X-Git-Url: https://git.liburcu.org/?a=blobdiff_plain;f=libust%2Fbuffers.c;h=c8976c7841834e82bd8fae53e0dbe8ca2783d51b;hb=d4d511d9f43d6d33c3a40b4aab69d976c4f93cec;hp=f3331d0899f523677be93f7c5e17b84007d9ee38;hpb=b5b073e22d34bec71259d39b8946354f170f01a9;p=ust.git diff --git a/libust/buffers.c b/libust/buffers.c index f3331d0..c8976c7 100644 --- a/libust/buffers.c +++ b/libust/buffers.c @@ -3,7 +3,7 @@ * LTTng userspace tracer buffering system * * Copyright (C) 2009 - Pierre-Marc Fournier (pierre-marc dot fournier at polymtl dot ca) - * Copyright (C) 2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * Copyright (C) 2008-2011 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -20,27 +20,151 @@ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ +/* + * Note: this code does not support the ref/noref flag and reader-owned + * subbuffer scheme needed for flight recorder mode. + */ + +#include #include #include #include #include -#include -#include +#include + +#include + #include "buffers.h" #include "channels.h" #include "tracer.h" #include "tracercore.h" -#include "usterr.h" +#include "usterr_signal_safe.h" + +struct ltt_reserve_switch_offsets { + long begin, end, old; + long begin_switch, end_switch_current, end_switch_old; + size_t before_hdr_pad, size; +}; + static DEFINE_MUTEX(ust_buffers_channels_mutex); -static LIST_HEAD(ust_buffers_channels); +static CDS_LIST_HEAD(ust_buffers_channels); -static int ust_buffers_init_buffer(struct ltt_trace_struct *trace, - struct ust_channel *ltt_chan, - struct ust_buffer *buf, - unsigned int n_subbufs); +static void ltt_force_switch(struct ust_buffer *buf, + enum force_switch_mode mode); + +static int get_n_cpus(void) +{ + int result; + static int n_cpus = 0; + + if(!n_cpus) { + /* On Linux, when some processors are offline + * _SC_NPROCESSORS_CONF counts the offline + * processors, whereas _SC_NPROCESSORS_ONLN + * does not. If we used _SC_NPROCESSORS_ONLN, + * getcpu() could return a value greater than + * this sysconf, in which case the arrays + * indexed by processor would overflow. + */ + result = sysconf(_SC_NPROCESSORS_CONF); + if(result == -1) { + return -1; + } + + n_cpus = result; + } + + return n_cpus; +} + +/** + * _ust_buffers_strncpy_fixup - Fix an incomplete string in a ltt_relay buffer. + * @buf : buffer + * @offset : offset within the buffer + * @len : length to write + * @copied: string actually copied + * @terminated: does string end with \0 + * + * Fills string with "X" if incomplete. + */ +void _ust_buffers_strncpy_fixup(struct ust_buffer *buf, size_t offset, + size_t len, size_t copied, int terminated) +{ + size_t buf_offset, cpy; + + if (copied == len) { + /* + * Deal with non-terminated string. + */ + assert(!terminated); + offset += copied - 1; + buf_offset = BUFFER_OFFSET(offset, buf->chan); + /* + * Underlying layer should never ask for writes across + * subbuffers. + */ + assert(buf_offset + < buf->chan->subbuf_size*buf->chan->subbuf_cnt); + ust_buffers_do_memset(buf->buf_data + buf_offset, '\0', 1); + return; + } + + /* + * Deal with incomplete string. + * Overwrite string's \0 with X too. + */ + cpy = copied - 1; + assert(terminated); + len -= cpy; + offset += cpy; + buf_offset = BUFFER_OFFSET(offset, buf->chan); + + /* + * Underlying layer should never ask for writes across subbuffers. + */ + assert(buf_offset + < buf->chan->subbuf_size*buf->chan->subbuf_cnt); + + ust_buffers_do_memset(buf->buf_data + buf_offset, + 'X', len); + + /* + * Overwrite last 'X' with '\0'. + */ + offset += len - 1; + buf_offset = BUFFER_OFFSET(offset, buf->chan); + /* + * Underlying layer should never ask for writes across subbuffers. + */ + assert(buf_offset + < buf->chan->subbuf_size*buf->chan->subbuf_cnt); + ust_buffers_do_memset(buf->buf_data + buf_offset, '\0', 1); +} + +static void ltt_buffer_begin(struct ust_buffer *buf, + u64 tsc, unsigned int subbuf_idx) +{ + struct ust_channel *channel = buf->chan; + struct ltt_subbuffer_header *header = + (struct ltt_subbuffer_header *) + ust_buffers_offset_address(buf, + subbuf_idx * buf->chan->subbuf_size); + + header->cycle_count_begin = tsc; + header->data_size = 0xFFFFFFFF; /* for recognizing crashed buffers */ + header->sb_size = 0xFFFFFFFF; /* for recognizing crashed buffers */ + /* + * No memory barrier needed to order data_data/sb_size vs commit count + * update, because commit count update contains a compiler barrier that + * ensures the order of the writes are OK from a program POV. It only + * matters for crash dump recovery which is not executed concurrently, + * so memory write order does not matter. + */ + ltt_write_trace_header(channel->trace, header); +} -static int ust_buffers_alloc_buf(struct ust_buffer *buf, size_t *size) +static int map_buf_data(struct ust_buffer *buf, size_t *size) { void *ptr; int result; @@ -48,17 +172,16 @@ static int ust_buffers_alloc_buf(struct ust_buffer *buf, size_t *size) *size = PAGE_ALIGN(*size); result = buf->shmid = shmget(getpid(), *size, IPC_CREAT | IPC_EXCL | 0700); - if(result == -1 && errno == EINVAL) { + if (result < 0 && errno == EINVAL) { ERR("shmget() returned EINVAL; maybe /proc/sys/kernel/shmmax should be increased."); return -1; - } - else if(result == -1) { + } else if (result < 0) { PERROR("shmget"); return -1; } ptr = shmat(buf->shmid, NULL, 0); - if(ptr == (void *) -1) { + if (ptr == (void *) -1) { perror("shmat"); goto destroy_shmem; } @@ -77,7 +200,7 @@ static int ust_buffers_alloc_buf(struct ust_buffer *buf, size_t *size) return 0; - destroy_shmem: +destroy_shmem: result = shmctl(buf->shmid, IPC_RMID, NULL); if(result == -1) { perror("shmctl"); @@ -86,281 +209,180 @@ static int ust_buffers_alloc_buf(struct ust_buffer *buf, size_t *size) return -1; } -static struct ust_buffer *ust_buffers_create_buf(struct ust_channel *channel) +static int open_buf(struct ust_channel *chan, int cpu) { - int result; + int result, fds[2]; + unsigned int j; + struct ust_trace *trace = chan->trace; + struct ust_buffer *buf = chan->buf[cpu]; + unsigned int n_subbufs = chan->subbuf_cnt; - result = ust_buffers_alloc_buf(channel->buf, &channel->alloc_size); - if(result) - goto free_buf; - ((struct ust_buffer *)channel->buf)->chan = channel; - kref_get(&channel->kref); - return channel->buf; + result = map_buf_data(buf, &chan->alloc_size); + if (result < 0) + return -1; -free_buf: - return NULL; -} + buf->commit_count = + zmalloc(sizeof(*buf->commit_count) * n_subbufs); + if (!buf->commit_count) + goto unmap_buf; -static void ust_buffers_destroy_channel(struct kref *kref) -{ - struct ust_channel *chan = container_of(kref, struct ust_channel, kref); - free(chan); -} + result = pipe(fds); + if (result < 0) { + PERROR("pipe"); + goto free_commit_count; + } + buf->data_ready_fd_read = fds[0]; + buf->data_ready_fd_write = fds[1]; -static void ust_buffers_destroy_buf(struct ust_buffer *buf) -{ - struct ust_channel *chan = buf->chan; - int result; + buf->cpu = cpu; + buf->chan = chan; - result = munmap(buf->buf_data, buf->buf_size); - if(result == -1) { - PERROR("munmap"); + uatomic_set(&buf->offset, ltt_subbuffer_header_size()); + uatomic_set(&buf->consumed, 0); + uatomic_set(&buf->active_readers, 0); + for (j = 0; j < n_subbufs; j++) { + uatomic_set(&buf->commit_count[j].cc, 0); + uatomic_set(&buf->commit_count[j].cc_sb, 0); } - free(buf); - kref_put(&chan->kref, ust_buffers_destroy_channel); -} - -/* called from kref_put */ -static void ust_buffers_remove_buf(struct kref *kref) -{ - struct ust_buffer *buf = container_of(kref, struct ust_buffer, kref); - ust_buffers_destroy_buf(buf); -} + ltt_buffer_begin(buf, trace->start_tsc, 0); -static struct ust_buffer *ust_buffers_open_buf(struct ust_channel *chan) -{ - struct ust_buffer *buf = NULL; - int err; + uatomic_add(&buf->commit_count[0].cc, ltt_subbuffer_header_size()); - buf = ust_buffers_create_buf(chan); - if (!buf) - return NULL; + uatomic_set(&buf->events_lost, 0); + uatomic_set(&buf->corrupted_subbuffers, 0); - kref_init(&buf->kref); + memset(buf->commit_seq, 0, sizeof(buf->commit_seq[0]) * n_subbufs); - err = ust_buffers_init_buffer(chan->trace, chan, buf, chan->subbuf_cnt); + return 0; - if (err) - return ERR_PTR(err); +free_commit_count: + free(buf->commit_count); - return buf; +unmap_buf: + if (shmdt(buf->buf_data) < 0) { + PERROR("shmdt failed"); + } - /* FIXME: decrementally destroy on error? */ + return -1; } -/** - * ust_buffers_close_buf - close a channel buffer - * @buf: buffer - */ -static void ust_buffers_close_buf(struct ust_buffer *buf) +static void close_buf(struct ust_buffer *buf) { - kref_put(&buf->kref, ust_buffers_remove_buf); + int result; + + result = shmdt(buf->buf_data); + if (result < 0) { + PERROR("shmdt"); + } + + result = close(buf->data_ready_fd_read); + if (result < 0) { + PERROR("close"); + } + + result = close(buf->data_ready_fd_write); + if (result < 0 && errno != EBADF) { + PERROR("close"); + } } -int ust_buffers_channel_open(struct ust_channel *chan, size_t subbuf_size, size_t subbuf_cnt) + +static int open_channel(struct ust_channel *chan, size_t subbuf_size, + size_t subbuf_cnt) { + int i; + int result; + if(subbuf_size == 0 || subbuf_cnt == 0) return -1; + /* Check that the subbuffer size is larger than a page. */ + WARN_ON_ONCE(subbuf_size < PAGE_SIZE); + + /* + * Make sure the number of subbuffers and subbuffer size are power of 2. + */ + WARN_ON_ONCE(hweight32(subbuf_size) != 1); + WARN_ON(hweight32(subbuf_cnt) != 1); + chan->version = UST_CHANNEL_VERSION; chan->subbuf_cnt = subbuf_cnt; chan->subbuf_size = subbuf_size; chan->subbuf_size_order = get_count_order(subbuf_size); - chan->alloc_size = FIX_SIZE(subbuf_size * subbuf_cnt); - kref_init(&chan->kref); + chan->alloc_size = subbuf_size * subbuf_cnt; - mutex_lock(&ust_buffers_channels_mutex); - chan->buf = ust_buffers_open_buf(chan); - if (!chan->buf) - goto error; - list_add(&chan->list, &ust_buffers_channels); - mutex_unlock(&ust_buffers_channels_mutex); + pthread_mutex_lock(&ust_buffers_channels_mutex); + for (i=0; i < chan->n_cpus; i++) { + result = open_buf(chan, i); + if (result == -1) + goto error; + } + cds_list_add(&chan->list, &ust_buffers_channels); + pthread_mutex_unlock(&ust_buffers_channels_mutex); return 0; - error: - kref_put(&chan->kref, ust_buffers_destroy_channel); - mutex_unlock(&ust_buffers_channels_mutex); + /* Error handling */ +error: + for(i--; i >= 0; i--) + close_buf(chan->buf[i]); + + pthread_mutex_unlock(&ust_buffers_channels_mutex); return -1; } -void ust_buffers_channel_close(struct ust_channel *chan) +static void close_channel(struct ust_channel *chan) { - if (!chan) + int i; + if(!chan) return; - mutex_lock(&ust_buffers_channels_mutex); - if (chan->buf) - ust_buffers_close_buf(chan->buf); - - list_del(&chan->list); - kref_put(&chan->kref, ust_buffers_destroy_channel); - mutex_unlock(&ust_buffers_channels_mutex); -} - -/* _ust_buffers_write() - * - * @buf: destination buffer - * @offset: offset in destination - * @src: source buffer - * @len: length of source - * @cpy: already copied - */ - -void _ust_buffers_write(struct ust_buffer *buf, size_t offset, - const void *src, size_t len, ssize_t cpy) -{ - do { - len -= cpy; - src += cpy; - offset += cpy; - WARN_ON(offset >= buf->buf_size); - - cpy = min_t(size_t, len, buf->buf_size - offset); - ust_buffers_do_copy(buf->buf_data + offset, src, cpy); - } while (unlikely(len != cpy)); -} - -/** - * ltt_buffers_offset_address - get address of a location within the buffer - * @buf : buffer - * @offset : offset within the buffer. - * - * Return the address where a given offset is located. - * Should be used to get the current subbuffer header pointer. Given we know - * it's never on a page boundary, it's safe to write directly to this address, - * as long as the write is never bigger than a page size. - */ -void *ltt_buffers_offset_address(struct ust_buffer *buf, size_t offset) -{ - return ((char *)buf->buf_data)+offset; -} - -/* - * ------- - */ - -/* - * Last TSC comparison functions. Check if the current TSC overflows - * LTT_TSC_BITS bits from the last TSC read. Reads and writes last_tsc - * atomically. - */ - -/* FIXME: does this test work properly? */ -#if (BITS_PER_LONG == 32) -static inline void save_last_tsc(struct ltt_channel_buf_struct *ltt_buf, - u64 tsc) -{ - ltt_buf->last_tsc = (unsigned long)(tsc >> LTT_TSC_BITS); -} - -static inline int last_tsc_overflow(struct ltt_channel_buf_struct *ltt_buf, - u64 tsc) -{ - unsigned long tsc_shifted = (unsigned long)(tsc >> LTT_TSC_BITS); - - if (unlikely((tsc_shifted - ltt_buf->last_tsc))) - return 1; - else - return 0; -} -#else -static inline void save_last_tsc(struct ust_buffer *ltt_buf, - u64 tsc) -{ - ltt_buf->last_tsc = (unsigned long)tsc; -} - -static inline int last_tsc_overflow(struct ust_buffer *ltt_buf, - u64 tsc) -{ - if (unlikely((tsc - ltt_buf->last_tsc) >> LTT_TSC_BITS)) - return 1; - else - return 0; -} -#endif - -/* - * A switch is done during tracing or as a final flush after tracing (so it - * won't write in the new sub-buffer). - */ -enum force_switch_mode { FORCE_ACTIVE, FORCE_FLUSH }; - -static void ust_buffers_destroy_buffer(struct ust_channel *ltt_chan); - -static void ltt_force_switch(struct ust_buffer *buf, - enum force_switch_mode mode); + pthread_mutex_lock(&ust_buffers_channels_mutex); + /* + * checking for chan->buf[i] being NULL or not is useless in + * practice because we allocate buffers for all possible cpus. + * However, should we decide to change this and only allocate + * for online cpus, this check becomes useful. + */ + for (i=0; in_cpus; i++) { + if (chan->buf[i]) + close_buf(chan->buf[i]); + } -/* - * Trace callbacks - */ -static void ltt_buffer_begin_callback(struct ust_buffer *buf, - u64 tsc, unsigned int subbuf_idx) -{ - struct ust_channel *channel = buf->chan; - struct ltt_subbuffer_header *header = - (struct ltt_subbuffer_header *) - ltt_buffers_offset_address(buf, - subbuf_idx * buf->chan->subbuf_size); + cds_list_del(&chan->list); - header->cycle_count_begin = tsc; - header->lost_size = 0xFFFFFFFF; /* for debugging */ - header->buf_size = buf->chan->subbuf_size; - ltt_write_trace_header(channel->trace, header); + pthread_mutex_unlock(&ust_buffers_channels_mutex); } /* * offset is assumed to never be 0 here : never deliver a completely empty * subbuffer. The lost size is between 0 and subbuf_size-1. */ -static notrace void ltt_buffer_end_callback(struct ust_buffer *buf, +static notrace void ltt_buffer_end(struct ust_buffer *buf, u64 tsc, unsigned int offset, unsigned int subbuf_idx) { struct ltt_subbuffer_header *header = (struct ltt_subbuffer_header *) - ltt_buffers_offset_address(buf, + ust_buffers_offset_address(buf, subbuf_idx * buf->chan->subbuf_size); + u32 data_size = SUBBUF_OFFSET(offset - 1, buf->chan) + 1; - header->lost_size = SUBBUF_OFFSET((buf->chan->subbuf_size - offset), - buf->chan); + header->sb_size = PAGE_ALIGN(data_size); header->cycle_count_end = tsc; - header->events_lost = local_read(&buf->events_lost); - header->subbuf_corrupt = local_read(&buf->corrupted_subbuffers); - -} - -void (*wake_consumer)(void *, int) = NULL; - -void relay_set_wake_consumer(void (*wake)(void *, int)) -{ - wake_consumer = wake; -} - -void relay_wake_consumer(void *arg, int finished) -{ - if(wake_consumer) - wake_consumer(arg, finished); -} - -static notrace void ltt_deliver(struct ust_buffer *buf, unsigned int subbuf_idx, - long commit_count) -{ - int result; - -//ust// #ifdef CONFIG_LTT_VMCORE - local_set(&buf->commit_seq[subbuf_idx], commit_count); -//ust// #endif - - /* wakeup consumer */ - result = write(buf->data_ready_fd_write, "1", 1); - if(result == -1) { - PERROR("write (in ltt_relay_buffer_flush)"); - ERR("this should never happen!"); + header->events_lost = uatomic_read(&buf->events_lost); + header->subbuf_corrupt = uatomic_read(&buf->corrupted_subbuffers); + if(unlikely(header->events_lost > 0)) { + DBG("Some events (%d) were lost in %s_%d", header->events_lost, buf->chan->channel_name, buf->cpu); } -//ust// atomic_set(<t_buf->wakeup_readers, 1); + /* + * Makes sure data_size write happens after write of the rest of the + * buffer end data, because data_size is used to identify a completely + * written subbuffer in a crash dump. + */ + cmm_barrier(); + header->data_size = data_size; } /* @@ -370,28 +392,42 @@ static notrace void ltt_buf_unfull(struct ust_buffer *buf, unsigned int subbuf_idx, long offset) { -//ust// struct ltt_channel_struct *ltt_channel = -//ust// (struct ltt_channel_struct *)buf->chan->private_data; -//ust// struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; -//ust// -//ust// ltt_relay_wake_writers(ltt_buf); } -int ust_buffers_do_get_subbuf(struct ust_buffer *buf, long *pconsumed_old) +/* + * Promote compiler cmm_barrier to a smp_mb(). + * For the specific LTTng case, this IPI call should be removed if the + * architecture does not reorder writes. This should eventually be provided by + * a separate architecture-specific infrastructure. + */ +//ust// static void remote_mb(void *info) +//ust// { +//ust// smp_mb(); +//ust// } + +int ust_buffers_get_subbuf(struct ust_buffer *buf, long *consumed) { struct ust_channel *channel = buf->chan; long consumed_old, consumed_idx, commit_count, write_offset; - consumed_old = atomic_long_read(&buf->consumed); +//ust// int retval; + + consumed_old = uatomic_read(&buf->consumed); consumed_idx = SUBBUF_INDEX(consumed_old, buf->chan); - commit_count = local_read(&buf->commit_count[consumed_idx]); + commit_count = uatomic_read(&buf->commit_count[consumed_idx].cc_sb); /* * Make sure we read the commit count before reading the buffer * data and the write offset. Correct consumed offset ordering * wrt commit count is insured by the use of cmpxchg to update * the consumed offset. */ - smp_rmb(); - write_offset = local_read(&buf->offset); + + /* + * Local rmb to match the remote wmb to read the commit count before the + * buffer data and the write offset. + */ + cmm_smp_rmb(); + + write_offset = uatomic_read(&buf->offset); /* * Check that the subbuffer we are trying to consume has been * already fully committed. @@ -412,28 +448,26 @@ int ust_buffers_do_get_subbuf(struct ust_buffer *buf, long *pconsumed_old) == 0) { return -EAGAIN; } + *consumed = consumed_old; - *pconsumed_old = consumed_old; return 0; } -int ust_buffers_do_put_subbuf(struct ust_buffer *buf, u32 uconsumed_old) +int ust_buffers_put_subbuf(struct ust_buffer *buf, unsigned long uconsumed_old) { long consumed_new, consumed_old; - consumed_old = atomic_long_read(&buf->consumed); + consumed_old = uatomic_read(&buf->consumed); consumed_old = consumed_old & (~0xFFFFFFFFL); consumed_old = consumed_old | uconsumed_old; consumed_new = SUBBUF_ALIGN(consumed_old, buf->chan); -//ust// spin_lock(<t_buf->full_lock); - if (atomic_long_cmpxchg(&buf->consumed, consumed_old, + if (uatomic_cmpxchg(&buf->consumed, consumed_old, consumed_new) != consumed_old) { /* We have been pushed by the writer : the last * buffer read _is_ corrupted! It can also * happen if this is a buffer we never got. */ -//ust// spin_unlock(<t_buf->full_lock); return -EIO; } else { /* tell the client that buffer is now unfull */ @@ -442,370 +476,418 @@ int ust_buffers_do_put_subbuf(struct ust_buffer *buf, u32 uconsumed_old) index = SUBBUF_INDEX(consumed_old, buf->chan); data = BUFFER_OFFSET(consumed_old, buf->chan); ltt_buf_unfull(buf, index, data); -//ust// spin_unlock(<t_buf->full_lock); } return 0; } -static void ltt_relay_print_subbuffer_errors( - struct ust_channel *channel, - long cons_off) +static int map_buf_structs(struct ust_channel *chan) { - struct ust_buffer *ltt_buf = channel->buf; - long cons_idx, commit_count, write_offset; + void *ptr; + int result; + size_t size; + int i; - cons_idx = SUBBUF_INDEX(cons_off, channel); - commit_count = local_read(<t_buf->commit_count[cons_idx]); - /* - * No need to order commit_count and write_offset reads because we - * execute after trace is stopped when there are no readers left. - */ - write_offset = local_read(<t_buf->offset); - WARN( "LTT : unread channel %s offset is %ld " - "and cons_off : %ld\n", - channel->channel_name, write_offset, cons_off); - /* Check each sub-buffer for non filled commit count */ - if (((commit_count - channel->subbuf_size) & channel->commit_count_mask) - - (BUFFER_TRUNC(cons_off, channel) >> channel->n_subbufs_order) != 0) { - ERR("LTT : %s : subbuffer %lu has non filled " - "commit count %lu.\n", - channel->channel_name, cons_idx, commit_count); + size = PAGE_ALIGN(1); + + for(i=0; in_cpus; i++) { + + result = chan->buf_struct_shmids[i] = shmget(getpid(), size, IPC_CREAT | IPC_EXCL | 0700); + if(result == -1) { + PERROR("shmget"); + goto destroy_previous; + } + + ptr = shmat(chan->buf_struct_shmids[i], NULL, 0); + if(ptr == (void *) -1) { + perror("shmat"); + goto destroy_shm; + } + + /* Already mark the shared memory for destruction. This will occur only + * when all users have detached. + */ + result = shmctl(chan->buf_struct_shmids[i], IPC_RMID, NULL); + if(result == -1) { + perror("shmctl"); + goto destroy_previous; + } + + chan->buf[i] = ptr; } - ERR("LTT : %s : commit count : %lu, subbuf size %zd\n", - channel->channel_name, commit_count, - channel->subbuf_size); -} -static void ltt_relay_print_errors(struct ltt_trace_struct *trace, - struct ust_channel *channel) -{ - struct ust_buffer *ltt_buf = channel->buf; - long cons_off; - - for (cons_off = atomic_long_read(<t_buf->consumed); - (SUBBUF_TRUNC(local_read(<t_buf->offset), - channel) - - cons_off) > 0; - cons_off = SUBBUF_ALIGN(cons_off, channel)) - ltt_relay_print_subbuffer_errors(channel, cons_off); -} + return 0; -static void ltt_relay_print_buffer_errors(struct ust_channel *channel) -{ - struct ltt_trace_struct *trace = channel->trace; - struct ust_buffer *ltt_buf = channel->buf; - - if (local_read(<t_buf->events_lost)) - printk(KERN_ALERT - "LTT : %s : %ld events lost " - "in %s channel.\n", - channel->channel_name, - local_read(<t_buf->events_lost), - channel->channel_name); - if (local_read(<t_buf->corrupted_subbuffers)) - printk(KERN_ALERT - "LTT : %s : %ld corrupted subbuffers " - "in %s channel.\n", - channel->channel_name, - local_read(<t_buf->corrupted_subbuffers), - channel->channel_name); - - ltt_relay_print_errors(trace, channel); + /* Jumping inside this loop occurs from within the other loop above with i as + * counter, so it unallocates the structures for the cpu = current_i down to + * zero. */ + for(; i>=0; i--) { + destroy_shm: + result = shmctl(chan->buf_struct_shmids[i], IPC_RMID, NULL); + if(result == -1) { + perror("shmctl"); + } + + destroy_previous: + continue; + } + + return -1; } -static void ltt_relay_release_channel(struct kref *kref) +static int unmap_buf_structs(struct ust_channel *chan) { - struct ust_channel *ltt_chan = container_of(kref, - struct ust_channel, kref); - free(ltt_chan->buf); + int i; + + for (i=0; i < chan->n_cpus; i++) { + if (shmdt(chan->buf[i]) < 0) { + PERROR("shmdt"); + } + } + return 0; } /* - * Create ltt buffer. + * Create channel. */ -//ust// static int ltt_relay_create_buffer(struct ltt_trace_struct *trace, -//ust// struct ltt_channel_struct *ltt_chan, struct rchan_buf *buf, -//ust// unsigned int cpu, unsigned int n_subbufs) -//ust// { -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_chan->buf, cpu); -//ust// unsigned int j; -//ust// -//ust// ltt_buf->commit_count = -//ust// kzalloc_node(sizeof(ltt_buf->commit_count) * n_subbufs, -//ust// GFP_KERNEL, cpu_to_node(cpu)); -//ust// if (!ltt_buf->commit_count) -//ust// return -ENOMEM; -//ust// kref_get(&trace->kref); -//ust// kref_get(&trace->ltt_transport_kref); -//ust// kref_get(<t_chan->kref); -//ust// local_set(<t_buf->offset, ltt_subbuffer_header_size()); -//ust// atomic_long_set(<t_buf->consumed, 0); -//ust// atomic_long_set(<t_buf->active_readers, 0); -//ust// for (j = 0; j < n_subbufs; j++) -//ust// local_set(<t_buf->commit_count[j], 0); -//ust// init_waitqueue_head(<t_buf->write_wait); -//ust// atomic_set(<t_buf->wakeup_readers, 0); -//ust// spin_lock_init(<t_buf->full_lock); -//ust// -//ust// ltt_buffer_begin_callback(buf, trace->start_tsc, 0); -//ust// /* atomic_add made on local variable on data that belongs to -//ust// * various CPUs : ok because tracing not started (for this cpu). */ -//ust// local_add(ltt_subbuffer_header_size(), <t_buf->commit_count[0]); -//ust// -//ust// local_set(<t_buf->events_lost, 0); -//ust// local_set(<t_buf->corrupted_subbuffers, 0); -//ust// -//ust// return 0; -//ust// } - -static int ust_buffers_init_buffer(struct ltt_trace_struct *trace, - struct ust_channel *ltt_chan, struct ust_buffer *buf, - unsigned int n_subbufs) +static int create_channel(const char *trace_name, struct ust_trace *trace, + const char *channel_name, struct ust_channel *chan, + unsigned int subbuf_size, unsigned int n_subbufs, int overwrite) { - unsigned int j; - int fds[2]; - int result; + int i, result; - buf->commit_count = - zmalloc(sizeof(buf->commit_count) * n_subbufs); - if (!buf->commit_count) - return -ENOMEM; - kref_get(&trace->kref); - kref_get(&trace->ltt_transport_kref); - kref_get(<t_chan->kref); - local_set(&buf->offset, ltt_subbuffer_header_size()); - atomic_long_set(&buf->consumed, 0); - atomic_long_set(&buf->active_readers, 0); - for (j = 0; j < n_subbufs; j++) - local_set(&buf->commit_count[j], 0); -//ust// init_waitqueue_head(&buf->write_wait); -//ust// atomic_set(&buf->wakeup_readers, 0); -//ust// spin_lock_init(&buf->full_lock); - - ltt_buffer_begin_callback(buf, trace->start_tsc, 0); - - local_add(ltt_subbuffer_header_size(), &buf->commit_count[0]); - - local_set(&buf->events_lost, 0); - local_set(&buf->corrupted_subbuffers, 0); + chan->trace = trace; + chan->overwrite = overwrite; + chan->n_subbufs_order = get_count_order(n_subbufs); + chan->commit_count_mask = (~0UL >> chan->n_subbufs_order); + chan->n_cpus = get_n_cpus(); - result = pipe(fds); - if(result == -1) { - PERROR("pipe"); - return -1; + /* These mappings should ideall be per-cpu, if somebody can do that + * from userspace, that would be cool! + */ + chan->buf = (void *) zmalloc(chan->n_cpus * sizeof(void *)); + if(chan->buf == NULL) { + goto error; } - buf->data_ready_fd_read = fds[0]; - buf->data_ready_fd_write = fds[1]; + chan->buf_struct_shmids = (int *) zmalloc(chan->n_cpus * sizeof(int)); + if(chan->buf_struct_shmids == NULL) + goto free_buf; - /* FIXME: do we actually need this? */ - result = fcntl(fds[0], F_SETFL, O_NONBLOCK); - if(result == -1) { - PERROR("fcntl"); + result = map_buf_structs(chan); + if(result != 0) { + goto free_buf_struct_shmids; } -//ust// buf->commit_seq = malloc(sizeof(buf->commit_seq) * n_subbufs); -//ust// if(!ltt_buf->commit_seq) { -//ust// return -1; -//ust// } - - /* FIXME: decrementally destroy on error */ + result = open_channel(chan, subbuf_size, n_subbufs); + if (result != 0) { + ERR("Cannot open channel for trace %s", trace_name); + goto unmap_buf_structs; + } return 0; + +unmap_buf_structs: + for (i=0; i < chan->n_cpus; i++) { + if (shmdt(chan->buf[i]) < 0) { + PERROR("shmdt bufstruct"); + } + } + +free_buf_struct_shmids: + free(chan->buf_struct_shmids); + +free_buf: + free(chan->buf); + +error: + return -1; } -/* FIXME: use this function */ -static void ust_buffers_destroy_buffer(struct ust_channel *ltt_chan) + +static void remove_channel(struct ust_channel *chan) { - struct ltt_trace_struct *trace = ltt_chan->trace; - struct ust_buffer *ltt_buf = ltt_chan->buf; - - kref_put(<t_chan->trace->ltt_transport_kref, - ltt_release_transport); - ltt_relay_print_buffer_errors(ltt_chan); -//ust// free(ltt_buf->commit_seq); - kfree(ltt_buf->commit_count); - ltt_buf->commit_count = NULL; - kref_put(<t_chan->kref, ltt_relay_release_channel); - kref_put(&trace->kref, ltt_release_trace); -//ust// wake_up_interruptible(&trace->kref_wq); + close_channel(chan); + + unmap_buf_structs(chan); + + free(chan->buf_struct_shmids); + + free(chan->buf); } -static void ltt_chan_alloc_ltt_buf(struct ust_channel *chan) +static void ltt_relay_async_wakeup_chan(struct ust_channel *ltt_channel) { - void *ptr; - int result; +} - /* Get one page */ - /* FIXME: increase size if we have a seq_commit array that overflows the page */ - size_t size = PAGE_ALIGN(1); +static void ltt_relay_finish_buffer(struct ust_channel *channel, unsigned int cpu) +{ + if (channel->buf[cpu]) { + struct ust_buffer *buf = channel->buf[cpu]; + ltt_force_switch(buf, FORCE_FLUSH); - result = chan->buf_shmid = shmget(getpid(), size, IPC_CREAT | IPC_EXCL | 0700); - if(chan->buf_shmid == -1) { - PERROR("shmget"); - return; + /* closing the pipe tells the consumer the buffer is finished */ + close(buf->data_ready_fd_write); } +} - ptr = shmat(chan->buf_shmid, NULL, 0); - if(ptr == (void *) -1) { - perror("shmat"); - goto destroy_shmem; - } - /* Already mark the shared memory for destruction. This will occur only - * when all users have detached. - */ - result = shmctl(chan->buf_shmid, IPC_RMID, NULL); - if(result == -1) { - perror("shmctl"); - return; +static void finish_channel(struct ust_channel *channel) +{ + unsigned int i; + + for (i=0; in_cpus; i++) { + ltt_relay_finish_buffer(channel, i); } +} - chan->buf = ptr; - return; +/* + * ltt_reserve_switch_old_subbuf: switch old subbuffer + * + * Concurrency safe because we are the last and only thread to alter this + * sub-buffer. As long as it is not delivered and read, no other thread can + * alter the offset, alter the reserve_count or call the + * client_buffer_end_callback on this sub-buffer. + * + * The only remaining threads could be the ones with pending commits. They will + * have to do the deliver themselves. Not concurrency safe in overwrite mode. + * We detect corrupted subbuffers with commit and reserve counts. We keep a + * corrupted sub-buffers count and push the readers across these sub-buffers. + * + * Not concurrency safe if a writer is stalled in a subbuffer and another writer + * switches in, finding out it's corrupted. The result will be than the old + * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer + * will be declared corrupted too because of the commit count adjustment. + * + * Note : offset_old should never be 0 here. + */ +static void ltt_reserve_switch_old_subbuf( + struct ust_channel *chan, struct ust_buffer *buf, + struct ltt_reserve_switch_offsets *offsets, u64 *tsc) +{ + long oldidx = SUBBUF_INDEX(offsets->old - 1, chan); + long commit_count, padding_size; - destroy_shmem: - result = shmctl(chan->buf_shmid, IPC_RMID, NULL); - if(result == -1) { - perror("shmctl"); - } + padding_size = chan->subbuf_size + - (SUBBUF_OFFSET(offsets->old - 1, chan) + 1); + ltt_buffer_end(buf, *tsc, offsets->old, oldidx); - return; + /* + * Must write slot data before incrementing commit count. + * This compiler barrier is upgraded into a cmm_smp_wmb() by the IPI + * sent by get_subbuf() when it does its cmm_smp_rmb(). + */ + cmm_smp_wmb(); + uatomic_add(&buf->commit_count[oldidx].cc, padding_size); + commit_count = uatomic_read(&buf->commit_count[oldidx].cc); + ltt_check_deliver(chan, buf, offsets->old - 1, commit_count, oldidx); + ltt_write_commit_counter(chan, buf, oldidx, + offsets->old, commit_count, padding_size); } /* - * Create channel. + * ltt_reserve_switch_new_subbuf: Populate new subbuffer. + * + * This code can be executed unordered : writers may already have written to the + * sub-buffer before this code gets executed, caution. The commit makes sure + * that this code is executed before the deliver of this sub-buffer. */ -static int ust_buffers_create_channel(const char *trace_name, struct ltt_trace_struct *trace, - const char *channel_name, struct ust_channel *ltt_chan, - unsigned int subbuf_size, unsigned int n_subbufs, int overwrite) +static void ltt_reserve_switch_new_subbuf( + struct ust_channel *chan, struct ust_buffer *buf, + struct ltt_reserve_switch_offsets *offsets, u64 *tsc) { - int err = 0; - int result; - - kref_init(<t_chan->kref); - - ltt_chan->trace = trace; - ltt_chan->buffer_begin = ltt_buffer_begin_callback; - ltt_chan->buffer_end = ltt_buffer_end_callback; - ltt_chan->overwrite = overwrite; - ltt_chan->n_subbufs_order = get_count_order(n_subbufs); - ltt_chan->commit_count_mask = (~0UL >> ltt_chan->n_subbufs_order); -//ust// ltt_chan->buf = percpu_alloc_mask(sizeof(struct ltt_channel_buf_struct), GFP_KERNEL, cpu_possible_map); - - ltt_chan_alloc_ltt_buf(ltt_chan); - -//ust// ltt_chan->buf = malloc(sizeof(struct ltt_channel_buf_struct)); - if (!ltt_chan->buf) - goto alloc_error; - /* FIXME: handle error of this call */ - result = ust_buffers_channel_open(ltt_chan, subbuf_size, n_subbufs); - if (result == -1) { - printk(KERN_ERR "LTT : Can't open channel for trace %s\n", - trace_name); - goto relay_open_error; - } + long beginidx = SUBBUF_INDEX(offsets->begin, chan); + long commit_count; - err = 0; - goto end; + ltt_buffer_begin(buf, *tsc, beginidx); -relay_open_error: -//ust// percpu_free(ltt_chan->buf); -alloc_error: - err = EPERM; -end: - return err; + /* + * Must write slot data before incrementing commit count. + * This compiler barrier is upgraded into a cmm_smp_wmb() by the IPI + * sent by get_subbuf() when it does its cmm_smp_rmb(). + */ + cmm_smp_wmb(); + uatomic_add(&buf->commit_count[beginidx].cc, ltt_subbuffer_header_size()); + commit_count = uatomic_read(&buf->commit_count[beginidx].cc); + /* Check if the written buffer has to be delivered */ + ltt_check_deliver(chan, buf, offsets->begin, commit_count, beginidx); + ltt_write_commit_counter(chan, buf, beginidx, + offsets->begin, commit_count, ltt_subbuffer_header_size()); } /* - * LTTng channel flush function. + * ltt_reserve_end_switch_current: finish switching current subbuffer + * + * Concurrency safe because we are the last and only thread to alter this + * sub-buffer. As long as it is not delivered and read, no other thread can + * alter the offset, alter the reserve_count or call the + * client_buffer_end_callback on this sub-buffer. * - * Must be called when no tracing is active in the channel, because of - * accesses across CPUs. + * The only remaining threads could be the ones with pending commits. They will + * have to do the deliver themselves. Not concurrency safe in overwrite mode. + * We detect corrupted subbuffers with commit and reserve counts. We keep a + * corrupted sub-buffers count and push the readers across these sub-buffers. + * + * Not concurrency safe if a writer is stalled in a subbuffer and another writer + * switches in, finding out it's corrupted. The result will be than the old + * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer + * will be declared corrupted too because of the commit count adjustment. */ -static notrace void ltt_relay_buffer_flush(struct ust_buffer *buf) +static void ltt_reserve_end_switch_current( + struct ust_channel *chan, + struct ust_buffer *buf, + struct ltt_reserve_switch_offsets *offsets, u64 *tsc) { - int result; + long endidx = SUBBUF_INDEX(offsets->end - 1, chan); + long commit_count, padding_size; -//ust// buf->finalized = 1; - ltt_force_switch(buf, FORCE_FLUSH); + padding_size = chan->subbuf_size + - (SUBBUF_OFFSET(offsets->end - 1, chan) + 1); - result = write(buf->data_ready_fd_write, "1", 1); - if(result == -1) { - PERROR("write (in ltt_relay_buffer_flush)"); - ERR("this should never happen!"); - } -} + ltt_buffer_end(buf, *tsc, offsets->end, endidx); -static void ltt_relay_async_wakeup_chan(struct ust_channel *ltt_channel) -{ -//ust// unsigned int i; -//ust// struct rchan *rchan = ltt_channel->trans_channel_data; -//ust// -//ust// for_each_possible_cpu(i) { -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_channel->buf, i); -//ust// -//ust// if (atomic_read(<t_buf->wakeup_readers) == 1) { -//ust// atomic_set(<t_buf->wakeup_readers, 0); -//ust// wake_up_interruptible(&rchan->buf[i]->read_wait); -//ust// } -//ust// } + /* + * Must write slot data before incrementing commit count. + * This compiler barrier is upgraded into a cmm_smp_wmb() by the IPI + * sent by get_subbuf() when it does its cmm_smp_rmb(). + */ + cmm_smp_wmb(); + uatomic_add(&buf->commit_count[endidx].cc, padding_size); + commit_count = uatomic_read(&buf->commit_count[endidx].cc); + ltt_check_deliver(chan, buf, + offsets->end - 1, commit_count, endidx); + ltt_write_commit_counter(chan, buf, endidx, + offsets->end, commit_count, padding_size); } -static void ltt_relay_finish_buffer(struct ust_channel *channel) +/* + * Returns : + * 0 if ok + * !0 if execution must be aborted. + */ +static int ltt_relay_try_switch_slow( + enum force_switch_mode mode, + struct ust_channel *chan, + struct ust_buffer *buf, + struct ltt_reserve_switch_offsets *offsets, + u64 *tsc) { -// int result; + long subbuf_index; + long reserve_commit_diff; - if (channel->buf) { - struct ust_buffer *buf = channel->buf; - ltt_relay_buffer_flush(buf); -//ust// ltt_relay_wake_writers(ltt_buf); - /* closing the pipe tells the consumer the buffer is finished */ - - //result = write(ltt_buf->data_ready_fd_write, "D", 1); - //if(result == -1) { - // PERROR("write (in ltt_relay_finish_buffer)"); - // ERR("this should never happen!"); - //} - close(buf->data_ready_fd_write); + offsets->begin = uatomic_read(&buf->offset); + offsets->old = offsets->begin; + offsets->begin_switch = 0; + offsets->end_switch_old = 0; + + *tsc = trace_clock_read64(); + + if (SUBBUF_OFFSET(offsets->begin, buf->chan) != 0) { + offsets->begin = SUBBUF_ALIGN(offsets->begin, buf->chan); + offsets->end_switch_old = 1; + } else { + /* we do not have to switch : buffer is empty */ + return -1; } + if (mode == FORCE_ACTIVE) + offsets->begin += ltt_subbuffer_header_size(); + /* + * Always begin_switch in FORCE_ACTIVE mode. + * Test new buffer integrity + */ + subbuf_index = SUBBUF_INDEX(offsets->begin, buf->chan); + reserve_commit_diff = + (BUFFER_TRUNC(offsets->begin, buf->chan) + >> chan->n_subbufs_order) + - (uatomic_read(&buf->commit_count[subbuf_index].cc_sb) + & chan->commit_count_mask); + if (reserve_commit_diff == 0) { + /* Next buffer not corrupted. */ + if (mode == FORCE_ACTIVE + && !chan->overwrite + && offsets->begin - uatomic_read(&buf->consumed) + >= chan->alloc_size) { + /* + * We do not overwrite non consumed buffers and we are + * full : ignore switch while tracing is active. + */ + return -1; + } + } else { + /* + * Next subbuffer corrupted. Force pushing reader even in normal + * mode + */ + } + offsets->end = offsets->begin; + return 0; } - -static void ltt_relay_finish_channel(struct ust_channel *channel) +/* + * Force a sub-buffer switch for a per-cpu buffer. This operation is + * completely reentrant : can be called while tracing is active with + * absolutely no lock held. + */ +void ltt_force_switch_lockless_slow(struct ust_buffer *buf, + enum force_switch_mode mode) { -//ust// unsigned int i; + struct ust_channel *chan = buf->chan; + struct ltt_reserve_switch_offsets offsets; + u64 tsc; -//ust// for_each_possible_cpu(i) - ltt_relay_finish_buffer(channel); -} + offsets.size = 0; -static void ltt_relay_remove_channel(struct ust_channel *channel) -{ - ust_buffers_channel_close(channel); - kref_put(&channel->kref, ltt_relay_release_channel); -} + DBG("Switching (forced) %s_%d", chan->channel_name, buf->cpu); + /* + * Perform retryable operations. + */ + do { + if (ltt_relay_try_switch_slow(mode, chan, buf, + &offsets, &tsc)) + return; + } while (uatomic_cmpxchg(&buf->offset, offsets.old, + offsets.end) != offsets.old); -struct ltt_reserve_switch_offsets { - long begin, end, old; - long begin_switch, end_switch_current, end_switch_old; - long commit_count, reserve_commit_diff; - size_t before_hdr_pad, size; -}; + /* + * Atomically update last_tsc. This update races against concurrent + * atomic updates, but the race will always cause supplementary full TSC + * events, never the opposite (missing a full TSC event when it would be + * needed). + */ + save_last_tsc(buf, tsc); + + /* + * Push the reader if necessary + */ + if (mode == FORCE_ACTIVE) { + ltt_reserve_push_reader(chan, buf, offsets.end - 1); + } + + /* + * Switch old subbuffer if needed. + */ + if (offsets.end_switch_old) { + ltt_reserve_switch_old_subbuf(chan, buf, &offsets, &tsc); + } + + /* + * Populate new subbuffer. + */ + if (mode == FORCE_ACTIVE) + ltt_reserve_switch_new_subbuf(chan, buf, &offsets, &tsc); +} /* * Returns : * 0 if ok * !0 if execution must be aborted. */ -static inline int ltt_relay_try_reserve( - struct ust_channel *channel, struct ust_buffer *buf, +static int ltt_relay_try_reserve_slow(struct ust_channel *chan, struct ust_buffer *buf, struct ltt_reserve_switch_offsets *offsets, size_t data_size, u64 *tsc, unsigned int *rflags, int largest_align) { - offsets->begin = local_read(&buf->offset); + long reserve_commit_diff; + + offsets->begin = uatomic_read(&buf->offset); offsets->old = offsets->begin; offsets->begin_switch = 0; offsets->end_switch_current = 0; @@ -815,59 +897,51 @@ static inline int ltt_relay_try_reserve( if (last_tsc_overflow(buf, *tsc)) *rflags = LTT_RFLAG_ID_SIZE_TSC; - if (SUBBUF_OFFSET(offsets->begin, buf->chan) == 0) { + if (unlikely(SUBBUF_OFFSET(offsets->begin, buf->chan) == 0)) { offsets->begin_switch = 1; /* For offsets->begin */ } else { - offsets->size = ust_get_header_size(channel, + offsets->size = ust_get_header_size(chan, offsets->begin, data_size, &offsets->before_hdr_pad, *rflags); offsets->size += ltt_align(offsets->begin + offsets->size, largest_align) + data_size; - if ((SUBBUF_OFFSET(offsets->begin, buf->chan) + offsets->size) - > buf->chan->subbuf_size) { + if (unlikely((SUBBUF_OFFSET(offsets->begin, buf->chan) + + offsets->size) > buf->chan->subbuf_size)) { offsets->end_switch_old = 1; /* For offsets->old */ offsets->begin_switch = 1; /* For offsets->begin */ } } - if (offsets->begin_switch) { + if (unlikely(offsets->begin_switch)) { long subbuf_index; - if (offsets->end_switch_old) + /* + * We are typically not filling the previous buffer completely. + */ + if (likely(offsets->end_switch_old)) offsets->begin = SUBBUF_ALIGN(offsets->begin, buf->chan); offsets->begin = offsets->begin + ltt_subbuffer_header_size(); /* Test new buffer integrity */ subbuf_index = SUBBUF_INDEX(offsets->begin, buf->chan); - offsets->reserve_commit_diff = - (BUFFER_TRUNC(offsets->begin, buf->chan) - >> channel->n_subbufs_order) - - (local_read(&buf->commit_count[subbuf_index]) - & channel->commit_count_mask); - if (offsets->reserve_commit_diff == 0) { - long consumed; - - consumed = atomic_long_read(&buf->consumed); - + reserve_commit_diff = + (BUFFER_TRUNC(offsets->begin, buf->chan) + >> chan->n_subbufs_order) + - (uatomic_read(&buf->commit_count[subbuf_index].cc_sb) + & chan->commit_count_mask); + if (likely(reserve_commit_diff == 0)) { /* Next buffer not corrupted. */ - if (!channel->overwrite && + if (unlikely(!chan->overwrite && (SUBBUF_TRUNC(offsets->begin, buf->chan) - - SUBBUF_TRUNC(consumed, buf->chan)) - >= channel->alloc_size) { - - long consumed_idx = SUBBUF_INDEX(consumed, buf->chan); - long commit_count = local_read(&buf->commit_count[consumed_idx]); - if(((commit_count - buf->chan->subbuf_size) & channel->commit_count_mask) - (BUFFER_TRUNC(consumed, buf->chan) >> channel->n_subbufs_order) != 0) { - WARN("Event dropped. Caused by non-committed event."); - } - else { - WARN("Event dropped. Caused by non-consumed buffer."); - } + - SUBBUF_TRUNC(uatomic_read( + &buf->consumed), + buf->chan)) + >= chan->alloc_size)) { /* * We do not overwrite non consumed buffers * and we are full : event is lost. */ - local_inc(&buf->events_lost); + uatomic_inc(&buf->events_lost); return -1; } else { /* @@ -878,24 +952,26 @@ static inline int ltt_relay_try_reserve( } } else { /* - * Next subbuffer corrupted. Force pushing reader even - * in normal mode. It's safe to write in this new - * subbuffer. + * Next subbuffer corrupted. Drop event in normal and + * overwrite mode. Caused by either a writer OOPS or + * too many nested writes over a reserve/commit pair. */ + uatomic_inc(&buf->events_lost); + return -1; } - offsets->size = ust_get_header_size(channel, + offsets->size = ust_get_header_size(chan, offsets->begin, data_size, &offsets->before_hdr_pad, *rflags); offsets->size += ltt_align(offsets->begin + offsets->size, largest_align) + data_size; - if ((SUBBUF_OFFSET(offsets->begin, buf->chan) + offsets->size) - > buf->chan->subbuf_size) { + if (unlikely((SUBBUF_OFFSET(offsets->begin, buf->chan) + + offsets->size) > buf->chan->subbuf_size)) { /* * Event too big for subbuffers, report error, don't * complete the sub-buffer switch. */ - local_inc(&buf->events_lost); + uatomic_inc(&buf->events_lost); return -1; } else { /* @@ -911,7 +987,7 @@ static inline int ltt_relay_try_reserve( } offsets->end = offsets->begin + offsets->size; - if ((SUBBUF_OFFSET(offsets->end, buf->chan)) == 0) { + if (unlikely((SUBBUF_OFFSET(offsets->end, buf->chan)) == 0)) { /* * The offset_end will fall at the very beginning of the next * subbuffer. @@ -921,252 +997,8 @@ static inline int ltt_relay_try_reserve( return 0; } -/* - * Returns : - * 0 if ok - * !0 if execution must be aborted. - */ -static inline int ltt_relay_try_switch( - enum force_switch_mode mode, - struct ust_channel *channel, - struct ust_buffer *buf, - struct ltt_reserve_switch_offsets *offsets, - u64 *tsc) -{ - long subbuf_index; - - offsets->begin = local_read(&buf->offset); - offsets->old = offsets->begin; - offsets->begin_switch = 0; - offsets->end_switch_old = 0; - - *tsc = trace_clock_read64(); - - if (SUBBUF_OFFSET(offsets->begin, buf->chan) != 0) { - offsets->begin = SUBBUF_ALIGN(offsets->begin, buf->chan); - offsets->end_switch_old = 1; - } else { - /* we do not have to switch : buffer is empty */ - return -1; - } - if (mode == FORCE_ACTIVE) - offsets->begin += ltt_subbuffer_header_size(); - /* - * Always begin_switch in FORCE_ACTIVE mode. - * Test new buffer integrity - */ - subbuf_index = SUBBUF_INDEX(offsets->begin, buf->chan); - offsets->reserve_commit_diff = - (BUFFER_TRUNC(offsets->begin, buf->chan) - >> channel->n_subbufs_order) - - (local_read(&buf->commit_count[subbuf_index]) - & channel->commit_count_mask); - if (offsets->reserve_commit_diff == 0) { - /* Next buffer not corrupted. */ - if (mode == FORCE_ACTIVE - && !channel->overwrite - && offsets->begin - atomic_long_read(&buf->consumed) - >= channel->alloc_size) { - /* - * We do not overwrite non consumed buffers and we are - * full : ignore switch while tracing is active. - */ - return -1; - } - } else { - /* - * Next subbuffer corrupted. Force pushing reader even in normal - * mode - */ - } - offsets->end = offsets->begin; - return 0; -} - -static inline void ltt_reserve_push_reader( - struct ust_channel *channel, - struct ust_buffer *buf, - struct ltt_reserve_switch_offsets *offsets) -{ - long consumed_old, consumed_new; - - do { - consumed_old = atomic_long_read(&buf->consumed); - /* - * If buffer is in overwrite mode, push the reader consumed - * count if the write position has reached it and we are not - * at the first iteration (don't push the reader farther than - * the writer). This operation can be done concurrently by many - * writers in the same buffer, the writer being at the farthest - * write position sub-buffer index in the buffer being the one - * which will win this loop. - * If the buffer is not in overwrite mode, pushing the reader - * only happens if a sub-buffer is corrupted. - */ - if ((SUBBUF_TRUNC(offsets->end-1, buf->chan) - - SUBBUF_TRUNC(consumed_old, buf->chan)) - >= channel->alloc_size) - consumed_new = SUBBUF_ALIGN(consumed_old, buf->chan); - else { - consumed_new = consumed_old; - break; - } - } while (atomic_long_cmpxchg(&buf->consumed, consumed_old, - consumed_new) != consumed_old); - - if (consumed_old != consumed_new) { - /* - * Reader pushed : we are the winner of the push, we can - * therefore reequilibrate reserve and commit. Atomic increment - * of the commit count permits other writers to play around - * with this variable before us. We keep track of - * corrupted_subbuffers even in overwrite mode : - * we never want to write over a non completely committed - * sub-buffer : possible causes : the buffer size is too low - * compared to the unordered data input, or there is a writer - * that died between the reserve and the commit. - */ - if (offsets->reserve_commit_diff) { - /* - * We have to alter the sub-buffer commit count. - * We do not deliver the previous subbuffer, given it - * was either corrupted or not consumed (overwrite - * mode). - */ - local_add(offsets->reserve_commit_diff, - &buf->commit_count[ - SUBBUF_INDEX(offsets->begin, - buf->chan)]); - if (!channel->overwrite - || offsets->reserve_commit_diff - != channel->subbuf_size) { - /* - * The reserve commit diff was not subbuf_size : - * it means the subbuffer was partly written to - * and is therefore corrupted. If it is multiple - * of subbuffer size and we are in flight - * recorder mode, we are skipping over a whole - * subbuffer. - */ - local_inc(&buf->corrupted_subbuffers); - } - } - } -} - - -/* - * ltt_reserve_switch_old_subbuf: switch old subbuffer - * - * Concurrency safe because we are the last and only thread to alter this - * sub-buffer. As long as it is not delivered and read, no other thread can - * alter the offset, alter the reserve_count or call the - * client_buffer_end_callback on this sub-buffer. - * - * The only remaining threads could be the ones with pending commits. They will - * have to do the deliver themselves. Not concurrency safe in overwrite mode. - * We detect corrupted subbuffers with commit and reserve counts. We keep a - * corrupted sub-buffers count and push the readers across these sub-buffers. - * - * Not concurrency safe if a writer is stalled in a subbuffer and another writer - * switches in, finding out it's corrupted. The result will be than the old - * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer - * will be declared corrupted too because of the commit count adjustment. - * - * Note : offset_old should never be 0 here. - */ -static inline void ltt_reserve_switch_old_subbuf( - struct ust_channel *channel, - struct ust_buffer *buf, - struct ltt_reserve_switch_offsets *offsets, u64 *tsc) -{ - long oldidx = SUBBUF_INDEX(offsets->old - 1, channel); - - channel->buffer_end(buf, *tsc, offsets->old, oldidx); - /* Must write buffer end before incrementing commit count */ - smp_wmb(); - offsets->commit_count = - local_add_return(channel->subbuf_size - - (SUBBUF_OFFSET(offsets->old - 1, channel) - + 1), - &buf->commit_count[oldidx]); - if ((BUFFER_TRUNC(offsets->old - 1, channel) - >> channel->n_subbufs_order) - - ((offsets->commit_count - channel->subbuf_size) - & channel->commit_count_mask) == 0) - ltt_deliver(buf, oldidx, offsets->commit_count); -} - -/* - * ltt_reserve_switch_new_subbuf: Populate new subbuffer. - * - * This code can be executed unordered : writers may already have written to the - * sub-buffer before this code gets executed, caution. The commit makes sure - * that this code is executed before the deliver of this sub-buffer. - */ -static /*inline*/ void ltt_reserve_switch_new_subbuf( - struct ust_channel *channel, - struct ust_buffer *buf, - struct ltt_reserve_switch_offsets *offsets, u64 *tsc) -{ - long beginidx = SUBBUF_INDEX(offsets->begin, channel); - - channel->buffer_begin(buf, *tsc, beginidx); - /* Must write buffer end before incrementing commit count */ - smp_wmb(); - offsets->commit_count = local_add_return(ltt_subbuffer_header_size(), - &buf->commit_count[beginidx]); - /* Check if the written buffer has to be delivered */ - if ((BUFFER_TRUNC(offsets->begin, channel) - >> channel->n_subbufs_order) - - ((offsets->commit_count - channel->subbuf_size) - & channel->commit_count_mask) == 0) - ltt_deliver(buf, beginidx, offsets->commit_count); -} - - -/* - * ltt_reserve_end_switch_current: finish switching current subbuffer - * - * Concurrency safe because we are the last and only thread to alter this - * sub-buffer. As long as it is not delivered and read, no other thread can - * alter the offset, alter the reserve_count or call the - * client_buffer_end_callback on this sub-buffer. - * - * The only remaining threads could be the ones with pending commits. They will - * have to do the deliver themselves. Not concurrency safe in overwrite mode. - * We detect corrupted subbuffers with commit and reserve counts. We keep a - * corrupted sub-buffers count and push the readers across these sub-buffers. - * - * Not concurrency safe if a writer is stalled in a subbuffer and another writer - * switches in, finding out it's corrupted. The result will be than the old - * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer - * will be declared corrupted too because of the commit count adjustment. - */ -static inline void ltt_reserve_end_switch_current( - struct ust_channel *channel, - struct ust_buffer *buf, - struct ltt_reserve_switch_offsets *offsets, u64 *tsc) -{ - long endidx = SUBBUF_INDEX(offsets->end - 1, channel); - - channel->buffer_end(buf, *tsc, offsets->end, endidx); - /* Must write buffer begin before incrementing commit count */ - smp_wmb(); - offsets->commit_count = - local_add_return(channel->subbuf_size - - (SUBBUF_OFFSET(offsets->end - 1, channel) - + 1), - &buf->commit_count[endidx]); - if ((BUFFER_TRUNC(offsets->end - 1, channel) - >> channel->n_subbufs_order) - - ((offsets->commit_count - channel->subbuf_size) - & channel->commit_count_mask) == 0) - ltt_deliver(buf, endidx, offsets->commit_count); -} - /** - * ltt_relay_reserve_slot - Atomic slot reservation in a LTTng buffer. + * ltt_relay_reserve_slot_lockless_slow - Atomic slot reservation in a buffer. * @trace: the trace structure to log to. * @ltt_channel: channel structure * @transport_data: data structure specific to ltt relay @@ -1179,30 +1011,24 @@ static inline void ltt_reserve_end_switch_current( * Return : -ENOSPC if not enough space, else returns 0. * It will take care of sub-buffer switching. */ -static notrace int ltt_relay_reserve_slot(struct ltt_trace_struct *trace, - struct ust_channel *channel, void **transport_data, - size_t data_size, size_t *slot_size, long *buf_offset, u64 *tsc, - unsigned int *rflags, int largest_align) +int ltt_reserve_slot_lockless_slow(struct ust_channel *chan, + struct ust_trace *trace, size_t data_size, + int largest_align, int cpu, + struct ust_buffer **ret_buf, + size_t *slot_size, long *buf_offset, + u64 *tsc, unsigned int *rflags) { - struct ust_buffer *buf = *transport_data = channel->buf; + struct ust_buffer *buf = *ret_buf = chan->buf[cpu]; struct ltt_reserve_switch_offsets offsets; - offsets.reserve_commit_diff = 0; offsets.size = 0; - /* - * Perform retryable operations. - */ - if (ltt_nesting > 4) { - local_inc(&buf->events_lost); - return -EPERM; - } do { - if (ltt_relay_try_reserve(channel, buf, &offsets, data_size, tsc, rflags, - largest_align)) + if (unlikely(ltt_relay_try_reserve_slow(chan, buf, &offsets, + data_size, tsc, rflags, largest_align))) return -ENOSPC; - } while (local_cmpxchg(&buf->offset, offsets.old, - offsets.end) != offsets.old); + } while (unlikely(uatomic_cmpxchg(&buf->offset, offsets.old, + offsets.end) != offsets.old)); /* * Atomically update last_tsc. This update races against concurrent @@ -1215,197 +1041,40 @@ static notrace int ltt_relay_reserve_slot(struct ltt_trace_struct *trace, /* * Push the reader if necessary */ - ltt_reserve_push_reader(channel, buf, &offsets); + ltt_reserve_push_reader(chan, buf, offsets.end - 1); /* * Switch old subbuffer if needed. */ - if (offsets.end_switch_old) - ltt_reserve_switch_old_subbuf(channel, buf, &offsets, tsc); + if (unlikely(offsets.end_switch_old)) { + ltt_reserve_switch_old_subbuf(chan, buf, &offsets, tsc); + DBG("Switching %s_%d", chan->channel_name, cpu); + } /* * Populate new subbuffer. */ - if (offsets.begin_switch) - ltt_reserve_switch_new_subbuf(channel, buf, &offsets, tsc); + if (unlikely(offsets.begin_switch)) + ltt_reserve_switch_new_subbuf(chan, buf, &offsets, tsc); - if (offsets.end_switch_current) - ltt_reserve_end_switch_current(channel, buf, &offsets, tsc); + if (unlikely(offsets.end_switch_current)) + ltt_reserve_end_switch_current(chan, buf, &offsets, tsc); *slot_size = offsets.size; *buf_offset = offsets.begin + offsets.before_hdr_pad; return 0; } -/* - * Force a sub-buffer switch for a per-cpu buffer. This operation is - * completely reentrant : can be called while tracing is active with - * absolutely no lock held. - * - * Note, however, that as a local_cmpxchg is used for some atomic - * operations, this function must be called from the CPU which owns the buffer - * for a ACTIVE flush. - */ -static notrace void ltt_force_switch(struct ust_buffer *buf, - enum force_switch_mode mode) -{ - struct ust_channel *channel = buf->chan; - struct ltt_reserve_switch_offsets offsets; - u64 tsc; - - offsets.reserve_commit_diff = 0; - offsets.size = 0; - - /* - * Perform retryable operations. - */ - do { - if (ltt_relay_try_switch(mode, channel, buf, &offsets, &tsc)) - return; - } while (local_cmpxchg(&buf->offset, offsets.old, - offsets.end) != offsets.old); - - /* - * Atomically update last_tsc. This update races against concurrent - * atomic updates, but the race will always cause supplementary full TSC - * events, never the opposite (missing a full TSC event when it would be - * needed). - */ - save_last_tsc(buf, tsc); - - /* - * Push the reader if necessary - */ - if (mode == FORCE_ACTIVE) - ltt_reserve_push_reader(channel, buf, &offsets); - - /* - * Switch old subbuffer if needed. - */ - if (offsets.end_switch_old) - ltt_reserve_switch_old_subbuf(channel, buf, &offsets, &tsc); - - /* - * Populate new subbuffer. - */ - if (mode == FORCE_ACTIVE) - ltt_reserve_switch_new_subbuf(channel, buf, &offsets, &tsc); -} - -static void ltt_relay_print_user_errors(struct ltt_trace_struct *trace, - unsigned int chan_index, size_t data_size, - struct user_dbg_data *dbg) -{ - struct ust_channel *channel; - struct ust_buffer *buf; - - channel = &trace->channels[chan_index]; - buf = channel->buf; - - printk(KERN_ERR "Error in LTT usertrace : " - "buffer full : event lost in blocking " - "mode. Increase LTT_RESERVE_CRITICAL.\n"); - printk(KERN_ERR "LTT nesting level is %u.\n", ltt_nesting); - printk(KERN_ERR "LTT avail size %lu.\n", - dbg->avail_size); - printk(KERN_ERR "avai write : %lu, read : %lu\n", - dbg->write, dbg->read); - - dbg->write = local_read(&buf->offset); - dbg->read = atomic_long_read(&buf->consumed); - - printk(KERN_ERR "LTT cur size %lu.\n", - dbg->write + LTT_RESERVE_CRITICAL + data_size - - SUBBUF_TRUNC(dbg->read, channel)); - printk(KERN_ERR "cur write : %lu, read : %lu\n", - dbg->write, dbg->read); -} - static struct ltt_transport ust_relay_transport = { .name = "ustrelay", .ops = { - .create_channel = ust_buffers_create_channel, - .finish_channel = ltt_relay_finish_channel, - .remove_channel = ltt_relay_remove_channel, + .create_channel = create_channel, + .finish_channel = finish_channel, + .remove_channel = remove_channel, .wakeup_channel = ltt_relay_async_wakeup_chan, -// .commit_slot = ltt_relay_commit_slot, - .reserve_slot = ltt_relay_reserve_slot, - .user_errors = ltt_relay_print_user_errors, }, }; -/* - * for flight recording. must be called after relay_commit. - * This function decrements de subbuffer's lost_size each time the commit count - * reaches back the reserve offset (module subbuffer size). It is useful for - * crash dump. - */ -static /* inline */ void ltt_write_commit_counter(struct ust_buffer *buf, - struct ust_buffer *ltt_buf, - long idx, long buf_offset, long commit_count, size_t data_size) -{ - long offset; - long commit_seq_old; - - offset = buf_offset + data_size; - - /* - * SUBBUF_OFFSET includes commit_count_mask. We can simply - * compare the offsets within the subbuffer without caring about - * buffer full/empty mismatch because offset is never zero here - * (subbuffer header and event headers have non-zero length). - */ - if (unlikely(SUBBUF_OFFSET(offset - commit_count, buf->chan))) - return; - - commit_seq_old = local_read(<t_buf->commit_seq[idx]); - while (commit_seq_old < commit_count) - commit_seq_old = local_cmpxchg(<t_buf->commit_seq[idx], - commit_seq_old, commit_count); -} - -/* - * Atomic unordered slot commit. Increments the commit count in the - * specified sub-buffer, and delivers it if necessary. - * - * Parameters: - * - * @ltt_channel : channel structure - * @transport_data: transport-specific data - * @buf_offset : offset following the event header. - * @data_size : size of the event data. - * @slot_size : size of the reserved slot. - */ -/* FIXME: make this function static inline in the .h! */ -/*static*/ /* inline */ notrace void ltt_commit_slot( - struct ust_channel *channel, - void **transport_data, long buf_offset, - size_t data_size, size_t slot_size) -{ - struct ust_buffer *buf = *transport_data; - long offset_end = buf_offset; - long endidx = SUBBUF_INDEX(offset_end - 1, channel); - long commit_count; - - /* Must write slot data before incrementing commit count */ - smp_wmb(); - commit_count = local_add_return(slot_size, - &buf->commit_count[endidx]); - /* Check if all commits have been done */ - if ((BUFFER_TRUNC(offset_end - 1, channel) - >> channel->n_subbufs_order) - - ((commit_count - channel->subbuf_size) - & channel->commit_count_mask) == 0) - ltt_deliver(buf, endidx, commit_count); - /* - * Update lost_size for each commit. It's needed only for extracting - * ltt buffers from vmcore, after crash. - */ - ltt_write_commit_counter(buf, buf, endidx, - buf_offset, commit_count, data_size); -} - - static char initialized = 0; void __attribute__((constructor)) init_ustrelay_transport(void) @@ -1416,7 +1085,78 @@ void __attribute__((constructor)) init_ustrelay_transport(void) } } -static void __attribute__((destructor)) ltt_relay_exit(void) +static void __attribute__((destructor)) ust_buffers_exit(void) { ltt_transport_unregister(&ust_relay_transport); } + +size_t ltt_write_event_header_slow(struct ust_channel *channel, + struct ust_buffer *buf, long buf_offset, + u16 eID, u32 event_size, + u64 tsc, unsigned int rflags) +{ + struct ltt_event_header header; + u16 small_size; + + switch (rflags) { + case LTT_RFLAG_ID_SIZE_TSC: + header.id_time = 29 << LTT_TSC_BITS; + break; + case LTT_RFLAG_ID_SIZE: + header.id_time = 30 << LTT_TSC_BITS; + break; + case LTT_RFLAG_ID: + header.id_time = 31 << LTT_TSC_BITS; + break; + default: + WARN_ON_ONCE(1); + header.id_time = 0; + break; + } + + header.id_time |= (u32)tsc & LTT_TSC_MASK; + ust_buffers_write(buf, buf_offset, &header, sizeof(header)); + buf_offset += sizeof(header); + + switch (rflags) { + case LTT_RFLAG_ID_SIZE_TSC: + small_size = (u16)min_t(u32, event_size, LTT_MAX_SMALL_SIZE); + ust_buffers_write(buf, buf_offset, + &eID, sizeof(u16)); + buf_offset += sizeof(u16); + ust_buffers_write(buf, buf_offset, + &small_size, sizeof(u16)); + buf_offset += sizeof(u16); + if (small_size == LTT_MAX_SMALL_SIZE) { + ust_buffers_write(buf, buf_offset, + &event_size, sizeof(u32)); + buf_offset += sizeof(u32); + } + buf_offset += ltt_align(buf_offset, sizeof(u64)); + ust_buffers_write(buf, buf_offset, + &tsc, sizeof(u64)); + buf_offset += sizeof(u64); + break; + case LTT_RFLAG_ID_SIZE: + small_size = (u16)min_t(u32, event_size, LTT_MAX_SMALL_SIZE); + ust_buffers_write(buf, buf_offset, + &eID, sizeof(u16)); + buf_offset += sizeof(u16); + ust_buffers_write(buf, buf_offset, + &small_size, sizeof(u16)); + buf_offset += sizeof(u16); + if (small_size == LTT_MAX_SMALL_SIZE) { + ust_buffers_write(buf, buf_offset, + &event_size, sizeof(u32)); + buf_offset += sizeof(u32); + } + break; + case LTT_RFLAG_ID: + ust_buffers_write(buf, buf_offset, + &eID, sizeof(u16)); + buf_offset += sizeof(u16); + break; + } + + return buf_offset; +}