[RFC XEN PATCH v4 06/41] acpi: probe valid PMEM regions via NFIT

[Haozhong Zhang <haozhong.zhang@intel.com>]

A PMEM region with failures (e.g., not properly flushed in the last
power cycle, or some blocks within it are borken) cannot be safely
used by Xen and guest. Scan the state flags of NVDIMM region mapping
structures in NFIT to check whether any failures happened to a PMEM
region. The recovery of those failure are left out of Xen (e.g. left
to the firmware or other management utilities on the bare metal).

Signed-off-by: Haozhong Zhang <haozhong.zhang@intel.com>
---
Cc: Jan Beulich <jbeulich@suse.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>
Cc: George Dunlap <George.Dunlap@eu.citrix.com>
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Tim Deegan <tim@xen.org>
Cc: Wei Liu <wei.liu2@citrix.com>

Changes in v4:
 * Scan memory mapping tables from SPA tables in acpi_nfit_register_pmem(),
   rather than in the reverse order.
---
 xen/arch/x86/acpi/boot.c  |   4 ++
 xen/drivers/acpi/nfit.c   | 176 +++++++++++++++++++++++++++++++++++++++++++++-
 xen/include/acpi/actbl1.h |  26 +++++++
 xen/include/xen/acpi.h    |   1 +
 4 files changed, 206 insertions(+), 1 deletion(-)

diff --git a/xen/arch/x86/acpi/boot.c b/xen/arch/x86/acpi/boot.c
index 8e6c96dcf6..f52a2c6dc5 100644
--- a/xen/arch/x86/acpi/boot.c
+++ b/xen/arch/x86/acpi/boot.c
@@ -732,5 +732,9 @@ int __init acpi_boot_init(void)
 
 	acpi_table_parse(ACPI_SIG_BGRT, acpi_invalidate_bgrt);
 
+#ifdef CONFIG_NVDIMM_PMEM
+	acpi_nfit_init();
+#endif
+
 	return 0;
 }
diff --git a/xen/drivers/acpi/nfit.c b/xen/drivers/acpi/nfit.c
index e099378ee0..0a44983aad 100644
--- a/xen/drivers/acpi/nfit.c
+++ b/xen/drivers/acpi/nfit.c
@@ -31,11 +31,166 @@ static const uint8_t nfit_spa_pmem_guid[] =
     0xac, 0x43, 0x0d, 0x33, 0x18, 0xb7, 0x8c, 0xdb,
 };
 
+struct nfit_spa_desc {
+    struct list_head link;
+    struct acpi_nfit_system_address *acpi_table;
+    struct list_head memdev_list;
+};
+
+struct nfit_memdev_desc {
+    struct list_head link;
+    struct acpi_nfit_memory_map *acpi_table;
+    struct list_head memdev_link;
+};
+
 struct acpi_nfit_desc {
     struct acpi_table_nfit *acpi_table;
+    struct list_head spa_list;
+    struct list_head memdev_list;
 };
 
-static struct acpi_nfit_desc nfit_desc;
+static struct acpi_nfit_desc nfit_desc = {
+    .spa_list = LIST_HEAD_INIT(nfit_desc.spa_list),
+    .memdev_list = LIST_HEAD_INIT(nfit_desc.memdev_list),
+};
+
+static void __init acpi_nfit_del_subtables(struct acpi_nfit_desc *desc)
+{
+    struct nfit_spa_desc *spa, *spa_next;
+    struct nfit_memdev_desc *memdev, *memdev_next;
+
+    list_for_each_entry_safe(spa, spa_next, &desc->spa_list, link)
+    {
+        list_del(&spa->link);
+        xfree(spa);
+    }
+    list_for_each_entry_safe (memdev, memdev_next, &desc->memdev_list, link)
+    {
+        list_del(&memdev->link);
+        xfree(memdev);
+    }
+}
+
+static int __init acpi_nfit_add_subtables(struct acpi_nfit_desc *desc)
+{
+    struct acpi_table_nfit *nfit_table = desc->acpi_table;
+    uint32_t hdr_offset = sizeof(*nfit_table);
+    uint32_t nfit_length = nfit_table->header.length;
+    struct acpi_nfit_header *hdr;
+    struct nfit_spa_desc *spa_desc;
+    struct nfit_memdev_desc *memdev_desc;
+    int ret = 0;
+
+#define INIT_DESC(desc, acpi_hdr, acpi_type, desc_list) \
+    do {                                                \
+        (desc) = xzalloc(typeof(*(desc)));              \
+        if ( unlikely(!(desc)) ) {                      \
+            ret = -ENOMEM;                              \
+            goto nomem;                                 \
+        }                                               \
+        (desc)->acpi_table = (acpi_type *)(acpi_hdr);   \
+        INIT_LIST_HEAD(&(desc)->link);                  \
+        list_add_tail(&(desc)->link, (desc_list));      \
+    } while ( 0 )
+
+    while ( hdr_offset < nfit_length )
+    {
+        hdr = (void *)nfit_table + hdr_offset;
+        hdr_offset += hdr->length;
+
+        switch ( hdr->type )
+        {
+        case ACPI_NFIT_TYPE_SYSTEM_ADDRESS:
+            INIT_DESC(spa_desc, hdr, struct acpi_nfit_system_address,
+                      &desc->spa_list);
+            break;
+
+        case ACPI_NFIT_TYPE_MEMORY_MAP:
+            INIT_DESC(memdev_desc, hdr, struct acpi_nfit_memory_map,
+                      &desc->memdev_list);
+            break;
+
+        default:
+            continue;
+        }
+    }
+
+#undef INIT_DESC
+
+    return 0;
+
+ nomem:
+    acpi_nfit_del_subtables(desc);
+
+    return ret;
+}
+
+static void __init acpi_nfit_link_subtables(struct acpi_nfit_desc *desc)
+{
+    struct nfit_spa_desc *spa_desc;
+    struct nfit_memdev_desc *memdev_desc;
+    uint16_t spa_idx;
+
+    list_for_each_entry(spa_desc, &desc->spa_list, link)
+    {
+        INIT_LIST_HEAD(&spa_desc->memdev_list);
+
+        spa_idx = spa_desc->acpi_table->range_index;
+
+        list_for_each_entry(memdev_desc, &desc->memdev_list, link)
+        {
+            if ( memdev_desc->acpi_table->range_index == spa_idx )
+                list_add_tail(&memdev_desc->memdev_link,
+                              &spa_desc->memdev_list);
+        }
+    }
+}
+
+static void __init acpi_nfit_register_pmem(struct acpi_nfit_desc *desc)
+{
+    struct nfit_spa_desc *spa_desc;
+    struct nfit_memdev_desc *memdev_desc;
+    struct acpi_nfit_system_address *spa;
+    unsigned long smfn, emfn;
+    bool failed;
+
+    list_for_each_entry(spa_desc, &desc->spa_list, link)
+    {
+        spa = spa_desc->acpi_table;
+
+        /* Skip non-pmem entry. */
+        if ( memcmp(spa->range_guid, nfit_spa_pmem_guid, 16) )
+            continue;
+
+        smfn = paddr_to_pfn(spa->address);
+        emfn = paddr_to_pfn(spa->address + spa->length);
+        failed = false;
+
+        list_for_each_entry(memdev_desc, &spa_desc->memdev_list, memdev_link)
+        {
+            if ( memdev_desc->acpi_table->flags &
+                 (ACPI_NFIT_MEM_SAVE_FAILED |
+                  ACPI_NFIT_MEM_RESTORE_FAILED |
+                  ACPI_NFIT_MEM_FLUSH_FAILED |
+                  ACPI_NFIT_MEM_NOT_ARMED |
+                  ACPI_NFIT_MEM_MAP_FAILED) )
+            {
+                failed = true;
+                break;
+            }
+        }
+
+        if ( failed )
+        {
+            printk(XENLOG_INFO
+                   "NFIT: detected failures on PMEM MFNs 0x%lx - 0x%lx, skipped\n",
+                   smfn, emfn);
+            continue;
+        }
+
+        printk(XENLOG_INFO "NFIT: PMEM MFNs 0x%lx - 0x%lx\n", smfn, emfn);
+    }
+}
 
 void __init acpi_nfit_boot_init(void)
 {
@@ -53,6 +208,25 @@ void __init acpi_nfit_boot_init(void)
                      PAGE_HYPERVISOR);
 }
 
+void __init acpi_nfit_init(void)
+{
+    if ( !nfit_desc.acpi_table )
+        return;
+
+    /* Collect all SPA and memory map sub-tables. */
+    if ( acpi_nfit_add_subtables(&nfit_desc) )
+    {
+        printk(XENLOG_ERR "NFIT: no memory for NFIT management\n");
+        return;
+    }
+
+    /* Link descriptors of SPA and memory map sub-tables. */
+    acpi_nfit_link_subtables(&nfit_desc);
+
+    /* Register valid pmem regions to Xen hypervisor. */
+    acpi_nfit_register_pmem(&nfit_desc);
+}
+
 /**
  * Search pmem regions overlapped with the specified address range.
  *
diff --git a/xen/include/acpi/actbl1.h b/xen/include/acpi/actbl1.h
index 94d8d7775c..037652916a 100644
--- a/xen/include/acpi/actbl1.h
+++ b/xen/include/acpi/actbl1.h
@@ -946,6 +946,32 @@ struct acpi_nfit_system_address {
 	u64 memory_mapping;
 };
 
+/* 1: Memory Device to System Address Range Map Structure */
+struct acpi_nfit_memory_map {
+	struct acpi_nfit_header header;
+	u32 device_handle;
+	u16 physical_id;
+	u16 region_id;
+	u16 range_index;
+	u16 region_index;
+	u64 region_size;
+	u64 region_offset;
+	u64 address;
+	u16 interleave_index;
+	u16 interleave_ways;
+	u16 flags;
+	u16 reserved;		/* Reserved, must be zero */
+};
+
+/* Flags in struct acpi_nfit_memory_map */
+#define ACPI_NFIT_MEM_SAVE_FAILED		(1)	/* 00: Last SAVE to Memory Device failed */
+#define ACPI_NFIT_MEM_RESTORE_FAILED	(1<<1)	/* 01: Last RESTORE from Memory Device failed */
+#define ACPI_NFIT_MEM_FLUSH_FAILED		(1<<2)	/* 02: Platform flush failed */
+#define ACPI_NFIT_MEM_NOT_ARMED			(1<<3)	/* 03: Memory Device is not armed */
+#define ACPI_NFIT_MEM_HEALTH_OBSERVED	(1<<4)	/* 04: Memory Device observed SMART/health events */
+#define ACPI_NFIT_MEM_HEALTH_ENABLED	(1<<5)	/* 05: SMART/health events enabled */
+#define ACPI_NFIT_MEM_MAP_FAILED		(1<<6)	/* 06: Mapping to SPA failed */
+
 /*******************************************************************************
  *
  * SBST - Smart Battery Specification Table
diff --git a/xen/include/xen/acpi.h b/xen/include/xen/acpi.h
index 1bd8f9f4e4..088f01255d 100644
--- a/xen/include/xen/acpi.h
+++ b/xen/include/xen/acpi.h
@@ -185,6 +185,7 @@ void acpi_nfit_boot_init(void);
 bool acpi_nfit_boot_search_pmem(unsigned long smfn, unsigned long emfn,
                                 unsigned long *ret_smfn,
                                 unsigned long *ret_emfn);
+void acpi_nfit_init(void);
 #endif /* CONFIG_NVDIMM_PMEM */
 
 #endif /*_LINUX_ACPI_H*/
-- 
2.15.1


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel