cippd_pmaint.c

<B>Digital的Unix操作系统VAX 4.2源码</B>

#ifndef	lint
static char *sccsid = "@(#)cippd_pmaint.c	4.1	(ULTRIX)	7/2/90";
#endif	lint

/************************************************************************
 *                                                                      *
 *                      Copyright (c) 1988 - 1989 by                    *
 *              Digital Equipment Corporation, Maynard, MA              *
 *                      All rights reserved.                            *
 *                                                                      *
 *   This software is furnished under a license and may be used and     *
 *   copied  only  in accordance with the terms of such license and     *
 *   with the  inclusion  of  the  above  copyright  notice.   This     *
 *   software  or  any  other copies thereof may not be provided or     *
 *   otherwise made available to any other person.  No title to and     *
 *   ownership of the software is hereby transferred.                   *
 *                                                                      *
 *   The information in this software is subject to change  without     *
 *   notice  and should not be construed as a commitment by Digital     *
 *   Equipment Corporation.                                             *
 *                                                                      *
 *   Digital assumes no responsibility for the use  or  reliability     *
 *   of its software on equipment which is not supplied by Digital.     *
 *                                                                      *
 ************************************************************************
 *
 *
 *   Facility:	Systems Communication Architecture
 *		Computer Interconnect Port-to-Port Driver
 *
 *   Abstract:	This module contains Computer Interconnect Port-to-Port
 *		Driver( CI PPD ) path maintenance service functions.  All
 *		port drivers which utilize the CI PPD for path establishment,
 *		maintenance, and termination should use these functions.
 *
 *   Creator:	Todd M. Katz	Creation Date:	May 6, 1985
 *
 *   Function/Routines:
 *
 *   cippd_crash_pb		Crash CI PPD Path Represented by Path Block
 *   cippd_get_pb		Retrieve Path Block
 *   cippd_open_pb		Transition Formative Path Block to Open State
 *   cippd_remove_pb		Remove Path Block from System-wide Databases
 *
 *   Modification History:
 *
 *   20-May-1989	Pete Keilty
 *	Added support for mips risc cpu's 
 *	Add include file ../h/systm.h this file has define for remque
 *	insque for mips.
 *
 *   06-Apr-1989	Pete Keilty
 *	Added include file smp_lock.h
 *
 *   17-Jan-1989	Todd M. Katz		TMK0004
 *	1. The PCCB address is now part of the interface to Get_port.
 *	2. The macro Scaaddr_lol() has been renamed to Scaaddr_low().  It now
 *	   accesses only the low order word( instead of low order longword ) of
 *	   a SCA system address.
 *	3. Include header file ../vaxmsi/msisysap.h.
 *
 *   17-Aug-1988	Todd M. Katz		TMK0003
 *	1. Modifications to cippd_get_pb().
 *		1) Update the conditions under which cippd_get_pb() considers
 *		   remote port station address invalid.  Formerly, for a remote
 *		   port station address to be considered invalid it only had to
 *		   exceed the software( CI PPD ) maximum port number(
 *		   CIPPD_MAXPATHS - 1 ) and be provided to the function in the
 *		   form of buffer.  Now, not only must both these criteria
 *		   still be met; but in addition, the remote port station
 *		   address must also exceed the hardware maximum port number of
 *		   the specified local port.
 *	   	2) Invoke an appropriate port driver specific routine to log
 *		   invalid remote port station addresses.  Previously the
 *		   logging was done by the CI PPD itself.
 *	2. Modifications to cippd_crash_pb():
 *		1) Initialize the CI PPD specific PCCB field elogopt.port_num
 *		   to EL_UNDEF when it is invoked to crash an non-existent
 *		   path.  This initialization is required by cippd_log_path(),
 *		   the routine invoked to log SCS and CI PPD path crash event
 *		   codes.
 *		2) Apply the path crash severity modifier( ESM_PC ) to the path
 *		   crash reason code whenever the path has not yet been failed.
 *		3) Cache the reason code within the PB only when the path has
 *		   not yet been failed.  Cache a general reason when the path
 *		   is formative.  Cache the specific path crash reason when the
 *		   path is open.  Previously, the path crash reason was always
 *		   being cached.  This could have lead to loss of the
 *		   legitimate reason when the routine is invoked twice in quick
 *		   succession on the same path.  
 *		4) The routine parameter scsbp ALWAYS points to a character
 *		   string of size NAME_SIZE instead of to the SCS header of a
 *		   datagram/message buffer whenever SCS invokes the routine
 *		   with a reason code of E_SYSAP.  This character string
 *		   consists of the name of the local SYSAP responsible for
 *		   crashing the path.
 *		5) Add a comment as to how individual port drivers go about
 *		   crashing paths following discover of physical path specific
 *		   errors.
 *	3. Refer to error logging as event logging.
 *
 *   02-Jun-1988	Ricky S. Palmer
 *	Removed inclusion of header file ../vaxmsi/msisysap.h
 *
 *   11-Apr-1988	Todd M. Katz		TMK0002
 *	1. Add use of Pccb_fork() macro in place of straight-line code.
 *	2. Modify cippd_crash_pb() to optionally panic on requests to crash
 *	   paths.  Error recovery is the default action taken; however,
 *	   panicing occurs whenever the CI PPD configuration variable
 *	   cippd_pc_panic has been appropriately set.  This variable may be set
 *	   to trigger panicing on any path crash request or on only requests to
 *	   crash open paths.
 *	3. CI PPD event logging is now split between the routines
 *	   cippd_log_path() and cippd_log_sys() to differentiate between
 *	   logging of path specific and system level events respectively.
 *	   Modify cippd_crash_pb() and cippd_get_pb() to invoke
 *	   cippd_log_path().
 *
 *   08-Jan-1988	Todd M. Katz		TMK0001
 *	Formated module, revised comments, increased robustness, made
 *	CI PPD and GVP completely independent from underlying port drivers,
 *	restructured code paths, and added SMP support.
 */

/* Libraries and Include Files.
 */
#include		"../h/types.h"
#include		"../h/param.h"
#include		"../h/systm.h"
#include		"../h/ksched.h"
#include		"../h/time.h"
#include		"../h/errlog.h"
#include		"../h/smp_lock.h"
#include		"../io/scs/sca.h"
#include		"../io/scs/scaparam.h"
#include		"../io/ci/cippdsysap.h"
#include		"../io/ci/cisysap.h"
#include		"../io/msi/msisysap.h"
#include		"../io/bi/bvpsysap.h"
#include		"../io/gvp/gvpsysap.h"
#include		"../io/uba/uqsysap.h"
#include		"../io/sysap/sysap.h"
#include		"../io/ci/cippdscs.h"
#include		"../io/ci/ciscs.h"
#include		"../io/msi/msiscs.h"
#include		"../io/bi/bvpscs.h"
#include		"../io/gvp/gvpscs.h"
#include		"../io/uba/uqscs.h"
#include		"../io/scs/scs.h"
#include		"../io/gvp/gvp.h"
#include		"../io/ci/cippd.h"

/* External Variables and Routines.
 */
extern	u_short		cippd_pc_panic;
extern	void		cippd_log_path(), ksched(), scs_dealloc_pb(),
			scs_dealloc_sb();

/*   Name:	cippd_crash_pb	- Crash CI PPD Path Represented by Path Block
 *
 *   Abstract:	This function crashes a specific CI PPD path to a known system
 *		and initiates the process by which it is disabled and cleaned
 *		up.  This routine may be sequentially invoked any number of
 *		times to crash a specific path; however, only its first
 *		invocation actually results in crashing of the path.  All
 *		subsequent invocations only log crash requests and dispose of
 *		optional buffers.  This situation persists until the next
 *		incarnation of the path which does not occur until after the
 *		path has been cleaned up, re-discovered, and re-established.
 *
 *		An optional message or datagram buffer can be provided.
 *		Alternatively an optional port command packet can be provided
 *		when this routine is invoked by the port drivers themselves.
 *		Three options exist for disposal of this optional buffer:
 *
 *		1. Buffers may be returned to the routine invocator.
 *		2. Buffers may be deallocated.
 *		3. Buffers may be inserted onto the appropriate free queue of
 *		   the appropriate local port.
 *
 *		Only datagram and message buffers may be disposed of by
 *		deallocation or placement onto an appropriate free queue.  PDs
 *		must ALWAYS request return of the buffers they provide when
 *		invoking this routine.
 *
 *		A pointer to a character string of size NAME_SIZE is ALWAYS
 *		provided in place of an optional message or datagram buffer
 *		whenever this routine is invoked with a reason code == E_SYSAP.
 *		This character string contains the name of the local SYSAP
 *		responsible for crashing the specified path.
 *
 *		The appropriate local port is crashed whenever the optional PB
 *		is not provided.
 *
 *		NOTE: Port drivers may invoke this routine following discovery
 *		      of a physical error associated with a specific path.
 *		      However, they must log the explicit physical error
 *		      themselves before invoking this routine with a "generic"
 *		      CI PPD event code( E_PD or SE_PD ) appropriate to the
 *		      severity of the physical error.
 *
 *		NOTE: When port drivers invoke this routine to crash a path,
 *		      the path may have already been disabled.  If so, they
 *		      must set the PB path_closed state bit BEFORE calling this
 *		      routine.
 *
 *		NOTE: All error recovery is bypassed in favor of immediately
 *		      panicing the system whenever the CI PPD configuration
 *		      variable flag cippd_pc_panic is appropriately set.
 *		      Normally, this flag is 0 and error recovery is the
 *		      default action taken.  It is set only when special
 *		      debugging activity is required and may be set to trigger
 *		      panicing on only requests to crash open paths or on
 *		      requests to crash any path, formative or open.
 *
 *   Inputs:
 *
 *   IPL_SCS			- Interrupt processor level
 *   cippd_pc_panic		- CI PPD path crash panic flag
 *   disposal			- RETURN_BUF or DEALLOC_BUF or RECEIVE_BUF
 *   pccb			- Port Command and Control Block pointer
 *   reason			- Reason for crashing virtual path
 *   scsbp			- Address of SCS header in buffer( OPTIONAL )
 *				- Address of local SYSAP name( OPTIONAL )
 *   pb				- Path Block pointer( OPTIONAL )
 *
 *   Outputs:
 *
 *   IPL_SCS			- Interrupt processor level
 *   pb				- Path Block pointer( OPTIONAL )
 *	pinfo.reason		-  Reason for crashing virtual path
 *   pccb			- Port Command and Control Block pointer
 *      ppd.cippd               -  CI PPD specific PCCB fields
 *          elogopt.port_num	-   Remote port number
 *	    elogopt.sysapname	-   Name of local SYSAP crashing path
 *
 *   SMP:	The PCCB is locked INTERNALLY whenever it was not locked
 *		EXTERNALLY prior to routine invocation.  Locking the PCCB
 *		synchronizes access; prevents premature PB deletion; and is
 *		required by cippd_log_path(), by PD routines which add
 *		datagrams or messages to appropriate local port free queues in
 *		this particular case( It is required in this particular case
 *		because the PB is locked ), and by the CI PPD finite state
 *		machine.  PCCB addresses are always valid because these data
 *		structures are never deleted once their corresponding ports
 *		have been initialized.
 *
 *		The PB is locked INTERNALLY( provided it exists ) whenever it
 *		was not locked EXTERNALLY prior to routine invocation.  Locking
 *		the PB allows exclusive access to PB contents and is required
 *		by the CI PPD finite state machine.  PB address validity must
 *		be EXTERNALLY guaranteed either through EXTERNALLY held locks
 *		or by some other mechanism.
 *
 *		Whenever the PB is locked EXTERNALLY the PCCB must also be
 *		locked EXTERNALLY.  This is because the latter occupies a
 *		higher relative position than the former within the SCA locking
 *		hierarchy.
 */
void
cippd_crash_pb( pccb, pb, reason, disposal, scsbp )
    PCCB		*pccb;
    register PB		*pb;
    register u_long	reason;
    u_long		disposal;
    register SCSH	*scsbp;
{
    register GVPPPDH	*cippdbp;
    register u_long	unlock_pb = 0, unlock_pccb = 0;

    /* Crashing of formative and established CI PPD paths is extremely similar
     * with only a few minor differences.  These are pointed out as they are
     * encountered.  Crashing of either type of path consists of the following
     * distinct steps:
     *
     * 1. The crash request is pre-processed.
     * 2. The path is disabled.
     * 3. The path is asynchronously cleaned up.
     *
     * The first step is executed each time this routine is invoked.  It is the
     * latter two steps which constitute crashing the path and are executed
     * only once per path incarnation.  The major benefits of this single
     * threading of path crashing are the assumptions made during path clean up
     * which follow directly from it.  These assumptions greatly reduce code
     * complexity and allow extension of the CI PPD to an SMP environment with
     * a minimum of pain.  They are pointed out as they occur.
     *
     * Pre-processing of crash requests includes:
     *
     * 1. Locking the PCCB whenever it was not locked EXTERNALLY.
     * 2. Locking the PB whenever one was provided and it was not locked
     *    EXTERNALLY.
     * 3. Applying the path crash severity modifier( ESM_PC ) to the path crash
     *    reason code.
     * 4. Caching the path crash reason within the PB for latter reference.
     * 5. Caching the name of the local SYSAP responsible for crashing the path
     *	  for event logging purposes.
     * 6. Logging the crash request.
     * 7. Optionally panicing the system.
     * 8. Disposing of the optional buffer as requested.
     *
     * Steps 3 is executed only if the path is still open as only open paths
     * can be actually crashed.  Formative paths are really just prematurely
     * terminated and failed paths have already been crashed.
     *
     * Step 4 is executed only if the path has not yet been crashed.  This step
     * must only be executed once per path incarnation( ie- only if the path