Skip to contentSkip to conversation

tracing: Fix synthetic event bug

PATCH In review Steven Rostedt 4 participants 8 messages Oct 12, 2022

Subsystems (2)
lore
Share & tools
lore

Steven Rostedt

author maintainer
#0 lore raw

v2 cover tracing: Fix synthetic event bug

The follow commands caused a crash:

  # cd /sys/kernel/tracing
  # echo 's:open char file[]' > dynamic_events
  # echo 'hist:keys=common_pid:file=filename:onchange($file).trace(open,$file)' > events/syscalls/sys_enter_openat/trigger'
  # echo 1 > events/synthetic/open/enable

BOOM!

The problem is that the synthetic event field "char file[]" will read the value given to it as a string without any memory checks to make sure the address is valid. The above example will pass in the user space address and the sythetic event code will happily call strlen() on it and then strscpy() where either one will cause an oops when accessing user space addresses.

Changes since v1: https://lore.kernel.org/all/[email protected]/

- Handle "(fault)" printing when there's a fault

Steven Rostedt (Google) (3):
tracing: Move duplicate code of trace_kprobe/eprobe.c into header
tracing: Add "(fault)" name injection to kernel probes
tracing: Fix reading strings from synthetic events

----

 kernel/trace/trace_eprobe.c       |  60 ++------------------
 kernel/trace/trace_events_synth.c |  23 ++++++--
 kernel/trace/trace_kprobe.c       |  60 ++------------------
 kernel/trace/trace_probe_kernel.h | 115 ++++++++++++++++++++++++++++++++++++++
 4 files changed, 142 insertions(+), 116 deletions(-)
 create mode 100644 kernel/trace/trace_probe_kernel.h
author maintainer
#1 lore raw

PATCH v2 1/3 tracing: Move duplicate code of trace_kprobe/eprobe.c into header

From: "Steven Rostedt (Google)" <[email protected]>

The functions:

fetch_store_strlen_user()
fetch_store_strlen()
fetch_store_string_user()
fetch_store_string()

are identical in both trace_kprobe.c and trace_eprobe.c. Move them into a new header file trace_probe_kernel.h to share it. This code will later be used by the synthetic events as well.

Marked for stable as a fix for a crash in synthetic events requires it.

Fixes bd82631d7ccdc ("tracing: Add support for dynamic strings to synthetic events")
---
 kernel/trace/trace_eprobe.c       | 60 ++-----------------
 kernel/trace/trace_kprobe.c       | 60 ++-----------------
 kernel/trace/trace_probe_kernel.h | 96 +++++++++++++++++++++++++++++++
 3 files changed, 106 insertions(+), 110 deletions(-)
 create mode 100644 kernel/trace/trace_probe_kernel.h
kernel/trace/trace_eprobe.c +5 -55
index c08bde9871ec..5dd0617e5df6 100644
@@ -16,6 +16,7 @@
 #include "trace_dynevent.h"
 #include "trace_probe.h"
 #include "trace_probe_tmpl.h"
+#include "trace_probe_kernel.h"
 
 #define EPROBE_EVENT_SYSTEM "eprobes"
 
@@ -456,29 +457,14 @@ NOKPROBE_SYMBOL(process_fetch_insn)
 static nokprobe_inline int
 fetch_store_strlen_user(unsigned long addr)
 {
-	const void __user *uaddr =  (__force const void __user *)addr;
-
-	return strnlen_user_nofault(uaddr, MAX_STRING_SIZE);
+	return kern_fetch_store_strlen_user(addr);
 }
 
 /* Return the length of string -- including null terminal byte */
 static nokprobe_inline int
 fetch_store_strlen(unsigned long addr)
 {
-	int ret, len = 0;
-	u8 c;
-
-#ifdef CONFIG_ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
-	if (addr < TASK_SIZE)
-		return fetch_store_strlen_user(addr);
-#endif
-
-	do {
-		ret = copy_from_kernel_nofault(&c, (u8 *)addr + len, 1);
-		len++;
-	} while (c && ret == 0 && len < MAX_STRING_SIZE);
-
-	return (ret < 0) ? ret : len;
+	return kern_fetch_store_strlen(addr);
 }
 
 /*
@@ -488,21 +474,7 @@ fetch_store_strlen(unsigned long addr)
 static nokprobe_inline int
 fetch_store_string_user(unsigned long addr, void *dest, void *base)
 {
-	const void __user *uaddr =  (__force const void __user *)addr;
-	int maxlen = get_loc_len(*(u32 *)dest);
-	void *__dest;
-	long ret;
-
-	if (unlikely(!maxlen))
-		return -ENOMEM;
-
-	__dest = get_loc_data(dest, base);
-
-	ret = strncpy_from_user_nofault(__dest, uaddr, maxlen);
-	if (ret >= 0)
-		*(u32 *)dest = make_data_loc(ret, __dest - base);
-
-	return ret;
+	return kern_fetch_store_string_user(addr, dest, base);
 }
 
 /*
@@ -512,29 +484,7 @@ fetch_store_string_user(unsigned long addr, void *dest, void *base)
 static nokprobe_inline int
 fetch_store_string(unsigned long addr, void *dest, void *base)
 {
-	int maxlen = get_loc_len(*(u32 *)dest);
-	void *__dest;
-	long ret;
-
-#ifdef CONFIG_ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
-	if ((unsigned long)addr < TASK_SIZE)
-		return fetch_store_string_user(addr, dest, base);
-#endif
-
-	if (unlikely(!maxlen))
-		return -ENOMEM;
-
-	__dest = get_loc_data(dest, base);
-
-	/*
-	 * Try to get string again, since the string can be changed while
-	 * probing.
-	 */
-	ret = strncpy_from_kernel_nofault(__dest, (void *)addr, maxlen);
-	if (ret >= 0)
-		*(u32 *)dest = make_data_loc(ret, __dest - base);
-
-	return ret;
+	return kern_fetch_store_string(addr, dest, base);
 }
 
 static nokprobe_inline int
kernel/trace/trace_kprobe.c +5 -55

90 lines not on this page · raw message

kernel/trace/trace_probe_kernel.h new file +96 -0

96 lines not on this page · raw message

Signature · 1 line
2.35.1
author maintainer
#2 lore raw

PATCH v2 2/3 tracing: Add "(fault)" name injection to kernel probes

In reply to Steven Rostedt

From: "Steven Rostedt (Google)" <[email protected]>

Have the specific functions for kernel probes that read strings to inject the "(fault)" name directly. trace_probes.c does this too (for uprobes) but as the code to read strings are going to be used by synthetic events (and perhaps other utilities), it simplifies the code by making sure those other uses do not need to implement the "(fault)" name injection as well.

Fixes bd82631d7ccdc ("tracing: Add support for dynamic strings to synthetic events")
---
 kernel/trace/trace_probe_kernel.h | 31 +++++++++++++++++++++++++------
 1 file changed, 25 insertions(+), 6 deletions(-)
kernel/trace/trace_probe_kernel.h +25 -6
index 1d43df29a1f8..77dbd9ff9782 100644
@@ -2,6 +2,8 @@
 #ifndef __TRACE_PROBE_KERNEL_H_
 #define __TRACE_PROBE_KERNEL_H_
 
+#define FAULT_STRING "(fault)"
+
 /*
  * This depends on trace_probe.h, but can not include it due to
  * the way trace_probe_tmpl.h is used by trace_kprobe.c and trace_eprobe.c.
@@ -13,8 +15,16 @@ static nokprobe_inline int
 kern_fetch_store_strlen_user(unsigned long addr)
 {
 	const void __user *uaddr =  (__force const void __user *)addr;
+	int ret;
 
-	return strnlen_user_nofault(uaddr, MAX_STRING_SIZE);
+	ret = strnlen_user_nofault(uaddr, MAX_STRING_SIZE);
+	/*
+	 * strnlen_user_nofault returns zero on fault, insert the
+	 * FAULT_STRING when that occurs.
+	 */
+	if (ret <= 0)
+		return strlen(FAULT_STRING) + 1;
+	return ret;
 }
 
 /* Return the length of string -- including null terminal byte */
@@ -34,7 +44,18 @@ kern_fetch_store_strlen(unsigned long addr)
 		len++;
 	} while (c && ret == 0 && len < MAX_STRING_SIZE);
 
-	return (ret < 0) ? ret : len;
+	/* For faults, return enough to hold the FAULT_STRING */
+	return (ret < 0) ? strlen(FAULT_STRING) + 1 : len;
+}
+
+static nokprobe_inline void set_data_loc(int ret, void *dest, void *__dest, void *base, int len)
+{
+	if (ret >= 0) {
+		*(u32 *)dest = make_data_loc(ret, __dest - base);
+	} else {
+		strscpy(__dest, FAULT_STRING, len);
+		ret = strlen(__dest) + 1;
+	}
 }
 
 /*
@@ -55,8 +76,7 @@ kern_fetch_store_string_user(unsigned long addr, void *dest, void *base)
 	__dest = get_loc_data(dest, base);
 
 	ret = strncpy_from_user_nofault(__dest, uaddr, maxlen);
-	if (ret >= 0)
-		*(u32 *)dest = make_data_loc(ret, __dest - base);
+	set_data_loc(ret, dest, __dest, base, maxlen);
 
 	return ret;
 }
@@ -87,8 +107,7 @@ kern_fetch_store_string(unsigned long addr, void *dest, void *base)
 	 * probing.
 	 */
 	ret = strncpy_from_kernel_nofault(__dest, (void *)addr, maxlen);
-	if (ret >= 0)
-		*(u32 *)dest = make_data_loc(ret, __dest - base);
+	set_data_loc(ret, dest, __dest, base, maxlen);
 
 	return ret;
 }
Signature · 1 line
2.35.1
#3 lore raw

RE: [PATCH v2 2/3] tracing: Add "(fault)" name injection to kernel probes

From: Steven Rostedt

Quoted 34 lines

Sent: 12 October 2022 11:41

Have the specific functions for kernel probes that read strings to inject the "(fault)" name directly. trace_probes.c does this too (for uprobes) but as the code to read strings are going to be used by synthetic events (and perhaps other utilities), it simplifies the code by making sure those other uses do not need to implement the "(fault)" name injection as well.

Fixes bd82631d7ccdc ("tracing: Add support for dynamic strings to synthetic events")
---
 kernel/trace/trace_probe_kernel.h | 31 +++++++++++++++++++++++++------
 1 file changed, 25 insertions(+), 6 deletions(-)
kernel/trace/trace_probe_kernel.h +3 -0
index 1d43df29a1f8..77dbd9ff9782 100644
@@ -2,6 +2,8 @@
 #ifndef __TRACE_PROBE_KERNEL_H_
 #define __TRACE_PROBE_KERNEL_H_
 
+#define FAULT_STRING "(fault)"
+
 /*
  * This depends on trace_probe.h, but can not include it due to
  * the way trace_probe_tmpl.h is used by trace_kprobe.c and trace_eprobe.c.
@@ -13,8 +15,16 @@ static nokprobe_inline int
 kern_fetch_store_strlen_user(unsigned long addr)
 {
 	const void __user *uaddr =  (__force const void __user *)addr;
+	int ret;
 

kernel/trace/trace_probe_kernel.h static nokprobe_inline int · line 28

Quoted 10 lines
kernel/trace/trace_probe_kernel.h +8 -1
-	return strnlen_user_nofault(uaddr, MAX_STRING_SIZE);
+	ret = strnlen_user_nofault(uaddr, MAX_STRING_SIZE);
+	/*
+	 * strnlen_user_nofault returns zero on fault, insert the
+	 * FAULT_STRING when that occurs.
+	 */
+	if (ret <= 0)
+		return strlen(FAULT_STRING) + 1;
+	return ret;
 }

Isn't that going to do the wrong thing if the user string is valid memory but just zero length??

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

author maintainer
#4 lore raw

Re: [PATCH v2 2/3] tracing: Add "(fault)" name injection to kernel probes

Quoted 6 lines · On Wed, 12 Oct 2022 12:34:45 +0000 David Laight <[email protected]> wrote:
Quoted 6 lines
kernel/trace/trace_probe_kernel.h +1 -0
@@ -13,8 +15,16 @@ static nokprobe_inline int
 kern_fetch_store_strlen_user(unsigned long addr)
 {
 	const void __user *uaddr =  (__force const void __user *)addr;
+	int ret;
 
Quoted 13 lines

kernel/trace/trace_probe_kernel.h static nokprobe_inline int · line 28

Quoted 10 lines
kernel/trace/trace_probe_kernel.h +8 -1
-	return strnlen_user_nofault(uaddr, MAX_STRING_SIZE);
+	ret = strnlen_user_nofault(uaddr, MAX_STRING_SIZE);
+	/*
+	 * strnlen_user_nofault returns zero on fault, insert the
+	 * FAULT_STRING when that occurs.
+	 */
+	if (ret <= 0)
+		return strlen(FAULT_STRING) + 1;
+	return ret;
 }  

Isn't that going to do the wrong thing if the user string is valid memory but just zero length??

I thought so at first (and was in the process of changing things because of that) until I saw the comment above this code:

/* Return the length of string -- including null terminal byte */

And looking the function of strnlen_user_nofault():

* Returns the size of the string INCLUDING the terminating NUL.

That is, it returns 1 on a zero length string and 0 on fault :-p

Yes, I think we should fix that API, but that's another story.

-- Steve

author maintainer
#5 lore raw

PATCH v2 3/3 tracing: Fix reading strings from synthetic events

In reply to Steven Rostedt

From: "Steven Rostedt (Google)" <[email protected]>

The follow commands caused a crash:

  # cd /sys/kernel/tracing
  # echo 's:open char file[]' > dynamic_events
  # echo 'hist:keys=common_pid:file=filename:onchange($file).trace(open,$file)' > events/syscalls/sys_enter_openat/trigger'
  # echo 1 > events/synthetic/open/enable

BOOM!

The problem is that the synthetic event field "char file[]" will read the value given to it as a string without any memory checks to make sure the address is valid. The above example will pass in the user space address and the sythetic event code will happily call strlen() on it and then strscpy() where either one will cause an oops when accessing user space addresses.

Use the helper functions from trace_kprobe and trace_eprobe that can read strings safely (and actually succeed when the address is from user space and the memory is mapped in).

Now the above can show:

     packagekitd-1721    [000] ...2.   104.597170: open: file=/usr/lib/rpm/fileattrs/cmake.attr
    in:imjournal-978     [006] ...2.   104.599642: open: file=/var/lib/rsyslog/imjournal.state.tmp
     packagekitd-1721    [000] ...2.   104.626308: open: file=/usr/lib/rpm/fileattrs/debuginfo.attr
Fixes bd82631d7ccdc ("tracing: Add support for dynamic strings to synthetic events")
---
 kernel/trace/trace_events_synth.c | 23 +++++++++++++++++------
 1 file changed, 17 insertions(+), 6 deletions(-)
kernel/trace/trace_events_synth.c +17 -6
index 5e8c07aef071..e310052dc83c 100644
@@ -17,6 +17,8 @@
 /* for gfp flag names */
 #include <linux/trace_events.h>
 #include <trace/events/mmflags.h>
+#include "trace_probe.h"
+#include "trace_probe_kernel.h"
 
 #include "trace_synth.h"
 
@@ -409,6 +411,7 @@ static unsigned int trace_string(struct synth_trace_event *entry,
 {
 	unsigned int len = 0;
 	char *str_field;
+	int ret;
 
 	if (is_dynamic) {
 		u32 data_offset;
@@ -417,19 +420,27 @@ static unsigned int trace_string(struct synth_trace_event *entry,
 		data_offset += event->n_u64 * sizeof(u64);
 		data_offset += data_size;
 
-		str_field = (char *)entry + data_offset;
-
-		len = strlen(str_val) + 1;
-		strscpy(str_field, str_val, len);
+		len = kern_fetch_store_strlen((unsigned long)str_val);
 
 		data_offset |= len << 16;
 		*(u32 *)&entry->fields[*n_u64] = data_offset;
 
+		ret = kern_fetch_store_string((unsigned long)str_val, &entry->fields[*n_u64], entry);
+
 		(*n_u64)++;
 	} else {
 		str_field = (char *)&entry->fields[*n_u64];
 
-		strscpy(str_field, str_val, STR_VAR_LEN_MAX);
+#ifdef CONFIG_ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
+		if ((unsigned long)str_val < TASK_SIZE)
+			ret = strncpy_from_user_nofault(str_field, str_val, STR_VAR_LEN_MAX);
+		else
+#endif
+			ret = strncpy_from_kernel_nofault(str_field, str_val, STR_VAR_LEN_MAX);
+
+		if (ret < 0)
+			strcpy(str_field, FAULT_STRING);
+
 		(*n_u64) += STR_VAR_LEN_MAX / sizeof(u64);
 	}
 
@@ -462,7 +473,7 @@ static notrace void trace_event_raw_event_synth(void *__data,
 		val_idx = var_ref_idx[field_pos];
 		str_val = (char *)(long)var_ref_vals[val_idx];
 
-		len = strlen(str_val) + 1;
+		len = kern_fetch_store_strlen((unsigned long)str_val);
 
 		fields_size += len;
 	}
Signature · 1 line
2.35.1
Acked-by Masami Hiramatsu (Google) on the v2 series
#6 lore raw
In reply to Steven Rostedt

Hi,

2022年10月12日(水) 11:45 Steven Rostedt <[email protected]>:

Quoted 11 lines

The follow commands caused a crash:

  # cd /sys/kernel/tracing
  # echo 's:open char file[]' > dynamic_events
  # echo 'hist:keys=common_pid:file=filename:onchange($file).trace(open,$file)' > events/syscalls/sys_enter_openat/trigger'
  # echo 1 > events/synthetic/open/enable

BOOM!

Quoted 8 lines

The problem is that the synthetic event field "char file[]" will read the value given to it as a string without any memory checks to make sure the address is valid. The above example will pass in the user space address and the sythetic event code will happily call strlen() on it and then strscpy() where either one will cause an oops when accessing user space addresses.

Changes since v1: https://lore.kernel.org/all/[email protected]/

Thanks, this series of patches looks good to me.

Thank you!

Quoted 15 lines

- Handle "(fault)" printing when there's a fault

Steven Rostedt (Google) (3):
tracing: Move duplicate code of trace_kprobe/eprobe.c into header
tracing: Add "(fault)" name injection to kernel probes
tracing: Fix reading strings from synthetic events

----

 kernel/trace/trace_eprobe.c       |  60 ++------------------
 kernel/trace/trace_events_synth.c |  23 ++++++--
 kernel/trace/trace_kprobe.c       |  60 ++------------------
 kernel/trace/trace_probe_kernel.h | 115 ++++++++++++++++++++++++++++++++++++++
 4 files changed, 142 insertions(+), 116 deletions(-)
 create mode 100644 kernel/trace/trace_probe_kernel.h
Signature · 2 lines
Masami Hiramatsu
mailto:[email protected]
Reviewed-by on the v2 series
#7 lore raw
In reply to Steven Rostedt

Hi Steve,

Quoted 20 lines · Steven Rostedt

The follow commands caused a crash:

  # cd /sys/kernel/tracing
  # echo 's:open char file[]' > dynamic_events
  # echo
'hist:keys=common_pid:file=filename:onchange($file).trace(open,$file)
' > events/syscalls/sys_enter_openat/trigger'
  # echo 1 > events/synthetic/open/enable

BOOM!

The problem is that the synthetic event field "char file[]" will read the value given to it as a string without any memory checks to make sure
the address is valid. The above example will pass in the user space address and the sythetic event code will happily call strlen() on it and then strscpy() where either one will cause an oops when accessing user space addresses.

Quoted 4 lines

Changes since v1:
https://lore.kernel.org/all/[email protected]/

 - Handle "(fault)" printing when there's a fault

Thanks for fixing the synthetic event string tracing bug, along with the other nice cleanup.

Quoted 15 lines

Steven Rostedt (Google) (3):
      tracing: Move duplicate code of trace_kprobe/eprobe.c into header
      tracing: Add "(fault)" name injection to kernel probes
      tracing: Fix reading strings from synthetic events

----
 kernel/trace/trace_eprobe.c       |  60 ++------------------ kernel/trace/trace_events_synth.c |  23 ++++++-- kernel/trace/trace_kprobe.c       |  60 ++------------------ kernel/trace/trace_probe_kernel.h | 115
++++++++++++++++++++++++++++++++++++++

 4 files changed, 142 insertions(+), 116 deletions(-)

 create mode 100644 kernel/trace/trace_probe_kernel.h