lkml.org 
[lkml]   [2015]   [Jun]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH v4 6/6] arch, x86: pmem api for ensuring durability of persistent memory updates
On Thu, Jun 11, 2015 at 2:19 PM, Dan Williams <dan.j.williams@intel.com> wrote:
> From: Ross Zwisler <ross.zwisler@linux.intel.com>
>
> Based on an original patch by Ross Zwisler [1].
>
> Writes to persistent memory have the potential to be posted to cpu
> cache, cpu write buffers, and platform write buffers (memory controller)
> before being committed to persistent media. Provide apis,
> memcpy_to_pmem(), sync_pmem(), and memremap_pmem(), to write data to
> pmem and assert that it is durable in PMEM (a persistent linear address
> range). A '__pmem' attribute is added so sparse can track proper usage
> of pointers to pmem.
>
> [1]: https://lists.01.org/pipermail/linux-nvdimm/2015-May/000932.html
>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: "H. Peter Anvin" <hpa@zytor.com>
> Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
> [djbw: various reworks]
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> ---
> arch/x86/Kconfig | 1
> arch/x86/include/asm/cacheflush.h | 36 +++++++++++++
> arch/x86/include/asm/io.h | 6 ++
> drivers/block/pmem.c | 75 +++++++++++++++++++++++++--
> include/linux/compiler.h | 2 +
> include/linux/pmem.h | 102 +++++++++++++++++++++++++++++++++++++
> lib/Kconfig | 3 +
> 7 files changed, 218 insertions(+), 7 deletions(-)
> create mode 100644 include/linux/pmem.h
>
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index f16caf7eac27..5dfb8f31ac48 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -28,6 +28,7 @@ config X86
> select ARCH_HAS_FAST_MULTIPLIER
> select ARCH_HAS_GCOV_PROFILE_ALL
> select ARCH_HAS_MEMREMAP
> + select ARCH_HAS_PMEM_API
> select ARCH_HAS_SG_CHAIN
> select ARCH_HAVE_NMI_SAFE_CMPXCHG
> select ARCH_MIGHT_HAVE_ACPI_PDC if ACPI
> diff --git a/arch/x86/include/asm/cacheflush.h b/arch/x86/include/asm/cacheflush.h
> index b6f7457d12e4..4d896487382c 100644
> --- a/arch/x86/include/asm/cacheflush.h
> +++ b/arch/x86/include/asm/cacheflush.h
> @@ -4,6 +4,7 @@
> /* Caches aren't brain-dead on the intel. */
> #include <asm-generic/cacheflush.h>
> #include <asm/special_insns.h>
> +#include <asm/uaccess.h>
>
> /*
> * The set_memory_* API can be used to change various attributes of a virtual
> @@ -108,4 +109,39 @@ static inline int rodata_test(void)
> }
> #endif
>
> +#ifdef ARCH_HAS_NOCACHE_UACCESS
> +static inline void arch_memcpy_to_pmem(void __pmem *dst, const void *src, size_t n)
> +{
> + /*
> + * We are copying between two kernel buffers, if
> + * __copy_from_user_inatomic_nocache() returns an error (page
> + * fault) we would have already taken an unhandled fault before
> + * the BUG_ON. The BUG_ON is simply here to satisfy
> + * __must_check and allow reuse of the common non-temporal store
> + * implementation for memcpy_to_pmem().
> + */
> + BUG_ON(__copy_from_user_inatomic_nocache((void __force *) dst,
> + (void __user *) src, n));

Ick. If we take a fault, we will lose the debugging information we
would otherwise have gotten unless we get lucky and get a usable CR2
value in the oops.

> +}
> +
> +static inline void arch_sync_pmem(void)
> +{
> + wmb();
> + pcommit_sfence();
> +}

This function is non-intuitive to me. It's really "arch-specific sync
pmem after one or more copies using arch_memcpy_to_pmem". If normal
stores or memcpy to non-WC memory is used instead, then it's
insufficient if the memory is WB and it's unnecessarily slow if the
memory is WT or UC (the first sfence isn't needed).

I would change the name and add documentation. I'd also add a comment
about the wmb() being an SFENCE to flush pending non-temporal writes.

--Andy


\
 
 \ /
  Last update: 2015-06-17 17:21    [W:0.113 / U:0.912 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site