xen: features and fixes for 4.3-rc0

- Convert xen-blkfront to the multiqueue API
 - [arm] Support binding event channels to different VCPUs.
 - [x86] Support > 512 GiB in a PV guests (off by default as such a
   guest cannot be migrated with the current toolstack).
 - [x86] PMU support for PV dom0 (limited support for using perf with
   Xen and other guests).
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1
 
 iQEcBAABAgAGBQJV7wIdAAoJEFxbo/MsZsTR0hEH/04HTKLKGnSJpZ5WbMPxqZxE
 UqGlvhvVWNAmFocZmbPcEi9T1qtcFrX5pM55JQr6UmAp3ovYsT2q1Q1kKaOaawks
 pSfc/YEH3oQW5VUQ9Lm9Ru5Z8Btox0WrzRREO92OF36UOgUOBOLkGsUfOwDinNIM
 lSk2djbYwDYAsoeC3PHB32wwMI//Lz6B/9ZVXcyL6ULynt1ULdspETjGnptRPZa7
 JTB5L4/soioKOn18HDwwOhKmvaFUPQv9Odnv7dc85XwZreajhM/KMu3qFbMDaF/d
 WVB1NMeCBdQYgjOrUjrmpyr5uTMySiQEG54cplrEKinfeZgKlEyjKvjcAfJfiac=
 =Ktjl
 -----END PGP SIGNATURE-----

Merge tag 'for-linus-4.3-rc0-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

Pull xen updates from David Vrabel:
 "Xen features and fixes for 4.3:

   - Convert xen-blkfront to the multiqueue API
   - [arm] Support binding event channels to different VCPUs.
   - [x86] Support > 512 GiB in a PV guests (off by default as such a
     guest cannot be migrated with the current toolstack).
   - [x86] PMU support for PV dom0 (limited support for using perf with
     Xen and other guests)"

* tag 'for-linus-4.3-rc0-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip: (33 commits)
  xen: switch extra memory accounting to use pfns
  xen: limit memory to architectural maximum
  xen: avoid another early crash of memory limited dom0
  xen: avoid early crash of memory limited dom0
  arm/xen: Remove helpers which are PV specific
  xen/x86: Don't try to set PCE bit in CR4
  xen/PMU: PMU emulation code
  xen/PMU: Intercept PMU-related MSR and APIC accesses
  xen/PMU: Describe vendor-specific PMU registers
  xen/PMU: Initialization code for Xen PMU
  xen/PMU: Sysfs interface for setting Xen PMU mode
  xen: xensyms support
  xen: remove no longer needed p2m.h
  xen: allow more than 512 GB of RAM for 64 bit pv-domains
  xen: move p2m list if conflicting with e820 map
  xen: add explicit memblock_reserve() calls for special pages
  mm: provide early_memremap_ro to establish read-only mapping
  xen: check for initrd conflicting with e820 map
  xen: check pre-allocated page tables for conflict with memory map
  xen: check for kernel memory conflicting with memory layout
  ...

drivers/xen/Kconfig

@@ -280,4 +280,15 @@ config XEN_ACPI
 	def_bool y
 	depends on X86 && ACPI
 
+config XEN_SYMS
+       bool "Xen symbols"
+       depends on X86 && XEN_DOM0 && XENFS
+       default y if KALLSYMS
+       help
+          Exports hypervisor symbols (along with their types and addresses) via
+          /proc/xen/xensyms file, similar to /proc/kallsyms
+
+config XEN_HAVE_VPMU
+       bool
+
 endmenu

drivers/xen/balloon.c

@@ -638,9 +638,9 @@ static int __init balloon_init(void)
 	 * regions (see arch/x86/xen/setup.c).
 	 */
 	for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++)
-		if (xen_extra_mem[i].size)
-			balloon_add_region(PFN_UP(xen_extra_mem[i].start),
-					   PFN_DOWN(xen_extra_mem[i].size));
+		if (xen_extra_mem[i].n_pfns)
+			balloon_add_region(xen_extra_mem[i].start_pfn,
+					   xen_extra_mem[i].n_pfns);
 
 	return 0;
 }

drivers/xen/events/events_base.c

@@ -1301,11 +1301,7 @@ static int rebind_irq_to_cpu(unsigned irq, unsigned tcpu)
 	if (!VALID_EVTCHN(evtchn))
 		return -1;
 
-	/*
-	 * Events delivered via platform PCI interrupts are always
-	 * routed to vcpu 0 and hence cannot be rebound.
-	 */
-	if (xen_hvm_domain() && !xen_have_vector_callback)
+	if (!xen_support_evtchn_rebind())
 		return -1;
 
 	/* Send future instances of this interrupt to other vcpu. */

drivers/xen/sys-hypervisor.c

@@ -20,6 +20,9 @@
 #include <xen/xenbus.h>
 #include <xen/interface/xen.h>
 #include <xen/interface/version.h>
+#ifdef CONFIG_XEN_HAVE_VPMU
+#include <xen/interface/xenpmu.h>
+#endif
 
 #define HYPERVISOR_ATTR_RO(_name) \
 static struct hyp_sysfs_attr  _name##_attr = __ATTR_RO(_name)
@@ -368,6 +371,126 @@ static void xen_properties_destroy(void)
 	sysfs_remove_group(hypervisor_kobj, &xen_properties_group);
 }
 
+#ifdef CONFIG_XEN_HAVE_VPMU
+struct pmu_mode {
+	const char *name;
+	uint32_t mode;
+};
+
+static struct pmu_mode pmu_modes[] = {
+	{"off", XENPMU_MODE_OFF},
+	{"self", XENPMU_MODE_SELF},
+	{"hv", XENPMU_MODE_HV},
+	{"all", XENPMU_MODE_ALL}
+};
+
+static ssize_t pmu_mode_store(struct hyp_sysfs_attr *attr,
+			      const char *buffer, size_t len)
+{
+	int ret;
+	struct xen_pmu_params xp;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(pmu_modes); i++) {
+		if (strncmp(buffer, pmu_modes[i].name, len - 1) == 0) {
+			xp.val = pmu_modes[i].mode;
+			break;
+		}
+	}
+
+	if (i == ARRAY_SIZE(pmu_modes))
+		return -EINVAL;
+
+	xp.version.maj = XENPMU_VER_MAJ;
+	xp.version.min = XENPMU_VER_MIN;
+	ret = HYPERVISOR_xenpmu_op(XENPMU_mode_set, &xp);
+	if (ret)
+		return ret;
+
+	return len;
+}
+
+static ssize_t pmu_mode_show(struct hyp_sysfs_attr *attr, char *buffer)
+{
+	int ret;
+	struct xen_pmu_params xp;
+	int i;
+	uint32_t mode;
+
+	xp.version.maj = XENPMU_VER_MAJ;
+	xp.version.min = XENPMU_VER_MIN;
+	ret = HYPERVISOR_xenpmu_op(XENPMU_mode_get, &xp);
+	if (ret)
+		return ret;
+
+	mode = (uint32_t)xp.val;
+	for (i = 0; i < ARRAY_SIZE(pmu_modes); i++) {
+		if (mode == pmu_modes[i].mode)
+			return sprintf(buffer, "%s\n", pmu_modes[i].name);
+	}
+
+	return -EINVAL;
+}
+HYPERVISOR_ATTR_RW(pmu_mode);
+
+static ssize_t pmu_features_store(struct hyp_sysfs_attr *attr,
+				  const char *buffer, size_t len)
+{
+	int ret;
+	uint32_t features;
+	struct xen_pmu_params xp;
+
+	ret = kstrtou32(buffer, 0, &features);
+	if (ret)
+		return ret;
+
+	xp.val = features;
+	xp.version.maj = XENPMU_VER_MAJ;
+	xp.version.min = XENPMU_VER_MIN;
+	ret = HYPERVISOR_xenpmu_op(XENPMU_feature_set, &xp);
+	if (ret)
+		return ret;
+
+	return len;
+}
+
+static ssize_t pmu_features_show(struct hyp_sysfs_attr *attr, char *buffer)
+{
+	int ret;
+	struct xen_pmu_params xp;
+
+	xp.version.maj = XENPMU_VER_MAJ;
+	xp.version.min = XENPMU_VER_MIN;
+	ret = HYPERVISOR_xenpmu_op(XENPMU_feature_get, &xp);
+	if (ret)
+		return ret;
+
+	return sprintf(buffer, "0x%x\n", (uint32_t)xp.val);
+}
+HYPERVISOR_ATTR_RW(pmu_features);
+
+static struct attribute *xen_pmu_attrs[] = {
+	&pmu_mode_attr.attr,
+	&pmu_features_attr.attr,
+	NULL
+};
+
+static const struct attribute_group xen_pmu_group = {
+	.name = "pmu",
+	.attrs = xen_pmu_attrs,
+};
+
+static int __init xen_pmu_init(void)
+{
+	return sysfs_create_group(hypervisor_kobj, &xen_pmu_group);
+}
+
+static void xen_pmu_destroy(void)
+{
+	sysfs_remove_group(hypervisor_kobj, &xen_pmu_group);
+}
+#endif
+
 static int __init hyper_sysfs_init(void)
 {
 	int ret;
@@ -390,7 +513,15 @@ static int __init hyper_sysfs_init(void)
 	ret = xen_properties_init();
 	if (ret)
 		goto prop_out;
-
+#ifdef CONFIG_XEN_HAVE_VPMU
+	if (xen_initial_domain()) {
+		ret = xen_pmu_init();
+		if (ret) {
+			xen_properties_destroy();
+			goto prop_out;
+		}
+	}
+#endif
 	goto out;
 
 prop_out:
@@ -407,6 +538,9 @@ out:
 
 static void __exit hyper_sysfs_exit(void)
 {
+#ifdef CONFIG_XEN_HAVE_VPMU
+	xen_pmu_destroy();
+#endif
 	xen_properties_destroy();
 	xen_compilation_destroy();
 	xen_sysfs_uuid_destroy();

drivers/xen/xenfs/Makefile

@@ -2,3 +2,4 @@ obj-$(CONFIG_XENFS) += xenfs.o
 
 xenfs-y			  = super.o
 xenfs-$(CONFIG_XEN_DOM0) += xenstored.o
+xenfs-$(CONFIG_XEN_SYMS) += xensyms.o

drivers/xen/xenfs/super.c

@@ -57,6 +57,9 @@ static int xenfs_fill_super(struct super_block *sb, void *data, int silent)
 		{ "privcmd", &xen_privcmd_fops, S_IRUSR|S_IWUSR },
 		{ "xsd_kva", &xsd_kva_file_ops, S_IRUSR|S_IWUSR},
 		{ "xsd_port", &xsd_port_file_ops, S_IRUSR|S_IWUSR},
+#ifdef CONFIG_XEN_SYMS
+		{ "xensyms", &xensyms_ops, S_IRUSR},
+#endif
 		{""},
 	};
 

drivers/xen/xenfs/xenfs.h

@@ -3,5 +3,6 @@
 
 extern const struct file_operations xsd_kva_file_ops;
 extern const struct file_operations xsd_port_file_ops;
+extern const struct file_operations xensyms_ops;
 
 #endif	/* _XENFS_XENBUS_H */

drivers/xen/xenfs/xensyms.c

@@ -0,0 +1,152 @@
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/seq_file.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/proc_fs.h>
+#include <linux/slab.h>
+#include <xen/interface/platform.h>
+#include <asm/xen/hypercall.h>
+#include <xen/xen-ops.h>
+#include "xenfs.h"
+
+
+#define XEN_KSYM_NAME_LEN 127 /* Hypervisor may have different name length */
+
+struct xensyms {
+	struct xen_platform_op op;
+	char *name;
+	uint32_t namelen;
+};
+
+/* Grab next output page from the hypervisor */
+static int xensyms_next_sym(struct xensyms *xs)
+{
+	int ret;
+	struct xenpf_symdata *symdata = &xs->op.u.symdata;
+	uint64_t symnum;
+
+	memset(xs->name, 0, xs->namelen);
+	symdata->namelen = xs->namelen;
+
+	symnum = symdata->symnum;
+
+	ret = HYPERVISOR_dom0_op(&xs->op);
+	if (ret < 0)
+		return ret;
+
+	/*
+	 * If hypervisor's symbol didn't fit into the buffer then allocate
+	 * a larger buffer and try again.
+	 */
+	if (unlikely(symdata->namelen > xs->namelen)) {
+		kfree(xs->name);
+
+		xs->namelen = symdata->namelen;
+		xs->name = kzalloc(xs->namelen, GFP_KERNEL);
+		if (!xs->name)
+			return -ENOMEM;
+
+		set_xen_guest_handle(symdata->name, xs->name);
+		symdata->symnum--; /* Rewind */
+
+		ret = HYPERVISOR_dom0_op(&xs->op);
+		if (ret < 0)
+			return ret;
+	}
+
+	if (symdata->symnum == symnum)
+		/* End of symbols */
+		return 1;
+
+	return 0;
+}
+
+static void *xensyms_start(struct seq_file *m, loff_t *pos)
+{
+	struct xensyms *xs = (struct xensyms *)m->private;
+
+	xs->op.u.symdata.symnum = *pos;
+
+	if (xensyms_next_sym(xs))
+		return NULL;
+
+	return m->private;
+}
+
+static void *xensyms_next(struct seq_file *m, void *p, loff_t *pos)
+{
+	struct xensyms *xs = (struct xensyms *)m->private;
+
+	xs->op.u.symdata.symnum = ++(*pos);
+
+	if (xensyms_next_sym(xs))
+		return NULL;
+
+	return p;
+}
+
+static int xensyms_show(struct seq_file *m, void *p)
+{
+	struct xensyms *xs = (struct xensyms *)m->private;
+	struct xenpf_symdata *symdata = &xs->op.u.symdata;
+
+	seq_printf(m, "%016llx %c %s\n", symdata->address,
+		   symdata->type, xs->name);
+
+	return 0;
+}
+
+static void xensyms_stop(struct seq_file *m, void *p)
+{
+}
+
+static const struct seq_operations xensyms_seq_ops = {
+	.start = xensyms_start,
+	.next = xensyms_next,
+	.show = xensyms_show,
+	.stop = xensyms_stop,
+};
+
+static int xensyms_open(struct inode *inode, struct file *file)
+{
+	struct seq_file *m;
+	struct xensyms *xs;
+	int ret;
+
+	ret = seq_open_private(file, &xensyms_seq_ops,
+			       sizeof(struct xensyms));
+	if (ret)
+		return ret;
+
+	m = file->private_data;
+	xs = (struct xensyms *)m->private;
+
+	xs->namelen = XEN_KSYM_NAME_LEN + 1;
+	xs->name = kzalloc(xs->namelen, GFP_KERNEL);
+	if (!xs->name) {
+		seq_release_private(inode, file);
+		return -ENOMEM;
+	}
+	set_xen_guest_handle(xs->op.u.symdata.name, xs->name);
+	xs->op.cmd = XENPF_get_symbol;
+	xs->op.u.symdata.namelen = xs->namelen;
+
+	return 0;
+}
+
+static int xensyms_release(struct inode *inode, struct file *file)
+{
+	struct seq_file *m = file->private_data;
+	struct xensyms *xs = (struct xensyms *)m->private;
+
+	kfree(xs->name);
+	return seq_release_private(inode, file);
+}
+
+const struct file_operations xensyms_ops = {
+	.open = xensyms_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = xensyms_release
+};