]> xenbits.xensource.com Git - qemu-xen-4.3-testing.git/commitdiff
exec.c: Clarify comment about tlb_flush() flush_global parameter
authorPeter Maydell <peter.maydell@linaro.org>
Tue, 17 Jan 2012 13:23:13 +0000 (13:23 +0000)
committerAnthony Liguori <aliguori@us.ibm.com>
Wed, 1 Feb 2012 20:45:01 +0000 (14:45 -0600)
Clarify the comment about tlb_flush()'s flush_global parameter,
so it is clearer what it does and why it is OK that the implementation
currently ignores it.

Reviewed-by: Andreas F=C3=A4rber <afaerber@suse.de>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
exec.c

diff --git a/exec.c b/exec.c
index 5b9eb9aa192f532e4d53fa80ae8a7c6f54a7ed10..b81677ade9c7c67add6c4adbf4ceeb3d0d5ffe26 100644 (file)
--- a/exec.c
+++ b/exec.c
@@ -1876,8 +1876,18 @@ static CPUTLBEntry s_cputlb_empty_entry = {
     .addend     = -1,
 };
 
-/* NOTE: if flush_global is true, also flush global entries (not
-   implemented yet) */
+/* NOTE:
+ * If flush_global is true (the usual case), flush all tlb entries.
+ * If flush_global is false, flush (at least) all tlb entries not
+ * marked global.
+ *
+ * Since QEMU doesn't currently implement a global/not-global flag
+ * for tlb entries, at the moment tlb_flush() will also flush all
+ * tlb entries in the flush_global == false case. This is OK because
+ * CPU architectures generally permit an implementation to drop
+ * entries from the TLB at any time, so flushing more entries than
+ * required is only an efficiency issue, not a correctness issue.
+ */
 void tlb_flush(CPUState *env, int flush_global)
 {
     int i;