[IA64-SGI] Shub2 BTE support - BTE recovery code
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / arch / ia64 / sn / kernel / huberror.c
blob5c39b43ba3c029604103bb1757dbe3d5db1faa5c
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.
6 * Copyright (C) 1992 - 1997, 2000,2002-2005 Silicon Graphics, Inc. All rights reserved.
7 */
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>
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)
27 struct hubdev_info *hubdev_info;
28 struct ia64_sal_retval ret_stuff;
29 nasid_t nasid;
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);
38 if ((int)ret_stuff.v0)
39 panic("hubii_eint_handler(): Fatal TIO Error");
41 if (is_shub1()) {
42 if (!(nasid & 1)) /* Not a TIO, handle CRB errors */
43 (void)hubiio_crb_error_handler(hubdev_info);
44 } else
45 bte_error_handler((unsigned long)NODEPDA(nasid_to_cnodeid(nasid)));
47 return IRQ_HANDLED;
51 * Free the hub CRB "crbnum" which encountered an error.
52 * Assumption is, error handling was successfully done,
53 * and we now want to return the CRB back to Hub for normal usage.
55 * In order to free the CRB, all that's needed is to de-allocate it
57 * Assumption:
58 * No other processor is mucking around with the hub control register.
59 * So, upper layer has to single thread this.
61 void hubiio_crb_free(struct hubdev_info *hubdev_info, int crbnum)
63 ii_icrb0_b_u_t icrbb;
66 * The hardware does NOT clear the mark bit, so it must get cleared
67 * here to be sure the error is not processed twice.
69 icrbb.ii_icrb0_b_regval = REMOTE_HUB_L(hubdev_info->hdi_nasid,
70 IIO_ICRB_B(crbnum));
71 icrbb.b_mark = 0;
72 REMOTE_HUB_S(hubdev_info->hdi_nasid, IIO_ICRB_B(crbnum),
73 icrbb.ii_icrb0_b_regval);
75 * Deallocate the register wait till hub indicates it's done.
77 REMOTE_HUB_S(hubdev_info->hdi_nasid, IIO_ICDR, (IIO_ICDR_PND | crbnum));
78 while (REMOTE_HUB_L(hubdev_info->hdi_nasid, IIO_ICDR) & IIO_ICDR_PND)
79 udelay(1);
84 * hubiio_crb_error_handler
86 * This routine gets invoked when a hub gets an error
87 * interrupt. So, the routine is running in interrupt context
88 * at error interrupt level.
89 * Action:
90 * It's responsible for identifying ALL the CRBs that are marked
91 * with error, and process them.
93 * If you find the CRB that's marked with error, map this to the
94 * reason it caused error, and invoke appropriate error handler.
96 * XXX Be aware of the information in the context register.
98 * NOTE:
99 * Use REMOTE_HUB_* macro instead of LOCAL_HUB_* so that the interrupt
100 * handler can be run on any node. (not necessarily the node
101 * corresponding to the hub that encountered error).
104 void hubiio_crb_error_handler(struct hubdev_info *hubdev_info)
106 nasid_t nasid;
107 ii_icrb0_a_u_t icrba; /* II CRB Register A */
108 ii_icrb0_b_u_t icrbb; /* II CRB Register B */
109 ii_icrb0_c_u_t icrbc; /* II CRB Register C */
110 ii_icrb0_d_u_t icrbd; /* II CRB Register D */
111 ii_icrb0_e_u_t icrbe; /* II CRB Register D */
112 int i;
113 int num_errors = 0; /* Num of errors handled */
114 ioerror_t ioerror;
116 nasid = hubdev_info->hdi_nasid;
119 * XXX - Add locking for any recovery actions
122 * Scan through all CRBs in the Hub, and handle the errors
123 * in any of the CRBs marked.
125 for (i = 0; i < IIO_NUM_CRBS; i++) {
126 /* Check this crb entry to see if it is in error. */
127 icrbb.ii_icrb0_b_regval = REMOTE_HUB_L(nasid, IIO_ICRB_B(i));
129 if (icrbb.b_mark == 0) {
130 continue;
133 icrba.ii_icrb0_a_regval = REMOTE_HUB_L(nasid, IIO_ICRB_A(i));
135 IOERROR_INIT(&ioerror);
137 /* read other CRB error registers. */
138 icrbc.ii_icrb0_c_regval = REMOTE_HUB_L(nasid, IIO_ICRB_C(i));
139 icrbd.ii_icrb0_d_regval = REMOTE_HUB_L(nasid, IIO_ICRB_D(i));
140 icrbe.ii_icrb0_e_regval = REMOTE_HUB_L(nasid, IIO_ICRB_E(i));
142 IOERROR_SETVALUE(&ioerror, errortype, icrbb.b_ecode);
144 /* Check if this error is due to BTE operation,
145 * and handle it separately.
147 if (icrbd.d_bteop ||
148 ((icrbb.b_initiator == IIO_ICRB_INIT_BTE0 ||
149 icrbb.b_initiator == IIO_ICRB_INIT_BTE1) &&
150 (icrbb.b_imsgtype == IIO_ICRB_IMSGT_BTE ||
151 icrbb.b_imsgtype == IIO_ICRB_IMSGT_SN1NET))) {
153 int bte_num;
155 if (icrbd.d_bteop)
156 bte_num = icrbc.c_btenum;
157 else /* b_initiator bit 2 gives BTE number */
158 bte_num = (icrbb.b_initiator & 0x4) >> 2;
160 hubiio_crb_free(hubdev_info, i);
162 bte_crb_error_handler(nasid_to_cnodeid(nasid), bte_num,
163 i, &ioerror, icrbd.d_bteop);
164 num_errors++;
165 continue;
171 * Function : hub_error_init
172 * Purpose : initialize the error handling requirements for a given hub.
173 * Parameters : cnode, the compact nodeid.
174 * Assumptions : Called only once per hub, either by a local cpu. Or by a
175 * remote cpu, when this hub is headless.(cpuless)
176 * Returns : None
178 void hub_error_init(struct hubdev_info *hubdev_info)
180 if (request_irq(SGI_II_ERROR, (void *)hub_eint_handler, SA_SHIRQ,
181 "SN_hub_error", (void *)hubdev_info))
182 printk("hub_error_init: Failed to request_irq for 0x%p\n",
183 hubdev_info);
184 return;
189 * Function : ice_error_init
190 * Purpose : initialize the error handling requirements for a given tio.
191 * Parameters : cnode, the compact nodeid.
192 * Assumptions : Called only once per tio.
193 * Returns : None
195 void ice_error_init(struct hubdev_info *hubdev_info)
197 if (request_irq
198 (SGI_TIO_ERROR, (void *)hub_eint_handler, SA_SHIRQ, "SN_TIO_error",
199 (void *)hubdev_info))
200 printk("ice_error_init: request_irq() error hubdev_info 0x%p\n",
201 hubdev_info);
202 return;