Skip to contentSkip to conversation

firewire: nosy: don't read packets bigger than requested

PATCH In review Jann Horn 3 participants 4 messages Jul 6, 2018 – Sep 18, 2018 Firewire

lore
Share & tools
lore

Jann Horn

author
#0 lore raw

PATCH 1/1 firewire: nosy: don't read packets bigger than requested

In general, accessing userspace memory beyond the length of the supplied buffer in VFS read/write handlers can lead to both kernel memory corruption (via kernel_read()/kernel_write(), which can e.g. be triggered via sys_splice()) and privilege escalation inside userspace.

Fixes 286468210d83 ("firewire: new driver: nosy - IEEE 1394 traffic sniffer")

---
No CC stable because this device shouldn't be available to unprivileged code by default and should be pretty rare.

 drivers/firewire/nosy.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/firewire/nosy.c +3 -2
index a128dd1126ae..732075fc312e 100644
@@ -161,11 +161,12 @@ packet_buffer_get(struct client *client, char __user *data, size_t user_length)
 	if (atomic_read(&buffer->size) == 0)
 		return -ENODEV;
 
-	/* FIXME: Check length <= user_length. */
-
 	end = buffer->data + buffer->capacity;
 	length = buffer->head->length;
 
+	if (length > user_length)
+		return -EINVAL;
+
 	if (&buffer->head->data[length] < end) {
 		if (copy_to_user(data, buffer->head->data, length))
 			return -EFAULT;
Signature · 1 line
2.18.0.399.gad0ab374a1-goog
author
#1 lore raw
Quoted 23 lines · Jann Horn

In general, accessing userspace memory beyond the length of the supplied buffer in VFS read/write handlers can lead to both kernel memory corruption (via kernel_read()/kernel_write(), which can e.g. be triggered via sys_splice()) and privilege escalation inside userspace.

Fixes 286468210d83 ("firewire: new driver: nosy - IEEE 1394 traffic sniffer")

---
No CC stable because this device shouldn't be available to unprivileged code by default and should be pretty rare.

 drivers/firewire/nosy.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/firewire/nosy.c +0 -1
index a128dd1126ae..732075fc312e 100644
@@ -161,11 +161,12 @@ packet_buffer_get(struct client *client, char __user *data, size_t user_length)
        if (atomic_read(&buffer->size) == 0)
                return -ENODEV;
 
-       /* FIXME: Check length <= user_length. */

drivers/firewire/nosy.c packet_buffer_get(struct client *client, char __user *data, size_t user_length) · line 172

Quoted 10 lines
drivers/firewire/nosy.c +3 -1
-
        end = buffer->data + buffer->capacity;
        length = buffer->head->length;
 
+       if (length > user_length)
+               return -EINVAL;
+
        if (&buffer->head->data[length] < end) {
                if (copy_to_user(data, buffer->head->data, length))
                        return -EFAULT;

Ping. I sent this about two months ago, I haven't received a reply, and from what I can tell, it hasn't landed in any tree so far...

#2 lore raw
Quoted 28 lines · Jann Horn
Quoted 27 lines · Jann Horn

In general, accessing userspace memory beyond the length of the supplied buffer in VFS read/write handlers can lead to both kernel memory corruption (via kernel_read()/kernel_write(), which can e.g. be triggered via sys_splice()) and privilege escalation inside userspace.

Fixes 286468210d83 ("firewire: new driver: nosy - IEEE 1394 traffic sniffer")

---
No CC stable because this device shouldn't be available to unprivileged code by default and should be pretty rare.

 drivers/firewire/nosy.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/firewire/nosy.c +0 -2
index a128dd1126ae..732075fc312e 100644
@@ -161,11 +161,12 @@ packet_buffer_get(struct client *client, char __user *data, size_t user_length)
        if (atomic_read(&buffer->size) == 0)
                return -ENODEV;
 
-       /* FIXME: Check length <= user_length. */
-
        end = buffer->data + buffer->capacity;
        length = buffer->head->length;
 
Quoted 10 lines

drivers/firewire/nosy.c packet_buffer_get(struct client *client, char __user *data, size_t user_length) · line 172

Quoted 6 lines
drivers/firewire/nosy.c +3 -0
+       if (length > user_length)
+               return -EINVAL;
+
        if (&buffer->head->data[length] < end) {
                if (copy_to_user(data, buffer->head->data, length))
                        return -EFAULT;

Ping. I sent this about two months ago, I haven't received a reply, and from what I can tell, it hasn't landed in any tree so far...

:(
I have that same problem with some Firewire documentation patches. I plan to ask someone else to merge my patches.

Signature · 1 line
~Randy
#3 lore raw
Quoted 9 lines · On Sep 03 Randy Dunlap wrote:
Quoted 8 lines · Jann Horn
Quoted 7 lines · Jann Horn

In general, accessing userspace memory beyond the length of the supplied buffer in VFS read/write handlers can lead to both kernel memory corruption (via kernel_read()/kernel_write(), which can e.g. be triggered via sys_splice()) and privilege escalation inside userspace.

Fixes 286468210d83 ("firewire: new driver: nosy - IEEE 1394 traffic sniffer")

[...]

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
 drivers/firewire/nosy.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

[...]

Quoted 7 lines
Quoted 3 lines

Ping. I sent this about two months ago, I haven't received a reply, and from what I can tell, it hasn't landed in any tree so far...

:(
I have that same problem with some Firewire documentation patches. I plan to ask someone else to merge my patches.

Jann,

sorry for not responding in July (was buried in other work and been effectively absent from maintainership for many months). And sorry for missing your ping in September (it must have been misplaced into the spam folder and I apparently overlooked it there).

This week is another one in which I will not be able to check your patch. Next week I will have a vacation of sorts and will use it to (a) review and merge your patch and (b) clean out my mailbox and update my mail sorting filters (long overdue after my mail service provider changed backends).

Again sorry, and thank you for your extraordinary patience.

Signature · 3 lines
Stefan Richter
-======---=- =--= =--=-
http://arcgraph.de/sr/