xref: /openbmc/linux/arch/powerpc/platforms/cell/ras.c (revision 85250a24)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Copyright 2006-2008, IBM Corporation.
4  */
5 
6 #undef DEBUG
7 
8 #include <linux/types.h>
9 #include <linux/kernel.h>
10 #include <linux/slab.h>
11 #include <linux/smp.h>
12 #include <linux/reboot.h>
13 #include <linux/kexec.h>
14 #include <linux/crash_dump.h>
15 #include <linux/of.h>
16 
17 #include <asm/kexec.h>
18 #include <asm/reg.h>
19 #include <asm/io.h>
20 #include <asm/machdep.h>
21 #include <asm/rtas.h>
22 #include <asm/cell-regs.h>
23 
24 #include "ras.h"
25 
26 
27 static void dump_fir(int cpu)
28 {
29 	struct cbe_pmd_regs __iomem *pregs = cbe_get_cpu_pmd_regs(cpu);
30 	struct cbe_iic_regs __iomem *iregs = cbe_get_cpu_iic_regs(cpu);
31 
32 	if (pregs == NULL)
33 		return;
34 
35 	/* Todo: do some nicer parsing of bits and based on them go down
36 	 * to other sub-units FIRs and not only IIC
37 	 */
38 	printk(KERN_ERR "Global Checkstop FIR    : 0x%016llx\n",
39 	       in_be64(&pregs->checkstop_fir));
40 	printk(KERN_ERR "Global Recoverable FIR  : 0x%016llx\n",
41 	       in_be64(&pregs->checkstop_fir));
42 	printk(KERN_ERR "Global MachineCheck FIR : 0x%016llx\n",
43 	       in_be64(&pregs->spec_att_mchk_fir));
44 
45 	if (iregs == NULL)
46 		return;
47 	printk(KERN_ERR "IOC FIR                 : 0x%016llx\n",
48 	       in_be64(&iregs->ioc_fir));
49 
50 }
51 
52 DEFINE_INTERRUPT_HANDLER(cbe_system_error_exception)
53 {
54 	int cpu = smp_processor_id();
55 
56 	printk(KERN_ERR "System Error Interrupt on CPU %d !\n", cpu);
57 	dump_fir(cpu);
58 	dump_stack();
59 }
60 
61 DEFINE_INTERRUPT_HANDLER(cbe_maintenance_exception)
62 {
63 	int cpu = smp_processor_id();
64 
65 	/*
66 	 * Nothing implemented for the maintenance interrupt at this point
67 	 */
68 
69 	printk(KERN_ERR "Unhandled Maintenance interrupt on CPU %d !\n", cpu);
70 	dump_stack();
71 }
72 
73 DEFINE_INTERRUPT_HANDLER(cbe_thermal_exception)
74 {
75 	int cpu = smp_processor_id();
76 
77 	/*
78 	 * Nothing implemented for the thermal interrupt at this point
79 	 */
80 
81 	printk(KERN_ERR "Unhandled Thermal interrupt on CPU %d !\n", cpu);
82 	dump_stack();
83 }
84 
85 static int cbe_machine_check_handler(struct pt_regs *regs)
86 {
87 	int cpu = smp_processor_id();
88 
89 	printk(KERN_ERR "Machine Check Interrupt on CPU %d !\n", cpu);
90 	dump_fir(cpu);
91 
92 	/* No recovery from this code now, lets continue */
93 	return 0;
94 }
95 
96 struct ptcal_area {
97 	struct list_head list;
98 	int nid;
99 	int order;
100 	struct page *pages;
101 };
102 
103 static LIST_HEAD(ptcal_list);
104 
105 static int ptcal_start_tok, ptcal_stop_tok;
106 
107 static int __init cbe_ptcal_enable_on_node(int nid, int order)
108 {
109 	struct ptcal_area *area;
110 	int ret = -ENOMEM;
111 	unsigned long addr;
112 
113 	if (is_kdump_kernel())
114 		rtas_call(ptcal_stop_tok, 1, 1, NULL, nid);
115 
116 	area = kmalloc(sizeof(*area), GFP_KERNEL);
117 	if (!area)
118 		goto out_err;
119 
120 	area->nid = nid;
121 	area->order = order;
122 	area->pages = __alloc_pages_node(area->nid,
123 						GFP_KERNEL|__GFP_THISNODE,
124 						area->order);
125 
126 	if (!area->pages) {
127 		printk(KERN_WARNING "%s: no page on node %d\n",
128 			__func__, area->nid);
129 		goto out_free_area;
130 	}
131 
132 	/*
133 	 * We move the ptcal area to the middle of the allocated
134 	 * page, in order to avoid prefetches in memcpy and similar
135 	 * functions stepping on it.
136 	 */
137 	addr = __pa(page_address(area->pages)) + (PAGE_SIZE >> 1);
138 	printk(KERN_DEBUG "%s: enabling PTCAL on node %d address=0x%016lx\n",
139 			__func__, area->nid, addr);
140 
141 	ret = -EIO;
142 	if (rtas_call(ptcal_start_tok, 3, 1, NULL, area->nid,
143 				(unsigned int)(addr >> 32),
144 				(unsigned int)(addr & 0xffffffff))) {
145 		printk(KERN_ERR "%s: error enabling PTCAL on node %d!\n",
146 				__func__, nid);
147 		goto out_free_pages;
148 	}
149 
150 	list_add(&area->list, &ptcal_list);
151 
152 	return 0;
153 
154 out_free_pages:
155 	__free_pages(area->pages, area->order);
156 out_free_area:
157 	kfree(area);
158 out_err:
159 	return ret;
160 }
161 
162 static int __init cbe_ptcal_enable(void)
163 {
164 	const u32 *size;
165 	struct device_node *np;
166 	int order, found_mic = 0;
167 
168 	np = of_find_node_by_path("/rtas");
169 	if (!np)
170 		return -ENODEV;
171 
172 	size = of_get_property(np, "ibm,cbe-ptcal-size", NULL);
173 	if (!size) {
174 		of_node_put(np);
175 		return -ENODEV;
176 	}
177 
178 	pr_debug("%s: enabling PTCAL, size = 0x%x\n", __func__, *size);
179 	order = get_order(*size);
180 	of_node_put(np);
181 
182 	/* support for malta device trees, with be@/mic@ nodes */
183 	for_each_node_by_type(np, "mic-tm") {
184 		cbe_ptcal_enable_on_node(of_node_to_nid(np), order);
185 		found_mic = 1;
186 	}
187 
188 	if (found_mic)
189 		return 0;
190 
191 	/* support for older device tree - use cpu nodes */
192 	for_each_node_by_type(np, "cpu") {
193 		const u32 *nid = of_get_property(np, "node-id", NULL);
194 		if (!nid) {
195 			printk(KERN_ERR "%s: node %pOF is missing node-id?\n",
196 					__func__, np);
197 			continue;
198 		}
199 		cbe_ptcal_enable_on_node(*nid, order);
200 		found_mic = 1;
201 	}
202 
203 	return found_mic ? 0 : -ENODEV;
204 }
205 
206 static int cbe_ptcal_disable(void)
207 {
208 	struct ptcal_area *area, *tmp;
209 	int ret = 0;
210 
211 	pr_debug("%s: disabling PTCAL\n", __func__);
212 
213 	list_for_each_entry_safe(area, tmp, &ptcal_list, list) {
214 		/* disable ptcal on this node */
215 		if (rtas_call(ptcal_stop_tok, 1, 1, NULL, area->nid)) {
216 			printk(KERN_ERR "%s: error disabling PTCAL "
217 					"on node %d!\n", __func__,
218 					area->nid);
219 			ret = -EIO;
220 			continue;
221 		}
222 
223 		/* ensure we can access the PTCAL area */
224 		memset(page_address(area->pages), 0,
225 				1 << (area->order + PAGE_SHIFT));
226 
227 		/* clean up */
228 		list_del(&area->list);
229 		__free_pages(area->pages, area->order);
230 		kfree(area);
231 	}
232 
233 	return ret;
234 }
235 
236 static int cbe_ptcal_notify_reboot(struct notifier_block *nb,
237 		unsigned long code, void *data)
238 {
239 	return cbe_ptcal_disable();
240 }
241 
242 static void cbe_ptcal_crash_shutdown(void)
243 {
244 	cbe_ptcal_disable();
245 }
246 
247 static struct notifier_block cbe_ptcal_reboot_notifier = {
248 	.notifier_call = cbe_ptcal_notify_reboot
249 };
250 
251 #ifdef CONFIG_PPC_IBM_CELL_RESETBUTTON
252 static int sysreset_hack;
253 
254 static int __init cbe_sysreset_init(void)
255 {
256 	struct cbe_pmd_regs __iomem *regs;
257 
258 	sysreset_hack = of_machine_is_compatible("IBM,CBPLUS-1.0");
259 	if (!sysreset_hack)
260 		return 0;
261 
262 	regs = cbe_get_cpu_pmd_regs(0);
263 	if (!regs)
264 		return 0;
265 
266 	/* Enable JTAG system-reset hack */
267 	out_be32(&regs->fir_mode_reg,
268 		in_be32(&regs->fir_mode_reg) |
269 		CBE_PMD_FIR_MODE_M8);
270 
271 	return 0;
272 }
273 device_initcall(cbe_sysreset_init);
274 
275 int cbe_sysreset_hack(void)
276 {
277 	struct cbe_pmd_regs __iomem *regs;
278 
279 	/*
280 	 * The BMC can inject user triggered system reset exceptions,
281 	 * but cannot set the system reset reason in srr1,
282 	 * so check an extra register here.
283 	 */
284 	if (sysreset_hack && (smp_processor_id() == 0)) {
285 		regs = cbe_get_cpu_pmd_regs(0);
286 		if (!regs)
287 			return 0;
288 		if (in_be64(&regs->ras_esc_0) & 0x0000ffff) {
289 			out_be64(&regs->ras_esc_0, 0);
290 			return 0;
291 		}
292 	}
293 	return 1;
294 }
295 #endif /* CONFIG_PPC_IBM_CELL_RESETBUTTON */
296 
297 static int __init cbe_ptcal_init(void)
298 {
299 	int ret;
300 	ptcal_start_tok = rtas_token("ibm,cbe-start-ptcal");
301 	ptcal_stop_tok = rtas_token("ibm,cbe-stop-ptcal");
302 
303 	if (ptcal_start_tok == RTAS_UNKNOWN_SERVICE
304 			|| ptcal_stop_tok == RTAS_UNKNOWN_SERVICE)
305 		return -ENODEV;
306 
307 	ret = register_reboot_notifier(&cbe_ptcal_reboot_notifier);
308 	if (ret)
309 		goto out1;
310 
311 	ret = crash_shutdown_register(&cbe_ptcal_crash_shutdown);
312 	if (ret)
313 		goto out2;
314 
315 	return cbe_ptcal_enable();
316 
317 out2:
318 	unregister_reboot_notifier(&cbe_ptcal_reboot_notifier);
319 out1:
320 	printk(KERN_ERR "Can't disable PTCAL, so not enabling\n");
321 	return ret;
322 }
323 
324 arch_initcall(cbe_ptcal_init);
325 
326 void __init cbe_ras_init(void)
327 {
328 	unsigned long hid0;
329 
330 	/*
331 	 * Enable System Error & thermal interrupts and wakeup conditions
332 	 */
333 
334 	hid0 = mfspr(SPRN_HID0);
335 	hid0 |= HID0_CBE_THERM_INT_EN | HID0_CBE_THERM_WAKEUP |
336 		HID0_CBE_SYSERR_INT_EN | HID0_CBE_SYSERR_WAKEUP;
337 	mtspr(SPRN_HID0, hid0);
338 	mb();
339 
340 	/*
341 	 * Install machine check handler. Leave setting of precise mode to
342 	 * what the firmware did for now
343 	 */
344 	ppc_md.machine_check_exception = cbe_machine_check_handler;
345 	mb();
346 
347 	/*
348 	 * For now, we assume that IOC_FIR is already set to forward some
349 	 * error conditions to the System Error handler. If that is not true
350 	 * then it will have to be fixed up here.
351 	 */
352 }
353