lkml.org 
[lkml]   [2017]   [Feb]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH V8 05/10] acpi: apei: handle SEA notification type for ARMv8
Hi Tyler,

On 01/02/17 17:16, Tyler Baicar wrote:
> ARM APEI extension proposal added SEA (Synchronous External Abort)
> notification type for ARMv8.
> Add a new GHES error source handling function for SEA. If an error
> source's notification type is SEA, then this function can be registered
> into the SEA exception handler. That way GHES will parse and report
> SEA exceptions when they occur.

It's worth adding details of the other things this patch changes, just to alert
busy reviewers, something like:
An SEA can interrupt code that had interrupts masked and is treated as an NMI.
To aid this the page of address space for mapping APEI buffers while in_nmi() is
always reserved, and ghes_ioremap_pfn_nmi() is changed to use the helper methods
to find the prot_t to map with in the same way as ghes_ioremap_pfn_irq().


> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 1117421..f92778d 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -53,6 +53,8 @@ config ARM64
> select HANDLE_DOMAIN_IRQ
> select HARDIRQS_SW_RESEND
> select HAVE_ACPI_APEI if (ACPI && EFI)
> + select HAVE_ACPI_APEI_SEA if (ACPI && EFI)

> + select HAVE_NMI if HAVE_ACPI_APEI_SEA

Nit: This section of the file is largely in alphabetical order, can we try to
keep it that way?!


> select HAVE_ALIGNED_STRUCT_PAGE if SLUB
> select HAVE_ARCH_AUDITSYSCALL
> select HAVE_ARCH_BITREVERSE
> diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
> index 9ae7e65..5a5a096 100644
> --- a/arch/arm64/mm/fault.c
> +++ b/arch/arm64/mm/fault.c
> @@ -30,6 +30,7 @@
> #include <linux/highmem.h>
> #include <linux/perf_event.h>
> #include <linux/preempt.h>

> +#include <linux/hardirq.h>

This header is already included by this file further up.


>
> #include <asm/bug.h>
> #include <asm/cpufeature.h>
> @@ -41,6 +42,8 @@
> #include <asm/pgtable.h>
> #include <asm/tlbflush.h>
>
> +#include <acpi/ghes.h>
> +
> static const char *fault_name(unsigned int esr);
>
> #ifdef CONFIG_KPROBES
> @@ -500,6 +503,14 @@ static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs)
> pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n",
> fault_name(esr), esr, addr);
>

> + /*
> + * synchronize_rcu() will wait for nmi_exit(), so no need to
> + * rcu_read_lock().
> + */

This comment should go against the use of RCU in ghes_notify_sea(), but there
should be something here to explain the surprise use of nmi. Something like:
Synchronous aborts may interrupt code which had interrupts masked. Before
calling out into the wider kernel tell the interested subsystems.


This should be wrapped in:
if (IS_ENABLED(HAVE_ACPI_APEI_SEA)) {

> + nmi_enter();
> + ghes_notify_sea();
> + nmi_exit();

}

To avoid breaking systems that don't have SEA configured.


> +
> info.si_signo = SIGBUS;
> info.si_errno = 0;
> info.si_code = 0;
> diff --git a/drivers/acpi/apei/Kconfig b/drivers/acpi/apei/Kconfig
> index b0140c8..3786ff1 100644
> --- a/drivers/acpi/apei/Kconfig
> +++ b/drivers/acpi/apei/Kconfig
> @@ -4,6 +4,20 @@ config HAVE_ACPI_APEI
> config HAVE_ACPI_APEI_NMI
> bool
>
> +config HAVE_ACPI_APEI_SEA
> + bool "APEI Synchronous External Abort logging/recovering support"
> + depends on ARM64

depends on CONFIG_ACPI_APEI_GHES
?

(I think this is what the kbuild robot has managed to miss out!)


> + help
> + This option should be enabled if the system supports
> + firmware first handling of SEA (Synchronous External Abort).
> + SEA happens with certain faults of data abort or instruction
> + abort synchronous exceptions on ARMv8 systems. If a system
> + supports firmware first handling of SEA, the platform analyzes
> + and handles hardware error notifications with SEA, and it may then

Nit: notifications from SEA,

> + form a HW error record for the OS to parse and handle. This
> + option allows the OS to look for such HW error record, and

Nit: 'HW'->hardware. This is spelled out for the other seven uses in the file.


> + take appropriate action.
> +
> config ACPI_APEI
> bool "ACPI Platform Error Interface (APEI)"
> select MISC_FILESYSTEMS
> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
> index b25e7cf..8756172 100644
> --- a/drivers/acpi/apei/ghes.c
> +++ b/drivers/acpi/apei/ghes.c
> @@ -114,11 +114,7 @@
> * Two virtual pages are used, one for IRQ/PROCESS context, the other for
> * NMI context (optionally).
> */
> -#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)
>
> @@ -156,11 +152,14 @@ static void ghes_ioremap_exit(void)
>
> static void __iomem *ghes_ioremap_pfn_nmi(u64 pfn)
> {
> - unsigned long vaddr;
> + unsigned long vaddr, paddr;
> + pgprot_t prot;
>
> vaddr = (unsigned long)GHES_IOREMAP_NMI_PAGE(ghes_ioremap_area->addr);
> - ioremap_page_range(vaddr, vaddr + PAGE_SIZE,
> - pfn << PAGE_SHIFT, PAGE_KERNEL);
> +

> + paddr = pfn << PAGE_SHIFT;

Physical addresses might not always fit in 'unsigned long'. phys_addr_t exists
to hide this nasty detail!

From arch/x86/Kconfig:
> config ARCH_PHYS_ADDR_T_64BIT
> def_bool y
> depends on X86_64 || X86_PAE

32bit x86 kernels configured with PAE define phys_addr_t to be u64.


> + prot = arch_apei_get_mem_attribute(paddr);
> + ioremap_page_range(vaddr, vaddr + PAGE_SIZE, paddr, prot);
>
> return (void __iomem *)vaddr;
> }
> @@ -767,6 +766,48 @@ static int ghes_notify_sci(struct notifier_block *this,
> .notifier_call = ghes_notify_sci,
> };
>
> +#ifdef CONFIG_HAVE_ACPI_APEI_SEA
> +static LIST_HEAD(ghes_sea);
> +
> +void ghes_notify_sea(void)
> +{
> + struct ghes *ghes;
> +

/*
* synchronize_rcu() will wait for nmi_exit(), so no need to
* rcu_read_lock().
*/
> + list_for_each_entry_rcu(ghes, &ghes_sea, list) {
> + ghes_proc(ghes);
> + }
> +}
> +
> +static int ghes_sea_add(struct ghes *ghes)
> +{
> + mutex_lock(&ghes_list_mutex);
> + list_add_rcu(&ghes->list, &ghes_sea);
> + mutex_unlock(&ghes_list_mutex);

> + return 0;

This function returns 0 or -ENOTSUPP, depending on CONFIG_HAVE_ACPI_APEI_SEA,
but ...


> +}
> +
> +static void ghes_sea_remove(struct ghes *ghes)
> +{
> + mutex_lock(&ghes_list_mutex);
> + list_del_rcu(&ghes->list);
> + mutex_unlock(&ghes_list_mutex);
> + synchronize_rcu();
> +}
> +#else /* CONFIG_HAVE_ACPI_APEI_SEA */
> +static inline int ghes_sea_add(struct ghes *ghes)
> +{
> + pr_err(GHES_PFX "ID: %d, trying to add SEA notification which is not supported\n",
> + ghes->generic->header.source_id);
> + return -ENOTSUPP;
> +}
> +
> +static inline void ghes_sea_remove(struct ghes *ghes)
> +{
> + pr_err(GHES_PFX "ID: %d, trying to remove SEA notification which is not supported\n",
> + ghes->generic->header.source_id);
> +}
> +#endif /* CONFIG_HAVE_ACPI_APEI_SEA */
> +
> #ifdef CONFIG_HAVE_ACPI_APEI_NMI
> /*
> * printk is not safe in NMI context. So in NMI handler, we allocate
> @@ -1012,6 +1053,14 @@ static int ghes_probe(struct platform_device *ghes_dev)
> case ACPI_HEST_NOTIFY_EXTERNAL:
> case ACPI_HEST_NOTIFY_SCI:
> break;
> + case ACPI_HEST_NOTIFY_SEA:
> + if (!IS_ENABLED(CONFIG_HAVE_ACPI_APEI_SEA)) {
> + pr_warn(GHES_PFX "Generic hardware error source: %d notified via SEA is not supported\n",
> + generic->header.source_id);
> + rc = -ENOTSUPP;
> + goto err;

... here we jump out of ghes_probe() if NOTIFY_SEA is used but the kernel wasn't
built with CONFIG_HAVE_ACPI_APEI_SEA....

> + }
> + 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",
> @@ -1023,6 +1072,13 @@ static int ghes_probe(struct platform_device *ghes_dev)
> pr_warning(GHES_PFX "Generic hardware error source: %d notified via local interrupt is not supported!\n",
> generic->header.source_id);
> goto err;
> + case ACPI_HEST_NOTIFY_GPIO:
> + case ACPI_HEST_NOTIFY_SEI:
> + case ACPI_HEST_NOTIFY_GSIV:
> + pr_warn(GHES_PFX "Generic hardware error source: %d notified via notification type %u is not supported\n",
> + generic->header.source_id, generic->header.source_id);
> + rc = -ENOTSUPP;
> + goto err;
> default:
> pr_warning(FW_WARN GHES_PFX "Unknown notification type: %u for generic hardware error source: %d\n",
> generic->notify.type, generic->header.source_id);
> @@ -1077,6 +1133,11 @@ static int ghes_probe(struct platform_device *ghes_dev)
> list_add_rcu(&ghes->list, &ghes_sci);
> mutex_unlock(&ghes_list_mutex);
> break;
> + case ACPI_HEST_NOTIFY_SEA:
> + rc = ghes_sea_add(ghes);

... so this error handling will never be needed.
ghes_nmi_add() returns void.

I guess the not-configured versions of the symbols need to exist for older
compilers that can't work out that this can never be called.


> + if (rc)
> + goto err_edac_unreg;
> + break;
> case ACPI_HEST_NOTIFY_NMI:
> ghes_nmi_add(ghes);
> break;
> @@ -1119,6 +1180,9 @@ static int ghes_remove(struct platform_device *ghes_dev)
> unregister_acpi_hed_notifier(&ghes_notifier_sci);
> mutex_unlock(&ghes_list_mutex);
> break;
> + case ACPI_HEST_NOTIFY_SEA:
> + ghes_sea_remove(ghes);
> + break;
> case ACPI_HEST_NOTIFY_NMI:
> ghes_nmi_remove(ghes);
> break;

> diff --git a/include/acpi/ghes.h b/include/acpi/ghes.h
> index 6ae318b..adf5455 100644
> --- a/include/acpi/ghes.h
> +++ b/include/acpi/ghes.h
> @@ -95,3 +95,5 @@ static inline void *acpi_hest_generic_data_payload(struct acpi_hest_generic_data
> (void *)(((struct acpi_hest_generic_data_v300 *)(gdata)) + 1) :
> gdata + 1;
> }
> +
> +void ghes_notify_sea(void);
>

This header file ought to have an include guard, could you add one?

I think the kbuild-robot has managed to configure SEA on, but GHES off so ghes.c
isn't included in the kernel. I will dig some more into this on Monday.



Thanks,

James



\
 
 \ /
  Last update: 2017-02-03 17:03    [W:0.303 / U:0.520 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site