!sysret, !sysret2 (hook SYSRET instruction execution)
Description of '!sysret, !sysret2' commands in HyperDbg.
Command
!sysret
!sysret2
Syntax
!sysret [pid ProcessId (hex)] [core CoreId (hex)] [imm IsImmediate (yesno)] [sc EnableShortCircuiting (onoff)] [stage CallingStage (prepostall)] [buffer PreAllocatedBuffer (hex)] [script { Script (string) }] [condition { Condition (hex) }] [code { Code (hex) }] [output {OutputName (string)}]
!sysret2 [pid ProcessId (hex)] [core CoreId (hex)] [imm IsImmediate (yesno)] [sc EnableShortCircuiting (onoff)] [stage CallingStage (prepostall)] [buffer PreAllocatedBuffer (hex)] [script { Script (string) }] [condition { Condition (hex) }] [code { Code (hex) }] [output {OutputName (string)}]
Description
Triggers when the debugging machine executes a sysret instruction or, in other words, when Windows tries to return to user-mode from a previous syscall.
When you enable this event, all sysret instructions from all processes will be monitored, and due to the limitation in hardware, you can't expect it to trigger for just one process. Still, you can configure the debugger to trigger the event for you in the case of a special process by adding pid xxto the command.
The difference between !sysret and !sysret2 is that we safely check the memory in the first command to see if the instruction that caused #UD is really an SYSRET or a SYSCALL. So, we access the memory in this command. However, we realized that older systems have problems with this way of memory access. In the second command, we just check for the RIP to see if it's a kernel address or a user address. Usually, this method works without error for several hours, but if one application generates a #UD, then a BSOD will happen. The second method is generally faster in speed, but we encourage you to use the first command and if your computer doesn't support the first command, then use the second command.
Parameters
[pid ProcessId (hex)] (optional)
Optional value to trigger the event in just a specific process. Add pid xx to your command; thus, the command will be executed if the process id is equal to xx. If you don't specify this option, then by default, you receive events on all processes.
Still, in the case of user-mode debugging, HyperDbg will apply it only to the current active debugging process (not all the processes). In that case, you can specify pid all to intercept events from the entire system.
[core CoreId (hex)] (optional)
Optional value to trigger the event in just a specific core. Add core xx to your command thus command will be executed if core id is equal to xx. If you don't specify this option, then by default, you receive events on all cores.
[imm IsImmediate (yesno)] (optional)
Optional value in which yes means the results (printed texts in scripts) should be delivered immediately to the debugger. no means that the results can be accumulated and delivered as a couple of messages when the buffer is full; thus, it's substantially faster, but it's not real-time. By default, this value is set to yes.
[sc EnableShortCircuiting (onoff)] (optional)
Optional value to ignore the emulation (skip execution) of the event. Add sc on to your command thus whenever the event is triggered, the effects and the execution of the actual event will be ignored. For more information, please read this article. If you don't specify this option, then by default, all the events will be emulated (executed). By default, this value is set to off.
[stage CallingStage (prepostall)] (optional)
Optional value to configure the calling stage of the event. To trigger the event before the emulation, include stage pre in your command. Conversely, using stage post will cause the event to be triggered after the emulation. Additionally, using stage all will trigger the event both before and after the emulation. For more information, please read this article. By default, this value is set to pre.
[buffer PreAllocatedBuffer (hex)] (optional)
Optional value which reserves a safe pre-allocated buffer to be accessed within the event codes.
[script { Script (string) }] (optional)
A HyperDbg script will be executed each time the event is triggered.
[condition { Condition (hex) }] (optional)
Optional hex assembly codes which check for conditions in assembly.
[code { Code (hex) }] (optional)
Optional hex assembly codes will be executed each time the event is triggered.
[output {OutputName (string)}] (optional)
Optional output resource name for forwarding events.
Context
As the Context ($context pseudo-register in the event's script, r8 in custom code, and rdx in condition code register) to the event trigger, HyperDbg sends the rip register of where executes the sysret instruction. Generally, it should be the same in value in Windows (just one sysret instruction is in Windows).
Short-circuiting
This event supports 'event short-circuiting', which means that you can configure HyperDbg to ignore its execution and its effects. For additional details, please refer to the article provided here.
Calling Stages
This event supports different calling stages. The 'pre' calling stage is triggered prior to the user transfer, whereas the 'post' calling stage is triggered subsequent to the user transfer. In addition, the 'all' calling stage will trigger the event in both cases. You can ignore the event in the 'pre' stage. For more information, please refer to the article provided here.
Debugger
This event supports three debugging mechanisms.
Break
Script
Custom Code
Please read "How to create a condition?" if you need a conditional event, a conditional event can be used in all "Break", "Script", and "Custom Code".
Break
You can use condition if you want to check the sysret parameters, for example, you can check the result of a previous syscall which now wants to return to user-mode by executing sysret instruction to see if it matches to your debugging logic or not and have a conditional sysret hooker and also you can change the parameters of a sysret.
Imagine we want to break on all sysret executions of a process id 0x490.
Script
Using the following command, you can use HyperDbg's Script Engine. You should replace the string between braces (HyperDbg Script Here) with your script. You can find script examples here.
The above command when messages don't need to be delivered immediately.
Script (From File)
If you saved your script into a file, then you can add file: instead of a script and append the file path to it. For example, the following examples show how you can run a script from file:c:\users\sina\desktop\script.txt.
You can use event forwarding to forward the event monitoring results from this event and other events to an external source, e.g., File, NamedPipe, or TCP Socket. This way, you can use HyperDbg as a monitoring tool and gather your target system's behavior and use it later or analyze it on other systems.
Custom Code
Please read "How to create an action?" to get an idea about how to run the custom buffer code in HyperDbg.
Your custom code will be executed in vmx-root mode. Take a look at this topic for more information. Running code in vmx-root is considered "unsafe".
Run Custom Code (Unconditional)
Monitoring process id 0x490 for sysret instruction execution and run 3 nops whenever the event is triggered. Take a look at Run Custom Code for more information.
Run Custom Code (Conditional)
Monitoring process ID 0x490 for sysret instruction execution and run 3 nops whenever the event condition is triggered and run 3 nops whenever the event is triggered. Take a look at Run Custom Code and how to create a condition for more information.
Keep in mind that a conditional event can be used in Breaking to Debugger and Running Script too.
IOCTL
This command uses the same method to send IOCTL for regular events.
As EventType use SYSCALL_HOOK_EFER_SYSRET in DEBUGGER_GENERAL_EVENT_DETAIL.
Design
Take a look at "Design of !syscall & !sysret" to see how it works.
Remarks
This command is not PatchGurad compatible, which means that PatchGuard detects this command and will cause BSOD; thus, make sure to turn it off (e.g., attaching a kernel-mode WinDbg debugger at the start of the Windows) before using this command. Disabling Driver Signature Enforcement alone won't turn off the PatchGuard.
This command makes your computer substantially slower.
This command creates an event. Starting from HyperDbg v0.7, events are guaranteed to keep the debuggee in a halt state (in the Debugger Mode); thus, nothing will change during its execution and the context (registers and memory) remain untouched. You can visit instant events for more information.
Requirements
Post-Nehalem Processor (EPT)
Related
Windows X86-64 System Call Table (XP/2003/Vista/2008/7/2012/8/10)
Last updated