diff options
Diffstat (limited to 'arch/powerpc/platforms/pseries')
-rw-r--r-- | arch/powerpc/platforms/pseries/Kconfig | 6 | ||||
-rw-r--r-- | arch/powerpc/platforms/pseries/kexec.c | 10 | ||||
-rw-r--r-- | arch/powerpc/platforms/pseries/lpar.c | 37 | ||||
-rw-r--r-- | arch/powerpc/platforms/pseries/ras.c | 102 |
4 files changed, 109 insertions, 46 deletions
diff --git a/arch/powerpc/platforms/pseries/Kconfig b/arch/powerpc/platforms/pseries/Kconfig index 5d1b743dbe7e..5b3da4b4ea79 100644 --- a/arch/powerpc/platforms/pseries/Kconfig +++ b/arch/powerpc/platforms/pseries/Kconfig | |||
@@ -10,7 +10,7 @@ config PPC_PSERIES | |||
10 | select RTAS_ERROR_LOGGING | 10 | select RTAS_ERROR_LOGGING |
11 | select PPC_UDBG_16550 | 11 | select PPC_UDBG_16550 |
12 | select PPC_NATIVE | 12 | select PPC_NATIVE |
13 | select PPC_PCI_CHOICE if EMBEDDED | 13 | select PPC_PCI_CHOICE if EXPERT |
14 | default y | 14 | default y |
15 | 15 | ||
16 | config PPC_SPLPAR | 16 | config PPC_SPLPAR |
@@ -24,9 +24,9 @@ config PPC_SPLPAR | |||
24 | two or more partitions. | 24 | two or more partitions. |
25 | 25 | ||
26 | config EEH | 26 | config EEH |
27 | bool "PCI Extended Error Handling (EEH)" if EMBEDDED | 27 | bool "PCI Extended Error Handling (EEH)" if EXPERT |
28 | depends on PPC_PSERIES && PCI | 28 | depends on PPC_PSERIES && PCI |
29 | default y if !EMBEDDED | 29 | default y if !EXPERT |
30 | 30 | ||
31 | config PSERIES_MSI | 31 | config PSERIES_MSI |
32 | bool | 32 | bool |
diff --git a/arch/powerpc/platforms/pseries/kexec.c b/arch/powerpc/platforms/pseries/kexec.c index 53cbd53d8740..77d38a5e2ff9 100644 --- a/arch/powerpc/platforms/pseries/kexec.c +++ b/arch/powerpc/platforms/pseries/kexec.c | |||
@@ -61,13 +61,3 @@ void __init setup_kexec_cpu_down_xics(void) | |||
61 | { | 61 | { |
62 | ppc_md.kexec_cpu_down = pseries_kexec_cpu_down_xics; | 62 | ppc_md.kexec_cpu_down = pseries_kexec_cpu_down_xics; |
63 | } | 63 | } |
64 | |||
65 | static int __init pseries_kexec_setup(void) | ||
66 | { | ||
67 | ppc_md.machine_kexec = default_machine_kexec; | ||
68 | ppc_md.machine_kexec_prepare = default_machine_kexec_prepare; | ||
69 | ppc_md.machine_crash_shutdown = default_machine_crash_shutdown; | ||
70 | |||
71 | return 0; | ||
72 | } | ||
73 | machine_device_initcall(pseries, pseries_kexec_setup); | ||
diff --git a/arch/powerpc/platforms/pseries/lpar.c b/arch/powerpc/platforms/pseries/lpar.c index 5d3ea9f60dd7..ca5d5898d320 100644 --- a/arch/powerpc/platforms/pseries/lpar.c +++ b/arch/powerpc/platforms/pseries/lpar.c | |||
@@ -713,6 +713,13 @@ EXPORT_SYMBOL(arch_free_page); | |||
713 | /* NB: reg/unreg are called while guarded with the tracepoints_mutex */ | 713 | /* NB: reg/unreg are called while guarded with the tracepoints_mutex */ |
714 | extern long hcall_tracepoint_refcount; | 714 | extern long hcall_tracepoint_refcount; |
715 | 715 | ||
716 | /* | ||
717 | * Since the tracing code might execute hcalls we need to guard against | ||
718 | * recursion. One example of this are spinlocks calling H_YIELD on | ||
719 | * shared processor partitions. | ||
720 | */ | ||
721 | static DEFINE_PER_CPU(unsigned int, hcall_trace_depth); | ||
722 | |||
716 | void hcall_tracepoint_regfunc(void) | 723 | void hcall_tracepoint_regfunc(void) |
717 | { | 724 | { |
718 | hcall_tracepoint_refcount++; | 725 | hcall_tracepoint_refcount++; |
@@ -725,12 +732,42 @@ void hcall_tracepoint_unregfunc(void) | |||
725 | 732 | ||
726 | void __trace_hcall_entry(unsigned long opcode, unsigned long *args) | 733 | void __trace_hcall_entry(unsigned long opcode, unsigned long *args) |
727 | { | 734 | { |
735 | unsigned long flags; | ||
736 | unsigned int *depth; | ||
737 | |||
738 | local_irq_save(flags); | ||
739 | |||
740 | depth = &__get_cpu_var(hcall_trace_depth); | ||
741 | |||
742 | if (*depth) | ||
743 | goto out; | ||
744 | |||
745 | (*depth)++; | ||
728 | trace_hcall_entry(opcode, args); | 746 | trace_hcall_entry(opcode, args); |
747 | (*depth)--; | ||
748 | |||
749 | out: | ||
750 | local_irq_restore(flags); | ||
729 | } | 751 | } |
730 | 752 | ||
731 | void __trace_hcall_exit(long opcode, unsigned long retval, | 753 | void __trace_hcall_exit(long opcode, unsigned long retval, |
732 | unsigned long *retbuf) | 754 | unsigned long *retbuf) |
733 | { | 755 | { |
756 | unsigned long flags; | ||
757 | unsigned int *depth; | ||
758 | |||
759 | local_irq_save(flags); | ||
760 | |||
761 | depth = &__get_cpu_var(hcall_trace_depth); | ||
762 | |||
763 | if (*depth) | ||
764 | goto out; | ||
765 | |||
766 | (*depth)++; | ||
734 | trace_hcall_exit(opcode, retval, retbuf); | 767 | trace_hcall_exit(opcode, retval, retbuf); |
768 | (*depth)--; | ||
769 | |||
770 | out: | ||
771 | local_irq_restore(flags); | ||
735 | } | 772 | } |
736 | #endif | 773 | #endif |
diff --git a/arch/powerpc/platforms/pseries/ras.c b/arch/powerpc/platforms/pseries/ras.c index a4fc6da87c2e..c55d7ad9c648 100644 --- a/arch/powerpc/platforms/pseries/ras.c +++ b/arch/powerpc/platforms/pseries/ras.c | |||
@@ -54,7 +54,8 @@ | |||
54 | static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX]; | 54 | static unsigned char ras_log_buf[RTAS_ERROR_LOG_MAX]; |
55 | static DEFINE_SPINLOCK(ras_log_buf_lock); | 55 | static DEFINE_SPINLOCK(ras_log_buf_lock); |
56 | 56 | ||
57 | static char mce_data_buf[RTAS_ERROR_LOG_MAX]; | 57 | static char global_mce_data_buf[RTAS_ERROR_LOG_MAX]; |
58 | static DEFINE_PER_CPU(__u64, mce_data_buf); | ||
58 | 59 | ||
59 | static int ras_get_sensor_state_token; | 60 | static int ras_get_sensor_state_token; |
60 | static int ras_check_exception_token; | 61 | static int ras_check_exception_token; |
@@ -196,12 +197,24 @@ static irqreturn_t ras_error_interrupt(int irq, void *dev_id) | |||
196 | return IRQ_HANDLED; | 197 | return IRQ_HANDLED; |
197 | } | 198 | } |
198 | 199 | ||
199 | /* Get the error information for errors coming through the | 200 | /* |
201 | * Some versions of FWNMI place the buffer inside the 4kB page starting at | ||
202 | * 0x7000. Other versions place it inside the rtas buffer. We check both. | ||
203 | */ | ||
204 | #define VALID_FWNMI_BUFFER(A) \ | ||
205 | ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \ | ||
206 | (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16)))) | ||
207 | |||
208 | /* | ||
209 | * Get the error information for errors coming through the | ||
200 | * FWNMI vectors. The pt_regs' r3 will be updated to reflect | 210 | * FWNMI vectors. The pt_regs' r3 will be updated to reflect |
201 | * the actual r3 if possible, and a ptr to the error log entry | 211 | * the actual r3 if possible, and a ptr to the error log entry |
202 | * will be returned if found. | 212 | * will be returned if found. |
203 | * | 213 | * |
204 | * The mce_data_buf does not have any locks or protection around it, | 214 | * If the RTAS error is not of the extended type, then we put it in a per |
215 | * cpu 64bit buffer. If it is the extended type we use global_mce_data_buf. | ||
216 | * | ||
217 | * The global_mce_data_buf does not have any locks or protection around it, | ||
205 | * if a second machine check comes in, or a system reset is done | 218 | * if a second machine check comes in, or a system reset is done |
206 | * before we have logged the error, then we will get corruption in the | 219 | * before we have logged the error, then we will get corruption in the |
207 | * error log. This is preferable over holding off on calling | 220 | * error log. This is preferable over holding off on calling |
@@ -210,20 +223,31 @@ static irqreturn_t ras_error_interrupt(int irq, void *dev_id) | |||
210 | */ | 223 | */ |
211 | static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs) | 224 | static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs) |
212 | { | 225 | { |
213 | unsigned long errdata = regs->gpr[3]; | ||
214 | struct rtas_error_log *errhdr = NULL; | ||
215 | unsigned long *savep; | 226 | unsigned long *savep; |
227 | struct rtas_error_log *h, *errhdr = NULL; | ||
228 | |||
229 | if (!VALID_FWNMI_BUFFER(regs->gpr[3])) { | ||
230 | printk(KERN_ERR "FWNMI: corrupt r3\n"); | ||
231 | return NULL; | ||
232 | } | ||
216 | 233 | ||
217 | if ((errdata >= 0x7000 && errdata < 0x7fff0) || | 234 | savep = __va(regs->gpr[3]); |
218 | (errdata >= rtas.base && errdata < rtas.base + rtas.size - 16)) { | 235 | regs->gpr[3] = savep[0]; /* restore original r3 */ |
219 | savep = __va(errdata); | 236 | |
220 | regs->gpr[3] = savep[0]; /* restore original r3 */ | 237 | /* If it isn't an extended log we can use the per cpu 64bit buffer */ |
221 | memset(mce_data_buf, 0, RTAS_ERROR_LOG_MAX); | 238 | h = (struct rtas_error_log *)&savep[1]; |
222 | memcpy(mce_data_buf, (char *)(savep + 1), RTAS_ERROR_LOG_MAX); | 239 | if (!h->extended) { |
223 | errhdr = (struct rtas_error_log *)mce_data_buf; | 240 | memcpy(&__get_cpu_var(mce_data_buf), h, sizeof(__u64)); |
241 | errhdr = (struct rtas_error_log *)&__get_cpu_var(mce_data_buf); | ||
224 | } else { | 242 | } else { |
225 | printk("FWNMI: corrupt r3\n"); | 243 | int len; |
244 | |||
245 | len = max_t(int, 8+h->extended_log_length, RTAS_ERROR_LOG_MAX); | ||
246 | memset(global_mce_data_buf, 0, RTAS_ERROR_LOG_MAX); | ||
247 | memcpy(global_mce_data_buf, h, len); | ||
248 | errhdr = (struct rtas_error_log *)global_mce_data_buf; | ||
226 | } | 249 | } |
250 | |||
227 | return errhdr; | 251 | return errhdr; |
228 | } | 252 | } |
229 | 253 | ||
@@ -235,7 +259,7 @@ static void fwnmi_release_errinfo(void) | |||
235 | { | 259 | { |
236 | int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL); | 260 | int ret = rtas_call(rtas_token("ibm,nmi-interlock"), 0, 1, NULL); |
237 | if (ret != 0) | 261 | if (ret != 0) |
238 | printk("FWNMI: nmi-interlock failed: %d\n", ret); | 262 | printk(KERN_ERR "FWNMI: nmi-interlock failed: %d\n", ret); |
239 | } | 263 | } |
240 | 264 | ||
241 | int pSeries_system_reset_exception(struct pt_regs *regs) | 265 | int pSeries_system_reset_exception(struct pt_regs *regs) |
@@ -259,31 +283,43 @@ int pSeries_system_reset_exception(struct pt_regs *regs) | |||
259 | * Return 1 if corrected (or delivered a signal). | 283 | * Return 1 if corrected (or delivered a signal). |
260 | * Return 0 if there is nothing we can do. | 284 | * Return 0 if there is nothing we can do. |
261 | */ | 285 | */ |
262 | static int recover_mce(struct pt_regs *regs, struct rtas_error_log * err) | 286 | static int recover_mce(struct pt_regs *regs, struct rtas_error_log *err) |
263 | { | 287 | { |
264 | int nonfatal = 0; | 288 | int recovered = 0; |
265 | 289 | ||
266 | if (err->disposition == RTAS_DISP_FULLY_RECOVERED) { | 290 | if (!(regs->msr & MSR_RI)) { |
291 | /* If MSR_RI isn't set, we cannot recover */ | ||
292 | recovered = 0; | ||
293 | |||
294 | } else if (err->disposition == RTAS_DISP_FULLY_RECOVERED) { | ||
267 | /* Platform corrected itself */ | 295 | /* Platform corrected itself */ |
268 | nonfatal = 1; | 296 | recovered = 1; |
269 | } else if ((regs->msr & MSR_RI) && | 297 | |
270 | user_mode(regs) && | 298 | } else if (err->disposition == RTAS_DISP_LIMITED_RECOVERY) { |
271 | err->severity == RTAS_SEVERITY_ERROR_SYNC && | 299 | /* Platform corrected itself but could be degraded */ |
272 | err->disposition == RTAS_DISP_NOT_RECOVERED && | 300 | printk(KERN_ERR "MCE: limited recovery, system may " |
273 | err->target == RTAS_TARGET_MEMORY && | 301 | "be degraded\n"); |
274 | err->type == RTAS_TYPE_ECC_UNCORR && | 302 | recovered = 1; |
275 | !(current->pid == 0 || is_global_init(current))) { | 303 | |
276 | /* Kill off a user process with an ECC error */ | 304 | } else if (user_mode(regs) && !is_global_init(current) && |
277 | printk(KERN_ERR "MCE: uncorrectable ecc error for pid %d\n", | 305 | err->severity == RTAS_SEVERITY_ERROR_SYNC) { |
278 | current->pid); | 306 | |
279 | /* XXX something better for ECC error? */ | 307 | /* |
280 | _exception(SIGBUS, regs, BUS_ADRERR, regs->nip); | 308 | * If we received a synchronous error when in userspace |
281 | nonfatal = 1; | 309 | * kill the task. Firmware may report details of the fail |
310 | * asynchronously, so we can't rely on the target and type | ||
311 | * fields being valid here. | ||
312 | */ | ||
313 | printk(KERN_ERR "MCE: uncorrectable error, killing task " | ||
314 | "%s:%d\n", current->comm, current->pid); | ||
315 | |||
316 | _exception(SIGBUS, regs, BUS_MCEERR_AR, regs->nip); | ||
317 | recovered = 1; | ||
282 | } | 318 | } |
283 | 319 | ||
284 | log_error((char *)err, ERR_TYPE_RTAS_LOG, !nonfatal); | 320 | log_error((char *)err, ERR_TYPE_RTAS_LOG, 0); |
285 | 321 | ||
286 | return nonfatal; | 322 | return recovered; |
287 | } | 323 | } |
288 | 324 | ||
289 | /* | 325 | /* |