Patent Application
20070294601
The systems and methods for watchdog processors in multicore systems in accordance with the present invention are further described with reference to the accompanying drawings in which:
Certain specific details are set forth in the following description and figures to provide a thorough understanding of various embodiments of the invention. Certain well-known details often associated with computing and software technology are not set forth in the following disclosure, however, to avoid unnecessarily obscuring the various embodiments of the invention. Further, those of ordinary skill in the relevant art will understand that they can practice other embodiments of the invention without one or more of the details described below. Finally, while various methods are described with reference to steps and sequences in the following disclosure, the description as such is for providing a clear implementation of embodiments of the invention, and the steps and sequences of steps should not be taken as required to practice this invention.
The concept of a watch-dog processor as a processor that performs a bus monitoring function is generally understood in the art. Watchdog processors have been used to monitor Input/Output (I/O) behavior and generate exceptions upon certain data values on a monitored bus. Watch-dog processors have not however previously been deployed in integrated multicore computer chips or otherwise integrated into multiprocessor processing systems as described herein, nor have watchdog processors been utilized according to the various techniques described herein. As opposed to monitoring a given data bus, in one aspect of the invention a watch-dog processor may be dedicated to a specific processor or group of processors. The watchdog can behave as a parasite that monitors the behavior of its target.
In one embodiment, a watch-dog processor may be used for security and specific performance applications in multicore systems. One objective of such watchdog processors is to observe the behavior of a given application and respond to unexpected behavior or report on certain performance parameters. As the rate of decrease in sizes of semiconductor features moderates in coming years, an increase of on-chip latency in super-linear manner with respect to interconnect length will emerge. As a result, computer chips are increasingly built as a network of relatively small functional units, cores, connected via a networking structure that comprises buses, routers, and relays.
Processes such as firewalls, malware scanners, device drivers, and peer-to-peer networking handlers can be executed on separate processors with dedicated or shared memory and with optimized datapaths. For example, a 100-million transistor processor can pack 3450 i8086 or 18 Pentium P6 processors; obviously a substantial computational power at high frequency clocks that is hard to equal by context switching a large number of processes and/or exploring better instruction level parallelism of individual threads using extreme pipelining or superscalar units but at low frequency clocks.
It will be appreciated that a multicore computer chip 200 such as that of
Components of chip 200 may be grouped into functional groups. For example, shared memory 203, caches 230, main CPU 210, crypto processor 240, watchdog processor 250, and key storage 295 may be components of a first functional unit. Such grouping is not necessary to practice the invention but will clarify the description by reducing the subset of components that must be discussed to describe operation of an exemplary watchdog processor as contemplated herein. Aspects of an exemplary functional group of a processor are illustrated in
At least one wire connection 305 can run between the watchdog 310 and an internal component 325 of a chip component such as the second processor 320. Additional wire connections may run to further internal components of the second processor 320, and additional wire connections may run to internal components of other chip components such as 330, 351-355, 360, etc. The watchdog 310 can monitor information in an internal component 325 via said wire connection 305.
While
Internal component 325 may be one or more of a variety of internal processor components that allow monitoring of behaviors which are generally known and appreciated by those of skill in the art. Candidate values and events to monitor are:
System calls/Child processes. The application or the user can set up policies for the watchdog 310 to obey when detecting suspicious behavior. Upon detection, both the parent and the child process (or system call) can be terminated or paused. For example, the watchdog 310 may detect a system call that launches a command shell; if unexpected, such a call is commonly a sign of system intrusion.
Program counter. For a given program, a specialist and/or a secure automated analysis mechanism may first identify all addresses to which a jump/branch or call/return instruction can go to at compilation time. Thus, any inconsistency with these addresses during program execution can be identified as a bug or intrusion.
Pointer access. The watchdog 310 could verify each pointer access against a heap map. It could build the heap map in parallel while the second processor 320 allocates memory. Each read access to uninitialized memory or deallocation of an uninitialized pointer could be identified by the watchdog 310.
To monitor the various values and events exemplified by those set forth above, watchdog 310 may monitor internal components such as a register, a stack pointer table, and a virtual memory table. Finally, other system parameters such as I/O behavior, detailed page fault statistics, communication to other processes/cores, etc. can be also detected and analyzed by the watch-dog 310. The results can be in this case served as application performance and communication profile to the operating system which could use it for optimized process-to-core assignment and scheduling.
It should be noted that in some configurations, watchdog 310 may be configured to monitor behaviors of functional groups 351-355 as a whole rather than the internal operations of a particular functional group 353. In such embodiments, a wire connection such as 305 may link watchdog 310 to an internal component of one or more of the functional groups 351-355, while a bus akin to 300 connects watchdog 310 to the various functional groups 351-355 instead of or in addition to processors 320 and 330.
In this embodiment, it may be useful for watchdog 310 to monitor many of the same behaviors as when watchdog 310 is monitoring internal component 305 of processor 320. For example, system calls/child processes, program counters, and pointer access pertaining to interactions between second processor 320 and third processor 330 are beneficially monitored by watchdog 310 by monitoring bus 300. Embodiments may further beneficially combine monitoring of an internal component 325 with monitoring processor interactions to achieve robust and effective multicore monitoring capabilities.
The watchdog 310 may be configured to enforce an interaction policy against the second processor 320 and/or third processor 330. Such a configuration can be understood with reference to
In one embodiment, an OS 470 may dictate a basic interaction policy 472 to be applied regardless of the process or particular application functions that may be running on second processor 320 or third processor 330. Similarly, a basic policy such as 480 may be universally applied regardless of processes running on second processor 420 or third processor 430. This is not to say that policies 472 and 480 coming from an OS 470 or from hardware 450 may not be conditionally applied. Various useful configurations of the invention may utilize, not utilize, or conditionally utilize hardware and OS processor policies as desired.
Additionally, an application 462 may comprise a processor interaction policy 464 in addition to the application functions 463 supplied with the application 464. Like other processor policies, the application processor policy 464 may provide either independent or supplemental interaction policies to be enforced when the one or more processes associated with the application are executing on a processor 420. For example, some applications may dictate that no system calls may be made and no child processes may be spawned during execution of a particular process. The watchdog 410 may accordingly watch for such a behavior when one of the processors 420 is executing such a process.
An interaction policy may further dictate what action is to be taken by the watchdog 410 when a disapproved behavior occurs. In one embodiment, the watchdog 410 may freeze the processor 420 or 430 that exhibited the disapproved behavior. Other less or more drastic measures are also available depending on a level of security that is desired.
A processor policy may evolve dynamically. Behaviors of a complex multiprocessor system such as chip 450 may be difficult to fully understand, even by experts in the field. In this regard, intelligent logic may be placed in an OS 470 or on chip 450 to learn over time which processor interaction behaviors are considered normal or, conversely, which interaction behaviors are considered abnormal. Policies such as 472 can be updated to reflect advances in knowledge. Another way to dynamically update processor policies 472 may be over a network. For example, as new security loopholes are discovered, a processor policy such as 472 may be updated via a network connection so watchdog 410 can effectively close discovered security loopholes.
The invention is operational with numerous general purpose or special purpose computing system environments or configurations. Examples of well known computing systems, environments, and/or configurations that may be suitable for use with the invention include, but are not limited to, personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, cell phones, Personal Digital Assistants (PDA), distributed computing environments that include any of the above systems or devices, and the like.
In light of the diverse computing environments that may be built according to the general frameworks of provided in
