Update skc_obj_alloc for spl kmem caches that are backed by Linux

Currently, for certain sizes and classes of allocations we use
SPL caches that are backed by caches in the Linux Slab allocator
to reduce fragmentation and increase utilization of memory. The
way things are implemented for these caches as of now though is
that we don't keep any statistics of the allocations that we
make from these caches.

This patch enables the tracking of allocated objects in those
SPL caches by making the trade-off of grabbing the cache lock
at every object allocation and free to update the respective
counter.

Additionally, this patch makes those caches visible in the
/proc/spl/kmem/slab special file.

As a side note, enabling the specific counter for those caches
enables SDB to create a more user-friendly interface than
/proc/spl/kmem/slab that can also cross-reference data from
slabinfo. Here is for example the output of one of those
caches in SDB that outputs the name of the underlying Linux
cache, the memory of SPL objects allocated in that cache,
and the percentage of those objects compared to all the
objects in it:
```
> spl_kmem_caches | filter obj.skc_name == "zio_buf_512" | pp
name        ...            source total_memory util
----------- ... ----------------- ------------ ----
zio_buf_512 ... kmalloc-512[SLUB]       16.9MB    8
```

Reviewed-by: Matt Ahrens <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Serapheim Dimitropoulos <[email protected]>
Closes #9474

module/spl/spl-kmem-cache.c

@@ -1453,6 +1453,17 @@ spl_kmem_cache_alloc(spl_kmem_cache_t *skc, int flags)
 			obj = kmem_cache_alloc(slc, kmem_flags_convert(flags));
 		} while ((obj == NULL) && !(flags & KM_NOSLEEP));
 
+		if (obj != NULL) {
+			/*
+			 * Even though we leave everything up to the
+			 * underlying cache we still keep track of
+			 * how many objects we've allocated in it for
+			 * better debuggability.
+			 */
+			spin_lock(&skc->skc_lock);
+			skc->skc_obj_alloc++;
+			spin_unlock(&skc->skc_lock);
+		}
 		goto ret;
 	}
 
@@ -1526,6 +1537,9 @@ spl_kmem_cache_free(spl_kmem_cache_t *skc, void *obj)
 	 */
 	if (skc->skc_flags & KMC_SLAB) {
 		kmem_cache_free(skc->skc_linux_cache, obj);
+		spin_lock(&skc->skc_lock);
+		skc->skc_obj_alloc--;
+		spin_unlock(&skc->skc_lock);
 		return;
 	}
 

module/spl/spl-proc.c

@@ -437,11 +437,29 @@ slab_seq_show(struct seq_file *f, void *p)
 
 	ASSERT(skc->skc_magic == SKC_MAGIC);
 
-	/*
-	 * Backed by Linux slab see /proc/slabinfo.
-	 */
-	if (skc->skc_flags & KMC_SLAB)
+	if (skc->skc_flags & KMC_SLAB) {
+		/*
+		 * This cache is backed by a generic Linux kmem cache which
+		 * has its own accounting. For these caches we only track
+		 * the number of active allocated objects that exist within
+		 * the underlying Linux slabs. For the overall statistics of
+		 * the underlying Linux cache please refer to /proc/slabinfo.
+		 */
+		spin_lock(&skc->skc_lock);
+		seq_printf(f, "%-36s  ", skc->skc_name);
+		seq_printf(f, "0x%05lx %9s %9lu %8s %8u  "
+		    "%5s %5s %5s  %5s %5lu %5s  %5s %5s %5s\n",
+		    (long unsigned)skc->skc_flags,
+		    "-",
+		    (long unsigned)(skc->skc_obj_size * skc->skc_obj_alloc),
+		    "-",
+		    (unsigned)skc->skc_obj_size,
+		    "-", "-", "-", "-",
+		    (long unsigned)skc->skc_obj_alloc,
+		    "-", "-", "-", "-");
+		spin_unlock(&skc->skc_lock);
 		return (0);
+	}
 
 	spin_lock(&skc->skc_lock);
 	seq_printf(f, "%-36s  ", skc->skc_name);
@@ -461,9 +479,7 @@ slab_seq_show(struct seq_file *f, void *p)
 	    (long unsigned)skc->skc_obj_deadlock,
 	    (long unsigned)skc->skc_obj_emergency,
 	    (long unsigned)skc->skc_obj_emergency_max);
-
 	spin_unlock(&skc->skc_lock);
-
 	return (0);
 }