OSDN Git Service

tcg: Do not set guard pages on the rx portion of code_gen_buffer
authorRichard Henderson <richard.henderson@linaro.org>
Sat, 20 Mar 2021 16:57:19 +0000 (10:57 -0600)
committerRichard Henderson <richard.henderson@linaro.org>
Wed, 24 Mar 2021 01:36:36 +0000 (19:36 -0600)
The rw portion of the buffer is the only one in which overruns
can be generated.  Allow the rx portion to be more completely
covered by huge pages.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
Message-Id: <20210320165720.1813545-2-richard.henderson@linaro.org>

tcg/tcg.c

index de91bb6..88c9e6f 100644 (file)
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -828,7 +828,6 @@ void tcg_region_init(void)
     size_t region_size;
     size_t n_regions;
     size_t i;
-    uintptr_t splitwx_diff;
 
     n_regions = tcg_n_regions();
 
@@ -858,8 +857,11 @@ void tcg_region_init(void)
     /* account for that last guard page */
     region.end -= page_size;
 
-    /* set guard pages */
-    splitwx_diff = tcg_splitwx_diff;
+    /*
+     * Set guard pages in the rw buffer, as that's the one into which
+     * buffer overruns could occur.  Do not set guard pages in the rx
+     * buffer -- let that one use hugepages throughout.
+     */
     for (i = 0; i < region.n; i++) {
         void *start, *end;
         int rc;
@@ -867,10 +869,6 @@ void tcg_region_init(void)
         tcg_region_bounds(i, &start, &end);
         rc = qemu_mprotect_none(end, page_size);
         g_assert(!rc);
-        if (splitwx_diff) {
-            rc = qemu_mprotect_none(end + splitwx_diff, page_size);
-            g_assert(!rc);
-        }
     }
 
     tcg_region_trees_init();