On Fri, 2017-09-29 at 13:56 +0200, Borislav Petkov wrote: > On Thu, Sep 28, 2017 at 11:06:42PM -0700, Ricardo Neri wrote: > > > > I agree. In fact, insn_get_seg_base() does not need insn at all. All it > > needs is > > a INAT_SEG_REG_* index. This would make things clear. UMIP (and callers that > > need to copy_from_user code can do insn_get_seg_base(regs, INAT_SEG_REG_CS). > > No > > insn needed. > Yap. > > > > > In fact, it is only the insn_get_addr_ref_xx() family of functions that does > I think you mean get_addr_ref_xx() here.
Yes. > > > > > Do you think the pseudocode above addresses your concerns? > > > > *insn_get_seg_base() will take a INAT_SEG_REG_* index > > *insn_get_ref_xx() receives an initialized insn that can check for NULL > > value. > > *a reworked resolve_seg_reg_idx will clearly check if it can use segment > > override prefixes and obtain them. If not, it will use default values. > Makes sense, but send me the final version to take a look at it too. I just sent a v9 with all these changes as they impacted several patches in the series. Thanks and BR, Ricardo
