RTEMS trap handling versus RISCV mtvec BASE and MODE

[Hesham Almatary]

Hello Jens,

On Tue, 1 Jun 2021 at 12:05, Schweikhardt, Jens (TSPCE3-TL4)
<Jens.Schweikhardt at tesat.de> wrote:
>
> hello world,
>
> I'm scratching my head over what may turn out to be a glitch in
> Volume 2, Privileged Spec v. 20190608 RISC-V
> https://github.com/riscv/riscv-isa-manual/releases/download/Ratified-IMFDQC-and-Priv-v1.11/riscv-privileged-20190608.pdf
>
> In particular, the way RTEMS for RISCV sets the machine trap vector
> in src/rtems/bsps/riscv/shared/start/start.S with
>
>     LADDR  t0, _RISCV_Exception_handler
>     csrw   mtvec, t0
>
> i.e. loads the 32 bit address _RISCV_Exception_handler into mtvec.
> This seems to contradict the layout oft he mtvec register documented in 3.1.7 as
>
> mtvec:   30 bits of BASE, 2 bits of MODE.
>
That might be where the confusion is. BASE is 32-bit, and as per the
same section "must always be aligned on a 4-byte boundary." So, you
always end up with BASE[1:0] bits = 0 (to be used for MODE bits).
Thus, you don't do any shifting of addresses.

>
> in conjunction with Table 3.5, MODE 0, "All exceptions set pc to BASE".
>
> This would set pc to _RISCV_Exception_handler *divided by 4* and probably crash.
>
> So either RTEMS writes the wrong value to mtvec, or the RISC-V docs should say "All exceptions set pc to BASE x 4".
> Am I missing something?
>
You might be confusing taking BASE[XLEN-1:2] (see Figure 3.8) with
shifting an address by 2. RTEMS aligns _RISCV_Exception_handler to 4
bytes as per the specification, so you'd always end up with an address
with zeros in the least significant 2 bits. So, BASE[1:0] will always
be zero. When you write that to mtvec, you'll effectively set the MODE
to 0 (Direct), and BASE[XLEN-1:2] will be placed as is in mtvec.

I hope that clarifies things up a bit.

>
> Jens
>
> ________________________________
>
> Tesat-Spacecom GmbH & Co. KG
> Sitz: Backnang; Registergericht: Amtsgericht Stuttgart HRA 270977
> Persoenlich haftender Gesellschafter: Tesat-Spacecom Geschaeftsfuehrungs GmbH;
> Sitz: Backnang; Registergericht: Amtsgericht Stuttgart HRB 271658;
> Geschaeftsfuehrung: Dr. Marc Steckling, Kerstin Basche, Ralf Zimmermann
>
> [banner]
> _______________________________________________
> users mailing list
> users at rtems.org
> http://lists.rtems.org/mailman/listinfo/users



-- 
Hesham