From: Vince Weaver <vincent.weaver@maine.edu>
To: trinity@vger.kernel.org
Subject: [patch] make perf_event_open() create system-wide events
Date: Tue, 13 May 2014 14:40:54 -0400 (EDT) [thread overview]
Message-ID: <alpine.DEB.2.10.1405131434170.24957@vincent-weaver-1.umelst.maine.edu> (raw)
This patch adds a create_mostly_valid_global_event case to the
perf_event_open() code.
Without this patch it is extremely unlikely that a global
(not per-process) event would be randomly created.
Most users will not notice a change, as you can only create global events
as root or else if /proc/sys/kernel/perf_event_paranoid is set to < 1
(the default is 1).
With this change some of the more obscure event types such as uncore,
northbridge, and RAPL power events are occasionally properly created.
Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
diff --git a/fuzzer/trinity_files/perf_event_open.c b/fuzzer/trinity_files/perf_event_open.c
index c4ccafd..7c20350 100644
--- a/fuzzer/trinity_files/perf_event_open.c
+++ b/fuzzer/trinity_files/perf_event_open.c
@@ -1089,6 +1089,36 @@ static void create_mostly_valid_sampling_event(struct perf_event_attr *attr,
}
+
+/* Creates a global event: one that is not per-process, but system-wide */
+/* To be valid must be created with pid=-1 and cpu being a valid CPU. */
+/* Also usually only root can create these unless */
+/* /proc/sys/kernel/perf_event_paranoid is less than 1. */
+/* Most custom PMU types (uncore/northbridge/RAPL) are covered here. */
+
+static void create_mostly_valid_global_event(struct perf_event_attr *attr,
+ int group_leader)
+{
+
+ attr->type = random_event_type();
+ attr->size = random_attr_size();
+ attr->config = random_event_config(&attr->type,
+ &attr->config1,
+ &attr->config2);
+
+ attr->read_format = random_read_format();
+
+ /* Bitfield parameters, mostly boolean */
+ attr->disabled = rand_bool();
+ attr->inherit = rand_bool();
+ if (group_leader) {
+ attr->pinned = rand_bool();
+ }
+
+ /* Not setting most other paramaters */
+ /* As they tend to be not valid in a global event */
+}
+
/* Creates a completely random event, unlikely to be valid */
static void create_random_event(struct perf_event_attr *attr)
{
@@ -1244,7 +1274,7 @@ void sanitise_perf_event_open(int childno)
shm->syscall[childno].a2 = pid;
/* set up attr structure */
- switch (rand() % 3) {
+ switch (rand() % 4) {
case 0:
create_mostly_valid_counting_event(attr,group_leader);
break;
@@ -1252,6 +1282,9 @@ void sanitise_perf_event_open(int childno)
create_mostly_valid_sampling_event(attr,group_leader);
break;
case 2:
+ create_mostly_valid_global_event(attr,group_leader);
+ break;
+ case 3:
create_random_event(attr);
break;
default:
next reply other threads:[~2014-05-13 18:40 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-05-13 18:40 Vince Weaver [this message]
2014-05-13 18:45 ` [patch] make perf_event_open() create system-wide events Vince Weaver
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=alpine.DEB.2.10.1405131434170.24957@vincent-weaver-1.umelst.maine.edu \
--to=vincent.weaver@maine.edu \
--cc=trinity@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).