ka6200.c

<B>Digital的Unix操作系统VAX 4.2源码</B>

#ifndef lint
static char *sccsid = "@(#)ka6200.c	4.3	ULTRIX	2/28/91";
#endif lint

/************************************************************************
 *									*
 *			Copyright (c) 1988 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.	*
 *									*
 ************************************************************************/

/***********************************************************************
 *
 * Modification History:
 *
 * 26-Feb-91  jas
 *	Mapped in XMI nodespace in ka6200conf().  Need to be able to
 *	access nodespace.
 *
 * 18-Jun-90  jas
 *      added support for XMA2 to xma routines.
 *
 * 30-Mar-90  jaw
 *	recover from hard errors on ident command (passive release).
 *
 * 29-Nov-89    Paul Grist
 *      modified ka6200machcheck() to call log_xmi_bierrors and log_xmierrors
 *      to look for any pending XBI or VAXBI errors. Added log_ka6200memerrs
 *      for outside callers to log xma errors (for log_bierrors).
 *        
 * 10-Nov-89	jaw
 *	move kmalloc of attach processors machdep structure to cca_startcpu.
 *
 * 24-May-89	darrell
 *	Changed the #include for cpuconf.h to find it in it's new home --
 *	sys/machine/common/cpuconf.h
 *
 * 24-May-89	darrell
 *	Removed the v_ prefix from all cpusw fields, removed cpup from any
 *	arguments being passed in function args.  cpup is now defined
 *	globally -- as part of the new cpusw.
 *
 *  26-Jan-89	jaw
 *	fix up start/stop cpu.
 *
 * 19-Aug-88	Tom Kong.  
 *		Changed ka6200machcheck so that instead of panicking,
 *		we try to simply terminating the user process that
 *		caused the machine check.  For several cases of machine
 *		checks, the instruction is simply retried and no process
 *		is killed.  Note that this only happens if KILL_USER is
 *		defined.  Otherwise the OS is killed.
 *
 *		Took out "extern int (*vax8800bivec[])()" since it isn't
 *		referenced in this file.
 *
 * 22-Jul-88	darrell Moved cca_send() and cca_setup() from this file
 *			to cvax.c so that it can be shared by VAX60
 *			(Firefox) machine dependent code.
 *
 * 07-Jun-88	darrell Removed ka650.h from include files and added
 *		        cvax.h.
 *
 * 26-Apr-88	jaw	fix to mcheck handler.
 *
 * 04-Feb-88	jaw 	work on soft error handling.
 *
 * 28-Jan-88	jaw 	Add logging for CRD's
 *
 **********************************************************************/

#include "../machine/pte.h"
#include "../h/param.h"
#include "../h/conf.h"
#include "../h/time.h"
#include "../h/errno.h"
#include "../h/systm.h"
#include "../h/types.h"
#include "../h/map.h"
#include "../h/buf.h"
#include "../h/errlog.h"
#include "../h/ioctl.h"
#include "../h/tty.h"
#include "../h/cpudata.h"
#include "../../machine/common/cpuconf.h"
#include "../h/kmalloc.h"
#include "../h/vmmac.h"

#include "../machine/cons.h"
#include "../machine/cons.h"
#include "../machine/cpu.h"
#include "../machine/clock.h"
#include "../machine/mtpr.h"
#include "../machine/mem.h"
#include "../machine/nexus.h"
#include "../machine/scb.h"
#include "../io/uba/ubareg.h"
#include "../io/uba/ubavar.h"
#include "../io/bi/buareg.h"
#include "../machine/sas/vmb.h"
#include "../machine/ka6200.h"
#include "../machine/cvax.h"

#include "../io/xmi/xmireg.h"
#include "../io/xmi/xmareg.h"

/*#define KILL_USER */
#ifdef KILL_USER
/*
 * Need u symbols, ...etc to terminate user process
 */
#include "../h/user.h"	
#include "../h/proc.h"
#include "../machine/psl.h"
#endif

extern struct bidata bidata[];
extern int cache_state;
extern int nNVAXBI;
extern char *ka6200_ip_addr;

struct ssc_regs cvqssc[1];		/* SSC regs */
struct v6200csr v6200csr[1];
char *mccvax[] = {
	"unknown machine check type code",		/* 0 */
	"CFPA protocol error",				/* 1 */
	"CFPA reserved instruction",			/* 2 */
	"CFPA protocol error",				/* 3 */
	"CFPA protocol error",				/* 4 */
	"process PTE in P0 space during TB miss",	/* 5 */
	"process PTE in P1 space during TB miss",	/* 6 */
	"process PTE in P0 space during M = 0",		/* 7 */
	"process PTE in P1 space during M = 0",		/* 8 */
	"hardware interrupt at unused IPL",		/* 9 */
	"undefined MOVC3 or MOVC5 state",		/* 10 */
	"cache/memory/bus read error",			/* 80 */
	"SCB, PCB or SPTE read error",			/* 81 */
	"cache/memory/bus write error",			/* 82 */
	"PCB or SPTE write error",			/* 83 */
};



/*
 * KA6200 machine check exception handler.
 * Environment:
 *	IPL		0x1f
 *	stack		Interrupt stack
 *	SCB vector <1:0> = 01.
 *
 * Parameter:
 *	mcf 		Points to machine check frame.  For Ka6200, the
 *			frame format:
 *				byte count	(0x10)
 *				mcheck code (1,2,3,4,5,6,7,8,9,a,80,81,82,or83)
 *				internal state info 1
 *				internal state info 2
 *				PC
 *				PSL
 * Returns:
 *	0
 */
ka6200machcheck (mcf)
register struct el_mc6200frame *mcf;
{
	register int cpunum;
	register struct xcp_reg *xcp_node;
	int xminode, recover;
	struct xmidata *xmidata;
	register struct el_mc6200frame *framep;
	register struct el_rec *elrp;
	struct el_mck *elmckp;
	register unsigned int mcode;
	long	time_now;
	struct	xcp_machdep_data *mchk_data; /* Pointer to an array		   */
	
	mchk_data =(struct xcp_machdep_data *) ( CURRENT_CPUDATA->cpu_machdep);

	time_now = mfpr(TODR);		/* Log the current time	*/

	recover = 0;			/* Assume we have to panic */

	cpunum=CURRENT_CPUDATA->cpu_num;

	if (mchk_data->mchk_in_progress == 0) {
		/* if no machine check in progress, we're ok */
		mchk_data->mchk_in_progress++;
		}
	else	{
		/* mcheck on another mcheck, we're in big trouble	*/
		asm("halt");
	}


	/*
	 * Case of the type of machine check, we may be able
	 * to recover from the machine check under certain conditions.
	 */
	mcode = ((struct mcframe *)mcf)->mc_summary;

	switch (mcode) {
	case 1:	/* CFPA protocol error */
	case 2:	/* CFPA reserved instruction */
	case 3:	/* CFPA unknown error */
	case 4:	/* CFPA unknown error */
	case 5:	/* calculated virtual address of process PTE in P0 space */
		/* TB miss flow */
	case 6:	/* calculated virtual address of process PTE in P1 space */
		/* TB miss flow */
	case 7:	/* calculated virtual address of process PTE in P0 space */
		/* M = 0 flow */
	case 8:	/* calculated virtual address of process PTE in P1 space */
		/* M = 0 flow */
	case 10:/* Impossible situations in microcode */

#ifdef KILL_USER
		/* 
		 * The above errors are not recoverable.  If we
		 * are running a user process, the user process should
		 * be killed. If we are running the OS, panic.
		 *
		 * Note: we assume we were running a user process during
		 * the machine check if the saved PSL on the stack indicate
		 * a current mode of USER, and that the process isn't "init".
		 */
		if (USERMODE(mcf->mc1_psl) && (u.u_procp->p_pid != 1)) {
			/* kill the user process */
			swkill(u.u_procp, "ka6200machcheck");
			recover = 1;	/* don't have to panic */
		}
#endif
		break;
		
			
	case 0x80: /* read bus error, normal read */
	case 0x81: /* read bus error, SPTE, PCB, or SCB read */
		 
		/*
		 * The instruction is restartable for the above errors if:
		 * 1) The saved PSL on the stack has FPD bit set, or
		 * 2) The VAX CAN'T RESTART bit on the mcheck frame is clear.
		 */
		if (((mcf->mc1_psl & 0x08000000) == 0x08000000)	||
		    ((mcf->mc1_internal_state2 & 0x8000) == 0)) {
			/*
			 * We can restart the instruction.
			 * Decide to allow restart if one or more of the 
			 * following is satisfied:
			 * 1) previous machine check was long ago.
			 * 2) previous machine check was different.
			 */
			if (time_now - mchk_data->time > 1000) {
				/* last mcheck was at least 10 secs ago */
				recover = 1;
			}

			if (mchk_data->code != mcf->mc1_summary)
				recover = 1;

			/*
			 * if recover == 1 (we are retrying the instruction)
			 * we should (in future) map out the bad physical 
			 * page here.
			 */
		}
		if (!recover)	{
			/*
			 * The instruction is not restartable, we
			 * need to kill the user process or the operating
			 * system.
			 */
#ifdef KILL_USER
			if (USERMODE(mcf->mc1_psl) && (u.u_procp->p_pid !=1)) {
				swkill(u.u_procp,"ka6200machcheck");
				recover = 1; /* kill user, but don't panic */
				/*
			 	 * In addition to killing the user process,
			 	 * we should map out the bad physical page in
			 	 * the future.  Do it here.
				 */
			}
#endif
		}
		break;
	case 0x82: /* write bus error, normal write */
	case 0x83: /* write bus error, SPTE or PCB write */
		/*
		 * The instruction is not restartable, we
		 * need to kill the user process or the operating
		 * system.
		 */

#ifdef KILL_USER
		if (USERMODE(mcf->mc1_psl) && (u.u_procp->p_pid !=1)) {
			swkill(u.u_procp,"ka6200machcheck");
			recover = 1; /* kill user, but don't panic */
			/*
			 * In addition to killing the user process,
			 * we should map out the bad physical page in
			 * the future.  Do it here.
			 */
		}
#endif
		break;
	case 9:	/* Interrupt controller requests an interrupt at an unused */
		/* IPL (IPL 0x18,0x19,and 0x1b)				   */
	default:   /* Unknown machine check type code */
		/* Panic on the above errors */
		break;
	}


	/*
	 * Log the machine check in the error log.
	 */
	xmidata = get_xmi(0);/* get pointer to xcp node that machine checked */
	xcp_node =(struct xcp_reg *)xmidata->xmivirt+(v6200csr->csr1 & 0xf);
	/* allocate a error log packet */
	elrp = ealloc((sizeof(struct el_mc6200frame)),
		recover ? EL_PRIHIGH:EL_PRISEVERE);

	if (elrp) {
		LSUBID(elrp,ELCT_MCK,ELMCKT_6200,cpu_subtype,cpunum,EL_UNDEF,
			mcf->mc1_summary);
		elmckp = &elrp->el_body.elmck;
		framep = (struct el_mc6200frame *) &elmckp->elmck_frame.
				el6200mcf.mc1_bcnt;
		framep->mc1_bcnt = mcf->mc1_bcnt;
		framep->mc1_summary = mcf->mc1_summary;
		framep->mc1_vap = mcf->mc1_vap;
		framep->mc1_internal_state1 = mcf->mc1_internal_state1;
		framep->mc1_internal_state2 = mcf->mc1_internal_state2;
		framep->mc1_pc = mcf->mc1_pc;
		framep->mc1_psl = mcf->mc1_psl;
		framep->xcp_dtype = xcp_node->xcp_dtype;
		framep->xcp_xbe = xcp_node->xcp_xbe;
		framep->xcp_csr2 = xcp_node->xcp_csr2;
		framep->xcp_csr1 = v6200csr->csr1;
		framep->xcp_mser = mfpr(MSER);
		EVALID(elrp);	/* Make error log packet valid */
	}

	/* look for any pending XMI errors */

	log_xmierrors(0,mcf->mc1_pc);
	log_xmi_bierrors(0,mcf->mc1_pc);


	/* 
	 * if not recovering then we print the machine check frame to 
	 * console device, and then panic.  
	 */
	if (!recover) {
		cprintf("cpu %x ",(v6200csr->csr1 & 0xf));
		if ( mcode > 10) {
			mcode = mcode - 0x75;
			if (mcode >14) mcode = 0;
	        } 
		cprintf("%s\n", mccvax[mcode]);
		cprintf("\tcode\t\t= %x\n", mcf->mc1_summary);
		cprintf("\tmost recent virtual addr\t=%x\n", mcf->mc1_vap);
		cprintf("\tinternal state 1\t=%x\n", mcf->mc1_internal_state1);
		cprintf("\tinternal state 2\t=%x\n", mcf->mc1_internal_state2);
		cprintf("\tpc\t\t= %x\n", mcf->mc1_pc);
		cprintf("\tpsl\t\t= %x\n\n", mcf->mc1_psl);
	
		xma_check_errors(xcp_node,EL_PRISEVERE);
		panic("mchk");
		} 
	else 	{
		/* We are trying to recover */
		ka6200_enable_cache();
		/*