patch-2.4.4 linux/arch/ia64/sn/io/huberror.c

Next file: linux/arch/ia64/sn/io/hubspc.c
Previous file: linux/arch/ia64/sn/io/hubdev.c
Back to the patch index
Back to the overall index

diff -u --recursive --new-file v2.4.3/linux/arch/ia64/sn/io/huberror.c linux/arch/ia64/sn/io/huberror.c
@@ -0,0 +1,475 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Alan Mayer
+ */
+
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <asm/smp.h>
+#include <asm/sn/sgi.h>
+#include <asm/sn/iograph.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/labelcl.h>
+#include <asm/sn/nodemask.h>
+#include <asm/sn/sn_private.h>
+#include <asm/sn/klconfig.h>
+#include <asm/sn/synergy.h>
+#include <asm/sn/sn_cpuid.h>
+#include <asm/sn/pci/pciio.h>
+#include <asm/sn/pci/pcibr.h>
+#include <asm/sn/xtalk/xtalk.h>
+#include <asm/sn/pci/pcibr_private.h>
+#include <asm/sn/intr.h>
+
+extern void hubni_eint_init(cnodeid_t cnode);
+extern void hubii_eint_init(cnodeid_t cnode);
+extern void hubii_eint_handler (int irq, void *arg, struct pt_regs *ep);
+extern void snia_error_intr_handler(int irq, void *devid, struct pt_regs *pt_regs);
+
+extern int maxcpus;
+
+#define HUB_ERROR_PERIOD        (120 * HZ)      /* 2 minutes */
+
+
+void
+hub_error_clear(nasid_t nasid)
+{
+	int i;
+	hubreg_t idsr;
+	int sn;
+
+	for(sn=0; sn<NUM_SUBNODES; sn++) {
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_PEND, -1);
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_STATUS0_A_CLR, -1);
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_STATUS0_B_CLR, -1);
+		REMOTE_HUB_PI_S(nasid, sn, PI_SPURIOUS_HDR_0, 0);
+		REMOTE_HUB_PI_S(nasid, sn, PI_SPURIOUS_HDR_1, 0);
+	}
+
+	REMOTE_HUB_L(nasid, MD_DIR_ERROR_CLR);
+	REMOTE_HUB_L(nasid, MD_MEM_ERROR_CLR);
+	REMOTE_HUB_L(nasid, MD_MISC1_ERROR_CLR);
+	REMOTE_HUB_L(nasid, MD_PROTOCOL_ERR_CLR);
+
+    /*
+     * Make sure spurious write response errors are cleared
+     * (values are from hub_set_prb())
+     */
+    for (i = 0; i <= HUB_WIDGET_ID_MAX - HUB_WIDGET_ID_MIN + 1; i++) {
+        iprb_t prb;
+
+	prb.iprb_regval = REMOTE_HUB_L(nasid, IIO_IOPRB_0 + (i * sizeof(hubreg_t)));
+
+        /* Clear out some fields */
+        prb.iprb_ovflow = 1;
+        prb.iprb_bnakctr = 0;
+        prb.iprb_anakctr = 0;
+
+	/*
+	 * PIO reads in fire-and-forget mode on bedrock 1.0 don't
+	 * frob the credit count properly, making the responses appear
+	 * spurious.  So don't use fire-and-forget mode.  Bug 761802.
+	 */
+        prb.iprb_ff = 0;        /* disable fire-and-forget mode by default */
+
+        prb.iprb_xtalkctr = 3;  /* approx. PIO credits for the widget */
+
+        REMOTE_HUB_S(nasid, IIO_IOPRB_0 + (i * sizeof(hubreg_t)), prb.iprb_regval);
+    }
+
+    REMOTE_HUB_S(nasid, IIO_IO_ERR_CLR, -1);
+    idsr = REMOTE_HUB_L(nasid, IIO_IIDSR);
+    REMOTE_HUB_S(nasid, IIO_IIDSR, (idsr & ~(IIO_IIDSR_SENT_MASK)));
+
+    REMOTE_HUB_L(nasid, NI_PORT_ERROR_CLEAR);
+    /* No need to clear NI_PORT_HEADER regs; they are continually overwritten*/
+
+    REMOTE_HUB_S(nasid, LB_ERROR_MASK_CLR, -1);
+    REMOTE_HUB_S(nasid, LB_ERROR_HDR1, 0);
+
+    /* Clear XB error regs, in order */
+    for (i = 0;
+         i <= XB_FIRST_ERROR_CLEAR - XB_POQ0_ERROR_CLEAR;
+         i += sizeof(hubreg_t)) {
+        REMOTE_HUB_S(nasid, XB_POQ0_ERROR_CLEAR + i, 0);
+    }
+}
+
+
+/*
+ * Function	: hub_error_init
+ * Purpose	: initialize the error handling requirements for a given hub.
+ * Parameters	: cnode, the compact nodeid.
+ * Assumptions	: Called only once per hub, either by a local cpu. Or by a 
+ *			remote cpu, when this hub is headless.(cpuless)
+ * Returns	: None
+ */
+
+void
+hub_error_init(cnodeid_t cnode)
+{
+	nasid_t nasid;
+
+    nasid = cnodeid_to_nasid(cnode);
+    hub_error_clear(nasid);
+
+#ifdef ajm
+    if (cnode == 0) {
+	/*
+	 * Allocate log for storing the node specific error info
+	 */
+	for (i = 0; i < numnodes; i++) {
+	    kl_error_log[i]  = kmem_zalloc_node(sizeof(sn0_error_log_t), 
+						KM_NOSLEEP, i);
+	    hub_err_count[i] = kmem_zalloc_node(sizeof(hub_errcnt_t),
+						VM_DIRECT | KM_NOSLEEP, i);
+	    ASSERT_ALWAYS(kl_error_log[i] && hub_err_count[i]);
+	}
+    }
+
+    /*
+     * Assumption: There will be only one cpu who will initialize
+     * a hub. we need to setup the ii and each pi error interrupts.
+     * The SN1 hub (bedrock) has two PI, one for up to two processors.
+     */
+
+    if (cpuid_to_cnodeid(smp_processor_id()) == cnode) { 
+	int generic_intr_mask = PI_ERR_GENERIC; /* These interrupts are sent to only 1 CPU per NODE */
+
+	ASSERT_ALWAYS(kl_error_log[cnode]);
+	ASSERT_ALWAYS(hub_err_count[cnode]);
+	MD_ERR_LOG_INIT(kl_error_log[cnode]);
+
+	/* One for each CPU */
+	recover_error_init(RECOVER_ERROR_TABLE(cnode, 0));
+	recover_error_init(RECOVER_ERROR_TABLE(cnode, 1));
+	recover_error_init(RECOVER_ERROR_TABLE(cnode, 2));
+	recover_error_init(RECOVER_ERROR_TABLE(cnode, 3));
+
+	/*
+	 * Setup error intr masks.
+	 */
+	for(sn=0; sn<NUM_SUBNODES; sn++) {
+		int cpuA_present = REMOTE_HUB_PI_L(nasid, sn, PI_CPU_ENABLE_A);
+		int cpuB_present = REMOTE_HUB_PI_L(nasid, sn, PI_CPU_ENABLE_B);
+
+		if (cpuA_present) {
+			if (cpuB_present) {		/* A && B */
+	    			REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_A,
+					(PI_FATAL_ERR_CPU_B | PI_MISC_ERR_CPU_A|generic_intr_mask));
+	    			REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_B,
+					(PI_FATAL_ERR_CPU_A | PI_MISC_ERR_CPU_B));
+
+			} else {			/* A && !B */
+	    			REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_A,
+					(PI_FATAL_ERR_CPU_A | PI_MISC_ERR_CPU_A|generic_intr_mask));
+			}
+			generic_intr_mask = 0;
+		} else {
+			if (cpuB_present) {		/* !A && B */
+	    			REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_B,
+					(PI_FATAL_ERR_CPU_B | PI_MISC_ERR_CPU_B|generic_intr_mask));
+				generic_intr_mask = 0;
+
+			} else {			/* !A && !B */
+				/* nothing to set up */
+			}
+		}
+	}
+
+	/*
+	 * Turn off UNCAC_UNCORR interrupt in the masks. Anyone interested
+	 * in these errors will peek at the int pend register to see if its
+	 * set.
+	 */ 
+	for(sn=0; sn<NUM_SUBNODES; sn++) {
+		misc = REMOTE_HUB_PI_L(nasid, sn, PI_ERR_INT_MASK_A);
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_A, (misc & ~PI_ERR_UNCAC_UNCORR_A));
+		misc = REMOTE_HUB_PI_L(nasid, sn, PI_ERR_INT_MASK_B);
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_B, (misc & ~PI_ERR_UNCAC_UNCORR_B));
+	}
+
+	/*
+	 * enable all error indicators to turn on, in case of errors.
+	 *
+	 * This is not good on single cpu node boards.
+	 **** LOCAL_HUB_S(PI_SYSAD_ERRCHK_EN, PI_SYSAD_CHECK_ALL);
+	 */
+	for(sn=0; sn<NUM_SUBNODES; sn++) {
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_STATUS1_A_CLR, 0);
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_STATUS1_B_CLR, 0);
+	}
+
+	/* Set up stack for each present processor */
+	for(sn=0; sn<NUM_SUBNODES; sn++) {
+		if (REMOTE_HUB_PI_L(nasid, sn, PI_CPU_PRESENT_A)) {
+	    	SN0_ERROR_LOG(cnode)->el_spool_cur_addr[0] =
+			SN0_ERROR_LOG(cnode)->el_spool_last_addr[0] =
+		    	REMOTE_HUB_PI_L(nasid, sn, PI_ERR_STACK_ADDR_A);
+		}
+	    
+		if (REMOTE_HUB_PI_L(nasid, sn, PI_CPU_PRESENT_B)) {
+	    	SN0_ERROR_LOG(cnode)->el_spool_cur_addr[1] =
+			SN0_ERROR_LOG(cnode)->el_spool_last_addr[1] =
+		    	REMOTE_HUB_PI_L(nasid, sn, PI_ERR_STACK_ADDR_B);
+		}
+	}
+
+
+	PI_SPOOL_SIZE_BYTES = 
+	    ERR_STACK_SIZE_BYTES(REMOTE_HUB_L(nasid, PI_ERR_STACK_SIZE));
+
+#ifdef BRINGUP
+/* BRINGUP: The following code looks like a check to make sure
+the prom set up the error spool correctly for 2 processors.  I
+don't think it is needed.  */
+	for(sn=0; sn<NUM_SUBNODES; sn++) {
+		if (REMOTE_HUB_PI_L(nasid, sn, PI_CPU_PRESENT_B)) {
+			__psunsigned_t addr_a = REMOTE_HUB_PI_L(nasid, sn, PI_ERR_STACK_ADDR_A);
+			__psunsigned_t addr_b = REMOTE_HUB_PI_L(nasid, sn, PI_ERR_STACK_ADDR_B);
+			if ((addr_a & ~0xff) == (addr_b & ~0xff)) {
+			    REMOTE_HUB_PI_S(nasid, sn, PI_ERR_STACK_ADDR_B, 	
+					addr_b + PI_SPOOL_SIZE_BYTES);
+	
+			    SN0_ERROR_LOG(cnode)->el_spool_cur_addr[1] =
+				SN0_ERROR_LOG(cnode)->el_spool_last_addr[1] =
+				    REMOTE_HUB_PI_L(nasid, sn, PI_ERR_STACK_ADDR_B);
+	
+		    }
+		}
+	}
+#endif /* BRINGUP */
+
+	/* programming our own hub. Enable error_int_pend intr.
+	 * If both present, CPU A takes CPU b's error interrupts and any
+	 * generic ones. CPU B takes CPU A error ints.
+	 */
+	if (cause_intr_connect (SRB_ERR_IDX,
+				(intr_func_t)(hubpi_eint_handler),
+				SR_ALL_MASK|SR_IE)) {
+	    cmn_err(ERR_WARN, 
+		    "hub_error_init: cause_intr_connect failed on %d", cnode);
+	}
+    }
+    else {
+	/* programming remote hub. The only valid reason that this
+	 * is called will be on headless hubs. No interrupts 
+	 */
+	for(sn=0; sn<NUM_SUBNODES; sn++) {
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_A, 0); /* not necessary */
+		REMOTE_HUB_PI_S(nasid, sn, PI_ERR_INT_MASK_B, 0); /* not necessary */
+	}
+    }
+#endif /* ajm */
+    /*
+     * Now setup the hub ii and ni error interrupt handler.
+     */
+
+    hubii_eint_init(cnode);
+    hubni_eint_init(cnode);
+
+#ifdef ajm
+    /*** XXX FIXME XXX resolve the following***/
+    /* INT_PEND1 bits set up for one hub only:
+     *	SHUTDOWN_INTR
+     *	MD_COR_ERR_INTR
+     *  COR_ERR_INTR_A and COR_ERR_INTR_B should be sent to the
+     *  appropriate CPU only.
+     */
+
+    if (cnode == 0) {
+	    error_consistency_check.eps_state = 0;
+	    error_consistency_check.eps_cpuid = -1;
+	    spinlock_init(&error_consistency_check.eps_lock, "error_dump_lock");
+    }
+#endif
+
+    nodepda->huberror_ticks = HUB_ERROR_PERIOD;
+    return;
+}
+
+/*
+ * Function	: hubii_eint_init
+ * Parameters	: cnode
+ * Purpose	: to initialize the hub iio error interrupt.
+ * Assumptions	: Called once per hub, by the cpu which will ultimately
+ *			handle this interrupt.
+ * Returns	: None.
+ */
+
+
+void
+hubii_eint_init(cnodeid_t cnode)
+{
+    int			bit, rv;
+    ii_iidsr_u_t    	hubio_eint;
+    hubinfo_t		hinfo; 
+    cpuid_t		intr_cpu;
+    devfs_handle_t 	hub_v;
+    ii_ilcsr_u_t	ilcsr;
+
+    hub_v = (devfs_handle_t)cnodeid_to_vertex(cnode);
+    ASSERT_ALWAYS(hub_v);
+    hubinfo_get(hub_v, &hinfo);
+
+    ASSERT(hinfo);
+    ASSERT(hinfo->h_cnodeid == cnode);
+
+    ilcsr.ii_ilcsr_regval = REMOTE_HUB_L(hinfo->h_nasid, IIO_ILCSR);
+
+    if ((ilcsr.ii_ilcsr_fld_s.i_llp_stat & 0x2) == 0) {
+	/* 
+	 * HUB II link is not up. 
+	 * Just disable LLP, and don't connect any interrupts.
+	 */
+	ilcsr.ii_ilcsr_fld_s.i_llp_en = 0;
+	REMOTE_HUB_S(hinfo->h_nasid, IIO_ILCSR, ilcsr.ii_ilcsr_regval);
+	return;
+    }
+    /* Select a possible interrupt target where there is a free interrupt
+     * bit and also reserve the interrupt bit for this IO error interrupt
+     */
+    intr_cpu = intr_heuristic(hub_v,0,INTRCONNECT_ANYBIT,II_ERRORINT,hub_v,
+			      "HUB IO error interrupt",&bit);
+    if (intr_cpu == CPU_NONE) {
+	printk("hubii_eint_init: intr_reserve_level failed, cnode %d", cnode);
+	return;
+    }
+	
+    rv = intr_connect_level(intr_cpu, bit, 0,(intr_func_t)(NULL),
+			    (void *)(long)hub_v, NULL);
+    synergy_intr_connect(bit, intr_cpu);
+    request_irq(bit_pos_to_irq(bit) + (intr_cpu << 8), hubii_eint_handler, 0, NULL, (void *)hub_v);
+    ASSERT_ALWAYS(rv >= 0);
+    hubio_eint.ii_iidsr_regval = 0;
+    hubio_eint.ii_iidsr_fld_s.i_enable = 1;
+    hubio_eint.ii_iidsr_fld_s.i_level = bit;/* Take the least significant bits*/
+    hubio_eint.ii_iidsr_fld_s.i_node = COMPACT_TO_NASID_NODEID(cnode);
+    hubio_eint.ii_iidsr_fld_s.i_pi_id = cpuid_to_subnode(intr_cpu);
+    REMOTE_HUB_S(hinfo->h_nasid, IIO_IIDSR, hubio_eint.ii_iidsr_regval);
+
+}
+
+void
+hubni_eint_init(cnodeid_t cnode)
+{
+    int intr_bit;
+    cpuid_t targ;
+
+
+    if ((targ = cnodeid_to_cpuid(cnode)) == CPU_NONE)
+	return;
+
+	/* The prom chooses which cpu gets these interrupts, but we
+	*  don't know which one it chose.  We will register all of the 
+	*  cpus to be sure.  This only costs us an irqaction per cpu.
+	*/
+    for (; targ < CPUS_PER_NODE; targ++) {
+	if (!cpu_enabled(targ) ) continue;
+	/* connect the INTEND1 bits. */
+	for (intr_bit = XB_ERROR; intr_bit <= MSC_PANIC_INTR; intr_bit++) {
+		intr_connect_level(targ, intr_bit, II_ERRORINT, NULL, NULL, NULL);
+	}
+	request_irq(SGI_HUB_ERROR_IRQ + (targ << 8), snia_error_intr_handler, 0, NULL, NULL);
+	/* synergy masks are initialized in the prom to enable all interrupts. */
+	/* We'll just leave them that way, here, for these interrupts. */
+    }
+}
+
+
+/*ARGSUSED*/
+void
+hubii_eint_handler (int irq, void *arg, struct pt_regs *ep)
+{
+    devfs_handle_t	hub_v;
+    hubinfo_t		hinfo; 
+    ii_wstat_u_t	wstat;
+    hubreg_t		idsr;
+
+	panic("Hubii interrupt\n");
+#ifdef ajm
+    /*
+     * If the NI has a problem, everyone has a problem.  We shouldn't
+     * even attempt to handle other errors when an NI error is present.
+     */
+    if (check_ni_errors()) {
+	hubni_error_handler("II interrupt", 1);
+	/* NOTREACHED */
+    }
+
+    /* two levels of casting avoids compiler warning.!! */
+    hub_v = (devfs_handle_t)(long)(arg); 
+    ASSERT(hub_v);
+
+    hubinfo_get(hub_v, &hinfo);
+    
+    /* 
+     * Identify the reason for error. 
+     */
+    wstat.ii_wstat_regval = REMOTE_HUB_L(hinfo->h_nasid, IIO_WSTAT);
+
+    if (wstat.ii_wstat_fld_s.w_crazy) {
+	char	*reason;
+	/*
+	 * We can do a couple of things here. 
+	 * Look at the fields TX_MX_RTY/XT_TAIL_TO/XT_CRD_TO to check
+	 * which of these caused the CRAZY bit to be set. 
+	 * You may be able to check if the Link is up really.
+	 */
+	if (wstat.ii_wstat_fld_s.w_tx_mx_rty)
+		reason = "Micro Packet Retry Timeout";
+	else if (wstat.ii_wstat_fld_s.w_xt_tail_to)
+		reason = "Crosstalk Tail Timeout";
+	else if (wstat.ii_wstat_fld_s.w_xt_crd_to)
+		reason = "Crosstalk Credit Timeout";
+	else {
+		hubreg_t	hubii_imem;
+		/*
+		 * Check if widget 0 has been marked as shutdown, or
+		 * if BTE 0/1 has been marked.
+		 */
+		hubii_imem = REMOTE_HUB_L(hinfo->h_nasid, IIO_IMEM);
+		if (hubii_imem & IIO_IMEM_W0ESD)
+			reason = "Hub Widget 0 has been Shutdown";
+		else if (hubii_imem & IIO_IMEM_B0ESD)
+			reason = "BTE 0 has been shutdown";
+		else if (hubii_imem & IIO_IMEM_B1ESD)
+			reason = "BTE 1 has been shutdown";
+		else	reason = "Unknown";
+	
+	}
+	/*
+	 * Note: we may never be able to print this, if the II talking
+	 * to Xbow which hosts the console is dead. 
+	 */
+	printk("Hub %d to Xtalk Link failed (II_ECRAZY) Reason: %s", 
+		hinfo->h_cnodeid, reason);
+    }
+
+    /* 
+     * It's a toss as to which one among PRB/CRB to check first. 
+     * Current decision is based on the severity of the errors. 
+     * IO CRB errors tend to be more severe than PRB errors.
+     *
+     * It is possible for BTE errors to have been handled already, so we
+     * may not see any errors handled here. 
+     */
+    (void)hubiio_crb_error_handler(hub_v, hinfo);
+    (void)hubiio_prb_error_handler(hub_v, hinfo);
+    /*
+     * If we reach here, it indicates crb/prb handlers successfully
+     * handled the error. So, re-enable II to send more interrupt
+     * and return.
+     */
+    REMOTE_HUB_S(hinfo->h_nasid, IIO_IECLR, 0xffffff);
+    idsr = REMOTE_HUB_L(hinfo->h_nasid, IIO_IIDSR) & ~IIO_IIDSR_SENT_MASK;
+    REMOTE_HUB_S(hinfo->h_nasid, IIO_IIDSR, idsr);
+#endif /* ajm */
+}

FUNET's LINUX-ADM group, linux-adm@nic.funet.fi
TCL-scripts by Sam Shen (who was at: slshen@lbl.gov)