arch/ia64/sn/kernel/huberror.c

   1 /*
   2  * This file is subject to the terms and conditions of the GNU General Public
   3  * License.  See the file "COPYING" in the main directory of this archive
   4  * for more details.
   5  *
   6  * Copyright (C) 1992 - 1997, 2000,2002-2004 Silicon Graphics, Inc. All rights reserved.
   7  */
   8
   9 #include <linux/types.h>
  10 #include <linux/interrupt.h>
  11 #include <linux/pci.h>
  12 #include <asm/delay.h>
  13 #include <asm/sn/sn_sal.h>
  14 #include "ioerror.h"
  15 #include <asm/sn/addrs.h>
  16 #include <asm/sn/shubio.h>
  17 #include <asm/sn/geo.h>
  18 #include "xtalk/xwidgetdev.h"
  19 #include "xtalk/hubdev.h"
  20 #include <asm/sn/bte.h>
  21
  22 void hubiio_crb_error_handler(struct hubdev_info *hubdev_info);
  23 extern void bte_crb_error_handler(cnodeid_t, int, int, ioerror_t *,
  24                                   int);
  25 static irqreturn_t hub_eint_handler(int irq, void *arg, struct pt_regs *ep)
  26 {
  27         struct hubdev_info *hubdev_info;
  28         struct ia64_sal_retval ret_stuff;
  29         nasid_t nasid;
  30
  31         ret_stuff.status = 0;
  32         ret_stuff.v0 = 0;
  33         hubdev_info = (struct hubdev_info *)arg;
  34         nasid = hubdev_info->hdi_nasid;
  35         SAL_CALL_NOLOCK(ret_stuff, SN_SAL_HUB_ERROR_INTERRUPT,
  36                         (u64) nasid, 0, 0, 0, 0, 0, 0);
  37
  38         if ((int)ret_stuff.v0)
  39                 panic("hubii_eint_handler(): Fatal TIO Error");
  40
  41         if (!(nasid & 1)) /* Not a TIO, handle CRB errors */
  42                 (void)hubiio_crb_error_handler(hubdev_info);
  43
  44         return IRQ_HANDLED;
  45 }
  46
  47 /*
  48  * Free the hub CRB "crbnum" which encountered an error.
  49  * Assumption is, error handling was successfully done,
  50  * and we now want to return the CRB back to Hub for normal usage.
  51  *
  52  * In order to free the CRB, all that's needed is to de-allocate it
  53  *
  54  * Assumption:
  55  *      No other processor is mucking around with the hub control register.
  56  *      So, upper layer has to single thread this.
  57  */
  58 void hubiio_crb_free(struct hubdev_info *hubdev_info, int crbnum)
  59 {
  60         ii_icrb0_b_u_t icrbb;
  61
  62         /*
  63          * The hardware does NOT clear the mark bit, so it must get cleared
  64          * here to be sure the error is not processed twice.
  65          */
  66         icrbb.ii_icrb0_b_regval = REMOTE_HUB_L(hubdev_info->hdi_nasid,
  67                                                IIO_ICRB_B(crbnum));
  68         icrbb.b_mark = 0;
  69         REMOTE_HUB_S(hubdev_info->hdi_nasid, IIO_ICRB_B(crbnum),
  70                      icrbb.ii_icrb0_b_regval);
  71         /*
  72          * Deallocate the register wait till hub indicates it's done.
  73          */
  74         REMOTE_HUB_S(hubdev_info->hdi_nasid, IIO_ICDR, (IIO_ICDR_PND | crbnum));
  75         while (REMOTE_HUB_L(hubdev_info->hdi_nasid, IIO_ICDR) & IIO_ICDR_PND)
  76                 udelay(1);
  77
  78 }
  79
  80 /*
  81  * hubiio_crb_error_handler
  82  *
  83  *      This routine gets invoked when a hub gets an error
  84  *      interrupt. So, the routine is running in interrupt context
  85  *      at error interrupt level.
  86  * Action:
  87  *      It's responsible for identifying ALL the CRBs that are marked
  88  *      with error, and process them.
  89  *
  90  *      If you find the CRB that's marked with error, map this to the
  91  *      reason it caused error, and invoke appropriate error handler.
  92  *
  93  *      XXX Be aware of the information in the context register.
  94  *
  95  * NOTE:
  96  *      Use REMOTE_HUB_* macro instead of LOCAL_HUB_* so that the interrupt
  97  *      handler can be run on any node. (not necessarily the node
  98  *      corresponding to the hub that encountered error).
  99  */
 100
 101 void hubiio_crb_error_handler(struct hubdev_info *hubdev_info)
 102 {
 103         nasid_t nasid;
 104         ii_icrb0_a_u_t icrba;   /* II CRB Register A */
 105         ii_icrb0_b_u_t icrbb;   /* II CRB Register B */
 106         ii_icrb0_c_u_t icrbc;   /* II CRB Register C */
 107         ii_icrb0_d_u_t icrbd;   /* II CRB Register D */
 108         ii_icrb0_e_u_t icrbe;   /* II CRB Register D */
 109         int i;
 110         int num_errors = 0;     /* Num of errors handled */
 111         ioerror_t ioerror;
 112
 113         nasid = hubdev_info->hdi_nasid;
 114
 115         /*
 116          * XXX - Add locking for any recovery actions
 117          */
 118         /*
 119          * Scan through all CRBs in the Hub, and handle the errors
 120          * in any of the CRBs marked.
 121          */
 122         for (i = 0; i < IIO_NUM_CRBS; i++) {
 123                 /* Check this crb entry to see if it is in error. */
 124                 icrbb.ii_icrb0_b_regval = REMOTE_HUB_L(nasid, IIO_ICRB_B(i));
 125
 126                 if (icrbb.b_mark == 0) {
 127                         continue;
 128                 }
 129
 130                 icrba.ii_icrb0_a_regval = REMOTE_HUB_L(nasid, IIO_ICRB_A(i));
 131
 132                 IOERROR_INIT(&ioerror);
 133
 134                 /* read other CRB error registers. */
 135                 icrbc.ii_icrb0_c_regval = REMOTE_HUB_L(nasid, IIO_ICRB_C(i));
 136                 icrbd.ii_icrb0_d_regval = REMOTE_HUB_L(nasid, IIO_ICRB_D(i));
 137                 icrbe.ii_icrb0_e_regval = REMOTE_HUB_L(nasid, IIO_ICRB_E(i));
 138
 139                 IOERROR_SETVALUE(&ioerror, errortype, icrbb.b_ecode);
 140
 141                 /* Check if this error is due to BTE operation,
 142                  * and handle it separately.
 143                  */
 144                 if (icrbd.d_bteop ||
 145                     ((icrbb.b_initiator == IIO_ICRB_INIT_BTE0 ||
 146                       icrbb.b_initiator == IIO_ICRB_INIT_BTE1) &&
 147                      (icrbb.b_imsgtype == IIO_ICRB_IMSGT_BTE ||
 148                       icrbb.b_imsgtype == IIO_ICRB_IMSGT_SN1NET))) {
 149
 150                         int bte_num;
 151
 152                         if (icrbd.d_bteop)
 153                                 bte_num = icrbc.c_btenum;
 154                         else    /* b_initiator bit 2 gives BTE number */
 155                                 bte_num = (icrbb.b_initiator & 0x4) >> 2;
 156
 157                         hubiio_crb_free(hubdev_info, i);
 158
 159                         bte_crb_error_handler(nasid_to_cnodeid(nasid), bte_num,
 160                                               i, &ioerror, icrbd.d_bteop);
 161                         num_errors++;
 162                         continue;
 163                 }
 164         }
 165 }
 166
 167 /*
 168  * Function     : hub_error_init
 169  * Purpose      : initialize the error handling requirements for a given hub.
 170  * Parameters   : cnode, the compact nodeid.
 171  * Assumptions  : Called only once per hub, either by a local cpu. Or by a
 172  *                      remote cpu, when this hub is headless.(cpuless)
 173  * Returns      : None
 174  */
 175 void hub_error_init(struct hubdev_info *hubdev_info)
 176 {
 177         if (request_irq(SGI_II_ERROR, (void *)hub_eint_handler, SA_SHIRQ,
 178                         "SN_hub_error", (void *)hubdev_info))
 179                 printk("hub_error_init: Failed to request_irq for 0x%p\n",
 180                     hubdev_info);
 181         return;
 182 }
 183
 184
 185 /*
 186  * Function     : ice_error_init
 187  * Purpose      : initialize the error handling requirements for a given tio.
 188  * Parameters   : cnode, the compact nodeid.
 189  * Assumptions  : Called only once per tio.
 190  * Returns      : None
 191  */
 192 void ice_error_init(struct hubdev_info *hubdev_info)
 193 {
 194         if (request_irq
 195             (SGI_TIO_ERROR, (void *)hub_eint_handler, SA_SHIRQ, "SN_TIO_error",
 196              (void *)hubdev_info))
 197                 printk("ice_error_init: request_irq() error hubdev_info 0x%p\n",
 198                        hubdev_info);
 199         return;
 200 }
 201