Add an API for edac device, for mulriple errors

----------------------------------------------------------------------

Review

Quoted 6 lines · Hanna Hawa

(quoted diff) +6 -0

+void edac_device_handle_ce(struct edac_device_ctl_info *edac_dev,
+			int inst_nr, int block_nr, const char *msg)
+{
+	__edac_device_handle_ce(edac_dev, 1, inst_nr, block_nr, msg);
+}
+EXPORT_SYMBOL_GPL(edac_device_handle_ce);

Eww, I don't like that: exporting the function *and* the __ counterpart. The user will get confused and that is unnecessary.

It is *not* the counterpart. The __* version already has the...

The first patch only adds functionality but keeps the abi. Thus it makes a backport easier.

----------------------------------------------------------------------

Quoted 1 line · Robert Richter

It is *not* the counterpart. The __* version already has the...

Lemme cut to the chase:

"Make the main workhorse the "count" functions which can log a @count of errors. Have the current APIs edac_device_handle_{ce,ue}() call the _count() variants and this way keep the exported symbols number unchanged."

I want to see exactly *two* pairs of functions:

edac_device_handle_{ce,ue}_count - logs a @count of errors
edac_device_handle_{ce,ue} - logs a single error

Not three pairs. I.e., the "__" versions are not needed.

Quoted 2 lines

The first patch only adds functionality but keeps the abi. Thus it makes a backport easier.

Works just the same with my version - single backport.

If you move to static inline for edac_device_handle_{ce,ue} the symbols vanish and this breaks the abi. That's why the split in two patches.

Your comment to not have a __ version as a third variant of the interface makes sense to me. But to keep ABI your patch still needs to be split.

Quoted 3 lines · Robert Richter

If you move to static inline for edac_device_handle_{ce,ue} the symbols vanish and this breaks the abi. That's why the split in two patches.

ABI issues do not concern upstream. And that coming from me working at a company who dance a lot to make ABI happy.

Also, I'm missing the reasoning why you use the ABI as an argument at
all: do you know of a particular case where people are thinking of backporting this or this is all hypothetical.

Quoted 3 lines

Your comment to not have a __ version as a third variant of the interface makes sense to me. But to keep ABI your patch still needs to be split.

Not really - normally, when you fix ABI issues with symbols disappearing, all of a sudden, you add dummy ones so that the ABI checker is happy.

Review

Quoted 6 lines · Hanna Hawa

(quoted diff) +6 -0

+void edac_device_handle_ce(struct edac_device_ctl_info *edac_dev,
+			int inst_nr, int block_nr, const char *msg)
+{
+	__edac_device_handle_ce(edac_dev, 1, inst_nr, block_nr, msg);
+}
+EXPORT_SYMBOL_GPL(edac_device_handle_ce);

Eww, I don't like that: exporting the function *and* the __ counterpart. The user will get confused and that is unnecessary.

See below for a better version. This way you solve the whole deal with a single patch.

---
From: Hanna Hawa <[email protected]>
Date: Mon, 23 Sep 2019 20:17:40 +0100
Subject: [PATCH] EDAC/device: Rework error logging API

Make the main workhorse the "count" functions which can log a @count of errors. Have the current APIs edac_device_handle_{ce,ue}() call the _count() variants and this way keep the exported symbols number unchanged.

[ bp: Rewrite. ]

Signed-off-by Hanna Hawa <[email protected]>

Signed-off-by Borislav Petkov <[email protected]>

Cc [email protected]

Cc [email protected]

Cc [email protected]

Cc James Morse <[email protected]>

Cc [email protected]

Cc linux-edac <[email protected]>

Cc Mauro Carvalho Chehab <[email protected]>

Cc [email protected]

Cc [email protected]

Cc Tony Luck <[email protected]>

Link https://lkml.kernel.org/r/[email protected] open here

---
  drivers/edac/edac_device.c | 44 ++++++++++++++++---------------
  drivers/edac/edac_device.h | 54 ++++++++++++++++++++++++++++++--------
  2 files changed, 66 insertions(+), 32 deletions(-)

drivers/edac/edac_device.c +0 -0

index 65cf2b9355c4..d4d8bed5b55d 100644
@@ -555,8 +555,9 @@ static inline int edac_device_get_panic_on_ue(struct edac_device_ctl_info
  	return edac_dev->panic_on_ue;

drivers/edac/edac_device.c +20 -19

@@ -582,23 +583,24 @@ void edac_device_handle_ce(struct edac_device_ctl_info *edac_dev,
 
  	if (instance->nr_blocks > 0) {
  		block = instance->blocks + block_nr;
-		block->counters.ce_count++;
+		block->counters.ce_count += count;
  	}
 
  	/* Propagate the count up the 'totals' tree */
-	instance->counters.ce_count++;
-	edac_dev->counters.ce_count++;
+	instance->counters.ce_count += count;
+	edac_dev->counters.ce_count += count;
 
  	if (edac_device_get_log_ce(edac_dev))
  		edac_device_printk(edac_dev, KERN_WARNING,
-				"CE: %s instance: %s block: %s '%s'\n",
-				edac_dev->ctl_name, instance->name,
-				block ? block->name : "N/A", msg);
+				   "CE: %s instance: %s block: %s count: %d '%s'\n",
+				   edac_dev->ctl_name, instance->name,
+				   block ? block->name : "N/A", count, msg);
  }
-EXPORT_SYMBOL_GPL(edac_device_handle_ce);
+EXPORT_SYMBOL_GPL(edac_device_handle_ce_count);
 
-void edac_device_handle_ue(struct edac_device_ctl_info *edac_dev,
-			int inst_nr, int block_nr, const char *msg)
+void edac_device_handle_ue_count(struct edac_device_ctl_info *edac_dev,
+				 unsigned int count, int inst_nr, int block_nr,
+				 const char *msg)
  {
  	struct edac_device_instance *instance;
  	struct edac_device_block *block = NULL;

@@ -624,22 +626,22 @@ void edac_device_handle_ue(struct edac_device_ctl_info *edac_dev,
 
  	if (instance->nr_blocks > 0) {
  		block = instance->blocks + block_nr;
-		block->counters.ue_count++;
+		block->counters.ue_count += count;
  	}
 
  	/* Propagate the count up the 'totals' tree */
-	instance->counters.ue_count++;
-	edac_dev->counters.ue_count++;
+	instance->counters.ue_count += count;
+	edac_dev->counters.ue_count += count;
 
  	if (edac_device_get_log_ue(edac_dev))
  		edac_device_printk(edac_dev, KERN_EMERG,
-				"UE: %s instance: %s block: %s '%s'\n",
-				edac_dev->ctl_name, instance->name,
-				block ? block->name : "N/A", msg);
+				   "UE: %s instance: %s block: %s count: %d '%s'\n",
+				   edac_dev->ctl_name, instance->name,
+				   block ? block->name : "N/A", count, msg);
 
  	if (edac_device_get_panic_on_ue(edac_dev))
-		panic("EDAC %s: UE instance: %s block %s '%s'\n",
-			edac_dev->ctl_name, instance->name,
-			block ? block->name : "N/A", msg);
+		panic("EDAC %s: UE instance: %s block %s count: %d '%s'\n",
+		      edac_dev->ctl_name, instance->name,
+		      block ? block->name : "N/A", count, msg);
  }
-EXPORT_SYMBOL_GPL(edac_device_handle_ue);
+EXPORT_SYMBOL_GPL(edac_device_handle_ue_count);

drivers/edac/edac_device.h +43 -11

index 1aaba74ae411..c4c0e0bdce14 100644
@@ -286,27 +286,60 @@ extern int edac_device_add_device(struct edac_device_ctl_info *edac_dev);
  extern struct edac_device_ctl_info *edac_device_del_device(struct device *dev);
 
  /**
- * edac_device_handle_ue():
- *	perform a common output and handling of an 'edac_dev' UE event
+ * Log correctable errors.
   *
   * @edac_dev: pointer to struct &edac_device_ctl_info
- * @inst_nr: number of the instance where the UE error happened
- * @block_nr: number of the block where the UE error happened
+ * @inst_nr: number of the instance where the CE error happened
+ * @count: Number of errors to log.
+ * @block_nr: number of the block where the CE error happened
+ * @msg: message to be printed
+ */
+void edac_device_handle_ce_count(struct edac_device_ctl_info *edac_dev,
+				 unsigned int count, int inst_nr, int block_nr,
+				 const char *msg);
+
+/**
+ * Log uncorrectable errors.
+ *
+ * @edac_dev: pointer to struct &edac_device_ctl_info
+ * @inst_nr: number of the instance where the CE error happened
+ * @count: Number of errors to log.
+ * @block_nr: number of the block where the CE error happened
   * @msg: message to be printed
   */
-extern void edac_device_handle_ue(struct edac_device_ctl_info *edac_dev,
-				int inst_nr, int block_nr, const char *msg);
+void edac_device_handle_ue_count(struct edac_device_ctl_info *edac_dev,
+				 unsigned int count, int inst_nr, int block_nr,
+				 const char *msg);
+
  /**
- * edac_device_handle_ce():
- *	perform a common output and handling of an 'edac_dev' CE event
+ * edac_device_handle_ce(): Log a single correctable error
   *
   * @edac_dev: pointer to struct &edac_device_ctl_info
   * @inst_nr: number of the instance where the CE error happened
   * @block_nr: number of the block where the CE error happened
   * @msg: message to be printed
   */
-extern void edac_device_handle_ce(struct edac_device_ctl_info *edac_dev,
-				int inst_nr, int block_nr, const char *msg);
+static inline void
+edac_device_handle_ce(struct edac_device_ctl_info *edac_dev, int inst_nr,
+		      int block_nr, const char *msg)
+{
+	edac_device_handle_ce_count(edac_dev, 1, inst_nr, block_nr, msg);
+}
+
+/**
+ * edac_device_handle_ue(): Log a single uncorrectable error
+ *
+ * @edac_dev: pointer to struct &edac_device_ctl_info
+ * @inst_nr: number of the instance where the UE error happened
+ * @block_nr: number of the block where the UE error happened
+ * @msg: message to be printed
+ */
+static inline void
+edac_device_handle_ue(struct edac_device_ctl_info *edac_dev, int inst_nr,
+		      int block_nr, const char *msg)
+{
+	edac_device_handle_ue_count(edac_dev, 1, inst_nr, block_nr, msg);
+}
 
  /**
   * edac_device_alloc_index: Allocate a unique device index number

@@ -316,5 +349,4 @@ extern void edac_device_handle_ce(struct edac_device_ctl_info *edac_dev,
   */
  extern int edac_device_alloc_index(void);
  extern const char *edac_layer_name[];
-
  #endif

Missing count check, same in *_ue_count():
if (count)

Add an API for EDAC device to report for multiple errors, and move the old report function to use the new API.

Changes from v3:
----------------
- Move count check to inline function
- Fix count variable describtion (Reported-by: kbuild test robot <[email protected]>)