Hi,
attached patch corrects a bug in cmpxchg emulation in the hypervisor.
BS2000 running as HVM-domain on 4 vcpus (no HAP) hit an error due to this bug
after several days.
Juergen
--
Juergen Gross Principal Developer Operating Systems
TSP ES&S SWE OS6 Telephone: +49 (0) 89 636 47950
Fujitsu Technolgy Solutions e-mail: juergen.gross@xxxxxxxxxxxxxx
Otto-Hahn-Ring 6 Internet: ts.fujitsu.com
D-81739 Muenchen Company details: ts.fujitsu.com/imprint.html
The cmpxchg emulation for accesses to page tables of guests doesn't handle
races correct.
ops->cmpxchg might return X86EMUL_CMPXCHG_FAILED if the addressed memory
location changed after checking the old contents. In this case ZF was not
changed and could remain 1 instead of being set to 0.
Signed-off-by: juergen.gross@xxxxxxxxxxxxxx
# HG changeset patch
# User juergen.gross@xxxxxxxxxxxxxx
# Date 1249540842 -7200
# Node ID 26adbdb6cb1d59d95e0a65b6a0d38fa8e95b9f51
# Parent 68e8b8379244e293c55875e7dc3692fc81d3d212
handle race on cmpxchg emulation
diff -r 68e8b8379244 -r 26adbdb6cb1d xen/arch/x86/x86_emulate/x86_emulate.c
--- a/xen/arch/x86/x86_emulate/x86_emulate.c Sun Aug 02 13:43:15 2009 +0100
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c Thu Aug 06 08:40:42 2009 +0200
@@ -4124,6 +4124,7 @@
op_bytes *= 2;
/* Get actual old value. */
+cmpxchg_failed:
for ( i = 0; i < (op_bytes/sizeof(long)); i++ )
if ( (rc = read_ulong(ea.mem.seg, ea.mem.off + i*sizeof(long),
&old[i], sizeof(long), ctxt, ops)) != 0 )
@@ -4151,10 +4152,13 @@
else
{
/* Expected == actual: attempt atomic cmpxchg and set ZF. */
- if ( (rc = ops->cmpxchg(ea.mem.seg, ea.mem.off, old,
- new, op_bytes, ctxt)) != 0 )
- goto done;
- _regs.eflags |= EFLG_ZF;
+ rc = ops->cmpxchg(ea.mem.seg, ea.mem.off, old, new, op_bytes,
ctxt);
+ if ( rc == 0 )
+ _regs.eflags |= EFLG_ZF;
+ else if ( rc == X86EMUL_CMPXCHG_FAILED )
+ goto cmpxchg_failed;
+ else
+ goto done;
}
break;
}
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|