aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorJesse Barnes <jbarnes@virtuousgeek.org>2009-07-11 16:48:03 -0400
committerEric Anholt <eric@anholt.net>2009-07-13 18:33:37 -0400
commit8a90523639f49dc4b4fa7ae47bb9c8ed73ea8577 (patch)
tree6b7142b32b16c2f87e38420568a39f92968692b3
parent832cc28d5bc676331e6376d940ae45d5937aa688 (diff)
drm/i915: refactor error detection & collection
This patch refactors the existing error detection and collection code, placing most of it in i915_handle_error(). Additionally, we introduce a work queue for scheduling post-crash tasks such as generating a uevent. Using the uevent facility, userspace should be able to capture a post-mortem dump for diagnostics. Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by: Ben Gamari <bgamari.foss@gmail.com> Signed-off-by: Eric Anholt <eric@anholt.net>
Diffstat
-rw-r--r--drivers/gpu/drm/i915/i915_drv.h1
-rw-r--r--drivers/gpu/drm/i915/i915_gem_debugfs.c2
-rw-r--r--drivers/gpu/drm/i915/i915_irq.c232
3 files changed, 161 insertions, 74 deletions
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index d08752875885..b05b44dd3bf6 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -229,6 +229,7 @@ typedef struct drm_i915_private {
229 229
230 spinlock_t error_lock; 230 spinlock_t error_lock;
231 struct drm_i915_error_state *first_error; 231 struct drm_i915_error_state *first_error;
232 struct work_struct error_work;
232 233
233 /* Register state */ 234 /* Register state */
234 u8 saveLBB; 235 u8 saveLBB;
diff --git a/drivers/gpu/drm/i915/i915_gem_debugfs.c b/drivers/gpu/drm/i915/i915_gem_debugfs.c
index 9a44bfcb8139..cb3b97405fbf 100644
--- a/drivers/gpu/drm/i915/i915_gem_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_gem_debugfs.c
@@ -343,6 +343,8 @@ static int i915_error_state(struct seq_file *m, void *unused)
343 343
344 error = dev_priv->first_error; 344 error = dev_priv->first_error;
345 345
346 seq_printf(m, "Time: %ld s %ld us\n", error->time.tv_sec,
347 error->time.tv_usec);
346 seq_printf(m, "EIR: 0x%08x\n", error->eir); 348 seq_printf(m, "EIR: 0x%08x\n", error->eir);
347 seq_printf(m, " PGTBL_ER: 0x%08x\n", error->pgtbl_er); 349 seq_printf(m, " PGTBL_ER: 0x%08x\n", error->pgtbl_er);
348 seq_printf(m, " INSTPM: 0x%08x\n", error->instpm); 350 seq_printf(m, " INSTPM: 0x%08x\n", error->instpm);
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 7ba23a69a0c0..f340b3fd54e6 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -290,6 +290,35 @@ irqreturn_t igdng_irq_handler(struct drm_device *dev)
290 return ret; 290 return ret;
291} 291}
292 292
293/**
294 * i915_error_work_func - do process context error handling work
295 * @work: work struct
296 *
297 * Fire an error uevent so userspace can see that a hang or error
298 * was detected.
299 */
300static void i915_error_work_func(struct work_struct *work)
301{
302 drm_i915_private_t *dev_priv = container_of(work, drm_i915_private_t,
303 error_work);
304 struct drm_device *dev = dev_priv->dev;
305 char *event_string = "ERROR=1";
306 char *envp[] = { event_string, NULL };
307
308 DRM_DEBUG("generating error event\n");
309
310 kobject_uevent_env(&dev->primary->kdev.kobj, KOBJ_CHANGE, envp);
311}
312
313/**
314 * i915_capture_error_state - capture an error record for later analysis
315 * @dev: drm device
316 *
317 * Should be called when an error is detected (either a hang or an error
318 * interrupt) to capture error state from the time of the error. Fills
319 * out a structure which becomes available in debugfs for user level tools
320 * to pick up.
321 */
293static void i915_capture_error_state(struct drm_device *dev) 322static void i915_capture_error_state(struct drm_device *dev)
294{ 323{
295 struct drm_i915_private *dev_priv = dev->dev_private; 324 struct drm_i915_private *dev_priv = dev->dev_private;
@@ -325,12 +354,137 @@ static void i915_capture_error_state(struct drm_device *dev)
325 error->acthd = I915_READ(ACTHD_I965); 354 error->acthd = I915_READ(ACTHD_I965);
326 } 355 }
327 356
357 do_gettimeofday(&error->time);
358
328 dev_priv->first_error = error; 359 dev_priv->first_error = error;
329 360
330out: 361out:
331 spin_unlock_irqrestore(&dev_priv->error_lock, flags); 362 spin_unlock_irqrestore(&dev_priv->error_lock, flags);
332} 363}
333 364
365/**
366 * i915_handle_error - handle an error interrupt
367 * @dev: drm device
368 *
369 * Do some basic checking of regsiter state at error interrupt time and
370 * dump it to the syslog. Also call i915_capture_error_state() to make
371 * sure we get a record and make it available in debugfs. Fire a uevent
372 * so userspace knows something bad happened (should trigger collection
373 * of a ring dump etc.).
374 */
375static void i915_handle_error(struct drm_device *dev)
376{
377 struct drm_i915_private *dev_priv = dev->dev_private;
378 u32 eir = I915_READ(EIR);
379 u32 pipea_stats = I915_READ(PIPEASTAT);
380 u32 pipeb_stats = I915_READ(PIPEBSTAT);
381
382 i915_capture_error_state(dev);
383
384 printk(KERN_ERR "render error detected, EIR: 0x%08x\n",
385 eir);
386
387 if (IS_G4X(dev)) {
388 if (eir & (GM45_ERROR_MEM_PRIV | GM45_ERROR_CP_PRIV)) {
389 u32 ipeir = I915_READ(IPEIR_I965);
390
391 printk(KERN_ERR " IPEIR: 0x%08x\n",
392 I915_READ(IPEIR_I965));
393 printk(KERN_ERR " IPEHR: 0x%08x\n",
394 I915_READ(IPEHR_I965));
395 printk(KERN_ERR " INSTDONE: 0x%08x\n",
396 I915_READ(INSTDONE_I965));
397 printk(KERN_ERR " INSTPS: 0x%08x\n",
398 I915_READ(INSTPS));
399 printk(KERN_ERR " INSTDONE1: 0x%08x\n",
400 I915_READ(INSTDONE1));
401 printk(KERN_ERR " ACTHD: 0x%08x\n",
402 I915_READ(ACTHD_I965));
403 I915_WRITE(IPEIR_I965, ipeir);
404 (void)I915_READ(IPEIR_I965);
405 }
406 if (eir & GM45_ERROR_PAGE_TABLE) {
407 u32 pgtbl_err = I915_READ(PGTBL_ER);
408 printk(KERN_ERR "page table error\n");
409 printk(KERN_ERR " PGTBL_ER: 0x%08x\n",
410 pgtbl_err);
411 I915_WRITE(PGTBL_ER, pgtbl_err);
412 (void)I915_READ(PGTBL_ER);
413 }
414 }
415
416 if (IS_I9XX(dev)) {
417 if (eir & I915_ERROR_PAGE_TABLE) {
418 u32 pgtbl_err = I915_READ(PGTBL_ER);
419 printk(KERN_ERR "page table error\n");
420 printk(KERN_ERR " PGTBL_ER: 0x%08x\n",
421 pgtbl_err);
422 I915_WRITE(PGTBL_ER, pgtbl_err);
423 (void)I915_READ(PGTBL_ER);
424 }
425 }
426
427 if (eir & I915_ERROR_MEMORY_REFRESH) {
428 printk(KERN_ERR "memory refresh error\n");
429 printk(KERN_ERR "PIPEASTAT: 0x%08x\n",
430 pipea_stats);
431 printk(KERN_ERR "PIPEBSTAT: 0x%08x\n",
432 pipeb_stats);
433 /* pipestat has already been acked */
434 }
435 if (eir & I915_ERROR_INSTRUCTION) {
436 printk(KERN_ERR "instruction error\n");
437 printk(KERN_ERR " INSTPM: 0x%08x\n",
438 I915_READ(INSTPM));
439 if (!IS_I965G(dev)) {
440 u32 ipeir = I915_READ(IPEIR);
441
442 printk(KERN_ERR " IPEIR: 0x%08x\n",
443 I915_READ(IPEIR));
444 printk(KERN_ERR " IPEHR: 0x%08x\n",
445 I915_READ(IPEHR));
446 printk(KERN_ERR " INSTDONE: 0x%08x\n",
447 I915_READ(INSTDONE));
448 printk(KERN_ERR " ACTHD: 0x%08x\n",
449 I915_READ(ACTHD));
450 I915_WRITE(IPEIR, ipeir);
451 (void)I915_READ(IPEIR);
452 } else {
453 u32 ipeir = I915_READ(IPEIR_I965);
454
455 printk(KERN_ERR " IPEIR: 0x%08x\n",
456 I915_READ(IPEIR_I965));
457 printk(KERN_ERR " IPEHR: 0x%08x\n",
458 I915_READ(IPEHR_I965));
459 printk(KERN_ERR " INSTDONE: 0x%08x\n",
460 I915_READ(INSTDONE_I965));
461 printk(KERN_ERR " INSTPS: 0x%08x\n",
462 I915_READ(INSTPS));
463 printk(KERN_ERR " INSTDONE1: 0x%08x\n",
464 I915_READ(INSTDONE1));
465 printk(KERN_ERR " ACTHD: 0x%08x\n",
466 I915_READ(ACTHD_I965));
467 I915_WRITE(IPEIR_I965, ipeir);
468 (void)I915_READ(IPEIR_I965);
469 }
470 }
471
472 I915_WRITE(EIR, eir);
473 (void)I915_READ(EIR);
474 eir = I915_READ(EIR);
475 if (eir) {
476 /*
477 * some errors might have become stuck,
478 * mask them.
479 */
480 DRM_ERROR("EIR stuck: 0x%08x, masking\n", eir);
481 I915_WRITE(EMR, I915_READ(EMR) | eir);
482 I915_WRITE(IIR, I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT);
483 }
484
485 schedule_work(&dev_priv->error_work);
486}
487
334irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) 488irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
335{ 489{
336 struct drm_device *dev = (struct drm_device *) arg; 490 struct drm_device *dev = (struct drm_device *) arg;
@@ -372,6 +526,9 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
372 pipea_stats = I915_READ(PIPEASTAT); 526 pipea_stats = I915_READ(PIPEASTAT);
373 pipeb_stats = I915_READ(PIPEBSTAT); 527 pipeb_stats = I915_READ(PIPEBSTAT);
374 528
529 if (iir & I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT)
530 i915_handle_error(dev);
531
375 /* 532 /*
376 * Clear the PIPE(A|B)STAT regs before the IIR 533 * Clear the PIPE(A|B)STAT regs before the IIR
377 */ 534 */
@@ -409,80 +566,6 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
409 I915_READ(PORT_HOTPLUG_STAT); 566 I915_READ(PORT_HOTPLUG_STAT);
410 } 567 }
411 568
412 if (iir & I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT) {
413 u32 eir = I915_READ(EIR);
414
415 i915_capture_error_state(dev);
416
417 printk(KERN_ERR "render error detected, EIR: 0x%08x\n",
418 eir);
419 if (eir & I915_ERROR_PAGE_TABLE) {
420 u32 pgtbl_err = I915_READ(PGTBL_ER);
421 printk(KERN_ERR "page table error\n");
422 printk(KERN_ERR " PGTBL_ER: 0x%08x\n",
423 pgtbl_err);
424 I915_WRITE(PGTBL_ER, pgtbl_err);
425 (void)I915_READ(PGTBL_ER);
426 }
427 if (eir & I915_ERROR_MEMORY_REFRESH) {
428 printk(KERN_ERR "memory refresh error\n");
429 printk(KERN_ERR "PIPEASTAT: 0x%08x\n",
430 pipea_stats);
431 printk(KERN_ERR "PIPEBSTAT: 0x%08x\n",
432 pipeb_stats);
433 /* pipestat has already been acked */
434 }
435 if (eir & I915_ERROR_INSTRUCTION) {
436 printk(KERN_ERR "instruction error\n");
437 printk(KERN_ERR " INSTPM: 0x%08x\n",
438 I915_READ(INSTPM));
439 if (!IS_I965G(dev)) {
440 u32 ipeir = I915_READ(IPEIR);
441
442 printk(KERN_ERR " IPEIR: 0x%08x\n",
443 I915_READ(IPEIR));
444 printk(KERN_ERR " IPEHR: 0x%08x\n",
445 I915_READ(IPEHR));
446 printk(KERN_ERR " INSTDONE: 0x%08x\n",
447 I915_READ(INSTDONE));
448 printk(KERN_ERR " ACTHD: 0x%08x\n",
449 I915_READ(ACTHD));
450 I915_WRITE(IPEIR, ipeir);
451 (void)I915_READ(IPEIR);
452 } else {
453 u32 ipeir = I915_READ(IPEIR_I965);
454
455 printk(KERN_ERR " IPEIR: 0x%08x\n",
456 I915_READ(IPEIR_I965));
457 printk(KERN_ERR " IPEHR: 0x%08x\n",
458 I915_READ(IPEHR_I965));
459 printk(KERN_ERR " INSTDONE: 0x%08x\n",
460 I915_READ(INSTDONE_I965));
461 printk(KERN_ERR " INSTPS: 0x%08x\n",
462 I915_READ(INSTPS));
463 printk(KERN_ERR " INSTDONE1: 0x%08x\n",
464 I915_READ(INSTDONE1));
465 printk(KERN_ERR " ACTHD: 0x%08x\n",
466 I915_READ(ACTHD_I965));
467 I915_WRITE(IPEIR_I965, ipeir);
468 (void)I915_READ(IPEIR_I965);
469 }
470 }
471
472 I915_WRITE(EIR, eir);
473 (void)I915_READ(EIR);
474 eir = I915_READ(EIR);
475 if (eir) {
476 /*
477 * some errors might have become stuck,
478 * mask them.
479 */
480 DRM_ERROR("EIR stuck: 0x%08x, masking\n", eir);
481 I915_WRITE(EMR, I915_READ(EMR) | eir);
482 I915_WRITE(IIR, I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT);
483 }
484 }
485
486 I915_WRITE(IIR, iir); 569 I915_WRITE(IIR, iir);
487 new_iir = I915_READ(IIR); /* Flush posted writes */ 570 new_iir = I915_READ(IIR); /* Flush posted writes */
488 571
@@ -830,6 +913,7 @@ void i915_driver_irq_preinstall(struct drm_device * dev)
830 atomic_set(&dev_priv->irq_received, 0); 913 atomic_set(&dev_priv->irq_received, 0);
831 914
832 INIT_WORK(&dev_priv->hotplug_work, i915_hotplug_work_func); 915 INIT_WORK(&dev_priv->hotplug_work, i915_hotplug_work_func);
916 INIT_WORK(&dev_priv->error_work, i915_error_work_func);
833 917
834 if (IS_IGDNG(dev)) { 918 if (IS_IGDNG(dev)) {
835 igdng_irq_preinstall(dev); 919 igdng_irq_preinstall(dev);
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-25 23:19:02 -0400