OSDN Git Service

percpu: Introduce DEFINE_PER_CPU_DECRYPTED
authorBrijesh Singh <brijesh.singh@amd.com>
Fri, 20 Oct 2017 14:30:57 +0000 (09:30 -0500)
committerThomas Gleixner <tglx@linutronix.de>
Tue, 7 Nov 2017 14:35:59 +0000 (15:35 +0100)
KVM guest defines three per-CPU variables (steal-time, apf_reason, and
kvm_pic_eoi) which are shared between a guest and a hypervisor.

When SEV is active, memory is encrypted with a guest-specific key, and if
the guest OS wants to share the memory region with the hypervisor then it
must clear the C-bit (i.e set decrypted) before sharing it.

DEFINE_PER_CPU_DECRYPTED can be used to define the per-CPU variables
which will be shared between a guest and a hypervisor.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Borislav Petkov <bp@suse.de>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Borislav Petkov <bp@suse.de>
Cc: linux-arch@vger.kernel.org
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: kvm@vger.kernel.org
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Christoph Lameter <cl@linux.com>
Link: https://lkml.kernel.org/r/20171020143059.3291-16-brijesh.singh@amd.com
include/asm-generic/vmlinux.lds.h
include/linux/percpu-defs.h

index 63e56f6..c58f380 100644 (file)
 #endif
 
 /*
+ * Memory encryption operates on a page basis. Since we need to clear
+ * the memory encryption mask for this section, it needs to be aligned
+ * on a page boundary and be a page-size multiple in length.
+ *
+ * Note: We use a separate section so that only this section gets
+ * decrypted to avoid exposing more than we wish.
+ */
+#ifdef CONFIG_AMD_MEM_ENCRYPT
+#define PERCPU_DECRYPTED_SECTION                                       \
+       . = ALIGN(PAGE_SIZE);                                           \
+       *(.data..percpu..decrypted)                                     \
+       . = ALIGN(PAGE_SIZE);
+#else
+#define PERCPU_DECRYPTED_SECTION
+#endif
+
+
+/*
  * Default discarded sections.
  *
  * Some archs want to discard exit text/data at runtime rather than
        . = ALIGN(cacheline);                                           \
        *(.data..percpu)                                                \
        *(.data..percpu..shared_aligned)                                \
+       PERCPU_DECRYPTED_SECTION                                        \
        VMLINUX_SYMBOL(__per_cpu_end) = .;
 
 /**
index 8f16299..2d2096b 100644 (file)
        DEFINE_PER_CPU_SECTION(type, name, "..read_mostly")
 
 /*
+ * Declaration/definition used for per-CPU variables that should be accessed
+ * as decrypted when memory encryption is enabled in the guest.
+ */
+#if defined(CONFIG_VIRTUALIZATION) && defined(CONFIG_AMD_MEM_ENCRYPT)
+
+#define DECLARE_PER_CPU_DECRYPTED(type, name)                          \
+       DECLARE_PER_CPU_SECTION(type, name, "..decrypted")
+
+#define DEFINE_PER_CPU_DECRYPTED(type, name)                           \
+       DEFINE_PER_CPU_SECTION(type, name, "..decrypted")
+#else
+#define DEFINE_PER_CPU_DECRYPTED(type, name)   DEFINE_PER_CPU(type, name)
+#endif
+
+/*
  * Intermodule exports for per-CPU variables.  sparse forgets about
  * address space across EXPORT_SYMBOL(), change EXPORT_SYMBOL() to
  * noop if __CHECKER__.