Leaked source code of windows server 2003
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

943 lines
32 KiB

/*
* HrProcessorEntry.c v0.10
* Generated in conjunction with Management Factory scripts:
* script version: SNMPv1, 0.16, Apr 25, 1996
* project: D:\TEMP\EXAMPLE\HOSTMIB
****************************************************************************
* *
* (C) Copyright 1995 DIGITAL EQUIPMENT CORPORATION *
* *
* This software is an unpublished work protected under the *
* the copyright laws of the United States of America, all *
* rights reserved. *
* *
* In the event this software is licensed for use by the United *
* States Government, all use, duplication or disclosure by the *
* United States Government is subject to restrictions as set *
* forth in either subparagraph (c)(1)(ii) of the Rights in *
* Technical Data And Computer Software Clause at DFARS *
* 252.227-7013, or the Commercial Computer Software Restricted *
* Rights Clause at FAR 52.221-19, whichever is applicable. *
* *
****************************************************************************
*
* Facility:
*
* Windows NT SNMP Extension Agent
*
* Abstract:
*
* This module contains the code for dealing with the get, set, and
* instance name routines for the HrProcessorEntry. Actual instrumentation code is
* supplied by the developer.
*
* Functions:
*
* A get and set routine for each attribute in the class.
*
* The routines for instances within the class.
*
* Author:
*
* D. D. Burns @ Webenable Inc
*
* Revision History:
*
* V1.00 - 04/28/97 D. D. Burns Genned: Thu Nov 07 16:42:19 1996
*
*/
#include <nt.h>
#include <ntrtl.h>
#include <nturtl.h>
#include <windows.h>
#include <malloc.h>
#include <snmp.h>
#include "mib.h"
#include "smint.h"
#include "hostmsmi.h"
#include "user.h" /* Developer supplied include file */
#include "HMCACHE.H" /* Cache-related definitions */
#include "HRDEVENT.H" /* HrDevice Table-related definitions */
#include <stdio.h>
/*
|==============================================================================
| "Processor-Information Buffer" Definition
|
| This definition defines a logical "Processor Information Block" where we
| can store all the information returned from an NtQuerySystemInformation()
| call that requests "SystemProcessorPerformanceInformation" for each running
| processor.
*/
typedef
struct pi_block {
struct pi_block *other; // Associated "other" buffer
LARGE_INTEGER sys_time; // Time when "pi_array" was last
// refreshed in 100ns ticks
SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION
*pi_array; // Array: One element per processor
DWORD pi_size; // Size in bytes of "pi_array" storage
}
PI_BLOCK;
/*
|==============================================================================
| "Processor-Information Buffer" Instances
|
| We create two instances of a Processor Information Buffer, one for the
| "oldest" and a second for "newest" samples of timer values. Maintaining
| two enables us to compute the average over time for processor loads.
|
| These blocks are initialized by code in "Gen_HrProcessor_Cache()" in
| this module.
|
| These buffers are refreshed in an alternating manner by function
| "hrProcessLoad_Refresh()" (in this module) that itself is invoked on a
| timer-driven basis. (See source for the function).
*/
static
PI_BLOCK pi_buf1; // First Buffer
static
PI_BLOCK pi_buf2; // Second Buffer
/*
|==============================================================================
| Oldest "Processor-Information Buffer"
|
| This cell points at one of the two PI_BLOCKs above. It always points to
| the buffer block that has the "oldest" data in it.
*/
static
PI_BLOCK *oldest_pi=NULL;
#if defined(PROC_CACHE) // For debug cache dump only
static
int processor_count;
#endif
/*
* GetHrProcessorFrwID
* The product ID of the firmware associated with the processor.
*
* Gets the value for HrProcessorFrwID.
*
* Arguments:
*
* outvalue address to return variable value
* accesss Reserved for future security use
* instance address of instance name as ordered native
* data type(s)
*
* Return Codes:
*
* Standard PDU error codes.
*
* SNMP_ERRORSTATUS_NOERROR Successful get
* SNMP_ERRORSTATUS_GENERR Catch-all failure code
* mibtget.c v0.10
*
| =============== From WebEnable Design Spec Rev 3 04/11/97==================
| hrProcessorFrwID
|
| ACCESS SYNTAX
| read-only ProductID
|
| "The product ID of the firmware associated with the processor."
|
| DISCUSSION:
|
| <POA-11> The underlying syntax of this attribute is Object Identifier. None
| of the documented Win32 API functions seem capable of reporting this value.
| We are allowed to report "unknownProductID" ("0.0") in liew of the real
| value, and this will be hardcoded unless an alternative is specified.
|
| RESOLVED >>>>>>>>
| <POA-11> Returning an unknown Product ID is acceptable.
| RESOLVED >>>>>>>>
|
|============================================================================
| 1.3.6.1.2.1.25.3.3.1.1.<instance>
| | | | |
| | | | *-hrProcessorFrwID
| | | *-hrProcessorEntry
| | *-hrProcessorTable
| *-hrDevice
*/
UINT
GetHrProcessorFrwID(
OUT ProductID *outvalue ,
IN Access_Credential *access ,
IN InstanceName *instance )
{
/*
| The deal on this attribute is that we'll never have a valid OID value
| for this attribute. Consequently, we always return the standard
| "unknown" OID value ("0.0") regardless of the instance value (which
| by now in the calling sequence of things has been validated anyway).
*/
if ( (outvalue->ids = SNMP_malloc(2 * sizeof( UINT ))) == NULL) {
return SNMP_ERRORSTATUS_GENERR;
}
outvalue->idLength = 2;
/*
| Load in the OID value for "unknown" for ProductID: "0.0"
*/
outvalue->ids[0] = 0;
outvalue->ids[1] = 0;
return SNMP_ERRORSTATUS_NOERROR ;
} /* end of GetHrProcessorFrwID() */
/*
* GetHrProcessorLoad
* The average, over the last minute, of the percentage of time that this
* processor was not idle.
*
* Gets the value for HrProcessorLoad.
*
* Arguments:
*
* outvalue address to return variable value
* accesss Reserved for future security use
* instance address of instance name as ordered native
* data type(s)
*
* Return Codes:
*
* Standard PDU error codes.
*
* SNMP_ERRORSTATUS_NOERROR Successful get
* SNMP_ERRORSTATUS_GENERR Catch-all failure code
* mibtget.c v0.10
*
| =============== From WebEnable Design Spec Rev 3 04/11/97==================
| hrProcessorLoad
|
| ACCESS SYNTAX
| read-only INTEGER (0..100)
|
| "The average, over the last minute, of the percentage of time that this
| processor was not idle."
|
| DISCUSSION:
|
| <POA-12> It seems likely to me that this performance statistic might be
| maintained or be derivable from performance information maintained in the
| Registry. Please describe.
|
| RESOLVED >>>>>>>>
| <POA-12> I think we should just use the PerfMon code for this.
| RESOLVED >>>>>>>>
|
|============================================================================
| We reference a continuously updated module-local cache of CPU time-usage
| info maintained in the buffer-blocks "pi_buf1" and "pi_buf2" defined
| at the beginning of this module. In the code below, we reach into these
| caches and compute the processor load for the processor specified.
|============================================================================
| 1.3.6.1.2.1.25.3.3.1.2.<instance>
| | | | |
| | | | *-hrProcessorLoad
| | | *-hrProcessorEntry
| | *-hrProcessorTable
| *-hrDevice
*/
UINT
GetHrProcessorLoad(
OUT Integer *outvalue ,
IN Access_Credential *access ,
IN InstanceName *instance )
{
ULONG index; /* As fetched from instance structure */
CACHEROW *row; /* Row entry fetched from cache */
ULONG p; /* Selected Processor (number from 0) */
SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION
*oldest, *newest;/* --> CPU data for "n" and "n+1minute" */
LONGLONG llDendiff; /* Difference Denominator */
LONGLONG llNewNum; /* Numerator of Newest Time-count */
LONGLONG llOldNum; /* Numerator of Oldest Time-count */
double fNum,fDen; /* Floated versions of LONGLONGs */
double fload; /* Percentage Load */
/*
| Grab the instance information
*/
index = GET_INSTANCE(0);
/*
| Use it to find the right entry in the cache
*/
if ((row = FindTableRow(index, &hrDevice_cache)) == NULL) {
return SNMP_ERRORSTATUS_GENERR;
}
/*
| By convention with "Gen_HrProcessor_Cache()" the cache initialization
| routine, the "hidden context" for devices which are "Processors" is
| the Processor Number, starting with 0.
*/
p = row->attrib_list[HIDDEN_CTX].u.unumber_value;
/*
| We compute the load using "SystemProcessorPerformanceInformation" that
| has been gathered for all processors in buffers maintained in "pi_buf1"
| and "pi_buf2".
|
| Obtain pointers to the "newest" and "oldest" slug of information for
| the specified processor out of "pi_buf1/2".
*/
oldest = &(oldest_pi->pi_array[p]);
newest = &(oldest_pi->other->pi_array[p]);
/*
| The performance info (as of this writing) we need comes from:
|
| typedef struct _SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION {
| LARGE_INTEGER IdleTime;
| LARGE_INTEGER KernelTime;
| LARGE_INTEGER UserTime;
| LARGE_INTEGER DpcTime; // DEVL only
| LARGE_INTEGER InterruptTime; // DEVL only
| ULONG InterruptCount;
| } ....
|
| where "IdleTime" is taken to be time spent by this processor in its
| idlethread.
|
| "KernelTime" is taken to be total time spent by processor in kernel
| mode code (including the idle thread).
|
| "UserTime" is taken to be total time spent by processor in user mode
| code.
|
| all in ticks of 100ns (one tenth of a millionth of a second).
|
| So total "Not-Idle" time is "(KernelTime-IdleTime) + UserTime" in ticks.
| Total time for the interval is the difference in the "sys_time" that is
| associated with each buffer ("oldest_pi->" and "oldest_pi->other->").
|
*/
llNewNum = (newest->KernelTime.QuadPart - newest->IdleTime.QuadPart)
+ newest->UserTime.QuadPart;
llOldNum = (oldest->KernelTime.QuadPart - oldest->IdleTime.QuadPart)
+ oldest->UserTime.QuadPart;
/* (Newest System-Time) - (Oldest System-Time) */
llDendiff = oldest_pi->other->sys_time.QuadPart - oldest_pi->sys_time.QuadPart;
/* If there will be no divide by 0 */
if ( llDendiff != 0 ) {
/*
| Now float these guys and convert to a percentage.
*/
fNum = (double) (llNewNum - llOldNum);
fDen = (double) llDendiff;
fload = (fNum / fDen) * 100.0;
}
else {
fload = 0.0;
}
*outvalue = (int) fload; // Truncate to integer
return SNMP_ERRORSTATUS_NOERROR ;
} /* end of GetHrProcessorLoad() */
/*
* HrProcessorEntryFindInstance
*
* This routine is used to verify that the specified instance is
* valid.
*
* Arguments:
*
* FullOid Address for the full oid - group, variable,
* and instance information
* instance Address for instance specification as an oid
*
* Return Codes:
*
* SNMP_ERRORSTATUS_NOERROR Instance found and valid
* SNMP_ERRORSTATUS_NOSUCHNAME Invalid instance
*
*/
UINT
HrProcessorEntryFindInstance( IN ObjectIdentifier *FullOid ,
IN OUT ObjectIdentifier *instance )
{
UINT tmp_instance ; /* Instance arc value */
CACHEROW *row; /* Row entry fetched from cache */
//
// Developer instrumentation code to find appropriate instance goes here.
// For non-tables, it is not necessary to modify this routine. However, if
// there is any context that needs to be set, it can be done here.
//
if ( FullOid->idLength <= HRPROCESSORENTRY_VAR_INDEX )
// No instance was specified
return SNMP_ERRORSTATUS_NOSUCHNAME ;
else if ( FullOid->idLength != HRPROCESSORENTRY_VAR_INDEX + 1 )
// Instance length is more than 1
return SNMP_ERRORSTATUS_NOSUCHNAME ;
else
// The only valid instance for a non-table are instance 0. If this
// is a non-table, the following code validates the instances. If this
// is a table, developer modification is necessary below.
tmp_instance = FullOid->ids[ HRPROCESSORENTRY_VAR_INDEX ] ;
/*
| For hrProcessorTable, the instance arc(s) is a single arc, and
| it must correctly select an entry in the hrDeviceTable cache.
|
| Check that here.
*/
if ( (row = FindTableRow(tmp_instance, &hrDevice_cache)) == NULL ) {
return SNMP_ERRORSTATUS_NOSUCHNAME ;
}
else
{
/*
| The instance arc selects an hrDeviceTable row entry, but is that
| entry actually for a device of type "Processor"?
|
| (We examine the last arc of the OID that specifies the device
| type in the row entry selected by the instance arc).
*/
if (row->attrib_list[HRDV_TYPE].u.unumber_value !=
HRDV_TYPE_LASTARC_PROCESSOR) {
return SNMP_ERRORSTATUS_NOSUCHNAME;
}
// the instance is valid. Create the instance portion of the OID
// to be returned from this call.
instance->ids[ 0 ] = tmp_instance ;
instance->idLength = 1 ;
}
return SNMP_ERRORSTATUS_NOERROR ;
} /* end of HrProcessorEntryFindInstance() */
/*
* HrProcessorEntryFindNextInstance
*
* This routine is called to get the next instance. If no instance
* was passed than return the first instance (1).
*
* Arguments:
*
* FullOid Address for the full oid - group, variable,
* and instance information
* instance Address for instance specification as an oid
*
* Return Codes:
*
* SNMP_ERRORSTATUS_NOERROR Instance found and valid
* SNMP_ERRORSTATUS_NOSUCHNAME Invalid instance
*
*/
UINT
HrProcessorEntryFindNextInstance( IN ObjectIdentifier *FullOid ,
IN OUT ObjectIdentifier *instance )
{
//
// Developer supplied code to find the next instance of class goes here.
// If this is a class with cardinality 1, no modification of this routine
// is necessary unless additional context needs to be set.
// If the FullOid does not specify an instance, then the only instance
// of the class is returned. If this is a table, the first row of the
// table is returned.
//
// If an instance is specified and this is a non-table class, then NOSUCHNAME
// is returned so that correct MIB rollover processing occurs. If this is
// a table, then the next instance is the one following the current instance.
// If there are no more instances in the table, return NOSUCHNAME.
//
CACHEROW *row;
ULONG tmp_instance;
if ( FullOid->idLength <= HRPROCESSORENTRY_VAR_INDEX )
{
/*
| Too short: must return the instance arc that selects the first
| entry in the table if there is one.
*/
tmp_instance = 0;
}
else {
/*
| There is at least one instance arc. Even if it is the only arc
| we use it as the "index" in a request for the "NEXT" one.
*/
tmp_instance = FullOid->ids[ HRPROCESSORENTRY_VAR_INDEX ] ;
}
/* Now go off and try to find the next instance in the table */
if ((row = FindNextTableRow(tmp_instance, &hrDevice_cache)) == NULL) {
return SNMP_ERRORSTATUS_NOSUCHNAME ;
}
/*
| The instance arc selects an hrDeviceTable row entry, but is that
| entry actually for a device of type "Processor"?
|
| (We examine the last arc of the OID that specifies the device
| type in the row entry selected by the instance arc).
*/
do {
if (row->attrib_list[HRDV_TYPE].u.unumber_value ==
HRDV_TYPE_LASTARC_PROCESSOR) {
/* Found an hrDeviceTable entry for the right device type */
break;
}
/* Step to the next row in the table */
row = GetNextTableRow( row );
}
while ( row != NULL );
/* If we found a proper device-type row . . . */
if ( row != NULL) {
instance->ids[ 0 ] = row->index ;
instance->idLength = 1 ;
}
else {
/*
| Fell off the end of the hrDeviceTable without finding a row
| entry that had the right device type.
*/
return SNMP_ERRORSTATUS_NOSUCHNAME ;
}
return SNMP_ERRORSTATUS_NOERROR ;
} /* end of HrProcessorEntryFindNextInstance() */
/*
* HrProcessorEntryConvertInstance
*
* This routine is used to convert the object id specification of an
* instance into an ordered native representation. The object id format
* is that object identifier that is returned from the Find Instance
* or Find Next Instance routines. It is NOT the full object identifier
* that contains the group and variable object ids as well. The native
* representation is an argc/argv-like structure that contains the
* ordered variables that define the instance. This is specified by
* the MIB's INDEX clause. See RFC 1212 for information about the INDEX
* clause.
*
*
* Arguments:
*
* oid_spec Address of the object id instance specification
* native_spec Address to return the ordered native instance
* specification
*
* Return Codes:
*
* SUCCESS Conversion complete successfully
* FAILURE Unable to convert object id into native format
*
*/
UINT
HrProcessorEntryConvertInstance( IN ObjectIdentifier *oid_spec ,
IN OUT InstanceName *native_spec )
{
static char *array; /* The address of this (char *) is passed back */
/* as though it were an array of length 1 of these */
/* types. */
static ULONG inst; /* The address of this ULONG is passed back */
/* (Obviously, no "free()" action is needed) */
/* We only expect the one arc in "oid_spec" */
inst = oid_spec->ids[0];
array = (char *) &inst;
native_spec->count = 1;
native_spec->array = &array;
return SUCCESS ;
} /* end of HrProcessorEntryConvertInstance() */
/*
* HrProcessorEntryFreeInstance
*
* This routine is used to free an ordered native representation of an
* instance name.
*
* Arguments:
*
* instance Address to return the ordered native instance
* specification
*
* Return Codes:
*
*
*/
void
HrProcessorEntryFreeInstance( IN OUT InstanceName *instance )
{
/* No action needed for hrProcessor Table */
} /* end of HrProcessorEntryFreeInstance() */
/*
| End of Generated Code
*/
/* Gen_HrProcessor_Cache - Gen. a initial cache for HrDevice PROCESSOR Table */
/* Gen_HrProcessor_Cache - Gen. a initial cache for HrDevice PROCESSOR Table */
/* Gen_HrProcessor_Cache - Gen. a initial cache for HrDevice PROCESSOR Table */
BOOL
Gen_HrProcessor_Cache(
ULONG type_arc
)
/*
| EXPLICIT INPUTS:
|
| "type_arc" is the number "n" to be used as the last arc in the
| device-type OID:
|
| 1.3.6.1.2.1.25.3.1.n
| | | |
| | | * Identifying arc for type
| | *-hrDeviceTypes (OIDs specifying device types)
| *-hrDevice
|
| for devices created by this cache-population routine.
|
| IMPLICIT INPUTS:
|
| None.
|
| OUTPUTS:
|
| On Success:
| Function returns TRUE indicating that the HrDevice cache has been fully
| populated with all rows required for Processor devices.
|
| On any Failure:
| Function returns FALSE (indicating "not enough storage" or other
| internal logic error).
|
| THE BIG PICTURE:
|
| At subagent startup time, the cache for each table in the MIB is
| populated with rows for each row in the table. This function is
| invoked by the start-up code in "Gen_HrDevice_Cache()" to
| populate the cache for the HrDevice table with processor-specific
| entries.
|
| OTHER THINGS TO KNOW:
|
| This function is loading entries into the existing HrDevice cache
| for devices of type "processor" as well as setting up what logically
| amounts to a "cache" of information used to compute the value of
| hrProcessorLoad.
|
| Specifically, this function initializes module-local cells that
| describe buffers containing processor-time information for each
| processor.
|
| This function holds a convention with the GET routines earlier in
| this module that the "HIDDEN_CTX" attribute for processors contains
| a number that can be used to select which processor information
| is to be returned. We generate this number here.
|============================================================================
| 1.3.6.1.2.1.25.3.3.1...
| | | |
| | | *hrProcessorEntry
| | *hrProcessorTable
| *-hrDevice
|
*/
{
SYSTEM_INFO sys_info; /* Filled in by GetSystemInfo for processors */
UINT i; /* Handy-Dandy loop index */
char *descr; /* Selected description string */
/* Acquire system information statistics */
GetSystemInfo(&sys_info);
/*
| Build a description based on the system info. We presume all processors
| are identical.
*/
switch (sys_info.wProcessorArchitecture) {
case PROCESSOR_ARCHITECTURE_INTEL:
switch (sys_info.wProcessorLevel) {
case 3: descr = "Intel 80386"; break;
case 4: descr = "Intel 80486"; break;
case 5: descr = "Intel Pentium"; break;
default: descr = "Intel"; break;
}
break;
case PROCESSOR_ARCHITECTURE_ALPHA:
switch (sys_info.wProcessorLevel) {
case 21064: descr = "Alpha 21064"; break;
case 21066: descr = "Alpha 21066"; break;
case 21164: descr = "Alpha 21164"; break;
default: descr = "DEC Alpha"; break;
}
break;
case PROCESSOR_ARCHITECTURE_MIPS:
switch (sys_info.wProcessorLevel) {
case 4: descr = "MIPS R4000"; break;
default: descr = "MIPS"; break;
}
break;
case PROCESSOR_ARCHITECTURE_PPC:
switch (sys_info.wProcessorLevel) {
case 1: descr = "PPC 601"; break;
case 3: descr = "PPC 603"; break;
case 4: descr = "PPC 604"; break;
case 6: descr = "PPC 603+"; break;
case 9: descr = "PPC 604+"; break;
case 20: descr = "PPC 620"; break;
default: descr = "PPC"; break;
}
break;
case PROCESSOR_ARCHITECTURE_UNKNOWN:
default:
descr = "Unknown Processor Type";
}
/*
| For every processor we have in the system, fill in a row in the hrDevice
| table.
*/
for (i = 0; i < sys_info.dwNumberOfProcessors; i += 1) {
/*
| "Hidden Context" is a generated number starting at 0 which we'll
| consider to be the processor number..
*/
if (AddHrDeviceRow(type_arc, // DeviceType OID Last-Arc
descr, // Device Description
&i, // Hidden Ctx "Processor #"
CA_NUMBER // Hidden Ctx type
) == NULL ) {
/* Something blew */
return ( FALSE );
}
}
/*
| Now initialize the PI_BLOCK instances needed to compute the hrProcessorLoad
| and the pointer to the PI_BLOCK instance that is to be considered the
| "oldest".
*/
/*
| Storage for both buffers. . . .*/
pi_buf1.pi_size = sys_info.dwNumberOfProcessors *
sizeof(SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION);
if ((pi_buf1.pi_array = (SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION *)
malloc(pi_buf1.pi_size)) == NULL) {
return ( FALSE ); // Out of Memory
}
pi_buf2.pi_size = sys_info.dwNumberOfProcessors *
sizeof(SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION);
if ((pi_buf2.pi_array = (SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION *)
malloc(pi_buf2.pi_size)) == NULL) {
return ( FALSE ); // Out of Memory
}
/*
| Now "hook" the two buffer blocks together so we can switch between them
| symmetrically.
*/
pi_buf1.other = &pi_buf2;
pi_buf2.other = &pi_buf1;
#if defined(PROC_CACHE) // For debug cache dump only
processor_count = sys_info.dwNumberOfProcessors;
#endif
/*
| Pretend the first is the oldest and get it refreshed.
*/
oldest_pi = &pi_buf1; // Select it
hrProcessLoad_Refresh(); // Refresh it, and select the other as "oldest"
SleepEx(1, FALSE); // Pause one millisecond to avoid divide by 0
hrProcessLoad_Refresh(); // Refresh the second and select other as "oldest"
/*
| Now each Processor Information Block contains full information (about
| all processors) separated in time by 1 millisecond. The "oldest" will
| be refreshed periodically every minute by the timer which is initialized
| via a call to "TrapInit()" made from source "mib.c" after the initialization
| of caches is complete. Once the timer begins ticking regularly, the time
| samples in these two PI_BLOCK buffers will differ by one minute, (the period
| of the timer) which is the period required by the definition of
| "hrProcessorLoad".
*/
return ( TRUE );
}
/* hrProcessLoad_Refresh - Processor Load Time-Information Refresh Routine */
/* hrProcessLoad_Refresh - Processor Load Time-Information Refresh Routine */
/* hrProcessLoad_Refresh - Processor Load Time-Information Refresh Routine */
void
hrProcessLoad_Refresh(
void
)
/*
| EXPLICIT INPUTS:
|
| None.
|
| IMPLICIT INPUTS:
|
| The "Processor-Information Buffer" specified by module-local cell
| "oldest_pi" is refreshed with new time information fetched from
| the kernel.
|
| OUTPUTS:
|
| On Success:
| The PI_Block specified by "oldest_pi" is refreshed and "oldest_pi"
| is set to point to the other PI_BLOCK.
|
| On any Failure:
| Function simply returns.
|
| THE BIG PICTURE:
|
| At subagent startup time, a timer is created by code in "TrapInit()"
| set to "tick" at an interval of one minute.
|
| Each time the timer goes off, the SNMP Agent calls the
| "SnmpExtensionTrap()" standard entry point for the sub agent. Rather
| than handle a trap, that function will invoke this function which
| gathers CPU performance data so that the hrProcessLoad value can be
| properly computed.
|
| OTHER THINGS TO KNOW:
|
| We alternate the buffer into which the newest CPU data is placed
| by simply changing "oldest_pi" (each time we're invoked) to point
| to the "other" buffer after we're done refreshing the oldest buffer.
| In this manner, we always have two buffers of Processor Load info
| allowing us to compute the load during the times associated with
| those two buffers.
*/
{
NTSTATUS ntstatus;
DWORD bytesused;
/* Get the current system-time in 100ns intervals . . .*/
ntstatus = NtQuerySystemTime (&oldest_pi->sys_time);
/*
| . . . and as rapidly thereafter refresh the oldest buffer with information
| on all processors */
ntstatus = NtQuerySystemInformation(SystemProcessorPerformanceInformation,
oldest_pi->pi_array,
oldest_pi->pi_size,
&bytesused);
#if defined(PROC_CACHE)
/* =========================== DEBUG DUMP ================================== */
{
FILE *pfile; /* Dump goes here */
time_t ltime; /* For debug message */
int i; /* Loop index */
SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION
*oldest, *newest;/* --> CPU data for "n" and "n+1minute" */
if ((pfile=fopen(PROC_FILE, "a+")) != NULL) {
time( &ltime);
fprintf(pfile, "\n=============== Open for appending: %s", ctime( &ltime ));
fprintf(pfile, "Periodic Refresh of \"oldest_pi\" @ %x\n", oldest_pi);
/* For each processor . . . */
for (i=0; i < processor_count; i += 1) {
fprintf(pfile, "For Processor %d:\n", i);
oldest = &(oldest_pi->pi_array[i]);
//newest = &(oldest_pi->other->pi_array[i]);
fprintf(pfile, " IdleTime = (HI) %x (LO) %x\n",
oldest->IdleTime.HighPart, oldest->IdleTime.LowPart);
fprintf(pfile, " KernelTime = (HI) %x (LO) %x\n",
oldest->KernelTime.HighPart, oldest->KernelTime.LowPart);
fprintf(pfile, " UserTime = (HI) %x (LO) %x\n",
oldest->UserTime.HighPart, oldest->UserTime.LowPart);
}
}
fclose(pfile);
}
/* =========================== DEBUG DUMP ================================== */
#endif
/* Now the other buffer contains the "oldest" data, so change "oldest_pi" */
oldest_pi = oldest_pi->other;
}