APEI is currently implemented so that it depends on x86 hardware.
The primary dependency is that GHES uses the x86 NMI for hardware
error notification and MCE for memory error handling. These patches
remove that dependency.
Other APEI features such as error reporting via external IRQ, error
serialization, or error injection, do not require changes to use them
on non-x86 architectures.
The following patch set eliminates the APEI Kconfig x86 dependency
by making these changes:
- treat NMI notification as GHES architecture - HAVE_ACPI_APEI_NMI
- group and wrap around #ifdef CONFIG_HAVE_ACPI_APEI_NMI code which
is used only for NMI path
- identify architectural boxes and abstract it accordingly (tlb flush and MCE)
- rework ioremap for both IRQ and NMI context
NMI code is kept in ghes.c file since NMI and IRQ context are tightly coupled.
Note, these patches introduce no functional changes for x86. The NMI notification
feature is hard selected for x86. Architectures that want to use this
feature should also provide NMI code infrastructure.
V1->V2
- address Borislav's comment
- abstract arch-specific calls instead of wrapping into the #ifdef
V2->V3
- address Robert's comment
- disable ACPI_APEI_NMI selection so that it is hard selected by arch Kconfig
- rename ACPI_APEI_NMI to ARCH_HAS_ACPI_APEI_NMI
V3->V4
- do not abstract NMI calls for archs which do not support it
- merge some of patches to make review process easier
V4->V5
- markers for preprocessor statements (#else, #endif) around long code block
Tomasz Nowicki (3):
apei, mce: Factor out APEI architecture specific MCE calls.
acpi, apei, ghes: Make NMI error notification to be GHES architecture
extension.
acpi, apei, ghes: Factor out ioremap virtual memory for IRQ and NMI
context.
arch/x86/Kconfig | 2 +
arch/x86/kernel/acpi/Makefile | 1 +
arch/x86/kernel/acpi/apei.c | 62 +++++++++++++++
drivers/acpi/apei/Kconfig | 8 +-
drivers/acpi/apei/apei-base.c | 13 ++++
drivers/acpi/apei/ghes.c | 173 +++++++++++++++++++++++++++---------------
drivers/acpi/apei/hest.c | 29 +------
include/acpi/apei.h | 4 +
include/linux/nmi.h | 4 +
9 files changed, 204 insertions(+), 92 deletions(-)
create mode 100644 arch/x86/kernel/acpi/apei.c
--
1.9.1
This commit abstracts MCE calls and provides weak corresponding default
implementation for those architectures which do not need arch specific
actions. Each platform willing to do additional architectural actions
should provides desired function definition. It allows us to avoid wrap
code into #ifdef in generic code and prevent new platform from introducing
dummy stub function too.
Initially, there are two APEI arch-specific calls:
- arch_apei_enable_cmcff()
- arch_apei_report_mem_error()
Both interact with MCE driver for X86 architecture.
Acked-by: Borislav Petkov <[email protected]>
Signed-off-by: Tomasz Nowicki <[email protected]>
---
arch/x86/kernel/acpi/Makefile | 1 +
arch/x86/kernel/acpi/apei.c | 56 +++++++++++++++++++++++++++++++++++++++++++
drivers/acpi/apei/apei-base.c | 13 ++++++++++
drivers/acpi/apei/ghes.c | 6 ++---
drivers/acpi/apei/hest.c | 29 +---------------------
include/acpi/apei.h | 3 +++
6 files changed, 76 insertions(+), 32 deletions(-)
create mode 100644 arch/x86/kernel/acpi/apei.c
diff --git a/arch/x86/kernel/acpi/Makefile b/arch/x86/kernel/acpi/Makefile
index 163b225..3242e59 100644
--- a/arch/x86/kernel/acpi/Makefile
+++ b/arch/x86/kernel/acpi/Makefile
@@ -1,5 +1,6 @@
obj-$(CONFIG_ACPI) += boot.o
obj-$(CONFIG_ACPI_SLEEP) += sleep.o wakeup_$(BITS).o
+obj-$(CONFIG_ACPI_APEI) += apei.o
ifneq ($(CONFIG_ACPI_PROCESSOR),)
obj-y += cstate.o
diff --git a/arch/x86/kernel/acpi/apei.c b/arch/x86/kernel/acpi/apei.c
new file mode 100644
index 0000000..12b13de
--- /dev/null
+++ b/arch/x86/kernel/acpi/apei.c
@@ -0,0 +1,56 @@
+/*
+ * Arch-specific APEI-related functions.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#include <acpi/apei.h>
+
+#include <asm/mce.h>
+
+int arch_apei_enable_cmcff(struct acpi_hest_header *hest_hdr, void *data)
+{
+#ifdef CONFIG_X86_MCE
+ int i;
+ struct acpi_hest_ia_corrected *cmc;
+ struct acpi_hest_ia_error_bank *mc_bank;
+
+ if (hest_hdr->type != ACPI_HEST_TYPE_IA32_CORRECTED_CHECK)
+ return 0;
+
+ cmc = (struct acpi_hest_ia_corrected *)hest_hdr;
+ if (!cmc->enabled)
+ return 0;
+
+ /*
+ * We expect HEST to provide a list of MC banks that report errors
+ * in firmware first mode. Otherwise, return non-zero value to
+ * indicate that we are done parsing HEST.
+ */
+ if (!(cmc->flags & ACPI_HEST_FIRMWARE_FIRST) ||
+ !cmc->num_hardware_banks)
+ return 1;
+
+ pr_info("HEST: Enabling Firmware First mode for corrected errors.\n");
+
+ mc_bank = (struct acpi_hest_ia_error_bank *)(cmc + 1);
+ for (i = 0; i < cmc->num_hardware_banks; i++, mc_bank++)
+ mce_disable_bank(mc_bank->bank_number);
+#endif
+ return 1;
+}
+
+void arch_apei_report_mem_error(int sev, struct cper_sec_mem_err *mem_err)
+{
+#ifdef CONFIG_X86_MCE
+ apei_mce_report_mem_error(sev, mem_err);
+#endif
+}
diff --git a/drivers/acpi/apei/apei-base.c b/drivers/acpi/apei/apei-base.c
index 8678dfe..2cd7bdd 100644
--- a/drivers/acpi/apei/apei-base.c
+++ b/drivers/acpi/apei/apei-base.c
@@ -745,6 +745,19 @@ struct dentry *apei_get_debugfs_dir(void)
}
EXPORT_SYMBOL_GPL(apei_get_debugfs_dir);
+int __weak arch_apei_enable_cmcff(struct acpi_hest_header *hest_hdr,
+ void *data)
+{
+ return 1;
+}
+EXPORT_SYMBOL_GPL(arch_apei_enable_cmcff);
+
+void __weak arch_apei_report_mem_error(int sev,
+ struct cper_sec_mem_err *mem_err)
+{
+}
+EXPORT_SYMBOL_GPL(arch_apei_report_mem_error);
+
int apei_osc_setup(void)
{
static u8 whea_uuid_str[] = "ed855e0c-6c90-47bf-a62a-26de0fc5ad5c";
diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
index dab7cb7..352170a 100644
--- a/drivers/acpi/apei/ghes.c
+++ b/drivers/acpi/apei/ghes.c
@@ -49,7 +49,7 @@
#include <linux/aer.h>
#include <acpi/ghes.h>
-#include <asm/mce.h>
+#include <acpi/apei.h>
#include <asm/tlbflush.h>
#include <asm/nmi.h>
@@ -455,9 +455,7 @@ static void ghes_do_proc(struct ghes *ghes,
mem_err = (struct cper_sec_mem_err *)(gdata+1);
ghes_edac_report_mem_error(ghes, sev, mem_err);
-#ifdef CONFIG_X86_MCE
- apei_mce_report_mem_error(sev, mem_err);
-#endif
+ arch_apei_report_mem_error(sev, mem_err);
ghes_handle_memory_failure(gdata, sev);
}
#ifdef CONFIG_ACPI_APEI_PCIEAER
diff --git a/drivers/acpi/apei/hest.c b/drivers/acpi/apei/hest.c
index f5e37f3..06e9b41 100644
--- a/drivers/acpi/apei/hest.c
+++ b/drivers/acpi/apei/hest.c
@@ -36,7 +36,6 @@
#include <linux/io.h>
#include <linux/platform_device.h>
#include <acpi/apei.h>
-#include <asm/mce.h>
#include "apei-internal.h"
@@ -128,33 +127,7 @@ EXPORT_SYMBOL_GPL(apei_hest_parse);
*/
static int __init hest_parse_cmc(struct acpi_hest_header *hest_hdr, void *data)
{
-#ifdef CONFIG_X86_MCE
- int i;
- struct acpi_hest_ia_corrected *cmc;
- struct acpi_hest_ia_error_bank *mc_bank;
-
- if (hest_hdr->type != ACPI_HEST_TYPE_IA32_CORRECTED_CHECK)
- return 0;
-
- cmc = (struct acpi_hest_ia_corrected *)hest_hdr;
- if (!cmc->enabled)
- return 0;
-
- /*
- * We expect HEST to provide a list of MC banks that report errors
- * in firmware first mode. Otherwise, return non-zero value to
- * indicate that we are done parsing HEST.
- */
- if (!(cmc->flags & ACPI_HEST_FIRMWARE_FIRST) || !cmc->num_hardware_banks)
- return 1;
-
- pr_info(HEST_PFX "Enabling Firmware First mode for corrected errors.\n");
-
- mc_bank = (struct acpi_hest_ia_error_bank *)(cmc + 1);
- for (i = 0; i < cmc->num_hardware_banks; i++, mc_bank++)
- mce_disable_bank(mc_bank->bank_number);
-#endif
- return 1;
+ return arch_apei_enable_cmcff(hest_hdr, data);
}
struct ghes_arr {
diff --git a/include/acpi/apei.h b/include/acpi/apei.h
index 04f349d..8a23c95 100644
--- a/include/acpi/apei.h
+++ b/include/acpi/apei.h
@@ -42,5 +42,8 @@ ssize_t erst_read(u64 record_id, struct cper_record_header *record,
size_t buflen);
int erst_clear(u64 record_id);
+int arch_apei_enable_cmcff(struct acpi_hest_header *hest_hdr, void *data);
+void arch_apei_report_mem_error(int sev, struct cper_sec_mem_err *mem_err);
+
#endif
#endif
--
1.9.1
GHES currently maps two pages with atomic_ioremap. From now
on, NMI is architectural depended so there is no need to allocate
an NMI page for platforms without NMI support.
To make it possible to not use a second page, swap the existing
page order so that the IRQ context page is first, and the optional
NMI context page is second. Then, use HAVE_ACPI_APEI_NMI to decide
how many pages are to be allocated.
Acked-by: Borislav Petkov <[email protected]>
Signed-off-by: Tomasz Nowicki <[email protected]>
---
arch/x86/kernel/acpi/apei.c | 6 ++++++
drivers/acpi/apei/ghes.c | 18 +++++++++++-------
include/acpi/apei.h | 1 +
3 files changed, 18 insertions(+), 7 deletions(-)
diff --git a/arch/x86/kernel/acpi/apei.c b/arch/x86/kernel/acpi/apei.c
index 12b13de..c280df6 100644
--- a/arch/x86/kernel/acpi/apei.c
+++ b/arch/x86/kernel/acpi/apei.c
@@ -15,6 +15,7 @@
#include <acpi/apei.h>
#include <asm/mce.h>
+#include <asm/tlbflush.h>
int arch_apei_enable_cmcff(struct acpi_hest_header *hest_hdr, void *data)
{
@@ -54,3 +55,8 @@ void arch_apei_report_mem_error(int sev, struct cper_sec_mem_err *mem_err)
apei_mce_report_mem_error(sev, mem_err);
#endif
}
+
+void arch_apei_flush_tlb_one(unsigned long addr)
+{
+ __flush_tlb_one(addr);
+}
diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
index 7fcf4d7..e05d84e7 100644
--- a/drivers/acpi/apei/ghes.c
+++ b/drivers/acpi/apei/ghes.c
@@ -105,12 +105,16 @@ static DEFINE_MUTEX(ghes_list_mutex);
*/
/*
- * Two virtual pages are used, one for NMI context, the other for
- * IRQ/PROCESS context
+ * Two virtual pages are used, one for IRQ/PROCESS context, the other for
+ * NMI context (optionally).
*/
-#define GHES_IOREMAP_PAGES 2
-#define GHES_IOREMAP_NMI_PAGE(base) (base)
-#define GHES_IOREMAP_IRQ_PAGE(base) ((base) + PAGE_SIZE)
+#ifdef CONFIG_HAVE_ACPI_APEI_NMI
+#define GHES_IOREMAP_PAGES 2
+#else
+#define GHES_IOREMAP_PAGES 1
+#endif
+#define GHES_IOREMAP_IRQ_PAGE(base) (base)
+#define GHES_IOREMAP_NMI_PAGE(base) ((base) + PAGE_SIZE)
/* virtual memory area for atomic ioremap */
static struct vm_struct *ghes_ioremap_area;
@@ -173,7 +177,7 @@ static void ghes_iounmap_nmi(void __iomem *vaddr_ptr)
BUG_ON(vaddr != (unsigned long)GHES_IOREMAP_NMI_PAGE(base));
unmap_kernel_range_noflush(vaddr, PAGE_SIZE);
- __flush_tlb_one(vaddr);
+ arch_apei_flush_tlb_one(vaddr);
}
static void ghes_iounmap_irq(void __iomem *vaddr_ptr)
@@ -183,7 +187,7 @@ static void ghes_iounmap_irq(void __iomem *vaddr_ptr)
BUG_ON(vaddr != (unsigned long)GHES_IOREMAP_IRQ_PAGE(base));
unmap_kernel_range_noflush(vaddr, PAGE_SIZE);
- __flush_tlb_one(vaddr);
+ arch_apei_flush_tlb_one(vaddr);
}
static int ghes_estatus_pool_init(void)
diff --git a/include/acpi/apei.h b/include/acpi/apei.h
index 8a23c95..76284bb 100644
--- a/include/acpi/apei.h
+++ b/include/acpi/apei.h
@@ -44,6 +44,7 @@ int erst_clear(u64 record_id);
int arch_apei_enable_cmcff(struct acpi_hest_header *hest_hdr, void *data);
void arch_apei_report_mem_error(int sev, struct cper_sec_mem_err *mem_err);
+void arch_apei_flush_tlb_one(unsigned long addr);
#endif
#endif
--
1.9.1
Currently APEI depends on x86 architecture. It is because of NMI hardware
error notification of GHES which is currently supported by x86 only.
However, many other APEI features can be still used perfectly by other
architectures.
This commit adds two symbols:
1. HAVE_ACPI_APEI for those archs which support APEI.
2. HAVE_ACPI_APEI_NMI which is used for NMI code isolation in ghes.c
file. NMI related data and functions are grouped so they can be wrapped
inside one #ifdef section. Appropriate function stubs are provided for
!NMI case.
Note there is no functional changes for x86 due to hard selected
HAVE_ACPI_APEI and HAVE_ACPI_APEI_NMI symbols.
Signed-off-by: Tomasz Nowicki <[email protected]>
---
arch/x86/Kconfig | 2 +
drivers/acpi/apei/Kconfig | 8 ++-
drivers/acpi/apei/ghes.c | 149 ++++++++++++++++++++++++++++++----------------
include/linux/nmi.h | 4 ++
4 files changed, 110 insertions(+), 53 deletions(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 3fc9b12..e7cca5d 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -132,6 +132,8 @@ config X86
select HAVE_CC_STACKPROTECTOR
select GENERIC_CPU_AUTOPROBE
select HAVE_ARCH_AUDITSYSCALL
+ select HAVE_ACPI_APEI
+ select HAVE_ACPI_APEI_NMI
config INSTRUCTION_DECODER
def_bool y
diff --git a/drivers/acpi/apei/Kconfig b/drivers/acpi/apei/Kconfig
index c4dac71..b0140c8 100644
--- a/drivers/acpi/apei/Kconfig
+++ b/drivers/acpi/apei/Kconfig
@@ -1,9 +1,15 @@
+config HAVE_ACPI_APEI
+ bool
+
+config HAVE_ACPI_APEI_NMI
+ bool
+
config ACPI_APEI
bool "ACPI Platform Error Interface (APEI)"
select MISC_FILESYSTEMS
select PSTORE
select UEFI_CPER
- depends on X86
+ depends on HAVE_ACPI_APEI
help
APEI allows to report errors (for example from the chipset)
to the operating system. This improves NMI handling
diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
index 352170a..7fcf4d7 100644
--- a/drivers/acpi/apei/ghes.c
+++ b/drivers/acpi/apei/ghes.c
@@ -47,11 +47,11 @@
#include <linux/genalloc.h>
#include <linux/pci.h>
#include <linux/aer.h>
+#include <linux/nmi.h>
#include <acpi/ghes.h>
#include <acpi/apei.h>
#include <asm/tlbflush.h>
-#include <asm/nmi.h>
#include "apei-internal.h"
@@ -86,8 +86,6 @@
bool ghes_disable;
module_param_named(disable, ghes_disable, bool, 0);
-static int ghes_panic_timeout __read_mostly = 30;
-
/*
* All error sources notified with SCI shares one notifier function,
* so they need to be linked and checked one by one. This is applied
@@ -97,16 +95,9 @@ static int ghes_panic_timeout __read_mostly = 30;
* list changing, not for traversing.
*/
static LIST_HEAD(ghes_sci);
-static LIST_HEAD(ghes_nmi);
static DEFINE_MUTEX(ghes_list_mutex);
/*
- * NMI may be triggered on any CPU, so ghes_nmi_lock is used for
- * mutual exclusion.
- */
-static DEFINE_RAW_SPINLOCK(ghes_nmi_lock);
-
-/*
* Because the memory area used to transfer hardware error information
* from BIOS to Linux can be determined only in NMI, IRQ or timer
* handler, but general ioremap can not be used in atomic context, so
@@ -130,18 +121,8 @@ static struct vm_struct *ghes_ioremap_area;
static DEFINE_RAW_SPINLOCK(ghes_ioremap_lock_nmi);
static DEFINE_SPINLOCK(ghes_ioremap_lock_irq);
-/*
- * printk is not safe in NMI context. So in NMI handler, we allocate
- * required memory from lock-less memory allocator
- * (ghes_estatus_pool), save estatus into it, put them into lock-less
- * list (ghes_estatus_llist), then delay printk into IRQ context via
- * irq_work (ghes_proc_irq_work). ghes_estatus_size_request record
- * required pool size by all NMI error source.
- */
static struct gen_pool *ghes_estatus_pool;
static unsigned long ghes_estatus_pool_size_request;
-static struct llist_head ghes_estatus_llist;
-static struct irq_work ghes_proc_irq_work;
struct ghes_estatus_cache *ghes_estatus_caches[GHES_ESTATUS_CACHES_SIZE];
static atomic_t ghes_estatus_cache_alloced;
@@ -249,11 +230,6 @@ static int ghes_estatus_pool_expand(unsigned long len)
return 0;
}
-static void ghes_estatus_pool_shrink(unsigned long len)
-{
- ghes_estatus_pool_size_request -= PAGE_ALIGN(len);
-}
-
static struct ghes *ghes_new(struct acpi_hest_generic *generic)
{
struct ghes *ghes;
@@ -732,6 +708,32 @@ static int ghes_notify_sci(struct notifier_block *this,
return ret;
}
+static struct notifier_block ghes_notifier_sci = {
+ .notifier_call = ghes_notify_sci,
+};
+
+#ifdef CONFIG_HAVE_ACPI_APEI_NMI
+/*
+ * printk is not safe in NMI context. So in NMI handler, we allocate
+ * required memory from lock-less memory allocator
+ * (ghes_estatus_pool), save estatus into it, put them into lock-less
+ * list (ghes_estatus_llist), then delay printk into IRQ context via
+ * irq_work (ghes_proc_irq_work). ghes_estatus_size_request record
+ * required pool size by all NMI error source.
+ */
+static struct llist_head ghes_estatus_llist;
+static struct irq_work ghes_proc_irq_work;
+
+/*
+ * NMI may be triggered on any CPU, so ghes_nmi_lock is used for
+ * mutual exclusion.
+ */
+static DEFINE_RAW_SPINLOCK(ghes_nmi_lock);
+
+static LIST_HEAD(ghes_nmi);
+
+static int ghes_panic_timeout __read_mostly = 30;
+
static struct llist_node *llist_nodes_reverse(struct llist_node *llnode)
{
struct llist_node *next, *tail = NULL;
@@ -875,10 +877,6 @@ out:
return ret;
}
-static struct notifier_block ghes_notifier_sci = {
- .notifier_call = ghes_notify_sci,
-};
-
static unsigned long ghes_esource_prealloc_size(
const struct acpi_hest_generic *generic)
{
@@ -894,11 +892,71 @@ static unsigned long ghes_esource_prealloc_size(
return prealloc_size;
}
+static void ghes_estatus_pool_shrink(unsigned long len)
+{
+ ghes_estatus_pool_size_request -= PAGE_ALIGN(len);
+}
+
+static void ghes_nmi_add(struct ghes *ghes)
+{
+ unsigned long len;
+
+ len = ghes_esource_prealloc_size(ghes->generic);
+ ghes_estatus_pool_expand(len);
+ mutex_lock(&ghes_list_mutex);
+ if (list_empty(&ghes_nmi))
+ register_nmi_handler(NMI_LOCAL, ghes_notify_nmi, 0, "ghes");
+ list_add_rcu(&ghes->list, &ghes_nmi);
+ mutex_unlock(&ghes_list_mutex);
+}
+
+static void ghes_nmi_remove(struct ghes *ghes)
+{
+ unsigned long len;
+
+ mutex_lock(&ghes_list_mutex);
+ list_del_rcu(&ghes->list);
+ if (list_empty(&ghes_nmi))
+ unregister_nmi_handler(NMI_LOCAL, "ghes");
+ mutex_unlock(&ghes_list_mutex);
+ /*
+ * To synchronize with NMI handler, ghes can only be
+ * freed after NMI handler finishes.
+ */
+ synchronize_rcu();
+ len = ghes_esource_prealloc_size(ghes->generic);
+ ghes_estatus_pool_shrink(len);
+}
+
+static void ghes_nmi_init_cxt(void)
+{
+ init_irq_work(&ghes_proc_irq_work, ghes_proc_in_irq);
+}
+#else /* CONFIG_HAVE_ACPI_APEI_NMI */
+static inline void ghes_nmi_add(struct ghes *ghes)
+{
+ pr_err(GHES_PFX "ID: %d, trying to add NMI notification which is not supported!\n",
+ ghes->generic->header.source_id);
+ BUG();
+}
+
+static inline void ghes_nmi_remove(struct ghes *ghes)
+{
+ pr_err(GHES_PFX "ID: %d, trying to remove NMI notification which is not supported!\n",
+ ghes->generic->header.source_id);
+ BUG();
+}
+
+static inline void ghes_nmi_init_cxt(void)
+{
+}
+#endif /* CONFIG_HAVE_ACPI_APEI_NMI */
+
static int ghes_probe(struct platform_device *ghes_dev)
{
struct acpi_hest_generic *generic;
struct ghes *ghes = NULL;
- unsigned long len;
+
int rc = -EINVAL;
generic = *(struct acpi_hest_generic **)ghes_dev->dev.platform_data;
@@ -909,7 +967,13 @@ static int ghes_probe(struct platform_device *ghes_dev)
case ACPI_HEST_NOTIFY_POLLED:
case ACPI_HEST_NOTIFY_EXTERNAL:
case ACPI_HEST_NOTIFY_SCI:
+ break;
case ACPI_HEST_NOTIFY_NMI:
+ if (!IS_ENABLED(CONFIG_HAVE_ACPI_APEI_NMI)) {
+ pr_warn(GHES_PFX "Generic hardware error source: %d notified via NMI interrupt is not supported!\n",
+ generic->header.source_id);
+ goto err;
+ }
break;
case ACPI_HEST_NOTIFY_LOCAL:
pr_warning(GHES_PFX "Generic hardware error source: %d notified via local interrupt is not supported!\n",
@@ -970,14 +1034,7 @@ static int ghes_probe(struct platform_device *ghes_dev)
mutex_unlock(&ghes_list_mutex);
break;
case ACPI_HEST_NOTIFY_NMI:
- len = ghes_esource_prealloc_size(generic);
- ghes_estatus_pool_expand(len);
- mutex_lock(&ghes_list_mutex);
- if (list_empty(&ghes_nmi))
- register_nmi_handler(NMI_LOCAL, ghes_notify_nmi, 0,
- "ghes");
- list_add_rcu(&ghes->list, &ghes_nmi);
- mutex_unlock(&ghes_list_mutex);
+ ghes_nmi_add(ghes);
break;
default:
BUG();
@@ -999,7 +1056,6 @@ static int ghes_remove(struct platform_device *ghes_dev)
{
struct ghes *ghes;
struct acpi_hest_generic *generic;
- unsigned long len;
ghes = platform_get_drvdata(ghes_dev);
generic = ghes->generic;
@@ -1020,18 +1076,7 @@ static int ghes_remove(struct platform_device *ghes_dev)
mutex_unlock(&ghes_list_mutex);
break;
case ACPI_HEST_NOTIFY_NMI:
- mutex_lock(&ghes_list_mutex);
- list_del_rcu(&ghes->list);
- if (list_empty(&ghes_nmi))
- unregister_nmi_handler(NMI_LOCAL, "ghes");
- mutex_unlock(&ghes_list_mutex);
- /*
- * To synchronize with NMI handler, ghes can only be
- * freed after NMI handler finishes.
- */
- synchronize_rcu();
- len = ghes_esource_prealloc_size(generic);
- ghes_estatus_pool_shrink(len);
+ ghes_nmi_remove(ghes);
break;
default:
BUG();
@@ -1075,7 +1120,7 @@ static int __init ghes_init(void)
return -EINVAL;
}
- init_irq_work(&ghes_proc_irq_work, ghes_proc_in_irq);
+ ghes_nmi_init_cxt();
rc = ghes_ioremap_init();
if (rc)
diff --git a/include/linux/nmi.h b/include/linux/nmi.h
index 447775e..1d2a6ab 100644
--- a/include/linux/nmi.h
+++ b/include/linux/nmi.h
@@ -63,4 +63,8 @@ extern int proc_dowatchdog(struct ctl_table *, int ,
void __user *, size_t *, loff_t *);
#endif
+#ifdef CONFIG_HAVE_ACPI_APEI_NMI
+#include <asm/nmi.h>
+#endif
+
#endif
--
1.9.1
On Tue, Jul 22, 2014 at 11:20:12AM +0200, Tomasz Nowicki wrote:
> Currently APEI depends on x86 architecture. It is because of NMI hardware
> error notification of GHES which is currently supported by x86 only.
> However, many other APEI features can be still used perfectly by other
> architectures.
>
> This commit adds two symbols:
> 1. HAVE_ACPI_APEI for those archs which support APEI.
> 2. HAVE_ACPI_APEI_NMI which is used for NMI code isolation in ghes.c
> file. NMI related data and functions are grouped so they can be wrapped
> inside one #ifdef section. Appropriate function stubs are provided for
> !NMI case.
>
> Note there is no functional changes for x86 due to hard selected
> HAVE_ACPI_APEI and HAVE_ACPI_APEI_NMI symbols.
>
> Signed-off-by: Tomasz Nowicki <[email protected]>
Acked-by: Borislav Petkov <[email protected]>
--
Regards/Gruss,
Boris.
Sent from a fat crate under my desk. Formatting is fine.
--
On Tue, Jul 22, 2014 at 2:20 AM, Tomasz Nowicki
<[email protected]> wrote:
> APEI is currently implemented so that it depends on x86 hardware.
> The primary dependency is that GHES uses the x86 NMI for hardware
> error notification and MCE for memory error handling. These patches
> remove that dependency.
>
> Other APEI features such as error reporting via external IRQ, error
> serialization, or error injection, do not require changes to use them
> on non-x86 architectures.
>
> The following patch set eliminates the APEI Kconfig x86 dependency
> by making these changes:
> - treat NMI notification as GHES architecture - HAVE_ACPI_APEI_NMI
> - group and wrap around #ifdef CONFIG_HAVE_ACPI_APEI_NMI code which
> is used only for NMI path
> - identify architectural boxes and abstract it accordingly (tlb flush and MCE)
> - rework ioremap for both IRQ and NMI context
>
> NMI code is kept in ghes.c file since NMI and IRQ context are tightly coupled.
>
> Note, these patches introduce no functional changes for x86. The NMI notification
> feature is hard selected for x86. Architectures that want to use this
> feature should also provide NMI code infrastructure.
Looks OK to me. I've thrown this series into a topic branch of the ras
git tree named "apei". When I push it to kernel.org in a few minutes
the build test robots will pounce and try all sorts of randconfig builds.
They'll send you (and me) e-mails if they find any corner cases.
-Tony
Many thanks Tony! And thanks goes to Borislav and Robert too.
Regards,
Tomasz
On 22.07.2014 23:08, Tony Luck wrote:
> On Tue, Jul 22, 2014 at 2:20 AM, Tomasz Nowicki
> <[email protected]> wrote:
>> APEI is currently implemented so that it depends on x86 hardware.
>> The primary dependency is that GHES uses the x86 NMI for hardware
>> error notification and MCE for memory error handling. These patches
>> remove that dependency.
>>
>> Other APEI features such as error reporting via external IRQ, error
>> serialization, or error injection, do not require changes to use them
>> on non-x86 architectures.
>>
>> The following patch set eliminates the APEI Kconfig x86 dependency
>> by making these changes:
>> - treat NMI notification as GHES architecture - HAVE_ACPI_APEI_NMI
>> - group and wrap around #ifdef CONFIG_HAVE_ACPI_APEI_NMI code which
>> is used only for NMI path
>> - identify architectural boxes and abstract it accordingly (tlb flush and MCE)
>> - rework ioremap for both IRQ and NMI context
>>
>> NMI code is kept in ghes.c file since NMI and IRQ context are tightly coupled.
>>
>> Note, these patches introduce no functional changes for x86. The NMI notification
>> feature is hard selected for x86. Architectures that want to use this
>> feature should also provide NMI code infrastructure.
>
> Looks OK to me. I've thrown this series into a topic branch of the ras
> git tree named "apei". When I push it to kernel.org in a few minutes
> the build test robots will pounce and try all sorts of randconfig builds.
> They'll send you (and me) e-mails if they find any corner cases.
>
> -Tony
>