bin: compile lttng-sessiond as C++
[lttng-tools.git] / src / bin / lttng-sessiond / kernel.c
diff --git a/src/bin/lttng-sessiond/kernel.c b/src/bin/lttng-sessiond/kernel.c
deleted file mode 100644 (file)
index ffbde69..0000000
+++ /dev/null
@@ -1,2544 +0,0 @@
-/*
- * Copyright (C) 2011 David Goulet <david.goulet@polymtl.ca>
- *
- * SPDX-License-Identifier: GPL-2.0-only
- *
- */
-
-#define _LGPL_SOURCE
-#include <fcntl.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include <unistd.h>
-#include <inttypes.h>
-#include <sys/types.h>
-
-#include <common/common.h>
-#include <common/hashtable/utils.h>
-#include <common/trace-chunk.h>
-#include <common/kernel-ctl/kernel-ctl.h>
-#include <common/kernel-ctl/kernel-ioctl.h>
-#include <common/sessiond-comm/sessiond-comm.h>
-#include <common/tracker.h>
-#include <common/utils.h>
-#include <lttng/event.h>
-#include <lttng/lttng-error.h>
-#include <lttng/tracker.h>
-
-#include <lttng/userspace-probe.h>
-#include <lttng/userspace-probe-internal.h>
-#include <lttng/condition/event-rule-matches.h>
-#include <lttng/condition/event-rule-matches-internal.h>
-#include <lttng/event-rule/event-rule.h>
-#include <lttng/event-rule/event-rule-internal.h>
-#include <lttng/event-rule/kernel-uprobe-internal.h>
-
-#include "event-notifier-error-accounting.h"
-#include "lttng-sessiond.h"
-#include "lttng-syscall.h"
-#include "condition-internal.h"
-#include "consumer.h"
-#include "kernel.h"
-#include "kernel-consumer.h"
-#include "kern-modules.h"
-#include "sessiond-config.h"
-#include "utils.h"
-#include "rotate.h"
-#include "modprobe.h"
-#include "tracker.h"
-#include "notification-thread-commands.h"
-
-/*
- * Key used to reference a channel between the sessiond and the consumer. This
- * is only read and updated with the session_list lock held.
- */
-static uint64_t next_kernel_channel_key;
-
-static const char *module_proc_lttng = "/proc/lttng";
-
-static int kernel_tracer_fd = -1;
-static int kernel_tracer_event_notifier_group_fd = -1;
-static int kernel_tracer_event_notifier_group_notification_fd = -1;
-static struct cds_lfht *kernel_token_to_event_notifier_rule_ht;
-
-/*
- * Add context on a kernel channel.
- *
- * Assumes the ownership of ctx.
- */
-int kernel_add_channel_context(struct ltt_kernel_channel *chan,
-               struct ltt_kernel_context *ctx)
-{
-       int ret;
-
-       LTTNG_ASSERT(chan);
-       LTTNG_ASSERT(ctx);
-
-       DBG("Adding context to channel %s", chan->channel->name);
-       ret = kernctl_add_context(chan->fd, &ctx->ctx);
-       if (ret < 0) {
-               switch (-ret) {
-               case ENOSYS:
-                       /* Exists but not available for this kernel */
-                       ret = LTTNG_ERR_KERN_CONTEXT_UNAVAILABLE;
-                       goto error;
-               case EEXIST:
-                       /* If EEXIST, we just ignore the error */
-                       ret = 0;
-                       goto end;
-               default:
-                       PERROR("add context ioctl");
-                       ret = LTTNG_ERR_KERN_CONTEXT_FAIL;
-                       goto error;
-               }
-       }
-       ret = 0;
-
-end:
-       cds_list_add_tail(&ctx->list, &chan->ctx_list);
-       ctx->in_list = true;
-       ctx = NULL;
-error:
-       if (ctx) {
-               trace_kernel_destroy_context(ctx);
-       }
-       return ret;
-}
-
-/*
- * Create a new kernel session, register it to the kernel tracer and add it to
- * the session daemon session.
- */
-int kernel_create_session(struct ltt_session *session)
-{
-       int ret;
-       struct ltt_kernel_session *lks;
-
-       LTTNG_ASSERT(session);
-
-       /* Allocate data structure */
-       lks = trace_kernel_create_session();
-       if (lks == NULL) {
-               ret = -1;
-               goto error;
-       }
-
-       /* Kernel tracer session creation */
-       ret = kernctl_create_session(kernel_tracer_fd);
-       if (ret < 0) {
-               PERROR("ioctl kernel create session");
-               goto error;
-       }
-
-       lks->fd = ret;
-       /* Prevent fd duplication after execlp() */
-       ret = fcntl(lks->fd, F_SETFD, FD_CLOEXEC);
-       if (ret < 0) {
-               PERROR("fcntl session fd");
-       }
-
-       lks->id = session->id;
-       lks->consumer_fds_sent = 0;
-       session->kernel_session = lks;
-
-       DBG("Kernel session created (fd: %d)", lks->fd);
-
-       /*
-        * This is necessary since the creation time is present in the session
-        * name when it is generated.
-        */
-       if (session->has_auto_generated_name) {
-               ret = kernctl_session_set_name(lks->fd, DEFAULT_SESSION_NAME);
-       } else {
-               ret = kernctl_session_set_name(lks->fd, session->name);
-       }
-       if (ret) {
-               WARN("Could not set kernel session name for session %" PRIu64 " name: %s",
-                       session->id, session->name);
-       }
-
-       ret = kernctl_session_set_creation_time(lks->fd, session->creation_time);
-       if (ret) {
-               WARN("Could not set kernel session creation time for session %" PRIu64 " name: %s",
-                       session->id, session->name);
-       }
-
-       return 0;
-
-error:
-       if (lks) {
-               trace_kernel_destroy_session(lks);
-               trace_kernel_free_session(lks);
-       }
-       return ret;
-}
-
-/*
- * Create a kernel channel, register it to the kernel tracer and add it to the
- * kernel session.
- */
-int kernel_create_channel(struct ltt_kernel_session *session,
-               struct lttng_channel *chan)
-{
-       int ret;
-       struct ltt_kernel_channel *lkc;
-
-       LTTNG_ASSERT(session);
-       LTTNG_ASSERT(chan);
-
-       /* Allocate kernel channel */
-       lkc = trace_kernel_create_channel(chan);
-       if (lkc == NULL) {
-               goto error;
-       }
-
-       DBG3("Kernel create channel %s with attr: %d, %" PRIu64 ", %" PRIu64 ", %u, %u, %d, %d",
-                       chan->name, lkc->channel->attr.overwrite,
-                       lkc->channel->attr.subbuf_size, lkc->channel->attr.num_subbuf,
-                       lkc->channel->attr.switch_timer_interval, lkc->channel->attr.read_timer_interval,
-                       lkc->channel->attr.live_timer_interval, lkc->channel->attr.output);
-
-       /* Kernel tracer channel creation */
-       ret = kernctl_create_channel(session->fd, &lkc->channel->attr);
-       if (ret < 0) {
-               PERROR("ioctl kernel create channel");
-               goto error;
-       }
-
-       /* Setup the channel fd */
-       lkc->fd = ret;
-       /* Prevent fd duplication after execlp() */
-       ret = fcntl(lkc->fd, F_SETFD, FD_CLOEXEC);
-       if (ret < 0) {
-               PERROR("fcntl session fd");
-       }
-
-       /* Add channel to session */
-       cds_list_add(&lkc->list, &session->channel_list.head);
-       session->channel_count++;
-       lkc->session = session;
-       lkc->key = ++next_kernel_channel_key;
-
-       DBG("Kernel channel %s created (fd: %d, key: %" PRIu64 ")",
-                       lkc->channel->name, lkc->fd, lkc->key);
-
-       return 0;
-
-error:
-       if (lkc) {
-               free(lkc->channel);
-               free(lkc);
-       }
-       return -1;
-}
-
-/*
- * Create a kernel event notifier group, register it to the kernel tracer and
- * add it to the kernel session.
- */
-static int kernel_create_event_notifier_group(int *event_notifier_group_fd)
-{
-       int ret;
-       int local_fd = -1;
-
-       LTTNG_ASSERT(event_notifier_group_fd);
-
-       /* Kernel event notifier group creation. */
-       ret = kernctl_create_event_notifier_group(kernel_tracer_fd);
-       if (ret < 0) {
-               PERROR("Failed to create kernel event notifier group");
-               ret = -1;
-               goto error;
-       }
-
-       local_fd = ret;
-
-       /* Prevent fd duplication after execlp(). */
-       ret = fcntl(local_fd, F_SETFD, FD_CLOEXEC);
-       if (ret < 0) {
-               PERROR("Failed to set FD_CLOEXEC on kernel event notifier group file descriptor: fd = %d",
-                               local_fd);
-               goto error;
-       }
-
-       DBG("Created kernel event notifier group: fd = %d", local_fd);
-       *event_notifier_group_fd = local_fd;
-       local_fd = -1;
-       ret = 0;
-error:
-       if (local_fd >= 0) {
-               ret = close(local_fd);
-               if (ret) {
-                       PERROR("Failed to close kernel event notifier group file descriptor: fd = %d",
-                                       local_fd);
-               }
-       }
-
-       return ret;
-}
-
-/*
- * Compute the offset of the instrumentation byte in the binary based on the
- * function probe location using the ELF lookup method.
- *
- * Returns 0 on success and set the offset out parameter to the offset of the
- * elf symbol
- * Returns -1 on error
- */
-static
-int extract_userspace_probe_offset_function_elf(
-               const struct lttng_userspace_probe_location *probe_location,
-               uid_t uid, gid_t gid, uint64_t *offset)
-{
-       int fd;
-       int ret = 0;
-       const char *symbol = NULL;
-       const struct lttng_userspace_probe_location_lookup_method *lookup = NULL;
-       enum lttng_userspace_probe_location_lookup_method_type lookup_method_type;
-
-       LTTNG_ASSERT(lttng_userspace_probe_location_get_type(probe_location) ==
-                       LTTNG_USERSPACE_PROBE_LOCATION_TYPE_FUNCTION);
-
-       lookup = lttng_userspace_probe_location_get_lookup_method(
-                       probe_location);
-       if (!lookup) {
-               ret = -1;
-               goto end;
-       }
-
-       lookup_method_type =
-                       lttng_userspace_probe_location_lookup_method_get_type(lookup);
-
-       LTTNG_ASSERT(lookup_method_type ==
-                       LTTNG_USERSPACE_PROBE_LOCATION_LOOKUP_METHOD_TYPE_FUNCTION_ELF);
-
-       symbol = lttng_userspace_probe_location_function_get_function_name(
-                       probe_location);
-       if (!symbol) {
-               ret = -1;
-               goto end;
-       }
-
-       fd = lttng_userspace_probe_location_function_get_binary_fd(probe_location);
-       if (fd < 0) {
-               ret = -1;
-               goto end;
-       }
-
-       ret = run_as_extract_elf_symbol_offset(fd, symbol, uid, gid, offset);
-       if (ret < 0) {
-               DBG("userspace probe offset calculation failed for "
-                               "function %s", symbol);
-               goto end;
-       }
-
-       DBG("userspace probe elf offset for %s is 0x%jd", symbol, (intmax_t)(*offset));
-end:
-       return ret;
-}
-
-/*
- * Compute the offsets of the instrumentation bytes in the binary based on the
- * tracepoint probe location using the SDT lookup method. This function
- * allocates the offsets buffer, the caller must free it.
- *
- * Returns 0 on success and set the offset out parameter to the offsets of the
- * SDT tracepoint.
- * Returns -1 on error.
- */
-static
-int extract_userspace_probe_offset_tracepoint_sdt(
-               const struct lttng_userspace_probe_location *probe_location,
-               uid_t uid, gid_t gid, uint64_t **offsets,
-               uint32_t *offsets_count)
-{
-       enum lttng_userspace_probe_location_lookup_method_type lookup_method_type;
-       const struct lttng_userspace_probe_location_lookup_method *lookup = NULL;
-       const char *probe_name = NULL, *provider_name = NULL;
-       int ret = 0;
-       int fd, i;
-
-       LTTNG_ASSERT(lttng_userspace_probe_location_get_type(probe_location) ==
-                       LTTNG_USERSPACE_PROBE_LOCATION_TYPE_TRACEPOINT);
-
-       lookup = lttng_userspace_probe_location_get_lookup_method(probe_location);
-       if (!lookup) {
-               ret = -1;
-               goto end;
-       }
-
-       lookup_method_type =
-                       lttng_userspace_probe_location_lookup_method_get_type(lookup);
-
-       LTTNG_ASSERT(lookup_method_type ==
-                       LTTNG_USERSPACE_PROBE_LOCATION_LOOKUP_METHOD_TYPE_TRACEPOINT_SDT);
-
-
-       probe_name = lttng_userspace_probe_location_tracepoint_get_probe_name(
-                       probe_location);
-       if (!probe_name) {
-               ret = -1;
-               goto end;
-       }
-
-       provider_name = lttng_userspace_probe_location_tracepoint_get_provider_name(
-                       probe_location);
-       if (!provider_name) {
-               ret = -1;
-               goto end;
-       }
-
-       fd = lttng_userspace_probe_location_tracepoint_get_binary_fd(probe_location);
-       if (fd < 0) {
-               ret = -1;
-               goto end;
-       }
-
-       ret = run_as_extract_sdt_probe_offsets(fd, provider_name, probe_name,
-                       uid, gid, offsets, offsets_count);
-       if (ret < 0) {
-               DBG("userspace probe offset calculation failed for sdt "
-                               "probe %s:%s", provider_name, probe_name);
-               goto end;
-       }
-
-       if (*offsets_count == 0) {
-               DBG("no userspace probe offset found");
-               goto end;
-       }
-
-       DBG("%u userspace probe SDT offsets found for %s:%s at:",
-                       *offsets_count, provider_name, probe_name);
-       for (i = 0; i < *offsets_count; i++) {
-               DBG("\t0x%jd", (intmax_t)((*offsets)[i]));
-       }
-end:
-       return ret;
-}
-
-static
-int userspace_probe_add_callsite(
-               const struct lttng_userspace_probe_location *location,
-               uid_t uid, gid_t gid, int fd)
-{
-       const struct lttng_userspace_probe_location_lookup_method *lookup_method = NULL;
-       enum lttng_userspace_probe_location_lookup_method_type type;
-       int ret;
-
-       lookup_method = lttng_userspace_probe_location_get_lookup_method(location);
-       if (!lookup_method) {
-               ret = -1;
-               goto end;
-       }
-
-       type = lttng_userspace_probe_location_lookup_method_get_type(lookup_method);
-       switch (type) {
-       case LTTNG_USERSPACE_PROBE_LOCATION_LOOKUP_METHOD_TYPE_FUNCTION_ELF:
-       {
-               struct lttng_kernel_abi_event_callsite callsite;
-               uint64_t offset;
-
-               ret = extract_userspace_probe_offset_function_elf(location,
-                               uid, gid, &offset);
-               if (ret) {
-                       ret = LTTNG_ERR_PROBE_LOCATION_INVAL;
-                       goto end;
-               }
-
-               callsite.u.uprobe.offset = offset;
-               ret = kernctl_add_callsite(fd, &callsite);
-               if (ret) {
-                       WARN("Failed to add callsite to ELF userspace probe.");
-                       ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       goto end;
-               }
-               break;
-       }
-       case LTTNG_USERSPACE_PROBE_LOCATION_LOOKUP_METHOD_TYPE_TRACEPOINT_SDT:
-       {
-               int i;
-               uint64_t *offsets = NULL;
-               uint32_t offsets_count;
-               struct lttng_kernel_abi_event_callsite callsite;
-
-               /*
-                * This call allocates the offsets buffer. This buffer must be freed
-                * by the caller
-                */
-               ret = extract_userspace_probe_offset_tracepoint_sdt(location,
-                               uid, gid, &offsets, &offsets_count);
-               if (ret) {
-                       ret = LTTNG_ERR_PROBE_LOCATION_INVAL;
-                       goto end;
-               }
-               for (i = 0; i < offsets_count; i++) {
-                       callsite.u.uprobe.offset = offsets[i];
-                       ret = kernctl_add_callsite(fd, &callsite);
-                       if (ret) {
-                               WARN("Failed to add callsite to SDT userspace probe");
-                               ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                               free(offsets);
-                               goto end;
-                       }
-               }
-               free(offsets);
-               break;
-       }
-       default:
-               ret = LTTNG_ERR_PROBE_LOCATION_INVAL;
-               goto end;
-       }
-end:
-       return ret;
-}
-
-/*
- * Extract the offsets of the instrumentation point for the different lookup
- * methods.
- */
-static
-int userspace_probe_event_add_callsites(struct lttng_event *ev,
-                       struct ltt_kernel_session *session, int fd)
-{
-       int ret;
-       const struct lttng_userspace_probe_location *location = NULL;
-
-       LTTNG_ASSERT(ev);
-       LTTNG_ASSERT(ev->type == LTTNG_EVENT_USERSPACE_PROBE);
-
-       location = lttng_event_get_userspace_probe_location(ev);
-       if (!location) {
-               ret = -1;
-               goto end;
-       }
-
-       ret = userspace_probe_add_callsite(location, session->uid, session->gid,
-               fd);
-       if (ret) {
-               WARN("Failed to add callsite to userspace probe event '%s'",
-                               ev->name);
-       }
-
-end:
-       return ret;
-}
-
-/*
- * Extract the offsets of the instrumentation point for the different look-up
- * methods.
- */
-static int userspace_probe_event_rule_add_callsites(
-               const struct lttng_event_rule *rule,
-               const struct lttng_credentials *creds,
-               int fd)
-{
-       int ret;
-       enum lttng_event_rule_status status;
-       enum lttng_event_rule_type event_rule_type;
-       const struct lttng_userspace_probe_location *location = NULL;
-
-       LTTNG_ASSERT(rule);
-       LTTNG_ASSERT(creds);
-
-       event_rule_type = lttng_event_rule_get_type(rule);
-       LTTNG_ASSERT(event_rule_type == LTTNG_EVENT_RULE_TYPE_KERNEL_UPROBE);
-
-       status = lttng_event_rule_kernel_uprobe_get_location(rule, &location);
-       if (status != LTTNG_EVENT_RULE_STATUS_OK || !location) {
-               ret = -1;
-               goto end;
-       }
-
-       ret = userspace_probe_add_callsite(location,
-                       lttng_credentials_get_uid(creds),
-                       lttng_credentials_get_gid(creds), fd);
-       if (ret) {
-               WARN("Failed to add callsite to user space probe object: fd = %d",
-                               fd);
-       }
-
-end:
-       return ret;
-}
-
-/*
- * Create a kernel event, enable it to the kernel tracer and add it to the
- * channel event list of the kernel session.
- * We own filter_expression and filter.
- */
-int kernel_create_event(struct lttng_event *ev,
-               struct ltt_kernel_channel *channel,
-               char *filter_expression,
-               struct lttng_bytecode *filter)
-{
-       int err, fd;
-       enum lttng_error_code ret;
-       struct ltt_kernel_event *event;
-
-       LTTNG_ASSERT(ev);
-       LTTNG_ASSERT(channel);
-
-       /* We pass ownership of filter_expression and filter */
-       ret = trace_kernel_create_event(ev, filter_expression,
-                       filter, &event);
-       if (ret != LTTNG_OK) {
-               goto error;
-       }
-
-       fd = kernctl_create_event(channel->fd, event->event);
-       if (fd < 0) {
-               switch (-fd) {
-               case EEXIST:
-                       ret = LTTNG_ERR_KERN_EVENT_EXIST;
-                       break;
-               case ENOSYS:
-                       WARN("Event type not implemented");
-                       ret = LTTNG_ERR_KERN_EVENT_ENOSYS;
-                       break;
-               case ENOENT:
-                       WARN("Event %s not found!", ev->name);
-                       ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       break;
-               default:
-                       ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       PERROR("create event ioctl");
-               }
-               goto free_event;
-       }
-
-       event->type = ev->type;
-       event->fd = fd;
-       /* Prevent fd duplication after execlp() */
-       err = fcntl(event->fd, F_SETFD, FD_CLOEXEC);
-       if (err < 0) {
-               PERROR("fcntl session fd");
-       }
-
-       if (filter) {
-               err = kernctl_filter(event->fd, filter);
-               if (err < 0) {
-                       switch (-err) {
-                       case ENOMEM:
-                               ret = LTTNG_ERR_FILTER_NOMEM;
-                               break;
-                       default:
-                               ret = LTTNG_ERR_FILTER_INVAL;
-                               break;
-                       }
-                       goto filter_error;
-               }
-       }
-
-       if (ev->type == LTTNG_EVENT_USERSPACE_PROBE) {
-               ret = userspace_probe_event_add_callsites(ev, channel->session,
-                               event->fd);
-               if (ret) {
-                       goto add_callsite_error;
-               }
-       }
-
-       err = kernctl_enable(event->fd);
-       if (err < 0) {
-               switch (-err) {
-               case EEXIST:
-                       ret = LTTNG_ERR_KERN_EVENT_EXIST;
-                       break;
-               default:
-                       PERROR("enable kernel event");
-                       ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       break;
-               }
-               goto enable_error;
-       }
-
-       /* Add event to event list */
-       cds_list_add(&event->list, &channel->events_list.head);
-       channel->event_count++;
-
-       DBG("Event %s created (fd: %d)", ev->name, event->fd);
-
-       return 0;
-
-add_callsite_error:
-enable_error:
-filter_error:
-       {
-               int closeret;
-
-               closeret = close(event->fd);
-               if (closeret) {
-                       PERROR("close event fd");
-               }
-       }
-free_event:
-       free(event);
-error:
-       return ret;
-}
-
-/*
- * Disable a kernel channel.
- */
-int kernel_disable_channel(struct ltt_kernel_channel *chan)
-{
-       int ret;
-
-       LTTNG_ASSERT(chan);
-
-       ret = kernctl_disable(chan->fd);
-       if (ret < 0) {
-               PERROR("disable chan ioctl");
-               goto error;
-       }
-
-       chan->enabled = 0;
-       DBG("Kernel channel %s disabled (fd: %d, key: %" PRIu64 ")",
-                       chan->channel->name, chan->fd, chan->key);
-
-       return 0;
-
-error:
-       return ret;
-}
-
-/*
- * Enable a kernel channel.
- */
-int kernel_enable_channel(struct ltt_kernel_channel *chan)
-{
-       int ret;
-
-       LTTNG_ASSERT(chan);
-
-       ret = kernctl_enable(chan->fd);
-       if (ret < 0 && ret != -EEXIST) {
-               PERROR("Enable kernel chan");
-               goto error;
-       }
-
-       chan->enabled = 1;
-       DBG("Kernel channel %s enabled (fd: %d, key: %" PRIu64 ")",
-                       chan->channel->name, chan->fd, chan->key);
-
-       return 0;
-
-error:
-       return ret;
-}
-
-/*
- * Enable a kernel event.
- */
-int kernel_enable_event(struct ltt_kernel_event *event)
-{
-       int ret;
-
-       LTTNG_ASSERT(event);
-
-       ret = kernctl_enable(event->fd);
-       if (ret < 0) {
-               switch (-ret) {
-               case EEXIST:
-                       ret = LTTNG_ERR_KERN_EVENT_EXIST;
-                       break;
-               default:
-                       PERROR("enable kernel event");
-                       break;
-               }
-               goto error;
-       }
-
-       event->enabled = 1;
-       DBG("Kernel event %s enabled (fd: %d)", event->event->name, event->fd);
-
-       return 0;
-
-error:
-       return ret;
-}
-
-/*
- * Disable a kernel event.
- */
-int kernel_disable_event(struct ltt_kernel_event *event)
-{
-       int ret;
-
-       LTTNG_ASSERT(event);
-
-       ret = kernctl_disable(event->fd);
-       if (ret < 0) {
-               PERROR("Failed to disable kernel event: name = '%s', fd = %d",
-                               event->event->name, event->fd);
-               goto error;
-       }
-
-       event->enabled = 0;
-       DBG("Kernel event %s disabled (fd: %d)", event->event->name, event->fd);
-
-       return 0;
-
-error:
-       return ret;
-}
-
-/*
- * Disable a kernel event notifier.
- */
-static
-int kernel_disable_event_notifier_rule(struct ltt_kernel_event_notifier_rule *event)
-{
-       int ret;
-
-       LTTNG_ASSERT(event);
-
-       rcu_read_lock();
-       cds_lfht_del(kernel_token_to_event_notifier_rule_ht, &event->ht_node);
-       rcu_read_unlock();
-
-       ret = kernctl_disable(event->fd);
-       if (ret < 0) {
-               PERROR("Failed to disable kernel event notifier: fd = %d, token = %" PRIu64,
-                               event->fd, event->token);
-               goto error;
-       }
-
-       event->enabled = 0;
-       DBG("Disabled kernel event notifier: fd = %d, token = %" PRIu64,
-                       event->fd, event->token);
-
-error:
-       return ret;
-}
-
-static
-struct process_attr_tracker *_kernel_get_process_attr_tracker(
-               struct ltt_kernel_session *session,
-               enum lttng_process_attr process_attr)
-{
-       switch (process_attr) {
-       case LTTNG_PROCESS_ATTR_PROCESS_ID:
-               return session->tracker_pid;
-       case LTTNG_PROCESS_ATTR_VIRTUAL_PROCESS_ID:
-               return session->tracker_vpid;
-       case LTTNG_PROCESS_ATTR_USER_ID:
-               return session->tracker_uid;
-       case LTTNG_PROCESS_ATTR_VIRTUAL_USER_ID:
-               return session->tracker_vuid;
-       case LTTNG_PROCESS_ATTR_GROUP_ID:
-               return session->tracker_gid;
-       case LTTNG_PROCESS_ATTR_VIRTUAL_GROUP_ID:
-               return session->tracker_vgid;
-       default:
-               return NULL;
-       }
-}
-
-const struct process_attr_tracker *kernel_get_process_attr_tracker(
-               struct ltt_kernel_session *session,
-               enum lttng_process_attr process_attr)
-{
-       return (const struct process_attr_tracker *)
-                       _kernel_get_process_attr_tracker(session, process_attr);
-}
-
-enum lttng_error_code kernel_process_attr_tracker_set_tracking_policy(
-               struct ltt_kernel_session *session,
-               enum lttng_process_attr process_attr,
-               enum lttng_tracking_policy policy)
-{
-       int ret;
-       enum lttng_error_code ret_code = LTTNG_OK;
-       struct process_attr_tracker *tracker =
-                       _kernel_get_process_attr_tracker(session, process_attr);
-       enum lttng_tracking_policy previous_policy;
-
-       if (!tracker) {
-               ret_code = LTTNG_ERR_INVALID;
-               goto end;
-       }
-
-       previous_policy = process_attr_tracker_get_tracking_policy(tracker);
-       ret = process_attr_tracker_set_tracking_policy(tracker, policy);
-       if (ret) {
-               ret_code = LTTNG_ERR_UNK;
-               goto end;
-       }
-
-       if (previous_policy == policy) {
-               goto end;
-       }
-
-       switch (policy) {
-       case LTTNG_TRACKING_POLICY_INCLUDE_ALL:
-               if (process_attr == LTTNG_PROCESS_ATTR_PROCESS_ID) {
-                       /*
-                        * Maintain a special case for the process ID process
-                        * attribute tracker as it was the only supported
-                        * attribute prior to 2.12.
-                        */
-                       ret = kernctl_track_pid(session->fd, -1);
-               } else {
-                       ret = kernctl_track_id(session->fd, process_attr, -1);
-               }
-               break;
-       case LTTNG_TRACKING_POLICY_EXCLUDE_ALL:
-       case LTTNG_TRACKING_POLICY_INCLUDE_SET:
-               /* fall-through. */
-               if (process_attr == LTTNG_PROCESS_ATTR_PROCESS_ID) {
-                       /*
-                        * Maintain a special case for the process ID process
-                        * attribute tracker as it was the only supported
-                        * attribute prior to 2.12.
-                        */
-                       ret = kernctl_untrack_pid(session->fd, -1);
-               } else {
-                       ret = kernctl_untrack_id(session->fd, process_attr, -1);
-               }
-               break;
-       default:
-               abort();
-       }
-       /* kern-ctl error handling */
-       switch (-ret) {
-       case 0:
-               ret_code = LTTNG_OK;
-               break;
-       case EINVAL:
-               ret_code = LTTNG_ERR_INVALID;
-               break;
-       case ENOMEM:
-               ret_code = LTTNG_ERR_NOMEM;
-               break;
-       case EEXIST:
-               ret_code = LTTNG_ERR_PROCESS_ATTR_EXISTS;
-               break;
-       default:
-               ret_code = LTTNG_ERR_UNK;
-               break;
-       }
-end:
-       return ret_code;
-}
-
-enum lttng_error_code kernel_process_attr_tracker_inclusion_set_add_value(
-               struct ltt_kernel_session *session,
-               enum lttng_process_attr process_attr,
-               const struct process_attr_value *value)
-{
-       int ret, integral_value;
-       enum lttng_error_code ret_code;
-       struct process_attr_tracker *tracker;
-       enum process_attr_tracker_status status;
-
-       /*
-        * Convert process attribute tracker value to the integral
-        * representation required by the kern-ctl API.
-        */
-       switch (process_attr) {
-       case LTTNG_PROCESS_ATTR_PROCESS_ID:
-       case LTTNG_PROCESS_ATTR_VIRTUAL_PROCESS_ID:
-               integral_value = (int) value->value.pid;
-               break;
-       case LTTNG_PROCESS_ATTR_USER_ID:
-       case LTTNG_PROCESS_ATTR_VIRTUAL_USER_ID:
-               if (value->type == LTTNG_PROCESS_ATTR_VALUE_TYPE_USER_NAME) {
-                       uid_t uid;
-
-                       ret_code = utils_user_id_from_name(
-                                       value->value.user_name, &uid);
-                       if (ret_code != LTTNG_OK) {
-                               goto end;
-                       }
-                       integral_value = (int) uid;
-               } else {
-                       integral_value = (int) value->value.uid;
-               }
-               break;
-       case LTTNG_PROCESS_ATTR_GROUP_ID:
-       case LTTNG_PROCESS_ATTR_VIRTUAL_GROUP_ID:
-               if (value->type == LTTNG_PROCESS_ATTR_VALUE_TYPE_GROUP_NAME) {
-                       gid_t gid;
-
-                       ret_code = utils_group_id_from_name(
-                                       value->value.group_name, &gid);
-                       if (ret_code != LTTNG_OK) {
-                               goto end;
-                       }
-                       integral_value = (int) gid;
-               } else {
-                       integral_value = (int) value->value.gid;
-               }
-               break;
-       default:
-               ret_code = LTTNG_ERR_INVALID;
-               goto end;
-       }
-
-       tracker = _kernel_get_process_attr_tracker(session, process_attr);
-       if (!tracker) {
-               ret_code = LTTNG_ERR_INVALID;
-               goto end;
-       }
-
-       status = process_attr_tracker_inclusion_set_add_value(tracker, value);
-       if (status != PROCESS_ATTR_TRACKER_STATUS_OK) {
-               switch (status) {
-               case PROCESS_ATTR_TRACKER_STATUS_EXISTS:
-                       ret_code = LTTNG_ERR_PROCESS_ATTR_EXISTS;
-                       break;
-               case PROCESS_ATTR_TRACKER_STATUS_INVALID_TRACKING_POLICY:
-                       ret_code = LTTNG_ERR_PROCESS_ATTR_TRACKER_INVALID_TRACKING_POLICY;
-                       break;
-               case PROCESS_ATTR_TRACKER_STATUS_ERROR:
-               default:
-                       ret_code = LTTNG_ERR_UNK;
-                       break;
-               }
-               goto end;
-       }
-
-       DBG("Kernel track %s %d for session id %" PRIu64,
-                       lttng_process_attr_to_string(process_attr),
-                       integral_value, session->id);
-       if (process_attr == LTTNG_PROCESS_ATTR_PROCESS_ID) {
-               /*
-                * Maintain a special case for the process ID process attribute
-                * tracker as it was the only supported attribute prior to 2.12.
-                */
-               ret = kernctl_track_pid(session->fd, integral_value);
-       } else {
-               ret = kernctl_track_id(
-                               session->fd, process_attr, integral_value);
-       }
-       if (ret == 0) {
-               ret_code = LTTNG_OK;
-               goto end;
-       }
-
-       kernel_wait_quiescent();
-
-       /* kern-ctl error handling */
-       switch (-ret) {
-       case 0:
-               ret_code = LTTNG_OK;
-               break;
-       case EINVAL:
-               ret_code = LTTNG_ERR_INVALID;
-               break;
-       case ENOMEM:
-               ret_code = LTTNG_ERR_NOMEM;
-               break;
-       case EEXIST:
-               ret_code = LTTNG_ERR_PROCESS_ATTR_EXISTS;
-               break;
-       default:
-               ret_code = LTTNG_ERR_UNK;
-               break;
-       }
-
-       /* Attempt to remove the value from the tracker. */
-       status = process_attr_tracker_inclusion_set_remove_value(
-                       tracker, value);
-       if (status != PROCESS_ATTR_TRACKER_STATUS_OK) {
-               ERR("Failed to roll-back the tracking of kernel %s process attribute %d while handling a kern-ctl error",
-                               lttng_process_attr_to_string(process_attr),
-                               integral_value);
-       }
-end:
-       return ret_code;
-}
-
-enum lttng_error_code kernel_process_attr_tracker_inclusion_set_remove_value(
-               struct ltt_kernel_session *session,
-               enum lttng_process_attr process_attr,
-               const struct process_attr_value *value)
-{
-       int ret, integral_value;
-       enum lttng_error_code ret_code;
-       struct process_attr_tracker *tracker;
-       enum process_attr_tracker_status status;
-
-       /*
-        * Convert process attribute tracker value to the integral
-        * representation required by the kern-ctl API.
-        */
-       switch (process_attr) {
-       case LTTNG_PROCESS_ATTR_PROCESS_ID:
-       case LTTNG_PROCESS_ATTR_VIRTUAL_PROCESS_ID:
-               integral_value = (int) value->value.pid;
-               break;
-       case LTTNG_PROCESS_ATTR_USER_ID:
-       case LTTNG_PROCESS_ATTR_VIRTUAL_USER_ID:
-               if (value->type == LTTNG_PROCESS_ATTR_VALUE_TYPE_USER_NAME) {
-                       uid_t uid;
-
-                       ret_code = utils_user_id_from_name(
-                                       value->value.user_name, &uid);
-                       if (ret_code != LTTNG_OK) {
-                               goto end;
-                       }
-                       integral_value = (int) uid;
-               } else {
-                       integral_value = (int) value->value.uid;
-               }
-               break;
-       case LTTNG_PROCESS_ATTR_GROUP_ID:
-       case LTTNG_PROCESS_ATTR_VIRTUAL_GROUP_ID:
-               if (value->type == LTTNG_PROCESS_ATTR_VALUE_TYPE_GROUP_NAME) {
-                       gid_t gid;
-
-                       ret_code = utils_group_id_from_name(
-                                       value->value.group_name, &gid);
-                       if (ret_code != LTTNG_OK) {
-                               goto end;
-                       }
-                       integral_value = (int) gid;
-               } else {
-                       integral_value = (int) value->value.gid;
-               }
-               break;
-       default:
-               ret_code = LTTNG_ERR_INVALID;
-               goto end;
-       }
-
-       tracker = _kernel_get_process_attr_tracker(session, process_attr);
-       if (!tracker) {
-               ret_code = LTTNG_ERR_INVALID;
-               goto end;
-       }
-
-       status = process_attr_tracker_inclusion_set_remove_value(
-                       tracker, value);
-       if (status != PROCESS_ATTR_TRACKER_STATUS_OK) {
-               switch (status) {
-               case PROCESS_ATTR_TRACKER_STATUS_MISSING:
-                       ret_code = LTTNG_ERR_PROCESS_ATTR_MISSING;
-                       break;
-               case PROCESS_ATTR_TRACKER_STATUS_INVALID_TRACKING_POLICY:
-                       ret_code = LTTNG_ERR_PROCESS_ATTR_TRACKER_INVALID_TRACKING_POLICY;
-                       break;
-               case PROCESS_ATTR_TRACKER_STATUS_ERROR:
-               default:
-                       ret_code = LTTNG_ERR_UNK;
-                       break;
-               }
-               goto end;
-       }
-
-       DBG("Kernel track %s %d for session id %" PRIu64,
-                       lttng_process_attr_to_string(process_attr),
-                       integral_value, session->id);
-       if (process_attr == LTTNG_PROCESS_ATTR_PROCESS_ID) {
-               /*
-                * Maintain a special case for the process ID process attribute
-                * tracker as it was the only supported attribute prior to 2.12.
-                */
-               ret = kernctl_untrack_pid(session->fd, integral_value);
-       } else {
-               ret = kernctl_untrack_id(
-                               session->fd, process_attr, integral_value);
-       }
-       if (ret == 0) {
-               ret_code = LTTNG_OK;
-               goto end;
-       }
-       kernel_wait_quiescent();
-
-       /* kern-ctl error handling */
-       switch (-ret) {
-       case 0:
-               ret_code = LTTNG_OK;
-               break;
-       case EINVAL:
-               ret_code = LTTNG_ERR_INVALID;
-               break;
-       case ENOMEM:
-               ret_code = LTTNG_ERR_NOMEM;
-               break;
-       case ENOENT:
-               ret_code = LTTNG_ERR_PROCESS_ATTR_MISSING;
-               break;
-       default:
-               ret_code = LTTNG_ERR_UNK;
-               break;
-       }
-
-       /* Attempt to add the value to the tracker. */
-       status = process_attr_tracker_inclusion_set_add_value(
-                       tracker, value);
-       if (status != PROCESS_ATTR_TRACKER_STATUS_OK) {
-               ERR("Failed to roll-back the tracking of kernel %s process attribute %d while handling a kern-ctl error",
-                               lttng_process_attr_to_string(process_attr),
-                               integral_value);
-       }
-end:
-       return ret_code;
-}
-
-/*
- * Create kernel metadata, open from the kernel tracer and add it to the
- * kernel session.
- */
-int kernel_open_metadata(struct ltt_kernel_session *session)
-{
-       int ret;
-       struct ltt_kernel_metadata *lkm = NULL;
-
-       LTTNG_ASSERT(session);
-
-       /* Allocate kernel metadata */
-       lkm = trace_kernel_create_metadata();
-       if (lkm == NULL) {
-               goto error;
-       }
-
-       /* Kernel tracer metadata creation */
-       ret = kernctl_open_metadata(session->fd, &lkm->conf->attr);
-       if (ret < 0) {
-               goto error_open;
-       }
-
-       lkm->fd = ret;
-       lkm->key = ++next_kernel_channel_key;
-       /* Prevent fd duplication after execlp() */
-       ret = fcntl(lkm->fd, F_SETFD, FD_CLOEXEC);
-       if (ret < 0) {
-               PERROR("fcntl session fd");
-       }
-
-       session->metadata = lkm;
-
-       DBG("Kernel metadata opened (fd: %d)", lkm->fd);
-
-       return 0;
-
-error_open:
-       trace_kernel_destroy_metadata(lkm);
-error:
-       return -1;
-}
-
-/*
- * Start tracing session.
- */
-int kernel_start_session(struct ltt_kernel_session *session)
-{
-       int ret;
-
-       LTTNG_ASSERT(session);
-
-       ret = kernctl_start_session(session->fd);
-       if (ret < 0) {
-               PERROR("ioctl start session");
-               goto error;
-       }
-
-       DBG("Kernel session started");
-
-       return 0;
-
-error:
-       return ret;
-}
-
-/*
- * Make a kernel wait to make sure in-flight probe have completed.
- */
-void kernel_wait_quiescent(void)
-{
-       int ret;
-       int fd = kernel_tracer_fd;
-
-       DBG("Kernel quiescent wait on %d", fd);
-
-       ret = kernctl_wait_quiescent(fd);
-       if (ret < 0) {
-               PERROR("wait quiescent ioctl");
-               ERR("Kernel quiescent wait failed");
-       }
-}
-
-/*
- *  Force flush buffer of metadata.
- */
-int kernel_metadata_flush_buffer(int fd)
-{
-       int ret;
-
-       DBG("Kernel flushing metadata buffer on fd %d", fd);
-
-       ret = kernctl_buffer_flush(fd);
-       if (ret < 0) {
-               ERR("Fail to flush metadata buffers %d (ret: %d)", fd, ret);
-       }
-
-       return 0;
-}
-
-/*
- * Force flush buffer for channel.
- */
-int kernel_flush_buffer(struct ltt_kernel_channel *channel)
-{
-       int ret;
-       struct ltt_kernel_stream *stream;
-
-       LTTNG_ASSERT(channel);
-
-       DBG("Flush buffer for channel %s", channel->channel->name);
-
-       cds_list_for_each_entry(stream, &channel->stream_list.head, list) {
-               DBG("Flushing channel stream %d", stream->fd);
-               ret = kernctl_buffer_flush(stream->fd);
-               if (ret < 0) {
-                       PERROR("ioctl");
-                       ERR("Fail to flush buffer for stream %d (ret: %d)",
-                                       stream->fd, ret);
-               }
-       }
-
-       return 0;
-}
-
-/*
- * Stop tracing session.
- */
-int kernel_stop_session(struct ltt_kernel_session *session)
-{
-       int ret;
-
-       LTTNG_ASSERT(session);
-
-       ret = kernctl_stop_session(session->fd);
-       if (ret < 0) {
-               goto error;
-       }
-
-       DBG("Kernel session stopped");
-
-       return 0;
-
-error:
-       return ret;
-}
-
-/*
- * Open stream of channel, register it to the kernel tracer and add it
- * to the stream list of the channel.
- *
- * Note: given that the streams may appear in random order wrt CPU
- * number (e.g. cpu hotplug), the index value of the stream number in
- * the stream name is not necessarily linked to the CPU number.
- *
- * Return the number of created stream. Else, a negative value.
- */
-int kernel_open_channel_stream(struct ltt_kernel_channel *channel)
-{
-       int ret;
-       struct ltt_kernel_stream *lks;
-
-       LTTNG_ASSERT(channel);
-
-       while ((ret = kernctl_create_stream(channel->fd)) >= 0) {
-               lks = trace_kernel_create_stream(channel->channel->name,
-                               channel->stream_count);
-               if (lks == NULL) {
-                       ret = close(ret);
-                       if (ret) {
-                               PERROR("close");
-                       }
-                       goto error;
-               }
-
-               lks->fd = ret;
-               /* Prevent fd duplication after execlp() */
-               ret = fcntl(lks->fd, F_SETFD, FD_CLOEXEC);
-               if (ret < 0) {
-                       PERROR("fcntl session fd");
-               }
-
-               lks->tracefile_size = channel->channel->attr.tracefile_size;
-               lks->tracefile_count = channel->channel->attr.tracefile_count;
-
-               /* Add stream to channel stream list */
-               cds_list_add(&lks->list, &channel->stream_list.head);
-               channel->stream_count++;
-
-               DBG("Kernel stream %s created (fd: %d, state: %d)", lks->name, lks->fd,
-                               lks->state);
-       }
-
-       return channel->stream_count;
-
-error:
-       return -1;
-}
-
-/*
- * Open the metadata stream and set it to the kernel session.
- */
-int kernel_open_metadata_stream(struct ltt_kernel_session *session)
-{
-       int ret;
-
-       LTTNG_ASSERT(session);
-
-       ret = kernctl_create_stream(session->metadata->fd);
-       if (ret < 0) {
-               PERROR("kernel create metadata stream");
-               goto error;
-       }
-
-       DBG("Kernel metadata stream created (fd: %d)", ret);
-       session->metadata_stream_fd = ret;
-       /* Prevent fd duplication after execlp() */
-       ret = fcntl(session->metadata_stream_fd, F_SETFD, FD_CLOEXEC);
-       if (ret < 0) {
-               PERROR("fcntl session fd");
-       }
-
-       return 0;
-
-error:
-       return -1;
-}
-
-/*
- * Get the event list from the kernel tracer and return the number of elements.
- */
-ssize_t kernel_list_events(struct lttng_event **events)
-{
-       int fd, ret;
-       char *event;
-       size_t nbmem, count = 0;
-       FILE *fp;
-       struct lttng_event *elist;
-
-       LTTNG_ASSERT(events);
-
-       fd = kernctl_tracepoint_list(kernel_tracer_fd);
-       if (fd < 0) {
-               PERROR("kernel tracepoint list");
-               goto error;
-       }
-
-       fp = fdopen(fd, "r");
-       if (fp == NULL) {
-               PERROR("kernel tracepoint list fdopen");
-               goto error_fp;
-       }
-
-       /*
-        * Init memory size counter
-        * See kernel-ctl.h for explanation of this value
-        */
-       nbmem = KERNEL_EVENT_INIT_LIST_SIZE;
-       elist = zmalloc(sizeof(struct lttng_event) * nbmem);
-       if (elist == NULL) {
-               PERROR("alloc list events");
-               count = -ENOMEM;
-               goto end;
-       }
-
-       while (fscanf(fp, "event { name = %m[^;]; };\n", &event) == 1) {
-               if (count >= nbmem) {
-                       struct lttng_event *new_elist;
-                       size_t new_nbmem;
-
-                       new_nbmem = nbmem << 1;
-                       DBG("Reallocating event list from %zu to %zu bytes",
-                                       nbmem, new_nbmem);
-                       new_elist = realloc(elist, new_nbmem * sizeof(struct lttng_event));
-                       if (new_elist == NULL) {
-                               PERROR("realloc list events");
-                               free(event);
-                               free(elist);
-                               count = -ENOMEM;
-                               goto end;
-                       }
-                       /* Zero the new memory */
-                       memset(new_elist + nbmem, 0,
-                               (new_nbmem - nbmem) * sizeof(struct lttng_event));
-                       nbmem = new_nbmem;
-                       elist = new_elist;
-               }
-               strncpy(elist[count].name, event, LTTNG_SYMBOL_NAME_LEN);
-               elist[count].name[LTTNG_SYMBOL_NAME_LEN - 1] = '\0';
-               elist[count].enabled = -1;
-               count++;
-               free(event);
-       }
-
-       *events = elist;
-       DBG("Kernel list events done (%zu events)", count);
-end:
-       ret = fclose(fp);       /* closes both fp and fd */
-       if (ret) {
-               PERROR("fclose");
-       }
-       return count;
-
-error_fp:
-       ret = close(fd);
-       if (ret) {
-               PERROR("close");
-       }
-error:
-       return -1;
-}
-
-/*
- * Get kernel version and validate it.
- */
-int kernel_validate_version(struct lttng_kernel_abi_tracer_version *version,
-               struct lttng_kernel_abi_tracer_abi_version *abi_version)
-{
-       int ret;
-
-       ret = kernctl_tracer_version(kernel_tracer_fd, version);
-       if (ret < 0) {
-               ERR("Failed to retrieve the lttng-modules version");
-               goto error;
-       }
-
-       /* Validate version */
-       if (version->major != VERSION_MAJOR) {
-               ERR("Kernel tracer major version (%d) is not compatible with lttng-tools major version (%d)",
-                       version->major, VERSION_MAJOR);
-               goto error_version;
-       }
-       ret = kernctl_tracer_abi_version(kernel_tracer_fd, abi_version);
-       if (ret < 0) {
-               ERR("Failed to retrieve lttng-modules ABI version");
-               goto error;
-       }
-       if (abi_version->major != LTTNG_KERNEL_ABI_MAJOR_VERSION) {
-               ERR("Kernel tracer ABI version (%d.%d) does not match the expected ABI major version (%d.*)",
-                       abi_version->major, abi_version->minor,
-                       LTTNG_KERNEL_ABI_MAJOR_VERSION);
-               goto error;
-       }
-       DBG2("Kernel tracer version validated (%d.%d, ABI %d.%d)",
-                       version->major, version->minor,
-                       abi_version->major, abi_version->minor);
-       return 0;
-
-error_version:
-       ret = -1;
-
-error:
-       ERR("Kernel tracer version check failed; kernel tracing will not be available");
-       return ret;
-}
-
-/*
- * Kernel work-arounds called at the start of sessiond main().
- */
-int init_kernel_workarounds(void)
-{
-       int ret;
-       FILE *fp;
-
-       /*
-        * boot_id needs to be read once before being used concurrently
-        * to deal with a Linux kernel race. A fix is proposed for
-        * upstream, but the work-around is needed for older kernels.
-        */
-       fp = fopen("/proc/sys/kernel/random/boot_id", "r");
-       if (!fp) {
-               goto end_boot_id;
-       }
-       while (!feof(fp)) {
-               char buf[37] = "";
-
-               ret = fread(buf, 1, sizeof(buf), fp);
-               if (ret < 0) {
-                       /* Ignore error, we don't really care */
-               }
-       }
-       ret = fclose(fp);
-       if (ret) {
-               PERROR("fclose");
-       }
-end_boot_id:
-       return 0;
-}
-
-/*
- * Teardown of a kernel session, keeping data required by destroy notifiers.
- */
-void kernel_destroy_session(struct ltt_kernel_session *ksess)
-{
-       struct lttng_trace_chunk *trace_chunk;
-
-       if (ksess == NULL) {
-               DBG3("No kernel session when tearing down session");
-               return;
-       }
-
-       DBG("Tearing down kernel session");
-       trace_chunk = ksess->current_trace_chunk;
-
-       /*
-        * Destroy channels on the consumer if at least one FD has been sent and we
-        * are in no output mode because the streams are in *no* monitor mode so we
-        * have to send a command to clean them up or else they leaked.
-        */
-       if (!ksess->output_traces && ksess->consumer_fds_sent) {
-               int ret;
-               struct consumer_socket *socket;
-               struct lttng_ht_iter iter;
-
-               /* For each consumer socket. */
-               rcu_read_lock();
-               cds_lfht_for_each_entry(ksess->consumer->socks->ht, &iter.iter,
-                               socket, node.node) {
-                       struct ltt_kernel_channel *chan;
-
-                       /* For each channel, ask the consumer to destroy it. */
-                       cds_list_for_each_entry(chan, &ksess->channel_list.head, list) {
-                               ret = kernel_consumer_destroy_channel(socket, chan);
-                               if (ret < 0) {
-                                       /* Consumer is probably dead. Use next socket. */
-                                       continue;
-                               }
-                       }
-               }
-               rcu_read_unlock();
-       }
-
-       /* Close any relayd session */
-       consumer_output_send_destroy_relayd(ksess->consumer);
-
-       trace_kernel_destroy_session(ksess);
-       lttng_trace_chunk_put(trace_chunk);
-}
-
-/* Teardown of data required by destroy notifiers. */
-void kernel_free_session(struct ltt_kernel_session *ksess)
-{
-       if (ksess == NULL) {
-               return;
-       }
-       trace_kernel_free_session(ksess);
-}
-
-/*
- * Destroy a kernel channel object. It does not do anything on the tracer side.
- */
-void kernel_destroy_channel(struct ltt_kernel_channel *kchan)
-{
-       struct ltt_kernel_session *ksess = NULL;
-
-       LTTNG_ASSERT(kchan);
-       LTTNG_ASSERT(kchan->channel);
-
-       DBG3("Kernel destroy channel %s", kchan->channel->name);
-
-       /* Update channel count of associated session. */
-       if (kchan->session) {
-               /* Keep pointer reference so we can update it after the destroy. */
-               ksess = kchan->session;
-       }
-
-       trace_kernel_destroy_channel(kchan);
-
-       /*
-        * At this point the kernel channel is not visible anymore. This is safe
-        * since in order to work on a visible kernel session, the tracing session
-        * lock (ltt_session.lock) MUST be acquired.
-        */
-       if (ksess) {
-               ksess->channel_count--;
-       }
-}
-
-/*
- * Take a snapshot for a given kernel session.
- *
- * Return LTTNG_OK on success or else return a LTTNG_ERR code.
- */
-enum lttng_error_code kernel_snapshot_record(
-               struct ltt_kernel_session *ksess,
-               const struct consumer_output *output, int wait,
-               uint64_t nb_packets_per_stream)
-{
-       int err, ret, saved_metadata_fd;
-       enum lttng_error_code status = LTTNG_OK;
-       struct consumer_socket *socket;
-       struct lttng_ht_iter iter;
-       struct ltt_kernel_metadata *saved_metadata;
-       char *trace_path = NULL;
-       size_t consumer_path_offset = 0;
-
-       LTTNG_ASSERT(ksess);
-       LTTNG_ASSERT(ksess->consumer);
-       LTTNG_ASSERT(output);
-
-       DBG("Kernel snapshot record started");
-
-       /* Save current metadata since the following calls will change it. */
-       saved_metadata = ksess->metadata;
-       saved_metadata_fd = ksess->metadata_stream_fd;
-
-       rcu_read_lock();
-
-       ret = kernel_open_metadata(ksess);
-       if (ret < 0) {
-               status = LTTNG_ERR_KERN_META_FAIL;
-               goto error;
-       }
-
-       ret = kernel_open_metadata_stream(ksess);
-       if (ret < 0) {
-               status = LTTNG_ERR_KERN_META_FAIL;
-               goto error_open_stream;
-       }
-
-       trace_path = setup_channel_trace_path(ksess->consumer,
-                       "", &consumer_path_offset);
-       if (!trace_path) {
-               status = LTTNG_ERR_INVALID;
-               goto error;
-       }
-       /* Send metadata to consumer and snapshot everything. */
-       cds_lfht_for_each_entry(output->socks->ht, &iter.iter,
-                       socket, node.node) {
-               struct ltt_kernel_channel *chan;
-
-               pthread_mutex_lock(socket->lock);
-               /* This stream must not be monitored by the consumer. */
-               ret = kernel_consumer_add_metadata(socket, ksess, 0);
-               pthread_mutex_unlock(socket->lock);
-               if (ret < 0) {
-                       status = LTTNG_ERR_KERN_META_FAIL;
-                       goto error_consumer;
-               }
-
-               /* For each channel, ask the consumer to snapshot it. */
-               cds_list_for_each_entry(chan, &ksess->channel_list.head, list) {
-                       status = consumer_snapshot_channel(socket, chan->key, output, 0,
-                                       ksess->uid, ksess->gid,
-                                       &trace_path[consumer_path_offset], wait,
-                                       nb_packets_per_stream);
-                       if (status != LTTNG_OK) {
-                               (void) kernel_consumer_destroy_metadata(socket,
-                                               ksess->metadata);
-                               goto error_consumer;
-                       }
-               }
-
-               /* Snapshot metadata, */
-               status = consumer_snapshot_channel(socket, ksess->metadata->key, output,
-                               1, ksess->uid, ksess->gid, &trace_path[consumer_path_offset],
-                               wait, 0);
-               if (status != LTTNG_OK) {
-                       goto error_consumer;
-               }
-
-               /*
-                * The metadata snapshot is done, ask the consumer to destroy it since
-                * it's not monitored on the consumer side.
-                */
-               (void) kernel_consumer_destroy_metadata(socket, ksess->metadata);
-       }
-
-error_consumer:
-       /* Close newly opened metadata stream. It's now on the consumer side. */
-       err = close(ksess->metadata_stream_fd);
-       if (err < 0) {
-               PERROR("close snapshot kernel");
-       }
-
-error_open_stream:
-       trace_kernel_destroy_metadata(ksess->metadata);
-error:
-       /* Restore metadata state.*/
-       ksess->metadata = saved_metadata;
-       ksess->metadata_stream_fd = saved_metadata_fd;
-       rcu_read_unlock();
-       free(trace_path);
-       return status;
-}
-
-/*
- * Get the syscall mask array from the kernel tracer.
- *
- * Return 0 on success else a negative value. In both case, syscall_mask should
- * be freed.
- */
-int kernel_syscall_mask(int chan_fd, char **syscall_mask, uint32_t *nr_bits)
-{
-       LTTNG_ASSERT(syscall_mask);
-       LTTNG_ASSERT(nr_bits);
-
-       return kernctl_syscall_mask(chan_fd, syscall_mask, nr_bits);
-}
-
-static
-int kernel_tracer_abi_greater_or_equal(unsigned int major, unsigned int minor)
-{
-       int ret;
-       struct lttng_kernel_abi_tracer_abi_version abi;
-
-       ret = kernctl_tracer_abi_version(kernel_tracer_fd, &abi);
-       if (ret < 0) {
-               ERR("Failed to retrieve lttng-modules ABI version");
-               goto error;
-       }
-
-       ret = abi.major > major || (abi.major == major && abi.minor >= minor);
-error:
-       return ret;
-}
-
-/*
- * Check for the support of the RING_BUFFER_SNAPSHOT_SAMPLE_POSITIONS via abi
- * version number.
- *
- * Return 1 on success, 0 when feature is not supported, negative value in case
- * of errors.
- */
-int kernel_supports_ring_buffer_snapshot_sample_positions(void)
-{
-       /*
-        * RING_BUFFER_SNAPSHOT_SAMPLE_POSITIONS was introduced in 2.3
-        */
-       return kernel_tracer_abi_greater_or_equal(2, 3);
-}
-
-/*
- * Check for the support of the packet sequence number via abi version number.
- *
- * Return 1 on success, 0 when feature is not supported, negative value in case
- * of errors.
- */
-int kernel_supports_ring_buffer_packet_sequence_number(void)
-{
-       /*
-        * Packet sequence number was introduced in LTTng 2.8,
-        * lttng-modules ABI 2.1.
-        */
-       return kernel_tracer_abi_greater_or_equal(2, 1);
-}
-
-/*
- * Check for the support of event notifiers via abi version number.
- *
- * Return 1 on success, 0 when feature is not supported, negative value in case
- * of errors.
- */
-int kernel_supports_event_notifiers(void)
-{
-       /*
-        * Event notifiers were introduced in LTTng 2.13, lttng-modules ABI 2.6.
-        */
-       return kernel_tracer_abi_greater_or_equal(2, 6);
-}
-
-/*
- * Rotate a kernel session.
- *
- * Return LTTNG_OK on success or else an LTTng error code.
- */
-enum lttng_error_code kernel_rotate_session(struct ltt_session *session)
-{
-       int ret;
-       enum lttng_error_code status = LTTNG_OK;
-       struct consumer_socket *socket;
-       struct lttng_ht_iter iter;
-       struct ltt_kernel_session *ksess = session->kernel_session;
-
-       LTTNG_ASSERT(ksess);
-       LTTNG_ASSERT(ksess->consumer);
-
-       DBG("Rotate kernel session %s started (session %" PRIu64 ")",
-                       session->name, session->id);
-
-       rcu_read_lock();
-
-       /*
-        * Note that this loop will end after one iteration given that there is
-        * only one kernel consumer.
-        */
-       cds_lfht_for_each_entry(ksess->consumer->socks->ht, &iter.iter,
-                       socket, node.node) {
-               struct ltt_kernel_channel *chan;
-
-               /* For each channel, ask the consumer to rotate it. */
-               cds_list_for_each_entry(chan, &ksess->channel_list.head, list) {
-                       DBG("Rotate kernel channel %" PRIu64 ", session %s",
-                                       chan->key, session->name);
-                       ret = consumer_rotate_channel(socket, chan->key,
-                                       ksess->uid, ksess->gid, ksess->consumer,
-                                       /* is_metadata_channel */ false);
-                       if (ret < 0) {
-                               status = LTTNG_ERR_ROTATION_FAIL_CONSUMER;
-                               goto error;
-                       }
-               }
-
-               /*
-                * Rotate the metadata channel.
-                */
-               ret = consumer_rotate_channel(socket, ksess->metadata->key,
-                               ksess->uid, ksess->gid, ksess->consumer,
-                               /* is_metadata_channel */ true);
-               if (ret < 0) {
-                       status = LTTNG_ERR_ROTATION_FAIL_CONSUMER;
-                       goto error;
-               }
-       }
-
-error:
-       rcu_read_unlock();
-       return status;
-}
-
-enum lttng_error_code kernel_create_channel_subdirectories(
-               const struct ltt_kernel_session *ksess)
-{
-       enum lttng_error_code ret = LTTNG_OK;
-       enum lttng_trace_chunk_status chunk_status;
-
-       rcu_read_lock();
-       LTTNG_ASSERT(ksess->current_trace_chunk);
-
-       /*
-        * Create the index subdirectory which will take care
-        * of implicitly creating the channel's path.
-        */
-       chunk_status = lttng_trace_chunk_create_subdirectory(
-                       ksess->current_trace_chunk,
-                       DEFAULT_KERNEL_TRACE_DIR "/" DEFAULT_INDEX_DIR);
-       if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) {
-               ret = LTTNG_ERR_CREATE_DIR_FAIL;
-               goto error;
-       }
-error:
-       rcu_read_unlock();
-       return ret;
-}
-
-/*
- * Setup necessary data for kernel tracer action.
- */
-int init_kernel_tracer(void)
-{
-       int ret;
-       bool is_root = !getuid();
-
-       /* Modprobe lttng kernel modules */
-       ret = modprobe_lttng_control();
-       if (ret < 0) {
-               goto error;
-       }
-
-       /* Open debugfs lttng */
-       kernel_tracer_fd = open(module_proc_lttng, O_RDWR);
-       if (kernel_tracer_fd < 0) {
-               DBG("Failed to open %s", module_proc_lttng);
-               goto error_open;
-       }
-
-       /* Validate kernel version */
-       ret = kernel_validate_version(&the_kernel_tracer_version,
-                       &the_kernel_tracer_abi_version);
-       if (ret < 0) {
-               goto error_version;
-       }
-
-       ret = modprobe_lttng_data();
-       if (ret < 0) {
-               goto error_modules;
-       }
-
-       ret = kernel_supports_ring_buffer_snapshot_sample_positions();
-       if (ret < 0) {
-               goto error_modules;
-       }
-       if (ret < 1) {
-               WARN("Kernel tracer does not support buffer monitoring. "
-                       "The monitoring timer of channels in the kernel domain "
-                       "will be set to 0 (disabled).");
-       }
-
-       ret = kernel_supports_event_notifiers();
-       if (ret < 0) {
-               ERR("Failed to check for kernel tracer event notifier support");
-               goto error_modules;
-       }
-       ret = kernel_create_event_notifier_group(&kernel_tracer_event_notifier_group_fd);
-       if (ret < 0) {
-               /* This is not fatal. */
-               WARN("Failed to create kernel event notifier group");
-               kernel_tracer_event_notifier_group_fd = -1;
-       } else {
-               enum event_notifier_error_accounting_status error_accounting_status;
-               enum lttng_error_code error_code_ret =
-                               kernel_create_event_notifier_group_notification_fd(
-                                               &kernel_tracer_event_notifier_group_notification_fd);
-
-               if (error_code_ret != LTTNG_OK) {
-                       goto error_modules;
-               }
-
-               error_accounting_status = event_notifier_error_accounting_register_kernel(
-                               kernel_tracer_event_notifier_group_fd);
-               if (error_accounting_status != EVENT_NOTIFIER_ERROR_ACCOUNTING_STATUS_OK) {
-                       ERR("Failed to initialize event notifier error accounting for kernel tracer");
-                       error_code_ret = LTTNG_ERR_EVENT_NOTIFIER_ERROR_ACCOUNTING;
-                       goto error_modules;
-               }
-
-               kernel_token_to_event_notifier_rule_ht = cds_lfht_new(
-                               DEFAULT_HT_SIZE, 1, 0,
-                               CDS_LFHT_AUTO_RESIZE | CDS_LFHT_ACCOUNTING,
-                               NULL);
-               if (!kernel_token_to_event_notifier_rule_ht) {
-                       goto error_token_ht;
-               }
-       }
-
-       DBG("Kernel tracer initialized: kernel tracer fd = %d, event notifier group fd = %d, event notifier group notification fd = %d",
-                       kernel_tracer_fd, kernel_tracer_event_notifier_group_fd,
-                       kernel_tracer_event_notifier_group_notification_fd);
-
-       ret = syscall_init_table(kernel_tracer_fd);
-       if (ret < 0) {
-               ERR("Unable to populate syscall table. Syscall tracing won't "
-                       "work for this session daemon.");
-       }
-
-       return 0;
-
-error_version:
-       modprobe_remove_lttng_control();
-       ret = close(kernel_tracer_fd);
-       if (ret) {
-               PERROR("Failed to close kernel tracer file descriptor: fd = %d",
-                               kernel_tracer_fd);
-       }
-
-       kernel_tracer_fd = -1;
-       return LTTNG_ERR_KERN_VERSION;
-
-
-error_token_ht:
-       ret = close(kernel_tracer_event_notifier_group_notification_fd);
-       if (ret) {
-               PERROR("Failed to close kernel tracer event notifier group notification file descriptor: fd = %d",
-                               kernel_tracer_event_notifier_group_notification_fd);
-       }
-
-       kernel_tracer_event_notifier_group_notification_fd = -1;
-
-error_modules:
-       ret = close(kernel_tracer_event_notifier_group_fd);
-       if (ret) {
-               PERROR("Failed to close kernel tracer event notifier group file descriptor: fd = %d",
-                               kernel_tracer_event_notifier_group_fd);
-       }
-
-       kernel_tracer_event_notifier_group_fd = -1;
-
-       ret = close(kernel_tracer_fd);
-       if (ret) {
-               PERROR("Failed to close kernel tracer file descriptor: fd = %d",
-                               kernel_tracer_fd);
-       }
-
-       kernel_tracer_fd = -1;
-
-error_open:
-       modprobe_remove_lttng_control();
-
-error:
-       WARN("No kernel tracer available");
-       kernel_tracer_fd = -1;
-       if (!is_root) {
-               return LTTNG_ERR_NEED_ROOT_SESSIOND;
-       } else {
-               return LTTNG_ERR_KERN_NA;
-       }
-}
-
-void cleanup_kernel_tracer(void)
-{
-       DBG2("Closing kernel event notifier group notification file descriptor");
-       if (kernel_tracer_event_notifier_group_notification_fd >= 0) {
-               int ret = notification_thread_command_remove_tracer_event_source(
-                               the_notification_thread_handle,
-                               kernel_tracer_event_notifier_group_notification_fd);
-               if (ret != LTTNG_OK) {
-                       ERR("Failed to remove kernel event notifier notification from notification thread");
-               }
-
-               ret = close(kernel_tracer_event_notifier_group_notification_fd);
-               if (ret) {
-                       PERROR("Failed to close kernel event notifier group notification file descriptor: fd = %d",
-                                       kernel_tracer_event_notifier_group_notification_fd);
-               }
-
-               kernel_tracer_event_notifier_group_notification_fd = -1;
-       }
-
-       if (kernel_token_to_event_notifier_rule_ht) {
-               const int ret = cds_lfht_destroy(
-                               kernel_token_to_event_notifier_rule_ht, NULL);
-               LTTNG_ASSERT(ret == 0);
-       }
-
-       DBG2("Closing kernel event notifier group file descriptor");
-       if (kernel_tracer_event_notifier_group_fd >= 0) {
-               const int ret = close(kernel_tracer_event_notifier_group_fd);
-
-               if (ret) {
-                       PERROR("Failed to close kernel event notifier group file descriptor: fd = %d",
-                                       kernel_tracer_event_notifier_group_fd);
-               }
-
-               kernel_tracer_event_notifier_group_fd = -1;
-       }
-
-       DBG2("Closing kernel fd");
-       if (kernel_tracer_fd >= 0) {
-               const int ret = close(kernel_tracer_fd);
-
-               if (ret) {
-                       PERROR("Failed to close kernel tracer file descriptor: fd = %d",
-                                       kernel_tracer_fd);
-               }
-
-               kernel_tracer_fd = -1;
-       }
-
-       free(syscall_table);
-}
-
-bool kernel_tracer_is_initialized(void)
-{
-       return kernel_tracer_fd >= 0;
-}
-
-/*
- *  Clear a kernel session.
- *
- * Return LTTNG_OK on success or else an LTTng error code.
- */
-enum lttng_error_code kernel_clear_session(struct ltt_session *session)
-{
-       int ret;
-       enum lttng_error_code status = LTTNG_OK;
-       struct consumer_socket *socket;
-       struct lttng_ht_iter iter;
-       struct ltt_kernel_session *ksess = session->kernel_session;
-
-       LTTNG_ASSERT(ksess);
-       LTTNG_ASSERT(ksess->consumer);
-
-       DBG("Clear kernel session %s (session %" PRIu64 ")",
-                       session->name, session->id);
-
-       rcu_read_lock();
-
-       if (ksess->active) {
-               ERR("Expecting inactive session %s (%" PRIu64 ")", session->name, session->id);
-               status = LTTNG_ERR_FATAL;
-               goto end;
-       }
-
-       /*
-        * Note that this loop will end after one iteration given that there is
-        * only one kernel consumer.
-        */
-       cds_lfht_for_each_entry(ksess->consumer->socks->ht, &iter.iter,
-                       socket, node.node) {
-               struct ltt_kernel_channel *chan;
-
-               /* For each channel, ask the consumer to clear it. */
-               cds_list_for_each_entry(chan, &ksess->channel_list.head, list) {
-                       DBG("Clear kernel channel %" PRIu64 ", session %s",
-                                       chan->key, session->name);
-                       ret = consumer_clear_channel(socket, chan->key);
-                       if (ret < 0) {
-                               goto error;
-                       }
-               }
-
-               if (!ksess->metadata) {
-                       /*
-                        * Nothing to do for the metadata.
-                        * This is a snapshot session.
-                        * The metadata is genererated on the fly.
-                        */
-                       continue;
-               }
-
-               /*
-                * Clear the metadata channel.
-                * Metadata channel is not cleared per se but we still need to
-                * perform a rotation operation on it behind the scene.
-                */
-               ret = consumer_clear_channel(socket, ksess->metadata->key);
-               if (ret < 0) {
-                       goto error;
-               }
-       }
-
-       goto end;
-error:
-       switch (-ret) {
-       case LTTCOMM_CONSUMERD_RELAYD_CLEAR_DISALLOWED:
-             status = LTTNG_ERR_CLEAR_RELAY_DISALLOWED;
-             break;
-       default:
-             status = LTTNG_ERR_CLEAR_FAIL_CONSUMER;
-             break;
-       }
-end:
-       rcu_read_unlock();
-       return status;
-}
-
-enum lttng_error_code kernel_create_event_notifier_group_notification_fd(
-               int *event_notifier_group_notification_fd)
-{
-       int local_fd = -1, ret;
-       enum lttng_error_code error_code_ret;
-
-       LTTNG_ASSERT(event_notifier_group_notification_fd);
-
-       ret = kernctl_create_event_notifier_group_notification_fd(
-                       kernel_tracer_event_notifier_group_fd);
-       if (ret < 0) {
-               PERROR("Failed to create kernel event notifier group notification file descriptor");
-               error_code_ret = LTTNG_ERR_EVENT_NOTIFIER_GROUP_NOTIFICATION_FD;
-               goto error;
-       }
-
-       local_fd = ret;
-
-       /* Prevent fd duplication after execlp(). */
-       ret = fcntl(local_fd, F_SETFD, FD_CLOEXEC);
-       if (ret < 0) {
-               PERROR("Failed to set FD_CLOEXEC on kernel event notifier group notification file descriptor: fd = %d",
-                               local_fd);
-               error_code_ret = LTTNG_ERR_EVENT_NOTIFIER_GROUP_NOTIFICATION_FD;
-               goto error;
-       }
-
-       DBG("Created kernel notifier group notification file descriptor: fd = %d",
-                       local_fd);
-       error_code_ret = LTTNG_OK;
-       *event_notifier_group_notification_fd = local_fd;
-       local_fd = -1;
-
-error:
-       if (local_fd >= 0) {
-               ret = close(local_fd);
-               if (ret) {
-                       PERROR("Failed to close kernel event notifier group notification file descriptor: fd = %d",
-                                       local_fd);
-               }
-       }
-
-       return error_code_ret;
-}
-
-enum lttng_error_code kernel_destroy_event_notifier_group_notification_fd(
-               int event_notifier_group_notification_fd)
-{
-       enum lttng_error_code ret_code = LTTNG_OK;
-
-       DBG("Closing event notifier group notification file descriptor: fd = %d",
-                       event_notifier_group_notification_fd);
-       if (event_notifier_group_notification_fd >= 0) {
-               const int ret = close(event_notifier_group_notification_fd);
-               if (ret) {
-                       PERROR("Failed to close event notifier group notification file descriptor: fd = %d",
-                                       event_notifier_group_notification_fd);
-               }
-       }
-
-       return ret_code;
-}
-
-static
-unsigned long hash_trigger(const struct lttng_trigger *trigger)
-{
-       const struct lttng_condition *condition =
-                       lttng_trigger_get_const_condition(trigger);
-
-       return lttng_condition_hash(condition);
-}
-
-static
-int match_trigger(struct cds_lfht_node *node, const void *key)
-{
-       const struct ltt_kernel_event_notifier_rule *event_notifier_rule;
-       const struct lttng_trigger *trigger = key;
-
-       event_notifier_rule = caa_container_of(node,
-                       const struct ltt_kernel_event_notifier_rule, ht_node);
-
-       return lttng_trigger_is_equal(trigger, event_notifier_rule->trigger);
-}
-
-static enum lttng_error_code kernel_create_event_notifier_rule(
-               struct lttng_trigger *trigger,
-               const struct lttng_credentials *creds, uint64_t token)
-{
-       int err, fd, ret = 0;
-       enum lttng_error_code error_code_ret;
-       enum lttng_condition_status condition_status;
-       enum lttng_condition_type condition_type;
-       enum lttng_event_rule_type event_rule_type;
-       struct ltt_kernel_event_notifier_rule *event_notifier_rule;
-       struct lttng_kernel_abi_event_notifier kernel_event_notifier = {};
-       unsigned int capture_bytecode_count = 0, i;
-       const struct lttng_condition *condition = NULL;
-       const struct lttng_event_rule *event_rule = NULL;
-       enum lttng_condition_status cond_status;
-
-       LTTNG_ASSERT(trigger);
-
-       condition = lttng_trigger_get_const_condition(trigger);
-       LTTNG_ASSERT(condition);
-
-       condition_type = lttng_condition_get_type(condition);
-       LTTNG_ASSERT(condition_type == LTTNG_CONDITION_TYPE_EVENT_RULE_MATCHES);
-
-       /* Does not acquire a reference. */
-       condition_status = lttng_condition_event_rule_matches_get_rule(
-                       condition, &event_rule);
-       LTTNG_ASSERT(condition_status == LTTNG_CONDITION_STATUS_OK);
-       LTTNG_ASSERT(event_rule);
-
-       event_rule_type = lttng_event_rule_get_type(event_rule);
-       LTTNG_ASSERT(event_rule_type != LTTNG_EVENT_RULE_TYPE_UNKNOWN);
-
-       error_code_ret = trace_kernel_create_event_notifier_rule(trigger, token,
-                       lttng_condition_event_rule_matches_get_error_counter_index(
-                                       condition),
-                       &event_notifier_rule);
-       if (error_code_ret != LTTNG_OK) {
-               goto error;
-       }
-
-       error_code_ret = trace_kernel_init_event_notifier_from_event_rule(
-                       event_rule, &kernel_event_notifier);
-       if (error_code_ret != LTTNG_OK) {
-               goto free_event;
-       }
-
-       kernel_event_notifier.event.token = event_notifier_rule->token;
-       kernel_event_notifier.error_counter_idx =
-                       lttng_condition_event_rule_matches_get_error_counter_index(
-                                       condition);
-
-       fd = kernctl_create_event_notifier(
-                       kernel_tracer_event_notifier_group_fd,
-                       &kernel_event_notifier);
-       if (fd < 0) {
-               switch (-fd) {
-               case EEXIST:
-                       error_code_ret = LTTNG_ERR_KERN_EVENT_EXIST;
-                       break;
-               case ENOSYS:
-                       WARN("Failed to create kernel event notifier: not notifier type not implemented");
-                       error_code_ret = LTTNG_ERR_KERN_EVENT_ENOSYS;
-                       break;
-               case ENOENT:
-                       WARN("Failed to create kernel event notifier: not found: name = '%s'",
-                                       kernel_event_notifier.event.name);
-                       error_code_ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       break;
-               default:
-                       PERROR("Failed to create kernel event notifier: error code = %d, name = '%s'",
-                                       fd, kernel_event_notifier.event.name);
-                       error_code_ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-               }
-               goto free_event;
-       }
-
-       event_notifier_rule->fd = fd;
-       /* Prevent fd duplication after execlp(). */
-       err = fcntl(event_notifier_rule->fd, F_SETFD, FD_CLOEXEC);
-       if (err < 0) {
-               PERROR("Failed to set FD_CLOEXEC on kernel event notifier file descriptor: fd = %d",
-                               fd);
-               error_code_ret = LTTNG_ERR_FATAL;
-               goto set_cloexec_error;
-       }
-
-       if (event_notifier_rule->filter) {
-               err = kernctl_filter(event_notifier_rule->fd, event_notifier_rule->filter);
-               if (err < 0) {
-                       switch (-err) {
-                       case ENOMEM:
-                               error_code_ret = LTTNG_ERR_FILTER_NOMEM;
-                               break;
-                       default:
-                               error_code_ret = LTTNG_ERR_FILTER_INVAL;
-                               break;
-                       }
-                       goto filter_error;
-               }
-       }
-
-       if (lttng_event_rule_get_type(event_rule) ==
-                       LTTNG_EVENT_RULE_TYPE_KERNEL_UPROBE) {
-               ret = userspace_probe_event_rule_add_callsites(
-                               event_rule, creds, event_notifier_rule->fd);
-               if (ret) {
-                       error_code_ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       goto add_callsite_error;
-               }
-       }
-
-       /* Set the capture bytecode if any. */
-       cond_status = lttng_condition_event_rule_matches_get_capture_descriptor_count(
-                       condition, &capture_bytecode_count);
-       LTTNG_ASSERT(cond_status == LTTNG_CONDITION_STATUS_OK);
-
-       for (i = 0; i < capture_bytecode_count; i++) {
-               const struct lttng_bytecode *capture_bytecode =
-                               lttng_condition_event_rule_matches_get_capture_bytecode_at_index(
-                                               condition, i);
-
-               if (capture_bytecode == NULL) {
-                       ERR("Unexpected NULL capture bytecode on condition");
-                       error_code_ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       goto capture_error;
-               }
-
-               ret = kernctl_capture(event_notifier_rule->fd, capture_bytecode);
-               if (ret < 0) {
-                       ERR("Failed to set capture bytecode on event notifier rule fd: fd = %d",
-                                       event_notifier_rule->fd);
-                       error_code_ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       goto capture_error;
-               }
-       }
-
-       err = kernctl_enable(event_notifier_rule->fd);
-       if (err < 0) {
-               switch (-err) {
-               case EEXIST:
-                       error_code_ret = LTTNG_ERR_KERN_EVENT_EXIST;
-                       break;
-               default:
-                       PERROR("enable kernel event notifier");
-                       error_code_ret = LTTNG_ERR_KERN_ENABLE_FAIL;
-                       break;
-               }
-               goto enable_error;
-       }
-
-       /* Add trigger to kernel token mapping in the hash table. */
-       rcu_read_lock();
-       cds_lfht_add(kernel_token_to_event_notifier_rule_ht, hash_trigger(trigger),
-                       &event_notifier_rule->ht_node);
-       rcu_read_unlock();
-
-       DBG("Created kernel event notifier: name = '%s', fd = %d",
-                       kernel_event_notifier.event.name,
-                       event_notifier_rule->fd);
-
-       return LTTNG_OK;
-
-capture_error:
-add_callsite_error:
-enable_error:
-set_cloexec_error:
-filter_error:
-       {
-               const int close_ret = close(event_notifier_rule->fd);
-
-               if (close_ret) {
-                       PERROR("Failed to close kernel event notifier file descriptor: fd = %d",
-                                       event_notifier_rule->fd);
-               }
-       }
-free_event:
-       free(event_notifier_rule);
-error:
-       return error_code_ret;
-}
-
-enum lttng_error_code kernel_register_event_notifier(
-               struct lttng_trigger *trigger,
-               const struct lttng_credentials *cmd_creds)
-{
-       enum lttng_error_code ret;
-       enum lttng_condition_status status;
-       enum lttng_domain_type domain_type;
-       const struct lttng_event_rule *event_rule;
-       const struct lttng_condition *const condition =
-                       lttng_trigger_get_const_condition(trigger);
-       const uint64_t token = lttng_trigger_get_tracer_token(trigger);
-
-       LTTNG_ASSERT(condition);
-
-       /* Does not acquire a reference to the event rule. */
-       status = lttng_condition_event_rule_matches_get_rule(
-                       condition, &event_rule);
-       LTTNG_ASSERT(status == LTTNG_CONDITION_STATUS_OK);
-
-       domain_type = lttng_event_rule_get_domain_type(event_rule);
-       LTTNG_ASSERT(domain_type == LTTNG_DOMAIN_KERNEL);
-
-       ret = kernel_create_event_notifier_rule(trigger, cmd_creds, token);
-       if (ret != LTTNG_OK) {
-               ERR("Failed to create kernel event notifier rule");
-       }
-
-       return ret;
-}
-
-enum lttng_error_code kernel_unregister_event_notifier(
-               const struct lttng_trigger *trigger)
-{
-       struct ltt_kernel_event_notifier_rule *token_event_rule_element;
-       struct cds_lfht_node *node;
-       struct cds_lfht_iter iter;
-       enum lttng_error_code error_code_ret;
-       int ret;
-
-       rcu_read_lock();
-
-       cds_lfht_lookup(kernel_token_to_event_notifier_rule_ht,
-                       hash_trigger(trigger), match_trigger, trigger, &iter);
-
-       node = cds_lfht_iter_get_node(&iter);
-       if (!node) {
-               error_code_ret = LTTNG_ERR_TRIGGER_NOT_FOUND;
-               goto error;
-       }
-
-       token_event_rule_element = caa_container_of(node,
-                       struct ltt_kernel_event_notifier_rule, ht_node);
-
-       ret = kernel_disable_event_notifier_rule(token_event_rule_element);
-       if (ret) {
-               error_code_ret = LTTNG_ERR_FATAL;
-               goto error;
-       }
-
-       trace_kernel_destroy_event_notifier_rule(token_event_rule_element);
-       error_code_ret = LTTNG_OK;
-
-error:
-       rcu_read_unlock();
-
-       return error_code_ret;
-}
-
-int kernel_get_notification_fd(void)
-{
-       return kernel_tracer_event_notifier_group_notification_fd;
-}
This page took 0.048263 seconds and 4 git commands to generate.