2 * Copyright (C) 2001 Dave Engebretsen IBM Corporation
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 * 2001/09/21 : engebret : Created with minimal EPOW and HW exception support.
24 #include <linux/errno.h>
25 #include <linux/threads.h>
26 #include <linux/kernel_stat.h>
27 #include <linux/signal.h>
28 #include <linux/sched.h>
29 #include <linux/ioport.h>
30 #include <linux/interrupt.h>
31 #include <linux/timex.h>
32 #include <linux/init.h>
33 #include <linux/slab.h>
34 #include <linux/pci.h>
35 #include <linux/delay.h>
36 #include <linux/irq.h>
37 #include <linux/random.h>
38 #include <linux/sysrq.h>
39 #include <linux/bitops.h>
41 #include <asm/uaccess.h>
42 #include <asm/system.h>
44 #include <asm/pgtable.h>
46 #include <asm/cache.h>
48 #include <asm/ptrace.h>
49 #include <asm/machdep.h>
53 static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX];
54 static DEFINE_SPINLOCK(ras_log_buf_lock);
56 char mce_data_buf[RTAS_ERROR_LOG_MAX]
58 /* This is true if we are using the firmware NMI handler (typically LPAR) */
59 extern int fwnmi_active;
61 static int ras_get_sensor_state_token;
62 static int ras_check_exception_token;
64 #define EPOW_SENSOR_TOKEN 9
65 #define EPOW_SENSOR_INDEX 0
66 #define RAS_VECTOR_OFFSET 0x500
68 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id,
69 struct pt_regs * regs);
70 static irqreturn_t ras_error_interrupt(int irq, void *dev_id,
71 struct pt_regs * regs);
75 static void request_ras_irqs(struct device_node *np, char *propname,
76 irqreturn_t (*handler)(int, void *, struct pt_regs *),
79 unsigned int *ireg, len, i;
82 ireg = (unsigned int *)get_property(np, propname, &len);
85 n_intr = prom_n_intr_cells(np);
86 len /= n_intr * sizeof(*ireg);
88 for (i = 0; i < len; i++) {
89 virq = virt_irq_create_mapping(*ireg);
91 printk(KERN_ERR "Unable to allocate interrupt "
92 "number for %s\n", np->full_name);
95 if (request_irq(irq_offset_up(virq), handler, 0, name, NULL)) {
96 printk(KERN_ERR "Unable to request interrupt %d for "
97 "%s\n", irq_offset_up(virq), np->full_name);
105 * Initialize handlers for the set of interrupts caused by hardware errors
106 * and power system events.
108 static int __init init_ras_IRQ(void)
110 struct device_node *np;
112 ras_get_sensor_state_token = rtas_token("get-sensor-state");
113 ras_check_exception_token = rtas_token("check-exception");
115 /* Internal Errors */
116 np = of_find_node_by_path("/event-sources/internal-errors");
118 request_ras_irqs(np, "open-pic-interrupt", ras_error_interrupt,
120 request_ras_irqs(np, "interrupts", ras_error_interrupt,
126 np = of_find_node_by_path("/event-sources/epow-events");
128 request_ras_irqs(np, "open-pic-interrupt", ras_epow_interrupt,
130 request_ras_irqs(np, "interrupts", ras_epow_interrupt,
137 __initcall(init_ras_IRQ);
140 * Handle power subsystem events (EPOW).
142 * Presently we just log the event has occurred. This should be fixed
143 * to examine the type of power failure and take appropriate action where
144 * the time horizon permits something useful to be done.
147 ras_epow_interrupt(int irq, void *dev_id, struct pt_regs * regs)
149 int status = 0xdeadbeef;
153 status = rtas_call(ras_get_sensor_state_token, 2, 2, &state,
154 EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX);
157 critical = 1; /* Time Critical */
161 spin_lock(&ras_log_buf_lock);
163 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
165 virt_irq_to_real(irq_offset_down(irq)),
166 RTAS_EPOW_WARNING | RTAS_POWERMGM_EVENTS,
167 critical, __pa(&ras_log_buf),
168 rtas_get_error_log_max());
170 udbg_printf("EPOW <0x%lx 0x%x 0x%x>\n",
171 *((unsigned long *)&ras_log_buf), status, state);
172 printk(KERN_WARNING "EPOW <0x%lx 0x%x 0x%x>\n",
173 *((unsigned long *)&ras_log_buf), status, state);
175 /* format and print the extended information */
176 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, 0);
178 spin_unlock(&ras_log_buf_lock);
183 * Handle hardware error interrupts.
185 * RTAS check-exception is called to collect data on the exception. If
186 * the error is deemed recoverable, we log a warning and return.
187 * For nonrecoverable errors, an error is logged and we stop all processing
188 * as quickly as possible in order to prevent propagation of the failure.
191 ras_error_interrupt(int irq, void *dev_id, struct pt_regs * regs)
193 struct rtas_error_log *rtas_elog;
194 int status = 0xdeadbeef;
197 spin_lock(&ras_log_buf_lock);
199 status = rtas_call(ras_check_exception_token, 6, 1, NULL,
201 virt_irq_to_real(irq_offset_down(irq)),
202 RTAS_INTERNAL_ERROR, 1 /*Time Critical */,
204 rtas_get_error_log_max());
206 rtas_elog = (struct rtas_error_log *)ras_log_buf;
208 if ((status == 0) && (rtas_elog->severity >= RTAS_SEVERITY_ERROR_SYNC))
213 /* format and print the extended information */
214 log_error(ras_log_buf, ERR_TYPE_RTAS_LOG, fatal);
217 udbg_printf("Fatal HW Error <0x%lx 0x%x>\n",
218 *((unsigned long *)&ras_log_buf), status);
219 printk(KERN_EMERG "Error: Fatal hardware error <0x%lx 0x%x>\n",
220 *((unsigned long *)&ras_log_buf), status);
223 /* Don't actually power off when debugging so we can test
224 * without actually failing while injecting errors.
225 * Error data will not be logged to syslog.
230 udbg_printf("Recoverable HW Error <0x%lx 0x%x>\n",
231 *((unsigned long *)&ras_log_buf), status);
233 "Warning: Recoverable hardware error <0x%lx 0x%x>\n",
234 *((unsigned long *)&ras_log_buf), status);
237 spin_unlock(&ras_log_buf_lock);
241 /* Get the error information for errors coming through the
242 * FWNMI vectors. The pt_regs' r3 will be updated to reflect
243 * the actual r3 if possible, and a ptr to the error log entry
244 * will be returned if found.
246 * The mce_data_buf does not have any locks or protection around it,
247 * if a second machine check comes in, or a system reset is done
248 * before we have logged the error, then we will get corruption in the
249 * error log. This is preferable over holding off on calling
250 * ibm,nmi-interlock which would result in us checkstopping if a
251 * second machine check did come in.
253 static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)
255 unsigned long errdata = regs->gpr[3];
256 struct rtas_error_log *errhdr = NULL;
257 unsigned long *savep;
259 if ((errdata >= 0x7000 && errdata < 0x7fff0) ||
260 (errdata >= rtas.base && errdata < rtas.base + rtas.size - 16)) {
261 savep = __va(errdata);
262 regs->gpr[3] = savep[0]; /* restore original r3 */
263 memset(mce_data_buf, 0, RTAS_ERROR_LOG_MAX);
264 memcpy(mce_data_buf, (char *)(savep + 1), RTAS_ERROR_LOG_MAX);
265 errhdr = (struct rtas_error_log *)mce_data_buf;
267 printk("FWNMI: corrupt r3\n");
272 /* Call this when done with the data returned by FWNMI_get_errinfo.
273 * It will release the saved data area for other CPUs in the
274 * partition to receive FWNMI errors.
276 static void fwnmi_release_errinfo(void)
278 int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL);
280 printk("FWNMI: nmi-interlock failed: %d\n", ret);
283 void pSeries_system_reset_exception(struct pt_regs *regs)
286 struct rtas_error_log *errhdr = fwnmi_get_errinfo(regs);
288 /* XXX Should look at FWNMI information */
290 fwnmi_release_errinfo();
295 * See if we can recover from a machine check exception.
296 * This is only called on power4 (or above) and only via
297 * the Firmware Non-Maskable Interrupts (fwnmi) handler
298 * which provides the error analysis for us.
300 * Return 1 if corrected (or delivered a signal).
301 * Return 0 if there is nothing we can do.
303 static int recover_mce(struct pt_regs *regs, struct rtas_error_log * err)
307 if (err->disposition == RTAS_DISP_FULLY_RECOVERED) {
308 /* Platform corrected itself */
310 } else if ((regs->msr & MSR_RI) &&
312 err->severity == RTAS_SEVERITY_ERROR_SYNC &&
313 err->disposition == RTAS_DISP_NOT_RECOVERED &&
314 err->target == RTAS_TARGET_MEMORY &&
315 err->type == RTAS_TYPE_ECC_UNCORR &&
316 !(current->pid == 0 || current->pid == 1)) {
317 /* Kill off a user process with an ECC error */
318 printk(KERN_ERR "MCE: uncorrectable ecc error for pid %d\n",
320 /* XXX something better for ECC error? */
321 _exception(SIGBUS, regs, BUS_ADRERR, regs->nip);
325 log_error((char *)err, ERR_TYPE_RTAS_LOG, !nonfatal);
331 * Handle a machine check.
333 * Note that on Power 4 and beyond Firmware Non-Maskable Interrupts (fwnmi)
334 * should be present. If so the handler which called us tells us if the
335 * error was recovered (never true if RI=0).
337 * On hardware prior to Power 4 these exceptions were asynchronous which
338 * means we can't tell exactly where it occurred and so we can't recover.
340 int pSeries_machine_check_exception(struct pt_regs *regs)
342 struct rtas_error_log *errp;
345 errp = fwnmi_get_errinfo(regs);
346 fwnmi_release_errinfo();
347 if (errp && recover_mce(regs, errp))