Automatically resource leak diagnosis and detecting process within the operating system

• US 7,774,741 B2
• Filed: 05/22/2006
• Issued: 08/10/2010
• Est. Priority Date: 05/22/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Sridharan, Baskar, Wohlgemuth, Jason Stewart, Heddaya, Abdelsalam A., Nallipogu, Elsie
• Primary Examiner(s): Khatri; Anil
• Application Number:
  US11/438,791
• Publication Number:
  US 20070271418A1
• Time in Patent Office: 1,541 Days
• Field of Search: 717100-102, 717124-132, 711/170
• US Class Current: 717/100
• CPC Class Codes: G06F 11/0748 in a remote unit communicating with a single-box computer node experiencing an error/fault remote testing G06F11/2294
• CPC Class Codes: G06F 11/0751 Error or fault detection not based on redundancy power supply failures G06F1/30; network fault management H04L41/06
• CPC Class Codes: G06F 11/0775 Content or structure details of the error report, e.g. specific table structure, specific error fields
• CPC Class Codes: G06F 12/023 Free address space management

Abstract

The resource leak detector attempts to identify the causes of resource leaks and fix the causes if possible. At a minimum, the located resource leaks are reported to an operating system manufacture for additional study and possible correction, either by the operating system manufacturer or the application manufacturer that is causing the resource leak. In an embodiment, in operation, a leak trigger for starting leak detection is started. If the leak trigger is activated, leak detection is started when an application begins. Resource allocations are tracked for the leaked resource and each leaked resource is matched to a corresponding allocation call stack. A type of the leaked resource may be identified. The leak detection and the leak trigger may be controlled according to at least one of a local policy and a global policy. The local and the global policy may be updated dynamically such as by using aggregated occurrences of applications that met the local policy or by matching a plurality of distinct types of leaked resources to the same call stack. A leak report may be generated of resource leak data where the leak report is stored locally or communicated to a local aggregation database or remote aggregation database. The leak report may include total leaked allocations, total size of leaked allocations, total outstanding allocations, total size of outstanding allocations and matched allocation call stacks. The leaked resource may be automatically reclaimed and a fix to the leaked resource may be created and applied automatically.

Claims

1. A method of determining a cause of a leaked resource in a computer system comprising a processor configured according to computer executable instructions and a memory physically configured to store computer executable instructions comprising physically configuring a processor for:

setting a leak trigger for starting leak detection;

if the leak trigger is activated;

starting leak detection when an application begins;

tracking resource allocations for the leaked resource; and

matching each leaked resource to a corresponding allocation call stack; and

if the leak trigger is not activated;

repeating the computer executable instructions;

generating a leak report of resource leak data wherein the leak report is stored at least one of locally or communicated to a local aggregation database or remote aggregation database; and

comprising at least one of automatically reclaiming the leaked resource and creating a fix to the leaked resource and applying the created fix automatically.

2. The method of claim 1, further comprising identifying a type of the leaked resource.

3. The method of claim 1, further comprising controlling at least one of the leak detection and the leak trigger according to at least one of a local policy and a global policy.

4. The method of claim 3, further comprising dynamically updating at least one of the local and the global policy.

5. The method of claim 3, further comprising updating the global policy using aggregated occurrences of applications that met the local policy.

6. The method of claim 1, further comprising matching a plurality of distinct types of leaked resources to the same call stack.

7. The method of claim 1, wherein the leak report comprises:

total leaked allocations;

total size of leaked allocations;

total outstanding allocations;

total size of outstanding allocations; and

matched allocation call stacks.

8. The method of claim 1, further comprising using the leak detector during code testing.

9. The method of claim 1, further comprising setting a global policy to enable leak detection comprising:

using a first communication channel to control whether to invoke a leak detector; and

using a second communication channel to communicate a leak report where the leak report is directed to the proper classification at a receiver of the leak report.

10. A method of analyzing leak report data for a computer system comprising a processor configured according to computer executable instructions and a memory physically configured to store computer executable instructions comprising physically configuring a processor for:

aggregating leak reports for the same component across at least one of multiple applications, users and machines;

providing the leak reports to developers of the applications;

comparing similar call stacks across reports from multiple users; and

performing data flow analysis on the source code identified in the stacks to identify the leaked local variables;

prioritizing leak reports based on at least one of the factors comprising application usage, application install base, size of the leak, number of components that use the targeted component, the number of leaked resource, the type of the leaked resource, the frequency of occurrence of the leaked resources.

11. The method of claim 10, further comprising performing control flow analysis to identify the leaked global variables.

12. A method of creating an algorithm to limit performance costs of a leak detector in a computer system comprising a processor configured according to computer executable instructions and a memory physically configured to store computer executable instructions comprising physically configuring a processor for:

limiting usage of computing overhead;

maintaining a unique call stack database and all associated data structures in a heap and the heap is maintained in the target process'"'"'s address space; and

storing in the call stack database only call stacks of outstanding blocks;

generating a leak report of resource leak data wherein the leak report is stored at least one of locally or communicated to a local aggregation database or remote aggregation database; and

comprising at least one of automatically reclaiming the leaked resource and creating a fix to the leaked resource and applying the created fix automatically.

13. The method of claim 12, wherein computing overhead comprises limiting at least one of CPU usage, memory usage and network usage.

14. The method of claim 12, further comprising limiting network usage by controlling the size of the files communicated by the method.

15. The method of claim 12, further comprising controlling CPU usage of the leak detector by setting a CPU usage quota by:

determining a maximum CPU cycles based on CPU frequency and computing CPU cycles for each stack tracing operation and deducted from the quota for the current time interval; and

computing the quota of CPU cycles to be spent for stack tracing in a fixed time interval.

16. The method of claim 12, further comprising limiting memory by limiting memory usage to a maximum amount.

17. The method of claim 12, further comprising when a resource is allocated or freed, the hash H of the resource identifier RID is computed using the formula:

H=SUM byte[k]*random[k] mod N(Σ

_k=0<sup>n−

1</sup>byte[k]*random[k]) % N,

where byte[k] is the kth byte of the RID,

random[k] is the kth slot in the random number table, and

N is size of S, chosen as a prime number.