Skip to content
Snippets Groups Projects
builtin-report.c 27.6 KiB
Newer Older
  • Learn to ignore specific revisions
  • /*
     * builtin-report.c
     *
     * Builtin report command: Analyze the perf.data input file,
     * look up and read DSOs and symbol information and display
     * a histogram of results, along various sorting keys.
     */
    
    #include "builtin.h"
    
    #include "util/color.h"
    
    #include "util/cache.h"
    
    #include <linux/rbtree.h>
    
    #include "util/string.h"
    
    #include "util/strlist.h"
    
    #include "perf.h"
    
    #include "util/debug.h"
    
    #include "util/header.h"
    
    
    #include "util/parse-options.h"
    #include "util/parse-events.h"
    
    
    #include "util/sort.h"
    
    static char		const *input_name = "perf.data";
    
    static char		*dso_list_str, *comm_list_str, *sym_list_str,
    			*col_width_list_str;
    
    static struct strlist	*dso_list, *comm_list, *sym_list;
    
    static int		force;
    
    static int		show_nr_samples;
    
    static int		show_threads;
    static struct perf_read_values	show_threads_values;
    
    
    static char		default_pretty_printing_style[] = "normal";
    static char		*pretty_printing_style = default_pretty_printing_style;
    
    
    static unsigned long	page_size;
    static unsigned long	mmap_window = 32;
    
    
    static int		exclude_other = 1;
    
    static char		callchain_default_opt[] = "fractal,0.5";
    
    
    static char		__cwd[PATH_MAX];
    static char		*cwd = __cwd;
    static int		cwdlen;
    
    
    static struct rb_root	threads;
    static struct thread	*last_match;
    
    
    static u64		sample_type;
    
    
    static size_t ipchain__fprintf_graph_line(FILE *fp, int depth, int depth_mask)
    {
    	int i;
    	size_t ret = 0;
    
    	ret += fprintf(fp, "%s", "                ");
    
    	for (i = 0; i < depth; i++)
    		if (depth_mask & (1 << i))
    			ret += fprintf(fp, "|          ");
    		else
    			ret += fprintf(fp, "           ");
    
    	ret += fprintf(fp, "\n");
    
    	return ret;
    }
    
    ipchain__fprintf_graph(FILE *fp, struct callchain_list *chain, int depth,
    		       int depth_mask, int count, u64 total_samples,
    		       int hits)
    {
    	int i;
    	size_t ret = 0;
    
    	ret += fprintf(fp, "%s", "                ");
    	for (i = 0; i < depth; i++) {
    		if (depth_mask & (1 << i))
    			ret += fprintf(fp, "|");
    		else
    			ret += fprintf(fp, " ");
    		if (!count && i == depth - 1) {
    			double percent;
    
    			percent = hits * 100.0 / total_samples;
    
    			ret += percent_color_fprintf(fp, "--%2.2f%%-- ", percent);
    
    		} else
    			ret += fprintf(fp, "%s", "          ");
    	}
    	if (chain->sym)
    		ret += fprintf(fp, "%s\n", chain->sym->name);
    	else
    		ret += fprintf(fp, "%p\n", (void *)(long)chain->ip);
    
    	return ret;
    }
    
    
    static struct symbol *rem_sq_bracket;
    static struct callchain_list rem_hits;
    
    static void init_rem_hits(void)
    {
    	rem_sq_bracket = malloc(sizeof(*rem_sq_bracket) + 6);
    	if (!rem_sq_bracket) {
    		fprintf(stderr, "Not enough memory to display remaining hits\n");
    		return;
    	}
    
    	strcpy(rem_sq_bracket->name, "[...]");
    	rem_hits.sym = rem_sq_bracket;
    }
    
    
    static size_t
    callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
    			u64 total_samples, int depth, int depth_mask)
    {
    	struct rb_node *node, *next;
    	struct callchain_node *child;
    	struct callchain_list *chain;
    	int new_depth_mask = depth_mask;
    
    	node = rb_first(&self->rb_root);
    	while (node) {
    
    		child = rb_entry(node, struct callchain_node, rb_node);
    
    
    		/*
    		 * The depth mask manages the output of pipes that show
    		 * the depth. We don't want to keep the pipes of the current
    
    		 * level for the last child of this depth.
    		 * Except if we have remaining filtered hits. They will
    		 * supersede the last child
    
    		if (!next && (callchain_param.mode != CHAIN_GRAPH_REL || !remaining))
    
    			new_depth_mask &= ~(1 << (depth - 1));
    
    		/*
    		 * But we keep the older depth mask for the line seperator
    		 * to keep the level link until we reach the last child
    		 */
    		ret += ipchain__fprintf_graph_line(fp, depth, depth_mask);
    		i = 0;
    		list_for_each_entry(chain, &child->val, list) {
    			if (chain->ip >= PERF_CONTEXT_MAX)
    				continue;
    			ret += ipchain__fprintf_graph(fp, chain, depth,
    						      new_depth_mask, i++,
    
    		ret += callchain__fprintf_graph(fp, child, new_total,
    
    						depth + 1,
    						new_depth_mask | (1 << depth));
    		node = next;
    	}
    
    
    	if (callchain_param.mode == CHAIN_GRAPH_REL &&
    		remaining && remaining != new_total) {
    
    		if (!rem_sq_bracket)
    			return ret;
    
    		new_depth_mask &= ~(1 << (depth - 1));
    
    		ret += ipchain__fprintf_graph(fp, &rem_hits, depth,
    					      new_depth_mask, 0, new_total,
    					      remaining);
    	}
    
    
    	return ret;
    }
    
    static size_t
    callchain__fprintf_flat(FILE *fp, struct callchain_node *self,
    			u64 total_samples)
    
    {
    	struct callchain_list *chain;
    	size_t ret = 0;
    
    	if (!self)
    		return 0;
    
    
    	ret += callchain__fprintf_flat(fp, self->parent, total_samples);
    
    	list_for_each_entry(chain, &self->val, list) {
    		if (chain->ip >= PERF_CONTEXT_MAX)
    			continue;
    		if (chain->sym)
    			ret += fprintf(fp, "                %s\n", chain->sym->name);
    		else
    			ret += fprintf(fp, "                %p\n",
    
    					(void *)(long)chain->ip);
    
    
    	return ret;
    }
    
    static size_t
    hist_entry_callchain__fprintf(FILE *fp, struct hist_entry *self,
    			      u64 total_samples)
    {
    	struct rb_node *rb_node;
    	struct callchain_node *chain;
    	size_t ret = 0;
    
    	rb_node = rb_first(&self->sorted_chain);
    	while (rb_node) {
    		double percent;
    
    		chain = rb_entry(rb_node, struct callchain_node, rb_node);
    		percent = chain->hit * 100.0 / total_samples;
    
    			ret += percent_color_fprintf(fp, "           %6.2f%%\n",
    						     percent);
    
    			ret += callchain__fprintf_flat(fp, chain, total_samples);
    
    			break;
    		case CHAIN_GRAPH_ABS: /* Falldown */
    		case CHAIN_GRAPH_REL:
    
    			ret += callchain__fprintf_graph(fp, chain,
    							total_samples, 1, 1);
    
    		case CHAIN_NONE:
    
    		ret += fprintf(fp, "\n");
    		rb_node = rb_next(rb_node);
    	}
    
    	return ret;
    }
    
    
    hist_entry__fprintf(FILE *fp, struct hist_entry *self, u64 total_samples)
    
    {
    	struct sort_entry *se;
    	size_t ret;
    
    
    	if (exclude_other && !self->parent)
    		return 0;
    
    
    		ret = percent_color_fprintf(fp,
    					    field_sep ? "%.2f" : "   %6.2f%%",
    					(self->count * 100.0) / total_samples);
    
    		ret = fprintf(fp, field_sep ? "%lld" : "%12lld ", self->count);
    
    	if (show_nr_samples) {
    		if (field_sep)
    			fprintf(fp, "%c%lld", *field_sep, self->count);
    		else
    			fprintf(fp, "%11lld", self->count);
    	}
    
    	list_for_each_entry(se, &hist_entry__sort_list, list) {
    
    		fprintf(fp, "%s", field_sep ?: "  ");
    		ret += se->print(fp, self, se->width ? *se->width : 0);
    
    	if (callchain)
    		hist_entry_callchain__fprintf(fp, self, total_samples);
    
    
    static void dso__calc_col_width(struct dso *self)
    {
    	if (!col_width_list_str && !field_sep &&
    	    (!dso_list || strlist__has_entry(dso_list, self->name))) {
    		unsigned int slen = strlen(self->name);
    		if (slen > dsos__col_width)
    			dsos__col_width = slen;
    	}
    
    	self->slen_calculated = 1;
    }
    
    
    static void thread__comm_adjust(struct thread *self)
    
    	char *comm = self->comm;
    
    
    	if (!col_width_list_str && !field_sep &&
    	    (!comm_list || strlist__has_entry(comm_list, comm))) {
    		unsigned int slen = strlen(comm);
    
    		if (slen > comms__col_width) {
    			comms__col_width = slen;
    			threads__col_width = slen + 6;
    		}
    	}
    
    }
    
    static int thread__set_comm_adjust(struct thread *self, const char *comm)
    {
    	int ret = thread__set_comm(self, comm);
    
    	if (ret)
    		return ret;
    
    	thread__comm_adjust(self);
    
    static struct symbol *
    
    resolve_symbol(struct thread *thread, struct map **mapp, u64 *ipp)
    
    {
    	struct map *map = mapp ? *mapp : NULL;
    
    	map = thread__find_map(thread, ip);
    	if (map != NULL) {
    
    		/*
    		 * We have to do this here as we may have a dso
    		 * with no symbol hit that has a name longer than
    		 * the ones with symbols sampled.
    		 */
    
    		if (!sort_dso.elide && !map->dso->slen_calculated)
    
    		if (mapp)
    			*mapp = map;
    got_map:
    		ip = map->map_ip(map, ip);
    	} else {
    		/*
    		 * If this is outside of all known maps,
    		 * and is a negative address, try to look it
    		 * up in the kernel dso, as it might be a
    
    		 * vsyscall or vdso (which executes in user-mode).
    		 *
    		 * XXX This is nasty, we should have a symbol list in
    		 * the "[vdso]" dso, but for now lets use the old
    		 * trick of looking in the whole kernel symbol list.
    
    		if ((long long)ip < 0) {
    			map = kernel_map;
    			if (mapp)
    				*mapp = map;
    		}
    
    	dump_printf(" ...... dso: %s\n",
    		    map ? map->dso->long_name : "<not found>");
    
    	dump_printf(" ...... map: %Lx -> %Lx\n", *ipp, ip);
    
    	return map ? map->dso->find_symbol(map->dso, ip) : NULL;
    
    static int call__match(struct symbol *sym)
    
    	if (sym->name && !regexec(&parent_regex, sym->name, 0, NULL, 0))
    
    static struct symbol **resolve_callchain(struct thread *thread, struct map *map,
    					 struct ip_callchain *chain,
    					 struct symbol **parent)
    
    {
    	u64 context = PERF_CONTEXT_MAX;
    
    	struct symbol **syms = NULL;
    
    
    	if (callchain) {
    		syms = calloc(chain->nr, sizeof(*syms));
    		if (!syms) {
    			fprintf(stderr, "Can't allocate memory for symbols\n");
    			exit(-1);
    		}
    	}
    
    	for (i = 0; i < chain->nr; i++) {
    		u64 ip = chain->ips[i];
    
    
    		if (ip >= PERF_CONTEXT_MAX) {
    			context = ip;
    			continue;
    		}
    
    		switch (context) {
    
    		case PERF_CONTEXT_HV:
    			break;
    
    		case PERF_CONTEXT_KERNEL:
    
    			sym = kernel_maps__find_symbol(ip, &map);
    
    			sym = resolve_symbol(thread, &map, &ip);
    
    			if (sort__has_parent && !*parent && call__match(sym))
    				*parent = sym;
    
    hist_entry__add(struct thread *thread, struct map *map,
    
    		struct symbol *sym, u64 ip, struct ip_callchain *chain,
    		char level, u64 count)
    
    	struct symbol **syms = NULL, *parent = NULL;
    	bool hit;
    
    	if ((sort__has_parent || callchain) && chain)
    
    		syms = resolve_callchain(thread, map, chain, &parent);
    
    	he = __hist_entry__add(thread, map, sym, parent,
    			       ip, count, level, &hit);
    	if (he == NULL)
    		return -ENOMEM;
    
    		if (!hit)
    			callchain_init(&he->callchain);
    
    		append_chain(&he->callchain, chain, syms);
    		free(syms);
    
    static size_t output__fprintf(FILE *fp, u64 total_samples)
    
    	struct hist_entry *pos;
    
    	struct rb_node *nd;
    	size_t ret = 0;
    
    	unsigned int width;
    	char *col_width = col_width_list_str;
    
    	int raw_printing_style;
    
    	raw_printing_style = !strcmp(pretty_printing_style, "raw");
    
    	fprintf(fp, "# Samples: %Ld\n", (u64)total_samples);
    
    	fprintf(fp, "#\n");
    
    	fprintf(fp, "# Overhead");
    
    	if (show_nr_samples) {
    		if (field_sep)
    			fprintf(fp, "%cSamples", *field_sep);
    		else
    			fputs("  Samples  ", fp);
    	}
    
    	list_for_each_entry(se, &hist_entry__sort_list, list) {
    
    		if (field_sep) {
    			fprintf(fp, "%c%s", *field_sep, se->header);
    
    		}
    		width = strlen(se->header);
    		if (se->width) {
    			if (col_width_list_str) {
    				if (col_width) {
    					*se->width = atoi(col_width);
    					col_width = strchr(col_width, ',');
    					if (col_width)
    						++col_width;
    				}
    			}
    			width = *se->width = max(*se->width, width);
    		}
    		fprintf(fp, "  %*s", width, se->header);
    
    	if (show_nr_samples)
    		fprintf(fp, " ..........");
    
    	list_for_each_entry(se, &hist_entry__sort_list, list) {
    
    		fprintf(fp, "  ");
    
    		if (se->width)
    			width = *se->width;
    		else
    			width = strlen(se->header);
    		for (i = 0; i < width; i++)
    
    	for (nd = rb_first(&output_hists); nd; nd = rb_next(nd)) {
    		pos = rb_entry(nd, struct hist_entry, rb_node);
    		ret += hist_entry__fprintf(fp, pos, total_samples);
    
    	if (sort_order == default_sort_order &&
    			parent_pattern == default_parent_pattern) {
    
    		fprintf(fp, "# (For a higher level overview, try: perf report --sort comm,dso)\n");
    
    	fprintf(fp, "\n");
    
    		perf_read_values_display(fp, &show_threads_values,
    					 raw_printing_style);
    
    static int validate_chain(struct ip_callchain *chain, event_t *event)
    
    {
    	unsigned int chain_size;
    
    	chain_size = event->header.size;
    	chain_size -= (unsigned long)&event->ip.__more_data - (unsigned long)event;
    
    
    	if (chain->nr*sizeof(u64) > chain_size)
    
    process_sample_event(event_t *event, unsigned long offset, unsigned long head)
    
    	u64 ip = event->ip.ip;
    	u64 period = 1;
    
    	struct map *map = NULL;
    
    	void *more_data = event->ip.__more_data;
    
    	struct ip_callchain *chain = NULL;
    
    	thread = threads__findnew(event->ip.pid, &threads, &last_match);
    
    
    	if (sample_type & PERF_SAMPLE_PERIOD) {
    
    		period = *(u64 *)more_data;
    		more_data += sizeof(u64);
    
    	dump_printf("%p [%p]: PERF_RECORD_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n",
    
    		(void *)(offset + head),
    		(void *)(long)(event->header.size),
    		event->header.misc,
    
    		event->ip.pid, event->ip.tid,
    
    	if (sample_type & PERF_SAMPLE_CALLCHAIN) {
    
    		dump_printf("... chain: nr:%Lu\n", chain->nr);
    
    		if (validate_chain(chain, event) < 0) {
    			eprintf("call-chain problem with event, skipping it.\n");
    			return 0;
    		}
    
    		if (dump_trace) {
    
    			for (i = 0; i < chain->nr; i++)
    
    				dump_printf("..... %2d: %016Lx\n", i, chain->ips[i]);
    
    	dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
    
    
    	if (thread == NULL) {
    
    		eprintf("problem processing %d event, skipping it.\n",
    
    			event->header.type);
    		return -1;
    	}
    
    	if (comm_list && !strlist__has_entry(comm_list, thread->comm))
    		return 0;
    
    
    	cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
    
    	if (cpumode == PERF_RECORD_MISC_KERNEL) {
    
    		sym = kernel_maps__find_symbol(ip, &map);
    		dump_printf(" ...... dso: %s\n",
    			    map ? map->dso->long_name : "<not found>");
    
    	} else if (cpumode == PERF_RECORD_MISC_USER) {
    
    		sym = resolve_symbol(thread, &map, &ip);
    
    	} else {
    		level = 'H';
    
    		dump_printf(" ...... dso: [hypervisor]\n");
    
    	if (dso_list &&
    	    (!map || !map->dso ||
    	     !(strlist__has_entry(dso_list, map->dso->short_name) ||
    	       (map->dso->short_name != map->dso->long_name &&
    		strlist__has_entry(dso_list, map->dso->long_name)))))
    		return 0;
    
    	if (sym_list && sym && !strlist__has_entry(sym_list, sym->name))
    		return 0;
    
    	if (hist_entry__add(thread, map, sym, ip,
    			    chain, level, period)) {
    		eprintf("problem incrementing symbol count, skipping event\n");
    		return -1;
    
    static int
    process_mmap_event(event_t *event, unsigned long offset, unsigned long head)
    {
    
    	struct map *map = map__new(&event->mmap, cwd, cwdlen);
    
    	thread = threads__findnew(event->mmap.pid, &threads, &last_match);
    
    
    	dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n",
    
    		(void *)(offset + head),
    		(void *)(long)(event->header.size),
    
    		(void *)(long)event->mmap.start,
    		(void *)(long)event->mmap.len,
    		(void *)(long)event->mmap.pgoff,
    		event->mmap.filename);
    
    	if (thread == NULL || map == NULL) {
    
    		dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n");
    
    	}
    
    	thread__insert_map(thread, map);
    	total_mmap++;
    
    	return 0;
    }
    
    static int
    process_comm_event(event_t *event, unsigned long offset, unsigned long head)
    {
    
    	struct thread *thread;
    
    	thread = threads__findnew(event->comm.pid, &threads, &last_match);
    
    	dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n",
    
    		(void *)(offset + head),
    		(void *)(long)(event->header.size),
    		event->comm.comm, event->comm.pid);
    
    	if (thread == NULL ||
    
    	    thread__set_comm_adjust(thread, event->comm.comm)) {
    
    		dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n");
    
    process_task_event(event_t *event, unsigned long offset, unsigned long head)
    
    	struct thread *thread;
    	struct thread *parent;
    
    	thread = threads__findnew(event->fork.pid, &threads, &last_match);
    	parent = threads__findnew(event->fork.ppid, &threads, &last_match);
    
    	dump_printf("%p [%p]: PERF_RECORD_%s: (%d:%d):(%d:%d)\n",
    
    		(void *)(offset + head),
    		(void *)(long)(event->header.size),
    
    		event->header.type == PERF_RECORD_FORK ? "FORK" : "EXIT",
    
    		event->fork.pid, event->fork.tid,
    		event->fork.ppid, event->fork.ptid);
    
    	/*
    	 * A thread clone will have the same PID for both
    	 * parent and child.
    	 */
    	if (thread == parent)
    		return 0;
    
    
    	if (event->header.type == PERF_RECORD_EXIT)
    
    
    	if (!thread || !parent || thread__fork(thread, parent)) {
    
    		dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n");
    
    static int
    process_lost_event(event_t *event, unsigned long offset, unsigned long head)
    {
    
    	dump_printf("%p [%p]: PERF_RECORD_LOST: id:%Ld: lost:%Ld\n",
    
    		(void *)(offset + head),
    		(void *)(long)(event->header.size),
    		event->lost.id,
    		event->lost.lost);
    
    	total_lost += event->lost.lost;
    
    	return 0;
    }
    
    
    static int
    process_read_event(event_t *event, unsigned long offset, unsigned long head)
    {
    
    
    	attr = perf_header__find_attr(event->read.id, header);
    
    		const char *name = attr ? __event_name(attr->type, attr->config)
    
    				   : "unknown";
    		perf_read_values_add_value(&show_threads_values,
    					   event->read.pid, event->read.tid,
    					   event->read.id,
    					   name,
    					   event->read.value);
    	}
    
    
    	dump_printf("%p [%p]: PERF_RECORD_READ: %d %d %s %Lu\n",
    
    			(void *)(offset + head),
    			(void *)(long)(event->header.size),
    			event->read.pid,
    			event->read.tid,
    
    			attr ? __event_name(attr->type, attr->config)
    			     : "FAIL",
    
    static int
    process_event(event_t *event, unsigned long offset, unsigned long head)
    {
    
    	trace_event(event);
    
    
    	switch (event->header.type) {
    
    		return process_sample_event(event, offset, head);
    
    
    		return process_mmap_event(event, offset, head);
    
    
    		return process_comm_event(event, offset, head);
    
    
    	case PERF_RECORD_FORK:
    	case PERF_RECORD_EXIT:
    
    		return process_task_event(event, offset, head);
    
    		return process_lost_event(event, offset, head);
    
    
    		return process_read_event(event, offset, head);
    
    
    	/*
    	 * We dont process them right now but they are fine:
    	 */
    
    	case PERF_RECORD_THROTTLE:
    	case PERF_RECORD_UNTHROTTLE:
    
    	default:
    		return -1;
    	}
    
    	return 0;
    }
    
    static int __cmd_report(void)
    {
    
    	int ret, rc = EXIT_FAILURE;
    
    	unsigned long offset = 0;
    
    	unsigned long head, shift;
    
    	struct stat input_stat;
    
    	event_t *event;
    	uint32_t size;
    
    	idle = register_idle_thread(&threads, &last_match);
    	thread__comm_adjust(idle);
    
    	if (show_threads)
    		perf_read_values_init(&show_threads_values);
    
    
    	input = open(input_name, O_RDONLY);
    	if (input < 0) {
    
    		fprintf(stderr, " failed to open file: %s", input_name);
    		if (!strcmp(input_name, "perf.data"))
    			fprintf(stderr, "  (try 'perf record' first)");
    		fprintf(stderr, "\n");
    
    	ret = fstat(input, &input_stat);
    
    	if (ret < 0) {
    		perror("failed to stat file");
    		exit(-1);
    	}
    
    
    	if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) {
    		fprintf(stderr, "file: %s not owned by current user or root\n", input_name);
    
    	if (!input_stat.st_size) {
    
    		fprintf(stderr, "zero-sized file, nothing to do!\n");
    		exit(0);
    	}
    
    
    	header = perf_header__read(input);
    	head = header->data_offset;
    
    	sample_type = perf_header__sample_type(header);
    
    	if (!(sample_type & PERF_SAMPLE_CALLCHAIN)) {
    		if (sort__has_parent) {
    			fprintf(stderr, "selected --sort parent, but no"
    					" callchain data. Did you call"
    					" perf record without -g?\n");
    			exit(-1);
    		}
    		if (callchain) {
    
    			fprintf(stderr, "selected -g but no callchain data."
    
    					" Did you call perf record without"
    					" -g?\n");
    			exit(-1);
    		}
    
    	} else if (callchain_param.mode != CHAIN_NONE && !callchain) {
    			callchain = 1;
    			if (register_callchain_param(&callchain_param) < 0) {
    				fprintf(stderr, "Can't register callchain"
    						" params\n");
    				exit(-1);
    			}
    
    	if (load_kernel() < 0) {
    		perror("failed to load kernel symbols");
    		return EXIT_FAILURE;
    	}
    
    	if (!full_paths) {
    		if (getcwd(__cwd, sizeof(__cwd)) == NULL) {
    			perror("failed to get the current directory");
    			return EXIT_FAILURE;
    		}
    		cwdlen = strlen(cwd);
    	} else {
    		cwd = NULL;
    		cwdlen = 0;
    	}
    
    
    	shift = page_size * (head / page_size);
    	offset += shift;
    	head -= shift;
    
    
    remap:
    	buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ,
    			   MAP_SHARED, input, offset);
    	if (buf == MAP_FAILED) {
    		perror("failed to mmap file");
    		exit(-1);
    	}
    
    more:
    	event = (event_t *)(buf + head);
    
    	size = event->header.size;
    	if (!size)
    		size = 8;
    
    	if (head + event->header.size >= page_size * mmap_window) {
    
    		int munmap_ret;
    
    		shift = page_size * (head / page_size);
    
    
    		munmap_ret = munmap(buf, page_size * mmap_window);
    		assert(munmap_ret == 0);
    
    
    		offset += shift;
    		head -= shift;
    		goto remap;
    	}
    
    	size = event->header.size;
    
    
    	dump_printf("\n%p [%p]: event: %d\n",
    
    			(void *)(offset + head),
    			(void *)(long)event->header.size,
    			event->header.type);
    
    
    	if (!size || process_event(event, offset, head) < 0) {
    
    
    		dump_printf("%p [%p]: skipping unknown header type: %d\n",
    
    			(void *)(offset + head),
    			(void *)(long)(event->header.size),
    			event->header.type);
    
    		total_unknown++;
    
    
    		/*
    		 * assume we lost track of the stream, check alignment, and
    		 * increment a single u64 in the hope to catch on again 'soon'.
    		 */
    
    		if (unlikely(head & 7))
    			head &= ~7ULL;