6 files changed, 24 insertions, 11 deletions

diff --git a/drivers/misc/sgi-gru/gru_instructions.h b/drivers/misc/sgi-gru/gru_instructions.h
index 2feb885ca96f..eb9140d32e62 100644
--- a/drivers/misc/sgi-gru/gru_instructions.h
+++ b/drivers/misc/sgi-gru/gru_instructions.h
@@ -253,32 +253,37 @@ struct gru_instruction {
 #define CBE_CAUSE_HA_RESPONSE_FATAL             (1 << 13)
 #define CBE_CAUSE_HA_RESPONSE_NON_FATAL         (1 << 14)
 #define CBE_CAUSE_ADDRESS_SPACE_DECODE_ERROR    (1 << 15)
-#define CBE_CAUSE_RESPONSE_DATA_ERROR           (1 << 16)
-#define CBE_CAUSE_PROTOCOL_STATE_DATA_ERROR     (1 << 17)
+#define CBE_CAUSE_PROTOCOL_STATE_DATA_ERROR     (1 << 16)
+#define CBE_CAUSE_RA_RESPONSE_DATA_ERROR        (1 << 17)
+#define CBE_CAUSE_HA_RESPONSE_DATA_ERROR        (1 << 18)
 
 /* CBE cbrexecstatus bits */
 #define CBR_EXS_ABORT_OCC_BIT                   0
 #define CBR_EXS_INT_OCC_BIT                     1
 #define CBR_EXS_PENDING_BIT                     2
 #define CBR_EXS_QUEUED_BIT                      3
-#define CBR_EXS_TLBHW_BIT                       4
+#define CBR_EXS_TLB_INVAL_BIT                   4
 #define CBR_EXS_EXCEPTION_BIT                   5
 
 #define CBR_EXS_ABORT_OCC                       (1 << CBR_EXS_ABORT_OCC_BIT)
 #define CBR_EXS_INT_OCC                         (1 << CBR_EXS_INT_OCC_BIT)
 #define CBR_EXS_PENDING                         (1 << CBR_EXS_PENDING_BIT)
 #define CBR_EXS_QUEUED                          (1 << CBR_EXS_QUEUED_BIT)
-#define CBR_EXS_TLBHW                           (1 << CBR_EXS_TLBHW_BIT)
+#define CBR_TLB_INVAL                           (1 << CBR_EXS_TLB_INVAL_BIT)
 #define CBR_EXS_EXCEPTION                       (1 << CBR_EXS_EXCEPTION_BIT)
 
 /*
  * Exceptions are retried for the following cases. If any OTHER bits are set
  * in ecause, the exception is not retryable.
  */
-#define EXCEPTION_RETRY_BITS (CBE_CAUSE_RESPONSE_DATA_ERROR |           \
-                              CBE_CAUSE_RA_REQUEST_TIMEOUT |            \
+#define EXCEPTION_RETRY_BITS (CBE_CAUSE_EXECUTION_HW_ERROR |            \
                               CBE_CAUSE_TLBHW_ERROR |                   \
-                              CBE_CAUSE_HA_REQUEST_TIMEOUT)
+                              CBE_CAUSE_RA_REQUEST_TIMEOUT |            \
+                              CBE_CAUSE_RA_RESPONSE_NON_FATAL |         \
+                              CBE_CAUSE_HA_RESPONSE_NON_FATAL |         \
+                              CBE_CAUSE_RA_RESPONSE_DATA_ERROR |        \
+                              CBE_CAUSE_HA_RESPONSE_DATA_ERROR          \
+                              )
 
 /* Message queue head structure */
 union gru_mesqhead {
diff --git a/drivers/misc/sgi-gru/grufault.c b/drivers/misc/sgi-gru/grufault.c
index 8443e90f9f6c..a489807613f8 100644
--- a/drivers/misc/sgi-gru/grufault.c
+++ b/drivers/misc/sgi-gru/grufault.c
@@ -339,8 +339,12 @@ static int gru_try_dropin(struct gru_thread_state *gts,
          * Might be a hardware race OR a stupid user. Ignore FMM because FMM
          * is a transient state.
          */
-        if (tfh->status != TFHSTATUS_EXCEPTION)
-                goto failnoexception;
+        if (tfh->status != TFHSTATUS_EXCEPTION) {
+                gru_flush_cache(tfh);
+                if (tfh->status != TFHSTATUS_EXCEPTION)
+                        goto failnoexception;
+                STAT(tfh_stale_on_fault);
+        }
         if (tfh->state == TFHSTATE_IDLE)
                 goto failidle;
         if (tfh->state == TFHSTATE_MISS_FMM && cb)
diff --git a/drivers/misc/sgi-gru/gruhandles.h b/drivers/misc/sgi-gru/gruhandles.h
index 9f41e2cc09d5..99ec82678f5c 100644
--- a/drivers/misc/sgi-gru/gruhandles.h
+++ b/drivers/misc/sgi-gru/gruhandles.h
@@ -255,7 +255,8 @@ struct gru_tlb_fault_handle {
         unsigned int state:3;
         unsigned int fill3:1;
 
-        unsigned int cause:7;
+        unsigned int cause:6;
+        unsigned int cb_int:1;
         unsigned int fill4:1;
 
         unsigned int indexway:12;       /* DW 0 - high 32 */
diff --git a/drivers/misc/sgi-gru/grukservices.c b/drivers/misc/sgi-gru/grukservices.c
index 7d7952b27e03..ba6fcd963f30 100644
--- a/drivers/misc/sgi-gru/grukservices.c
+++ b/drivers/misc/sgi-gru/grukservices.c
@@ -406,7 +406,8 @@ static int gru_retry_exception(void *cb)
                         return CBS_IDLE;
 
                 gru_get_cb_exception_detail(cb, &excdet);
-                if (excdet.ecause & ~EXCEPTION_RETRY_BITS)
+                if ((excdet.ecause & ~EXCEPTION_RETRY_BITS) ||
+                                (excdet.cbrexecstatus & CBR_EXS_ABORT_OCC))
                         break;
                 if (retry-- == 0)
                         break;
diff --git a/drivers/misc/sgi-gru/gruprocfs.c b/drivers/misc/sgi-gru/gruprocfs.c
index 6ef4cb4b84c8..b5b9cf5c182e 100644
--- a/drivers/misc/sgi-gru/gruprocfs.c
+++ b/drivers/misc/sgi-gru/gruprocfs.c
@@ -89,6 +89,7 @@ static int statistics_show(struct seq_file *s, void *p)
         printstat(s, tlb_dropin_fail_fmm);
         printstat(s, tlb_dropin_fail_no_exception);
         printstat(s, tlb_dropin_fail_no_exception_war);
+        printstat(s, tfh_stale_on_fault);
         printstat(s, mmu_invalidate_range);
         printstat(s, mmu_invalidate_page);
         printstat(s, mmu_clear_flush_young);
diff --git a/drivers/misc/sgi-gru/grutables.h b/drivers/misc/sgi-gru/grutables.h
index 6dfb3e69411f..246c63883ebf 100644
--- a/drivers/misc/sgi-gru/grutables.h
+++ b/drivers/misc/sgi-gru/grutables.h
@@ -212,6 +212,7 @@ struct gru_stats_s {
         atomic_long_t tlb_dropin_fail_fmm;
         atomic_long_t tlb_dropin_fail_no_exception;
         atomic_long_t tlb_dropin_fail_no_exception_war;
+        atomic_long_t tfh_stale_on_fault;
         atomic_long_t mmu_invalidate_range;
         atomic_long_t mmu_invalidate_page;
         atomic_long_t mmu_clear_flush_young;