xref: /openbmc/linux/arch/powerpc/platforms/cell/ras.c (revision f3539c12)
1 /*
2  * Copyright 2006-2008, IBM Corporation.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public License
6  * as published by the Free Software Foundation; either version
7  * 2 of the License, or (at your option) any later version.
8  */
9 
10 #undef DEBUG
11 
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/smp.h>
16 #include <linux/reboot.h>
17 #include <linux/kexec.h>
18 #include <linux/crash_dump.h>
19 
20 #include <asm/kexec.h>
21 #include <asm/reg.h>
22 #include <asm/io.h>
23 #include <asm/prom.h>
24 #include <asm/machdep.h>
25 #include <asm/rtas.h>
26 #include <asm/cell-regs.h>
27 
28 #include "ras.h"
29 
30 
31 static void dump_fir(int cpu)
32 {
33 	struct cbe_pmd_regs __iomem *pregs = cbe_get_cpu_pmd_regs(cpu);
34 	struct cbe_iic_regs __iomem *iregs = cbe_get_cpu_iic_regs(cpu);
35 
36 	if (pregs == NULL)
37 		return;
38 
39 	/* Todo: do some nicer parsing of bits and based on them go down
40 	 * to other sub-units FIRs and not only IIC
41 	 */
42 	printk(KERN_ERR "Global Checkstop FIR    : 0x%016llx\n",
43 	       in_be64(&pregs->checkstop_fir));
44 	printk(KERN_ERR "Global Recoverable FIR  : 0x%016llx\n",
45 	       in_be64(&pregs->checkstop_fir));
46 	printk(KERN_ERR "Global MachineCheck FIR : 0x%016llx\n",
47 	       in_be64(&pregs->spec_att_mchk_fir));
48 
49 	if (iregs == NULL)
50 		return;
51 	printk(KERN_ERR "IOC FIR                 : 0x%016llx\n",
52 	       in_be64(&iregs->ioc_fir));
53 
54 }
55 
56 void cbe_system_error_exception(struct pt_regs *regs)
57 {
58 	int cpu = smp_processor_id();
59 
60 	printk(KERN_ERR "System Error Interrupt on CPU %d !\n", cpu);
61 	dump_fir(cpu);
62 	dump_stack();
63 }
64 
65 void cbe_maintenance_exception(struct pt_regs *regs)
66 {
67 	int cpu = smp_processor_id();
68 
69 	/*
70 	 * Nothing implemented for the maintenance interrupt at this point
71 	 */
72 
73 	printk(KERN_ERR "Unhandled Maintenance interrupt on CPU %d !\n", cpu);
74 	dump_stack();
75 }
76 
77 void cbe_thermal_exception(struct pt_regs *regs)
78 {
79 	int cpu = smp_processor_id();
80 
81 	/*
82 	 * Nothing implemented for the thermal interrupt at this point
83 	 */
84 
85 	printk(KERN_ERR "Unhandled Thermal interrupt on CPU %d !\n", cpu);
86 	dump_stack();
87 }
88 
89 static int cbe_machine_check_handler(struct pt_regs *regs)
90 {
91 	int cpu = smp_processor_id();
92 
93 	printk(KERN_ERR "Machine Check Interrupt on CPU %d !\n", cpu);
94 	dump_fir(cpu);
95 
96 	/* No recovery from this code now, lets continue */
97 	return 0;
98 }
99 
100 struct ptcal_area {
101 	struct list_head list;
102 	int nid;
103 	int order;
104 	struct page *pages;
105 };
106 
107 static LIST_HEAD(ptcal_list);
108 
109 static int ptcal_start_tok, ptcal_stop_tok;
110 
111 static int __init cbe_ptcal_enable_on_node(int nid, int order)
112 {
113 	struct ptcal_area *area;
114 	int ret = -ENOMEM;
115 	unsigned long addr;
116 
117 	if (is_kdump_kernel())
118 		rtas_call(ptcal_stop_tok, 1, 1, NULL, nid);
119 
120 	area = kmalloc(sizeof(*area), GFP_KERNEL);
121 	if (!area)
122 		goto out_err;
123 
124 	area->nid = nid;
125 	area->order = order;
126 	area->pages = __alloc_pages_node(area->nid,
127 						GFP_KERNEL|__GFP_THISNODE,
128 						area->order);
129 
130 	if (!area->pages) {
131 		printk(KERN_WARNING "%s: no page on node %d\n",
132 			__func__, area->nid);
133 		goto out_free_area;
134 	}
135 
136 	/*
137 	 * We move the ptcal area to the middle of the allocated
138 	 * page, in order to avoid prefetches in memcpy and similar
139 	 * functions stepping on it.
140 	 */
141 	addr = __pa(page_address(area->pages)) + (PAGE_SIZE >> 1);
142 	printk(KERN_DEBUG "%s: enabling PTCAL on node %d address=0x%016lx\n",
143 			__func__, area->nid, addr);
144 
145 	ret = -EIO;
146 	if (rtas_call(ptcal_start_tok, 3, 1, NULL, area->nid,
147 				(unsigned int)(addr >> 32),
148 				(unsigned int)(addr & 0xffffffff))) {
149 		printk(KERN_ERR "%s: error enabling PTCAL on node %d!\n",
150 				__func__, nid);
151 		goto out_free_pages;
152 	}
153 
154 	list_add(&area->list, &ptcal_list);
155 
156 	return 0;
157 
158 out_free_pages:
159 	__free_pages(area->pages, area->order);
160 out_free_area:
161 	kfree(area);
162 out_err:
163 	return ret;
164 }
165 
166 static int __init cbe_ptcal_enable(void)
167 {
168 	const u32 *size;
169 	struct device_node *np;
170 	int order, found_mic = 0;
171 
172 	np = of_find_node_by_path("/rtas");
173 	if (!np)
174 		return -ENODEV;
175 
176 	size = of_get_property(np, "ibm,cbe-ptcal-size", NULL);
177 	if (!size) {
178 		of_node_put(np);
179 		return -ENODEV;
180 	}
181 
182 	pr_debug("%s: enabling PTCAL, size = 0x%x\n", __func__, *size);
183 	order = get_order(*size);
184 	of_node_put(np);
185 
186 	/* support for malta device trees, with be@/mic@ nodes */
187 	for_each_node_by_type(np, "mic-tm") {
188 		cbe_ptcal_enable_on_node(of_node_to_nid(np), order);
189 		found_mic = 1;
190 	}
191 
192 	if (found_mic)
193 		return 0;
194 
195 	/* support for older device tree - use cpu nodes */
196 	for_each_node_by_type(np, "cpu") {
197 		const u32 *nid = of_get_property(np, "node-id", NULL);
198 		if (!nid) {
199 			printk(KERN_ERR "%s: node %s is missing node-id?\n",
200 					__func__, np->full_name);
201 			continue;
202 		}
203 		cbe_ptcal_enable_on_node(*nid, order);
204 		found_mic = 1;
205 	}
206 
207 	return found_mic ? 0 : -ENODEV;
208 }
209 
210 static int cbe_ptcal_disable(void)
211 {
212 	struct ptcal_area *area, *tmp;
213 	int ret = 0;
214 
215 	pr_debug("%s: disabling PTCAL\n", __func__);
216 
217 	list_for_each_entry_safe(area, tmp, &ptcal_list, list) {
218 		/* disable ptcal on this node */
219 		if (rtas_call(ptcal_stop_tok, 1, 1, NULL, area->nid)) {
220 			printk(KERN_ERR "%s: error disabling PTCAL "
221 					"on node %d!\n", __func__,
222 					area->nid);
223 			ret = -EIO;
224 			continue;
225 		}
226 
227 		/* ensure we can access the PTCAL area */
228 		memset(page_address(area->pages), 0,
229 				1 << (area->order + PAGE_SHIFT));
230 
231 		/* clean up */
232 		list_del(&area->list);
233 		__free_pages(area->pages, area->order);
234 		kfree(area);
235 	}
236 
237 	return ret;
238 }
239 
240 static int cbe_ptcal_notify_reboot(struct notifier_block *nb,
241 		unsigned long code, void *data)
242 {
243 	return cbe_ptcal_disable();
244 }
245 
246 static void cbe_ptcal_crash_shutdown(void)
247 {
248 	cbe_ptcal_disable();
249 }
250 
251 static struct notifier_block cbe_ptcal_reboot_notifier = {
252 	.notifier_call = cbe_ptcal_notify_reboot
253 };
254 
255 #ifdef CONFIG_PPC_IBM_CELL_RESETBUTTON
256 static int sysreset_hack;
257 
258 static int __init cbe_sysreset_init(void)
259 {
260 	struct cbe_pmd_regs __iomem *regs;
261 
262 	sysreset_hack = of_machine_is_compatible("IBM,CBPLUS-1.0");
263 	if (!sysreset_hack)
264 		return 0;
265 
266 	regs = cbe_get_cpu_pmd_regs(0);
267 	if (!regs)
268 		return 0;
269 
270 	/* Enable JTAG system-reset hack */
271 	out_be32(&regs->fir_mode_reg,
272 		in_be32(&regs->fir_mode_reg) |
273 		CBE_PMD_FIR_MODE_M8);
274 
275 	return 0;
276 }
277 device_initcall(cbe_sysreset_init);
278 
279 int cbe_sysreset_hack(void)
280 {
281 	struct cbe_pmd_regs __iomem *regs;
282 
283 	/*
284 	 * The BMC can inject user triggered system reset exceptions,
285 	 * but cannot set the system reset reason in srr1,
286 	 * so check an extra register here.
287 	 */
288 	if (sysreset_hack && (smp_processor_id() == 0)) {
289 		regs = cbe_get_cpu_pmd_regs(0);
290 		if (!regs)
291 			return 0;
292 		if (in_be64(&regs->ras_esc_0) & 0x0000ffff) {
293 			out_be64(&regs->ras_esc_0, 0);
294 			return 0;
295 		}
296 	}
297 	return 1;
298 }
299 #endif /* CONFIG_PPC_IBM_CELL_RESETBUTTON */
300 
301 int __init cbe_ptcal_init(void)
302 {
303 	int ret;
304 	ptcal_start_tok = rtas_token("ibm,cbe-start-ptcal");
305 	ptcal_stop_tok = rtas_token("ibm,cbe-stop-ptcal");
306 
307 	if (ptcal_start_tok == RTAS_UNKNOWN_SERVICE
308 			|| ptcal_stop_tok == RTAS_UNKNOWN_SERVICE)
309 		return -ENODEV;
310 
311 	ret = register_reboot_notifier(&cbe_ptcal_reboot_notifier);
312 	if (ret)
313 		goto out1;
314 
315 	ret = crash_shutdown_register(&cbe_ptcal_crash_shutdown);
316 	if (ret)
317 		goto out2;
318 
319 	return cbe_ptcal_enable();
320 
321 out2:
322 	unregister_reboot_notifier(&cbe_ptcal_reboot_notifier);
323 out1:
324 	printk(KERN_ERR "Can't disable PTCAL, so not enabling\n");
325 	return ret;
326 }
327 
328 arch_initcall(cbe_ptcal_init);
329 
330 void __init cbe_ras_init(void)
331 {
332 	unsigned long hid0;
333 
334 	/*
335 	 * Enable System Error & thermal interrupts and wakeup conditions
336 	 */
337 
338 	hid0 = mfspr(SPRN_HID0);
339 	hid0 |= HID0_CBE_THERM_INT_EN | HID0_CBE_THERM_WAKEUP |
340 		HID0_CBE_SYSERR_INT_EN | HID0_CBE_SYSERR_WAKEUP;
341 	mtspr(SPRN_HID0, hid0);
342 	mb();
343 
344 	/*
345 	 * Install machine check handler. Leave setting of precise mode to
346 	 * what the firmware did for now
347 	 */
348 	ppc_md.machine_check_exception = cbe_machine_check_handler;
349 	mb();
350 
351 	/*
352 	 * For now, we assume that IOC_FIR is already set to forward some
353 	 * error conditions to the System Error handler. If that is not true
354 	 * then it will have to be fixed up here.
355 	 */
356 }
357