Variably-sized kernel memory stacks

• US 5,950,221 A
• Filed: 02/06/1997
• Issued: 09/07/1999
• Est. Priority Date: 02/06/1997
• Status: Expired due to Term

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Odinak, Gilad, Draves, Richard P., Cutshall, Scott
• Primary Examiner(s): Swann, Tod R.
• Assistant Examiner(s): Langjahr, David
• Application Number:
  US08/795,588
• Time in Patent Office: 943 Days
• Field of Search: 395/678, 395/677, 395/737, 711/6, 711/100, 711/202, 707/9
• US Class Current: 711/100
• CPC Class Codes: G06F 12/0223 User address space allocation, e.g. contiguous or non contiguous base addressing
• CPC Class Codes: G06F 9/5016 the resource being the memory
• CPC Class Codes: Y10S 707/99939 Privileged access

Abstract

The invention includes a computer system having a processor that executes program instructions in privileged and non-privileged execution modes. A user stack is used when the processor is executing in the non-privileged execution mode. A kernel memory stack is used when the processor is executing in the privileged execution mode. The kernel memory stack can grow and shrink dynamically as it is used by its associated thread, through the use of allocate-on-demand memory. A stack overflow handler is executed from within the kernel to resolve allocate-on-demand faults. The stack handler uses only fixed-size memory stacks.

Claims

1. In a computer system having privileged and non-privileged execution modes, a method comprising the following steps:

using a variably-sized kernel memory stack in the privileged execution mode;

executing a stack overflow handler that dynamically allocates additional memory for the variably-sized kernel memory stack;

using a fixed-size memory stack for the stack overflow handler.

2. A method as recited in claim 1, wherein the stack overflow handler dynamically allocates additional memory for the variably-sized kernel memory stack in response to allocate-on-demand memory faults.

3. A method as recited in claim 1, further comprising a step of initiating the stack overflow handler in response to memory faults.

4. A method as recited in claim 1, wherein the stack overflow handler comprises a fault handler and a helper thread, the method further comprising:

initiating the fault handler in response to memory faults;

the fault handler signaling the helper thread to allocate additional memory for the variably-sized kernel memory stack;

the helper thread allocating additional memory for the variably-sized kernel memory stack;

the fault handler and the helper thread using one or more fixed-size memory stacks.

5. A method as recited in claim 1, wherein the stack overflow handler comprises a helper thread, the method further comprising:

signaling the helper thread to allocate additional memory for the variably-sized kernel memory stack;

the helper thread allocating additional memory for the variably-sized kernel memory stack;

the helper thread using said fixed-size memory stack.

6. A method as recited in claim 1, wherein the stack overflow handler comprises a memory fault handler that uses the fixed-size memory stack.

7. A method as recited in claim 1, wherein the stack overflow handler comprises a helper thread that uses the fixed-size memory stack.

8. A method as recited in claim 1, wherein the stack overflow handler comprises a memory fault handler that uses a first fixed-size memory stack and a helper thread that uses a second fixed-size memory stack.

9. A method as recited in claim 1, wherein the steps execute on a microprocessor having a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode, and wherein the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault, the method comprising the following additional steps:

executing a kernel function in the level B execution mode, the kernel function using the variably-sized kernel memory stack;

switching to the level A execution mode in response to a memory fault that occurs while the kernel function is executing.

10. A method as recited in claim 1, wherein the steps execute on a microprocessor having a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode, and wherein the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault, the method comprising the following additional steps:

executing a kernel function in the level B execution mode, the kernel function using the variably-sized kernel memory stack;

switching to the level A execution mode in response to a memory fault that occurs while the kernel function is executing;

using the fixed-size memory stack when executing in the level A execution mode.

11. A method as recited in claim 1, further comprising a step of using a task gate to switch to the fixed-size memory stack in response to a memory fault that occurs while the kernel function is executing.

12. A method as recited in claim 1, and further comprising a step of periodically de-allocating unused memory from the variably-sized kernel memory stack.

13. A method as recited in claim 1, and further comprising a step of non-periodically de-allocating unused memory from the variably-sized kernel memory stack.

14. A method as recited in claim 1, and further comprising a step of de-allocating unused memory from the variably-sized kernel memory stack in response to running out of available physical memory.

15. A method as recited in claim 1, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, the method further comprising:

identifying critical sections of a kernel function during which additional memory cannot be allocated for the kernel memory stack;

probing the variably-sized kernel memory stack beyond the kernel memory stack pointer before entering said critical sections of the kernel function.

16. A method as recited in claim 1, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, the method further comprising:

identifying critical sections of a kernel function during which additional memory cannot be allocated for the variably-sized kernel memory stack;

before entering a critical section, advancing the stack pointer to a new memory page if required to ensure that there is sufficient allocated memory for the variably-sized kernel memory stack during the critical section;

writing to the variably-sized kernel memory stack at the location pointed to by the stack pointer.

17. A method as recited in claim 1, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, the method further comprising:

identifying critical sections of a kernel function in which additional memory cannot be allocated for the variably-sized kernel memory stack;

probing the variably-sized kernel memory stack beyond the kernel memory stack pointer before entering said critical sections of the kernel function;

omitting the probing step before entering a critical section that is nested within another critical section.

18. A method as recited in claim 1, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, the method further comprising:

identifying critical sections of a kernel function in which additional memory cannot be allocated for the variably-sized kernel memory stack;

incrementing a probe count when entering said critical sections of the kernel function;

decrementing the probe count when leaving said critical sections of the kernel function;

before entering a critical section of the kernel function, evaluating the probe count to determine whether the critical section is nested within another critical section;

probing the variably-sized kernel memory stack beyond the kernel memory stack pointer only before entering non-nested critical sections of the kernel function.

19. A method as recited in claim 1, wherein the stack overflow handler comprises a helper thread that uses said fixed-size memory stack, the method further comprising a step of initiating multiple helper threads to allocate memory for multiple variably-sized kernel memory stacks.

20. In a computer system having privileged and non-privileged execution modes, a method comprising the following steps:

using a variably-sized kernel memory stack in the privileged execution mode;

executing a stack overflow handler, the stack overflow handler allocating physical memory for the variably-sized kernel memory stack in response to accessing an address in the variably-sized kernel memory stack that has not been allocated;

switching to a fixed size memory stack while executing the stack overflow handler.

21. A method as recited in claim 20, wherein the stack overflow handler comprises a fault handler and a helper thread, the method further comprising:

initiating the fault handler in response to memory faults;

the fault handler detecting the accessing of an address that has not been allocated and in response signaling the helper thread to allocate additional memory for the variably-sized kernel memory stack;

the helper thread allocating additional memory for the variably-sized kernel memory stack;

the fault handler and the helper thread using one or more fixed-size memory stacks.

22. A method as recited in claim 20, wherein the stack overflow handler comprises a helper thread, the method further comprising:

signaling the helper thread to allocate additional memory for the variably-sized kernel memory stack;

the helper thread allocating additional memory for the variably-sized kernel memory stack;

the helper thread using said fixed-size memory stack.

23. A method as recited in claim 20, wherein the steps execute on a microprocessor having a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode, and wherein the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault, the method comprising the following additional steps:

executing a kernel function in the level B execution mode, the kernel function using the variably-sized kernel memory stack;

switching to the level A execution mode in response to a memory fault that occurs while the kernel function is executing.

24. A method as recited in claim 20, and further comprising a step of periodically de-allocating unused memory from the variably-sized kernel memory stack.

25. A method as recited in claim 20, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, the method further comprising:

identifying critical sections of a kernel function during which additional memory cannot be allocated for the variably-sized kernel memory stack;

probing the variably-sized kernel memory stack beyond the kernel memory stack pointer before entering said critical sections of the kernel function.

26. A method as recited in claim 20, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, the method further comprising:

identifying critical sections of a kernel function in which additional memory cannot be allocated for the variably-sized kernel memory stack;

probing the variably-sized kernel memory stack beyond the kernel memory stack pointer before entering said critical sections of the kernel function;

omitting the probing step before entering a critical section that is nested within another critical section.

27. In a computer system having privileged and non-privileged execution modes, a method comprising the following steps:

executing a user process or thread in the non-privileged execution mode, the user process or thread having a user memory stack;

invoking a kernel process or thread from the user process or thread, the kernel process or thread executing in the privileged execution mode, the kernel process or thread having a variably-sized kernel memory stack;

initiating a memory fault handler in response to memory faults;

the memory fault handler notifying a helper thread of memory faults that result from using out-of-bounds kernel memory stack addresses;

the helper thread allocating additional memory for the variably-sized kernel memory stack in response to being notified by the memory fault handler;

using one or more fixed-size memory stacks for the memory fault handler and the helper thread.

28. A method as recited in claim 27, wherein the memory fault handler uses a first fixed-size memory stack and the helper thread uses a second fixed-size memory stack.

29. A method as recited in claim 27, wherein the steps execute on a microprocessor having a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode, and wherein the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault, the method comprising the following additional steps:

executing a kernel function in the level B execution mode, said kernel function using the variably-sized kernel memory stack;

executing the memory fault handler in the level A execution mode;

switching to said one or more fixed-size memory stacks when executing in the level A execution mode.

30. A method as recited in claim 27, wherein the steps execute on a microprocessor having a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode, and wherein the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault, the method comprising the following additional steps:

executing a kernel function in the level B execution mode, the kernel function using the variably-sized kernel memory stack;

executing the memory fault handler in the level A execution mode;

the microprocessor automatically switching to said one or more fixed-size memory stacks when switching to the level A execution mode.

31. A computer system comprising:

a processor having privileged and non-privileged execution modes;

a user memory stack that is used when the processor is in the non-privileged execution mode;

a variably-sized kernel memory stack that is used when the processor is in the privileged execution mode;

a stack overflow handler that is executed by the processor to dynamically allocate additional memory for the variably-sized kernel memory stack;

at least one fixed-size memory stack that is used by the stack overflow handler.

32. A computer system as recited in claim 31, wherein the stack overflow handler comprises a memory fault handler that is initiated in response to memory faults.

33. A computer system as recited in claim 31, wherein the stack overflow handler comprises a helper thread that can be notified to allocate additional memory for the variably-sized kernel memory stack.

34. A computer system as recited in claim 31, wherein the stack overflow handler comprises a memory fault handler that is initiated in response to memory faults and a helper thread that is responsive to notifications from the memory fault handler to allocate additional memory for the variably-sized kernel memory stack;

the at least one fixed-size memory stack comprising a first fixed-size memory stack that is used by the memory fault handler and a second fixed-size memory stack that is used by the helper thread.

35. A computer system as recited in claim 31, further comprising:

a kernel memory stack pointer associated with the variably-sized kernel memory stack;

wherein a kernel function is configured to probe the variably-sized kernel memory stack beyond the kernel memory stack pointer before entering critical sections of the kernel function during which additional memory cannot be allocated for the variably-sized memory stack.

36. A computer system as recited in claim 31, further comprising:

to a kernel memory stack pointer associated with the variably-sized kernel memory stack;

a probe count that is incremented when entering critical sections of a kernel function during which additional memory cannot be allocated for the variably-sized kernel memory stack, the probe count being decremented when leaving said critical sections of the kernel function;

the probe count being used to avoid memory stack probing when a critical section of the kernel function is nested within another critical section of the kernel function.

37. A computer system as recited in claim 31, wherein the stack overflow handler comprises multiple helper threads.

38. A computer system comprising:

a processor having privileged and non-privileged execution modes;

a user process or thread having a user memory stack, the user process or thread executing in the non-privileged execution mode;

a kernel process or thread having a variably-sized kernel memory stack, the kernel process or thread executing in the privileged execution mode;

a memory fault handler that is initiated in response to memory faults;

a helper thread that is responsive to notifications from the memory fault handler to allocate additional memory for the variably-sized kernel memory stack;

a first fixed-size memory stack that is used by the memory fault handler; and

a second fixed-size memory stack that is used by the helper thread.

39. A computer system as recited in claim 38, wherein:

the processor has a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode;

the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault;

a kernel function is executed by the processor in the level B execution mode, the kernel function using the variably-sized kernel memory stack;

the memory fault handler is executed by the processor in the level A execution mode;

the microprocessor automatically switches to said one or more fixed-size memory stacks when switching to the level A execution mode.

40. A computer-readable medium having instructions for performing steps in conjunction with a computer having privileged and non-privileged execution modes, the steps comprising:

using a variably-sized kernel memory stack in the privileged execution mode;

executing a stack overflow handler that dynamically allocates additional memory for the variably-sized kernel memory stack;

using a fixed-size memory stack for the stack overflow handler.

41. A computer-readable medium as recited in claim 40, said instructions performing an additional step of initiating the stack overflow handler in response to memory faults.

42. A computer-readable medium as recited in claim 40, wherein the stack overflow handler comprises a fault handler and a helper thread, said instructions performing additional steps comprising:

initiating the fault handler in response to memory faults;

the fault handler signaling the helper thread to allocate additional memory for the variably-sized kernel memory stack;

the helper thread allocating additional memory for the variably-sized kernel memory stack;

the fault handler and the helper thread using one or more fixed-size memory stacks.

43. A computer-readable medium as recited in claim 40, wherein the stack overflow handler comprises a helper thread, said instructions performing additional steps comprising:

signaling the helper thread to allocate additional memory for the variably-sized kernel memory stack;

the helper thread allocating additional memory for the variably-sized kernel memory stack;

the helper thread using said fixed-size memory stack.

44. A computer-readable medium as recited in claim 40, wherein the stack overflow handler comprises a memory fault handler that uses the fixed-size memory stack.

45. A computer-readable medium as recited in claim 40, wherein the stack overflow handler comprises a helper thread that uses the fixed-size memory stack.

46. A computer-readable medium as recited in claim 40, wherein the stack overflow handler comprises a memory fault handler that uses a first fixed-size memory stack and a helper thread that uses a second fixed-size memory stack.

47. A computer-readable medium as recited in claim 40, wherein the instructions are for execution on a microprocessor having a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode, and wherein the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault, the instructions performing additional steps comprising:

executing a kernel function in the level B execution mode, the kernel function using the variably-sized kernel memory stack;

switching to the level A execution mode in response to a memory fault that occurs while the kernel function is executing.

48. A computer-readable medium as recited in claim 40, wherein the instructions are for execution on a microprocessor having a plurality of privileged execution modes including a level A execution mode and a level B execution mode, the level A execution mode allowing more privileges than the level B execution mode, and wherein the microprocessor when in the level A execution mode automatically pushes data to a currently active memory stack in response to a memory fault, the instructions performing additional steps comprising:

executing a kernel function in the level B execution mode, the kernel function using the variably-sized kernel memory stack;

switching to the level A execution mode in response to a memory fault that occurs while the kernel function is executing;

using the fixed-size memory stack when executing in the level A execution mode.

49. A computer-readable medium as recited in claim 40, said instructions performing an additional step of periodically de-allocating unused memory from the variably-sized kernel memory stack.

50. A computer-readable medium as recited in claim 40, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, said instructions performing additional steps comprising:

identifying critical sections of a kernel function during which additional memory cannot be allocated for the variably-sized kernel memory stack;

probing the kernel memory stack beyond the kernel memory stack pointer before entering said critical sections of the kernel function.

51. A computer-readable medium as recited in claim 40, wherein there is a kernel memory stack pointer associated with the variably-sized kernel memory stack, said instructions performing additional steps comprising:

identifying critical sections of a kernel function in which additional memory cannot be allocated for the variably-sized kernel memory stack;

probing the variably-sized kernel memory stack beyond the kernel memory stack pointer before entering said critical sections of the kernel function;

omitting the probing step before entering a critical section that is nested within another critical section.

52. A computer-readable medium as recited in claim 40, wherein there is a kernel memory stack pointer associated with the kernel memory stack, said instructions performing additional steps comprising:

identifying critical sections of a kernel function in which additional memory cannot be allocated for the kernel memory stack;

incrementing a probe count when entering said critical sections of the kernel function;

decrementing the probe count when leaving said critical sections of the kernel function;

before entering a critical section of the kernel function, evaluating the probe count to determine whether the critical section is nested within another critical section;

probing beyond the kernel memory stack pointer only before entering non-nested critical sections of the kernel function.

53. A computer-readable medium as recited in claim 40, wherein the stack overflow handler comprises a helper thread that uses said fixed-size memory stack, said instructions performing additional steps comprising a step of initiating multiple helper threads to allocate memory for multiple variably-sized kernel memory stacks.

54. In a computer system, an operating system having a non-privileged execution mode and a privileged execution mode, comprising:

a user process that executes in the non-privileged execution mode and invokes both privileged and non-privileged procedure calls;

a dynamically resizable user memory stack for storing data associated with an invocation of a non-privileged procedure call;

a kernel process that executes in the privileged execution mode and executes privileged procedure calls invoked by the user process;

a variably-sized kernel memory stack for storing data associated with the privileged procedure call, wherein the variably-sized kernel memory stack has an allocated size, wherein the allocated size increases dynamically in response to stack utilization.

55. In a computer system, and operating system as recited in claim 54, wherein the user and kernel processes are execution threads.

56. In a computer system, and operating system as recited in claim 54, further comprising:

a stack overflow handler that dynamically allocates additional memory for the variably-sized kernel memory stack;

at least one fixed-size memory stack that is used by the stack overflow handler.