Messages in this thread | | | From | Bandan Das <> | Subject | Re: [PATCH 09/13] KVM: x86: save/load state on SMM switch | Date | Tue, 12 May 2015 19:56:15 -0400 |
| |
Paolo Bonzini <pbonzini@redhat.com> writes:
> On 06/05/2015 19:55, Bandan Das wrote: >> Valid point. But I would say cryptic names will always be better then >> hex offsets. It's a one time pain in the neck to get the defines >> right. >> >> I found this header - >> ftp://ftp.xskernel.org/soft/linux-src/bochs-20080511/cpu/smm.h >> >> Maybe we can pick it up ? I can sanity check the list and >> make sure the offsets are correct if you prefer. > > Thanks, I like that header. :) No need to sanity check yourself. > > I'll have to adjust to define both 32-bit and 64-bit offsets. Ok to use > just SMRAM32_OFS_xyz and SMRAM64_OFS_xyz? No need for LO32 and HI32, in > particular. Agreed, that sounds like the right thing to do.
> The one problem is that I am doing math on the offsets in some places. > It would suck to save/restore 16 registers manually. Any suggestions? > I will probably send v1 for review without the offsets.
I think it's ok just to replace the offset number with a name. If you are concerned about readability, we could probably wrap the "for loops" in the math calculations with a more serious sounding macro eg. "#define foreach_smm_offset(val, start_offset, jump, count)" :)
> Paolo
| |