[PATCH] fix include order in mthca_memfree.c
[linux-2.6] / drivers / s390 / s390mach.c
1 /*
2  *  drivers/s390/s390mach.c
3  *   S/390 machine check handler
4  *
5  *  S390 version
6  *    Copyright (C) 2000 IBM Deutschland Entwicklung GmbH, IBM Corporation
7  *    Author(s): Ingo Adlung (adlung@de.ibm.com)
8  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
9  */
10
11 #include <linux/config.h>
12 #include <linux/init.h>
13 #include <linux/sched.h>
14 #include <linux/errno.h>
15 #include <linux/workqueue.h>
16
17 #include <asm/lowcore.h>
18
19 #include "s390mach.h"
20
21 #define DBG printk
22 // #define DBG(args,...) do {} while (0);
23
24 static struct semaphore m_sem;
25
26 extern int css_process_crw(int);
27 extern int chsc_process_crw(void);
28 extern int chp_process_crw(int, int);
29 extern void css_reiterate_subchannels(void);
30
31 extern struct workqueue_struct *slow_path_wq;
32 extern struct work_struct slow_path_work;
33
34 static void
35 s390_handle_damage(char *msg)
36 {
37         printk(KERN_EMERG "%s\n", msg);
38 #ifdef CONFIG_SMP
39         smp_send_stop();
40 #endif
41         disabled_wait((unsigned long) __builtin_return_address(0));
42 }
43
44 /*
45  * Retrieve CRWs and call function to handle event.
46  *
47  * Note : we currently process CRWs for io and chsc subchannels only
48  */
49 static int
50 s390_collect_crw_info(void *param)
51 {
52         struct crw crw;
53         int ccode, ret, slow;
54         struct semaphore *sem;
55
56         sem = (struct semaphore *)param;
57         /* Set a nice name. */
58         daemonize("kmcheck");
59 repeat:
60         down_interruptible(sem);
61         slow = 0;
62         while (1) {
63                 ccode = stcrw(&crw);
64                 if (ccode != 0)
65                         break;
66                 DBG(KERN_DEBUG "crw_info : CRW reports slct=%d, oflw=%d, "
67                     "chn=%d, rsc=%X, anc=%d, erc=%X, rsid=%X\n",
68                     crw.slct, crw.oflw, crw.chn, crw.rsc, crw.anc,
69                     crw.erc, crw.rsid);
70                 /* Check for overflows. */
71                 if (crw.oflw) {
72                         pr_debug("%s: crw overflow detected!\n", __FUNCTION__);
73                         css_reiterate_subchannels();
74                         slow = 1;
75                         continue;
76                 }
77                 switch (crw.rsc) {
78                 case CRW_RSC_SCH:
79                         pr_debug("source is subchannel %04X\n", crw.rsid);
80                         ret = css_process_crw (crw.rsid);
81                         if (ret == -EAGAIN)
82                                 slow = 1;
83                         break;
84                 case CRW_RSC_MONITOR:
85                         pr_debug("source is monitoring facility\n");
86                         break;
87                 case CRW_RSC_CPATH:
88                         pr_debug("source is channel path %02X\n", crw.rsid);
89                         switch (crw.erc) {
90                         case CRW_ERC_IPARM: /* Path has come. */
91                                 ret = chp_process_crw(crw.rsid, 1);
92                                 break;
93                         case CRW_ERC_PERRI: /* Path has gone. */
94                         case CRW_ERC_PERRN:
95                                 ret = chp_process_crw(crw.rsid, 0);
96                                 break;
97                         default:
98                                 pr_debug("Don't know how to handle erc=%x\n",
99                                          crw.erc);
100                                 ret = 0;
101                         }
102                         if (ret == -EAGAIN)
103                                 slow = 1;
104                         break;
105                 case CRW_RSC_CONFIG:
106                         pr_debug("source is configuration-alert facility\n");
107                         break;
108                 case CRW_RSC_CSS:
109                         pr_debug("source is channel subsystem\n");
110                         ret = chsc_process_crw();
111                         if (ret == -EAGAIN)
112                                 slow = 1;
113                         break;
114                 default:
115                         pr_debug("unknown source\n");
116                         break;
117                 }
118         }
119         if (slow)
120                 queue_work(slow_path_wq, &slow_path_work);
121         goto repeat;
122         return 0;
123 }
124
125 /*
126  * machine check handler.
127  */
128 void
129 s390_do_machine_check(void)
130 {
131         struct mci *mci;
132
133         mci = (struct mci *) &S390_lowcore.mcck_interruption_code;
134
135         if (mci->sd)            /* system damage */
136                 s390_handle_damage("received system damage machine check\n");
137
138         if (mci->pd)            /* instruction processing damage */
139                 s390_handle_damage("received instruction processing "
140                                    "damage machine check\n");
141
142         if (mci->se)            /* storage error uncorrected */
143                 s390_handle_damage("received storage error uncorrected "
144                                    "machine check\n");
145
146         if (mci->sc)            /* storage error corrected */
147                 printk(KERN_WARNING
148                        "received storage error corrected machine check\n");
149
150         if (mci->ke)            /* storage key-error uncorrected */
151                 s390_handle_damage("received storage key-error uncorrected "
152                                    "machine check\n");
153
154         if (mci->ds && mci->fa) /* storage degradation */
155                 s390_handle_damage("received storage degradation machine "
156                                    "check\n");
157
158         if (mci->cp)            /* channel report word pending */
159                 up(&m_sem);
160
161 #ifdef CONFIG_MACHCHK_WARNING
162 /*
163  * The warning may remain for a prolonged period on the bare iron.
164  * (actually till the machine is powered off, or until the problem is gone)
165  * So we just stop listening for the WARNING MCH and prevent continuously
166  * being interrupted.  One caveat is however, that we must do this per
167  * processor and cannot use the smp version of ctl_clear_bit().
168  * On VM we only get one interrupt per virtally presented machinecheck.
169  * Though one suffices, we may get one interrupt per (virtual) processor.
170  */
171         if (mci->w) {   /* WARNING pending ? */
172                 static int mchchk_wng_posted = 0;
173                 /*
174                  * Use single machine clear, as we cannot handle smp right now
175                  */
176                 __ctl_clear_bit(14, 24);        /* Disable WARNING MCH */
177                 if (xchg(&mchchk_wng_posted, 1) == 0)
178                         kill_proc(1, SIGPWR, 1);
179         }
180 #endif
181 }
182
183 /*
184  * s390_init_machine_check
185  *
186  * initialize machine check handling
187  */
188 static int
189 machine_check_init(void)
190 {
191         init_MUTEX_LOCKED(&m_sem);
192         ctl_clear_bit(14, 25);  /* disable damage MCH */
193         ctl_set_bit(14, 26);    /* enable degradation MCH */
194         ctl_set_bit(14, 27);    /* enable system recovery MCH */
195 #ifdef CONFIG_MACHCHK_WARNING
196         ctl_set_bit(14, 24);    /* enable warning MCH */
197 #endif
198         return 0;
199 }
200
201 /*
202  * Initialize the machine check handler really early to be able to
203  * catch all machine checks that happen during boot
204  */
205 arch_initcall(machine_check_init);
206
207 /*
208  * Machine checks for the channel subsystem must be enabled
209  * after the channel subsystem is initialized
210  */
211 static int __init
212 machine_check_crw_init (void)
213 {
214         kernel_thread(s390_collect_crw_info, &m_sem, CLONE_FS|CLONE_FILES);
215         ctl_set_bit(14, 28);    /* enable channel report MCH */
216         return 0;
217 }
218
219 device_initcall (machine_check_crw_init);