Fix: eliminate timestamp overlap between packets
[lttng-ust.git] / libringbuffer / frontend_internal.h
index 1fb08dfdec90d3f25e5f7ce85f39c010386badf4..8a0f78f32ee774e19e853b3c7c511957ed7adb85 100644 (file)
@@ -1,13 +1,28 @@
-#ifndef _LINUX_RING_BUFFER_FRONTEND_INTERNAL_H
-#define _LINUX_RING_BUFFER_FRONTEND_INTERNAL_H
+#ifndef _LTTNG_RING_BUFFER_FRONTEND_INTERNAL_H
+#define _LTTNG_RING_BUFFER_FRONTEND_INTERNAL_H
 
 /*
- * linux/ringbuffer/frontend_internal.h
- *
- * (C) Copyright 2005-2010 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ * libringbuffer/frontend_internal.h
  *
  * Ring Buffer Library Synchronization Header (internal helpers).
  *
+ * Copyright (C) 2005-2012 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; only
+ * version 2.1 of the License.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ *
+ *
  * Author:
  *     Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
  *
  */
 
 #include <urcu/compiler.h>
+#include <urcu/tls-compat.h>
+#include <signal.h>
+#include <pthread.h>
 
-#include <ust/ringbuffer-config.h>
+#include <lttng/ringbuffer-config.h>
 #include "backend_types.h"
 #include "frontend_types.h"
 #include "shm.h"
@@ -85,8 +103,8 @@ unsigned long subbuf_index(unsigned long offset, struct channel *chan)
 
 #if (CAA_BITS_PER_LONG == 32)
 static inline
-void save_last_tsc(const struct lib_ring_buffer_config *config,
-                  struct lib_ring_buffer *buf, u64 tsc)
+void save_last_tsc(const struct lttng_ust_lib_ring_buffer_config *config,
+                  struct lttng_ust_lib_ring_buffer *buf, uint64_t tsc)
 {
        if (config->tsc_bits == 0 || config->tsc_bits == 64)
                return;
@@ -98,8 +116,8 @@ void save_last_tsc(const struct lib_ring_buffer_config *config,
 }
 
 static inline
-int last_tsc_overflow(const struct lib_ring_buffer_config *config,
-                     struct lib_ring_buffer *buf, u64 tsc)
+int last_tsc_overflow(const struct lttng_ust_lib_ring_buffer_config *config,
+                     struct lttng_ust_lib_ring_buffer *buf, uint64_t tsc)
 {
        unsigned long tsc_shifted;
 
@@ -107,7 +125,7 @@ int last_tsc_overflow(const struct lib_ring_buffer_config *config,
                return 0;
 
        tsc_shifted = (unsigned long)(tsc >> config->tsc_bits);
-       if (unlikely(tsc_shifted
+       if (caa_unlikely(tsc_shifted
                     - (unsigned long)v_read(config, &buf->last_tsc)))
                return 1;
        else
@@ -115,8 +133,8 @@ int last_tsc_overflow(const struct lib_ring_buffer_config *config,
 }
 #else
 static inline
-void save_last_tsc(const struct lib_ring_buffer_config *config,
-                  struct lib_ring_buffer *buf, u64 tsc)
+void save_last_tsc(const struct lttng_ust_lib_ring_buffer_config *config,
+                  struct lttng_ust_lib_ring_buffer *buf, uint64_t tsc)
 {
        if (config->tsc_bits == 0 || config->tsc_bits == 64)
                return;
@@ -125,13 +143,13 @@ void save_last_tsc(const struct lib_ring_buffer_config *config,
 }
 
 static inline
-int last_tsc_overflow(const struct lib_ring_buffer_config *config,
-                     struct lib_ring_buffer *buf, u64 tsc)
+int last_tsc_overflow(const struct lttng_ust_lib_ring_buffer_config *config,
+                     struct lttng_ust_lib_ring_buffer *buf, uint64_t tsc)
 {
        if (config->tsc_bits == 0 || config->tsc_bits == 64)
                return 0;
 
-       if (unlikely((tsc - v_read(config, &buf->last_tsc))
+       if (caa_unlikely((tsc - v_read(config, &buf->last_tsc))
                     >> config->tsc_bits))
                return 1;
        else
@@ -140,17 +158,17 @@ int last_tsc_overflow(const struct lib_ring_buffer_config *config,
 #endif
 
 extern
-int lib_ring_buffer_reserve_slow(struct lib_ring_buffer_ctx *ctx);
+int lib_ring_buffer_reserve_slow(struct lttng_ust_lib_ring_buffer_ctx *ctx);
 
 extern
-void lib_ring_buffer_switch_slow(struct lib_ring_buffer *buf,
+void lib_ring_buffer_switch_slow(struct lttng_ust_lib_ring_buffer *buf,
                                 enum switch_mode mode,
-                                struct shm_handle *handle);
+                                struct lttng_ust_shm_handle *handle);
 
 /* Buffer write helpers */
 
 static inline
-void lib_ring_buffer_reserve_push_reader(struct lib_ring_buffer *buf,
+void lib_ring_buffer_reserve_push_reader(struct lttng_ust_lib_ring_buffer *buf,
                                         struct channel *chan,
                                         unsigned long offset)
 {
@@ -167,32 +185,32 @@ void lib_ring_buffer_reserve_push_reader(struct lib_ring_buffer *buf,
                 * write position sub-buffer index in the buffer being the one
                 * which will win this loop.
                 */
-               if (unlikely(subbuf_trunc(offset, chan)
+               if (caa_unlikely(subbuf_trunc(offset, chan)
                              - subbuf_trunc(consumed_old, chan)
                             >= chan->backend.buf_size))
                        consumed_new = subbuf_align(consumed_old, chan);
                else
                        return;
-       } while (unlikely(uatomic_cmpxchg(&buf->consumed, consumed_old,
+       } while (caa_unlikely(uatomic_cmpxchg(&buf->consumed, consumed_old,
                                              consumed_new) != consumed_old));
 }
 
 static inline
-void lib_ring_buffer_vmcore_check_deliver(const struct lib_ring_buffer_config *config,
-                                         struct lib_ring_buffer *buf,
+void lib_ring_buffer_vmcore_check_deliver(const struct lttng_ust_lib_ring_buffer_config *config,
+                                         struct lttng_ust_lib_ring_buffer *buf,
                                          unsigned long commit_count,
                                          unsigned long idx,
-                                         struct shm_handle *handle)
+                                         struct lttng_ust_shm_handle *handle)
 {
        if (config->oops == RING_BUFFER_OOPS_CONSISTENCY)
                v_set(config, &shmp_index(handle, buf->commit_hot, idx)->seq, commit_count);
 }
 
 static inline
-int lib_ring_buffer_poll_deliver(const struct lib_ring_buffer_config *config,
-                                struct lib_ring_buffer *buf,
+int lib_ring_buffer_poll_deliver(const struct lttng_ust_lib_ring_buffer_config *config,
+                                struct lttng_ust_lib_ring_buffer *buf,
                                 struct channel *chan,
-                                struct shm_handle *handle)
+                                struct lttng_ust_shm_handle *handle)
 {
        unsigned long consumed_old, consumed_idx, commit_count, write_offset;
 
@@ -232,18 +250,18 @@ int lib_ring_buffer_poll_deliver(const struct lib_ring_buffer_config *config,
 }
 
 static inline
-int lib_ring_buffer_pending_data(const struct lib_ring_buffer_config *config,
-                                struct lib_ring_buffer *buf,
+int lib_ring_buffer_pending_data(const struct lttng_ust_lib_ring_buffer_config *config,
+                                struct lttng_ust_lib_ring_buffer *buf,
                                 struct channel *chan)
 {
        return !!subbuf_offset(v_read(config, &buf->offset), chan);
 }
 
 static inline
-unsigned long lib_ring_buffer_get_data_size(const struct lib_ring_buffer_config *config,
-                                           struct lib_ring_buffer *buf,
+unsigned long lib_ring_buffer_get_data_size(const struct lttng_ust_lib_ring_buffer_config *config,
+                                           struct lttng_ust_lib_ring_buffer *buf,
                                            unsigned long idx,
-                                           struct shm_handle *handle)
+                                           struct lttng_ust_shm_handle *handle)
 {
        return subbuffer_get_data_size(config, &buf->backend, idx, handle);
 }
@@ -254,10 +272,10 @@ unsigned long lib_ring_buffer_get_data_size(const struct lib_ring_buffer_config
  * This is a very specific ftrace use-case, so we keep this as "internal" API.
  */
 static inline
-int lib_ring_buffer_reserve_committed(const struct lib_ring_buffer_config *config,
-                                     struct lib_ring_buffer *buf,
+int lib_ring_buffer_reserve_committed(const struct lttng_ust_lib_ring_buffer_config *config,
+                                     struct lttng_ust_lib_ring_buffer *buf,
                                      struct channel *chan,
-                                     struct shm_handle *handle)
+                                     struct lttng_ust_shm_handle *handle)
 {
        unsigned long offset, idx, commit_count;
 
@@ -283,20 +301,92 @@ int lib_ring_buffer_reserve_committed(const struct lib_ring_buffer_config *confi
 }
 
 static inline
-void lib_ring_buffer_check_deliver(const struct lib_ring_buffer_config *config,
-                                  struct lib_ring_buffer *buf,
+void lib_ring_buffer_wakeup(struct lttng_ust_lib_ring_buffer *buf,
+               struct lttng_ust_shm_handle *handle)
+{
+       int wakeup_fd = shm_get_wakeup_fd(handle, &buf->self._ref);
+       sigset_t sigpipe_set, pending_set, old_set;
+       int ret, sigpipe_was_pending = 0;
+
+       if (wakeup_fd < 0)
+               return;
+
+       /*
+        * Wake-up the other end by writing a null byte in the pipe
+        * (non-blocking).  Important note: Because writing into the
+        * pipe is non-blocking (and therefore we allow dropping wakeup
+        * data, as long as there is wakeup data present in the pipe
+        * buffer to wake up the consumer), the consumer should perform
+        * the following sequence for waiting:
+        * 1) empty the pipe (reads).
+        * 2) check if there is data in the buffer.
+        * 3) wait on the pipe (poll).
+        *
+        * Discard the SIGPIPE from write(), not disturbing any SIGPIPE
+        * that might be already pending. If a bogus SIGPIPE is sent to
+        * the entire process concurrently by a malicious user, it may
+        * be simply discarded.
+        */
+       ret = sigemptyset(&pending_set);
+       assert(!ret);
+       /*
+        * sigpending returns the mask of signals that are _both_
+        * blocked for the thread _and_ pending for either the thread or
+        * the entire process.
+        */
+       ret = sigpending(&pending_set);
+       assert(!ret);
+       sigpipe_was_pending = sigismember(&pending_set, SIGPIPE);
+       /*
+        * If sigpipe was pending, it means it was already blocked, so
+        * no need to block it.
+        */
+       if (!sigpipe_was_pending) {
+               ret = sigemptyset(&sigpipe_set);
+               assert(!ret);
+               ret = sigaddset(&sigpipe_set, SIGPIPE);
+               assert(!ret);
+               ret = pthread_sigmask(SIG_BLOCK, &sigpipe_set, &old_set);
+               assert(!ret);
+       }
+       do {
+               ret = write(wakeup_fd, "", 1);
+       } while (ret == -1L && errno == EINTR);
+       if (ret == -1L && errno == EPIPE && !sigpipe_was_pending) {
+               struct timespec timeout = { 0, 0 };
+               do {
+                       ret = sigtimedwait(&sigpipe_set, NULL,
+                               &timeout);
+               } while (ret == -1L && errno == EINTR);
+       }
+       if (!sigpipe_was_pending) {
+               ret = pthread_sigmask(SIG_SETMASK, &old_set, NULL);
+               assert(!ret);
+       }
+}
+
+/*
+ * Receive end of subbuffer TSC as parameter. It has been read in the
+ * space reservation loop of either reserve or switch, which ensures it
+ * progresses monotonically with event records in the buffer. Therefore,
+ * it ensures that the end timestamp of a subbuffer is <= begin
+ * timestamp of the following subbuffers.
+ */
+static inline
+void lib_ring_buffer_check_deliver(const struct lttng_ust_lib_ring_buffer_config *config,
+                                  struct lttng_ust_lib_ring_buffer *buf,
                                   struct channel *chan,
                                   unsigned long offset,
                                   unsigned long commit_count,
                                   unsigned long idx,
-                                  struct shm_handle *handle)
+                                  struct lttng_ust_shm_handle *handle,
+                                  uint64_t tsc)
 {
        unsigned long old_commit_count = commit_count
                                         - chan->backend.subbuf_size;
-       u64 tsc;
 
        /* Check if all commits have been done */
-       if (unlikely((buf_trunc(offset, chan) >> chan->backend.num_subbuf_order)
+       if (caa_unlikely((buf_trunc(offset, chan) >> chan->backend.num_subbuf_order)
                     - (old_commit_count & chan->commit_count_mask) == 0)) {
                /*
                 * If we succeeded at updating cc_sb below, we are the subbuffer
@@ -324,7 +414,13 @@ void lib_ring_buffer_check_deliver(const struct lib_ring_buffer_config *config,
                 * The subbuffer size is least 2 bytes (minimum size: 1 page).
                 * This guarantees that old_commit_count + 1 != commit_count.
                 */
-               if (likely(v_cmpxchg(config, &shmp_index(handle, buf->commit_cold, idx)->cc_sb,
+
+               /*
+                * Order prior updates to reserve count prior to the
+                * commit_cold cc_sb update.
+                */
+               cmm_smp_wmb();
+               if (caa_likely(v_cmpxchg(config, &shmp_index(handle, buf->commit_cold, idx)->cc_sb,
                                         old_commit_count, old_commit_count + 1)
                           == old_commit_count)) {
                        /*
@@ -333,7 +429,6 @@ void lib_ring_buffer_check_deliver(const struct lib_ring_buffer_config *config,
                         * and any other writer trying to access this subbuffer
                         * in this state is required to drop records.
                         */
-                       tsc = config->cb.ring_buffer_clock_read(chan);
                        v_add(config,
                              subbuffer_get_records_count(config,
                                                          &buf->backend,
@@ -369,6 +464,11 @@ void lib_ring_buffer_check_deliver(const struct lib_ring_buffer_config *config,
                        /* End of exclusive subbuffer access */
                        v_set(config, &shmp_index(handle, buf->commit_cold, idx)->cc_sb,
                              commit_count);
+                       /*
+                        * Order later updates to reserve count after
+                        * the commit cold cc_sb update.
+                        */
+                       cmm_smp_wmb();
                        lib_ring_buffer_vmcore_check_deliver(config, buf,
                                                 commit_count, idx, handle);
 
@@ -378,35 +478,8 @@ void lib_ring_buffer_check_deliver(const struct lib_ring_buffer_config *config,
                        if (config->wakeup == RING_BUFFER_WAKEUP_BY_WRITER
                            && uatomic_read(&buf->active_readers)
                            && lib_ring_buffer_poll_deliver(config, buf, chan, handle)) {
-                               int wakeup_fd = shm_get_wakeup_fd(handle, &buf->self._ref);
-
-                               if (wakeup_fd >= 0) {
-                                       int ret;
-                                       /*
-                                        * Wake-up the other end by
-                                        * writing a null byte in the
-                                        * pipe (non-blocking).
-                                        * Important note: Because
-                                        * writing into the pipe is
-                                        * non-blocking (and therefore
-                                        * we allow dropping wakeup
-                                        * data, as long as there is
-                                        * wakeup data present in the
-                                        * pipe buffer to wake up the
-                                        * consumer), the consumer
-                                        * should perform the following
-                                        * sequence for waiting:
-                                        * 1) empty the pipe (reads).
-                                        * 2) check if there is data in
-                                        *    the buffer.
-                                        * 3) wait on the pipe (poll).
-                                        */
-                                       do {
-                                               ret = write(wakeup_fd, "", 1);
-                                       } while (ret == -1L && errno == EINTR);
-                               }
+                               lib_ring_buffer_wakeup(buf, handle);
                        }
-
                }
        }
 }
@@ -420,14 +493,14 @@ void lib_ring_buffer_check_deliver(const struct lib_ring_buffer_config *config,
  * useful for crash dump.
  */
 static inline
-void lib_ring_buffer_write_commit_counter(const struct lib_ring_buffer_config *config,
-                                         struct lib_ring_buffer *buf,
+void lib_ring_buffer_write_commit_counter(const struct lttng_ust_lib_ring_buffer_config *config,
+                                         struct lttng_ust_lib_ring_buffer *buf,
                                          struct channel *chan,
                                          unsigned long idx,
                                          unsigned long buf_offset,
                                          unsigned long commit_count,
                                          size_t slot_size,
-                                         struct shm_handle *handle)
+                                         struct lttng_ust_shm_handle *handle)
 {
        unsigned long offset, commit_seq_old;
 
@@ -442,7 +515,7 @@ void lib_ring_buffer_write_commit_counter(const struct lib_ring_buffer_config *c
         * buffer full/empty mismatch because offset is never zero here
         * (subbuffer header and record headers have non-zero length).
         */
-       if (unlikely(subbuf_offset(offset - commit_count, chan)))
+       if (caa_unlikely(subbuf_offset(offset - commit_count, chan)))
                return;
 
        commit_seq_old = v_read(config, &shmp_index(handle, buf->commit_hot, idx)->seq);
@@ -451,14 +524,14 @@ void lib_ring_buffer_write_commit_counter(const struct lib_ring_buffer_config *c
                                           commit_seq_old, commit_count);
 }
 
-extern int lib_ring_buffer_create(struct lib_ring_buffer *buf,
+extern int lib_ring_buffer_create(struct lttng_ust_lib_ring_buffer *buf,
                                  struct channel_backend *chanb, int cpu,
-                                 struct shm_handle *handle,
+                                 struct lttng_ust_shm_handle *handle,
                                  struct shm_object *shmobj);
-extern void lib_ring_buffer_free(struct lib_ring_buffer *buf,
-                                struct shm_handle *handle);
+extern void lib_ring_buffer_free(struct lttng_ust_lib_ring_buffer *buf,
+                                struct lttng_ust_shm_handle *handle);
 
 /* Keep track of trap nesting inside ring buffer code */
-extern __thread unsigned int lib_ring_buffer_nesting;
+extern DECLARE_URCU_TLS(unsigned int, lib_ring_buffer_nesting);
 
-#endif /* _LINUX_RING_BUFFER_FRONTEND_INTERNAL_H */
+#endif /* _LTTNG_RING_BUFFER_FRONTEND_INTERNAL_H */
This page took 0.030216 seconds and 4 git commands to generate.