On Sun, Aug 12, 2018 at 6:17 AM Paolo Bonzini <pbonz...@redhat.com> wrote:
> On 12/08/2018 05:07, Andrew Oates via Qemu-devel wrote:
> > Currently call gates are always treated as 32-bit gates. In IA-32e mode
> > (either compatibility or 64-bit submode), system segment descriptors are
> > always 64-bit. Treating them as 32-bit has the expected unfortunate
> > effect: only the lower 32 bits of the offset are loaded, the stack
> > pointer is truncated, a bad new stack pointer is loaded from the TSS (if
> > switching privilege levels), etc.
> >
> > This change adds support for 64-bit call gate to the lcall instruction.
> > Additionally, there should be a check for non-canonical stack pointers,
> > but I've omitted that since there doesn't seem to be checks for
> > non-canonical addresses in this code elsewhere.
>
> Thanks. It's also missing a check for task gates in 64-bit mode and,
> since we are at it, we should also implement ljmp to call gates.
> Something like this (untested):
>
Yeah, I was going to do ljmp in a followup. I'll go ahead and add it to
this patch.
>
> diff --git a/target/i386/seg_helper.c b/target/i386/seg_helper.c
> index d3b31f3c1b..0b2efe2cae 100644
> --- a/target/i386/seg_helper.c
> +++ b/target/i386/seg_helper.c
> @@ -1645,6 +1645,13 @@ void helper_ljmp_protected(CPUX86State *env, int
> new_cs, target_ulong new_eip,
> rpl = new_cs & 3;
> cpl = env->hflags & HF_CPL_MASK;
> type = (e2 >> DESC_TYPE_SHIFT) & 0xf;
> +#ifdef TARGET_X86_64
> + if (env->efer & MSR_EFER_LMA) {
> + if (type != 12) {
> + raise_exception_err_ra(env, EXCP0D_GPF, gate_cs & 0xfffc,
> GETPC());
> + }
> + }
> +#endif
> switch (type) {
> case 1: /* 286 TSS */
> case 9: /* 386 TSS */
> @@ -1666,6 +1673,21 @@ void helper_ljmp_protected(CPUX86State *env, int
> new_cs, target_ulong new_eip,
> new_eip = (e1 & 0xffff);
> if (type == 12) {
> new_eip |= (e2 & 0xffff0000);
> +#ifdef TARGET_X86_64
> + if (env->efer & MSR_EFER_LMA) {
> + /* load the upper 8 bytes of the 64-bit call gate */
> + if (load_segment_ra(env, &e1, &e2, new_cs + 8,
> GETPC())) {
> + raise_exception_err_ra(env, EXCP0D_GPF, new_cs &
> 0xfffc,
> + GETPC());
> + }
> + type = (e2 >> DESC_TYPE_SHIFT) & 0x1f;
> + if (type != 0) {
> + raise_exception_err_ra(env, EXCP0D_GPF, gate_cs &
> 0xfffc,
> + GETPC());
> + }
> + new_eip |= ((target_ulong)e1) << 32;
> + }
> +#endif
> }
> if (load_segment_ra(env, &e1, &e2, gate_cs, GETPC()) != 0) {
> raise_exception_err_ra(env, EXCP0D_GPF, gate_cs & 0xfffc,
> GETPC());
> @@ -1812,6 +1834,13 @@ void helper_lcall_protected(CPUX86State *env, int
> new_cs, target_ulong new_eip,
> type = (e2 >> DESC_TYPE_SHIFT) & 0x1f;
> dpl = (e2 >> DESC_DPL_SHIFT) & 3;
> rpl = new_cs & 3;
> +#ifdef TARGET_X86_64
> + if (env->efer & MSR_EFER_LMA) {
> + if (type != 12) {
> + raise_exception_err_ra(env, EXCP0D_GPF, gate_cs & 0xfffc,
> GETPC());
> + }
> + }
> +#endif
> switch (type) {
> case 1: /* available 286 TSS */
> case 9: /* available 386 TSS */
>
>
> Out of curiosity, what OS is using 64-bit call gates? :)
>
As far as I know, no important ones :) Just something I've been hacking
around with in my spare time. The rational thing to do would have been to
just stop using 64-bit call gates, but I figured it would be more fun to
fix it for everyone.
>
> Paolo
>
> > I've left the raise_exception_err_ra lines unwapped at 80 columns to
> > match the style in the rest of the file.
> >
> > Signed-off-by: Andrew Oates <aoa...@google.com>
> > ---
> > target/i386/seg_helper.c | 145 ++++++++++++++++++++++++++++-----------
> > 1 file changed, 106 insertions(+), 39 deletions(-)
>
>
