[lkml]   [2020]   [Apr]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Patch in this message
Subject[PATCH v2][for-next] tracing: Do not allocate buffer in trace_find_next_entry() in atomic

Changes since v1:

I don't usually rebase my for-next branch, but its only a single patch,
and Masami pointed out a silly mistake in the first one. That
ftrace_dump() never initialized the temp_size value, and which caused the
WARN_ON_ONCE() added to also trigger.

-- Steve


Head SHA1: 8e99cf91b99bb30e16727f10ad6828741c0e992f

Steven Rostedt (VMware) (1):
tracing: Do not allocate buffer in trace_find_next_entry() in atomic

kernel/trace/trace.c | 21 ++++++++++++++++++++-
1 file changed, 20 insertions(+), 1 deletion(-)
commit 8e99cf91b99bb30e16727f10ad6828741c0e992f
Author: Steven Rostedt (VMware) <>
Date: Wed Apr 1 22:44:46 2020 -0400

tracing: Do not allocate buffer in trace_find_next_entry() in atomic

When dumping out the trace data in latency format, a check is made to peek
at the next event to compare its timestamp to the current one, and if the
delta is of a greater size, it will add a marker showing so. But to do this,
it needs to save the current event otherwise peeking at the next event will
remove the current event. To save the event, a temp buffer is used, and if
the event is bigger than the temp buffer, the temp buffer is freed and a
bigger buffer is allocated.

This allocation is a problem when called in atomic context. The only way
this gets called via atomic context is via ftrace_dump(). Thus, use a static
buffer of 128 bytes (which covers most events), and if the event is bigger
than that, simply return NULL. The callers of trace_find_next_entry() need
to handle a NULL case, as that's what would happen if the allocation failed.


Fixes: ff895103a84ab ("tracing: Save off entry when peeking at next entry")
Reported-by: kernel test robot <>
Signed-off-by: Steven Rostedt (VMware) <>

diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index 6519b7afc499..8d2b98812625 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -3472,6 +3472,9 @@ __find_next_entry(struct trace_iterator *iter, int *ent_cpu,
return next;

+static char static_temp_buf[STATIC_TEMP_BUF_SIZE];
/* Find the next real entry, without updating the iterator itself */
struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
int *ent_cpu, u64 *ent_ts)
@@ -3480,13 +3483,26 @@ struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
int ent_size = iter->ent_size;
struct trace_entry *entry;

+ /*
+ * If called from ftrace_dump(), then the iter->temp buffer
+ * will be the static_temp_buf and not created from kmalloc.
+ * If the entry size is greater than the buffer, we can
+ * not save it. Just return NULL in that case. This is only
+ * used to add markers when two consecutive events' time
+ * stamps have a large delta. See trace_print_lat_context()
+ */
+ if (iter->temp == static_temp_buf &&
+ STATIC_TEMP_BUF_SIZE < ent_size)
+ return NULL;
* The __find_next_entry() may call peek_next_entry(), which may
* call ring_buffer_peek() that may make the contents of iter->ent
* undefined. Need to copy iter->ent now.
if (iter->ent && iter->ent != iter->temp) {
- if (!iter->temp || iter->temp_size < iter->ent_size) {
+ if ((!iter->temp || iter->temp_size < iter->ent_size) &&
+ !WARN_ON_ONCE(iter->temp == static_temp_buf)) {
iter->temp = kmalloc(iter->ent_size, GFP_KERNEL);
if (!iter->temp)
@@ -9203,6 +9219,9 @@ void ftrace_dump(enum ftrace_dump_mode oops_dump_mode)

/* Simulate the iterator */
+ /* Can not use kmalloc for iter.temp */
+ iter.temp = static_temp_buf;
+ iter.temp_size = STATIC_TEMP_BUF_SIZE;

for_each_tracing_cpu(cpu) {
atomic_inc(&per_cpu_ptr(iter.array_buffer->data, cpu)->disabled);
 \ /
  Last update: 2020-04-03 17:36    [W:0.055 / U:0.652 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site