From 7440172974e85b1828bdd84ac6b23b5bcad9c5eb Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 6 Nov 2018 18:44:52 -0800 Subject: tracing: Replace synchronize_sched() and call_rcu_sched() Now that synchronize_rcu() waits for preempt-disable regions of code as well as RCU read-side critical sections, synchronize_sched() can be replaced by synchronize_rcu(). Similarly, call_rcu_sched() can be replaced by call_rcu(). This commit therefore makes these changes. Signed-off-by: Paul E. McKenney Cc: Ingo Molnar Cc: Acked-by: Steven Rostedt (VMware) --- kernel/trace/ring_buffer.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'kernel/trace/ring_buffer.c') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 65bd4616220d..4f3247a53259 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -1834,7 +1834,7 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size, * There could have been a race between checking * record_disable and incrementing it. */ - synchronize_sched(); + synchronize_rcu(); for_each_buffer_cpu(buffer, cpu) { cpu_buffer = buffer->buffers[cpu]; rb_check_pages(cpu_buffer); @@ -3151,7 +3151,7 @@ static bool rb_per_cpu_empty(struct ring_buffer_per_cpu *cpu_buffer) * This prevents all writes to the buffer. Any attempt to write * to the buffer after this will fail and return NULL. * - * The caller should call synchronize_sched() after this. + * The caller should call synchronize_rcu() after this. */ void ring_buffer_record_disable(struct ring_buffer *buffer) { @@ -3253,7 +3253,7 @@ bool ring_buffer_record_is_set_on(struct ring_buffer *buffer) * This prevents all writes to the buffer. Any attempt to write * to the buffer after this will fail and return NULL. * - * The caller should call synchronize_sched() after this. + * The caller should call synchronize_rcu() after this. */ void ring_buffer_record_disable_cpu(struct ring_buffer *buffer, int cpu) { @@ -4191,7 +4191,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_read_prepare); void ring_buffer_read_prepare_sync(void) { - synchronize_sched(); + synchronize_rcu(); } EXPORT_SYMBOL_GPL(ring_buffer_read_prepare_sync); @@ -4363,7 +4363,7 @@ void ring_buffer_reset_cpu(struct ring_buffer *buffer, int cpu) atomic_inc(&cpu_buffer->record_disabled); /* Make sure all commits have finished */ - synchronize_sched(); + synchronize_rcu(); raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); @@ -4496,7 +4496,7 @@ int ring_buffer_swap_cpu(struct ring_buffer *buffer_a, goto out; /* - * We can't do a synchronize_sched here because this + * We can't do a synchronize_rcu here because this * function can be called in atomic context. * Normally this will be called from the same CPU as cpu. * If not it's up to the caller to protect this. -- cgit v1.2.3 From 2c2b0a78b373908926e4683ea5571332f63c0eb5 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Thu, 29 Nov 2018 20:32:26 -0500 Subject: ring-buffer: Add percentage of ring buffer full to wake up reader Instead of just waiting for a page to be full before waking up a pending reader, allow the reader to pass in a "percentage" of pages that have content before waking up a reader. This should help keep the process of reading the events not cause wake ups that constantly cause reading of the buffer. Signed-off-by: Steven Rostedt (VMware) --- kernel/trace/ring_buffer.c | 71 ++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 66 insertions(+), 5 deletions(-) (limited to 'kernel/trace/ring_buffer.c') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 65bd4616220d..9edb628603ab 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -487,6 +487,9 @@ struct ring_buffer_per_cpu { local_t dropped_events; local_t committing; local_t commits; + local_t pages_touched; + local_t pages_read; + size_t shortest_full; unsigned long read; unsigned long read_bytes; u64 write_stamp; @@ -529,6 +532,41 @@ struct ring_buffer_iter { u64 read_stamp; }; +/** + * ring_buffer_nr_pages - get the number of buffer pages in the ring buffer + * @buffer: The ring_buffer to get the number of pages from + * @cpu: The cpu of the ring_buffer to get the number of pages from + * + * Returns the number of pages used by a per_cpu buffer of the ring buffer. + */ +size_t ring_buffer_nr_pages(struct ring_buffer *buffer, int cpu) +{ + return buffer->buffers[cpu]->nr_pages; +} + +/** + * ring_buffer_nr_pages_dirty - get the number of used pages in the ring buffer + * @buffer: The ring_buffer to get the number of pages from + * @cpu: The cpu of the ring_buffer to get the number of pages from + * + * Returns the number of pages that have content in the ring buffer. + */ +size_t ring_buffer_nr_dirty_pages(struct ring_buffer *buffer, int cpu) +{ + size_t read; + size_t cnt; + + read = local_read(&buffer->buffers[cpu]->pages_read); + cnt = local_read(&buffer->buffers[cpu]->pages_touched); + /* The reader can read an empty page, but not more than that */ + if (cnt < read) { + WARN_ON_ONCE(read > cnt + 1); + return 0; + } + + return cnt - read; +} + /* * rb_wake_up_waiters - wake up tasks waiting for ring buffer input * @@ -556,7 +594,7 @@ static void rb_wake_up_waiters(struct irq_work *work) * as data is added to any of the @buffer's cpu buffers. Otherwise * it will wait for data to be added to a specific cpu buffer. */ -int ring_buffer_wait(struct ring_buffer *buffer, int cpu, bool full) +int ring_buffer_wait(struct ring_buffer *buffer, int cpu, int full) { struct ring_buffer_per_cpu *uninitialized_var(cpu_buffer); DEFINE_WAIT(wait); @@ -571,7 +609,7 @@ int ring_buffer_wait(struct ring_buffer *buffer, int cpu, bool full) if (cpu == RING_BUFFER_ALL_CPUS) { work = &buffer->irq_work; /* Full only makes sense on per cpu reads */ - full = false; + full = 0; } else { if (!cpumask_test_cpu(cpu, buffer->cpumask)) return -ENODEV; @@ -623,15 +661,22 @@ int ring_buffer_wait(struct ring_buffer *buffer, int cpu, bool full) !ring_buffer_empty_cpu(buffer, cpu)) { unsigned long flags; bool pagebusy; + size_t nr_pages; + size_t dirty; if (!full) break; raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); pagebusy = cpu_buffer->reader_page == cpu_buffer->commit_page; + nr_pages = cpu_buffer->nr_pages; + dirty = ring_buffer_nr_dirty_pages(buffer, cpu); + if (!cpu_buffer->shortest_full || + cpu_buffer->shortest_full < full) + cpu_buffer->shortest_full = full; raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); - - if (!pagebusy) + if (!pagebusy && + (!nr_pages || (dirty * 100) > full * nr_pages)) break; } @@ -1054,6 +1099,7 @@ static void rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer, old_write = local_add_return(RB_WRITE_INTCNT, &next_page->write); old_entries = local_add_return(RB_WRITE_INTCNT, &next_page->entries); + local_inc(&cpu_buffer->pages_touched); /* * Just make sure we have seen our old_write and synchronize * with any interrupts that come in. @@ -2603,6 +2649,16 @@ rb_wakeups(struct ring_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer) pagebusy = cpu_buffer->reader_page == cpu_buffer->commit_page; if (!pagebusy && cpu_buffer->irq_work.full_waiters_pending) { + size_t nr_pages; + size_t dirty; + size_t full; + + full = cpu_buffer->shortest_full; + nr_pages = cpu_buffer->nr_pages; + dirty = ring_buffer_nr_dirty_pages(buffer, cpu_buffer->cpu); + if (full && nr_pages && (dirty * 100) <= full * nr_pages) + return; + cpu_buffer->irq_work.wakeup_full = true; cpu_buffer->irq_work.full_waiters_pending = false; /* irq_work_queue() supplies it's own memory barriers */ @@ -3732,13 +3788,15 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer) goto spin; /* - * Yeah! We succeeded in replacing the page. + * Yay! We succeeded in replacing the page. * * Now make the new head point back to the reader page. */ rb_list_head(reader->list.next)->prev = &cpu_buffer->reader_page->list; rb_inc_page(cpu_buffer, &cpu_buffer->head_page); + local_inc(&cpu_buffer->pages_read); + /* Finally update the reader page to the new head */ cpu_buffer->reader_page = reader; cpu_buffer->reader_page->read = 0; @@ -4334,6 +4392,9 @@ rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) local_set(&cpu_buffer->entries, 0); local_set(&cpu_buffer->committing, 0); local_set(&cpu_buffer->commits, 0); + local_set(&cpu_buffer->pages_touched, 0); + local_set(&cpu_buffer->pages_read, 0); + cpu_buffer->shortest_full = 0; cpu_buffer->read = 0; cpu_buffer->read_bytes = 0; -- cgit v1.2.3 From 03329f9939781041424edd29487b9603a704fcd9 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Thu, 29 Nov 2018 21:38:42 -0500 Subject: tracing: Add tracefs file buffer_percentage Add a "buffer_percentage" file, that allows users to specify how much of the buffer (percentage of pages) need to be filled before waking up a task blocked on a per cpu trace_pipe_raw file. Signed-off-by: Steven Rostedt (VMware) --- kernel/trace/ring_buffer.c | 39 +++++++++++++++++++++++---------------- 1 file changed, 23 insertions(+), 16 deletions(-) (limited to 'kernel/trace/ring_buffer.c') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 9edb628603ab..5434c16f2192 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -489,6 +489,7 @@ struct ring_buffer_per_cpu { local_t commits; local_t pages_touched; local_t pages_read; + long last_pages_touch; size_t shortest_full; unsigned long read; unsigned long read_bytes; @@ -2632,7 +2633,9 @@ static void rb_commit(struct ring_buffer_per_cpu *cpu_buffer, static __always_inline void rb_wakeups(struct ring_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer) { - bool pagebusy; + size_t nr_pages; + size_t dirty; + size_t full; if (buffer->irq_work.waiters_pending) { buffer->irq_work.waiters_pending = false; @@ -2646,24 +2649,27 @@ rb_wakeups(struct ring_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer) irq_work_queue(&cpu_buffer->irq_work.work); } - pagebusy = cpu_buffer->reader_page == cpu_buffer->commit_page; + if (cpu_buffer->last_pages_touch == local_read(&cpu_buffer->pages_touched)) + return; - if (!pagebusy && cpu_buffer->irq_work.full_waiters_pending) { - size_t nr_pages; - size_t dirty; - size_t full; + if (cpu_buffer->reader_page == cpu_buffer->commit_page) + return; - full = cpu_buffer->shortest_full; - nr_pages = cpu_buffer->nr_pages; - dirty = ring_buffer_nr_dirty_pages(buffer, cpu_buffer->cpu); - if (full && nr_pages && (dirty * 100) <= full * nr_pages) - return; + if (!cpu_buffer->irq_work.full_waiters_pending) + return; - cpu_buffer->irq_work.wakeup_full = true; - cpu_buffer->irq_work.full_waiters_pending = false; - /* irq_work_queue() supplies it's own memory barriers */ - irq_work_queue(&cpu_buffer->irq_work.work); - } + cpu_buffer->last_pages_touch = local_read(&cpu_buffer->pages_touched); + + full = cpu_buffer->shortest_full; + nr_pages = cpu_buffer->nr_pages; + dirty = ring_buffer_nr_dirty_pages(buffer, cpu_buffer->cpu); + if (full && nr_pages && (dirty * 100) <= full * nr_pages) + return; + + cpu_buffer->irq_work.wakeup_full = true; + cpu_buffer->irq_work.full_waiters_pending = false; + /* irq_work_queue() supplies it's own memory barriers */ + irq_work_queue(&cpu_buffer->irq_work.work); } /* @@ -4394,6 +4400,7 @@ rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) local_set(&cpu_buffer->commits, 0); local_set(&cpu_buffer->pages_touched, 0); local_set(&cpu_buffer->pages_read, 0); + cpu_buffer->last_pages_touch = 0; cpu_buffer->shortest_full = 0; cpu_buffer->read = 0; cpu_buffer->read_bytes = 0; -- cgit v1.2.3 From d325c402964e7c63db94e9138c530832269a1297 Mon Sep 17 00:00:00 2001 From: Miroslav Benes Date: Fri, 28 Dec 2018 14:38:47 +0100 Subject: ring-buffer: Remove unused function ring_buffer_page_len() Commit 6b7e633fe9c2 ("tracing: Remove extra zeroing out of the ring buffer page") removed the only caller of ring_buffer_page_len(). The function is now unused and may be removed. Link: http://lkml.kernel.org/r/20181228133847.106177-1-mbenes@suse.cz Signed-off-by: Miroslav Benes Signed-off-by: Steven Rostedt (VMware) --- kernel/trace/ring_buffer.c | 14 -------------- 1 file changed, 14 deletions(-) (limited to 'kernel/trace/ring_buffer.c') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 06e864a334bb..9a91479bbbfe 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -353,20 +353,6 @@ static void rb_init_page(struct buffer_data_page *bpage) local_set(&bpage->commit, 0); } -/** - * ring_buffer_page_len - the size of data on the page. - * @page: The page to read - * - * Returns the amount of data on the page, including buffer page header. - */ -size_t ring_buffer_page_len(void *page) -{ - struct buffer_data_page *bpage = page; - - return (local_read(&bpage->commit) & ~RB_MISSED_FLAGS) - + BUF_PAGE_HDR_SIZE; -} - /* * Also stolen from mm/slob.c. Thanks to Mathieu Desnoyers for pointing * this issue out. -- cgit v1.2.3 From 31b265b3baaf55f209229888b7ffea523ddab366 Mon Sep 17 00:00:00 2001 From: Douglas Anderson Date: Fri, 8 Mar 2019 11:32:04 -0800 Subject: tracing: kdb: Fix ftdump to not sleep As reported back in 2016-11 [1], the "ftdump" kdb command triggers a BUG for "sleeping function called from invalid context". kdb's "ftdump" command wants to call ring_buffer_read_prepare() in atomic context. A very simple solution for this is to add allocation flags to ring_buffer_read_prepare() so kdb can call it without triggering the allocation error. This patch does that. Note that in the original email thread about this, it was suggested that perhaps the solution for kdb was to either preallocate the buffer ahead of time or create our own iterator. I'm hoping that this alternative of adding allocation flags to ring_buffer_read_prepare() can be considered since it means I don't need to duplicate more of the core trace code into "trace_kdb.c" (for either creating my own iterator or re-preparing a ring allocator whose memory was already allocated). NOTE: another option for kdb is to actually figure out how to make it reuse the existing ftrace_dump() function and totally eliminate the duplication. This sounds very appealing and actually works (the "sr z" command can be seen to properly dump the ftrace buffer). The downside here is that ftrace_dump() fully consumes the trace buffer. Unless that is changed I'd rather not use it because it means "ftdump | grep xyz" won't be very useful to search the ftrace buffer since it will throw away the whole trace on the first grep. A future patch to dump only the last few lines of the buffer will also be hard to implement. [1] https://lkml.kernel.org/r/20161117191605.GA21459@google.com Link: http://lkml.kernel.org/r/20190308193205.213659-1-dianders@chromium.org Reported-by: Brian Norris Signed-off-by: Douglas Anderson Signed-off-by: Steven Rostedt (VMware) --- kernel/trace/ring_buffer.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'kernel/trace/ring_buffer.c') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 9a91479bbbfe..41b6f96e5366 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -4191,6 +4191,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_consume); * ring_buffer_read_prepare - Prepare for a non consuming read of the buffer * @buffer: The ring buffer to read from * @cpu: The cpu buffer to iterate over + * @flags: gfp flags to use for memory allocation * * This performs the initial preparations necessary to iterate * through the buffer. Memory is allocated, buffer recording @@ -4208,7 +4209,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_consume); * This overall must be paired with ring_buffer_read_finish. */ struct ring_buffer_iter * -ring_buffer_read_prepare(struct ring_buffer *buffer, int cpu) +ring_buffer_read_prepare(struct ring_buffer *buffer, int cpu, gfp_t flags) { struct ring_buffer_per_cpu *cpu_buffer; struct ring_buffer_iter *iter; @@ -4216,7 +4217,7 @@ ring_buffer_read_prepare(struct ring_buffer *buffer, int cpu) if (!cpumask_test_cpu(cpu, buffer->cpumask)) return NULL; - iter = kmalloc(sizeof(*iter), GFP_KERNEL); + iter = kmalloc(sizeof(*iter), flags); if (!iter) return NULL; -- cgit v1.2.3 From d6097c9e4454adf1f8f2c9547c2fa6060d55d952 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 23 Apr 2019 22:03:18 +0200 Subject: trace: Fix preempt_enable_no_resched() abuse Unless the very next line is schedule(), or implies it, one must not use preempt_enable_no_resched(). It can cause a preemption to go missing and thereby cause arbitrary delays, breaking the PREEMPT=y invariant. Link: http://lkml.kernel.org/r/20190423200318.GY14281@hirez.programming.kicks-ass.net Cc: Waiman Long Cc: Linus Torvalds Cc: Ingo Molnar Cc: Will Deacon Cc: Thomas Gleixner Cc: the arch/x86 maintainers Cc: Davidlohr Bueso Cc: Tim Chen Cc: huang ying Cc: Roman Gushchin Cc: Alexei Starovoitov Cc: Daniel Borkmann Cc: stable@vger.kernel.org Fixes: 2c2d7329d8af ("tracing/ftrace: use preempt_enable_no_resched_notrace in ring_buffer_time_stamp()") Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Steven Rostedt (VMware) --- kernel/trace/ring_buffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel/trace/ring_buffer.c') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 41b6f96e5366..4ee8d8aa3d0f 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -762,7 +762,7 @@ u64 ring_buffer_time_stamp(struct ring_buffer *buffer, int cpu) preempt_disable_notrace(); time = rb_time_stamp(buffer); - preempt_enable_no_resched_notrace(); + preempt_enable_notrace(); return time; } -- cgit v1.2.3