hw/display/qxl: Suppress clang-7 warning about misaligned atomic operation

If QEMU is compiled with clang-7 it results in the warning:

hw/display/qxl.c:1884:19: error: misaligned or large atomic operation
may incur significant performance penalty [-Werror,-Watomic-alignment]
    old_pending = atomic_fetch_or(&d->ram->int_pending, le_events);
                  ^

This is because the Spice headers forgot to define the QXLRam struct
with the '__aligned__(4)' attribute.  clang 7 and newer will thus
warn that the access here to int_pending might not be 4-aligned
(because the QXLRam object d->ram points at might start at a
misaligned address).  In fact we set up d->ram in init_qxl_ram() so
it always starts at a 4K boundary, so we know the atomic access here
is OK.

Newer Spice versions (with Spice commit
beda5ec7a6848be20c0cac2a9a8ef2a41e8069c1) will fix the bug;
for older Spice versions, work around it by telling the compiler
explicitly that the alignment is OK using __builtin_assume_aligned().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180927155538.699-1-peter.maydell@linaro.org
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

diff --git a/hw/display/qxl.c b/hw/display/qxl.c
index 747986478f2e4295673996ed537e79d9a035fd73..e628cf128613a7a1f164f5f3467b50e691f1baf2 100644 (file)
--- a/hw/display/qxl.c
+++ b/hw/display/qxl.c
@@ -1893,7 +1893,31 @@ static void qxl_send_events(PCIQXLDevice *d, uint32_t events)
         trace_qxl_send_events_vm_stopped(d->id, events);
         return;
     }
-    old_pending = atomic_fetch_or(&d->ram->int_pending, le_events);
+    /*
+     * Older versions of Spice forgot to define the QXLRam struct
+     * with the '__aligned__(4)' attribute. clang 7 and newer will
+     * thus warn that atomic_fetch_or(&d->ram->int_pending, ...)
+     * might be a misaligned atomic access, and will generate an
+     * out-of-line call for it, which results in a link error since
+     * we don't currently link against libatomic.
+     *
+     * In fact we set up d->ram in init_qxl_ram() so it always starts
+     * at a 4K boundary, so we know that &d->ram->int_pending is
+     * naturally aligned for a uint32_t. Newer Spice versions
+     * (with Spice commit beda5ec7a6848be20c0cac2a9a8ef2a41e8069c1)
+     * will fix the bug directly. To deal with older versions,
+     * we tell the compiler to assume the address really is aligned.
+     * Any compiler which cares about the misalignment will have
+     * __builtin_assume_aligned.
+     */
+#ifdef HAS_ASSUME_ALIGNED
+#define ALIGNED_UINT32_PTR(P) ((uint32_t *)__builtin_assume_aligned(P, 4))
+#else
+#define ALIGNED_UINT32_PTR(P) ((uint32_t *)P)
+#endif
+
+    old_pending = atomic_fetch_or(ALIGNED_UINT32_PTR(&d->ram->int_pending),
+                                  le_events);
     if ((old_pending & le_events) == le_events) {
         return;
     }

diff --git a/include/qemu/compiler.h b/include/qemu/compiler.h
index 5843812710c556266c8e61e3895870b96e4807f7..bf47e7bee4e7ecc1d37837aadaee0c09c842dacf 100644 (file)
--- a/include/qemu/compiler.h
+++ b/include/qemu/compiler.h
@@ -122,6 +122,15 @@
 #ifndef __has_feature
 #define __has_feature(x) 0 /* compatibility with non-clang compilers */
 #endif
+
+#ifndef __has_builtin
+#define __has_builtin(x) 0 /* compatibility with non-clang compilers */
+#endif
+
+#if __has_builtin(__builtin_assume_aligned) || QEMU_GNUC_PREREQ(4, 7)
+#define HAS_ASSUME_ALIGNED
+#endif
+
 /* Implement C11 _Generic via GCC builtins.  Example:
  *
  *    QEMU_GENERIC(x, (float, sinf), (long double, sinl), sin) (x)