linux-kernel - [PATCH] tracing: missing information in trace binary format

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [day] [month] [year] [list]
Message-ID: <BA096D1AF8AE9F408CA552890F7C14010ACD1A43E1@wiems02002.ceu.corp.heidelberg.com>
Date:	Fri, 19 Nov 2010 14:36:33 +0100
From:	"Husterer, Thomas RD-6.2.3" <Thomas.Husterer@...delberg.com>
To:	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: [PATCH] tracing: missing information in trace binary format


i am working with the latest patch of linux-rt (linux-2.6.33.7-rt29) and try to use
the tracer functionality with some userspace tools.
This works great when i use the ascii output of the trace file.
But parsing is not very easy due to dynamic field width and different types and length of seperators.
Additionaly this parser is not very stable over the time.


Fortunately there is a binary format (with trace_option 'bin') which can be
interpeted very fast and easy. But it is missing some data in this format.

1. There is a field value 'next_cpu' which is not written to the output.
2. There are two different event types (ctx and wake) which cannot be differentiated.
3. The Trace Output contains an ascii header and the binary body which cannot be
   separated very well.

Now I have implemented some proposals in the patch below.
1. Added the missing value 'next_cpu' (4Bytes)
2. Added one Byte which defines the event-type ctx:'>', wake:'+'
3. Insertet a uniq Sting betwen the ascii header and the binary body
   '#BinaryStart: Typ: 2 Reclen: 33\n'
   The Sting contains some info about the binary type (for future usage)
   and the length of one entry in Bytes
   and one terminating linefeed at the end.
   Behind this linefeed starts the binary data.

The patch works in linux-2.6.33.7-rt29 and linux-2.6.33.7.
Would it be possible to add this or a similar solution to the mainline kernel ?
I am not a kernel developer, and i am curently not subscribed to the kernel mailing
list but i think the solution described here could be useful for many other users.

Please CC: replies also to thomas.husterer@...delberg.com

best regards
Thomas Husterer (thus)

diff -ur linux-2.6.33.7-rt29.orig/kernel/trace/trace.c linux-2.6.33.7-rt29/kernel/trace/trace.c
--- linux-2.6.33.7-rt29.orig/kernel/trace/trace.c       2010-11-19 11:02:34.000000000 +0100
+++ linux-2.6.33.7-rt29/kernel/trace/trace.c    2010-11-19 12:08:14.000000000 +0100
@@ -2066,6 +2066,10 @@
                        if (!(trace_flags & TRACE_ITER_VERBOSE))
                                print_func_help_header(m);
                }
+                if (trace_flags & TRACE_ITER_BIN) {
+                       seq_puts(m, "#BinaryStart: Typ: 2 Reclen: 33\n");
+                }
+
        } else if (iter->leftover) {
                /*
                 * If we filled the seq_file buffer earlier, we
diff -ur linux-2.6.33.7-rt29.orig/kernel/trace/trace_output.c linux-2.6.33.7-rt29/kernel/trace/trace_output.c
--- linux-2.6.33.7-rt29.orig/kernel/trace/trace_output.c        2010-08-02 19:27:18.000000000 +0200
+++ linux-2.6.33.7-rt29/kernel/trace/trace_output.c     2010-11-19 11:24:24.000000000 +0100
@@ -995,7 +995,7 @@
 }

 static enum print_line_t trace_ctxwake_bin(struct trace_iterator *iter,
-                                          int flags)
+                                          char S)
 {
        struct ctx_switch_entry *field;
        struct trace_seq *s = &iter->seq;
@@ -1008,16 +1008,30 @@
        SEQ_PUT_FIELD_RET(s, field->next_pid);
        SEQ_PUT_FIELD_RET(s, field->next_prio);
        SEQ_PUT_FIELD_RET(s, field->next_state);
+       SEQ_PUT_FIELD_RET(s, field->next_cpu);
+       SEQ_PUT_FIELD_RET(s, S);

        return TRACE_TYPE_HANDLED;
 }

+static enum print_line_t trace_ctx_bin(struct trace_iterator *iter,
+                                          int flags)
+{
+        return trace_ctxwake_bin(iter,'>');
+}
+
+static enum print_line_t trace_wake_bin(struct trace_iterator *iter,
+                                          int flags)
+{
+        return trace_ctxwake_bin(iter,'+');
+}
+
 static struct trace_event trace_ctx_event = {
        .type           = TRACE_CTX,
        .trace          = trace_ctx_print,
        .raw            = trace_ctx_raw,
        .hex            = trace_ctx_hex,
-       .binary         = trace_ctxwake_bin,
+       .binary         = trace_ctx_bin,
 };

 static struct trace_event trace_wake_event = {
@@ -1025,7 +1039,7 @@
        .trace          = trace_wake_print,
        .raw            = trace_wake_raw,
        .hex            = trace_wake_hex,
-       .binary         = trace_ctxwake_bin,
+       .binary         = trace_wake_bin,
 };

 /* TRACE_SPECIAL */

Confidentiality note:
The information in this email and any attachment may contain confidential and proprietary information of Heidelberger Druckmaschinen AG and/or its affiliates and may be privileged or otherwise protected from disclosure. If you are not the intended recipient, you are hereby notified that any review, reliance or distribution by others or forwarding without express permission is strictly prohibited and may cause liability. In case you have received this message due to an error in transmission, we kindly ask you to notify the sender immediately and to delete this email and any attachment from your system.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/