[PATCH v3 0/5] pseries: Move memory hotplug to the kernel

[PATCH v3 0/5] pseries: Move memory hotplug to the kernel

[Nathan Fontenot]

In order to improve hotplug support for cpus, memory, and pci devices in
the PowerVM and PowerKVM environments, the handling of hotplugging devices
should be updated to perform the hotplug operation entirely in the kernel.
This patch set begins to address this by providing a framework and a common
entry point for hotplug operations for the PowerVM anbd PowerKVM environments.
This patch set also moves memory hotplug into the kernel. Patches to follow
will do the same for cpus and pci devices.

To provide background, the current handling of memory hotplug is
done by the drmgr command in user space. This command is invoked when memory
add/remove requests are made at the HMC and conveyed to a partition
through the RSCT framework. The drmgr command then performs parts
of the hotplug in user-space and makes requests to the kernel to perform
other pieces. This is not really ideal, we can do everything in the
kernel and do it faster.

In this patch set the handling of hotplug requests for pseries kernels
uses rtas hotplug events to communicate the hotplug request. For PowerKVM
systems the rtas hotplug request is communicated to the guest by qemu
using the ras epow interrupt, the current scheme also used for pci hotplug.
For PowerVM systems the drmgr command will be updated to communicate the
hotplug request by writing to /sys/kernel/dlpar. The format for hotplug
requests for this file is of the form "<resource> <action> <id_type> <id>".
The kernel will parse this request and create a rtas hotplug section that
can be passed to a common kernel hotplug entry point for PowerVM and PowerKVM
environments.

-Nathan

Patch 1/5
- Add definition of hotplug rtas event sections.

Patch 2/5
- export the dlpar_[acquire|release]_drc() routines.

Patch 3/5
- Create the new /sys/kernel/dpar interface

Patch 4/5
- Implement memory hotplug add in the kernel.

Patch 5/5
- Implement memory hotplug remove in the kernel.

 include/asm/rtas.h                 |   26 ++
 platforms/pseries/dlpar.c          |  118 ++++++++-
 platforms/pseries/hotplug-memory.c |  477 ++++++++++++++++++++++++++++++++++++-
 platforms/pseries/pseries.h        |   12 
 4 files changed, 629 insertions(+), 4 deletions(-)

[PATCH v3 1/5]pseries: Define rtas hotplug event sections

[Nathan Fontenot]

In order to handle device hotplug in the kernel on pseries the hotplug
request will be communicated in the kernel in the form of a
rtas hotplug event. This patch adds the definition of rtas hotplug event
sections.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
---
 arch/powerpc/include/asm/rtas.h |   26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/arch/powerpc/include/asm/rtas.h b/arch/powerpc/include/asm/rtas.h
index 2e23e92..eddc846 100644
--- a/arch/powerpc/include/asm/rtas.h
+++ b/arch/powerpc/include/asm/rtas.h
@@ -273,6 +273,7 @@ inline uint32_t rtas_ext_event_company_id(struct rtas_ext_event_log_v6 *ext_log)
 #define PSERIES_ELOG_SECT_ID_MANUFACT_INFO	(('M' << 8) | 'I')
 #define PSERIES_ELOG_SECT_ID_CALL_HOME		(('C' << 8) | 'H')
 #define PSERIES_ELOG_SECT_ID_USER_DEF		(('U' << 8) | 'D')
+#define PSERIES_ELOG_SECT_ID_HOTPLUG		(('H' << 8) | 'P')
 
 /* Vendor specific Platform Event Log Format, Version 6, section header */
 struct pseries_errorlog {
@@ -296,6 +297,31 @@ inline uint16_t pseries_errorlog_length(struct pseries_errorlog *sect)
 	return be16_to_cpu(sect->length);
 }
 
+/* RTAS pseries hotplug errorlog section */
+struct pseries_hp_errorlog {
+	u8	resource;
+	u8	action;
+	u8	id_type;
+	u8	reserved;
+	union {
+		__be32	drc_index;
+		__be32	drc_count;
+		char	drc_name[1];
+	} _drc_u;
+};
+
+#define PSERIES_HP_ELOG_RESOURCE_CPU	1
+#define PSERIES_HP_ELOG_RESOURCE_MEM	2
+#define PSERIES_HP_ELOG_RESOURCE_SLOT	3
+#define PSERIES_HP_ELOG_RESOURCE_PHB	4
+
+#define PSERIES_HP_ELOG_ACTION_ADD	1
+#define PSERIES_HP_ELOG_ACTION_REMOVE	2
+
+#define PSERIES_HP_ELOG_ID_DRC_NAME	1
+#define PSERIES_HP_ELOG_ID_DRC_INDEX	2
+#define PSERIES_HP_ELOG_ID_DRC_COUNT	3
+
 struct pseries_errorlog *get_pseries_errorlog(struct rtas_error_log *log,
 					      uint16_t section_id);
 

[PATCH v3 2/5] pseries: Export the acquire/release drc index routines

[Nathan Fontenot]

Export the routines to acquire and release a drc index.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
---
 arch/powerpc/platforms/pseries/pseries.h |    2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/powerpc/platforms/pseries/pseries.h b/arch/powerpc/platforms/pseries/pseries.h
index 1796c54..239bee5 100644
--- a/arch/powerpc/platforms/pseries/pseries.h
+++ b/arch/powerpc/platforms/pseries/pseries.h
@@ -60,6 +60,8 @@ extern struct device_node *dlpar_configure_connector(__be32,
 						struct device_node *);
 extern int dlpar_attach_node(struct device_node *);
 extern int dlpar_detach_node(struct device_node *);
+int dlpar_acquire_drc(u32 drc_index);
+int dlpar_release_drc(u32 drc_index);
 
 /* PCI root bridge prepare function override for pseries */
 struct pci_host_bridge;

[PATCH v3 3/5] pseries: Create new device hotplug entry point

[Nathan Fontenot]

The current hotplug (or dlpar) of devices (the process is generally the
same for memory, cpu, and pci) on PowerVM systems is initiated
from the HMC, which communicates the request to the partitions through
the RSCT framework. The RSCT framework then invokes the drmgr command.
The drmgr command performs the hotplug operation by doing some pieces,
such as most of the rtas calls and device tree parsing, in userspace
and make requests to the kernel to online/offline the device, update the
device tree and add/remove the device.

For PowerKVM the approach for device hotplug is to follow what is currently
being done for pci hotplug. A hotplug request is initiated from the host.
QEMU then generates an EPOW interrupt to the guest which causes the guest
to make the rtas,check-exception call. In QEMU, the rtas,check-exception call
returns a rtas hotplug event to the guest.

Please note that the current pci hotplug path for PowerKVM involves the
kernel receiving the rtas hotplug event, passing it to rtas_errd in
userspace, and having rtas_errd invoke drmgr. The drmgr command then
handles the request as described above for PowerVM systems.

There is no need for this circuitous route, we should just handle the entire
hotplug of devices in the kernel. What I am planning is to enable this
by moving the code to handle hotplug from drmgr into the kernel to
provide a single path for handling device hotplug for both PowerVM and
PowerKVM systems. This patch provides the common iframework and entry point.
For PowerKVM a future update to the kernel rtas code will recognize rtas
hotplug events returned from rtas,check-exception calls and use the common
entry point to handle hotplug of the device.

For PowerVM systems, This patch creates /sys/kernel/dlpar that can be
used by the drmgr command to initiate hotplug requests. In order to do
this a string of the format "<resource> <action> <id_type> <id>" is
written to this file. The string consists of a resource (cpu, memory, pci,
phb), an action (add or remove), an id_type (count, drc index, drc name),
and the corresponding id. The kernel will parse the string and create a
rtas hotplug section that can be passed to the common entry point for
handling hotplug requests.

It should be noted that there is no chance of updating how we receive
hotplug (dlpar) requests from the HMC on PowerVM systems.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
---
Changes from previous version:
- The interface between user-space and the kernel is a text command
  instead of a binary blob written to the sysfs file. The file created
  is stil /sys/kernel/dlpar.
- Corrected extraneous and missing "break"'s.
- Converted the rtas hotplug section to cpu format before attempting
  hotplug operations instead of converting as needed.

 arch/powerpc/platforms/pseries/dlpar.c          |  118 +++++++++++++++++++++++
 arch/powerpc/platforms/pseries/hotplug-memory.c |   19 ++++
 arch/powerpc/platforms/pseries/pseries.h        |   10 ++
 3 files changed, 145 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/platforms/pseries/dlpar.c b/arch/powerpc/platforms/pseries/dlpar.c
index c22bb1b..b4b1109 100644
--- a/arch/powerpc/platforms/pseries/dlpar.c
+++ b/arch/powerpc/platforms/pseries/dlpar.c
@@ -10,6 +10,8 @@
  * 2 as published by the Free Software Foundation.
  */
 
+#define pr_fmt(fmt)	"dlpar: " fmt
+
 #include <linux/kernel.h>
 #include <linux/notifier.h>
 #include <linux/spinlock.h>
@@ -535,13 +537,125 @@ static ssize_t dlpar_cpu_release(const char *buf, size_t count)
 	return count;
 }
 
+#endif /* CONFIG_ARCH_CPU_PROBE_RELEASE */
+
+static int handle_dlpar_errorlog(struct pseries_hp_errorlog *hp_elog)
+{
+	int rc;
+
+	/* pseries error logs are in BE format, convert to cpu type */
+	switch (hp_elog->id_type) {
+	case PSERIES_HP_ELOG_ID_DRC_COUNT:
+		hp_elog->_drc_u.drc_count =
+					be32_to_cpu(hp_elog->_drc_u.drc_count);
+		break;
+	case PSERIES_HP_ELOG_ID_DRC_INDEX:
+		hp_elog->_drc_u.drc_index =
+					be32_to_cpu(hp_elog->_drc_u.drc_index);
+	}
+
+	switch (hp_elog->resource) {
+	case PSERIES_HP_ELOG_RESOURCE_MEM:
+		rc = dlpar_memory(hp_elog);
+		break;
+	default:
+		pr_warn_ratelimited("Invalid resource (%d) specified\n",
+				    hp_elog->resource);
+		rc = -EINVAL;
+	}
+
+	return rc;
+}
+
+static ssize_t dlpar_store(struct class *class, struct class_attribute *attr,
+			   const char *buf, size_t count)
+{
+	struct pseries_hp_errorlog *hp_elog;
+	const char *arg;
+	int rc;
+
+	hp_elog = kzalloc(sizeof(*hp_elog), GFP_KERNEL);
+	if (!hp_elog) {
+		rc = -ENOMEM;
+		goto dlpar_store_out;
+	}
+
+	/* Parse out the request from the user, this will be in the form
+	 * <resource> <action> <id_type> <id>
+	 */
+	arg = buf;
+	if (!strncmp(arg, "memory", 6)) {
+		hp_elog->resource = PSERIES_HP_ELOG_RESOURCE_MEM;
+		arg += strlen("memory ");
+	} else {
+		pr_err("Invalid resource specified: \"%s\"\n", buf);
+		rc = -EINVAL;
+		goto dlpar_store_out;
+	}
+
+	if (!strncmp(arg, "add", 3)) {
+		hp_elog->action = PSERIES_HP_ELOG_ACTION_ADD;
+		arg += strlen("add ");
+	} else if (!strncmp(arg, "remove", 6)) {
+		hp_elog->action = PSERIES_HP_ELOG_ACTION_REMOVE;
+		arg += strlen("remove ");
+	} else {
+		pr_err("Invalid action specified: \"%s\"\n", buf);
+		rc = -EINVAL;
+		goto dlpar_store_out;
+	}
+
+	if (!strncmp(arg, "index", 5)) {
+		u32 index;
+
+		hp_elog->id_type = PSERIES_HP_ELOG_ID_DRC_INDEX;
+		arg += strlen("index ");
+		if (kstrtou32(arg, 0, &index)) {
+			rc = -EINVAL;
+			pr_err("Invalid drc_index specified: \"%s\"\n", buf);
+			goto dlpar_store_out;
+		}
+
+		hp_elog->_drc_u.drc_index = cpu_to_be32(index);
+	} else if (!strncmp(arg, "count", 5)) {
+		u32 count;
+
+		hp_elog->id_type = PSERIES_HP_ELOG_ID_DRC_COUNT;
+		arg += strlen("count ");
+		if (kstrtou32(arg, 0, &count)) {
+			rc = -EINVAL;
+			pr_err("Invalid count specified: \"%s\"\n", buf);
+			goto dlpar_store_out;
+		}
+
+		hp_elog->_drc_u.drc_count = cpu_to_be32(count);
+	} else {
+		pr_err("Invalid id_type specified: \"%s\"\n", buf);
+		rc = -EINVAL;
+		goto dlpar_store_out;
+	}
+
+	rc = handle_dlpar_errorlog(hp_elog);
+
+dlpar_store_out:
+	kfree(hp_elog);
+	return rc ? rc : count;
+}
+
+static CLASS_ATTR(dlpar, S_IWUSR, NULL, dlpar_store);
+
 static int __init pseries_dlpar_init(void)
 {
+	int rc;
+
+#ifdef CONFIG_ARCH_CPU_PROBE_RELEASE
 	ppc_md.cpu_probe = dlpar_cpu_probe;
 	ppc_md.cpu_release = dlpar_cpu_release;
+#endif /* CONFIG_ARCH_CPU_PROBE_RELEASE */
 
-	return 0;
+	rc = sysfs_create_file(kernel_kobj, &class_attr_dlpar.attr);
+
+	return rc;
 }
 machine_device_initcall(pseries, pseries_dlpar_init);
 
-#endif /* CONFIG_ARCH_CPU_PROBE_RELEASE */
diff --git a/arch/powerpc/platforms/pseries/hotplug-memory.c b/arch/powerpc/platforms/pseries/hotplug-memory.c
index fa41f0d..211d0bf 100644
--- a/arch/powerpc/platforms/pseries/hotplug-memory.c
+++ b/arch/powerpc/platforms/pseries/hotplug-memory.c
@@ -9,6 +9,8 @@
  *      2 of the License, or (at your option) any later version.
  */
 
+#define pr_fmt(fmt)	"pseries-hotplug-mem: " fmt
+
 #include <linux/of.h>
 #include <linux/of_address.h>
 #include <linux/memblock.h>
@@ -134,6 +136,23 @@ static inline int pseries_remove_mem_node(struct device_node *np)
 }
 #endif /* CONFIG_MEMORY_HOTREMOVE */
 
+int dlpar_memory(struct pseries_hp_errorlog *hp_elog)
+{
+	int rc = 0;
+
+	lock_device_hotplug();
+
+	switch (hp_elog->action) {
+	default:
+		pr_err("Invalid action (%d) specified\n", hp_elog->action);
+		rc = -EINVAL;
+		break;
+	}
+
+	unlock_device_hotplug();
+	return rc;
+}
+
 static int pseries_add_mem_node(struct device_node *np)
 {
 	const char *type;
diff --git a/arch/powerpc/platforms/pseries/pseries.h b/arch/powerpc/platforms/pseries/pseries.h
index 239bee5..40e0339 100644
--- a/arch/powerpc/platforms/pseries/pseries.h
+++ b/arch/powerpc/platforms/pseries/pseries.h
@@ -11,6 +11,7 @@
 #define _PSERIES_PSERIES_H
 
 #include <linux/interrupt.h>
+#include <asm/rtas.h>
 
 struct device_node;
 
@@ -63,6 +64,15 @@ extern int dlpar_detach_node(struct device_node *);
 int dlpar_acquire_drc(u32 drc_index);
 int dlpar_release_drc(u32 drc_index);
 
+#ifdef CONFIG_MEMORY_HOTPLUG
+int dlpar_memory(struct pseries_hp_errorlog *hp_elog);
+#else
+static inline int dlpar_memory(struct pseries_hp_errorlog *hp_elog)
+{
+	return -EOPNOTSUPP;
+}
+#endif
+
 /* PCI root bridge prepare function override for pseries */
 struct pci_host_bridge;
 int pseries_root_bridge_prepare(struct pci_host_bridge *bridge);

[PATCH v3 4/5] pseries: Implement memory hotplug add in the kernel

[Nathan Fontenot]

This patch adds the ability to do memory hotplug add in the kernel.

Currently the operation to hotplug add memory is handled by the drmgr
command which performs the operation by performing some work in user-space
and making requests to the kernel to handle other pieces. By moving all
of the work to the kernel we can do the add faster, and provide a common
code path to do memory hotplug for both the PowerVM and PowerKVM environments.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
---
Changes from previous version:
- converted uint32_t to u32
- removed un-needed phys_addr/page_block checking
- Updated the messaging to properly report the success/failure of memory adds
 
 arch/powerpc/platforms/pseries/hotplug-memory.c |  266 +++++++++++++++++++++++
 1 file changed, 265 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/platforms/pseries/hotplug-memory.c b/arch/powerpc/platforms/pseries/hotplug-memory.c
index 211d0bf..f5eec0f 100644
--- a/arch/powerpc/platforms/pseries/hotplug-memory.c
+++ b/arch/powerpc/platforms/pseries/hotplug-memory.c
@@ -16,6 +16,7 @@
 #include <linux/memblock.h>
 #include <linux/memory.h>
 #include <linux/memory_hotplug.h>
+#include <linux/slab.h>
 
 #include <asm/firmware.h>
 #include <asm/machdep.h>
@@ -23,6 +24,8 @@
 #include <asm/sparsemem.h>
 #include "pseries.h"
 
+static bool rtas_hp_event;
+
 unsigned long pseries_memory_block_size(void)
 {
 	struct device_node *np;
@@ -66,6 +69,67 @@ unsigned long pseries_memory_block_size(void)
 	return memblock_size;
 }
 
+static void dlpar_free_drconf_property(struct property *prop)
+{
+	kfree(prop->name);
+	kfree(prop->value);
+	kfree(prop);
+}
+
+static struct property *dlpar_clone_drconf_property(struct device_node *dn)
+{
+	struct property *prop, *new_prop;
+	struct of_drconf_cell *lmbs;
+	u32 num_lmbs, *p;
+	int i;
+
+	prop = of_find_property(dn, "ibm,dynamic-memory", NULL);
+	if (!prop)
+		return NULL;
+
+	new_prop = kzalloc(sizeof(*new_prop), GFP_KERNEL);
+	if (!new_prop)
+		return NULL;
+
+	new_prop->name = kstrdup(prop->name, GFP_KERNEL);
+	new_prop->value = kmalloc(prop->length, GFP_KERNEL);
+	if (!new_prop->name || !new_prop->value) {
+		dlpar_free_drconf_property(new_prop);
+		return NULL;
+	}
+
+	memcpy(new_prop->value, prop->value, prop->length);
+	new_prop->length = prop->length;
+
+	/* Convert the property to cpu endian-ness */
+	p = new_prop->value;
+	*p = be32_to_cpu(*p);
+
+	num_lmbs = *p++;
+	lmbs = (struct of_drconf_cell *)p;
+
+	for (i = 0; i < num_lmbs; i++) {
+		lmbs[i].base_addr = be64_to_cpu(lmbs[i].base_addr);
+		lmbs[i].drc_index = be32_to_cpu(lmbs[i].drc_index);
+		lmbs[i].flags = be32_to_cpu(lmbs[i].flags);
+	}
+
+	return new_prop;
+}
+
+static struct memory_block *lmb_to_memblock(struct of_drconf_cell *lmb)
+{
+	unsigned long section_nr;
+	struct mem_section *mem_sect;
+	struct memory_block *mem_block;
+
+	section_nr = pfn_to_section_nr(PFN_DOWN(lmb->base_addr));
+	mem_sect = __nr_to_section(section_nr);
+
+	mem_block = find_memory_block(mem_sect);
+	return mem_block;
+}
+
 #ifdef CONFIG_MEMORY_HOTREMOVE
 static int pseries_remove_memblock(unsigned long base, unsigned int memblock_size)
 {
@@ -136,19 +200,216 @@ static inline int pseries_remove_mem_node(struct device_node *np)
 }
 #endif /* CONFIG_MEMORY_HOTREMOVE */
 
+static int dlpar_add_lmb(struct of_drconf_cell *lmb)
+{
+	struct memory_block *mem_block;
+	unsigned long block_sz;
+	int nid, rc;
+
+	if (lmb->flags & DRCONF_MEM_ASSIGNED)
+		return -EINVAL;
+
+	block_sz = memory_block_size_bytes();
+
+	rc = dlpar_acquire_drc(lmb->drc_index);
+	if (rc)
+		return rc;
+
+	/* Find the node id for this address */
+	nid = memory_add_physaddr_to_nid(lmb->base_addr);
+
+	/* Add the memory */
+	rc = add_memory(nid, lmb->base_addr, block_sz);
+	if (rc) {
+		dlpar_release_drc(lmb->drc_index);
+		return rc;
+	}
+
+	/* Register this block of memory */
+	rc = memblock_add(lmb->base_addr, block_sz);
+	if (rc) {
+		remove_memory(nid, lmb->base_addr, block_sz);
+		dlpar_release_drc(lmb->drc_index);
+		return rc;
+	}
+
+	mem_block = lmb_to_memblock(lmb);
+	if (!mem_block) {
+		remove_memory(nid, lmb->base_addr, block_sz);
+		dlpar_release_drc(lmb->drc_index);
+		return -EINVAL;
+	}
+
+	rc = device_online(&mem_block->dev);
+	put_device(&mem_block->dev);
+	if (rc) {
+		remove_memory(nid, lmb->base_addr, block_sz);
+		dlpar_release_drc(lmb->drc_index);
+		return rc;
+	}
+
+	lmb->flags |= DRCONF_MEM_ASSIGNED;
+	return 0;
+}
+
+static int dlpar_memory_add_by_count(u32 lmbs_to_add, struct property *prop)
+{
+	struct of_drconf_cell *lmbs;
+	u32 num_lmbs, *p;
+	int lmbs_available = 0;
+	int lmbs_added = 0;
+	int i, rc;
+
+	pr_info("Attempting to hot-add %d LMB(s)\n", lmbs_to_add);
+
+	if (lmbs_to_add == 0)
+		return -EINVAL;
+
+	p = prop->value;
+	num_lmbs = *p++;
+	lmbs = (struct of_drconf_cell *)p;
+
+	/* Validate that there are enough LMBs to satisfy the request */
+	for (i = 0; i < num_lmbs; i++) {
+		if (!(lmbs[i].flags & DRCONF_MEM_ASSIGNED))
+			lmbs_available++;
+	}
+
+	if (lmbs_available < lmbs_to_add)
+		return -EINVAL;
+
+	for (i = 0; i < num_lmbs && lmbs_to_add != lmbs_added; i++) {
+		rc = dlpar_add_lmb(&lmbs[i]);
+		if (rc)
+			continue;
+
+		lmbs_added++;
+
+		/* Mark this lmb so we can remove it later if all of the
+		 * requested LMBs cannot be added.
+		 */
+		lmbs[i].reserved = 1;
+	}
+
+	if (lmbs_added != lmbs_to_add) {
+		/* TODO: remove added lmbs */
+		rc = -EINVAL;
+	} else {
+		for (i = 0; i < num_lmbs; i++) {
+			if (!lmbs[i].reserved)
+				continue;
+
+			pr_info("Memory at %llx (drc index %x) was hot-added\n",
+				lmbs[i].base_addr, lmbs[i].drc_index);
+			lmbs[i].reserved = 0;
+		}
+	}
+
+	return rc;
+}
+
+static int dlpar_memory_add_by_index(u32 drc_index, struct property *prop)
+{
+	struct of_drconf_cell *lmbs;
+	u32 num_lmbs, *p;
+	int i, lmb_found;
+	int rc;
+
+	pr_info("Attempting to hot-add LMB, drc index %x\n", drc_index);
+
+	p = prop->value;
+	num_lmbs = *p++;
+	lmbs = (struct of_drconf_cell *)p;
+
+	lmb_found = 0;
+	for (i = 0; i < num_lmbs; i++) {
+		if (lmbs[i].drc_index == drc_index) {
+			lmb_found = 1;
+			rc = dlpar_add_lmb(&lmbs[i]);
+			break;
+		}
+	}
+
+	if (!lmb_found)
+		rc = -EINVAL;
+
+	if (rc)
+		pr_info("Failed to hot-add memory, drc index %x\n", drc_index);
+	else
+		pr_info("Memory at %llx (drc index %x) was hot-added\n",
+			lmbs[i].base_addr, drc_index);
+
+	return rc;
+}
+
+static void dlpar_update_drconf_property(struct device_node *dn,
+					 struct property *prop)
+{
+	struct of_drconf_cell *lmbs;
+	u32 num_lmbs, *p;
+	int i;
+
+	/* Convert the property back to BE */
+	p = prop->value;
+	num_lmbs = *p;
+	*p = cpu_to_be32(*p);
+	p++;
+
+	lmbs = (struct of_drconf_cell *)p;
+	for (i = 0; i < num_lmbs; i++) {
+		lmbs[i].base_addr = cpu_to_be64(lmbs[i].base_addr);
+		lmbs[i].drc_index = cpu_to_be32(lmbs[i].drc_index);
+		lmbs[i].flags = cpu_to_be32(lmbs[i].flags);
+	}
+
+	rtas_hp_event = true;
+	of_update_property(dn, prop);
+	rtas_hp_event = false;
+}
+
 int dlpar_memory(struct pseries_hp_errorlog *hp_elog)
 {
-	int rc = 0;
+	struct device_node *dn;
+	struct property *prop;
+	u32 count, drc_index;
+	int rc;
+
+	count = hp_elog->_drc_u.drc_count;
+	drc_index = hp_elog->_drc_u.drc_index;
 
 	lock_device_hotplug();
 
+	dn = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
+	if (!dn)
+		return -EINVAL;
+
+	prop = dlpar_clone_drconf_property(dn);
+	if (!prop) {
+		of_node_put(dn);
+		return -EINVAL;
+	}
+
 	switch (hp_elog->action) {
+	case PSERIES_HP_ELOG_ACTION_ADD:
+		if (hp_elog->id_type == PSERIES_HP_ELOG_ID_DRC_COUNT)
+			rc = dlpar_memory_add_by_count(count, prop);
+		else if (hp_elog->id_type == PSERIES_HP_ELOG_ID_DRC_INDEX)
+			rc = dlpar_memory_add_by_index(drc_index, prop);
+		else
+			rc = -EINVAL;
+		break;
 	default:
 		pr_err("Invalid action (%d) specified\n", hp_elog->action);
 		rc = -EINVAL;
 		break;
 	}
 
+	if (rc)
+		dlpar_free_drconf_property(prop);
+	else
+		dlpar_update_drconf_property(dn, prop);
+
+	of_node_put(dn);
 	unlock_device_hotplug();
 	return rc;
 }
@@ -193,6 +454,9 @@ static int pseries_update_drconf_memory(struct of_reconfig_data *pr)
 	__be32 *p;
 	int i, rc = -EINVAL;
 
+	if (rtas_hp_event)
+		return 0;
+
 	memblock_size = pseries_memory_block_size();
 	if (!memblock_size)
 		return -EINVAL;

[PATCH v3 5/5] pseries: Implement memory hotplug remove in the kernel

[Nathan Fontenot]

This patch adds the ability to do memory hotplug remove in the kernel.

Currently the operation to hotplug remove memory is handled by the drmgr
command which performs the operation by performing some work in user-space
and making requests to the kernel to handle other pieces. By moving all
of the work to the kernel we can do the remove faster, and provide a common
code path to do memory hotplug for both the PowerVM and PowerKVM environments.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
---
Changes from previous cversion:
Changes from previous version:
- converted uint32_t to u32
- Updated the messaging to properly report the success/failure of memory adds

 arch/powerpc/platforms/pseries/hotplug-memory.c |  192 +++++++++++++++++++++++
 1 file changed, 191 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/platforms/pseries/hotplug-memory.c b/arch/powerpc/platforms/pseries/hotplug-memory.c
index f5eec0f..742ef88 100644
--- a/arch/powerpc/platforms/pseries/hotplug-memory.c
+++ b/arch/powerpc/platforms/pseries/hotplug-memory.c
@@ -188,6 +188,173 @@ static int pseries_remove_mem_node(struct device_node *np)
 	pseries_remove_memblock(base, lmb_size);
 	return 0;
 }
+
+static bool lmb_is_removable(struct of_drconf_cell *lmb)
+{
+	int i, scns_per_block;
+	int rc = 1;
+	unsigned long pfn, block_sz;
+	u64 phys_addr;
+
+	if (!(lmb->flags & DRCONF_MEM_ASSIGNED))
+		return false;
+
+	block_sz = memory_block_size_bytes();
+	scns_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;
+	phys_addr = lmb->base_addr;
+
+	for (i = 0; i < scns_per_block; i++) {
+		pfn = PFN_DOWN(phys_addr);
+		if (!pfn_present(pfn))
+			continue;
+
+		rc &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
+		phys_addr += MIN_MEMORY_BLOCK_SIZE;
+	}
+
+	return rc ? true : false;
+}
+
+static int dlpar_add_lmb(struct of_drconf_cell *);
+
+static int dlpar_remove_lmb(struct of_drconf_cell *lmb)
+{
+	struct memory_block *mem_block;
+	unsigned long block_sz;
+	int nid, rc;
+
+	if (!lmb_is_removable(lmb))
+		return -EINVAL;
+
+	mem_block = lmb_to_memblock(lmb);
+	if (!mem_block)
+		return -EINVAL;
+
+	rc = device_offline(&mem_block->dev);
+	put_device(&mem_block->dev);
+	if (rc)
+		return rc;
+
+	block_sz = pseries_memory_block_size();
+	nid = memory_add_physaddr_to_nid(lmb->base_addr);
+
+	remove_memory(nid, lmb->base_addr, block_sz);
+
+	/* Update memory regions for memory remove */
+	memblock_remove(lmb->base_addr, block_sz);
+
+	dlpar_release_drc(lmb->drc_index);
+
+	lmb->flags &= ~DRCONF_MEM_ASSIGNED;
+	return 0;
+}
+
+static int dlpar_memory_remove_by_count(u32 lmbs_to_remove,
+					struct property *prop)
+{
+	struct of_drconf_cell *lmbs;
+	int lmbs_removed = 0;
+	int lmbs_available = 0;
+	u32 num_lmbs, *p;
+	int i, rc;
+
+	pr_info("Attempting to hot-remove %d LMB(s)\n", lmbs_to_remove);
+
+	if (lmbs_to_remove == 0)
+		return -EINVAL;
+
+	p = prop->value;
+	num_lmbs = *p++;
+	lmbs = (struct of_drconf_cell *)p;
+
+	/* Validate that there are enough LMBs to satisfy the request */
+	for (i = 0; i < num_lmbs; i++) {
+		if (lmbs[i].flags & DRCONF_MEM_ASSIGNED)
+			lmbs_available++;
+	}
+
+	if (lmbs_available < lmbs_to_remove)
+		return -EINVAL;
+
+	for (i = 0; i < num_lmbs && lmbs_removed < lmbs_to_remove; i++) {
+		rc = dlpar_remove_lmb(&lmbs[i]);
+		if (rc)
+			continue;
+
+		lmbs_removed++;
+
+		/* Mark this lmb so we can add it later if all of the
+		 * requested LMBs cannot be removed.
+		 */
+		lmbs[i].reserved = 1;
+	}
+
+	if (lmbs_removed != lmbs_to_remove) {
+		pr_err("Memory hot-remove failed, adding LMB's back\n");
+
+		for (i = 0; i < num_lmbs; i++) {
+			if (!lmbs[i].reserved)
+				continue;
+
+			rc = dlpar_add_lmb(&lmbs[i]);
+			if (rc)
+				pr_err("Failed to add LMB back, drc index %x\n",
+				       lmbs[i].drc_index);
+
+			lmbs[i].reserved = 0;
+		}
+
+		rc = -EINVAL;
+	} else {
+		for (i = 0; i < num_lmbs; i++) {
+			if (!lmbs[i].reserved)
+				continue;
+
+			pr_info("Memory at %llx was hot-removed\n",
+				lmbs[i].base_addr);
+
+			lmbs[i].reserved = 0;
+		}
+		rc = 0;
+	}
+
+	return rc;
+}
+
+static int dlpar_memory_remove_by_index(u32 drc_index, struct property *prop)
+{
+	struct of_drconf_cell *lmbs;
+	u32 num_lmbs, *p;
+	int lmb_found;
+	int i, rc;
+
+	pr_info("Attempting to hot-remove LMB, drc index %x\n", drc_index);
+
+	p = prop->value;
+	num_lmbs = *p++;
+	lmbs = (struct of_drconf_cell *)p;
+
+	lmb_found = 0;
+	for (i = 0; i < num_lmbs; i++) {
+		if (lmbs[i].drc_index == drc_index) {
+			lmb_found = 1;
+			rc = dlpar_remove_lmb(&lmbs[i]);
+			break;
+		}
+	}
+
+	if (!lmb_found)
+		rc = -EINVAL;
+
+	if (rc)
+		pr_info("Failed to hot-remove memory at %llx\n",
+			lmbs[i].base_addr);
+	else
+		pr_info("Memory at %llx was hot-removed\n", lmbs[i].base_addr);
+
+	return rc;
+}
+
 #else
 static inline int pseries_remove_memblock(unsigned long base,
 					  unsigned int memblock_size)
@@ -198,6 +365,11 @@ static inline int pseries_remove_mem_node(struct device_node *np)
 {
 	return 0;
 }
+static inline int dlpar_memory_remove(struct pseries_hp_errorlog *hp_elog)
+{
+	return -EOPNOTSUPP;
+}
+
 #endif /* CONFIG_MEMORY_HOTREMOVE */
 
 static int dlpar_add_lmb(struct of_drconf_cell *lmb)
@@ -292,7 +464,17 @@ static int dlpar_memory_add_by_count(u32 lmbs_to_add, struct property *prop)
 	}
 
 	if (lmbs_added != lmbs_to_add) {
-		/* TODO: remove added lmbs */
+		pr_err("Memory hot-add failed, removing any added LMBs\n");
+
+		for (i = 0; i < num_lmbs; i++) {
+			if (!lmbs[i].reserved)
+				continue;
+
+			rc = dlpar_remove_lmb(&lmbs[i]);
+			if (rc)
+				pr_err("Failed to remove LMB, drc index %x\n",
+				       be32_to_cpu(lmbs[i].drc_index));
+		}
 		rc = -EINVAL;
 	} else {
 		for (i = 0; i < num_lmbs; i++) {
@@ -398,6 +580,14 @@ int dlpar_memory(struct pseries_hp_errorlog *hp_elog)
 		else
 			rc = -EINVAL;
 		break;
+	case PSERIES_HP_ELOG_ACTION_REMOVE:
+		if (hp_elog->id_type == PSERIES_HP_ELOG_ID_DRC_COUNT)
+			rc = dlpar_memory_remove_by_count(count, prop);
+		else if (hp_elog->id_type == PSERIES_HP_ELOG_ID_DRC_INDEX)
+			rc = dlpar_memory_remove_by_index(drc_index, prop);
+		else
+			rc = -EINVAL;
+		break;
 	default:
 		pr_err("Invalid action (%d) specified\n", hp_elog->action);
 		rc = -EINVAL;