usb: serialize transfers against close to fix use-after-free - #89
Open
amaanq wants to merge 1 commit into
Open
Conversation
Owner
|
Thanks for digging into this. I don't like that the release ends up waiting on a transfer though, and moving that wait into a coroutine means it now happens after |
Owner
|
@amaanq I believe that the correct fix for this issue would be something like Please test it and feel free to adjust it. |
Author
|
Thanks for the quick reply. I've played with your patch a bit, and it doesn't actually fix the UAF, but rather it just shifts the crash site, that being a segv in The reason your version still crashes is that // frameworks/base/core/jni/android_hardware_UsbDeviceConnection.cpp
struct usb_device* device = get_device_from_object(env, thiz);
if (!device) {
ALOGE("device is closed in native_request_wait");
return NULL;
}
// ...
request = usb_request_wait(device, timeout);And close() frees before it nulls the field: struct usb_device* device = get_device_from_object(env, thiz);
if (device) {
usb_device_close(device);
env->SetLongField(thiz, field_context, 0);
}Meaning, any thread that already got past the null check is holding a pointer that gets freed under it. I do agree not deferring Please take a look and let me know your thoughts |
Owner
|
Have you tried just adding a cancel before the close() in the block? |
Author
|
Yeah I tried that, it still crashes at Tombstone |
Owner
|
Sorry. I meant |
Author
I just tried this as well, and it still crashes unfortunately. |
Closing a transport mid-transfer frees the native `usb_device` while the transfer is still using it, and neither `bulkTransfer` nor `requestWait` checks that the connection is still open once it has read the pointer. I hit this on my phone with MTE enabled, where I ran into a SIGSEGV as soon as the freed struct was read. Each transfer now runs under a lock, and `close()` cancels the in-flight one before waiting on that same lock, so everything is released by the time `close()` returns. Co-authored-by: mimi89999 <michel@lebihan.pl>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closing a transport mid-transfer frees the native
usb_devicewhile thetransfer is still using it, and neither
bulkTransfernorrequestWaitchecksthat the connection is still open once it has read the pointer. I hit this on my
phone with MTE enabled, where I ran into a SIGSEGV as soon as the freed struct
was read.
sendRawnow holds a lock across the whole exchange, andclose()cancels thein-flight transfer before waiting on that same lock, so everything is released by
the time
close()returns.I've been running this patch on my phone now and have not run into any more MTE issues. Please note that I did use an LLM to help me diagnose this and propose a fix, but I've reviewed the code myself & tested it locally with a reproducer that trigger the UAF reliably, and this patch no longer triggers it.
Full tombstone
Thanks