end the number of dimensions for the structure ptr + */ +void ebpf_create_global_dimension(void *ptr, int end) +{ + netdata_publish_syscall_t *move = ptr; + + int i = 0; + while (move && i < end) { + ebpf_write_global_dimension(move->name, move->dimension, move->algorithm); + + move = move->next; + i++; + } +} + +/** + * Call write_chart_cmd to create the charts + * + * @param type the chart type + * @param id the chart id + * @param units the axis label + * @param family the group name used to attach the chart on dashaboard + * @param order the order number of the specified chart + * @param ncd a pointer to a function called to create dimensions + * @param move a pointer for a structure that has the dimensions + * @param end number of dimensions for the chart created + */ +void ebpf_create_chart(char *type, + char *id, + char *title, + char *units, + char *family, + int order, + void (*ncd)(void *, int), + void *move, + int end) +{ + ebpf_write_chart_cmd(type, id, title, units, family, "line", order); + + ncd(move, end); +} + +/** + * Create charts on apps submenu + * + * @param id the chart id + * @param title the value displayed on vertical axis. + * @param units the value displayed on vertical axis. + * @param family Submenu that the chart will be attached on dashboard. + * @param order the chart order + * @param algorithm the algorithm used by dimension + * @param root structure used to create the dimensions. + */ +void ebpf_create_charts_on_apps(char *id, char *title, char *units, char *family, int order, + char *algorithm, struct target *root) +{ + struct target *w; + ebpf_write_chart_cmd(NETDATA_APPS_FAMILY, id, title, units, family, "stacked", order); + + for (w = root; w; w = w->next) { + if (unlikely(w->exposed)) + fprintf(stdout, "DIMENSION %s '' %s 1 1\n", w->name, algorithm); + } +} + +/***************************************************************** + * + * FUNCTIONS TO DEFINE OPTIONS + * + *****************************************************************/ + +/** + * Define labels used to generate charts + * + * @param is structure with information about number of calls made for a function. + * @param pio structure used to generate charts. + * @param dim a pointer for the dimensions name + * @param name a pointer for the tensor with the name of the functions. + * @param algorithm a vector with the algorithms used to make the charts + * @param end the number of elements in the previous 4 arguments. + */ +void ebpf_global_labels(netdata_syscall_stat_t *is, netdata_publish_syscall_t *pio, char **dim, + char **name, int *algorithm, int end) +{ + int i; + + netdata_syscall_stat_t *prev = NULL; + netdata_publish_syscall_t *publish_prev = NULL; + for (i = 0; i < end; i++) { + if (prev) { + prev->next = &is[i]; + } + prev = &is[i]; + + pio[i].dimension = dim[i]; + pio[i].name = name[i]; + pio[i].algorithm = strdupz(ebpf_algorithms[algorithm[i]]); + if (publish_prev) { + publish_prev->next = &pio[i]; + } + publish_prev = &pio[i]; + } +} + +/** + * Define thread mode for all ebpf program. + * + * @param lmode the mode that will be used for them. + */ +static inline void ebpf_set_thread_mode(netdata_run_mode_t lmode) +{ + int i; + for (i = 0; ebpf_modules[i].thread_name; i++) { + ebpf_modules[i].mode = lmode; + } +} + +/** + * Enable specific charts selected by user. + * + * @param em the structure that will be changed + * @param enable the status about the apps charts. + */ +static inline void ebpf_enable_specific_chart(struct ebpf_module *em, int enable) +{ + em->enabled = 1; + if (!enable) { + em->apps_charts = 1; + } + em->global_charts = 1; +} + +/** + * Enable all charts + * + * @param apps what is the current status of apps + */ +static inline void ebpf_enable_all_charts(int apps) +{ + int i; + for (i = 0; ebpf_modules[i].thread_name; i++) { + ebpf_enable_specific_chart(&ebpf_modules[i], apps); + } +} + +/** + * Enable the specified chart group + * + * @param idx the index of ebpf_modules that I am enabling + * @param disable_apps should I keep apps charts? + */ +static inline void ebpf_enable_chart(int idx, int disable_apps) +{ + int i; + for (i = 0; ebpf_modules[i].thread_name; i++) { + if (i == idx) { + ebpf_enable_specific_chart(&ebpf_modules[i], disable_apps); + break; + } + } +} + +/** + * Disable APPs + * + * Disable charts for apps loading only global charts. + */ +static inline void ebpf_disable_apps() +{ + int i; + for (i = 0; ebpf_modules[i].thread_name; i++) { + ebpf_modules[i].apps_charts = 0; + } +} + +/** + * Print help on standard error for user knows how to use the collector. + */ +void ebpf_print_help() +{ + const time_t t = time(NULL); + struct tm ct; + struct tm *test = localtime_r(&t, &ct); + int year; + if (test) + year = ct.tm_year; + else + year = 0; + + fprintf(stderr, + "\n" + " Netdata ebpf.plugin %s\n" + " Copyright (C) 2016-%d Costa Tsaousis <costa@tsaousis.gr>\n" + " Released under GNU General Public License v3 or later.\n" + " All rights reserved.\n" + "\n" + " This program is a data collector plugin for netdata.\n" + "\n" + " Available command line options:\n" + "\n" + " SECONDS set the data collection frequency.\n" + "\n" + " --help or -h show this help.\n" + "\n" + " --version or -v show software version.\n" + "\n" + " --global or -g disable charts per application.\n" + "\n" + " --all or -a Enable all chart groups (global and apps), unless -g is also given.\n" + "\n" + " --net or -n Enable network viewer charts.\n" + "\n" + " --process or -p Enable charts related to process run time.\n" + "\n" + " --return or -r Run the collector in return mode.\n" + "\n", + VERSION, + (year >= 116) ? year + 1900 : 2020); +} + +/***************************************************************** + * + * AUXILIAR FUNCTIONS USED DURING INITIALIZATION + * + *****************************************************************/ + +/** + * Is ip inside the range + * + * Check if the ip is inside a IP range + * + * @param rfirst the first ip address of the range + * @param rlast the last ip address of the range + * @param cmpfirst the first ip to compare + * @param cmplast the last ip to compare + * @param family the IP family + * + * @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) +{ + if (family == AF_INET) { + if (ntohl(rfirst->addr32[0]) <= ntohl(cmpfirst->addr32[0]) && + ntohl(rlast->addr32[0]) >= ntohl(cmplast->addr32[0])) + return 1; + } else { + if (memcmp(rfirst->addr8, cmpfirst->addr8, sizeof(union netdata_ip_t)) <= 0 && + memcmp(rlast->addr8, cmplast->addr8, sizeof(union netdata_ip_t)) >= 0) { + return 1; + } + + } + return 0; +} + + +/** + * Fill IP list + * + * @param out a pointer to the link list. + * @param in the structure that will be linked. + */ +static inline void 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 (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)) { + info("The range/value (%s) is inside the range/value (%s) already inserted, it will be ignored.", + in->value, move->value); + freez(in->value); + freez(in); + return; + } + store = move; + move = move->next; + } + + store->next = in; + } else { + *out = in; + } + +#ifdef NETDATA_INTERNAL_CHECKS + char first[512], 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); + } else { + if (inet_ntop(AF_INET6, in->first.addr8, first, INET6_ADDRSTRLEN) && + inet_ntop(AF_INET6, in->last.addr8, last, INET6_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); + } +#endif +} + +/** + * Read Local Ports + * + * Parse /proc/net/{tcp,udp} and get the ports Linux is listening. + * + * @param filename the proc file to parse. + * @param proto is the magic number associated to the protocol file we are reading. + */ +static void read_local_ports(char *filename, uint8_t proto) +{ + procfile *ff = procfile_open(filename, " \t:", PROCFILE_FLAG_DEFAULT); + if (!ff) + return; + + ff = procfile_readall(ff); + if (!ff) + return; + + size_t lines = procfile_lines(ff), l; + for(l = 0; l < lines ;l++) { + size_t words = procfile_linewords(ff, l); + // This is header or end of file + if (unlikely(words < 14)) + continue; + + // https://elixir.bootlin.com/linux/v5.7.8/source/include/net/tcp_states.h + // 0A = TCP_LISTEN + if (strcmp("0A", procfile_lineword(ff, l, 5))) + continue; + + // Read local port + uint16_t port = (uint16_t)strtol(procfile_lineword(ff, l, 2), NULL, 16); + update_listen_table(htons(port), proto); + } + + procfile_close(ff); +} + +/** + * Read Local addresseses + * + * Read the local address from the interfaces. + */ +static void read_local_addresses() +{ + struct ifaddrs *ifaddr, *ifa; + if (getifaddrs(&ifaddr) == -1) { + error("Cannot get the local IP addresses, it is no possible to do separation between inbound and outbound connections"); + return; + } + + char *notext = { "No text representation" }; + for (ifa = ifaddr; ifa != NULL; ifa = ifa->ifa_next) { + if (ifa->ifa_addr == NULL) + continue; + + if ((ifa->ifa_addr->sa_family != AF_INET) && (ifa->ifa_addr->sa_family != AF_INET6)) + continue; + + ebpf_network_viewer_ip_list_t *w = callocz(1, sizeof(ebpf_network_viewer_ip_list_t)); + + int family = ifa->ifa_addr->sa_family; + w->ver = (uint8_t) family; + char text[INET6_ADDRSTRLEN]; + if (family == AF_INET) { + struct sockaddr_in *in = (struct sockaddr_in*) ifa->ifa_addr; + + w->first.addr32[0] = in->sin_addr.s_addr; + w->last.addr32[0] = in->sin_addr.s_addr; + + if (inet_ntop(AF_INET, w->first.addr8, text, INET_ADDRSTRLEN)) { + w->value = strdupz(text); + w->hash = simple_hash(text); + } else { + w->value = strdupz(notext); + w->hash = simple_hash(notext); + } + } else { + struct sockaddr_in6 *in6 = (struct sockaddr_in6*) ifa->ifa_addr; + + memcpy(w->first.addr8, (void *)&in6->sin6_addr, sizeof(struct in6_addr)); + memcpy(w->last.addr8, (void *)&in6->sin6_addr, sizeof(struct in6_addr)); + + if (inet_ntop(AF_INET6, w->first.addr8, text, INET_ADDRSTRLEN)) { + w->value = strdupz(text); + w->hash = simple_hash(text); + } else { + w->value = strdupz(notext); + w->hash = simple_hash(notext); + } + } + + fill_ip_list((family == AF_INET)?&network_viewer_opt.ipv4_local_ip:&network_viewer_opt.ipv6_local_ip, + w, + "selector"); + } + + freeifaddrs(ifaddr); +} + +/** + * Start Ptherad Variable + * + * This function starts all pthread variables. + * + * @return It returns 0 on success and -1. + */ +int ebpf_start_pthread_variables() +{ + pthread_mutex_init(&lock, NULL); + pthread_mutex_init(&collect_data_mutex, NULL); + + if (pthread_cond_init(&collect_data_cond_var, NULL)) { + thread_finished++; + error("Cannot start conditional variable to control Apps charts."); + return -1; + } + + return 0; +} + +/** + * Allocate the vectors used for all threads. + */ +static void ebpf_allocate_common_vectors() +{ + all_pids = callocz((size_t)pid_max, sizeof(struct pid_stat *)); + global_process_stat = callocz((size_t)ebpf_nprocs, sizeof(ebpf_process_stat_t)); +} + +/** + * Fill the ebpf_data structure with default values + * + * @param ef the pointer to set default values + */ +void fill_ebpf_data(ebpf_data_t *ef) +{ + memset(ef, 0, sizeof(ebpf_data_t)); + ef->kernel_string = kernel_string; + ef->running_on_kernel = running_on_kernel; + ef->map_fd = callocz(EBPF_MAX_MAPS, sizeof(int)); + ef->isrh = isrh; +} + +/** + * Define how to load the ebpf programs + * + * @param ptr the option given by users + */ +static inline void how_to_load(char *ptr) +{ + if (!strcasecmp(ptr, "return")) + ebpf_set_thread_mode(MODE_RETURN); + else if (!strcasecmp(ptr, "entry")) + ebpf_set_thread_mode(MODE_ENTRY); + else + error("the option %s for \"ebpf load mode\" is not a valid option.", ptr); +} + +/** + * Fill Port list + * + * @param out a pointer to the link list. + * @param in the structure that will be linked. + */ +static inline void fill_port_list(ebpf_network_viewer_port_list_t **out, ebpf_network_viewer_port_list_t *in) +{ + if (likely(*out)) { + ebpf_network_viewer_port_list_t *move = *out, *store = *out; + uint16_t first = ntohs(in->first); + uint16_t last = ntohs(in->last); + while (move) { + uint16_t cmp_first = ntohs(move->first); + uint16_t cmp_last = ntohs(move->last); + if (cmp_first <= first && first <= cmp_last && + cmp_first <= last && last <= cmp_last ) { + info("The range/value (%u, %u) is inside the range/value (%u, %u) already inserted, it will be ignored.", + first, last, cmp_first, cmp_last); + freez(in->value); + freez(in); + return; + } else if (first <= cmp_first && cmp_first <= last && + first <= cmp_last && cmp_last <= last) { + info("The range (%u, %u) is bigger than previous range (%u, %u) already inserted, the previous will be ignored.", + first, last, cmp_first, cmp_last); + freez(move->value); + move->value = in->value; + move->first = in->first; + move->last = in->last; + freez(in); + return; + } + + store = move; + move = move->next; + } + + store->next = in; + } else { + *out = in; + } + +#ifdef NETDATA_INTERNAL_CHECKS + info("Adding values %s( %u, %u) to %s port list used on network viewer", + in->value, ntohs(in->first), ntohs(in->last), + (*out == network_viewer_opt.included_port)?"included":"excluded"); +#endif +} + +/** + * Fill port list + * + * Fill an allocated port list with the range given + * + * @param out a pointer to store the link list + * @param range the informed range for the user. + */ +static void parse_port_list(void **out, char *range) +{ + int first, last; + ebpf_network_viewer_port_list_t **list = (ebpf_network_viewer_port_list_t **)out; + + char *copied = strdupz(range); + if (*range == '*' && *(range+1) == '\0') { + first = 1; + last = 65535; + + clean_port_structure(list); + goto fillenvpl; + } + + char *end = range; + //Move while I cannot find a separator + while (*end && *end != ':' && *end != '-') end++; + + //It has a range + if (likely(*end)) { + *end++ = '\0'; + if (*end == '!') { + info("The exclusion cannot be in the second part of the range, the range %s will be ignored.", copied); + freez(copied); + return; + } + last = str2i((const char *)end); + } else { + last = 0; + } + + first = str2i((const char *)range); + if (first < NETDATA_MINIMUM_PORT_VALUE || first > NETDATA_MAXIMUM_PORT_VALUE) { + info("The first port %d of the range \"%s\" is invalid and it will be ignored!", first, copied); + freez(copied); + return; + } + + if (!last) + last = first; + + if (last < NETDATA_MINIMUM_PORT_VALUE || last > NETDATA_MAXIMUM_PORT_VALUE) { + info("The second port %d of the range \"%s\" is invalid and the whole range will be ignored!", last, copied); + freez(copied); + return; + } + + if (first > last) { + info("The specified order %s is wrong, the smallest value is always the first, it will be ignored!", copied); + freez(copied); + return; + } + + ebpf_network_viewer_port_list_t *w; +fillenvpl: + w = callocz(1, sizeof(ebpf_network_viewer_port_list_t)); + w->value = copied; + w->hash = simple_hash(copied); + w->first = (uint16_t)htons((uint16_t)first); + w->last = (uint16_t)htons((uint16_t)last); + w->cmp_first = (uint16_t)first; + w->cmp_last = (uint16_t)last; + + fill_port_list(list, w); +} + +/** + * Parse Service List + * + * @param out a pointer to store the link list + * @param service the service used to create the structure that will be linked. + */ +static void parse_service_list(void **out, char *service) +{ + ebpf_network_viewer_port_list_t **list = (ebpf_network_viewer_port_list_t **)out; + struct servent *serv = getservbyname((const char *)service, "tcp"); + if (!serv) + serv = getservbyname((const char *)service, "udp"); + + if (!serv) { + info("Cannot resolv the service '%s' with protocols TCP and UDP, it will be ignored", service); + return; + } + + ebpf_network_viewer_port_list_t *w = callocz(1, sizeof(ebpf_network_viewer_port_list_t)); + w->value = strdupz(service); + w->hash = simple_hash(service); + + w->first = w->last = (uint16_t)serv->s_port; + + fill_port_list(list, w); +} + +/** + * Netmask + * + * Copied from iprange (https://github.com/firehol/iprange/blob/master/iprange.h) + * + * @param prefix create the netmask based in the CIDR value. + * + * @return + */ +static inline in_addr_t netmask(int prefix) { + + if (prefix == 0) + return (~((in_addr_t) - 1)); + else + return (in_addr_t)(~((1 << (32 - prefix)) - 1)); + +} + +/** + * Broadcast + * + * Copied from iprange (https://github.com/firehol/iprange/blob/master/iprange.h) + * + * @param addr is the ip address + * @param prefix is the CIDR value. + * + * @return It returns the last address of the range + */ +static inline in_addr_t broadcast(in_addr_t addr, int prefix) +{ + return (addr | ~netmask(prefix)); +} + +/** + * Network + * + * Copied from iprange (https://github.com/firehol/iprange/blob/master/iprange.h) + * + * @param addr is the ip address + * @param prefix is the CIDR value. + * + * @return It returns the first address of the range. + */ +static inline in_addr_t ipv4_network(in_addr_t addr, int prefix) +{ + return (addr & netmask(prefix)); +} + +/** + * IP to network long + * + * @param dst the vector to store the result + * @param ip the source ip given by our users. + * @param domain the ip domain (IPV4 or IPV6) + * @param source the original string + * + * @return it returns 0 on success and -1 otherwise. + */ +static inline int ip2nl(uint8_t *dst, char *ip, int domain, char *source) +{ + if (inet_pton(domain, ip, dst) <= 0) { + error("The address specified (%s) is invalid ", source); + return -1; + } + + return 0; +} + +/** + * Get IPV6 Last Address + * + * @param out the address to store the last address. + * @param in the address used to do the math. + * @param prefix number of bits used to calculate the address + */ +static void get_ipv6_last_addr(union netdata_ip_t *out, union netdata_ip_t *in, uint64_t prefix) +{ + uint64_t mask,tmp; + uint64_t ret[2]; + memcpy(ret, in->addr32, sizeof(union netdata_ip_t)); + + if (prefix == 128) { + memcpy(out->addr32, in->addr32, sizeof(union netdata_ip_t)); + return; + } else if (!prefix) { + ret[0] = ret[1] = 0xFFFFFFFFFFFFFFFF; + memcpy(out->addr32, ret, sizeof(union netdata_ip_t)); + return; + } else if (prefix <= 64) { + ret[1] = 0xFFFFFFFFFFFFFFFFULL; + + tmp = be64toh(ret[0]); + if (prefix > 0) { + mask = 0xFFFFFFFFFFFFFFFFULL << (64 - prefix); + tmp |= ~mask; + } + ret[0] = htobe64(tmp); + } else { + mask = 0xFFFFFFFFFFFFFFFFULL << (128 - prefix); + tmp = be64toh(ret[1]); + tmp |= ~mask; + ret[1] = htobe64(tmp); + } + + memcpy(out->addr32, ret, sizeof(union netdata_ip_t)); +} + +/** + * Calculate ipv6 first address + * + * @param out the address to store the first address. + * @param in the address used to do the math. + * @param prefix number of bits used to calculate the address + */ +static void get_ipv6_first_addr(union netdata_ip_t *out, union netdata_ip_t *in, uint64_t prefix) +{ + uint64_t mask,tmp; + uint64_t ret[2]; + + memcpy(ret, in->addr32, sizeof(union netdata_ip_t)); + + if (prefix == 128) { + memcpy(out->addr32, in->addr32, sizeof(union netdata_ip_t)); + return; + } else if (!prefix) { + ret[0] = ret[1] = 0; + memcpy(out->addr32, ret, sizeof(union netdata_ip_t)); + return; + } else if (prefix <= 64) { + ret[1] = 0ULL; + + tmp = be64toh(ret[0]); + if (prefix > 0) { + mask = 0xFFFFFFFFFFFFFFFFULL << (64 - prefix); + tmp &= mask; + } + ret[0] = htobe64(tmp); + } else { + mask = 0xFFFFFFFFFFFFFFFFULL << (128 - prefix); + tmp = be64toh(ret[1]); + tmp &= mask; + ret[1] = htobe64(tmp); + } + + memcpy(out->addr32, ret, sizeof(union netdata_ip_t)); +} + +/** + * Parse IP List + * + * Parse IP list and link it. + * + * @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) +{ + ebpf_network_viewer_ip_list_t **list = (ebpf_network_viewer_ip_list_t **)out; + + char *ipdup = strdupz(ip); + union netdata_ip_t first = { }; + union netdata_ip_t last = { }; + char *is_ipv6; + if (*ip == '*' && *(ip+1) == '\0') { + memset(first.addr8, 0, sizeof(first.addr8)); + memset(last.addr8, 0xFF, sizeof(last.addr8)); + + is_ipv6 = ip; + + clean_ip_structure(list); + goto storethisip; + } + + char *end = ip; + // Move while I cannot find a separator + while (*end && *end != '/' && *end != '-') end++; + + // We will use only the classic IPV6 for while, but we could consider the base 85 in a near future + // https://tools.ietf.org/html/rfc1924 + is_ipv6 = strchr(ip, ':'); + + int select; + if (*end && !is_ipv6) { // IPV4 range + select = (*end == '/') ? 0 : 1; + *end++ = '\0'; + if (*end == '!') { + info("The exclusion cannot be in the second part of the range %s, it will be ignored.", ipdup); + goto cleanipdup; + } + + if (!select) { // CIDR + select = ip2nl(first.addr8, ip, AF_INET, ipdup); + if (select) + goto cleanipdup; + + select = (int) str2i(end); + if (select < NETDATA_MINIMUM_IPV4_CIDR || select > NETDATA_MAXIMUM_IPV4_CIDR) { + info("The specified CIDR %s is not valid, the IP %s will be ignored.", end, ip); + goto cleanipdup; + } + + last.addr32[0] = htonl(broadcast(ntohl(first.addr32[0]), select)); + // This was added to remove + // https://app.codacy.com/manual/netdata/netdata/pullRequest?prid=5810941&bid=19021977 + UNUSED(last.addr32[0]); + + uint32_t ipv4_test = htonl(ipv4_network(ntohl(first.addr32[0]), select)); + if (first.addr32[0] != ipv4_test) { + first.addr32[0] = ipv4_test; + struct in_addr ipv4_convert; + ipv4_convert.s_addr = ipv4_test; + char ipv4_msg[INET_ADDRSTRLEN]; + if(inet_ntop(AF_INET, &ipv4_convert, ipv4_msg, INET_ADDRSTRLEN)) + info("The network value of CIDR %s was updated for %s .", ipdup, ipv4_msg); + } + } else { // Range + select = ip2nl(first.addr8, ip, AF_INET, ipdup); + if (select) + goto cleanipdup; + + select = ip2nl(last.addr8, end, AF_INET, ipdup); + if (select) + goto cleanipdup; + } + + if (htonl(first.addr32[0]) > htonl(last.addr32[0])) { + info("The specified range %s is invalid, the second address is smallest than the first, it will be ignored.", + ipdup); + goto cleanipdup; + } + } else if (is_ipv6) { // IPV6 + if (!*end) { // Unique + select = ip2nl(first.addr8, ip, AF_INET6, ipdup); + if (select) + goto cleanipdup; + + memcpy(last.addr8, first.addr8, sizeof(first.addr8)); + } else if (*end == '-') { + *end++ = 0x00; + if (*end == '!') { + info("The exclusion cannot be in the second part of the range %s, it will be ignored.", ipdup); + goto cleanipdup; + } + + select = ip2nl(first.addr8, ip, AF_INET6, ipdup); + if (select) + goto cleanipdup; + + select = ip2nl(last.addr8, end, AF_INET6, ipdup); + if (select) + goto cleanipdup; + } else { // CIDR + *end++ = 0x00; + if (*end == '!') { + info("The exclusion cannot be in the second part of the range %s, it will be ignored.", ipdup); + goto cleanipdup; + } + + select = str2i(end); + if (select < 0 || select > 128) { + info("The CIDR %s is not valid, the address %s will be ignored.", end, ip); + goto cleanipdup; + } + + uint64_t prefix = (uint64_t)select; + select = ip2nl(first.addr8, ip, AF_INET6, ipdup); + if (select) + goto cleanipdup; + + get_ipv6_last_addr(&last, &first, prefix); + + union netdata_ip_t ipv6_test; + get_ipv6_first_addr(&ipv6_test, &first, prefix); + + if (memcmp(first.addr8, ipv6_test.addr8, sizeof(union netdata_ip_t)) != 0) { + memcpy(first.addr8, ipv6_test.addr8, sizeof(union netdata_ip_t)); + + struct in6_addr ipv6_convert; + memcpy(ipv6_convert.s6_addr, ipv6_test.addr8, sizeof(union netdata_ip_t)); + + char ipv6_msg[INET6_ADDRSTRLEN]; + if(inet_ntop(AF_INET6, &ipv6_convert, ipv6_msg, INET6_ADDRSTRLEN)) + info("The network value of CIDR %s was updated for %s .", ipdup, ipv6_msg); + } + } + + if ((be64toh(*(uint64_t *)&first.addr32[2]) > be64toh(*(uint64_t *)&last.addr32[2]) && + !memcmp(first.addr32, last.addr32, 2*sizeof(uint32_t))) || + (be64toh(*(uint64_t *)&first.addr32) > be64toh(*(uint64_t *)&last.addr32)) ) { + info("The specified range %s is invalid, the second address is smallest than the first, it will be ignored.", + ipdup); + goto cleanipdup; + } + } else { // Unique ip + select = ip2nl(first.addr8, ip, AF_INET, ipdup); + if (select) + goto cleanipdup; + + memcpy(last.addr8, first.addr8, sizeof(first.addr8)); + } + + ebpf_network_viewer_ip_list_t *store; + +storethisip: + store = callocz(1, sizeof(ebpf_network_viewer_ip_list_t)); + store->value = ipdup; + store->hash = simple_hash(ipdup); + store->ver = (uint8_t)(!is_ipv6)?AF_INET:AF_INET6; + memcpy(store->first.addr8, first.addr8, sizeof(first.addr8)); + memcpy(store->last.addr8, last.addr8, sizeof(last.addr8)); + + fill_ip_list(list, store, "socket"); + return; + +cleanipdup: + freez(ipdup); +} + +/** + * Parse IP Range + * + * Parse the IP ranges given and create Network Viewer IP Structure + * + * @param ptr is a pointer with the text to parse. + */ +static void parse_ips(char *ptr) +{ + // No value + if (unlikely(!ptr)) + return; + + while (likely(ptr)) { + // Move forward until next valid character + while (isspace(*ptr)) ptr++; + + // No valid value found + if (unlikely(!*ptr)) + return; + + // Find space that ends the list + char *end = strchr(ptr, ' '); + if (end) { + *end++ = '\0'; + } + + int neg = 0; + if (*ptr == '!') { + neg++; + ptr++; + } + + if (isascii(*ptr)) { // Parse port + parse_ip_list((!neg)?(void **)&network_viewer_opt.included_ips:(void **)&network_viewer_opt.excluded_ips, + ptr); + } + + ptr = end; + } +} + + +/** + * Parse Port Range + * + * Parse the port ranges given and create Network Viewer Port Structure + * + * @param ptr is a pointer with the text to parse. + */ +static void parse_ports(char *ptr) +{ + // No value + if (unlikely(!ptr)) + return; + + while (likely(ptr)) { + // Move forward until next valid character + while (isspace(*ptr)) ptr++; + + // No valid value found + if (unlikely(!*ptr)) + return; + + // Find space that ends the list + char *end = strchr(ptr, ' '); + if (end) { + *end++ = '\0'; + } + + int neg = 0; + if (*ptr == '!') { + neg++; + ptr++; + } + + if (isdigit(*ptr)) { // Parse port + parse_port_list((!neg)?(void **)&network_viewer_opt.included_port:(void **)&network_viewer_opt.excluded_port, + ptr); + } else if (isalpha(*ptr)) { // Parse service + parse_service_list((!neg)?(void **)&network_viewer_opt.included_port:(void **)&network_viewer_opt.excluded_port, + ptr); + } else if (*ptr == '*') { // All + parse_port_list((!neg)?(void **)&network_viewer_opt.included_port:(void **)&network_viewer_opt.excluded_port, + ptr); + } + + ptr = end; + } +} + +/** + * Link hostname + * + * @param out is the output link list + * @param in the hostname to add to list. + */ +static void link_hostname(ebpf_network_viewer_hostname_list_t **out, ebpf_network_viewer_hostname_list_t *in) +{ + if (likely(*out)) { + ebpf_network_viewer_hostname_list_t *move = *out; + for (; move->next ; move = move->next ) { + if (move->hash == in->hash && !strcmp(move->value, in->value)) { + info("The hostname %s was already inserted, it will be ignored.", in->value); + freez(in->value); + simple_pattern_free(in->value_pattern); + freez(in); + return; + } + } + + move->next = in; + } else { + *out = in; + } +#ifdef NETDATA_INTERNAL_CHECKS + info("Adding value %s to %s hostname list used on network viewer", + in->value, + (*out == network_viewer_opt.included_hostnames)?"included":"excluded"); +#endif +} + +/** + * Link Hostnames + * + * Parse the list of hostnames to create the link list. + * This is not associated with the IP, because simple patterns like *example* cannot be resolved to IP. + * + * @param out is the output link list + * @param parse is a pointer with the text to parser. + */ +static void link_hostnames(char *parse) +{ + // No value + if (unlikely(!parse)) + return; + + while (likely(parse)) { + // Find the first valid value + while (isspace(*parse)) parse++; + + // No valid value found + if (unlikely(!*parse)) + return; + + // Find space that ends the list + char *end = strchr(parse, ' '); + if (end) { + *end++ = '\0'; + } + + int neg = 0; + if (*parse == '!') { + neg++; + parse++; + } + + ebpf_network_viewer_hostname_list_t *hostname = callocz(1 , sizeof(ebpf_network_viewer_hostname_list_t)); + hostname->value = strdupz(parse); + hostname->hash = simple_hash(parse); + hostname->value_pattern = simple_pattern_create(parse, NULL, SIMPLE_PATTERN_EXACT); + + link_hostname((!neg)?&network_viewer_opt.included_hostnames:&network_viewer_opt.excluded_hostnames, + hostname); + + parse = end; + } +} + +/** + * Read max dimension. + * + * Netdata plot two dimensions per connection, so it is necessary to adjust the values. + */ +static void read_max_dimension() +{ + int maxdim ; + maxdim = (int) appconfig_get_number(&collector_config, + EBPF_NETWORK_VIEWER_SECTION, + "maximum dimensions", + NETDATA_NV_CAP_VALUE); + if (maxdim < 0) { + error("'maximum dimensions = %d' must be a positive number, Netdata will change for default value %ld.", + maxdim, NETDATA_NV_CAP_VALUE); + maxdim = NETDATA_NV_CAP_VALUE; + } + + maxdim /= 2; + if (!maxdim) { + info("The number of dimensions is too small (%u), we are setting it to minimum 2", network_viewer_opt.max_dim); + network_viewer_opt.max_dim = 1; + } + + network_viewer_opt.max_dim = (uint32_t)maxdim; +} + +/** + * Parse network viewer section + */ +static void parse_network_viewer_section() +{ + read_max_dimension(); + + network_viewer_opt.hostname_resolution_enabled = appconfig_get_boolean(&collector_config, + EBPF_NETWORK_VIEWER_SECTION, + "resolve hostnames", + CONFIG_BOOLEAN_NO); + + network_viewer_opt.service_resolution_enabled = appconfig_get_boolean(&collector_config, + EBPF_NETWORK_VIEWER_SECTION, + "resolve service names", + CONFIG_BOOLEAN_NO); + + char *value = appconfig_get(&collector_config, EBPF_NETWORK_VIEWER_SECTION, + "ports", NULL); + parse_ports(value); + + if (network_viewer_opt.hostname_resolution_enabled) { + value = appconfig_get(&collector_config, EBPF_NETWORK_VIEWER_SECTION, "hostnames", NULL); + link_hostnames(value); + } else { + info("Name resolution is disabled, collector will not parser \"hostnames\" list."); + } + + value = appconfig_get(&collector_config, 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); +} + +/** + * Link dimension name + * + * Link user specified names inside a link list. + * + * @param port the port number associated to the dimension name. + * @param hash the calculated hash for the dimension name. + * @param name the dimension name. + */ +static void link_dimension_name(char *port, uint32_t hash, char *value) +{ + int test = str2i(port); + if (test < NETDATA_MINIMUM_PORT_VALUE || test > NETDATA_MAXIMUM_PORT_VALUE){ + error("The dimension given (%s = %s) has an invalid value and it will be ignored.", port, value); + return; + } + + ebpf_network_viewer_dim_name_t *w; + w = callocz(1, sizeof(ebpf_network_viewer_dim_name_t)); + + w->name = strdupz(value); + w->hash = hash; + + w->port = (uint16_t) htons(test); + + ebpf_network_viewer_dim_name_t *names = network_viewer_opt.names; + if (unlikely(!names)) { + network_viewer_opt.names = w; + } else { + for (; names->next; names = names->next) { + if (names->port == w->port) { + info("Dupplicated definition for a service, the name %s will be ignored. ", names->name); + freez(names->name); + names->name = w->name; + names->hash = w->hash; + freez(w); + return; + } + } + names->next = w; + } + +#ifdef NETDATA_INTERNAL_CHECKS + info("Adding values %s( %u) to dimension name list used on network viewer", w->name, htons(w->port)); +#endif +} + +/** + * Parse service Name section. + * + * This function gets the values that will be used to overwrite dimensions. + */ +static void parse_service_name_section() +{ + struct section *co = appconfig_get_section(&collector_config, EBPF_SERVICE_NAME_SECTION); + if (co) { + struct config_option *cv; + for (cv = co->values; cv ; cv = cv->next) { + link_dimension_name(cv->name, cv->hash, cv->value); + } + } + + // Always associated the default port to Netdata + ebpf_network_viewer_dim_name_t *names = network_viewer_opt.names; + if (names) { + uint16_t default_port = htons(19999); + while (names) { + if (names->port == default_port) + return; + + names = names->next; + } + } + + char *port_string = getenv("NETDATA_LISTEN_PORT"); + if (port_string) + link_dimension_name(port_string, simple_hash(port_string), "Netdata"); +} + +/** + * Read collector values + * + * @param disable_apps variable to store information related to apps. + */ +static void read_collector_values(int *disable_apps) +{ + // Read global section + char *value; + if (appconfig_exists(&collector_config, EBPF_GLOBAL_SECTION, "load")) // Backward compatibility + value = appconfig_get(&collector_config, EBPF_GLOBAL_SECTION, "load", "entry"); + else + value = appconfig_get(&collector_config, EBPF_GLOBAL_SECTION, "ebpf load mode", "entry"); + + how_to_load(value); + + // This is kept to keep compatibility + uint32_t enabled = appconfig_get_boolean(&collector_config, EBPF_GLOBAL_SECTION, "disable apps", + CONFIG_BOOLEAN_NO); + if (!enabled) { + // Apps is a positive sentence, so we need to invert the values to disable apps. + enabled = appconfig_get_boolean(&collector_config, EBPF_GLOBAL_SECTION, "apps", + CONFIG_BOOLEAN_YES); + enabled = (enabled == CONFIG_BOOLEAN_NO)?CONFIG_BOOLEAN_YES:CONFIG_BOOLEAN_NO; + } + *disable_apps = (int)enabled; + + // Read ebpf programs section + enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, + ebpf_modules[0].config_name, CONFIG_BOOLEAN_YES); + int started = 0; + if (enabled) { + ebpf_enable_chart(EBPF_MODULE_PROCESS_IDX, *disable_apps); + started++; + } + + // This is kept to keep compatibility + enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, "network viewer", + CONFIG_BOOLEAN_NO); + if (!enabled) + enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, ebpf_modules[1].config_name, + CONFIG_BOOLEAN_NO); + + if (enabled) { + ebpf_enable_chart(EBPF_MODULE_SOCKET_IDX, *disable_apps); + // Read network viewer section if network viewer is enabled + parse_network_viewer_section(); + parse_service_name_section(); + started++; + } + + // This is kept to keep compatibility + enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, "network connection monitoring", + CONFIG_BOOLEAN_NO); + if (!enabled) + enabled = appconfig_get_boolean(&collector_config, EBPF_PROGRAMS_SECTION, "network connections", + CONFIG_BOOLEAN_NO); + ebpf_modules[1].optional = enabled; + + if (!started){ + ebpf_enable_all_charts(*disable_apps); + // Read network viewer section + parse_network_viewer_section(); + parse_service_name_section(); + } +} + +/** + * Load collector config + * + * @param path the path where the file ebpf.conf is stored. + * @param disable_apps variable to store the information about apps plugin status. + * + * @return 0 on success and -1 otherwise. + */ +static int load_collector_config(char *path, int *disable_apps) +{ + char lpath[4096]; + + snprintf(lpath, 4095, "%s/%s", path, "ebpf.conf"); + + if (!appconfig_load(&collector_config, lpath, 0, NULL)) + return -1; + + read_collector_values(disable_apps); + + return 0; +} + +/** + * Set global variables reading environment variables + */ +void set_global_variables() +{ + // Get environment variables + ebpf_plugin_dir = getenv("NETDATA_PLUGINS_DIR"); + if (!ebpf_plugin_dir) + ebpf_plugin_dir = PLUGINS_DIR; + + ebpf_user_config_dir = getenv("NETDATA_USER_CONFIG_DIR"); + if (!ebpf_user_config_dir) + ebpf_user_config_dir = CONFIG_DIR; + + ebpf_stock_config_dir = getenv("NETDATA_STOCK_CONFIG_DIR"); + if (!ebpf_stock_config_dir) + ebpf_stock_config_dir = LIBCONFIG_DIR; + + ebpf_configured_log_dir = getenv("NETDATA_LOG_DIR"); + if (!ebpf_configured_log_dir) + ebpf_configured_log_dir = LOG_DIR; + + ebpf_nprocs = (int)sysconf(_SC_NPROCESSORS_ONLN); + if (ebpf_nprocs > NETDATA_MAX_PROCESSOR) { + ebpf_nprocs = NETDATA_MAX_PROCESSOR; + } + + isrh = get_redhat_release(); + pid_max = get_system_pid_max(); +} + +/** + * Parse arguments given from user. + * + * @param argc the number of arguments + * @param argv the pointer to the arguments + */ +static void parse_args(int argc, char **argv) +{ + int enabled = 0; + int disable_apps = 0; + int freq = 0; + int option_index = 0; + static struct option long_options[] = { + {"help", no_argument, 0, 'h' }, + {"version", no_argument, 0, 'v' }, + {"global", no_argument, 0, 'g' }, + {"all", no_argument, 0, 'a' }, + {"net", no_argument, 0, 'n' }, + {"process", no_argument, 0, 'p' }, + {"return", no_argument, 0, 'r' }, + {0, 0, 0, 0} + }; + + memset(&network_viewer_opt, 0, sizeof(network_viewer_opt)); + network_viewer_opt.max_dim = NETDATA_NV_CAP_VALUE; + + if (argc > 1) { + int n = (int)str2l(argv[1]); + if (n > 0) { + freq = n; + } + } + + while (1) { + int c = getopt_long(argc, argv, "hvganpr", long_options, &option_index); + if (c == -1) + break; + + switch (c) { + case 'h': { + ebpf_print_help(); + exit(0); + } + case 'v': { + printf("ebpf.plugin %s\n", VERSION); + exit(0); + } + case 'g': { + disable_apps = 1; + ebpf_disable_apps(); +#ifdef NETDATA_INTERNAL_CHECKS + info( + "EBPF running with global chart group, because it was started with the option \"--global\" or \"-g\"."); +#endif + break; + } + case 'a': { + ebpf_enable_all_charts(disable_apps); +#ifdef NETDATA_INTERNAL_CHECKS + info("EBPF running with all chart groups, because it was started with the option \"--all\" or \"-a\"."); +#endif + break; + } + case 'n': { + enabled = 1; + ebpf_enable_chart(EBPF_MODULE_SOCKET_IDX, disable_apps); +#ifdef NETDATA_INTERNAL_CHECKS + info("EBPF enabling \"NET\" charts, because it was started with the option \"--net\" or \"-n\"."); +#endif + break; + } + case 'p': { + enabled = 1; + ebpf_enable_chart(EBPF_MODULE_PROCESS_IDX, disable_apps); +#ifdef NETDATA_INTERNAL_CHECKS + info( + "EBPF enabling \"PROCESS\" charts, because it was started with the option \"--process\" or \"-p\"."); +#endif + break; + } + case 'r': { + ebpf_set_thread_mode(MODE_RETURN); +#ifdef NETDATA_INTERNAL_CHECKS + info("EBPF running in \"return\" mode, because it was started with the option \"--return\" or \"-r\"."); +#endif + break; + } + default: { + break; + } + } + } + + if (freq > 0) { + update_every = freq; + } + + if (load_collector_config(ebpf_user_config_dir, &disable_apps)) { + info( + "Does not have a configuration file inside `%s/ebpf.conf. It will try to load stock file.", + ebpf_user_config_dir); + if (load_collector_config(ebpf_stock_config_dir, &disable_apps)) { + info("Does not have a stock file. It is starting with default options."); + } else { + enabled = 1; + } + } else { + enabled = 1; + } + + if (!enabled) { + ebpf_enable_all_charts(disable_apps); +#ifdef NETDATA_INTERNAL_CHECKS + info("EBPF running with all charts, because neither \"-n\" or \"-p\" was given."); +#endif + } + + if (disable_apps) + return; + + // Load apps_groups.conf + if (ebpf_read_apps_groups_conf( + &apps_groups_default_target, &apps_groups_root_target, ebpf_user_config_dir, "groups")) { + info( + "Cannot read process groups configuration file '%s/apps_groups.conf'. Will try '%s/apps_groups.conf'", + ebpf_user_config_dir, ebpf_stock_config_dir); + if (ebpf_read_apps_groups_conf( + &apps_groups_default_target, &apps_groups_root_target, ebpf_stock_config_dir, "groups")) { + error( + "Cannot read process groups '%s/apps_groups.conf'. There are no internal defaults. Failing.", + ebpf_stock_config_dir); + thread_finished++; + ebpf_exit(1); + } + } else + info("Loaded config file '%s/apps_groups.conf'", ebpf_user_config_dir); +} + +/***************************************************************** + * + * COLLECTOR ENTRY POINT + * + *****************************************************************/ + +/** + * Entry point + * + * @param argc the number of arguments + * @param argv the pointer to the arguments + * + * @return it returns 0 on success and another integer otherwise + */ +int main(int argc, char **argv) +{ + set_global_variables(); + parse_args(argc, argv); + + running_on_kernel = get_kernel_version(kernel_string, 63); + if (!has_condition_to_run(running_on_kernel)) { + error("The current collector cannot run on this kernel."); + return 2; + } + + if (!am_i_running_as_root()) { + error( + "ebpf.plugin should either run as root (now running with uid %u, euid %u) or have special capabilities..", + (unsigned int)getuid(), (unsigned int)geteuid()); + return 3; + } + + // set name + program_name = "ebpf.plugin"; + + // disable syslog + error_log_syslog = 0; + + // set errors flood protection to 100 logs per hour + error_log_errors_per_period = 100; + error_log_throttle_period = 3600; + + struct rlimit r = { RLIM_INFINITY, RLIM_INFINITY }; + if (setrlimit(RLIMIT_MEMLOCK, &r)) { + error("Setrlimit(RLIMIT_MEMLOCK)"); + return 4; + } + + signal(SIGINT, ebpf_exit); + signal(SIGTERM, ebpf_exit); + signal(SIGPIPE, ebpf_exit); + + if (ebpf_start_pthread_variables()) { + thread_finished++; + error("Cannot start mutex to control overall charts."); + ebpf_exit(5); + } + + ebpf_allocate_common_vectors(); + + read_local_addresses(); + read_local_ports("/proc/net/tcp", IPPROTO_TCP); + read_local_ports("/proc/net/tcp6", IPPROTO_TCP); + read_local_ports("/proc/net/udp", IPPROTO_UDP); + read_local_ports("/proc/net/udp6", IPPROTO_UDP); + + struct netdata_static_thread ebpf_threads[] = { + {"EBPF PROCESS", NULL, NULL, 1, NULL, NULL, ebpf_modules[0].start_routine}, + {"EBPF SOCKET" , NULL, NULL, 1, NULL, NULL, ebpf_modules[1].start_routine}, + {NULL , NULL, NULL, 0, NULL, NULL, NULL} + }; + + //clean_loaded_events(); + + int i; + for (i = 0; ebpf_threads[i].name != NULL; i++) { + struct netdata_static_thread *st = &ebpf_threads[i]; + st->thread = mallocz(sizeof(netdata_thread_t)); + + ebpf_module_t *em = &ebpf_modules[i]; + em->thread_id = i; + netdata_thread_create(st->thread, st->name, NETDATA_THREAD_OPTION_JOINABLE, st->start_routine, em); + } + + for (i = 0; ebpf_threads[i].name != NULL; i++) { + struct netdata_static_thread *st = &ebpf_threads[i]; + netdata_thread_join(*st->thread, NULL); + } + + thread_finished++; + ebpf_exit(0); + + return 0; +} diff --git a/collectors/ebpf.plugin/ebpf.conf b/collectors/ebpf.plugin/ebpf.conf new file mode 100644 index 0000000..3a5b773 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf.conf @@ -0,0 +1,45 @@ +# +# Global options +# +# The `ebpf load mode` option accepts the following values : +# `entry` : The eBPF collector only monitors calls for the functions, and does not show charts related to errors. +# `return : In the `return` mode, the eBPF collector monitors the same kernel functions as `entry`, but also creates +# new charts for the return of these functions, such as errors. +# +# The eBPF collector also creates charts for each running application through an integration with the `apps plugin`. +# If you want to disable the integration with `apps.plugin` along with the above charts, change the setting `apps` to +# 'no'. +# +[global] + ebpf load mode = entry + apps = yes + +# +# eBPF Programs +# +# The eBPF collector enables and runs the following eBPF programs by default: +# +# `process` : This eBPF program creates charts that show information about process creation, VFS IO, and +# files removed. +# `socket` : This eBPF program creates charts with information about `TCP` and `UDP` functions, including the +# bandwidth consumed by each. +[ebpf programs] + process = yes + socket = yes + network connections = no + +# +# Network Connection +# +# This is a feature with status WIP(Work in Progress) +# +[network connections] + maximum dimensions = 50 + resolve hostnames = no + resolve service names = no + ports = * + 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 + hostnames = * + +[service name] + 19999 = Netdata diff --git a/collectors/ebpf.plugin/ebpf.h b/collectors/ebpf.plugin/ebpf.h new file mode 100644 index 0000000..35013c2 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf.h @@ -0,0 +1,207 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#ifndef NETDATA_COLLECTOR_EBPF_H +#define NETDATA_COLLECTOR_EBPF_H 1 + +#ifndef __FreeBSD__ +#include <linux/perf_event.h> +#endif +#include <stdint.h> +#include <errno.h> +#include <signal.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <unistd.h> +#include <dlfcn.h> + +#include <fcntl.h> +#include <ctype.h> +#include <dirent.h> + +// From libnetdata.h +#include "libnetdata/threads/threads.h" +#include "libnetdata/locks/locks.h" +#include "libnetdata/avl/avl.h" +#include "libnetdata/clocks/clocks.h" +#include "libnetdata/config/appconfig.h" +#include "libnetdata/ebpf/ebpf.h" +#include "libnetdata/procfile/procfile.h" +#include "daemon/main.h" + +#include "ebpf_apps.h" + +typedef struct netdata_syscall_stat { + unsigned long bytes; // total number of bytes + uint64_t call; // total number of calls + uint64_t ecall; // number of calls that returned error + struct netdata_syscall_stat *next; // Link list +} netdata_syscall_stat_t; + +typedef uint64_t netdata_idx_t; + +typedef struct netdata_publish_syscall { + char *dimension; + char *name; + char *algorithm; + unsigned long nbyte; + unsigned long pbyte; + uint64_t ncall; + uint64_t pcall; + uint64_t nerr; + uint64_t perr; + struct netdata_publish_syscall *next; +} netdata_publish_syscall_t; + +typedef struct netdata_publish_vfs_common { + long write; + long read; + + long running; + long zombie; +} netdata_publish_vfs_common_t; + +typedef struct netdata_error_report { + char comm[16]; + __u32 pid; + + int type; + int err; +} netdata_error_report_t; + +extern ebpf_module_t ebpf_modules[]; +#define EBPF_MODULE_PROCESS_IDX 0 +#define EBPF_MODULE_SOCKET_IDX 1 + +// Copied from musl header +#ifndef offsetof +#if __GNUC__ > 3 +#define offsetof(type, member) __builtin_offsetof(type, member) +#else +#define offsetof(type, member) ((size_t)((char *)&(((type *)0)->member) - (char *)0)) +#endif +#endif + +// Chart defintions +#define NETDATA_EBPF_FAMILY "ebpf" + +// Log file +#define NETDATA_DEVELOPER_LOG_FILE "developer.log" + +// Maximum number of processors monitored on perf events +#define NETDATA_MAX_PROCESSOR 512 + +// Kernel versions calculated with the formula: +// R = MAJOR*65536 + MINOR*256 + PATCH +#define NETDATA_KERNEL_V5_3 328448 +#define NETDATA_KERNEL_V4_15 265984 + +#define EBPF_SYS_CLONE_IDX 11 +#define EBPF_MAX_MAPS 32 + +enum ebpf_algorithms_list { + NETDATA_EBPF_ABSOLUTE_IDX, + NETDATA_EBPF_INCREMENTAL_IDX +}; + +// Threads +extern void *ebpf_process_thread(void *ptr); +extern void *ebpf_socket_thread(void *ptr); + +// Common variables +extern pthread_mutex_t lock; +extern int close_ebpf_plugin; +extern int ebpf_nprocs; +extern int running_on_kernel; +extern char *ebpf_plugin_dir; +extern char kernel_string[64]; + +extern pthread_mutex_t collect_data_mutex; +extern pthread_cond_t collect_data_cond_var; + +// Common functions +extern void ebpf_global_labels(netdata_syscall_stat_t *is, + netdata_publish_syscall_t *pio, + char **dim, + char **name, + int *algorithm, + int end); + +extern void ebpf_write_chart_cmd(char *type, + char *id, + char *title, + char *units, + char *family, + char *charttype, + int order); + +extern void ebpf_write_global_dimension(char *name, char *id, char *algorithm); + +extern void ebpf_create_global_dimension(void *ptr, int end); + +extern void ebpf_create_chart(char *type, + char *id, + char *title, + char *units, + char *family, + int order, + void (*ncd)(void *, int), + void *move, + int end); + +extern void write_begin_chart(char *family, char *name); + +extern void write_chart_dimension(char *dim, long long value); + +extern void write_count_chart(char *name, char *family, netdata_publish_syscall_t *move, uint32_t end); + +extern void write_err_chart(char *name, char *family, netdata_publish_syscall_t *move, int end); + +extern void write_io_chart(char *chart, char *family, char *dwrite, long long vwrite, + char *dread, long long vread); + +extern void fill_ebpf_data(ebpf_data_t *ef); + +extern void ebpf_create_charts_on_apps(char *name, + char *title, + char *units, + char *family, + int order, + char *algorithm, + struct target *root); + +extern void write_end_chart(); + +extern void ebpf_cleanup_publish_syscall(netdata_publish_syscall_t *nps); + +#define EBPF_GLOBAL_SECTION "global" +#define EBPF_PROGRAMS_SECTION "ebpf programs" +#define EBPF_NETWORK_VIEWER_SECTION "network connections" +#define EBPF_SERVICE_NAME_SECTION "service name" + +#define EBPF_COMMON_DIMENSION_CALL "calls/s" +#define EBPF_COMMON_DIMENSION_BITS "kilobits/s" +#define EBPF_COMMON_DIMENSION_BYTES "bytes/s" +#define EBPF_COMMON_DIMENSION_DIFFERENCE "difference" +#define EBPF_COMMON_DIMENSION_PACKETS "packets" + +// Common variables +extern char *ebpf_user_config_dir; +extern char *ebpf_stock_config_dir; +extern int debug_enabled; +extern struct pid_stat *root_of_pids; +extern char *ebpf_algorithms[]; + +// Socket functions and variables +// Common functions +extern void ebpf_socket_create_apps_charts(ebpf_module_t *em, struct target *root); +extern collected_number get_value_from_structure(char *basis, size_t offset); +extern struct pid_stat *root_of_pids; +extern ebpf_process_stat_t *global_process_stat; +extern size_t all_pids_count; +extern int update_every; +extern uint32_t finalized_threads; + +#define EBPF_MAX_SYNCHRONIZATION_TIME 300 + +#endif /* NETDATA_COLLECTOR_EBPF_H */ diff --git a/collectors/ebpf.plugin/ebpf_apps.c b/collectors/ebpf.plugin/ebpf_apps.c new file mode 100644 index 0000000..844ce23 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf_apps.c @@ -0,0 +1,1082 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#include "ebpf.h" +#include "ebpf_socket.h" +#include "ebpf_apps.h" + +// ---------------------------------------------------------------------------- +// internal flags +// handled in code (automatically set) + +static int proc_pid_cmdline_is_needed = 0; // 1 when we need to read /proc/cmdline + +/***************************************************************** + * + * FUNCTIONS USED TO READ HASH TABLES + * + *****************************************************************/ + +/** + * Read statistic hash table. + * + * @param ep the output structure. + * @param fd the file descriptor mapped from kernel ring. + * @param pid the index used to select the data. + * @param bpf_map_lookup_elem a pointer for the function used to read data. + * + * @return It returns 0 when the data was copied and -1 otherwise + */ +int ebpf_read_hash_table(void *ep, int fd, uint32_t pid) +{ + if (!ep) + return -1; + + if (!bpf_map_lookup_elem(fd, &pid, ep)) + return 0; + + return -1; +} + +/** + * Read socket statistic + * + * Read information from kernel ring to user ring. + * + * @param ep the table with all process stats values. + * @param fd the file descriptor mapped from kernel + * @param ef a pointer for the functions mapped from dynamic library + * @param pids the list of pids associated to a target. + * + * @return + */ +size_t read_bandwidth_statistic_using_pid_on_target(ebpf_bandwidth_t **ep, int fd, struct pid_on_target *pids) +{ + size_t count = 0; + while (pids) { + uint32_t current_pid = pids->pid; + if (!ebpf_read_hash_table(ep[current_pid], fd, current_pid)) + count++; + + pids = pids->next; + } + + return count; +} + +/** + * Read bandwidth statistic using hash table + * + * @param out the output tensor that will receive the information. + * @param fd the file descriptor that has the data + * @param bpf_map_lookup_elem a pointer for the function to read the data + * @param bpf_map_get_next_key a pointer fo the function to read the index. + */ +size_t read_bandwidth_statistic_using_hash_table(ebpf_bandwidth_t **out, int fd) +{ + size_t count = 0; + uint32_t key = 0; + uint32_t next_key = 0; + + while (bpf_map_get_next_key(fd, &key, &next_key) == 0) { + ebpf_bandwidth_t *eps = out[next_key]; + if (!eps) { + eps = callocz(1, sizeof(ebpf_process_stat_t)); + out[next_key] = eps; + } + ebpf_read_hash_table(eps, fd, next_key); + } + + return count; +} + +/***************************************************************** + * + * FUNCTIONS CALLED FROM COLLECTORS + * + *****************************************************************/ + +/** + * Am I running as Root + * + * Verify the user that is running the collector. + * + * @return It returns 1 for root and 0 otherwise. + */ +int am_i_running_as_root() +{ + uid_t uid = getuid(), euid = geteuid(); + + if (uid == 0 || euid == 0) { + return 1; + } + + return 0; +} + +/** + * Reset the target values + * + * @param root the pointer to the chain that will be reseted. + * + * @return it returns the number of structures that was reseted. + */ +size_t zero_all_targets(struct target *root) +{ + struct target *w; + size_t count = 0; + + for (w = root; w; w = w->next) { + count++; + + if (unlikely(w->root_pid)) { + struct pid_on_target *pid_on_target = w->root_pid; + + while (pid_on_target) { + struct pid_on_target *pid_on_target_to_free = pid_on_target; + pid_on_target = pid_on_target->next; + free(pid_on_target_to_free); + } + + w->root_pid = NULL; + } + } + + return count; +} + +/** + * Clean the allocated structures + * + * @param agrt the pointer to be cleaned. + */ +void clean_apps_groups_target(struct target *agrt) +{ + struct target *current_target; + while (agrt) { + current_target = agrt; + agrt = current_target->target; + + freez(current_target); + } +} + +/** + * Find or create a new target + * there are targets that are just aggregated to other target (the second argument) + * + * @param id + * @param target + * @param name + * + * @return It returns the target on success and NULL otherwise + */ +struct target *get_apps_groups_target(struct target **agrt, const char *id, struct target *target, const char *name) +{ + int tdebug = 0, thidden = target ? target->hidden : 0, ends_with = 0; + const char *nid = id; + + // extract the options + while (nid[0] == '-' || nid[0] == '+' || nid[0] == '*') { + if (nid[0] == '-') + thidden = 1; + if (nid[0] == '+') + tdebug = 1; + if (nid[0] == '*') + ends_with = 1; + nid++; + } + uint32_t hash = simple_hash(id); + + // find if it already exists + struct target *w, *last = *agrt; + for (w = *agrt; w; w = w->next) { + if (w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0) + return w; + + last = w; + } + + // find an existing target + if (unlikely(!target)) { + while (*name == '-') { + if (*name == '-') + thidden = 1; + name++; + } + + for (target = *agrt; target != NULL; target = target->next) { + if (!target->target && strcmp(name, target->name) == 0) + break; + } + } + + if (target && target->target) + fatal( + "Internal Error: request to link process '%s' to target '%s' which is linked to target '%s'", id, + target->id, target->target->id); + + w = callocz(1, sizeof(struct target)); + strncpyz(w->id, nid, MAX_NAME); + w->idhash = simple_hash(w->id); + + if (unlikely(!target)) + // copy the name + strncpyz(w->name, name, MAX_NAME); + else + // copy the id + strncpyz(w->name, nid, MAX_NAME); + + strncpyz(w->compare, nid, MAX_COMPARE_NAME); + size_t len = strlen(w->compare); + if (w->compare[len - 1] == '*') { + w->compare[len - 1] = '\0'; + w->starts_with = 1; + } + w->ends_with = ends_with; + + if (w->starts_with && w->ends_with) + proc_pid_cmdline_is_needed = 1; + + w->comparehash = simple_hash(w->compare); + w->comparelen = strlen(w->compare); + + w->hidden = thidden; +#ifdef NETDATA_INTERNAL_CHECKS + w->debug_enabled = tdebug; +#else + if (tdebug) + fprintf(stderr, "apps.plugin has been compiled without debugging\n"); +#endif + w->target = target; + + // append it, to maintain the order in apps_groups.conf + if (last) + last->next = w; + else + *agrt = w; + + return w; +} + +/** + * Read the apps_groups.conf file + * + * @param agrt a pointer to apps_group_root_target + * @param path the directory to search apps_%s.conf + * @param file the word to complement the file name. + * + * @return It returns 0 on succcess and -1 otherwise + */ +int ebpf_read_apps_groups_conf(struct target **agdt, struct target **agrt, const char *path, const char *file) +{ + char filename[FILENAME_MAX + 1]; + + snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", path, file); + + // ---------------------------------------- + + procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT); + if (!ff) + return -1; + + procfile_set_quotes(ff, "'\""); + + ff = procfile_readall(ff); + if (!ff) + return -1; + + size_t line, lines = procfile_lines(ff); + + for (line = 0; line < lines; line++) { + size_t word, words = procfile_linewords(ff, line); + if (!words) + continue; + + char *name = procfile_lineword(ff, line, 0); + if (!name || !*name) + continue; + + // find a possibly existing target + struct target *w = NULL; + + // loop through all words, skipping the first one (the name) + for (word = 0; word < words; word++) { + char *s = procfile_lineword(ff, line, word); + if (!s || !*s) + continue; + if (*s == '#') + break; + + // is this the first word? skip it + if (s == name) + continue; + + // add this target + struct target *n = get_apps_groups_target(agrt, s, w, name); + if (!n) { + error("Cannot create target '%s' (line %zu, word %zu)", s, line, word); + continue; + } + + // just some optimization + // to avoid searching for a target for each process + if (!w) + w = n->target ? n->target : n; + } + } + + procfile_close(ff); + + *agdt = get_apps_groups_target(agrt, "p+!o@w#e$i^r&7*5(-i)l-o_", NULL, "other"); // match nothing + if (!*agdt) + fatal("Cannot create default target"); + + struct target *ptr = *agdt; + if (ptr->target) + *agdt = ptr->target; + + return 0; +} + +// the minimum PID of the system +// this is also the pid of the init process +#define INIT_PID 1 + +// ---------------------------------------------------------------------------- +// string lengths + +#define MAX_COMPARE_NAME 100 +#define MAX_NAME 100 +#define MAX_CMDLINE 16384 + +struct pid_stat **all_pids = NULL; // to avoid allocations, we pre-allocate the + // the entire pid space. +struct pid_stat *root_of_pids = NULL; // global list of all processes running + +size_t all_pids_count = 0; // the number of processes running + +struct target + *apps_groups_default_target = NULL, // the default target + *apps_groups_root_target = NULL, // apps_groups.conf defined + *users_root_target = NULL, // users + *groups_root_target = NULL; // user groups + +size_t apps_groups_targets_count = 0; // # of apps_groups.conf targets + +// ---------------------------------------------------------------------------- +// internal counters + +static size_t + // global_iterations_counter = 1, + calls_counter = 0, + // file_counter = 0, + // filenames_allocated_counter = 0, + // inodes_changed_counter = 0, + // links_changed_counter = 0, + targets_assignment_counter = 0; + +// ---------------------------------------------------------------------------- +// debugging + +// log each problem once per process +// log flood protection flags (log_thrown) +#define PID_LOG_IO 0x00000001 +#define PID_LOG_STATUS 0x00000002 +#define PID_LOG_CMDLINE 0x00000004 +#define PID_LOG_FDS 0x00000008 +#define PID_LOG_STAT 0x00000010 + +int debug_enabled = 0; + +#ifdef NETDATA_INTERNAL_CHECKS + +#define debug_log(fmt, args...) \ + do { \ + if (unlikely(debug_enabled)) \ + debug_log_int(fmt, ##args); \ + } while (0) + +#else + +static inline void debug_log_dummy(void) +{ +} +#define debug_log(fmt, args...) debug_log_dummy() + +#endif + +/** + * Managed log + * + * Store log information if it is necessary. + * + * @param p the pid stat structure + * @param log the log id + * @param status the return from a function. + * + * @return It returns the status value. + */ +static inline int managed_log(struct pid_stat *p, uint32_t log, int status) +{ + if (unlikely(!status)) { + // error("command failed log %u, errno %d", log, errno); + + if (unlikely(debug_enabled || errno != ENOENT)) { + if (unlikely(debug_enabled || !(p->log_thrown & log))) { + p->log_thrown |= log; + switch (log) { + case PID_LOG_IO: + error( + "Cannot process %s/proc/%d/io (command '%s')", netdata_configured_host_prefix, p->pid, + p->comm); + break; + + case PID_LOG_STATUS: + error( + "Cannot process %s/proc/%d/status (command '%s')", netdata_configured_host_prefix, p->pid, + p->comm); + break; + + case PID_LOG_CMDLINE: + error( + "Cannot process %s/proc/%d/cmdline (command '%s')", netdata_configured_host_prefix, p->pid, + p->comm); + break; + + case PID_LOG_FDS: + error( + "Cannot process entries in %s/proc/%d/fd (command '%s')", netdata_configured_host_prefix, + p->pid, p->comm); + break; + + case PID_LOG_STAT: + break; + + default: + error("unhandled error for pid %d, command '%s'", p->pid, p->comm); + break; + } + } + } + errno = 0; + } else if (unlikely(p->log_thrown & log)) { + // error("unsetting log %u on pid %d", log, p->pid); + p->log_thrown &= ~log; + } + + return status; +} + +/** + * Get PID entry + * + * Get or allocate the PID entry for the specifid pid. + * + * @param pid the pid to search the data. + * + * @return It returns the pid entry structure + */ +static inline struct pid_stat *get_pid_entry(pid_t pid) +{ + if (unlikely(all_pids[pid])) + return all_pids[pid]; + + struct pid_stat *p = callocz(1, sizeof(struct pid_stat)); + + if (likely(root_of_pids)) + root_of_pids->prev = p; + + p->next = root_of_pids; + root_of_pids = p; + + p->pid = pid; + + all_pids[pid] = p; + all_pids_count++; + + return p; +} + +/** + * Assign the PID to a target. + * + * @param p the pid_stat structure to assign for a target. + */ +static inline void assign_target_to_pid(struct pid_stat *p) +{ + targets_assignment_counter++; + + uint32_t hash = simple_hash(p->comm); + size_t pclen = strlen(p->comm); + + struct target *w; + for (w = apps_groups_root_target; w; w = w->next) { + // if(debug_enabled || (p->target && p->target->debug_enabled)) debug_log_int("\t\tcomparing '%s' with '%s'", w->compare, p->comm); + + // find it - 4 cases: + // 1. the target is not a pattern + // 2. the target has the prefix + // 3. the target has the suffix + // 4. the target is something inside cmdline + + if (unlikely( + ((!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm)) || + (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen)) || + (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen])) || + (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && p->cmdline && strstr(p->cmdline, w->compare))))) { + if (w->target) + p->target = w->target; + else + p->target = w; + + if (debug_enabled || (p->target && p->target->debug_enabled)) + debug_log_int("%s linked to target %s", p->comm, p->target->name); + + break; + } + } +} + +// ---------------------------------------------------------------------------- +// update pids from proc + +/** + * Read cmd line from /proc/PID/cmdline + * + * @param p the pid_stat_structure. + * + * @return It returns 1 on success and 0 otherwise. + */ +static inline int read_proc_pid_cmdline(struct pid_stat *p) +{ + static char cmdline[MAX_CMDLINE + 1]; + + if (unlikely(!p->cmdline_filename)) { + char filename[FILENAME_MAX + 1]; + snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", netdata_configured_host_prefix, p->pid); + p->cmdline_filename = strdupz(filename); + } + + int fd = open(p->cmdline_filename, procfile_open_flags, 0666); + if (unlikely(fd == -1)) + goto cleanup; + + ssize_t i, bytes = read(fd, cmdline, MAX_CMDLINE); + close(fd); + + if (unlikely(bytes < 0)) + goto cleanup; + + cmdline[bytes] = '\0'; + for (i = 0; i < bytes; i++) { + if (unlikely(!cmdline[i])) + cmdline[i] = ' '; + } + + if (p->cmdline) + freez(p->cmdline); + p->cmdline = strdupz(cmdline); + + debug_log("Read file '%s' contents: %s", p->cmdline_filename, p->cmdline); + + return 1; + +cleanup: + // copy the command to the command line + if (p->cmdline) + freez(p->cmdline); + p->cmdline = strdupz(p->comm); + return 0; +} + +/** + * Read information from /proc/PID/stat and /proc/PID/cmdline + * Assign target to pid + * + * @param p the pid stat structure to store the data. + * @param ptr an useless argument. + */ +static inline int read_proc_pid_stat(struct pid_stat *p, void *ptr) +{ + UNUSED(ptr); + + static procfile *ff = NULL; + + if (unlikely(!p->stat_filename)) { + char filename[FILENAME_MAX + 1]; + snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", netdata_configured_host_prefix, p->pid); + p->stat_filename = strdupz(filename); + } + + int set_quotes = (!ff) ? 1 : 0; + + struct stat statbuf; + if (stat(p->stat_filename, &statbuf)) + return 0; + + ff = procfile_reopen(ff, p->stat_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO); + if (unlikely(!ff)) + return 0; + + if (unlikely(set_quotes)) + procfile_set_open_close(ff, "(", ")"); + + ff = procfile_readall(ff); + if (unlikely(!ff)) + return 0; + + p->last_stat_collected_usec = p->stat_collected_usec; + p->stat_collected_usec = now_monotonic_usec(); + calls_counter++; + + char *comm = procfile_lineword(ff, 0, 1); + p->ppid = (int32_t)str2pid_t(procfile_lineword(ff, 0, 3)); + + if (strcmp(p->comm, comm) != 0) { + if (unlikely(debug_enabled)) { + if (p->comm[0]) + debug_log("\tpid %d (%s) changed name to '%s'", p->pid, p->comm, comm); + else + debug_log("\tJust added %d (%s)", p->pid, comm); + } + + strncpyz(p->comm, comm, MAX_COMPARE_NAME); + + // /proc/<pid>/cmdline + if (likely(proc_pid_cmdline_is_needed)) + managed_log(p, PID_LOG_CMDLINE, read_proc_pid_cmdline(p)); + + assign_target_to_pid(p); + } + + if (unlikely(debug_enabled || (p->target && p->target->debug_enabled))) + debug_log_int( + "READ PROC/PID/STAT: %s/proc/%d/stat, process: '%s' on target '%s' (dt=%llu)", + netdata_configured_host_prefix, p->pid, p->comm, (p->target) ? p->target->name : "UNSET", + p->stat_collected_usec - p->last_stat_collected_usec); + + return 1; +} + +/** + * Collect data for PID + * + * @param pid the current pid that we are working + * @param ptr a NULL value + * + * @return It returns 1 on succcess and 0 otherwise + */ +static inline int collect_data_for_pid(pid_t pid, void *ptr) +{ + if (unlikely(pid < 0 || pid > pid_max)) { + error("Invalid pid %d read (expected %d to %d). Ignoring process.", pid, 0, pid_max); + return 0; + } + + struct pid_stat *p = get_pid_entry(pid); + if (unlikely(!p || p->read)) + return 0; + p->read = 1; + + if (unlikely(!managed_log(p, PID_LOG_STAT, read_proc_pid_stat(p, ptr)))) + // there is no reason to proceed if we cannot get its status + return 0; + + // check its parent pid + if (unlikely(p->ppid < 0 || p->ppid > pid_max)) { + error("Pid %d (command '%s') states invalid parent pid %d. Using 0.", pid, p->comm, p->ppid); + p->ppid = 0; + } + + // mark it as updated + p->updated = 1; + p->keep = 0; + p->keeploops = 0; + + return 1; +} + +/** + * Fill link list of parents with children PIDs + */ +static inline void link_all_processes_to_their_parents(void) +{ + struct pid_stat *p, *pp; + + // link all children to their parents + // and update children count on parents + for (p = root_of_pids; p; p = p->next) { + // for each process found + + p->sortlist = 0; + p->parent = NULL; + + if (unlikely(!p->ppid)) { + p->parent = NULL; + continue; + } + + pp = all_pids[p->ppid]; + if (likely(pp)) { + p->parent = pp; + pp->children_count++; + + if (unlikely(debug_enabled || (p->target && p->target->debug_enabled))) + debug_log_int( + "child %d (%s, %s) on target '%s' has parent %d (%s, %s).", p->pid, p->comm, + p->updated ? "running" : "exited", (p->target) ? p->target->name : "UNSET", pp->pid, pp->comm, + pp->updated ? "running" : "exited"); + } else { + p->parent = NULL; + debug_log("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid); + } + } +} + +/** + * Aggregate PIDs to targets. + */ +static void apply_apps_groups_targets_inheritance(void) +{ + struct pid_stat *p = NULL; + + // children that do not have a target + // inherit their target from their parent + int found = 1, loops = 0; + while (found) { + if (unlikely(debug_enabled)) + loops++; + found = 0; + for (p = root_of_pids; p; p = p->next) { + // if this process does not have a target + // and it has a parent + // and its parent has a target + // then, set the parent's target to this process + if (unlikely(!p->target && p->parent && p->parent->target)) { + p->target = p->parent->target; + found++; + + if (debug_enabled || (p->target && p->target->debug_enabled)) + debug_log_int( + "TARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s).", p->target->name, + p->pid, p->comm, p->parent->pid, p->parent->comm); + } + } + } + + // find all the procs with 0 childs and merge them to their parents + // repeat, until nothing more can be done. + int sortlist = 1; + found = 1; + while (found) { + if (unlikely(debug_enabled)) + loops++; + found = 0; + + for (p = root_of_pids; p; p = p->next) { + if (unlikely(!p->sortlist && !p->children_count)) + p->sortlist = sortlist++; + + if (unlikely( + !p->children_count // if this process does not have any children + && !p->merged // and is not already merged + && p->parent // and has a parent + && p->parent->children_count // and its parent has children + // and the target of this process and its parent is the same, + // or the parent does not have a target + && (p->target == p->parent->target || !p->parent->target) && + p->ppid != INIT_PID // and its parent is not init + )) { + // mark it as merged + p->parent->children_count--; + p->merged = 1; + + // the parent inherits the child's target, if it does not have a target itself + if (unlikely(p->target && !p->parent->target)) { + p->parent->target = p->target; + + if (debug_enabled || (p->target && p->target->debug_enabled)) + debug_log_int( + "TARGET INHERITANCE: %s is inherited by %d (%s) from its child %d (%s).", p->target->name, + p->parent->pid, p->parent->comm, p->pid, p->comm); + } + + found++; + } + } + + debug_log("TARGET INHERITANCE: merged %d processes", found); + } + + // init goes always to default target + if (all_pids[INIT_PID]) + all_pids[INIT_PID]->target = apps_groups_default_target; + + // pid 0 goes always to default target + if (all_pids[0]) + all_pids[0]->target = apps_groups_default_target; + + // give a default target on all top level processes + if (unlikely(debug_enabled)) + loops++; + for (p = root_of_pids; p; p = p->next) { + // if the process is not merged itself + // then is is a top level process + if (unlikely(!p->merged && !p->target)) + p->target = apps_groups_default_target; + + // make sure all processes have a sortlist + if (unlikely(!p->sortlist)) + p->sortlist = sortlist++; + } + + if (all_pids[1]) + all_pids[1]->sortlist = sortlist++; + + // give a target to all merged child processes + found = 1; + while (found) { + if (unlikely(debug_enabled)) + loops++; + found = 0; + for (p = root_of_pids; p; p = p->next) { + if (unlikely(!p->target && p->merged && p->parent && p->parent->target)) { + p->target = p->parent->target; + found++; + + if (debug_enabled || (p->target && p->target->debug_enabled)) + debug_log_int( + "TARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s) at phase 2.", + p->target->name, p->pid, p->comm, p->parent->pid, p->parent->comm); + } + } + } + + debug_log("apply_apps_groups_targets_inheritance() made %d loops on the process tree", loops); +} + +/** + * Update target timestamp. + * + * @param root the targets that will be updated. + */ +static inline void post_aggregate_targets(struct target *root) +{ + struct target *w; + for (w = root; w; w = w->next) { + if (w->collected_starttime) { + if (!w->starttime || w->collected_starttime < w->starttime) { + w->starttime = w->collected_starttime; + } + } else { + w->starttime = 0; + } + } +} + +/** + * Remove PID from the link list. + * + * @param pid the PID that will be removed. + */ +static inline void del_pid_entry(pid_t pid) +{ + struct pid_stat *p = all_pids[pid]; + + if (unlikely(!p)) { + error("attempted to free pid %d that is not allocated.", pid); + return; + } + + debug_log("process %d %s exited, deleting it.", pid, p->comm); + + if (root_of_pids == p) + root_of_pids = p->next; + + if (p->next) + p->next->prev = p->prev; + if (p->prev) + p->prev->next = p->next; + + freez(p->stat_filename); + freez(p->status_filename); + freez(p->io_filename); + freez(p->cmdline_filename); + freez(p->cmdline); + freez(p); + + all_pids[pid] = NULL; + all_pids_count--; +} + +/** + * Remove PIDs when they are not running more. + */ +void cleanup_exited_pids() +{ + struct pid_stat *p = NULL; + + for (p = root_of_pids; p;) { + if (!p->updated && (!p->keep || p->keeploops > 0)) { + if (unlikely(debug_enabled && (p->keep || p->keeploops))) + debug_log(" > CLEANUP cannot keep exited process %d (%s) anymore - removing it.", p->pid, p->comm); + + pid_t r = p->pid; + p = p->next; + del_pid_entry(r); + + // Clean process structure + freez(global_process_stats[r]); + global_process_stats[r] = NULL; + + freez(current_apps_data[r]); + current_apps_data[r] = NULL; + + // Clean socket structures + if (socket_bandwidth_curr) { + freez(socket_bandwidth_curr[r]); + socket_bandwidth_curr[r] = NULL; + } + } else { + if (unlikely(p->keep)) + p->keeploops++; + p->keep = 0; + p = p->next; + } + } +} + +/** + * Read proc filesystem for the first time. + * + * @return It returns 0 on success and -1 otherwise. + */ +static inline void read_proc_filesystem() +{ + char dirname[FILENAME_MAX + 1]; + + snprintfz(dirname, FILENAME_MAX, "%s/proc", netdata_configured_host_prefix); + DIR *dir = opendir(dirname); + if (!dir) + return; + + struct dirent *de = NULL; + + while ((de = readdir(dir))) { + char *endptr = de->d_name; + + if (unlikely(de->d_type != DT_DIR || de->d_name[0] < '0' || de->d_name[0] > '9')) + continue; + + pid_t pid = (pid_t)strtoul(de->d_name, &endptr, 10); + + // make sure we read a valid number + if (unlikely(endptr == de->d_name || *endptr != '\0')) + continue; + + collect_data_for_pid(pid, NULL); + } + closedir(dir); +} + +/** + * Aggregated PID on target + * + * @param w the target output + * @param p the pid with information to update + * @param o never used + */ +static inline void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) +{ + UNUSED(o); + + if (unlikely(!p->updated)) { + // the process is not running + return; + } + + if (unlikely(!w)) { + error("pid %d %s was left without a target!", p->pid, p->comm); + return; + } + + w->processes++; + struct pid_on_target *pid_on_target = mallocz(sizeof(struct pid_on_target)); + pid_on_target->pid = p->pid; + pid_on_target->next = w->root_pid; + w->root_pid = pid_on_target; +} + +/** + * Collect data for all process + * + * Read data from hash table and store it in appropriate vectors. + * It also creates the link between targets and PIDs. + * + * @param tbl_pid_stats_fd The mapped file descriptor for the hash table. + */ +void collect_data_for_all_processes(int tbl_pid_stats_fd) +{ + struct pid_stat *pids = root_of_pids; // global list of all processes running + while (pids) { + if (pids->updated_twice) { + pids->read = 0; // mark it as not read, so that collect_data_for_pid() will read it + pids->updated = 0; + pids->merged = 0; + pids->children_count = 0; + pids->parent = NULL; + } else { + if (pids->updated) + pids->updated_twice = 1; + } + + pids = pids->next; + } + + read_proc_filesystem(); + + uint32_t key; + pids = root_of_pids; // global list of all processes running + // while (bpf_map_get_next_key(tbl_pid_stats_fd, &key, &next_key) == 0) { + while (pids) { + key = pids->pid; + ebpf_process_stat_t *w = global_process_stats[key]; + if (!w) { + w = mallocz(sizeof(ebpf_process_stat_t)); + global_process_stats[key] = w; + } + + if (bpf_map_lookup_elem(tbl_pid_stats_fd, &key, w)) { + // Clean Process structures + freez(w); + global_process_stats[key] = NULL; + + freez(current_apps_data[key]); + current_apps_data[key] = NULL; + + // Clean socket structures + if (socket_bandwidth_curr) { + freez(socket_bandwidth_curr[key]); + socket_bandwidth_curr[key] = NULL; + } + + pids = pids->next; + continue; + } + + pids = pids->next; + } + + link_all_processes_to_their_parents(); + + apply_apps_groups_targets_inheritance(); + + apps_groups_targets_count = zero_all_targets(apps_groups_root_target); + + // this has to be done, before the cleanup + // // concentrate everything on the targets + for (pids = root_of_pids; pids; pids = pids->next) + aggregate_pid_on_target(pids->target, pids, NULL); + + post_aggregate_targets(apps_groups_root_target); +} diff --git a/collectors/ebpf.plugin/ebpf_apps.h b/collectors/ebpf.plugin/ebpf_apps.h new file mode 100644 index 0000000..f8cb7ac --- /dev/null +++ b/collectors/ebpf.plugin/ebpf_apps.h @@ -0,0 +1,430 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#ifndef NETDATA_EBPF_APPS_H +#define NETDATA_EBPF_APPS_H 1 + +#include "libnetdata/threads/threads.h" +#include "libnetdata/locks/locks.h" +#include "libnetdata/avl/avl.h" +#include "libnetdata/clocks/clocks.h" +#include "libnetdata/config/appconfig.h" +#include "libnetdata/ebpf/ebpf.h" + +#define NETDATA_APPS_FAMILY "apps" +#define NETDATA_APPS_FILE_GROUP "ebpf file" +#define NETDATA_APPS_VFS_GROUP "ebpf vfs" +#define NETDATA_APPS_PROCESS_GROUP "ebpf process" +#define NETDATA_APPS_NET_GROUP "ebpf net" + +#include "ebpf_process.h" + +#define MAX_COMPARE_NAME 100 +#define MAX_NAME 100 + +// ---------------------------------------------------------------------------- +// process_pid_stat +// +// Fields read from the kernel ring for a specific PID +// +typedef struct process_pid_stat { + uint64_t pid_tgid; // Unique identifier + uint32_t pid; // process id + + // Count number of calls done for specific function + uint32_t open_call; + uint32_t write_call; + uint32_t writev_call; + uint32_t read_call; + uint32_t readv_call; + uint32_t unlink_call; + uint32_t exit_call; + uint32_t release_call; + uint32_t fork_call; + uint32_t clone_call; + uint32_t close_call; + + // Count number of bytes written or read + uint64_t write_bytes; + uint64_t writev_bytes; + uint64_t readv_bytes; + uint64_t read_bytes; + + // Count number of errors for the specified function + uint32_t open_err; + uint32_t write_err; + uint32_t writev_err; + uint32_t read_err; + uint32_t readv_err; + uint32_t unlink_err; + uint32_t fork_err; + uint32_t clone_err; + uint32_t close_err; +} process_pid_stat_t; + +// ---------------------------------------------------------------------------- +// socket_bandwidth +// +// Fields read from the kernel ring for a specific PID +// +typedef struct socket_bandwidth { + uint64_t first; + uint64_t ct; + uint64_t sent; + uint64_t received; + unsigned char removed; +} socket_bandwidth_t; + +// ---------------------------------------------------------------------------- +// pid_stat +// +// structure to store data for each process running +// see: man proc for the description of the fields + +struct pid_fd { + int fd; + +#ifndef __FreeBSD__ + ino_t inode; + char *filename; + uint32_t link_hash; + size_t cache_iterations_counter; + size_t cache_iterations_reset; +#endif +}; + +struct target { + char compare[MAX_COMPARE_NAME + 1]; + uint32_t comparehash; + size_t comparelen; + + char id[MAX_NAME + 1]; + uint32_t idhash; + + char name[MAX_NAME + 1]; + + uid_t uid; + gid_t gid; + + /* These variables are not necessary for eBPF collector + kernel_uint_t minflt; + kernel_uint_t cminflt; + kernel_uint_t majflt; + kernel_uint_t cmajflt; + kernel_uint_t utime; + kernel_uint_t stime; + kernel_uint_t gtime; + kernel_uint_t cutime; + kernel_uint_t cstime; + kernel_uint_t cgtime; + kernel_uint_t num_threads; + // kernel_uint_t rss; + + kernel_uint_t status_vmsize; + kernel_uint_t status_vmrss; + kernel_uint_t status_vmshared; + kernel_uint_t status_rssfile; + kernel_uint_t status_rssshmem; + kernel_uint_t status_vmswap; + + kernel_uint_t io_logical_bytes_read; + kernel_uint_t io_logical_bytes_written; + // kernel_uint_t io_read_calls; + // kernel_uint_t io_write_calls; + kernel_uint_t io_storage_bytes_read; + kernel_uint_t io_storage_bytes_written; + // kernel_uint_t io_cancelled_write_bytes; + + int *target_fds; + int target_fds_size; + + kernel_uint_t openfiles; + kernel_uint_t openpipes; + kernel_uint_t opensockets; + kernel_uint_t openinotifies; + kernel_uint_t openeventfds; + kernel_uint_t opentimerfds; + kernel_uint_t opensignalfds; + kernel_uint_t openeventpolls; + kernel_uint_t openother; + */ + + kernel_uint_t starttime; + kernel_uint_t collected_starttime; + + /* + kernel_uint_t uptime_min; + kernel_uint_t uptime_sum; + kernel_uint_t uptime_max; + */ + + unsigned int processes; // how many processes have been merged to this + int exposed; // if set, we have sent this to netdata + int hidden; // if set, we set the hidden flag on the dimension + int debug_enabled; + int ends_with; + int starts_with; // if set, the compare string matches only the + // beginning of the command + + struct pid_on_target *root_pid; // list of aggregated pids for target debugging + + struct target *target; // the one that will be reported to netdata + struct target *next; +}; + +extern struct target *apps_groups_default_target; +extern struct target *apps_groups_root_target; +extern struct target *users_root_target; +extern struct target *groups_root_target; + +struct pid_stat { + int32_t pid; + char comm[MAX_COMPARE_NAME + 1]; + char *cmdline; + + uint32_t log_thrown; + + // char state; + int32_t ppid; + + // int32_t pgrp; + // int32_t session; + // int32_t tty_nr; + // int32_t tpgid; + // uint64_t flags; + + /* + // these are raw values collected + kernel_uint_t minflt_raw; + kernel_uint_t cminflt_raw; + kernel_uint_t majflt_raw; + kernel_uint_t cmajflt_raw; + kernel_uint_t utime_raw; + kernel_uint_t stime_raw; + kernel_uint_t gtime_raw; // guest_time + kernel_uint_t cutime_raw; + kernel_uint_t cstime_raw; + kernel_uint_t cgtime_raw; // cguest_time + + // these are rates + kernel_uint_t minflt; + kernel_uint_t cminflt; + kernel_uint_t majflt; + kernel_uint_t cmajflt; + kernel_uint_t utime; + kernel_uint_t stime; + kernel_uint_t gtime; + kernel_uint_t cutime; + kernel_uint_t cstime; + kernel_uint_t cgtime; + + // int64_t priority; + // int64_t nice; + int32_t num_threads; + // int64_t itrealvalue; + kernel_uint_t collected_starttime; + // kernel_uint_t vsize; + // kernel_uint_t rss; + // kernel_uint_t rsslim; + // kernel_uint_t starcode; + // kernel_uint_t endcode; + // kernel_uint_t startstack; + // kernel_uint_t kstkesp; + // kernel_uint_t kstkeip; + // uint64_t signal; + // uint64_t blocked; + // uint64_t sigignore; + // uint64_t sigcatch; + // uint64_t wchan; + // uint64_t nswap; + // uint64_t cnswap; + // int32_t exit_signal; + // int32_t processor; + // uint32_t rt_priority; + // uint32_t policy; + // kernel_uint_t delayacct_blkio_ticks; + + uid_t uid; + gid_t gid; + + kernel_uint_t status_vmsize; + kernel_uint_t status_vmrss; + kernel_uint_t status_vmshared; + kernel_uint_t status_rssfile; + kernel_uint_t status_rssshmem; + kernel_uint_t status_vmswap; +#ifndef __FreeBSD__ + ARL_BASE *status_arl; +#endif + + kernel_uint_t io_logical_bytes_read_raw; + kernel_uint_t io_logical_bytes_written_raw; + // kernel_uint_t io_read_calls_raw; + // kernel_uint_t io_write_calls_raw; + kernel_uint_t io_storage_bytes_read_raw; + kernel_uint_t io_storage_bytes_written_raw; + // kernel_uint_t io_cancelled_write_bytes_raw; + + kernel_uint_t io_logical_bytes_read; + kernel_uint_t io_logical_bytes_written; + // kernel_uint_t io_read_calls; + // kernel_uint_t io_write_calls; + kernel_uint_t io_storage_bytes_read; + kernel_uint_t io_storage_bytes_written; + // kernel_uint_t io_cancelled_write_bytes; + */ + + struct pid_fd *fds; // array of fds it uses + size_t fds_size; // the size of the fds array + + int children_count; // number of processes directly referencing this + unsigned char keep : 1; // 1 when we need to keep this process in memory even after it exited + int keeploops; // increases by 1 every time keep is 1 and updated 0 + unsigned char updated : 1; // 1 when the process is currently running + unsigned char updated_twice : 1; // 1 when the process was running in the previous iteration + unsigned char merged : 1; // 1 when it has been merged to its parent + unsigned char read : 1; // 1 when we have already read this process for this iteration + + int sortlist; // higher numbers = top on the process tree + + // each process gets a unique number + + struct target *target; // app_groups.conf targets + struct target *user_target; // uid based targets + struct target *group_target; // gid based targets + + usec_t stat_collected_usec; + usec_t last_stat_collected_usec; + + usec_t io_collected_usec; + usec_t last_io_collected_usec; + + kernel_uint_t uptime; + + char *fds_dirname; // the full directory name in /proc/PID/fd + + char *stat_filename; + char *status_filename; + char *io_filename; + char *cmdline_filename; + + struct pid_stat *parent; + struct pid_stat *prev; + struct pid_stat *next; +}; + +// ---------------------------------------------------------------------------- +// target +// +// target is the structure that processes are aggregated to be reported +// to netdata. +// +// - Each entry in /etc/apps_groups.conf creates a target. +// - Each user and group used by a process in the system, creates a target. +struct pid_on_target { + int32_t pid; + struct pid_on_target *next; +}; + +// ---------------------------------------------------------------------------- +// Structures used to read information from kernel ring +typedef struct ebpf_process_stat { + uint64_t pid_tgid; + uint32_t pid; + + //Counter + uint32_t open_call; + uint32_t write_call; + uint32_t writev_call; + uint32_t read_call; + uint32_t readv_call; + uint32_t unlink_call; + uint32_t exit_call; + uint32_t release_call; + uint32_t fork_call; + uint32_t clone_call; + uint32_t close_call; + + //Accumulator + uint64_t write_bytes; + uint64_t writev_bytes; + uint64_t readv_bytes; + uint64_t read_bytes; + + //Counter + uint32_t open_err; + uint32_t write_err; + uint32_t writev_err; + uint32_t read_err; + uint32_t readv_err; + uint32_t unlink_err; + uint32_t fork_err; + uint32_t clone_err; + uint32_t close_err; + + uint8_t removeme; +} ebpf_process_stat_t; + +typedef struct ebpf_bandwidth { + uint32_t pid; + + uint64_t first; // First timestamp + uint64_t ct; // Last timestamp + uint64_t bytes_sent; // Bytes sent + uint64_t bytes_received; // Bytes received + uint64_t call_tcp_sent; // Number of times tcp_sendmsg was called + uint64_t call_tcp_received; // Number of times tcp_cleanup_rbuf was called + uint64_t retransmit; // Number of times tcp_retransmit was called + uint64_t call_udp_sent; // Number of times udp_sendmsg was called + uint64_t call_udp_received; // Number of times udp_recvmsg was called +} ebpf_bandwidth_t; + +/** + * Internal function used to write debug messages. + * + * @param fmt the format to create the message. + * @param ... the arguments to fill the format. + */ +static inline void debug_log_int(const char *fmt, ...) +{ + va_list args; + + fprintf(stderr, "apps.plugin: "); + va_start(args, fmt); + vfprintf(stderr, fmt, args); + va_end(args); + + fputc('\n', stderr); +} + +// ---------------------------------------------------------------------------- +// Exported variabled and functions +// +extern struct pid_stat **all_pids; + +extern int ebpf_read_apps_groups_conf(struct target **apps_groups_default_target, + struct target **apps_groups_root_target, + const char *path, + const char *file); + +extern void clean_apps_groups_target(struct target *apps_groups_root_target); + +extern size_t zero_all_targets(struct target *root); + +extern int am_i_running_as_root(); + +extern void cleanup_exited_pids(); + +extern int ebpf_read_hash_table(void *ep, int fd, uint32_t pid); + +extern size_t read_processes_statistic_using_pid_on_target(ebpf_process_stat_t **ep, + int fd, + struct pid_on_target *pids); + +extern size_t read_bandwidth_statistic_using_pid_on_target(ebpf_bandwidth_t **ep, int fd, struct pid_on_target *pids); + +extern void collect_data_for_all_processes(int tbl_pid_stats_fd); + +extern ebpf_process_stat_t **global_process_stats; +extern ebpf_process_publish_apps_t **current_apps_data; + +#endif /* NETDATA_EBPF_APPS_H */ diff --git a/collectors/ebpf.plugin/ebpf_kernel_reject_list.txt b/collectors/ebpf.plugin/ebpf_kernel_reject_list.txt new file mode 100644 index 0000000..d56b216 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf_kernel_reject_list.txt @@ -0,0 +1 @@ +Ubuntu 4.18.0-13. diff --git a/collectors/ebpf.plugin/ebpf_process.c b/collectors/ebpf.plugin/ebpf_process.c new file mode 100644 index 0000000..27e39d1 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf_process.c @@ -0,0 +1,1057 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#include <sys/resource.h> + +#include "ebpf.h" +#include "ebpf_process.h" + +/***************************************************************** + * + * GLOBAL VARIABLES + * + *****************************************************************/ + +static char *process_dimension_names[NETDATA_KEY_PUBLISH_PROCESS_END] = { "open", "close", "delete", "read", "write", + "process", "task", "process", "thread" }; +static char *process_id_names[NETDATA_KEY_PUBLISH_PROCESS_END] = { "do_sys_open", "__close_fd", "vfs_unlink", + "vfs_read", "vfs_write", "do_exit", + "release_task", "_do_fork", "sys_clone" }; +static char *status[] = { "process", "zombie" }; + +static netdata_idx_t *process_hash_values = NULL; +static netdata_syscall_stat_t *process_aggregated_data = NULL; +static netdata_publish_syscall_t *process_publish_aggregated = NULL; + +static ebpf_data_t process_data; + +ebpf_process_stat_t **global_process_stats = NULL; +ebpf_process_publish_apps_t **current_apps_data = NULL; + +int process_enabled = 0; + +static int *map_fd = NULL; +static struct bpf_object *objects = NULL; +static struct bpf_link **probe_links = NULL; + +/***************************************************************** + * + * PROCESS DATA AND SEND TO NETDATA + * + *****************************************************************/ + +/** + * Update publish structure before to send data to Netdata. + * + * @param publish the first output structure with independent dimensions + * @param pvc the second output structure with correlated dimensions + * @param input the structure with the input data. + */ +static void ebpf_update_global_publish( + netdata_publish_syscall_t *publish, netdata_publish_vfs_common_t *pvc, netdata_syscall_stat_t *input) +{ + netdata_publish_syscall_t *move = publish; + int selector = NETDATA_KEY_PUBLISH_PROCESS_OPEN; + while (move) { + // Until NETDATA_KEY_PUBLISH_PROCESS_READ we are creating accumulators, so it is possible + // to use incremental charts, but after this we will do some math with the values, so we are storing + // absolute values + if (selector < NETDATA_KEY_PUBLISH_PROCESS_READ) { + move->ncall = input->call; + move->nbyte = input->bytes; + move->nerr = input->ecall; + } else { + move->ncall = (input->call > move->pcall) ? input->call - move->pcall : move->pcall - input->call; + move->nbyte = (input->bytes > move->pbyte) ? input->bytes - move->pbyte : move->pbyte - input->bytes; + move->nerr = (input->ecall > move->nerr) ? input->ecall - move->perr : move->perr - input->ecall; + + move->pcall = input->call; + move->pbyte = input->bytes; + move->perr = input->ecall; + } + + input = input->next; + move = move->next; + selector++; + } + + pvc->write = -((long)publish[NETDATA_KEY_PUBLISH_PROCESS_WRITE].nbyte); + pvc->read = (long)publish[NETDATA_KEY_PUBLISH_PROCESS_READ].nbyte; + + pvc->running = (long)publish[NETDATA_KEY_PUBLISH_PROCESS_FORK].ncall - (long)publish[NETDATA_KEY_PUBLISH_PROCESS_CLONE].ncall; + publish[NETDATA_KEY_PUBLISH_PROCESS_RELEASE_TASK].ncall = -publish[NETDATA_KEY_PUBLISH_PROCESS_RELEASE_TASK].ncall; + pvc->zombie = (long)publish[NETDATA_KEY_PUBLISH_PROCESS_EXIT].ncall + (long)publish[NETDATA_KEY_PUBLISH_PROCESS_RELEASE_TASK].ncall; +} + +/** + * Call the necessary functions to create a chart. + * + * @param family the chart family + * @param move the pointer with the values that will be published + */ +static void write_status_chart(char *family, netdata_publish_vfs_common_t *pvc) +{ + write_begin_chart(family, NETDATA_PROCESS_STATUS_NAME); + + write_chart_dimension(status[0], (long long)pvc->running); + write_chart_dimension(status[1], (long long)pvc->zombie); + + write_end_chart(); +} + +/** + * Send data to Netdata calling auxiliar functions. + * + * @param em the structure with thread information + */ +static void ebpf_process_send_data(ebpf_module_t *em) +{ + netdata_publish_vfs_common_t pvc; + ebpf_update_global_publish(process_publish_aggregated, &pvc, process_aggregated_data); + + write_count_chart( + NETDATA_FILE_OPEN_CLOSE_COUNT, NETDATA_EBPF_FAMILY, process_publish_aggregated, 2); + + write_count_chart( + NETDATA_VFS_FILE_CLEAN_COUNT, NETDATA_EBPF_FAMILY, &process_publish_aggregated[NETDATA_DEL_START], 1); + + write_count_chart( + NETDATA_VFS_FILE_IO_COUNT, NETDATA_EBPF_FAMILY, &process_publish_aggregated[NETDATA_IN_START_BYTE], 2); + + write_count_chart( + NETDATA_EXIT_SYSCALL, NETDATA_EBPF_FAMILY, &process_publish_aggregated[NETDATA_EXIT_START], 2); + write_count_chart( + NETDATA_PROCESS_SYSCALL, NETDATA_EBPF_FAMILY, &process_publish_aggregated[NETDATA_PROCESS_START], 2); + + write_status_chart(NETDATA_EBPF_FAMILY, &pvc); + if (em->mode < MODE_ENTRY) { + write_err_chart( + NETDATA_FILE_OPEN_ERR_COUNT, NETDATA_EBPF_FAMILY, process_publish_aggregated, 2); + write_err_chart( + NETDATA_VFS_FILE_ERR_COUNT, NETDATA_EBPF_FAMILY, &process_publish_aggregated[2], NETDATA_VFS_ERRORS); + write_err_chart( + NETDATA_PROCESS_ERROR_NAME, NETDATA_EBPF_FAMILY, &process_publish_aggregated[NETDATA_PROCESS_START], 2); + } + + write_io_chart(NETDATA_VFS_IO_FILE_BYTES, NETDATA_EBPF_FAMILY, + process_id_names[NETDATA_KEY_PUBLISH_PROCESS_WRITE], (long long) pvc.write, + process_id_names[NETDATA_KEY_PUBLISH_PROCESS_READ], (long long)pvc.read); +} + +/** + * Sum values for pid + * + * @param root the structure with all available PIDs + * + * @param offset the address that we are reading + * + * @return it returns the sum of all PIDs + */ +long long ebpf_process_sum_values_for_pids(struct pid_on_target *root, size_t offset) +{ + long long ret = 0; + while (root) { + int32_t pid = root->pid; + ebpf_process_publish_apps_t *w = current_apps_data[pid]; + if (w) { + ret += get_value_from_structure((char *)w, offset); + } + + root = root->next; + } + + return ret; +} + +/** + * Remove process pid + * + * Remove from PID task table when task_release was called. + */ +void ebpf_process_remove_pids() +{ + struct pid_stat *pids = root_of_pids; + int pid_fd = map_fd[0]; + while (pids) { + uint32_t pid = pids->pid; + ebpf_process_stat_t *w = global_process_stats[pid]; + if (w) { + if (w->removeme) { + freez(w); + global_process_stats[pid] = NULL; + bpf_map_delete_elem(pid_fd, &pid); + } + } + + pids = pids->next; + } +} + +/** + * Send data to Netdata calling auxiliar functions. + * + * @param em the structure with thread information + * @param root the target list. + */ +void ebpf_process_send_apps_data(ebpf_module_t *em, struct target *root) +{ + struct target *w; + collected_number value; + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_FILE_OPEN); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids(w->root_pid, offsetof(ebpf_process_publish_apps_t, call_sys_open)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + if (em->mode < MODE_ENTRY) { + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_FILE_OPEN_ERROR); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids( + w->root_pid, offsetof(ebpf_process_publish_apps_t, ecall_sys_open)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + } + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_FILE_CLOSED); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = + ebpf_process_sum_values_for_pids(w->root_pid, offsetof(ebpf_process_publish_apps_t, call_close_fd)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + if (em->mode < MODE_ENTRY) { + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_FILE_CLOSE_ERROR); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids( + w->root_pid, offsetof(ebpf_process_publish_apps_t, ecall_close_fd)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + } + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_FILE_DELETED); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = + ebpf_process_sum_values_for_pids(w->root_pid, offsetof(ebpf_process_publish_apps_t, call_vfs_unlink)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_VFS_WRITE_CALLS); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids( + w->root_pid, offsetof(ebpf_process_publish_apps_t, call_write)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + if (em->mode < MODE_ENTRY) { + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_VFS_WRITE_CALLS_ERROR); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids( + w->root_pid, offsetof(ebpf_process_publish_apps_t, ecall_write)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + } + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_VFS_READ_CALLS); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = + ebpf_process_sum_values_for_pids(w->root_pid, offsetof(ebpf_process_publish_apps_t, call_read)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + if (em->mode < MODE_ENTRY) { + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_VFS_READ_CALLS_ERROR); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids( + w->root_pid, offsetof(ebpf_process_publish_apps_t, ecall_read)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + } + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_VFS_WRITE_BYTES); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids( + w->root_pid, offsetof(ebpf_process_publish_apps_t, bytes_written)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_VFS_READ_BYTES); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids( + w->root_pid, offsetof(ebpf_process_publish_apps_t, bytes_read)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_TASK_PROCESS); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = + ebpf_process_sum_values_for_pids(w->root_pid, offsetof(ebpf_process_publish_apps_t, call_do_fork)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_TASK_THREAD); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = + ebpf_process_sum_values_for_pids(w->root_pid, offsetof(ebpf_process_publish_apps_t, call_sys_clone)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_SYSCALL_APPS_TASK_CLOSE); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_process_sum_values_for_pids(w->root_pid, offsetof(ebpf_process_publish_apps_t, + call_release_task)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + ebpf_process_remove_pids(); +} + +/***************************************************************** + * + * READ INFORMATION FROM KERNEL RING + * + *****************************************************************/ + +/** + * Read the hash table and store data to allocated vectors. + */ +static void read_hash_global_tables() +{ + uint64_t idx; + netdata_idx_t res[NETDATA_GLOBAL_VECTOR]; + + netdata_idx_t *val = process_hash_values; + for (idx = 0; idx < NETDATA_GLOBAL_VECTOR; idx++) { + if (!bpf_map_lookup_elem(map_fd[1], &idx, val)) { + uint64_t total = 0; + int i; + int end = (running_on_kernel < NETDATA_KERNEL_V4_15) ? 1 : ebpf_nprocs; + for (i = 0; i < end; i++) + total += val[i]; + + res[idx] = total; + } else { + res[idx] = 0; + } + } + + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_OPEN].call = res[NETDATA_KEY_CALLS_DO_SYS_OPEN]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_CLOSE].call = res[NETDATA_KEY_CALLS_CLOSE_FD]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_UNLINK].call = res[NETDATA_KEY_CALLS_VFS_UNLINK]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_READ].call = res[NETDATA_KEY_CALLS_VFS_READ] + res[NETDATA_KEY_CALLS_VFS_READV]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_WRITE].call = res[NETDATA_KEY_CALLS_VFS_WRITE] + res[NETDATA_KEY_CALLS_VFS_WRITEV]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_EXIT].call = res[NETDATA_KEY_CALLS_DO_EXIT]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_RELEASE_TASK].call = res[NETDATA_KEY_CALLS_RELEASE_TASK]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_FORK].call = res[NETDATA_KEY_CALLS_DO_FORK]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_CLONE].call = res[NETDATA_KEY_CALLS_SYS_CLONE]; + + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_OPEN].ecall = res[NETDATA_KEY_ERROR_DO_SYS_OPEN]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_CLOSE].ecall = res[NETDATA_KEY_ERROR_CLOSE_FD]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_UNLINK].ecall = res[NETDATA_KEY_ERROR_VFS_UNLINK]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_READ].ecall = res[NETDATA_KEY_ERROR_VFS_READ] + res[NETDATA_KEY_ERROR_VFS_READV]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_WRITE].ecall = res[NETDATA_KEY_ERROR_VFS_WRITE] + res[NETDATA_KEY_ERROR_VFS_WRITEV]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_FORK].ecall = res[NETDATA_KEY_ERROR_DO_FORK]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_CLONE].ecall = res[NETDATA_KEY_ERROR_SYS_CLONE]; + + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_WRITE].bytes = (uint64_t)res[NETDATA_KEY_BYTES_VFS_WRITE] + + (uint64_t)res[NETDATA_KEY_BYTES_VFS_WRITEV]; + process_aggregated_data[NETDATA_KEY_PUBLISH_PROCESS_READ].bytes = (uint64_t)res[NETDATA_KEY_BYTES_VFS_READ] + + (uint64_t)res[NETDATA_KEY_BYTES_VFS_READV]; +} + +/** + * Read the hash table and store data to allocated vectors. + */ +static void ebpf_process_update_apps_data() +{ + struct pid_stat *pids = root_of_pids; + while (pids) { + uint32_t current_pid = pids->pid; + ebpf_process_stat_t *ps = global_process_stats[current_pid]; + if (!ps) { + pids = pids->next; + continue; + } + + ebpf_process_publish_apps_t *cad = current_apps_data[current_pid]; + if (!cad) { + cad = callocz(1, sizeof(ebpf_process_publish_apps_t)); + current_apps_data[current_pid] = cad; + } + + //Read data + cad->call_sys_open = ps->open_call; + cad->call_close_fd = ps->close_call; + cad->call_vfs_unlink = ps->unlink_call; + cad->call_read = ps->read_call + ps->readv_call; + cad->call_write = ps->write_call + ps->writev_call; + cad->call_do_exit = ps->exit_call; + cad->call_release_task = ps->release_call; + cad->call_do_fork = ps->fork_call; + cad->call_sys_clone = ps->clone_call; + + cad->ecall_sys_open = ps->open_err; + cad->ecall_close_fd = ps->close_err; + cad->ecall_vfs_unlink = ps->unlink_err; + cad->ecall_read = ps->read_err + ps->readv_err; + cad->ecall_write = ps->write_err + ps->writev_err; + cad->ecall_do_fork = ps->fork_err; + cad->ecall_sys_clone = ps->clone_err; + + cad->bytes_written = (uint64_t)ps->write_bytes + (uint64_t)ps->write_bytes; + cad->bytes_read = (uint64_t)ps->read_bytes + (uint64_t)ps->readv_bytes; + + pids = pids->next; + } +} + +/***************************************************************** + * + * FUNCTIONS TO CREATE CHARTS + * + *****************************************************************/ + +/** + * Create IO chart + * + * @param family the chart family + * @param name the chart name + * @param axis the axis label + * @param web the group name used to attach the chart on dashaboard + * @param order the order number of the specified chart + * @param algorithm the algorithm used to make the charts. + */ +static void ebpf_create_io_chart(char *family, char *name, char *axis, char *web, int order, int algorithm) +{ + printf("CHART %s.%s '' 'Bytes written and read' '%s' '%s' '' line %d %d\n", + family, + name, + axis, + web, + order, + update_every); + + printf("DIMENSION %s %s %s 1 1\n", + process_id_names[NETDATA_KEY_PUBLISH_PROCESS_READ], + process_dimension_names[NETDATA_KEY_PUBLISH_PROCESS_READ], + ebpf_algorithms[algorithm]); + printf("DIMENSION %s %s %s 1 1\n", + process_id_names[NETDATA_KEY_PUBLISH_PROCESS_WRITE], + process_dimension_names[NETDATA_KEY_PUBLISH_PROCESS_WRITE], + ebpf_algorithms[algorithm]); +} + +/** + * Create process status chart + * + * @param family the chart family + * @param name the chart name + * @param axis the axis label + * @param web the group name used to attach the chart on dashaboard + * @param order the order number of the specified chart + */ +static void ebpf_process_status_chart(char *family, char *name, char *axis, + char *web, char *algorithm, int order) +{ + printf("CHART %s.%s '' 'Process not closed' '%s' '%s' '' line %d %d ''\n", + family, + name, + axis, + web, + order, + update_every); + + printf("DIMENSION %s '' %s 1 1\n", status[0], algorithm); + printf("DIMENSION %s '' %s 1 1\n", status[1], algorithm); +} + +/** + * Create global charts + * + * Call ebpf_create_chart to create the charts for the collector. + * + * @param em a pointer to the structure with the default values. + */ +static void ebpf_create_global_charts(ebpf_module_t *em) +{ + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_FILE_OPEN_CLOSE_COUNT, + "Open and close calls", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_FILE_GROUP, + 21000, + ebpf_create_global_dimension, + process_publish_aggregated, + 2); + + if (em->mode < MODE_ENTRY) { + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_FILE_OPEN_ERR_COUNT, + "Open fails", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_FILE_GROUP, + 21001, + ebpf_create_global_dimension, + process_publish_aggregated, + 2); + } + + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_VFS_FILE_CLEAN_COUNT, + "Remove files", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_VFS_GROUP, + 21002, + ebpf_create_global_dimension, + &process_publish_aggregated[NETDATA_DEL_START], + 1); + + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_VFS_FILE_IO_COUNT, + "Calls to IO", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_VFS_GROUP, + 21003, + ebpf_create_global_dimension, + &process_publish_aggregated[NETDATA_IN_START_BYTE], + 2); + + ebpf_create_io_chart(NETDATA_EBPF_FAMILY, + NETDATA_VFS_IO_FILE_BYTES, EBPF_COMMON_DIMENSION_BYTES, + NETDATA_VFS_GROUP, + 21004, + NETDATA_EBPF_ABSOLUTE_IDX); + + if (em->mode < MODE_ENTRY) { + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_VFS_FILE_ERR_COUNT, + "Fails to write or read", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_VFS_GROUP, + 21005, + ebpf_create_global_dimension, + &process_publish_aggregated[2], + NETDATA_VFS_ERRORS); + } + + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_PROCESS_SYSCALL, + "Start process", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_PROCESS_GROUP, + 21006, + ebpf_create_global_dimension, + &process_publish_aggregated[NETDATA_PROCESS_START], + 2); + + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_EXIT_SYSCALL, + "Exit process", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_PROCESS_GROUP, + 21007, + ebpf_create_global_dimension, + &process_publish_aggregated[NETDATA_EXIT_START], + 2); + + ebpf_process_status_chart(NETDATA_EBPF_FAMILY, + NETDATA_PROCESS_STATUS_NAME, + EBPF_COMMON_DIMENSION_DIFFERENCE, + NETDATA_PROCESS_GROUP, + ebpf_algorithms[NETDATA_EBPF_ABSOLUTE_IDX], + 21008); + + if (em->mode < MODE_ENTRY) { + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_PROCESS_ERROR_NAME, + "Fails to create process", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_PROCESS_GROUP, + 21009, + ebpf_create_global_dimension, + &process_publish_aggregated[NETDATA_PROCESS_START], + 2); + } +} + +/** + * Create process apps charts + * + * Call ebpf_create_chart to create the charts on apps submenu. + * + * @param em a pointer to the structure with the default values. + * @param root a pointer for the targets. + */ +static void ebpf_process_create_apps_charts(ebpf_module_t *em, struct target *root) +{ + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_FILE_OPEN, + "Number of open files", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_FILE_GROUP, + 20061, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + if (em->mode < MODE_ENTRY) { + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_FILE_OPEN_ERROR, + "Fails to open files", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_FILE_GROUP, + 20062, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + } + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_FILE_CLOSED, + "Files closed", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_FILE_GROUP, + 20063, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + if (em->mode < MODE_ENTRY) { + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_FILE_CLOSE_ERROR, + "Fails to close files", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_FILE_GROUP, + 20064, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + } + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_FILE_DELETED, + "Files deleted", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_VFS_GROUP, + 20065, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_VFS_WRITE_CALLS, + "Write to disk", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_VFS_GROUP, + 20066, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + apps_groups_root_target); + + if (em->mode < MODE_ENTRY) { + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_VFS_WRITE_CALLS_ERROR, + "Fails to write", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_VFS_GROUP, + 20067, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + } + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_VFS_READ_CALLS, + "Read from disk", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_VFS_GROUP, + 20068, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + if (em->mode < MODE_ENTRY) { + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_VFS_READ_CALLS_ERROR, + "Fails to read", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_VFS_GROUP, + 20069, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + } + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_VFS_WRITE_BYTES, + "Bytes written on disk", EBPF_COMMON_DIMENSION_BYTES, + NETDATA_APPS_VFS_GROUP, + 20070, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_VFS_READ_BYTES, + "Bytes read from disk", EBPF_COMMON_DIMENSION_BYTES, + NETDATA_APPS_VFS_GROUP, + 20071, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_TASK_PROCESS, + "Process started", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_PROCESS_GROUP, + 20072, + ebpf_algorithms[NETDATA_EBPF_ABSOLUTE_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_TASK_THREAD, + "Threads started", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_PROCESS_GROUP, + 20073, + ebpf_algorithms[NETDATA_EBPF_ABSOLUTE_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_SYSCALL_APPS_TASK_CLOSE, + "Tasks closed", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_PROCESS_GROUP, + 20074, + ebpf_algorithms[NETDATA_EBPF_ABSOLUTE_IDX], + root); +} + +/** + * Create apps charts + * + * Call ebpf_create_chart to create the charts on apps submenu. + * + * @param em a pointer to the structure with the default values. + * @param root a pointer for the targets. + */ +static void ebpf_create_apps_charts(ebpf_module_t *em, struct target *root) +{ + struct target *w; + int newly_added = 0; + + for (w = root; w; w = w->next) { + if (w->target) + continue; + + if (unlikely(w->processes && (debug_enabled || w->debug_enabled))) { + struct pid_on_target *pid_on_target; + + fprintf( + stderr, "ebpf.plugin: target '%s' has aggregated %u process%s:", w->name, w->processes, + (w->processes == 1) ? "" : "es"); + + for (pid_on_target = w->root_pid; pid_on_target; pid_on_target = pid_on_target->next) { + fprintf(stderr, " %d", pid_on_target->pid); + } + + fputc('\n', stderr); + } + + if (!w->exposed && w->processes) { + newly_added++; + w->exposed = 1; + if (debug_enabled || w->debug_enabled) + debug_log_int("%s just added - regenerating charts.", w->name); + } + } + + if (!newly_added) + return; + + if (ebpf_modules[EBPF_MODULE_PROCESS_IDX].apps_charts) + ebpf_process_create_apps_charts(em, root); + + if (ebpf_modules[EBPF_MODULE_SOCKET_IDX].apps_charts) + ebpf_socket_create_apps_charts(NULL, root); +} + +/***************************************************************** + * + * FUNCTIONS WITH THE MAIN LOOP + * + *****************************************************************/ + +/** + * Main loop for this collector. + * + * @param step the number of microseconds used with heart beat + * @param em the structure with thread information + */ +static void process_collector(usec_t step, ebpf_module_t *em) +{ + heartbeat_t hb; + heartbeat_init(&hb); + int publish_global = em->global_charts; + int apps_enabled = em->apps_charts; + int pid_fd = map_fd[0]; + while (!close_ebpf_plugin) { + usec_t dt = heartbeat_next(&hb, step); + (void)dt; + + read_hash_global_tables(); + + pthread_mutex_lock(&collect_data_mutex); + cleanup_exited_pids(); + collect_data_for_all_processes(pid_fd); + + ebpf_create_apps_charts(em, apps_groups_root_target); + + pthread_cond_broadcast(&collect_data_cond_var); + pthread_mutex_unlock(&collect_data_mutex); + + int publish_apps = 0; + if (apps_enabled && all_pids_count > 0) { + publish_apps = 1; + ebpf_process_update_apps_data(); + } + + pthread_mutex_lock(&lock); + if (publish_global) { + ebpf_process_send_data(em); + } + + if (publish_apps) { + ebpf_process_send_apps_data(em, apps_groups_root_target); + } + pthread_mutex_unlock(&lock); + + fflush(stdout); + } +} + +/***************************************************************** + * + * FUNCTIONS TO CLOSE THE THREAD + * + *****************************************************************/ + +void clean_global_memory() { + int pid_fd = map_fd[0]; + struct pid_stat *pids = root_of_pids; + while (pids) { + uint32_t pid = pids->pid; + freez(global_process_stats[pid]); + + bpf_map_delete_elem(pid_fd, &pid); + freez(current_apps_data[pid]); + + pids = pids->next; + } +} + +void clean_pid_on_target(struct pid_on_target *ptr) { + while (ptr) { + struct pid_on_target *next = ptr->next; + freez(ptr); + + ptr = next; + } +} + +void clean_apps_structures(struct target *ptr) { + struct target *agdt = ptr; + while (agdt) { + struct target *next = agdt->next; + clean_pid_on_target(agdt->root_pid); + freez(agdt); + + agdt = next; + } +} + +/** + * Clean up the main thread. + * + * @param ptr thread data. + */ +static void ebpf_process_cleanup(void *ptr) +{ + UNUSED(ptr); + + heartbeat_t hb; + heartbeat_init(&hb); + uint32_t tick = 50*USEC_PER_MS; + while (!finalized_threads) { + usec_t dt = heartbeat_next(&hb, tick); + UNUSED(dt); + } + + freez(process_aggregated_data); + ebpf_cleanup_publish_syscall(process_publish_aggregated); + freez(process_publish_aggregated); + freez(process_hash_values); + + clean_global_memory(); + freez(global_process_stats); + freez(current_apps_data); + + clean_apps_structures(apps_groups_root_target); + freez(process_data.map_fd); + + struct bpf_program *prog; + size_t i = 0 ; + bpf_object__for_each_program(prog, objects) { + bpf_link__destroy(probe_links[i]); + i++; + } + bpf_object__close(objects); +} + +/***************************************************************** + * + * FUNCTIONS TO START THREAD + * + *****************************************************************/ + +/** + * Allocate vectors used with this thread. + * We are not testing the return, because callocz does this and shutdown the software + * case it was not possible to allocate. + * + * @param length is the length for the vectors used inside the collector. + */ +static void ebpf_process_allocate_global_vectors(size_t length) +{ + process_aggregated_data = callocz(length, sizeof(netdata_syscall_stat_t)); + process_publish_aggregated = callocz(length, sizeof(netdata_publish_syscall_t)); + process_hash_values = callocz(ebpf_nprocs, sizeof(netdata_idx_t)); + + global_process_stats = callocz((size_t)pid_max, sizeof(ebpf_process_stat_t *)); + current_apps_data = callocz((size_t)pid_max, sizeof(ebpf_process_publish_apps_t *)); +} + +static void change_syscalls() +{ + static char *lfork = { "do_fork" }; + process_id_names[7] = lfork; +} + +/** + * Set local variables + * + */ +static void set_local_pointers() +{ + map_fd = process_data.map_fd; + + if (process_data.isrh >= NETDATA_MINIMUM_RH_VERSION && process_data.isrh < NETDATA_RH_8) + change_syscalls(); +} + +/***************************************************************** + * + * EBPF PROCESS THREAD + * + *****************************************************************/ + +/** + * + */ +static void wait_for_all_threads_die() +{ + ebpf_modules[EBPF_MODULE_PROCESS_IDX].enabled = 0; + + heartbeat_t hb; + heartbeat_init(&hb); + + int max = 10; + int i; + for (i = 0; i < max; i++) { + heartbeat_next(&hb, 200000); + + size_t j, counter = 0, compare = 0; + for (j = 0; ebpf_modules[j].thread_name; j++) { + if (!ebpf_modules[j].enabled) + counter++; + + compare++; + } + + if (counter == compare) + break; + } +} + +/** + * Process thread + * + * Thread used to generate process charts. + * + * @param ptr a pointer to `struct ebpf_module` + * + * @return It always return NULL + */ +void *ebpf_process_thread(void *ptr) +{ + netdata_thread_cleanup_push(ebpf_process_cleanup, ptr); + + ebpf_module_t *em = (ebpf_module_t *)ptr; + process_enabled = em->enabled; + fill_ebpf_data(&process_data); + + pthread_mutex_lock(&lock); + ebpf_process_allocate_global_vectors(NETDATA_MAX_MONITOR_VECTOR); + + if (ebpf_update_kernel(&process_data)) { + pthread_mutex_unlock(&lock); + goto endprocess; + } + + set_local_pointers(); + probe_links = ebpf_load_program(ebpf_plugin_dir, em, kernel_string, &objects, process_data.map_fd); + if (!probe_links) { + pthread_mutex_unlock(&lock); + goto endprocess; + } + + int algorithms[NETDATA_KEY_PUBLISH_PROCESS_END] = { + NETDATA_EBPF_INCREMENTAL_IDX, NETDATA_EBPF_INCREMENTAL_IDX,NETDATA_EBPF_INCREMENTAL_IDX, //open, close, unlink + NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX, + NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX + }; + + ebpf_global_labels( + process_aggregated_data, process_publish_aggregated, process_dimension_names, process_id_names, + algorithms, NETDATA_MAX_MONITOR_VECTOR); + + if (process_enabled) { + ebpf_create_global_charts(em); + } + + pthread_mutex_unlock(&lock); + + process_collector((usec_t)(em->update_time * USEC_PER_SEC), em); + +endprocess: + wait_for_all_threads_die(); + netdata_thread_cleanup_pop(1); + return NULL; +} diff --git a/collectors/ebpf.plugin/ebpf_process.h b/collectors/ebpf.plugin/ebpf_process.h new file mode 100644 index 0000000..aa6ed66 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf_process.h @@ -0,0 +1,139 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#ifndef NETDATA_EBPF_PROCESS_H +#define NETDATA_EBPF_PROCESS_H 1 + +// Groups used on Dashboard +#define NETDATA_FILE_GROUP "File" +#define NETDATA_VFS_GROUP "VFS" +#define NETDATA_PROCESS_GROUP "Process" + +// Internal constants +#define NETDATA_GLOBAL_VECTOR 24 +#define NETDATA_MAX_MONITOR_VECTOR 9 +#define NETDATA_VFS_ERRORS 3 + +// Map index +#define NETDATA_DEL_START 2 +#define NETDATA_IN_START_BYTE 3 +#define NETDATA_EXIT_START 5 +#define NETDATA_PROCESS_START 7 + +// Global chart name +#define NETDATA_FILE_OPEN_CLOSE_COUNT "file_descriptor" +#define NETDATA_FILE_OPEN_ERR_COUNT "file_error" +#define NETDATA_VFS_FILE_CLEAN_COUNT "deleted_objects" +#define NETDATA_VFS_FILE_IO_COUNT "io" +#define NETDATA_VFS_FILE_ERR_COUNT "io_error" + +#define NETDATA_EXIT_SYSCALL "exit" +#define NETDATA_PROCESS_SYSCALL "process_thread" +#define NETDATA_PROCESS_ERROR_NAME "task_error" +#define NETDATA_PROCESS_STATUS_NAME "process_status" + +#define NETDATA_VFS_IO_FILE_BYTES "io_bytes" + +// Charts created on Apps submenu +#define NETDATA_SYSCALL_APPS_FILE_OPEN "file_open" +#define NETDATA_SYSCALL_APPS_FILE_CLOSED "file_closed" +#define NETDATA_SYSCALL_APPS_FILE_DELETED "file_deleted" +#define NETDATA_SYSCALL_APPS_VFS_WRITE_CALLS "vfs_write_call" +#define NETDATA_SYSCALL_APPS_VFS_READ_CALLS "vfs_read_call" +#define NETDATA_SYSCALL_APPS_VFS_WRITE_BYTES "vfs_write_bytes" +#define NETDATA_SYSCALL_APPS_VFS_READ_BYTES "vfs_read_bytes" +#define NETDATA_SYSCALL_APPS_TASK_PROCESS "process_create" +#define NETDATA_SYSCALL_APPS_TASK_THREAD "thread_create" +#define NETDATA_SYSCALL_APPS_TASK_CLOSE "task_close" + +// Charts created on Apps submenu, if and only if, the return mode is active + +#define NETDATA_SYSCALL_APPS_FILE_OPEN_ERROR "file_open_error" +#define NETDATA_SYSCALL_APPS_FILE_CLOSE_ERROR "file_close_error" +#define NETDATA_SYSCALL_APPS_VFS_WRITE_CALLS_ERROR "vfs_write_error" +#define NETDATA_SYSCALL_APPS_VFS_READ_CALLS_ERROR "vfs_read_error" + +// Index from kernel +typedef enum ebpf_process_index { + NETDATA_KEY_CALLS_DO_SYS_OPEN, + NETDATA_KEY_ERROR_DO_SYS_OPEN, + + NETDATA_KEY_CALLS_VFS_WRITE, + NETDATA_KEY_ERROR_VFS_WRITE, + NETDATA_KEY_BYTES_VFS_WRITE, + + NETDATA_KEY_CALLS_VFS_READ, + NETDATA_KEY_ERROR_VFS_READ, + NETDATA_KEY_BYTES_VFS_READ, + + NETDATA_KEY_CALLS_VFS_UNLINK, + NETDATA_KEY_ERROR_VFS_UNLINK, + + NETDATA_KEY_CALLS_DO_EXIT, + + NETDATA_KEY_CALLS_RELEASE_TASK, + + NETDATA_KEY_CALLS_DO_FORK, + NETDATA_KEY_ERROR_DO_FORK, + + NETDATA_KEY_CALLS_CLOSE_FD, + NETDATA_KEY_ERROR_CLOSE_FD, + + NETDATA_KEY_CALLS_SYS_CLONE, + NETDATA_KEY_ERROR_SYS_CLONE, + + NETDATA_KEY_CALLS_VFS_WRITEV, + NETDATA_KEY_ERROR_VFS_WRITEV, + NETDATA_KEY_BYTES_VFS_WRITEV, + + NETDATA_KEY_CALLS_VFS_READV, + NETDATA_KEY_ERROR_VFS_READV, + NETDATA_KEY_BYTES_VFS_READV + +} ebpf_process_index_t; + +// This enum acts as an index for publish vector. +// Do not change the enum order because we use +// different algorithms to make charts with incremental +// values (the three initial positions) and absolute values +// (the remaining charts). +typedef enum netdata_publish_process { + NETDATA_KEY_PUBLISH_PROCESS_OPEN, + NETDATA_KEY_PUBLISH_PROCESS_CLOSE, + NETDATA_KEY_PUBLISH_PROCESS_UNLINK, + NETDATA_KEY_PUBLISH_PROCESS_READ, + NETDATA_KEY_PUBLISH_PROCESS_WRITE, + NETDATA_KEY_PUBLISH_PROCESS_EXIT, + NETDATA_KEY_PUBLISH_PROCESS_RELEASE_TASK, + NETDATA_KEY_PUBLISH_PROCESS_FORK, + NETDATA_KEY_PUBLISH_PROCESS_CLONE, + + NETDATA_KEY_PUBLISH_PROCESS_END +} netdata_publish_process_t; + +typedef struct ebpf_process_publish_apps { + // Number of calls during the last read + uint64_t call_sys_open; + uint64_t call_close_fd; + uint64_t call_vfs_unlink; + uint64_t call_read; + uint64_t call_write; + uint64_t call_do_exit; + uint64_t call_release_task; + uint64_t call_do_fork; + uint64_t call_sys_clone; + + // Number of errors during the last read + uint64_t ecall_sys_open; + uint64_t ecall_close_fd; + uint64_t ecall_vfs_unlink; + uint64_t ecall_read; + uint64_t ecall_write; + uint64_t ecall_do_fork; + uint64_t ecall_sys_clone; + + // Number of bytes during the last read + uint64_t bytes_written; + uint64_t bytes_read; +} ebpf_process_publish_apps_t; + +#endif /* NETDATA_EBPF_PROCESS_H */ diff --git a/collectors/ebpf.plugin/ebpf_socket.c b/collectors/ebpf.plugin/ebpf_socket.c new file mode 100644 index 0000000..7fbc244 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf_socket.c @@ -0,0 +1,1920 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#include <sys/resource.h> + +#include "ebpf.h" +#include "ebpf_socket.h" + +/***************************************************************** + * + * GLOBAL VARIABLES + * + *****************************************************************/ + +static char *socket_dimension_names[NETDATA_MAX_SOCKET_VECTOR] = { "sent", "received", "close", "sent", + "received", "retransmitted" }; +static char *socket_id_names[NETDATA_MAX_SOCKET_VECTOR] = { "tcp_sendmsg", "tcp_cleanup_rbuf", "tcp_close", + "udp_sendmsg", "udp_recvmsg", "tcp_retransmit_skb" }; + +static netdata_idx_t *socket_hash_values = NULL; +static netdata_syscall_stat_t *socket_aggregated_data = NULL; +static netdata_publish_syscall_t *socket_publish_aggregated = NULL; + +static ebpf_data_t socket_data; + +ebpf_socket_publish_apps_t **socket_bandwidth_curr = NULL; +static ebpf_bandwidth_t *bandwidth_vector = NULL; + +static int socket_apps_created = 0; +pthread_mutex_t nv_mutex; +int wait_to_plot = 0; +int read_thread_closed = 1; + +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; + +ebpf_network_viewer_port_list_t *listen_ports = NULL; + +static int *map_fd = NULL; +static struct bpf_object *objects = NULL; +static struct bpf_link **probe_links = NULL; + +/***************************************************************** + * + * PROCESS DATA AND SEND TO NETDATA + * + *****************************************************************/ + +/** + * Update publish structure before to send data to Netdata. + * + * @param publish the first output structure with independent dimensions + * @param tcp structure to store IO from tcp sockets + * @param udp structure to store IO from udp sockets + * @param input the structure with the input data. + */ +static void ebpf_update_global_publish( + netdata_publish_syscall_t *publish, netdata_publish_vfs_common_t *tcp, netdata_publish_vfs_common_t *udp, + netdata_syscall_stat_t *input) +{ + netdata_publish_syscall_t *move = publish; + while (move) { + if (input->call != move->pcall) { + // This condition happens to avoid initial values with dimensions higher than normal values. + if (move->pcall) { + move->ncall = (input->call > move->pcall) ? input->call - move->pcall : move->pcall - input->call; + move->nbyte = (input->bytes > move->pbyte) ? input->bytes - move->pbyte : move->pbyte - input->bytes; + move->nerr = (input->ecall > move->nerr) ? input->ecall - move->perr : move->perr - input->ecall; + } else { + move->ncall = 0; + move->nbyte = 0; + move->nerr = 0; + } + + move->pcall = input->call; + move->pbyte = input->bytes; + move->perr = input->ecall; + } else { + move->ncall = 0; + move->nbyte = 0; + move->nerr = 0; + } + + input = input->next; + move = move->next; + } + + tcp->write = -(long)publish[0].nbyte; + tcp->read = (long)publish[1].nbyte; + + udp->write = -(long)publish[3].nbyte; + udp->read = (long)publish[4].nbyte; +} + +/** + * Update Network Viewer plot data + * + * @param plot the structure where the data will be stored + * @param sock the last update from the socket + */ +static inline void update_nv_plot_data(netdata_plot_values_t *plot, netdata_socket_t *sock) +{ + 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; + plot->plot_recv_bytes = sock->recv_bytes; + plot->plot_sent_bytes = sock->sent_bytes; + plot->plot_retransmit = sock->retransmit; + } + + sock->recv_packets = 0; + sock->sent_packets = 0; + sock->recv_bytes = 0; + sock->sent_bytes = 0; + sock->retransmit = 0; +} + +/** + * Calculate Network Viewer Plot + * + * Do math with collected values before to plot data. + */ +static inline void calculate_nv_plot() +{ + uint32_t i; + uint32_t end = inbound_vectors.next; + for (i = 0; i < end; i++) { + update_nv_plot_data(&inbound_vectors.plot[i].plot, &inbound_vectors.plot[i].sock); + } + inbound_vectors.max_plot = end; + + // The 'Other' dimension is always calculated for the chart to have at least one dimension + update_nv_plot_data(&inbound_vectors.plot[inbound_vectors.last].plot, + &inbound_vectors.plot[inbound_vectors.last].sock); + + end = outbound_vectors.next; + for (i = 0; i < end; i++) { + update_nv_plot_data(&outbound_vectors.plot[i].plot, &outbound_vectors.plot[i].sock); + } + 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); +} + +/** + * Network viewer send bytes + * + * @param ptr the structure with values to plot + * @param chart the chart name. + */ +static inline void ebpf_socket_nv_send_bytes(netdata_vector_plot_t *ptr, char *chart) +{ + uint32_t i; + uint32_t end = ptr->last_plot; + netdata_socket_plot_t *w = ptr->plot; + collected_number value; + + write_begin_chart(NETDATA_EBPF_FAMILY, chart); + for (i = 0; i < end; i++) { + value = ((collected_number) w[i].plot.plot_sent_bytes); + write_chart_dimension(w[i].dimension_sent, value); + value = (collected_number) w[i].plot.plot_recv_bytes; + write_chart_dimension(w[i].dimension_recv, value); + } + + i = ptr->last; + value = ((collected_number) w[i].plot.plot_sent_bytes); + write_chart_dimension(w[i].dimension_sent, value); + value = (collected_number) w[i].plot.plot_recv_bytes; + write_chart_dimension(w[i].dimension_recv, value); + write_end_chart(); +} + +/** + * Network Viewer Send packets + * + * @param ptr the structure with values to plot + * @param chart the chart name. + */ +static inline void ebpf_socket_nv_send_packets(netdata_vector_plot_t *ptr, char *chart) +{ + uint32_t i; + uint32_t end = ptr->last_plot; + netdata_socket_plot_t *w = ptr->plot; + collected_number value; + + write_begin_chart(NETDATA_EBPF_FAMILY, chart); + for (i = 0; i < end; i++) { + value = ((collected_number)w[i].plot.plot_sent_packets); + write_chart_dimension(w[i].dimension_sent, value); + value = (collected_number) w[i].plot.plot_recv_packets; + write_chart_dimension(w[i].dimension_recv, value); + } + + i = ptr->last; + value = ((collected_number)w[i].plot.plot_sent_packets); + write_chart_dimension(w[i].dimension_sent, value); + value = (collected_number)w[i].plot.plot_recv_packets; + write_chart_dimension(w[i].dimension_recv, value); + write_end_chart(); +} + +/** + * Network Viewer Send Retransmit + * + * @param ptr the structure with values to plot + * @param chart the chart name. + */ +static inline void ebpf_socket_nv_send_retransmit(netdata_vector_plot_t *ptr, char *chart) +{ + uint32_t i; + uint32_t end = ptr->last_plot; + netdata_socket_plot_t *w = ptr->plot; + collected_number value; + + write_begin_chart(NETDATA_EBPF_FAMILY, chart); + for (i = 0; i < end; i++) { + value = (collected_number) w[i].plot.plot_retransmit; + write_chart_dimension(w[i].dimension_retransmit, value); + } + + i = ptr->last; + value = (collected_number)w[i].plot.plot_retransmit; + write_chart_dimension(w[i].dimension_retransmit, value); + write_end_chart(); +} + +/** + * Send network viewer data + * + * @param ptr the pointer to plot data + */ +static void ebpf_socket_send_nv_data(netdata_vector_plot_t *ptr) +{ + if (!ptr->flags) + return; + + if (ptr == (netdata_vector_plot_t *)&outbound_vectors) { + ebpf_socket_nv_send_bytes(ptr, NETDATA_NV_OUTBOUND_BYTES); + fflush(stdout); + + ebpf_socket_nv_send_packets(ptr, NETDATA_NV_OUTBOUND_PACKETS); + fflush(stdout); + + ebpf_socket_nv_send_retransmit(ptr, NETDATA_NV_OUTBOUND_RETRANSMIT); + fflush(stdout); + } else { + ebpf_socket_nv_send_bytes(ptr, NETDATA_NV_INBOUND_BYTES); + fflush(stdout); + + ebpf_socket_nv_send_packets(ptr, NETDATA_NV_INBOUND_PACKETS); + fflush(stdout); + } +} + +/** + * Send data to Netdata calling auxiliar functions. + * + * @param em the structure with thread information + */ +static void ebpf_socket_send_data(ebpf_module_t *em) +{ + netdata_publish_vfs_common_t common_tcp; + netdata_publish_vfs_common_t common_udp; + ebpf_update_global_publish(socket_publish_aggregated, &common_tcp, &common_udp, socket_aggregated_data); + + // We read bytes from function arguments, but bandiwdth is given in bits, + // so we need to multiply by 8 to convert for the final value. + write_count_chart( + NETDATA_TCP_FUNCTION_COUNT, NETDATA_EBPF_FAMILY, socket_publish_aggregated, 3); + write_io_chart( + NETDATA_TCP_FUNCTION_BITS, NETDATA_EBPF_FAMILY, socket_id_names[0], common_tcp.write*8/1000, + socket_id_names[1], common_tcp.read*8/1000); + if (em->mode < MODE_ENTRY) { + write_err_chart( + NETDATA_TCP_FUNCTION_ERROR, NETDATA_EBPF_FAMILY, socket_publish_aggregated, 2); + } + write_count_chart( + NETDATA_TCP_RETRANSMIT, NETDATA_EBPF_FAMILY, &socket_publish_aggregated[NETDATA_RETRANSMIT_START], 1); + + write_count_chart( + NETDATA_UDP_FUNCTION_COUNT, NETDATA_EBPF_FAMILY, &socket_publish_aggregated[NETDATA_UDP_START], 2); + write_io_chart( + NETDATA_UDP_FUNCTION_BITS, NETDATA_EBPF_FAMILY, + socket_id_names[3],(long long)common_udp.write*8/100, + socket_id_names[4], (long long)common_udp.read*8/1000); + if (em->mode < MODE_ENTRY) { + write_err_chart( + NETDATA_UDP_FUNCTION_ERROR, NETDATA_EBPF_FAMILY, &socket_publish_aggregated[NETDATA_UDP_START], 2); + } +} + +/** + * Sum values for pid + * + * @param root the structure with all available PIDs + * + * @param offset the address that we are reading + * + * @return it returns the sum of all PIDs + */ +long long ebpf_socket_sum_values_for_pids(struct pid_on_target *root, size_t offset) +{ + long long ret = 0; + while (root) { + int32_t pid = root->pid; + ebpf_socket_publish_apps_t *w = socket_bandwidth_curr[pid]; + if (w) { + ret += get_value_from_structure((char *)w, offset); + } + + root = root->next; + } + + return ret; +} + +/** + * Send data to Netdata calling auxiliar functions. + * + * @param em the structure with thread information + * @param root the target list. + */ +void ebpf_socket_send_apps_data(ebpf_module_t *em, struct target *root) +{ + UNUSED(em); + if (!socket_apps_created) + return; + + struct target *w; + collected_number value; + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_NET_APPS_BANDWIDTH_SENT); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_socket_sum_values_for_pids(w->root_pid, offsetof(ebpf_socket_publish_apps_t, + bytes_sent)); + // We multiply by 0.008, because we read bytes, but we display bits + write_chart_dimension(w->name, ((value)*8)/1000); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_NET_APPS_BANDWIDTH_RECV); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_socket_sum_values_for_pids(w->root_pid, offsetof(ebpf_socket_publish_apps_t, + bytes_received)); + // We multiply by 0.008, because we read bytes, but we display bits + write_chart_dimension(w->name, ((value)*8)/1000); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_NET_APPS_BANDWIDTH_TCP_SEND_CALLS); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_socket_sum_values_for_pids(w->root_pid, offsetof(ebpf_socket_publish_apps_t, + call_tcp_sent)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_NET_APPS_BANDWIDTH_TCP_RECV_CALLS); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_socket_sum_values_for_pids(w->root_pid, offsetof(ebpf_socket_publish_apps_t, + call_tcp_received)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_NET_APPS_BANDWIDTH_TCP_RETRANSMIT); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_socket_sum_values_for_pids(w->root_pid, offsetof(ebpf_socket_publish_apps_t, + retransmit)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_NET_APPS_BANDWIDTH_UDP_SEND_CALLS); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_socket_sum_values_for_pids(w->root_pid, offsetof(ebpf_socket_publish_apps_t, + call_udp_sent)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + + write_begin_chart(NETDATA_APPS_FAMILY, NETDATA_NET_APPS_BANDWIDTH_UDP_RECV_CALLS); + for (w = root; w; w = w->next) { + if (unlikely(w->exposed && w->processes)) { + value = ebpf_socket_sum_values_for_pids(w->root_pid, offsetof(ebpf_socket_publish_apps_t, + call_udp_received)); + write_chart_dimension(w->name, value); + } + } + write_end_chart(); + +} + +/***************************************************************** + * + * FUNCTIONS TO CREATE CHARTS + * + *****************************************************************/ + +/** + * Create global charts + * + * Call ebpf_create_chart to create the charts for the collector. + * + * @param em a pointer to the structure with the default values. + */ +static void ebpf_create_global_charts(ebpf_module_t *em) +{ + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_TCP_FUNCTION_COUNT, + "Calls to internal functions", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_SOCKET_GROUP, + 21070, + ebpf_create_global_dimension, + socket_publish_aggregated, + 3); + + ebpf_create_chart(NETDATA_EBPF_FAMILY, NETDATA_TCP_FUNCTION_BITS, + "TCP bandwidth", EBPF_COMMON_DIMENSION_BITS, + NETDATA_SOCKET_GROUP, + 21071, + ebpf_create_global_dimension, + socket_publish_aggregated, + 3); + + if (em->mode < MODE_ENTRY) { + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_TCP_FUNCTION_ERROR, + "TCP errors", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_SOCKET_GROUP, + 21072, + ebpf_create_global_dimension, + socket_publish_aggregated, + 2); + } + + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_TCP_RETRANSMIT, + "Packages retransmitted", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_SOCKET_GROUP, + 21073, + ebpf_create_global_dimension, + &socket_publish_aggregated[NETDATA_RETRANSMIT_START], + 1); + + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_UDP_FUNCTION_COUNT, + "UDP calls", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_SOCKET_GROUP, + 21074, + ebpf_create_global_dimension, + &socket_publish_aggregated[NETDATA_UDP_START], + 2); + + ebpf_create_chart(NETDATA_EBPF_FAMILY, NETDATA_UDP_FUNCTION_BITS, + "UDP bandwidth", EBPF_COMMON_DIMENSION_BITS, + NETDATA_SOCKET_GROUP, + 21075, + ebpf_create_global_dimension, + &socket_publish_aggregated[NETDATA_UDP_START], + 2); + + if (em->mode < MODE_ENTRY) { + ebpf_create_chart(NETDATA_EBPF_FAMILY, + NETDATA_UDP_FUNCTION_ERROR, + "UDP errors", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_SOCKET_GROUP, + 21076, + ebpf_create_global_dimension, + &socket_publish_aggregated[NETDATA_UDP_START], + 2); + } +} + +/** + * Create apps charts + * + * Call ebpf_create_chart to create the charts on apps submenu. + * + * @param em a pointer to the structure with the default values. + */ +void ebpf_socket_create_apps_charts(ebpf_module_t *em, struct target *root) +{ + UNUSED(em); + ebpf_create_charts_on_apps(NETDATA_NET_APPS_BANDWIDTH_SENT, + "Bytes sent", EBPF_COMMON_DIMENSION_BITS, + NETDATA_APPS_NET_GROUP, + 20080, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_NET_APPS_BANDWIDTH_RECV, + "bytes received", EBPF_COMMON_DIMENSION_BITS, + NETDATA_APPS_NET_GROUP, + 20081, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_NET_APPS_BANDWIDTH_TCP_SEND_CALLS, + "Calls for tcp_sendmsg", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_NET_GROUP, + 20082, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_NET_APPS_BANDWIDTH_TCP_RECV_CALLS, + "Calls for tcp_cleanup_rbuf", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_NET_GROUP, + 20083, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_NET_APPS_BANDWIDTH_TCP_RETRANSMIT, + "Calls for tcp_retransmit", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_NET_GROUP, + 20084, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_NET_APPS_BANDWIDTH_UDP_SEND_CALLS, + "Calls for udp_sendmsg", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_NET_GROUP, + 20085, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + ebpf_create_charts_on_apps(NETDATA_NET_APPS_BANDWIDTH_UDP_RECV_CALLS, + "Calls for udp_recvmsg", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_APPS_NET_GROUP, + 20086, + ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX], + root); + + socket_apps_created = 1; +} + +/** + * Create network viewer chart + * + * Create common charts. + * + * @param id the chart id + * @param title the chart title + * @param units the units label + * @param family the group name used to attach the chart on dashaboard + * @param order the chart order + * @param ptr the plot structure with values. + */ +static void ebpf_socket_create_nv_chart(char *id, char *title, char *units, + char *family, int order, netdata_vector_plot_t *ptr) +{ + ebpf_write_chart_cmd(NETDATA_EBPF_FAMILY, + id, + title, + units, + family, + "stacked", + order); + + uint32_t i; + uint32_t end = ptr->last_plot; + netdata_socket_plot_t *w = ptr->plot; + for (i = 0; i < end; i++) { + fprintf(stdout, "DIMENSION %s '' incremental -1 1\n", w[i].dimension_sent); + fprintf(stdout, "DIMENSION %s '' incremental 1 1\n", w[i].dimension_recv); + } + + end = ptr->last; + fprintf(stdout, "DIMENSION %s '' incremental -1 1\n", w[end].dimension_sent); + fprintf(stdout, "DIMENSION %s '' incremental 1 1\n", w[end].dimension_recv); +} + +/** + * Create network viewer retransmit + * + * Create a specific chart. + * + * @param id the chart id + * @param title the chart title + * @param units the units label + * @param family the group name used to attach the chart on dashaboard + * @param order the chart order + * @param ptr the plot structure with values. + */ +static void ebpf_socket_create_nv_retransmit(char *id, char *title, char *units, + char *family, int order, netdata_vector_plot_t *ptr) +{ + ebpf_write_chart_cmd(NETDATA_EBPF_FAMILY, + id, + title, + units, + family, + "stacked", + order); + + uint32_t i; + uint32_t end = ptr->last_plot; + netdata_socket_plot_t *w = ptr->plot; + for (i = 0; i < end; i++) { + fprintf(stdout, "DIMENSION %s '' incremental 1 1\n", w[i].dimension_retransmit); + } + + end = ptr->last; + fprintf(stdout, "DIMENSION %s '' incremental 1 1\n", w[end].dimension_retransmit); +} + +/** + * Create Network Viewer charts + * + * Recreate the charts when new sockets are created. + * + * @param ptr a pointer for inbound or outbound vectors. + */ +static void ebpf_socket_create_nv_charts(netdata_vector_plot_t *ptr) +{ + // We do not have new sockets, so we do not need move forward + if (ptr->max_plot == ptr->last_plot) + return; + + ptr->last_plot = ptr->max_plot; + + if (ptr == (netdata_vector_plot_t *)&outbound_vectors) { + ebpf_socket_create_nv_chart(NETDATA_NV_OUTBOUND_BYTES, + "Outbound connections (bytes).", EBPF_COMMON_DIMENSION_BYTES, + NETDATA_NETWORK_CONNECTIONS_GROUP, + 21080, + ptr); + + ebpf_socket_create_nv_chart(NETDATA_NV_OUTBOUND_PACKETS, + "Outbound connections (packets)", + EBPF_COMMON_DIMENSION_PACKETS, + NETDATA_NETWORK_CONNECTIONS_GROUP, + 21082, + ptr); + + ebpf_socket_create_nv_retransmit(NETDATA_NV_OUTBOUND_RETRANSMIT, + "Retransmitted packets", + EBPF_COMMON_DIMENSION_CALL, + NETDATA_NETWORK_CONNECTIONS_GROUP, + 21083, + ptr); + } else { + ebpf_socket_create_nv_chart(NETDATA_NV_INBOUND_BYTES, + "Inbound connections (bytes)", EBPF_COMMON_DIMENSION_BYTES, + NETDATA_NETWORK_CONNECTIONS_GROUP, + 21084, + ptr); + + ebpf_socket_create_nv_chart(NETDATA_NV_INBOUND_PACKETS, + "Inbound connections (packets)", + EBPF_COMMON_DIMENSION_PACKETS, + NETDATA_NETWORK_CONNECTIONS_GROUP, + 21085, + ptr); + } + + ptr->flags |= NETWORK_VIEWER_CHARTS_CREATED; +} + +/***************************************************************** + * + * READ INFORMATION FROM KERNEL RING + * + *****************************************************************/ + +/** + * Is specific ip inside the range + * + * Check if the ip is inside a IP range previously defined + * + * @param cmp the IP to compare + * @param family the IP family + * + * @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) +{ + if (!network_viewer_opt.excluded_ips && !network_viewer_opt.included_ips) + return 1; + + uint32_t ipv4_test = ntohl(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]) ) + return 0; + } else { + if (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 0; + } + } + move = move->next; + } + + 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) + return 1; + } else { + if (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; + } + } + move = move->next; + } + + return 0; +} + +/** + * Is port inside range + * + * Verify if the cmp port is inside the range [first, last]. + * This function expects only the last parameter as big endian. + * + * @param cmp the value to compare + * + * @return It returns 1 when cmp is inside and 0 otherwise. + */ +static int is_port_inside_range(uint16_t cmp) +{ + // We do not have restrictions for ports. + if (!network_viewer_opt.excluded_port && !network_viewer_opt.included_port) + return 1; + + // Test if port is excluded + ebpf_network_viewer_port_list_t *move = network_viewer_opt.excluded_port; + cmp = htons(cmp); + while (move) { + if (move->cmp_first <= cmp && cmp <= move->cmp_last) + return 0; + + move = move->next; + } + + // Test if the port is inside allowed range + move = network_viewer_opt.included_port; + while (move) { + if (move->cmp_first <= cmp && cmp <= move->cmp_last) + return 1; + + move = move->next; + } + + return 0; +} + +/** + * Hostname matches pattern + * + * @param cmp the value to compare + * + * @return It returns 1 when the value matches and zero otherwise. + */ +int hostname_matches_pattern(char *cmp) +{ + if (!network_viewer_opt.included_hostnames && !network_viewer_opt.excluded_hostnames) + return 1; + + ebpf_network_viewer_hostname_list_t *move = network_viewer_opt.excluded_hostnames; + while (move) { + if (simple_pattern_matches(move->value_pattern, cmp)) + return 0; + + move = move->next; + } + + move = network_viewer_opt.included_hostnames; + while (move) { + if (simple_pattern_matches(move->value_pattern, cmp)) + return 1; + + move = move->next; + } + + + return 0; +} + +/** + * Is socket allowed? + * + * Compare destination addresses and destination ports to define next steps + * + * @param key the socket read from kernel ring + * @param family the family used to compare IPs (AF_INET and AF_INET6) + * + * @return It returns 1 if this socket is inside the ranges and 0 otherwise. + */ +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); +} + +/** + * Compare sockets + * + * Compare destination address and destination port. + * We do not compare source port, because it is random. + * We also do not compare source address, because inbound and outbound connections are stored in separated AVL trees. + * + * @param a pointer to netdata_socket_plot + * @param b pointer to netdata_socket_plot + * + * @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) +{ + struct netdata_socket_plot *val1 = a; + struct netdata_socket_plot *val2 = b; + int cmp; + + // 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 { + 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)); + } + } + } 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 { + 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)); + } + } + } + + return cmp; +} + +/** + * Build dimension name + * + * Fill dimension name vector with values given + * + * @param dimname the output vector + * @param hostname the hostname for the socket. + * @param service_name the service used to connect. + * @param proto the protocol used in this connection + * @param family is this IPV4(AF_INET) or IPV6(AF_INET6) + * + * @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) +{ + return snprintf(dimname, CONFIG_MAX_NAME - 7, (family == AF_INET)?"%s:%s:%s_":"%s:%s:[%s]_", + service_name, proto, + hostname); +} + +/** + * Fill inbound dimension name + * + * Mount the dimension name with the input given + * + * @param dimname the output vector + * @param service_name the service used to connect. + * @param proto the protocol used in this connection + * + * @return it returns the size of the data copied on success and -1 otherwise. + */ +static inline int build_inbound_dimension_name(char *dimname, char *service_name, char *proto) +{ + return snprintf(dimname, CONFIG_MAX_NAME - 7, "%s:%s_", service_name, + proto); +} + +/** + * Fill Resolved Name + * + * Fill the resolved name structure with the value given. + * The hostname is the largest value possible, if it is necessary to cut some value, it must be cut. + * + * @param ptr the output vector + * @param hostname the hostname resolved or IP. + * @param length the length for the hostname. + * @param service_name the service name associated to the connection + * @param is_outbound the is this an outbound connection + */ +static inline void fill_resolved_name(netdata_socket_plot_t *ptr, char *hostname, size_t length, + char *service_name, int is_outbound) +{ + if (length < NETDATA_MAX_NETWORK_COMBINED_LENGTH) + ptr->resolved_name = strdupz(hostname); + else { + length = NETDATA_MAX_NETWORK_COMBINED_LENGTH; + ptr->resolved_name = mallocz( NETDATA_MAX_NETWORK_COMBINED_LENGTH + 1); + memcpy(ptr->resolved_name, hostname, length); + ptr->resolved_name[length] = '\0'; + } + + char dimname[CONFIG_MAX_NAME]; + int size; + char *protocol; + if (ptr->sock.protocol == IPPROTO_UDP) { + protocol = "UDP"; + } else if (ptr->sock.protocol == IPPROTO_TCP) { + protocol = "TCP"; + } else { + protocol = "ALL"; + } + + if (is_outbound) + size = build_outbound_dimension_name(dimname, hostname, service_name, protocol, ptr->family); + else + size = build_inbound_dimension_name(dimname,service_name, protocol); + + if (size > 0) { + strcpy(&dimname[size], "sent"); + dimname[size + 4] = '\0'; + ptr->dimension_sent = strdupz(dimname); + + strcpy(&dimname[size], "recv"); + ptr->dimension_recv = strdupz(dimname); + + dimname[size - 1] = '\0'; + ptr->dimension_retransmit = strdupz(dimname); + } +} + +/** + * Mount dimension names + * + * Fill the vector names after to resolve the addresses + * + * @param ptr a pointer to the structure where the values are stored. + * @param is_outbound is a outbound ptr value? + * + * @return It returns 1 if the name is valid and 0 otherwise. + */ +int fill_names(netdata_socket_plot_t *ptr, int is_outbound) +{ + char hostname[NI_MAXHOST], service_name[NI_MAXSERV]; + if (ptr->resolved) + return 1; + + int ret; + static int resolve_name = -1; + static int resolve_service = -1; + if (resolve_name == -1) + resolve_name = network_viewer_opt.hostname_resolution_enabled; + + if (resolve_service == -1) + resolve_service = network_viewer_opt.service_resolution_enabled; + + netdata_socket_idx_t *idx = &ptr->index; + + char *errname = { "Not resolved" }; + // Resolve Name + if (ptr->family == AF_INET) { //IPV4 + struct sockaddr_in myaddr; + memset(&myaddr, 0 , sizeof(myaddr)); + + myaddr.sin_family = ptr->family; + if (is_outbound) { + myaddr.sin_port = idx->dport; + myaddr.sin_addr.s_addr = idx->daddr.addr32[0]; + } else { + myaddr.sin_port = idx->sport; + myaddr.sin_addr.s_addr = idx->saddr.addr32[0]; + } + + ret = (!resolve_name)?-1:getnameinfo((struct sockaddr *)&myaddr, sizeof(myaddr), hostname, + sizeof(hostname), service_name, sizeof(service_name), NI_NAMEREQD); + + if (!ret && !resolve_service) { + snprintf(service_name, sizeof(service_name), "%u", ntohs(myaddr.sin_port)); + } + + if (ret) { + // I cannot resolve the name, I will use the IP + if (!inet_ntop(AF_INET, &myaddr.sin_addr.s_addr, hostname, NI_MAXHOST)) { + strncpy(hostname, errname, 13); + } + + snprintf(service_name, sizeof(service_name), "%u", ntohs(myaddr.sin_port)); + ret = 1; + } + } else { // IPV6 + struct sockaddr_in6 myaddr6; + memset(&myaddr6, 0 , sizeof(myaddr6)); + + myaddr6.sin6_family = AF_INET6; + if (is_outbound) { + myaddr6.sin6_port = idx->dport; + memcpy(myaddr6.sin6_addr.s6_addr, idx->daddr.addr8, sizeof(union netdata_ip_t)); + } else { + myaddr6.sin6_port = idx->sport; + memcpy(myaddr6.sin6_addr.s6_addr, idx->saddr.addr8, sizeof(union netdata_ip_t)); + } + + ret = (!resolve_name)?-1:getnameinfo((struct sockaddr *)&myaddr6, sizeof(myaddr6), hostname, + sizeof(hostname), service_name, sizeof(service_name), NI_NAMEREQD); + + if (!ret && !resolve_service) { + snprintf(service_name, sizeof(service_name), "%u", ntohs(myaddr6.sin6_port)); + } + + if (ret) { + // I cannot resolve the name, I will use the IP + if (!inet_ntop(AF_INET6, myaddr6.sin6_addr.s6_addr, hostname, NI_MAXHOST)) { + strncpy(hostname, errname, 13); + } + + snprintf(service_name, sizeof(service_name), "%u", ntohs(myaddr6.sin6_port)); + + ret = 1; + } + } + + fill_resolved_name(ptr, hostname, + strlen(hostname) + strlen(service_name)+ NETDATA_DOTS_PROTOCOL_COMBINED_LENGTH, + service_name, is_outbound); + + if (resolve_name && !ret) + ret = hostname_matches_pattern(hostname); + + ptr->resolved++; + + return ret; +} + +/** + * Fill last Network Viewer Dimension + * + * Fill the unique dimension that is always plotted. + * + * @param ptr the pointer for the last dimension + * @param is_outbound is this an inbound structure? + */ +static void fill_last_nv_dimension(netdata_socket_plot_t *ptr, int is_outbound) +{ + char hostname[NI_MAXHOST], service_name[NI_MAXSERV]; + char *other = { "other" }; + // We are also copying the NULL bytes to avoid warnings in new compilers + strncpy(hostname, other, 6); + strncpy(service_name, other, 6); + + ptr->family = AF_INET; + ptr->sock.protocol = 255; + ptr->flags = (!is_outbound)?NETDATA_INBOUND_DIRECTION:NETDATA_OUTBOUND_DIRECTION; + + fill_resolved_name(ptr, hostname, 10 + NETDATA_DOTS_PROTOCOL_COMBINED_LENGTH, service_name, is_outbound); + +#ifdef NETDATA_INTERNAL_CHECKS + info("Last %s dimension added: ID = %u, IP = OTHER, NAME = %s, DIM1 = %s, DIM2 = %s, DIM3 = %s", + (is_outbound)?"outbound":"inbound", network_viewer_opt.max_dim - 1, ptr->resolved_name, + ptr->dimension_recv, ptr->dimension_sent, ptr->dimension_retransmit); +#endif +} + +/** + * Update Socket Data + * + * Update the socket information with last collected data + * + * @param sock + * @param lvalues + */ +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; +} + +/** + * Store socket inside avl + * + * Store the socket values inside the avl tree. + * + * @param out the structure with information used to plot charts. + * @param lvalues Values read from socket ring. + * @param lindex the index information, the real socket. + * @param family the family associated to the socket + * @param flags the connection flags + */ +static void store_socket_inside_avl(netdata_vector_plot_t *out, netdata_socket_t *lvalues, + netdata_socket_idx_t *lindex, int family, uint32_t flags) +{ + netdata_socket_plot_t test, *ret ; + + memcpy(&test.index, lindex, sizeof(netdata_socket_idx_t)); + test.flags = flags; + + ret = (netdata_socket_plot_t *) avl_search_lock(&out->tree, (avl *)&test); + if (ret) { + if (lvalues->ct > ret->plot.last_time) { + update_socket_data(&ret->sock, lvalues); + } + } else { + uint32_t curr = out->next; + uint32_t last = out->last; + + netdata_socket_plot_t *w = &out->plot[curr]; + + int resolved; + if (curr == last) { + if (lvalues->ct > w->plot.last_time) { + update_socket_data(&w->sock, lvalues); + } + return; + } else { + memcpy(&w->sock, lvalues, sizeof(netdata_socket_t)); + memcpy(&w->index, lindex, sizeof(netdata_socket_idx_t)); + w->family = family; + + resolved = fill_names(w, out != (netdata_vector_plot_t *)&inbound_vectors); + } + + if (!resolved) { + freez(w->resolved_name); + freez(w->dimension_sent); + freez(w->dimension_recv); + freez(w->dimension_retransmit); + + memset(w, 0, sizeof(netdata_socket_plot_t)); + + return; + } + + w->flags = flags; + netdata_socket_plot_t *check ; + check = (netdata_socket_plot_t *) avl_insert_lock(&out->tree, (avl *)w); + if (check != w) + error("Internal error, cannot insert the AVL tree."); + +#ifdef NETDATA_INTERNAL_CHECKS + char iptext[INET6_ADDRSTRLEN]; + if (inet_ntop(family, &w->index.daddr.addr8, iptext, sizeof(iptext))) + info("New %s dimension added: ID = %u, IP = %s, NAME = %s, DIM1 = %s, DIM2 = %s, DIM3 = %s", + (out == &inbound_vectors)?"inbound":"outbound", curr, iptext, w->resolved_name, + w->dimension_recv, w->dimension_sent, w->dimension_retransmit); +#endif + curr++; + if (curr > last) + curr = last; + out->next = curr; + } +} + +/** + * Compare Vector to store + * + * Compare input values with local address to select table to store. + * + * @param direction store inbound and outbound direction. + * @param cmp index read from hash table. + * @param proto the protocol read. + * + * @return It returns the structure with address to compare. + */ +netdata_vector_plot_t * select_vector_to_store(uint32_t *direction, netdata_socket_idx_t *cmp, uint8_t proto) +{ + if (!listen_ports) { + *direction = NETDATA_OUTBOUND_DIRECTION; + return &outbound_vectors; + } + + ebpf_network_viewer_port_list_t *move_ports = listen_ports; + while (move_ports) { + if (move_ports->protocol == proto && move_ports->first == cmp->sport) { + *direction = NETDATA_INBOUND_DIRECTION; + return &inbound_vectors; + } + + move_ports = move_ports->next; + } + + *direction = NETDATA_OUTBOUND_DIRECTION; + return &outbound_vectors; +} + +/** + * Hash accumulator + * + * @param values the values used to calculate the data. + * @param key the key to store data. + * @param removesock check if this socket must be removed . + * @param family the connection family + * @param end the values size. + */ +static void hash_accumulator(netdata_socket_t *values, netdata_socket_idx_t *key, int *removesock, int family, int end) +{ + uint64_t bsent = 0, brecv = 0, psent = 0, precv = 0; + uint16_t retransmit = 0; + int i; + uint8_t protocol = values[0].protocol; + uint64_t ct = values[0].ct; + for (i = 1; i < end; i++) { + netdata_socket_t *w = &values[i]; + + precv += w->recv_packets; + psent += w->sent_packets; + brecv += w->recv_bytes; + bsent += w->sent_bytes; + retransmit += w->retransmit; + + if (!protocol) + protocol = w->protocol; + + if (w->ct > ct) + ct = w->ct; + + *removesock += (int)w->removeme; + } + + values[0].recv_packets += precv; + values[0].sent_packets += psent; + values[0].recv_bytes += brecv; + values[0].sent_bytes += bsent; + values[0].retransmit += retransmit; + values[0].removeme += (uint8_t)*removesock; + 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); + } +} + +/** + * Read socket hash table + * + * 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 + * + * @return it returns 0 on success and -1 otherwise. + */ +static void read_socket_hash_table(int fd, int family, int network_connection) +{ + if (wait_to_plot) + return; + + netdata_socket_idx_t key = {}; + netdata_socket_idx_t next_key; + netdata_socket_idx_t removeme; + int removesock = 0; + + netdata_socket_t *values = socket_values; + size_t length = ebpf_nprocs*sizeof(netdata_socket_t); + int test, end = (running_on_kernel < NETDATA_KERNEL_V4_15) ? 1 : ebpf_nprocs; + + while (bpf_map_get_next_key(fd, &key, &next_key) == 0) { + // We need to reset the values when we are working on kernel 4.15 or newer, because kernel does not create + // values for specific processor unless it is used to store data. As result of this behavior one the next socket + // can have values from the previous one. + memset(values, 0, length); + test = bpf_map_lookup_elem(fd, &key, values); + if (test < 0) { + key = next_key; + continue; + } + + if (removesock) + bpf_map_delete_elem(fd, &removeme); + + if (network_connection) { + removesock = 0; + hash_accumulator(values, &key, &removesock, family, end); + } + + if (removesock) + removeme = key; + + key = next_key; + } + + if (removesock) + bpf_map_delete_elem(fd, &removeme); + + test = bpf_map_lookup_elem(fd, &next_key, values); + if (test < 0) { + return; + } + + if (network_connection) { + removesock = 0; + hash_accumulator(values, &next_key, &removesock, family, end); + } + + if (removesock) + bpf_map_delete_elem(fd, &next_key); +} + +/** + * Update listen table + * + * Update link list when it is necessary. + * + * @param value the ports we are listen to. + * @param proto the protocol used with port connection. + */ +void update_listen_table(uint16_t value, uint8_t proto) +{ + ebpf_network_viewer_port_list_t *w; + if (likely(listen_ports)) { + ebpf_network_viewer_port_list_t *move = listen_ports, *store = listen_ports; + while (move) { + if (move->protocol == proto && move->first == value) + return; + + store = move; + move = move->next; + } + + w = callocz(1, sizeof(ebpf_network_viewer_port_list_t)); + w->first = value; + w->protocol = proto; + store->next = w; + } else { + w = callocz(1, sizeof(ebpf_network_viewer_port_list_t)); + w->first = value; + w->protocol = proto; + + listen_ports = w; + } + +#ifdef NETDATA_INTERNAL_CHECKS + info("The network viewer is monitoring inbound connections for port %u", ntohs(value)); +#endif +} + +/** + * Read listen table + * + * Read the table with all ports that we are listen on host. + */ +static void read_listen_table() +{ + uint16_t key = 0; + uint16_t next_key; + + int fd = map_fd[NETDATA_SOCKET_LISTEN_TABLE]; + uint8_t value; + while (bpf_map_get_next_key(fd, &key, &next_key) == 0) { + int test = bpf_map_lookup_elem(fd, &key, &value); + if (test < 0) { + key = next_key; + continue; + } + + // The correct protocol must come from kernel + update_listen_table(htons(key), (key == 53)?IPPROTO_UDP:IPPROTO_TCP); + + key = next_key; + } + + if (next_key) { + // The correct protocol must come from kernel + update_listen_table(htons(next_key), (key == 53)?IPPROTO_UDP:IPPROTO_TCP); + } +} + +/** + * 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_socket_read_hash(void *ptr) +{ + ebpf_module_t *em = (ebpf_module_t *)ptr; + + read_thread_closed = 0; + heartbeat_t hb; + heartbeat_init(&hb); + usec_t step = NETDATA_SOCKET_READ_SLEEP_MS; + int fd_ipv4 = map_fd[NETDATA_SOCKET_IPV4_HASH_TABLE]; + int fd_ipv6 = map_fd[NETDATA_SOCKET_IPV6_HASH_TABLE]; + int network_connection = em->optional; + while (!close_ebpf_plugin) { + usec_t dt = heartbeat_next(&hb, step); + (void)dt; + + 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; + pthread_mutex_unlock(&nv_mutex); + } + + read_thread_closed = 1; + return NULL; +} + +/** + * Read the hash table and store data to allocated vectors. + */ +static void read_hash_global_tables() +{ + uint64_t idx; + netdata_idx_t res[NETDATA_SOCKET_COUNTER]; + + netdata_idx_t *val = socket_hash_values; + int fd = map_fd[NETDATA_SOCKET_GLOBAL_HASH_TABLE]; + for (idx = 0; idx < NETDATA_SOCKET_COUNTER; idx++) { + if (!bpf_map_lookup_elem(fd, &idx, val)) { + uint64_t total = 0; + int i; + int end = (running_on_kernel < NETDATA_KERNEL_V4_15) ? 1 : ebpf_nprocs; + for (i = 0; i < end; i++) + total += val[i]; + + res[idx] = total; + } else { + res[idx] = 0; + } + } + + socket_aggregated_data[0].call = res[NETDATA_KEY_CALLS_TCP_SENDMSG]; + socket_aggregated_data[1].call = res[NETDATA_KEY_CALLS_TCP_CLEANUP_RBUF]; + socket_aggregated_data[2].call = res[NETDATA_KEY_CALLS_TCP_CLOSE]; + socket_aggregated_data[3].call = res[NETDATA_KEY_CALLS_UDP_RECVMSG]; + socket_aggregated_data[4].call = res[NETDATA_KEY_CALLS_UDP_SENDMSG]; + socket_aggregated_data[5].call = res[NETDATA_KEY_TCP_RETRANSMIT]; + + socket_aggregated_data[0].ecall = res[NETDATA_KEY_ERROR_TCP_SENDMSG]; + socket_aggregated_data[1].ecall = res[NETDATA_KEY_ERROR_TCP_CLEANUP_RBUF]; + socket_aggregated_data[3].ecall = res[NETDATA_KEY_ERROR_UDP_RECVMSG]; + socket_aggregated_data[4].ecall = res[NETDATA_KEY_ERROR_UDP_SENDMSG]; + + socket_aggregated_data[0].bytes = res[NETDATA_KEY_BYTES_TCP_SENDMSG]; + socket_aggregated_data[1].bytes = res[NETDATA_KEY_BYTES_TCP_CLEANUP_RBUF]; + socket_aggregated_data[3].bytes = res[NETDATA_KEY_BYTES_UDP_RECVMSG]; + socket_aggregated_data[4].bytes = res[NETDATA_KEY_BYTES_UDP_SENDMSG]; +} + +/** + * Fill publish apps when necessary. + * + * @param current_pid the PID that I am updating + * @param eb the structure with data read from memory. + */ +void ebpf_socket_fill_publish_apps(uint32_t current_pid, ebpf_bandwidth_t *eb) +{ + ebpf_socket_publish_apps_t *curr = socket_bandwidth_curr[current_pid]; + if (!curr) { + curr = callocz(1, sizeof(ebpf_socket_publish_apps_t)); + socket_bandwidth_curr[current_pid] = curr; + } + + curr->bytes_sent = eb->bytes_sent; + curr->bytes_received = eb->bytes_received; + curr->call_tcp_sent = eb->call_tcp_sent; + curr->call_tcp_received = eb->call_tcp_received; + curr->retransmit = eb->retransmit; + curr->call_udp_sent = eb->call_udp_sent; + curr->call_udp_received = eb->call_udp_received; +} + +/** + * Bandwidth accumulator. + * + * @param out the vector with the values to sum + */ +void ebpf_socket_bandwidth_accumulator(ebpf_bandwidth_t *out) +{ + int i, end = (running_on_kernel >= NETDATA_KERNEL_V4_15) ? ebpf_nprocs : 1; + ebpf_bandwidth_t *total = &out[0]; + for (i = 1; i < end; i++) { + ebpf_bandwidth_t *move = &out[i]; + total->bytes_sent += move->bytes_sent; + total->bytes_received += move->bytes_received; + total->call_tcp_sent += move->call_tcp_sent; + total->call_tcp_received += move->call_tcp_received; + total->retransmit += move->retransmit; + total->call_udp_sent += move->call_udp_sent; + total->call_udp_received += move->call_udp_received; + } +} + +/** + * Update the apps data reading information from the hash table + */ +static void ebpf_socket_update_apps_data() +{ + int fd = map_fd[NETDATA_SOCKET_APPS_HASH_TABLE]; + ebpf_bandwidth_t *eb = bandwidth_vector; + uint32_t key; + struct pid_stat *pids = root_of_pids; + while (pids) { + key = pids->pid; + + if (bpf_map_lookup_elem(fd, &key, eb)) { + pids = pids->next; + continue; + } + + ebpf_socket_bandwidth_accumulator(eb); + + ebpf_socket_fill_publish_apps(key, eb); + + pids = pids->next; + } +} + +/***************************************************************** + * + * FUNCTIONS WITH THE MAIN LOOP + * + *****************************************************************/ + +struct netdata_static_thread socket_threads = {"EBPF SOCKET READ", + NULL, NULL, 1, NULL, + NULL, ebpf_socket_read_hash }; + +/** + * 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) +{ + UNUSED(em); + UNUSED(step); + heartbeat_t hb; + heartbeat_init(&hb); + + socket_threads.thread = mallocz(sizeof(netdata_thread_t)); + + netdata_thread_create(socket_threads.thread, socket_threads.name, + NETDATA_THREAD_OPTION_JOINABLE, ebpf_socket_read_hash, em); + + int socket_apps_enabled = ebpf_modules[EBPF_MODULE_SOCKET_IDX].apps_charts; + int socket_global_enabled = ebpf_modules[EBPF_MODULE_SOCKET_IDX].global_charts; + int network_connection = em->optional; + while (!close_ebpf_plugin) { + pthread_mutex_lock(&collect_data_mutex); + pthread_cond_wait(&collect_data_cond_var, &collect_data_mutex); + + if (socket_global_enabled) + read_hash_global_tables(); + + if (socket_apps_enabled) + ebpf_socket_update_apps_data(); + + calculate_nv_plot(); + + pthread_mutex_lock(&lock); + if (socket_global_enabled) + ebpf_socket_send_data(em); + + if (socket_apps_enabled) + ebpf_socket_send_apps_data(em, apps_groups_root_target); + + fflush(stdout); + + if (network_connection) { + // We are calling fflush many times, because when we have a lot of dimensions + // we began to have not expected outputs and Netdata closed the plugin. + pthread_mutex_lock(&nv_mutex); + ebpf_socket_create_nv_charts(&inbound_vectors); + fflush(stdout); + ebpf_socket_send_nv_data(&inbound_vectors); + + ebpf_socket_create_nv_charts(&outbound_vectors); + fflush(stdout); + ebpf_socket_send_nv_data(&outbound_vectors); + wait_to_plot = 0; + pthread_mutex_unlock(&nv_mutex); + + } + + pthread_mutex_unlock(&collect_data_mutex); + pthread_mutex_unlock(&lock); + + } +} + +/***************************************************************** + * + * FUNCTIONS TO CLOSE THE THREAD + * + *****************************************************************/ + + +/** + * Clean internal socket plot + * + * Clean all structures allocated with strdupz. + * + * @param ptr the pointer with addresses to clean. + */ +static inline void clean_internal_socket_plot(netdata_socket_plot_t *ptr) +{ + freez(ptr->dimension_recv); + freez(ptr->dimension_sent); + freez(ptr->resolved_name); + freez(ptr->dimension_retransmit); +} + +/** + * Clean socket plot + * + * Clean the allocated data for inbound and outbound vectors. + */ +static void clean_allocated_socket_plot() +{ + uint32_t i; + uint32_t end = inbound_vectors.last; + netdata_socket_plot_t *plot = inbound_vectors.plot; + for (i = 0; i < end; i++) { + clean_internal_socket_plot(&plot[i]); + } + + clean_internal_socket_plot(&plot[inbound_vectors.last]); + + end = outbound_vectors.last; + plot = outbound_vectors.plot; + for (i = 0; i < end; i++) { + clean_internal_socket_plot(&plot[i]); + } + clean_internal_socket_plot(&plot[outbound_vectors.last]); +} + +/** + * Clean netowrk ports allocated during initializaion. + * + * @param ptr a pointer to the link list. + */ +static void clean_network_ports(ebpf_network_viewer_port_list_t *ptr) +{ + if (unlikely(!ptr)) + return; + + while (ptr) { + ebpf_network_viewer_port_list_t *next = ptr->next; + freez(ptr->value); + freez(ptr); + ptr = next; + } +} + +/** + * Clean service names + * + * Clean the allocated link list that stores names. + * + * @param names the link list. + */ +static void clean_service_names(ebpf_network_viewer_dim_name_t *names) +{ + if (unlikely(!names)) + return; + + while (names) { + ebpf_network_viewer_dim_name_t *next = names->next; + freez(names->name); + freez(names); + names = next; + } +} + +/** + * Clean hostnames + * + * @param hostnames the hostnames to clean + */ +static void clean_hostnames(ebpf_network_viewer_hostname_list_t *hostnames) +{ + if (unlikely(!hostnames)) + return; + + while (hostnames) { + ebpf_network_viewer_hostname_list_t *next = hostnames->next; + freez(hostnames->value); + simple_pattern_free(hostnames->value_pattern); + freez(hostnames); + hostnames = next; + } +} + +void clean_thread_structures() { + struct pid_stat *pids = root_of_pids; + while (pids) { + freez(socket_bandwidth_curr[pids->pid]); + + pids = pids->next; + } +} + +/** + * Clean up the main thread. + * + * @param ptr thread data. + */ +static void ebpf_socket_cleanup(void *ptr) +{ + ebpf_module_t *em = (ebpf_module_t *)ptr; + if (!em->enabled) + return; + + heartbeat_t hb; + heartbeat_init(&hb); + uint32_t tick = 2*USEC_PER_MS; + while (!read_thread_closed) { + usec_t dt = heartbeat_next(&hb, tick); + UNUSED(dt); + } + + freez(socket_aggregated_data); + ebpf_cleanup_publish_syscall(socket_publish_aggregated); + freez(socket_publish_aggregated); + freez(socket_hash_values); + + clean_thread_structures(); + freez(socket_bandwidth_curr); + freez(bandwidth_vector); + + freez(socket_values); + clean_allocated_socket_plot(); + freez(inbound_vectors.plot); + freez(outbound_vectors.plot); + + clean_port_structure(&listen_ports); + + ebpf_modules[EBPF_MODULE_SOCKET_IDX].enabled = 0; + + clean_network_ports(network_viewer_opt.included_port); + clean_network_ports(network_viewer_opt.excluded_port); + clean_service_names(network_viewer_opt.names); + clean_hostnames(network_viewer_opt.included_hostnames); + clean_hostnames(network_viewer_opt.excluded_hostnames); + + pthread_mutex_destroy(&nv_mutex); + freez(socket_data.map_fd); + + freez(socket_threads.thread); + + struct bpf_program *prog; + size_t i = 0 ; + bpf_object__for_each_program(prog, objects) { + bpf_link__destroy(probe_links[i]); + i++; + } + bpf_object__close(objects); + finalized_threads = 1; +} + +/***************************************************************** + * + * FUNCTIONS TO START THREAD + * + *****************************************************************/ + +/** + * Allocate vectors used with this thread. + * We are not testing the return, because callocz does this and shutdown the software + * case it was not possible to allocate. + * + * @param length is the length for the vectors used inside the collector. + */ +static void ebpf_socket_allocate_global_vectors(size_t length) +{ + socket_aggregated_data = callocz(length, sizeof(netdata_syscall_stat_t)); + socket_publish_aggregated = callocz(length, sizeof(netdata_publish_syscall_t)); + socket_hash_values = callocz(ebpf_nprocs, sizeof(netdata_idx_t)); + + socket_bandwidth_curr = callocz((size_t)pid_max, sizeof(ebpf_socket_publish_apps_t *)); + 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)); +} + +/** + * Set local function pointers, this function will never be compiled with static libraries + */ +static void set_local_pointers() +{ + map_fd = socket_data.map_fd; +} + +/** + * Initialize Inbound and Outbound + * + * Initialize the common outbound and inbound sockets. + */ +static void initialize_inbound_outbound() +{ + inbound_vectors.last = network_viewer_opt.max_dim - 1; + outbound_vectors.last = inbound_vectors.last; + fill_last_nv_dimension(&inbound_vectors.plot[inbound_vectors.last], 0); + fill_last_nv_dimension(&outbound_vectors.plot[outbound_vectors.last], 1); +} + +/***************************************************************** + * + * EBPF SOCKET THREAD + * + *****************************************************************/ + +/** + * Socket thread + * + * Thread used to generate socket charts. + * + * @param ptr a pointer to `struct ebpf_module` + * + * @return It always return NULL + */ +void *ebpf_socket_thread(void *ptr) +{ + netdata_thread_cleanup_push(ebpf_socket_cleanup, ptr); + + avl_init_lock(&inbound_vectors.tree, compare_sockets); + avl_init_lock(&outbound_vectors.tree, compare_sockets); + + ebpf_module_t *em = (ebpf_module_t *)ptr; + fill_ebpf_data(&socket_data); + + if (!em->enabled) + goto endsocket; + + if (pthread_mutex_init(&nv_mutex, NULL)) { + error("Cannot initialize local mutex"); + goto endsocket; + } + pthread_mutex_lock(&lock); + + ebpf_socket_allocate_global_vectors(NETDATA_MAX_SOCKET_VECTOR); + initialize_inbound_outbound(); + + if (ebpf_update_kernel(&socket_data)) { + pthread_mutex_unlock(&lock); + goto endsocket; + } + + set_local_pointers(); + probe_links = ebpf_load_program(ebpf_plugin_dir, em, kernel_string, &objects, socket_data.map_fd); + if (!probe_links) { + pthread_mutex_unlock(&lock); + goto endsocket; + } + + int algorithms[NETDATA_MAX_SOCKET_VECTOR] = { + NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX, + NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX, NETDATA_EBPF_ABSOLUTE_IDX + }; + ebpf_global_labels( + socket_aggregated_data, socket_publish_aggregated, socket_dimension_names, socket_id_names, + algorithms, NETDATA_MAX_SOCKET_VECTOR); + + ebpf_create_global_charts(em); + + finalized_threads = 0; + pthread_mutex_unlock(&lock); + + socket_collector((usec_t)(em->update_time * USEC_PER_SEC), em); + +endsocket: + netdata_thread_cleanup_pop(1); + return NULL; +} diff --git a/collectors/ebpf.plugin/ebpf_socket.h b/collectors/ebpf.plugin/ebpf_socket.h new file mode 100644 index 0000000..1316c00 --- /dev/null +++ b/collectors/ebpf.plugin/ebpf_socket.h @@ -0,0 +1,275 @@ +// SPDX-License-Identifier: GPL-3.0-or-later +#ifndef NETDATA_EBPF_SOCKET_H +#define NETDATA_EBPF_SOCKET_H 1 +#include <stdint.h> +#include "libnetdata/avl/avl.h" + +// Vector indexes +#define NETDATA_MAX_SOCKET_VECTOR 6 +#define NETDATA_UDP_START 3 +#define NETDATA_RETRANSMIT_START 5 + +#define NETDATA_SOCKET_APPS_HASH_TABLE 0 +#define NETDATA_SOCKET_IPV4_HASH_TABLE 1 +#define NETDATA_SOCKET_IPV6_HASH_TABLE 2 +#define NETDATA_SOCKET_GLOBAL_HASH_TABLE 4 +#define NETDATA_SOCKET_LISTEN_TABLE 5 + +#define NETDATA_SOCKET_READ_SLEEP_MS 800000ULL + +typedef enum ebpf_socket_idx { + NETDATA_KEY_CALLS_TCP_SENDMSG, + NETDATA_KEY_ERROR_TCP_SENDMSG, + NETDATA_KEY_BYTES_TCP_SENDMSG, + + NETDATA_KEY_CALLS_TCP_CLEANUP_RBUF, + NETDATA_KEY_ERROR_TCP_CLEANUP_RBUF, + NETDATA_KEY_BYTES_TCP_CLEANUP_RBUF, + + NETDATA_KEY_CALLS_TCP_CLOSE, + + NETDATA_KEY_CALLS_UDP_RECVMSG, + NETDATA_KEY_ERROR_UDP_RECVMSG, + NETDATA_KEY_BYTES_UDP_RECVMSG, + + NETDATA_KEY_CALLS_UDP_SENDMSG, + NETDATA_KEY_ERROR_UDP_SENDMSG, + NETDATA_KEY_BYTES_UDP_SENDMSG, + + NETDATA_KEY_TCP_RETRANSMIT, + + NETDATA_SOCKET_COUNTER +} ebpf_socket_index_t; + +#define NETDATA_SOCKET_GROUP "Socket" +#define NETDATA_NETWORK_CONNECTIONS_GROUP "Network connections" + +// Global chart name +#define NETDATA_TCP_FUNCTION_COUNT "tcp_functions" +#define NETDATA_TCP_FUNCTION_BITS "total_tcp_bandwidth" +#define NETDATA_TCP_FUNCTION_ERROR "tcp_error" +#define NETDATA_TCP_RETRANSMIT "tcp_retransmit" +#define NETDATA_UDP_FUNCTION_COUNT "udp_functions" +#define NETDATA_UDP_FUNCTION_BITS "total_udp_bandwidth" +#define NETDATA_UDP_FUNCTION_ERROR "udp_error" + +// Charts created on Apps submenu +#define NETDATA_NET_APPS_BANDWIDTH_SENT "total_bandwidth_sent" +#define NETDATA_NET_APPS_BANDWIDTH_RECV "total_bandwidth_recv" +#define NETDATA_NET_APPS_BANDWIDTH_TCP_SEND_CALLS "bandwidth_tcp_send" +#define NETDATA_NET_APPS_BANDWIDTH_TCP_RECV_CALLS "bandwidth_tcp_recv" +#define NETDATA_NET_APPS_BANDWIDTH_TCP_RETRANSMIT "bandwidth_tcp_retransmit" +#define NETDATA_NET_APPS_BANDWIDTH_UDP_SEND_CALLS "bandwidth_udp_send" +#define NETDATA_NET_APPS_BANDWIDTH_UDP_RECV_CALLS "bandwidth_udp_recv" + +// Network viewer charts +#define NETDATA_NV_OUTBOUND_BYTES "outbound_bytes" +#define NETDATA_NV_OUTBOUND_PACKETS "outbound_packets" +#define NETDATA_NV_OUTBOUND_RETRANSMIT "outbound_retransmit" +#define NETDATA_NV_INBOUND_BYTES "inbound_bytes" +#define NETDATA_NV_INBOUND_PACKETS "inbound_packets" + +// Port range +#define NETDATA_MINIMUM_PORT_VALUE 1 +#define NETDATA_MAXIMUM_PORT_VALUE 65535 + +#define NETDATA_MINIMUM_IPV4_CIDR 0 +#define NETDATA_MAXIMUM_IPV4_CIDR 32 + +typedef struct ebpf_socket_publish_apps { + // Data read + uint64_t bytes_sent; // Bytes sent + uint64_t bytes_received; // Bytes received + uint64_t call_tcp_sent; // Number of times tcp_sendmsg was called + uint64_t call_tcp_received; // Number of times tcp_cleanup_rbuf was called + uint64_t retransmit; // Number of times tcp_retransmit was called + uint64_t call_udp_sent; // Number of times udp_sendmsg was called + uint64_t call_udp_received; // Number of times udp_recvmsg was called + + // Publish information. + uint64_t publish_sent_bytes; + uint64_t publish_received_bytes; + uint64_t publish_tcp_sent; + uint64_t publish_tcp_received; + uint64_t publish_retransmit; + uint64_t publish_udp_sent; + uint64_t publish_udp_received; +} ebpf_socket_publish_apps_t; + +typedef struct ebpf_network_viewer_dimension_names { + char *name; + uint32_t hash; + + uint16_t port; + + struct ebpf_network_viewer_dimension_names *next; +} ebpf_network_viewer_dim_name_t ; + +typedef struct ebpf_network_viewer_port_list { + char *value; + uint32_t hash; + + uint16_t first; + uint16_t last; + + uint16_t cmp_first; + uint16_t cmp_last; + + uint8_t protocol; + struct ebpf_network_viewer_port_list *next; +} ebpf_network_viewer_port_list_t; + +/** + * Union used to store ip addresses + */ +union netdata_ip_t { + uint8_t addr8[16]; + uint16_t addr16[8]; + uint32_t addr32[4]; + uint64_t addr64[2]; +}; + +typedef struct ebpf_network_viewer_ip_list { + char *value; // IP value + uint32_t hash; // IP hash + + uint8_t ver; // IP version + + union netdata_ip_t first; // The IP address informed + union netdata_ip_t last; // The IP address informed + + struct ebpf_network_viewer_ip_list *next; +} ebpf_network_viewer_ip_list_t; + +typedef struct ebpf_network_viewer_hostname_list { + char *value; // IP value + uint32_t hash; // IP hash + + SIMPLE_PATTERN *value_pattern; + + struct ebpf_network_viewer_hostname_list *next; +} ebpf_network_viewer_hostname_list_t; + +#define NETDATA_NV_CAP_VALUE 50L +typedef struct ebpf_network_viewer_options { + uint32_t max_dim; // Store value read from 'maximum dimensions' + + uint32_t hostname_resolution_enabled; + uint32_t service_resolution_enabled; + + ebpf_network_viewer_port_list_t *excluded_port; + ebpf_network_viewer_port_list_t *included_port; + + ebpf_network_viewer_dim_name_t *names; + + ebpf_network_viewer_ip_list_t *excluded_ips; + ebpf_network_viewer_ip_list_t *included_ips; + + ebpf_network_viewer_hostname_list_t *excluded_hostnames; + ebpf_network_viewer_hostname_list_t *included_hostnames; + + ebpf_network_viewer_ip_list_t *ipv4_local_ip; + ebpf_network_viewer_ip_list_t *ipv6_local_ip; +} ebpf_network_viewer_options_t; + +extern ebpf_network_viewer_options_t network_viewer_opt; + +/** + * Structure to store socket information + */ +typedef struct netdata_socket { + uint64_t recv_packets; + uint64_t sent_packets; + uint64_t recv_bytes; + uint64_t sent_bytes; + uint64_t first; // First timestamp + uint64_t ct; // Current timestamp + uint16_t retransmit; // It is never used with UDP + uint8_t protocol; + uint8_t removeme; + uint32_t reserved; +} netdata_socket_t __attribute__((__aligned__(8))); + + +typedef struct netdata_plot_values { + // Values used in the previous iteration + uint64_t recv_packets; + uint64_t sent_packets; + uint64_t recv_bytes; + uint64_t sent_bytes; + uint16_t retransmit; + + uint64_t last_time; + + // Values used to plot + uint64_t plot_recv_packets; + uint64_t plot_sent_packets; + uint64_t plot_recv_bytes; + uint64_t plot_sent_bytes; + uint16_t plot_retransmit; +} netdata_plot_values_t; + +/** + * Index used together previous structure + */ +typedef struct netdata_socket_idx { + union netdata_ip_t saddr; + uint16_t sport; + union netdata_ip_t daddr; + uint16_t dport; +} netdata_socket_idx_t __attribute__((__aligned__(8))); + +// Next values were defined according getnameinfo(3) +#define NETDATA_MAX_NETWORK_COMBINED_LENGTH 1018 +#define NETDATA_DOTS_PROTOCOL_COMBINED_LENGTH 5 // :TCP: +#define NETDATA_DIM_LENGTH_WITHOUT_SERVICE_PROTOCOL 979 + +#define NETDATA_INBOUND_DIRECTION (uint32_t)1 +#define NETDATA_OUTBOUND_DIRECTION (uint32_t)2 +/** + * Allocate the maximum number of structures in the beginning, this can force the collector to use more memory + * in the long term, on the other had it is faster. + */ +typedef struct netdata_socket_plot { + // Search + avl avl; + netdata_socket_idx_t index; + + // Current data + netdata_socket_t sock; + + // Previous values and values used to write on chart. + netdata_plot_values_t plot; + + int family; // AF_INET or AF_INET6 + char *resolved_name; // Resolve only in the first call + unsigned char resolved; + + char *dimension_sent; + char *dimension_recv; + char *dimension_retransmit; + + uint32_t flags; +} netdata_socket_plot_t; + +#define NETWORK_VIEWER_CHARTS_CREATED (uint32_t)1 +typedef struct netdata_vector_plot { + netdata_socket_plot_t *plot; // Vector used to plot charts + + avl_tree_lock tree; // AVL tree to speed up search + uint32_t last; // The 'other' dimension, the last chart accepted. + uint32_t next; // The next position to store in the vector. + uint32_t max_plot; // Max number of elements to plot. + uint32_t last_plot; // Last element plot + + uint32_t flags; // Flags + +} netdata_vector_plot_t; + +extern void clean_port_structure(ebpf_network_viewer_port_list_t **clean); +extern ebpf_network_viewer_port_list_t *listen_ports; +extern void update_listen_table(uint16_t value, uint8_t proto); + +extern ebpf_socket_publish_apps_t **socket_bandwidth_curr; + +#endif diff --git a/collectors/ebpf.plugin/reset_netdata_trace.sh.in b/collectors/ebpf.plugin/reset_netdata_trace.sh.in new file mode 100644 index 0000000..51d981e --- /dev/null +++ b/collectors/ebpf.plugin/reset_netdata_trace.sh.in @@ -0,0 +1,9 @@ +#!/bin/bash + +KPROBE_FILE="/sys/kernel/debug/tracing/kprobe_events" + +DATA="$(grep _netdata_ $KPROBE_FILE| cut -d' ' -f1 | cut -d: -f2)" + +for I in $DATA; do + echo "-:$I" > $KPROBE_FILE 2>/dev/null; +done |