Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757301Ab2F1Awq (ORCPT ); Wed, 27 Jun 2012 20:52:46 -0400 Received: from hrndva-omtalb.mail.rr.com ([71.74.56.122]:10791 "EHLO hrndva-omtalb.mail.rr.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752560Ab2F1Awp (ORCPT ); Wed, 27 Jun 2012 20:52:45 -0400 X-Authority-Analysis: v=2.0 cv=ZuBv2qHG c=1 sm=0 a=ZycB6UtQUfgMyuk2+PxD7w==:17 a=XQbtiDEiEegA:10 a=pf8V27uzMmYA:10 a=5SG0PmZfjMsA:10 a=Q9fys5e9bTEA:10 a=meVymXHHAAAA:8 a=ayC55rCoAAAA:8 a=20KFwNOVAAAA:8 a=SKBzyUd3OZRTJhgoc0oA:9 a=PUjeQqilurYA:10 a=jEp0ucaQiEUA:10 a=jeBq3FmKZ4MA:10 a=dM68ZfjBud_p8Hj6:21 a=jIy4PLGBer9pg5vY:21 a=ZycB6UtQUfgMyuk2+PxD7w==:117 X-Cloudmark-Score: 0 X-Originating-IP: 74.67.80.29 Message-ID: <1340844763.16702.54.camel@gandalf.stny.rr.com> Subject: Re: big allocation failure during tracing init. From: Steven Rostedt To: Dave Jones Cc: Linux Kernel Date: Wed, 27 Jun 2012 20:52:43 -0400 In-Reply-To: <20120622193650.GA20098@redhat.com> References: <20120622193650.GA20098@redhat.com> Content-Type: text/plain; charset="ISO-8859-15" X-Mailer: Evolution 3.2.2-1+b1 Content-Transfer-Encoding: 7bit Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5951 Lines: 181 On Fri, 2012-06-22 at 15:36 -0400, Dave Jones wrote: > I guess the problem is on kernels built with CONFIG_MAXCPUS, struct trace_iterator > can be pretty large due to this embedded in it.. > > struct ring_buffer_iter *buffer_iter[NR_CPUS]; > > Any thoughts on how to change that to a smaller dynamic data structure, > without adding too much overhead to tracing ? (I'm assuming a linked-list walk > for eg would be excessive?) > Does this patch fix it for you? -- Steve commit 0339abcabbef5f1fc57a8af046861a04e0a0ef9f Author: Steven Rostedt Date: Wed Jun 27 20:46:14 2012 -0400 tracing: Remove NR_CPUS array from trace_iterator Replace the NR_CPUS array of buffer_iter from the trace_iterator with an allocated array. This will just create an array of possible CPUS instead of the max number specified. Added a helper function called 'trace_buffer_iter()' that returns the buffer_iter item or NULL if it is not defined or the array was not allocated. Some routines do not require the array (tracing_open_pipe() for one). Reported-by: Dave Jones Signed-off-by: Steven Rostedt diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h index 1aff183..af961d6 100644 --- a/include/linux/ftrace_event.h +++ b/include/linux/ftrace_event.h @@ -65,7 +65,7 @@ struct trace_iterator { void *private; int cpu_file; struct mutex mutex; - struct ring_buffer_iter *buffer_iter[NR_CPUS]; + struct ring_buffer_iter **buffer_iter; unsigned long iter_flags; /* trace_seq for __print_flags() and __print_symbolic() etc. */ diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index a8d5654..85073f9 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -1710,9 +1710,11 @@ EXPORT_SYMBOL_GPL(trace_vprintk); static void trace_iterator_increment(struct trace_iterator *iter) { + struct ring_buffer_iter *buf_iter = trace_buffer_iter(iter, iter->cpu); + iter->idx++; - if (iter->buffer_iter[iter->cpu]) - ring_buffer_read(iter->buffer_iter[iter->cpu], NULL); + if (buf_iter) + ring_buffer_read(buf_iter, NULL); } static struct trace_entry * @@ -1720,7 +1722,7 @@ peek_next_entry(struct trace_iterator *iter, int cpu, u64 *ts, unsigned long *lost_events) { struct ring_buffer_event *event; - struct ring_buffer_iter *buf_iter = iter->buffer_iter[cpu]; + struct ring_buffer_iter *buf_iter = trace_buffer_iter(iter, cpu); if (buf_iter) event = ring_buffer_iter_peek(buf_iter, ts); @@ -1858,10 +1860,10 @@ void tracing_iter_reset(struct trace_iterator *iter, int cpu) tr->data[cpu]->skipped_entries = 0; - if (!iter->buffer_iter[cpu]) + buf_iter = trace_buffer_iter(iter, cpu); + if (!buf_iter) return; - buf_iter = iter->buffer_iter[cpu]; ring_buffer_iter_reset(buf_iter); /* @@ -2207,13 +2209,15 @@ static enum print_line_t print_bin_fmt(struct trace_iterator *iter) int trace_empty(struct trace_iterator *iter) { + struct ring_buffer_iter *buf_iter; int cpu; /* If we are looking at one CPU buffer, only check that one */ if (iter->cpu_file != TRACE_PIPE_ALL_CPU) { cpu = iter->cpu_file; - if (iter->buffer_iter[cpu]) { - if (!ring_buffer_iter_empty(iter->buffer_iter[cpu])) + buf_iter = trace_buffer_iter(iter, cpu); + if (buf_iter) { + if (!ring_buffer_iter_empty(buf_iter)) return 0; } else { if (!ring_buffer_empty_cpu(iter->tr->buffer, cpu)) @@ -2223,8 +2227,9 @@ int trace_empty(struct trace_iterator *iter) } for_each_tracing_cpu(cpu) { - if (iter->buffer_iter[cpu]) { - if (!ring_buffer_iter_empty(iter->buffer_iter[cpu])) + buf_iter = trace_buffer_iter(iter, cpu); + if (buf_iter) { + if (!ring_buffer_iter_empty(buf_iter)) return 0; } else { if (!ring_buffer_empty_cpu(iter->tr->buffer, cpu)) @@ -2383,6 +2388,8 @@ __tracing_open(struct inode *inode, struct file *file) if (!iter) return ERR_PTR(-ENOMEM); + iter->buffer_iter = kzalloc(sizeof(*iter->buffer_iter) * num_possible_cpus(), + GFP_KERNEL); /* * We make a copy of the current tracer to avoid concurrent * changes on it while we are reading. @@ -2443,6 +2450,7 @@ __tracing_open(struct inode *inode, struct file *file) fail: mutex_unlock(&trace_types_lock); kfree(iter->trace); + kfree(iter->buffer_iter); seq_release_private(inode, file); return ERR_PTR(-ENOMEM); } @@ -2483,6 +2491,7 @@ static int tracing_release(struct inode *inode, struct file *file) mutex_destroy(&iter->mutex); free_cpumask_var(iter->started); kfree(iter->trace); + kfree(iter->buffer_iter); seq_release_private(inode, file); return 0; } diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 5aec220..55e1f7f 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -317,6 +317,14 @@ struct tracer { #define TRACE_PIPE_ALL_CPU -1 +static inline struct ring_buffer_iter * +trace_buffer_iter(struct trace_iterator *iter, int cpu) +{ + if (iter->buffer_iter && iter->buffer_iter[cpu]) + return iter->buffer_iter[cpu]; + return NULL; +} + int tracer_init(struct tracer *t, struct trace_array *tr); int tracing_is_enabled(void); void trace_wake_up(void); diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c index a7d2a4c..ce27c8b 100644 --- a/kernel/trace/trace_functions_graph.c +++ b/kernel/trace/trace_functions_graph.c @@ -538,7 +538,7 @@ get_return_for_leaf(struct trace_iterator *iter, next = &data->ret; } else { - ring_iter = iter->buffer_iter[iter->cpu]; + ring_iter = trace_buffer_iter(iter, iter->cpu); /* First peek to compare current entry and the next one */ if (ring_iter) -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/