[Qemu-devel] [PATCH] Fix userland ELF load failure when no .bss is present

[Thiemo Seufer <ths@networkno.de>]

Hello All,

the appended patch fixes the case where a ELF Linux binary has a
zero-sized .bss, or none at all.


Thiemo


Index: qemu-work/linux-user/elfload.c
===================================================================
--- qemu-work.orig/linux-user/elfload.c	2006-12-12 18:25:00.000000000 +0000
+++ qemu-work/linux-user/elfload.c	2006-12-12 18:33:08.000000000 +0000
@@ -553,10 +553,13 @@
 /* We need to explicitly zero any fractional pages after the data
    section (i.e. bss).  This would contain the junk from the file that
    should not be in memory. */
-static void padzero(unsigned long elf_bss)
+static void padzero(unsigned long elf_bss, unsigned long last_bss)
 {
         unsigned long nbyte;
 
+	if (elf_bss >= last_bss)
+		return;
+
         /* XXX: this is really a hack : if the real host page size is
            smaller than the target page size, some pages after the end
            of the file may not be mapped. A better fix would be to
@@ -798,7 +801,7 @@
 	 * that there are zeromapped pages up to and including the last
 	 * bss page.
 	 */
-	padzero(elf_bss);
+	padzero(elf_bss, last_bss);
 	elf_bss = TARGET_ELF_PAGESTART(elf_bss + qemu_host_page_size - 1); /* What we have mapped so far */
 
 	/* Map the last of the bss segment */
@@ -1227,7 +1230,7 @@
        sections */
     set_brk(elf_bss, elf_brk);
 
-    padzero(elf_bss);
+    padzero(elf_bss, elf_brk);
 
 #if 0
     printf("(start_brk) %x\n" , info->start_brk);