[PATCH 0/5] Set hstatus.GVA for traps going to HS-mode

[Andrew Jones]

On Thu, Aug 04, 2022 at 11:15:50AM +0800, Vivian Wang wrote:
> The privileged spec specifies that on a trap to HS-mode, hstatus.GVA
> should be set to 1 if stval is written with a guest virtual address, and
> to 0 otherwise.
> 
> Implement this by adding a field to sbi_trap_info called gva, which
> indicates if tval is a guest virtual address, and set it appropriately.
> When redirecting traps to HS-mode, set hstatus.GVA to trap->gva.
> 
> Patch 4 is a bugfix. The hypervisor CSRs should be set for all traps to
> HS-mode, not just those coming from VS/VU-mode. Since it is closely
> related to this series and makes the structure for patch 5 clearer, I
> have included it in this series.
> 
> Changes since RFC:
> 
> - Fixed names used in sign-off lines
> - Added more detailed commit messages
> - Separated out the hypervisor CSR bugfix into a separate patch
> 
> Vivian Wang (5):
>   include: Add mstatus[h].GVA encodings
>   lib: sbi_trap: Save mstatus[h].GVA in trap->gva
>   lib: sbi: Set gva when creating sbi_trap_info
>   lib: sbi_trap: Set hypervisor CSRs for HS-mode
>   lib: sbi_trap: Set hstatus.GVA when going to HS-mode
> 
>  include/sbi/riscv_encoding.h  |  4 ++++
>  include/sbi/sbi_trap.h        |  6 +++++-
>  lib/sbi/sbi_expected_trap.S   | 17 +++++++++++++++--
>  lib/sbi/sbi_illegal_insn.c    |  1 +
>  lib/sbi/sbi_misaligned_ldst.c |  2 ++
>  lib/sbi/sbi_trap.c            | 21 ++++++++++++++++-----
>  6 files changed, 43 insertions(+), 8 deletions(-)
>

FYI,

At least on my end, it looks like patches 2-5 are in-reply-to patch 1
instead of the cover-letter, which causes the thread to be a bit odd.

Thanks,
drew