dmaengine: idxd: add configuration for concurrent work descriptor processing

mainline inclusion
from mainline-v6.1
commit 1f273752
category: feature
bugzilla: https://gitee.com/openeuler/intel-kernel/issues/I596WO
CVE: NA

Intel-SIG: commit 1f273752 dmaengine: idxd: add configuration for concurrent work descriptor processing.
Incremental backporting patches for DSA/IAA on Intel Xeon platform.

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

Add sysfs knob to allow control of the number of work descriptors that can
be concurrently processed by an engine in the group as a fraction of the
Maximum Work Descriptors in Progress value specified in ENGCAP register.
This control knob is part of toggle for QoS control.
Signed-off-by: NDave Jiang <dave.jiang@intel.com>
Co-developed-by: NFenghua Yu <fenghua.yu@intel.com>
Signed-off-by: NFenghua Yu <fenghua.yu@intel.com>
Link: https://lore.kernel.org/r/20220917161222.2835172-5-fenghua.yu@intel.comSigned-off-by: NVinod Koul <vkoul@kernel.org>
Signed-off-by: NXiaochen Shen <xiaochen.shen@intel.com>

Documentation/ABI/stable/sysfs-driver-dma-idxd

@@ -266,3 +266,15 @@ Contact:	dmaengine@vger.kernel.org
 Description:	Indicates the number of Read Buffers reserved for the use of
 		engines in the group. See DSA spec v1.2 9.2.18 GRPCFG Read Buffers
 		Reserved.
+
+What:		/sys/bus/dsa/devices/group<m>.<n>/desc_progress_limit
+Date:		Sept 14, 2022
+KernelVersion:	6.0.0
+Contact:	dmaengine@vger.kernel.org
+Description:	Allows control of the number of work descriptors that can be
+		concurrently processed by an engine in the group as a fraction
+		of the Maximum Work Descriptors in Progress value specified in
+		the ENGCAP register. The acceptable values are 0 (default),
+		1 (1/2 of max value), 2 (1/4 of the max value), and 3 (1/8 of
+		the max value). It's visible only on platforms that support
+		the capability.

drivers/dma/idxd/device.c

@@ -709,6 +709,7 @@ static void idxd_groups_clear_state(struct idxd_device *idxd)
 			group->tc_a = -1;
 			group->tc_b = -1;
 		}
+		group->desc_progress_limit = 0;
 	}
 }
 
@@ -765,10 +766,10 @@ static void idxd_group_config_write(struct idxd_group *group)
 
 	/* setup GRPFLAGS */
 	grpcfg_offset = GRPFLGCFG_OFFSET(idxd, group->id);
-	iowrite32(group->grpcfg.flags.bits, idxd->reg_base + grpcfg_offset);
-	dev_dbg(dev, "GRPFLAGS flags[%d: %#x]: %#x\n",
+	iowrite64(group->grpcfg.flags.bits, idxd->reg_base + grpcfg_offset);
+	dev_dbg(dev, "GRPFLAGS flags[%d: %#x]: %#llx\n",
 		group->id, grpcfg_offset,
-		ioread32(idxd->reg_base + grpcfg_offset));
+		ioread64(idxd->reg_base + grpcfg_offset));
 }
 
 static int idxd_groups_config_write(struct idxd_device *idxd)
@@ -929,6 +930,8 @@ static void idxd_group_flags_setup(struct idxd_device *idxd)
 			group->grpcfg.flags.rdbufs_allowed = group->rdbufs_allowed;
 		else
 			group->grpcfg.flags.rdbufs_allowed = idxd->max_rdbufs;
+
+		group->grpcfg.flags.desc_progress_limit = group->desc_progress_limit;
 	}
 }
 
@@ -1111,8 +1114,8 @@ static void idxd_group_load_config(struct idxd_group *group)
 	}
 
 	grpcfg_offset = GRPFLGCFG_OFFSET(idxd, group->id);
-	group->grpcfg.flags.bits = ioread32(idxd->reg_base + grpcfg_offset);
-	dev_dbg(dev, "GRPFLAGS flags[%d: %#x]: %#x\n",
+	group->grpcfg.flags.bits = ioread64(idxd->reg_base + grpcfg_offset);
+	dev_dbg(dev, "GRPFLAGS flags[%d: %#x]: %#llx\n",
 		group->id, grpcfg_offset, group->grpcfg.flags.bits);
 }
 

drivers/dma/idxd/idxd.h

@@ -96,6 +96,7 @@ struct idxd_group {
 	u8 rdbufs_reserved;
 	int tc_a;
 	int tc_b;
+	int desc_progress_limit;
 };
 
 struct idxd_pmu {

drivers/dma/idxd/registers.h

@@ -68,7 +68,8 @@ union group_cap_reg {
 		u64 total_rdbufs:8;	/* formerly total_tokens */
 		u64 rdbuf_ctrl:1;	/* formerly token_en */
 		u64 rdbuf_limit:1;	/* formerly token_limit */
-		u64 rsvd:46;
+		u64 progress_limit:1;	/* descriptor and batch descriptor */
+		u64 rsvd:45;
 	};
 	u64 bits;
 } __packed;
@@ -288,16 +289,18 @@ union msix_perm {
 
 union group_flags {
 	struct {
-		u32 tc_a:3;
-		u32 tc_b:3;
-		u32 rsvd:1;
-		u32 use_rdbuf_limit:1;
-		u32 rdbufs_reserved:8;
-		u32 rsvd2:4;
-		u32 rdbufs_allowed:8;
-		u32 rsvd3:4;
+		u64 tc_a:3;
+		u64 tc_b:3;
+		u64 rsvd:1;
+		u64 use_rdbuf_limit:1;
+		u64 rdbufs_reserved:8;
+		u64 rsvd2:4;
+		u64 rdbufs_allowed:8;
+		u64 rsvd3:4;
+		u64 desc_progress_limit:2;
+		u64 rsvd4:30;
 	};
-	u32 bits;
+	u64 bits;
 } __packed;
 
 struct grpcfg {

drivers/dma/idxd/sysfs.c

@@ -443,6 +443,37 @@ static struct device_attribute dev_attr_group_traffic_class_b =
 		__ATTR(traffic_class_b, 0644, group_traffic_class_b_show,
 		       group_traffic_class_b_store);
 
+static ssize_t group_desc_progress_limit_show(struct device *dev,
+					      struct device_attribute *attr,
+					      char *buf)
+{
+	struct idxd_group *group = confdev_to_group(dev);
+
+	return sysfs_emit(buf, "%d\n", group->desc_progress_limit);
+}
+
+static ssize_t group_desc_progress_limit_store(struct device *dev,
+					       struct device_attribute *attr,
+					       const char *buf, size_t count)
+{
+	struct idxd_group *group = confdev_to_group(dev);
+	int val, rc;
+
+	rc = kstrtoint(buf, 10, &val);
+	if (rc < 0)
+		return -EINVAL;
+
+	if (val & ~GENMASK(1, 0))
+		return -EINVAL;
+
+	group->desc_progress_limit = val;
+	return count;
+}
+
+static struct device_attribute dev_attr_group_desc_progress_limit =
+		__ATTR(desc_progress_limit, 0644, group_desc_progress_limit_show,
+		       group_desc_progress_limit_store);
+
 static struct attribute *idxd_group_attributes[] = {
 	&dev_attr_group_work_queues.attr,
 	&dev_attr_group_engines.attr,
@@ -454,11 +485,33 @@ static struct attribute *idxd_group_attributes[] = {
 	&dev_attr_group_read_buffers_reserved.attr,
 	&dev_attr_group_traffic_class_a.attr,
 	&dev_attr_group_traffic_class_b.attr,
+	&dev_attr_group_desc_progress_limit.attr,
 	NULL,
 };
 
+static bool idxd_group_attr_progress_limit_invisible(struct attribute *attr,
+						     struct idxd_device *idxd)
+{
+	return attr == &dev_attr_group_desc_progress_limit.attr &&
+	       !idxd->hw.group_cap.progress_limit;
+}
+
+static umode_t idxd_group_attr_visible(struct kobject *kobj,
+				       struct attribute *attr, int n)
+{
+	struct device *dev = container_of(kobj, struct device, kobj);
+	struct idxd_group *group = confdev_to_group(dev);
+	struct idxd_device *idxd = group->idxd;
+
+	if (idxd_group_attr_progress_limit_invisible(attr, idxd))
+		return 0;
+
+	return attr->mode;
+}
+
 static const struct attribute_group idxd_group_attribute_group = {
 	.attrs = idxd_group_attributes,
+	.is_visible = idxd_group_attr_visible,
 };
 
 static const struct attribute_group *idxd_group_attribute_groups[] = {