Thread (55 messages) 55 messages, 4 authors, 2025-12-16

Re: [PATCH 06/19] coresight: trbe: Refactor syndrome decoding

From: Suzuki K Poulose <suzuki.poulose@arm.com>
Date: 2025-12-02 11:06:52
Also in: linux-perf-users, lkml

Hi Leo

A couple of minor nits below. Otherwise looks like a nice cleanup


On 01/12/2025 11:21, Leo Yan wrote:
It gives priority to TRBSR_EL1.EA (external abort); an external abort
will immediately bail out and return an error.

Next, the syndrome decoding is refactored based on two levels of
information: the EC (Event Class) bits and the BSC (Trace Buffer Status
Code) bits.

If TRBSR_EL1.EC==0b000000, the driver continues parsing TRBSR_EL1.BSC to
identify the specific trace buffer event.  Otherwise, any non-zero
TRBSR_EL1.EC is treated as an error.

For error cases, the driver prints an error string and dumps registers
for debugging.
minor nit: Please avoid describing the code, which cod does well.
Briefly mention what you are doing, giving any additional contexts
that may not be evident from the code.

e.g:

"Add support for decoding the syndrome for TRBE, providing a verbose
description of the code for Fatal/unhandled events.
While at it add the new definitions from the latest Arm ARM for
EC and BSC"



No additional checks are required for wrap mode beyond verifying the
TRBSR_EL1.WRAP bit, even on units with overwrite errata, as this bit
quoted hunk ↗ jump to hunk
reliably indicates a buffer wrap.

Signed-off-by: Leo Yan <leo.yan@arm.com>
---
  drivers/hwtracing/coresight/coresight-trbe.c | 60 +++++++++++++++++++++-------
  drivers/hwtracing/coresight/coresight-trbe.h |  8 ++--
  2 files changed, 51 insertions(+), 17 deletions(-)
diff --git a/drivers/hwtracing/coresight/coresight-trbe.c b/drivers/hwtracing/coresight/coresight-trbe.c
index 9e565122816949b37b8ff5e4ba04cfbc317c6f25..28e2bfa68074f19ccaa4a737d00af577aea818fe 100644
--- a/drivers/hwtracing/coresight/coresight-trbe.c
+++ b/drivers/hwtracing/coresight/coresight-trbe.c
@@ -643,29 +643,61 @@ static void trbe_enable_hw(struct trbe_buf *buf)
  static enum trbe_fault_action trbe_get_fault_act(struct perf_output_handle *handle,
  						 u64 trbsr)
  {
+	const char *err_str;
  	int ec = get_trbe_ec(trbsr);
  	int bsc = get_trbe_bsc(trbsr);
-	struct trbe_buf *buf = etm_perf_sink_config(handle);
-	struct trbe_cpudata *cpudata = buf->cpudata;
  
  	WARN_ON(is_trbe_running(trbsr));
-	if (is_trbe_trg(trbsr) || is_trbe_abort(trbsr))
-		return TRBE_FAULT_ACT_FATAL;
  
-	if ((ec == TRBE_EC_STAGE1_ABORT) || (ec == TRBE_EC_STAGE2_ABORT))
-		return TRBE_FAULT_ACT_FATAL;
+	if (is_trbe_abort(trbsr)) {
+		err_str = "External abort";
+		goto out_fatal;
+	}
  
-	/*
-	 * If the trbe is affected by TRBE_WORKAROUND_OVERWRITE_FILL_MODE,
-	 * it might write data after a WRAP event in the fill mode.
-	 * Thus the check TRBPTR == TRBBASER will not be honored.
-	 */
-	if ((is_trbe_wrap(trbsr) && (ec == TRBE_EC_OTHERS) && (bsc == TRBE_BSC_FILLED)) &&
-	    (trbe_may_overwrite_in_fill_mode(cpudata) ||
-	     get_trbe_write_pointer() == get_trbe_base_pointer()))
+	switch (ec) {
+	case TRBE_EC_OTHERS:
+		break;
+	case TRBE_EC_BUF_MGMT_IMPL:
+		err_str = "Unexpected implemented management";
minor nit: "Unexpected IMPDEF buffer management event"


Suzuki
quoted hunk ↗ jump to hunk
+		goto out_fatal;
+	case TRBE_EC_GP_CHECK_FAULT:
+		err_str = "Granule Protection Check fault";
+		goto out_fatal;
+	case TRBE_EC_STAGE1_ABORT:
+		err_str = "Stage 1 data abort";
+		goto out_fatal;
+	case TRBE_EC_STAGE2_ABORT:
+		err_str = "Stage 2 data abort";
+		goto out_fatal;
+	default:
+		err_str = "Unknown error code";
+		goto out_fatal;
+	}
+
+	switch (bsc) {
+	case TRBE_BSC_NOT_STOPPED:
+		break;
+	case TRBE_BSC_FILLED:
+		break;
+	case TRBE_BSC_TRIGGERED:
+		err_str = "Unexpected trigger status";
+		goto out_fatal;
+	default:
+		err_str = "Unexpected buffer status code";
+		goto out_fatal;
+	}
+
+	if (is_trbe_wrap(trbsr))
  		return TRBE_FAULT_ACT_WRAP;
  
  	return TRBE_FAULT_ACT_SPURIOUS;
+
+out_fatal:
+	pr_err_ratelimited("%s on CPU %d [TRBSR=0x%016llx, TRBPTR=0x%016llx, TRBLIMITR=0x%016llx]\n",
+			   err_str, smp_processor_id(), trbsr,
+			   read_sysreg_s(SYS_TRBPTR_EL1),
+			   read_sysreg_s(SYS_TRBLIMITR_EL1));
+	return TRBE_FAULT_ACT_FATAL;
  }
  
  static unsigned long trbe_get_trace_size(struct perf_output_handle *handle,
diff --git a/drivers/hwtracing/coresight/coresight-trbe.h b/drivers/hwtracing/coresight/coresight-trbe.h
index 45202c48accec7c86ba56130e2737bc2d1830fae..d7f7cd763c0c7139cf322b7336ee563073e3bea0 100644
--- a/drivers/hwtracing/coresight/coresight-trbe.h
+++ b/drivers/hwtracing/coresight/coresight-trbe.h
@@ -35,9 +35,11 @@ static inline bool is_trbe_enabled(void)
  	return trblimitr & TRBLIMITR_EL1_E;
  }
  
-#define TRBE_EC_OTHERS		0
-#define TRBE_EC_STAGE1_ABORT	36
-#define TRBE_EC_STAGE2_ABORT	37
+#define TRBE_EC_OTHERS			0x0
+#define TRBE_EC_GP_CHECK_FAULT		0X1e
+#define TRBE_EC_BUF_MGMT_IMPL		0x1f
+#define TRBE_EC_STAGE1_ABORT		0x24
+#define TRBE_EC_STAGE2_ABORT		0x25
  
  static inline int get_trbe_ec(u64 trbsr)
  {
  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help