Kideliverapc 64 Bit#
Windows 64 bit and Linux kernels use the table based exception processing to locate a handler for an instruction that caused an exception. The interesting difference between Windows and Linux kernels is in Windows mechanism to unwind a call stack, aka Frame Unwind. So when an interrupt processing code returnsĮxecution to an interrupted code the code resumes at the beginning ofĮxInterlockedPopEntrySList ( namely ExpInterlockedPopEntrySListResume ).Ĩ2acbdf1 0fb7416c movzx eax,word ptr Ĩ2acbdf5 8b5168 mov edx,dword ptr Ĩ2acbdfc 7511 jne nt!KiCheckForSListAddress+0x1e (82acbe0f)Ĩ2acbdfe b8f4dda882 mov eax,offset nt!ExpInterlockedPopEntrySListResumeĨ2acbe05 7222 jb nt!KiCheckForSListAddress+0x38 (82acbe29)Ĩ2acbe07 81fa1fdea882 cmp edx,offset nt!ExpInterlockedPopEntrySListEndĨ2acbe0d eb15 jmp nt!KiCheckForSListAddress+0x33 (82acbe24)Ĩ2acbe13 7514 jne nt!KiCheckForSListAddress+0x38 (82acbe29)Ĩ2acbe15 a1ac69bb82 mov eax,dword ptr [nt!KeUserPopEntrySListResumeĨ2acbe1c 720b jb nt!KiCheckForSListAddress+0x38 (82acbe29)Ĩ2acbe1e 3b15a469bb82 cmp edx,dword ptr [nt!KeUserPopEntrySListEndĨ2acbe24 7703 ja nt!KiCheckForSListAddress+0x38 (82acbe29)Ĩ2acbe26 894168 mov dword ptr ,eax X64) of a trap frame to restart SList pop operation if interrupt happened insideĮxInterlockedPopEntrySList. This routine is called atĭISPATCH_LEVEL before returning from an interrupt and it fixes the EIP(RIP for There is a routine KiCheckForSListAddress. I looked at the code again and found that interrupt processing code has a fixupįor SList. In for the ABA avoidance, but the primitives are in the Win32 API as well andĪny driver can employ custom pool allocator. I would have speculated that per-thread kernel allocator behavior was factored In the light of the above concerns, is SList on a pre-Windows 8 圆4ĭeployment really safe for all workloads? Is there some protection against hypervisor interference?ģ. Does the Windows scheduler protect against ABA by, e.g., restartingĢ. To further complicate, the kernelĬould run on a vcpu, creating time discontinuities.ġ.
Physical cores, this couldn't plausibly occur.
I wonder whether, depending on the number of threads and This means that 512 operations can completeĬoncurrently (without progress from particular thread) until an ABA problem To my knowledge, pre-Windows 8 圆4 implementations of SList use 9-bit sequence Safety of using ExInterlockedPopEntrySList
Kideliverapc drivers#
I believe this topic on ExInterlockedPopEntrySList might be interesting for Windows drivers developers.
0 kommentar(er)
