[Xen-devel] [PATCHv4 3/8] x86: provide add_sized()

[David Vrabel <david.vrabel@xxxxxxxxxx>]

add_sized(ptr, inc) adds inc to the value at ptr using only the correct
size of loads and stores for the type of *ptr.  The add is /not/ atomic.

This is needed for ticket locks to ensure the increment of the head ticket
does not affect the tail ticket.

Signed-off-by: David Vrabel <david.vrabel@xxxxxxxxxx>
---
 xen/include/asm-x86/atomic.h |   28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/xen/include/asm-x86/atomic.h b/xen/include/asm-x86/atomic.h
index aadded0..d97ea54 100644
--- a/xen/include/asm-x86/atomic.h
+++ b/xen/include/asm-x86/atomic.h
@@ -14,6 +14,14 @@ static inline void name(volatile type *addr, type val) \
 { asm volatile("mov" size " %1,%0": "=m" (*(volatile type *)addr) \
 :reg (val) barrier); }
 
+#define build_add_sized(name, size, type, reg) \
+    static inline void name(volatile type *addr, type val)              \
+    {                                                                   \
+        asm volatile("add" size " %1,%0"                                \
+                     : "=m" (*(volatile type *)addr)                    \
+                     : reg (val));                                      \
+    }
+
 build_read_atomic(read_u8_atomic, "b", uint8_t, "=q", )
 build_read_atomic(read_u16_atomic, "w", uint16_t, "=r", )
 build_read_atomic(read_u32_atomic, "l", uint32_t, "=r", )
@@ -25,8 +33,14 @@ build_write_atomic(write_u32_atomic, "l", uint32_t, "r", )
 build_read_atomic(read_u64_atomic, "q", uint64_t, "=r", )
 build_write_atomic(write_u64_atomic, "q", uint64_t, "r", )
 
+build_add_sized(add_u8_sized, "b", uint8_t, "qi")
+build_add_sized(add_u16_sized, "w", uint16_t, "ri")
+build_add_sized(add_u32_sized, "l", uint32_t, "ri")
+build_add_sized(add_u64_sized, "q", uint64_t, "ri")
+
 #undef build_read_atomic
 #undef build_write_atomic
+#undef build_add_sized
 
 void __bad_atomic_size(void);
 
@@ -55,6 +69,20 @@ void __bad_atomic_size(void);
     __x;                                                                \
 })
 
+#define add_sized(p, x) ({                                              \
+    typeof(*(p)) __x = (x);                                             \
+    unsigned long x_ = (unsigned long)__x;                              \
+    switch ( sizeof(*(p)) )                                             \
+        {                                                               \
+        case 1: add_u8_sized((uint8_t *)(p), (uint8_t)x_); break;       \
+        case 2: add_u16_sized((uint16_t *)(p), (uint16_t)x_); break;    \
+        case 4: add_u32_sized((uint32_t *)(p), (uint32_t)x_); break;    \
+        case 8: add_u64_sized((uint64_t *)(p), (uint64_t)x_); break;    \
+        default: __bad_atomic_size(); break;                            \
+        }                                                               \
+    __x;                                                                \
+})
+
 /*
  * NB. I've pushed the volatile qualifier into the operations. This allows
  * fast accessors such as _atomic_read() and _atomic_set() which don't give
-- 
1.7.10.4


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel