fixed null segment validation (aka x86_64 regression bug)
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@1708 c046a42c-6fe2-441c-8c8c-71466251a162
This commit is contained in:
parent
d3e9db933f
commit
cd072e01d8
@ -1986,7 +1986,14 @@ static inline void validate_seg(int seg_reg, int cpl)
|
||||
{
|
||||
int dpl;
|
||||
uint32_t e2;
|
||||
|
||||
|
||||
/* XXX: on x86_64, we do not want to nullify FS and GS because
|
||||
they may still contain a valid base. I would be interested to
|
||||
know how a real x86_64 CPU behaves */
|
||||
if ((seg_reg == R_FS || seg_reg == R_GS) &&
|
||||
(env->segs[seg_reg].selector & 0xfffc) == 0)
|
||||
return;
|
||||
|
||||
e2 = env->segs[seg_reg].flags;
|
||||
dpl = (e2 >> DESC_DPL_SHIFT) & 3;
|
||||
if (!(e2 & DESC_CS_MASK) || !(e2 & DESC_C_MASK)) {
|
||||
|
Loading…
Reference in New Issue
Block a user