Skip to contentSkip to conversation

Add support for i2c tracepoints

PATCH Mathieu Desnoyers 2 participants 6 messages Oct 3, 2016 – Oct 4, 2016 5 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 10 lines · ----- On Oct 3, 2016, at 2:21 PM, Simon Marchi [email protected] wrote:

This patch teaches lttng-modules about the i2c tracepoints in the Linux kernel.

It contains the following tracepoints:

* i2c_write
* i2c_read
* i2c_reply
* i2c_result

Quoted 5 lines

I translated the fields and assignments from the kernel's include/trace/events/i2c.h as well as I could. I also tried building this module against a kernel without CONFIG_I2C, and it built fine (the required types are unconditionally defined). So I don't think any "#if CONFIG_I2C" or similar are required.

Hi Simon,

The global approach taken in this patch looks good. However, it raises the question of sensitive information. I'm pretty sure i2c buffers may contain keyboard keystrokes, which is sensitive info (passwords). It's the same with read/write system call payloads, network application-level payload, and HID keystrokes. Back in the days of lttng 0.x, I made sure it would not allow users to gather a trace with sensitive information by mistake. You really had to clearly enable gathering of such sensitive information.

One possibility in lttng-modules 2.x would be to introduce a module parameter for each lttng-modules probe that would specify whether sensitive information should be traced. I could be specified by the system administrator at module load time. The default would be that no sensitive information is gathered.

I think we should *not* make this a per-session/dynamic option, because it's really something that the machine administrator should decide on a machine-by-machine basis. In production environments, this would be expected to disallow gathering sensitive info (default).

In specific use-cases where security is no concern (e.g. embedded board during development), this could be explicitly enabled by the developer with root access by passing the module flag at load time.

What do you think ?

Thanks,

Mathieu

Quoted 178 lines

---

instrumentation/events/lttng-module/i2c.h | 89 +++++++++++++++++++++++++++++++
probes/Kbuild                             |  1 +
probes/lttng-probe-i2c.c                  | 47 ++++++++++++++++
3 files changed, 137 insertions(+)

create mode 100644 instrumentation/events/lttng-module/i2c.h create mode 100644 probes/lttng-probe-i2c.c

instrumentation/events/lttng-module/i2c.h new file +89 -0
new file mode 100644
index 0000000..68ce80d
@@ -0,0 +1,89 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM i2c
+
+#if !defined(LTTNG_TRACE_I2C_H) || defined(TRACE_HEADER_MULTI_READ)
+#define LTTNG_TRACE_I2C_H
+
+#include <probes/lttng-tracepoint-event.h>
+
+#ifndef _TRACE_I2C_DEF_
+#define _TRACE_I2C_DEF_
+
+#endif /* _TRACE_I2C_DEF_ */
+
+/*
+ * __i2c_transfer() write request
+ */
+LTTNG_TRACEPOINT_EVENT(i2c_write,
+
+	TP_PROTO(const struct i2c_adapter *adap, const struct i2c_msg *msg, int num),
+
+	TP_ARGS(adap, msg, num),
+
+	TP_FIELDS(
+		ctf_integer(int, adapter_nr, adap->nr)
+		ctf_integer(__u16, msg_nr, num)
+		ctf_integer(__u16, addr, msg->addr)
+		ctf_integer(__u16, flags, msg->flags)
+		ctf_integer(__u16, len, msg->len)
+		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+	)
+)
+
+/*
+ * __i2c_transfer() read request
+ */
+LTTNG_TRACEPOINT_EVENT(i2c_read,
+
+	TP_PROTO(const struct i2c_adapter *adap, const struct i2c_msg *msg, int num),
+
+	TP_ARGS(adap, msg, num),
+
+	TP_FIELDS(
+		ctf_integer(int, adapter_nr, adap->nr)
+		ctf_integer(__u16, msg_nr, num)
+		ctf_integer(__u16, addr, msg->addr)
+		ctf_integer(__u16, flags, msg->flags)
+		ctf_integer(__u16, len, msg->len)
+	)
+)
+
+/*
+ * __i2c_transfer() read reply
+ */
+LTTNG_TRACEPOINT_EVENT(i2c_reply,
+
+	TP_PROTO(const struct i2c_adapter *adap, const struct i2c_msg *msg, int num),
+
+	TP_ARGS(adap, msg, num),
+
+	TP_FIELDS(
+		ctf_integer(int, adapter_nr, adap->nr)
+		ctf_integer(__u16, msg_nr, num)
+		ctf_integer(__u16, addr, msg->addr)
+		ctf_integer(__u16, flags, msg->flags)
+		ctf_integer(__u16, len, msg->len)
+		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+	)
+)
+
+/*
+ * __i2c_transfer() result
+ */
+LTTNG_TRACEPOINT_EVENT(i2c_result,
+
+	TP_PROTO(const struct i2c_adapter *adap, int num, int ret),
+
+	TP_ARGS(adap, num, ret),
+
+	TP_FIELDS(
+		ctf_integer(int, adapter_nr, adap->nr)
+		ctf_integer(__u16, nr_msgs, num)
+		ctf_integer(__s16, ret, ret)
+	)
+)
+
+#endif /*  LTTNG_TRACE_I2C_H */
+
+/* This part must be outside protection */
+#include <probes/define_trace.h>
probes/Kbuild +0 -0
index 8ae9a6b..ca7c0b1 100644
@@ -5,6 +5,7 @@ include $(TOP_LTTNG_MODULES_DIR)/Makefile.ABI.workarounds
 ccflags-y += -I$(TOP_LTTNG_MODULES_DIR)
 

obj-$(CONFIG_LTTNG) += lttng-probe-sched.o
+obj-$(CONFIG_LTTNG) += lttng-probe-i2c.o
obj-$(CONFIG_LTTNG) += lttng-probe-irq.o
obj-$(CONFIG_LTTNG) += lttng-probe-timer.o
obj-$(CONFIG_LTTNG) += lttng-probe-kmem.o

probes/lttng-probe-i2c.c new file +47 -1
new file mode 100644
index 0000000..94ebdc0
@@ -0,0 +1,47 @@
+/*
+ * probes/lttng-probe-i2c.c
+ *
+ * LTTng i2c probes.
+ *
+ * Copyright (C) 2012 Mathieu Desnoyers <[email protected]>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; only
+ * version 2.1 of the License.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <linux/module.h>
+#include <lttng-tracer.h>
+
+/*
+ * Create the tracepoint static inlines from the kernel to validate that our
+ * trace event macros match the kernel we run on.
+ */
+#include <trace/events/i2c.h>
+
+/*
+ * Create LTTng tracepoint probes.
+ */
+#define LTTNG_PACKAGE_BUILD
+#define CREATE_TRACE_POINTS
+#define TRACE_INCLUDE_PATH instrumentation/events/lttng-module
+
+#include <instrumentation/events/lttng-module/i2c.h>
+
+MODULE_LICENSE("GPL and additional rights");
+MODULE_AUTHOR("Mathieu Desnoyers <[email protected]>");
+MODULE_DESCRIPTION("LTTng i2c probes");
+MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
+	__stringify(LTTNG_MODULES_MINOR_VERSION) "."
+	__stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
+	LTTNG_MODULES_EXTRAVERSION);
--

2.9.3

_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

Signature · 7 lines
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw
Quoted 43 lines · Mathieu Desnoyers
Quoted 10 lines · ----- On Oct 3, 2016, at 2:21 PM, Simon Marchi [email protected] wrote:

This patch teaches lttng-modules about the i2c tracepoints in the Linux kernel.

It contains the following tracepoints:

* i2c_write
* i2c_read
* i2c_reply
* i2c_result

Quoted 5 lines

I translated the fields and assignments from the kernel's include/trace/events/i2c.h as well as I could. I also tried building this module against a kernel without CONFIG_I2C, and it built fine (the required types are unconditionally defined). So I don't think any "#if CONFIG_I2C" or similar are required.

Hi Simon,

The global approach taken in this patch looks good. However, it raises the question of sensitive information. I'm pretty sure i2c buffers may contain keyboard keystrokes, which is sensitive info (passwords). It's the same with read/write system call payloads, network application-level payload, and HID keystrokes. Back in the days of lttng 0.x, I made sure it would not allow users to gather a trace with sensitive information by mistake. You really had to clearly enable gathering of such sensitive information.

One possibility in lttng-modules 2.x would be to introduce a module parameter for each lttng-modules probe that would specify whether sensitive information should be traced. I could be specified by the system administrator at module load time. The default would be that no sensitive information is gathered.

I think we should *not* make this a per-session/dynamic option, because it's really something that the machine administrator should decide on a machine-by-machine basis. In production environments, this would be expected to disallow gathering sensitive info (default).

In specific use-cases where security is no concern (e.g. embedded board during development), this could be explicitly enabled by the developer with root access by passing the module flag at load time.

Quoted 5 lines

What do you think ?

Thanks,

Mathieu

It makes sense. I just tested it and it works fine, here's a prototype.

On a development machine, it's easy to enable once and for all by putting

options lttng-probe-i2c trace_sensitive_data=1

in /etc/modprobe.d/lttng.conf for example.

Here I have put the knob in the i2c module itself, which means that if we have this situation in other modules, we'll have to copy the code that defines the parameter. That gives us a module-level granularity, which can be an upside (more control over which modules collect sensitive data) or a downside (more complex, need to pass the option too all modules).

I assume it would be possible to put the parameter in the lttng-tracer module, on which all other modules depend, in order to have an all-or-nothing approach.

One cool thing about module parameters is that they can be modified dynamically by writing to sysfs (/sys/module/lttng_probe_i2c/parameters/trace_sensitive_data). So you can change the setting mid-trace, I tried it and it works.

If the gathering of sensitive info is off, the sequence of bytes is just empty. Since you have the length of the message as a different field, the length of the sequence does not match the former. The trace viewer can probably use this mismatch as a sign that the data was omitted, and that it might not have been an actual empty message.

From 69aa1c3d73b5b0dbf031b90c6e38d4b2a0ad8699 Mon Sep 17 00:00:00 2001
From: Simon Marchi <[email protected]>
Date: Tue, 4 Oct 2016 11:38:19 -0400
Subject: [PATCH] Provide knob to control tracing of i2c buffer contents

---
 instrumentation/events/lttng-module/i2c.h | 6 ++++--
 probes/lttng-probe-i2c.c                  | 7 +++++++
 2 files changed, 11 insertions(+), 2 deletions(-)
instrumentation/events/lttng-module/i2c.h +4 -2
index 68ce80d..1ede276 100644
@@ -26,7 +26,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_write,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)
 	)
 )
 
@@ -63,7 +64,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_reply,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)
 	)
 )
 
probes/lttng-probe-i2c.c +7 -0
index 94ebdc0..fd30a73 100644
@@ -21,6 +21,7 @@
  */
 
 #include <linux/module.h>
+#include <linux/moduleparam.h>
 #include <lttng-tracer.h>
 
 /*
@@ -36,6 +37,12 @@
 #define CREATE_TRACE_POINTS
 #define TRACE_INCLUDE_PATH instrumentation/events/lttng-module
 
+static int trace_sensitive_data = 0;
+module_param(trace_sensitive_data, int, 0644);
+MODULE_PARM_DESC(trace_sensitive_data, "Whether to trace possibly sensitive "
+				       "data (i2c buffer contents) or not (1 or "
+				       "0, default: 0).");
+
 #include <instrumentation/events/lttng-module/i2c.h>
 
 MODULE_LICENSE("GPL and additional rights");
Signature · 8 lines
2.10.1



_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

A message that is not in the archive (sent off-list or not yet received): [email protected]

#5 lore raw
Quoted 73 lines · ----- On Oct 4, 2016, at 1:08 PM, Simon Marchi [email protected] wrote:
Quoted 43 lines · Mathieu Desnoyers
Quoted 10 lines · ----- On Oct 3, 2016, at 2:21 PM, Simon Marchi [email protected] wrote:

This patch teaches lttng-modules about the i2c tracepoints in the Linux kernel.

It contains the following tracepoints:

* i2c_write
* i2c_read
* i2c_reply
* i2c_result

Quoted 5 lines

I translated the fields and assignments from the kernel's include/trace/events/i2c.h as well as I could. I also tried building this module against a kernel without CONFIG_I2C, and it built fine (the required types are unconditionally defined). So I don't think any "#if CONFIG_I2C" or similar are required.

Hi Simon,

The global approach taken in this patch looks good. However, it raises the question of sensitive information. I'm pretty sure i2c buffers may contain keyboard keystrokes, which is sensitive info (passwords). It's the same with read/write system call payloads, network application-level payload, and HID keystrokes. Back in the days of lttng 0.x, I made sure it would not allow users to gather a trace with sensitive information by mistake. You really had to clearly enable gathering of such sensitive information.

One possibility in lttng-modules 2.x would be to introduce a module parameter for each lttng-modules probe that would specify whether sensitive information should be traced. I could be specified by the system administrator at module load time. The default would be that no sensitive information is gathered.

I think we should *not* make this a per-session/dynamic option, because it's really something that the machine administrator should decide on a machine-by-machine basis. In production environments, this would be expected to disallow gathering sensitive info (default).

In specific use-cases where security is no concern (e.g. embedded board during development), this could be explicitly enabled by the developer with root access by passing the module flag at load time.

Quoted 5 lines

What do you think ?

Thanks,

Mathieu

It makes sense. I just tested it and it works fine, here's a prototype.

On a development machine, it's easy to enable once and for all by putting

options lttng-probe-i2c trace_sensitive_data=1

in /etc/modprobe.d/lttng.conf for example.

Here I have put the knob in the i2c module itself, which means that if we have this situation in other modules, we'll have to copy the code that defines the parameter. That gives us a module-level granularity, which can be an upside (more control over which modules collect sensitive data) or a downside (more complex, need to pass the option too all modules).

I assume it would be possible to put the parameter in the lttng-tracer module, on which all other modules depend, in order to have an all-or-nothing approach.

One cool thing about module parameters is that they can be modified dynamically by writing to sysfs (/sys/module/lttng_probe_i2c/parameters/trace_sensitive_data).
So you can change the setting mid-trace, I tried it and it works.

Quoted 5 lines

If the gathering of sensitive info is off, the sequence of bytes is just empty. Since you have the length of the message as a different field, the length of the sequence does not match the former. The trace viewer can probably use this mismatch as a sign that the data was omitted, and that it might not have been an actual empty message.

Perhaps we could rename the option to "trace_sensitive_payload" ?

I think we should keep it at the probe level, like you do in your prototype.

Quoted 18 lines

From 69aa1c3d73b5b0dbf031b90c6e38d4b2a0ad8699 Mon Sep 17 00:00:00 2001
From: Simon Marchi <[email protected]>
Date: Tue, 4 Oct 2016 11:38:19 -0400
Subject: [PATCH] Provide knob to control tracing of i2c buffer contents

---
instrumentation/events/lttng-module/i2c.h | 6 ++++--
probes/lttng-probe-i2c.c | 7 +++++++

2 files changed, 11 insertions(+), 2 deletions(-)
instrumentation/events/lttng-module/i2c.h +0 -0
index 68ce80d..1ede276 100644

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_write, · line 30

Quoted 6 lines
instrumentation/events/lttng-module/i2c.h +2 -1
@@ -26,7 +26,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_write,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)

Since "trace_sensitive_data" can be changed at runtime concurrently with tracing, we should load it with LOAD_ONCE(), keep it in a local variable (TRACEPOINT_EVENT_CODE, TP_locvar and TP_code), and use that in as argument. Otherwise the compiler can perform two fetch of that variable, and its value may change in between, thus leading to a NULL pointer dereference (OOPS).

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_reply, · line 71

Quoted 13 lines
instrumentation/events/lttng-module/i2c.h +2 -1
 	)
 )
 
@@ -63,7 +64,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_reply,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)
 	)
 )
 

Same here.

Can you also fold it with your i2c patch ?

Thanks,

Mathieu

Quoted 27 lines
probes/lttng-probe-i2c.c +7 -1
index 94ebdc0..fd30a73 100644
@@ -21,6 +21,7 @@
 */
 
 #include <linux/module.h>
+#include <linux/moduleparam.h>
 #include <lttng-tracer.h>
 
 /*
@@ -36,6 +37,12 @@
 #define CREATE_TRACE_POINTS
 #define TRACE_INCLUDE_PATH instrumentation/events/lttng-module
 
+static int trace_sensitive_data = 0;
+module_param(trace_sensitive_data, int, 0644);
+MODULE_PARM_DESC(trace_sensitive_data, "Whether to trace possibly sensitive "
+				       "data (i2c buffer contents) or not (1 or "
+				       "0, default: 0).");
+
 #include <instrumentation/events/lttng-module/i2c.h>
 
 MODULE_LICENSE("GPL and additional rights");
--

2.10.1

Signature · 7 lines
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

A message that is not in the archive (sent off-list or not yet received): [email protected]

#7 lore raw
Quoted 75 lines · ----- On Oct 4, 2016, at 1:37 PM, Mathieu Desnoyers…
Quoted 73 lines · ----- On Oct 4, 2016, at 1:08 PM, Simon Marchi [email protected] wrote:
Quoted 43 lines · Mathieu Desnoyers
Quoted 10 lines · ----- On Oct 3, 2016, at 2:21 PM, Simon Marchi [email protected] wrote:

This patch teaches lttng-modules about the i2c tracepoints in the Linux kernel.

It contains the following tracepoints:

* i2c_write
* i2c_read
* i2c_reply
* i2c_result

Quoted 5 lines

I translated the fields and assignments from the kernel's include/trace/events/i2c.h as well as I could. I also tried building this module against a kernel without CONFIG_I2C, and it built fine (the required types are unconditionally defined). So I don't think any "#if CONFIG_I2C" or similar are required.

Hi Simon,

The global approach taken in this patch looks good. However, it raises the question of sensitive information. I'm pretty sure i2c buffers may contain keyboard keystrokes, which is sensitive info (passwords). It's the same with read/write system call payloads, network application-level payload, and HID keystrokes. Back in the days of lttng 0.x, I made sure it would not allow users to gather a trace with sensitive information by mistake. You really had to clearly enable gathering of such sensitive information.

One possibility in lttng-modules 2.x would be to introduce a module parameter for each lttng-modules probe that would specify whether sensitive information should be traced. I could be specified by the system administrator at module load time. The default would be that no sensitive information is gathered.

I think we should *not* make this a per-session/dynamic option, because it's really something that the machine administrator should decide on a machine-by-machine basis. In production environments, this would be expected to disallow gathering sensitive info (default).

In specific use-cases where security is no concern (e.g. embedded board during development), this could be explicitly enabled by the developer with root access by passing the module flag at load time.

Quoted 5 lines

What do you think ?

Thanks,

Mathieu

It makes sense. I just tested it and it works fine, here's a prototype.

On a development machine, it's easy to enable once and for all by putting

options lttng-probe-i2c trace_sensitive_data=1

in /etc/modprobe.d/lttng.conf for example.

Here I have put the knob in the i2c module itself, which means that if we have this situation in other modules, we'll have to copy the code that defines the parameter. That gives us a module-level granularity, which can be an upside (more control over which modules collect sensitive data) or a downside (more complex, need to pass the option too all modules).

I assume it would be possible to put the parameter in the lttng-tracer module, on which all other modules depend, in order to have an all-or-nothing approach.

One cool thing about module parameters is that they can be modified dynamically by writing to sysfs (/sys/module/lttng_probe_i2c/parameters/trace_sensitive_data).
So you can change the setting mid-trace, I tried it and it works.

Quoted 7 lines
Quoted 5 lines

If the gathering of sensitive info is off, the sequence of bytes is just empty. Since you have the length of the message as a different field, the length of the sequence does not match the former. The trace viewer can probably use this mismatch as a sign that the data was omitted, and that it might not have been an actual empty message.

Perhaps we could rename the option to "trace_sensitive_payload" ?

Or: allow_sensitive_payload_extraction ?

Thanks,

Mathieu

Quoted 90 lines

I think we should keep it at the probe level, like you do in your prototype.

Quoted 18 lines

From 69aa1c3d73b5b0dbf031b90c6e38d4b2a0ad8699 Mon Sep 17 00:00:00 2001
From: Simon Marchi <[email protected]>
Date: Tue, 4 Oct 2016 11:38:19 -0400
Subject: [PATCH] Provide knob to control tracing of i2c buffer contents

---
instrumentation/events/lttng-module/i2c.h | 6 ++++--
probes/lttng-probe-i2c.c | 7 +++++++

2 files changed, 11 insertions(+), 2 deletions(-)
instrumentation/events/lttng-module/i2c.h +0 -0
index 68ce80d..1ede276 100644

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_write, · line 30

Quoted 6 lines
instrumentation/events/lttng-module/i2c.h +2 -1
@@ -26,7 +26,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_write,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)

Since "trace_sensitive_data" can be changed at runtime concurrently with tracing, we should load it with LOAD_ONCE(), keep it in a local variable (TRACEPOINT_EVENT_CODE, TP_locvar and TP_code), and use that in as argument. Otherwise the compiler can perform two fetch of that variable, and its value may change in between, thus leading to a NULL pointer dereference (OOPS).

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_reply, · line 71

Quoted 13 lines
instrumentation/events/lttng-module/i2c.h +2 -1
 	)
 )
 
@@ -63,7 +64,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_reply,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)
 	)
 )
 

Same here.

Can you also fold it with your i2c patch ?

Thanks,

Mathieu

Quoted 26 lines
probes/lttng-probe-i2c.c +7 -1
index 94ebdc0..fd30a73 100644
@@ -21,6 +21,7 @@
 */
 
 #include <linux/module.h>
+#include <linux/moduleparam.h>
 #include <lttng-tracer.h>
 
 /*
@@ -36,6 +37,12 @@
 #define CREATE_TRACE_POINTS
 #define TRACE_INCLUDE_PATH instrumentation/events/lttng-module
 
+static int trace_sensitive_data = 0;
+module_param(trace_sensitive_data, int, 0644);
+MODULE_PARM_DESC(trace_sensitive_data, "Whether to trace possibly sensitive "
+				       "data (i2c buffer contents) or not (1 or "
+				       "0, default: 0).");
+
 #include <instrumentation/events/lttng-module/i2c.h>
 
 MODULE_LICENSE("GPL and additional rights");
--
Quoted 1 line

2.10.1

--
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Signature · 7 lines
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
#8 lore raw
In reply to a message not in the archive
Quoted 1 line · Mathieu Desnoyers

Perhaps we could rename the option to "trace_sensitive_payload" ?

Of all those:

- trace_sensitive_data
- trace_sensitive_payload
- allow_sensitive_payload_extraction
- extract_sensitive_payload

I prefer "extract_sensitive_payload". But it's your call, just tell me which one you prefer.

Quoted 25 lines

I think we should keep it at the probe level, like you do in your prototype.

Quoted 23 lines

From 69aa1c3d73b5b0dbf031b90c6e38d4b2a0ad8699 Mon Sep 17 00:00:00 2001
From: Simon Marchi <[email protected]>
Date: Tue, 4 Oct 2016 11:38:19 -0400
Subject: [PATCH] Provide knob to control tracing of i2c buffer contents

---
instrumentation/events/lttng-module/i2c.h | 6 ++++--
probes/lttng-probe-i2c.c | 7 +++++++

2 files changed, 11 insertions(+), 2 deletions(-)
instrumentation/events/lttng-module/i2c.h +1 -1
index 68ce80d..1ede276 100644
@@ -26,7 +26,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_write,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
Quoted 8 lines

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_write, · line 30

Quoted 1 line
instrumentation/events/lttng-module/i2c.h +1 -0
+				 __u16, trace_sensitive_data ? msg->len : 0)

Since "trace_sensitive_data" can be changed at runtime concurrently with tracing, we should load it with LOAD_ONCE(), keep it in a local variable (TRACEPOINT_EVENT_CODE, TP_locvar and TP_code), and use that in as argument. Otherwise the compiler can perform two fetch of that variable, and its value may change in between, thus leading to a NULL pointer dereference (OOPS).

Ahh good catch. I'll look up those things you mentioned, but do you have a usage pattern in mind that we could use here? Where would the local variable be? I know of gcc's statement expression, we could use that, but do you have something else in mind?

Quoted 16 lines

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_reply, · line 71

Quoted 13 lines
instrumentation/events/lttng-module/i2c.h +2 -1
 	)
 )
 
@@ -63,7 +64,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_reply,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)
 	)
 )
 

Same here.

Quoted 1 line

Can you also fold it with your i2c patch ?

I'll send a v2 of the original patch once it's clear how to do the load once thingy.

Thanks!

Simon

_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

A message that is not in the archive (sent off-list or not yet received): [email protected]

#10 lore raw
Quoted 5 lines · ----- On Oct 4, 2016, at 3:48 PM, Simon Marchi [email protected] wrote:
Quoted 1 line · Mathieu Desnoyers

Perhaps we could rename the option to "trace_sensitive_payload" ?

Of all those:

Quoted 7 lines

- trace_sensitive_data
- trace_sensitive_payload
- allow_sensitive_payload_extraction
- extract_sensitive_payload

I prefer "extract_sensitive_payload". But it's your call, just tell me which one you prefer.

Let's go for "extract_sensitive_payload".

Quoted 35 lines
Quoted 26 lines

I think we should keep it at the probe level, like you do in your prototype.

Quoted 23 lines

From 69aa1c3d73b5b0dbf031b90c6e38d4b2a0ad8699 Mon Sep 17 00:00:00 2001
From: Simon Marchi <[email protected]>
Date: Tue, 4 Oct 2016 11:38:19 -0400
Subject: [PATCH] Provide knob to control tracing of i2c buffer contents

---
instrumentation/events/lttng-module/i2c.h | 6 ++++--
probes/lttng-probe-i2c.c | 7 +++++++

2 files changed, 11 insertions(+), 2 deletions(-)
instrumentation/events/lttng-module/i2c.h +1 -1
index 68ce80d..1ede276 100644
@@ -26,7 +26,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_write,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
Quoted 8 lines

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_write, · line 30

Quoted 1 line
instrumentation/events/lttng-module/i2c.h +1 -0
+				 __u16, trace_sensitive_data ? msg->len : 0)

Since "trace_sensitive_data" can be changed at runtime concurrently with tracing, we should load it with LOAD_ONCE(), keep it in a local variable (TRACEPOINT_EVENT_CODE, TP_locvar and TP_code), and use that in as argument. Otherwise the compiler can perform two fetch of that variable, and its value may change in between, thus leading to a NULL pointer dereference (OOPS).

Quoted 4 lines

Ahh good catch. I'll look up those things you mentioned, but do you have a usage pattern in mind that we could use here? Where would the local variable be? I know of gcc's statement expression, we could use that, but do you have something else in mind?

See http://lttng.org/docs/#doc-lttng-tracepoint-event-code

Note that there are slight changes to TP_code() in 2.8. See source for details.

Quoted 17 lines
Quoted 17 lines
Quoted 5 lines
instrumentation/events/lttng-module/i2c.h +0 -0
 	)
 )
 

instrumentation/events/lttng-module/i2c.h LTTNG_TRACEPOINT_EVENT(i2c_reply, · line 71

Quoted 9 lines
instrumentation/events/lttng-module/i2c.h +2 -1
@@ -63,7 +64,8 @@ LTTNG_TRACEPOINT_EVENT(i2c_reply,
 		ctf_integer(__u16, addr, msg->addr)
 		ctf_integer(__u16, flags, msg->flags)
 		ctf_integer(__u16, len, msg->len)
-		ctf_sequence_hex(__u8, buf, msg->buf, __u16, msg->len)
+		ctf_sequence_hex(__u8, buf, trace_sensitive_data ? msg->buf : NULL,
+				 __u16, trace_sensitive_data ? msg->len : 0)
 	)
 )
 

Same here.

Quoted 4 lines
Quoted 1 line

Can you also fold it with your i2c patch ?

I'll send a v2 of the original patch once it's clear how to do the load once thingy.

Allright, thanks!

Mathieu

Quoted 4 lines

Thanks!

Simon

Signature · 7 lines
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev