Skip to contentSkip to conversation

sim: validate IMS private identity

PATCH Superseded Sergey Matyukevich 2 participants 5 messages Jan 15, 2021

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 sim: validate IMS private identity

Make sure that IMS private identity is a valid UTF8 string before setting sim->impi field. Otherwise ofono may crash on dbus assert when SIM properties are reported via org.ofono.SimManager interface.

---
 src/sim.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
src/sim.c +2 -1
index 33e1245f..2a663e2d 100644
@@ -1664,7 +1664,8 @@ static void impi_read_cb(int ok, int total_length, int record,
 		return;
 	}
 
-	sim->impi = g_strndup((const char *)data + 2, data[1]);
+	if (g_utf8_validate((const char *)data + 2, data[1], NULL))
+		sim->impi = g_strndup((const char *)data + 2, data[1]);
 }
 
 static void discover_apps_cb(const struct ofono_error *error,
#1 lore raw

Hi Sergey,

Quoted 12 lines · Sergey Matyukevich

Make sure that IMS private identity is a valid UTF8 string before setting sim->impi field. Otherwise ofono may crash on dbus assert when SIM properties are reported via org.ofono.SimManager interface.

---
  src/sim.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
src/sim.c +0 -0
index 33e1245f..2a663e2d 100644

src/sim.c static void impi_read_cb(int ok, int total_length, int record, · line 1668

Quoted 6 lines
src/sim.c +2 -1
@@ -1664,7 +1664,8 @@ static void impi_read_cb(int ok, int total_length, int record,
  		return;
  	}
 
-	sim->impi = g_strndup((const char *)data + 2, data[1]);
+	if (g_utf8_validate((const char *)data + 2, data[1], NULL))
+		sim->impi = g_strndup((const char *)data + 2, data[1]);

I assume this code path was tested with a file containing embedded NULs as that is the only way it would have worked.

glib docs [1] say:
"Note that g_utf8_validate() returns FALSE if max_len is positive and any of the max_len bytes are nul."

So I think the above logic would flag such a file as invalid, no?

Quoted 7 lines
src/sim.c +0 -0
  }
 
  static void discover_apps_cb(const struct ofono_error *error,

_______________________________________________
ofono mailing list -- ofono(a)ofono.org
To unsubscribe send an email to ofono-leave(a)ofono.org

Regards,
-Denis

[1]
https://developer.gnome.org/glib/stable/glib-Unicode-Manipulation.html#g-utf8-validate

#2 lore raw
Quoted 21 lines
Quoted 12 lines

Make sure that IMS private identity is a valid UTF8 string before setting sim->impi field. Otherwise ofono may crash on dbus assert when SIM properties are reported via org.ofono.SimManager interface.

---
  src/sim.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
src/sim.c +0 -0
index 33e1245f..2a663e2d 100644

src/sim.c static void impi_read_cb(int ok, int total_length, int record, · line 1667

Quoted 5 lines
src/sim.c +2 -1
@@ -1664,7 +1664,8 @@ static void impi_read_cb(int ok, int total_length, int record,
  		return;
  	}
-	sim->impi = g_strndup((const char *)data + 2, data[1]);
+	if (g_utf8_validate((const char *)data + 2, data[1], NULL))
+		sim->impi = g_strndup((const char *)data + 2, data[1]);

I assume this code path was tested with a file containing embedded NULs as that is the only way it would have worked.

Quoted 5 lines

glib docs [1] say:
"Note that g_utf8_validate() returns FALSE if max_len is positive and any of the max_len bytes are nul."

So I think the above logic would flag such a file as invalid, no?

No, I tested using modem with attached SIM/eSIM. TLV data object appears to be well-formed, but the contents is all padding 0xff bytes. Could you please clarify your concern ? I assume we can not rely on the content being properly null terminated string.

Regards,
Sergey

#3 lore raw

Hi Sergey,

Quoted 7 lines · Sergey Matyukevich
Quoted 7 lines
Quoted 7 lines

Make sure that IMS private identity is a valid UTF8 string before setting sim->impi field. Otherwise ofono may crash on dbus assert when SIM properties are reported via org.ofono.SimManager interface.

---
   src/sim.c | 3 ++-
   1 file changed, 2 insertions(+), 1 deletion(-)
Quoted 13 lines
Quoted 5 lines
Quoted 5 lines
src/sim.c +0 -0
index 33e1245f..2a663e2d 100644
@@ -1664,7 +1664,8 @@ static void impi_read_cb(int ok, int total_length, int record,
Quoted 8 lines

src/sim.c static void impi_read_cb(int ok, int total_length, int record, · line 1668

Quoted 5 lines
src/sim.c +2 -1
   		return;
   	}
-	sim->impi = g_strndup((const char *)data + 2, data[1]);
+	if (g_utf8_validate((const char *)data + 2, data[1], NULL))
+		sim->impi = g_strndup((const char *)data + 2, data[1]);

I assume this code path was tested with a file containing embedded NULs as that is the only way it would have worked.

Ignore the last part of the above sentence. What I'm trying to say is that:

We in theory have two possibilities:

1. file with a string 'foo', no null:
0x80 0x03 'f' 'o' 'o'

2. file with a string 'foo' and null:
0x80 0x04 'f' 'o' 'o'

I suspect the spec really wants 1, but maybe it can be interpreted that 2 is also a possibility?

The present logic should work for either of the above, but not what you have, i.e.:

0x80 0x03 0xff 0xff 0xff

Quoted 7 lines
Quoted 6 lines

glib docs [1] say:
"Note that g_utf8_validate() returns FALSE if max_len is positive and any of the max_len bytes are nul."

So I think the above logic would flag such a file as invalid, no?

...but g_utf8_validate as invoked in this patch would flag possibility 2 as invalid...

Quoted 5 lines

No, I tested using modem with attached SIM/eSIM. TLV data object appears to be well-formed, but the contents is all padding 0xff bytes. Could you please clarify your concern ? I assume we can not rely on the content being properly null terminated string.

Yeah, for the case you have this logic works just fine.

Regards,
-Denis

#4 lore raw

Hi Denis,

Quoted 43 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Make sure that IMS private identity is a valid UTF8 string before setting sim->impi field. Otherwise ofono may crash on dbus assert when SIM properties are reported via org.ofono.SimManager interface.

---
   src/sim.c | 3 ++-
   1 file changed, 2 insertions(+), 1 deletion(-)
Quoted 13 lines
Quoted 5 lines
Quoted 5 lines
src/sim.c +0 -0
index 33e1245f..2a663e2d 100644
@@ -1664,7 +1664,8 @@ static void impi_read_cb(int ok, int total_length, int record,
Quoted 8 lines

src/sim.c static void impi_read_cb(int ok, int total_length, int record, · line 1668

Quoted 5 lines
src/sim.c +2 -1
   		return;
   	}
-	sim->impi = g_strndup((const char *)data + 2, data[1]);
+	if (g_utf8_validate((const char *)data + 2, data[1], NULL))
+		sim->impi = g_strndup((const char *)data + 2, data[1]);

I assume this code path was tested with a file containing embedded NULs as that is the only way it would have worked.

Ignore the last part of the above sentence. What I'm trying to say is that:

We in theory have two possibilities:

1. file with a string 'foo', no null:
0x80 0x03 'f' 'o' 'o'

2. file with a string 'foo' and null:
0x80 0x04 'f' 'o' 'o'

I suspect the spec really wants 1, but maybe it can be interpreted that 2 is also a possibility?

The present logic should work for either of the above, but not what you have, i.e.:

0x80 0x03 0xff 0xff 0xff

Quoted 5 lines
Quoted 5 lines

glib docs [1] say:
"Note that g_utf8_validate() returns FALSE if max_len is positive and any of the max_len bytes are nul."

Quoted 5 lines
Quoted 2 lines
Quoted 1 line

So I think the above logic would flag such a file as invalid, no?

...but g_utf8_validate as invoked in this patch would flag possibility 2 as invalid...

True. Thanks for detailed clarification. Indeed, both cases needs to be supported. Let me double-check and come back with v3.

Regards,
Sergey