OpenSBI Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Yu-Chien Peter Lin <peter.lin@sifive.com>
To: Anup Patel <anup@brainfault.org>
Cc: opensbi@lists.infradead.org, pawandeep.oza@oss.qualcomm.com,
	rkrcmar@qti.qualcomm.com, raymond.mao@riscstar.com,
	robin.randhawa@sifive.com, samuel.holland@sifive.com,
	zong.li@sifive.com, greentime.hu@sifive.com
Subject: Re: [PATCH v5 4/9] lib: sbi_hart: lock mwid CSR for RoT immutability
Date: Wed, 18 Nov 2026 10:00:47 +0800	[thread overview]
Message-ID: <av0GzyFfzDvasVEW@plin-1878> (raw)
In-Reply-To: <CAAhSdy2=JhtXg_h==eiet7nJ-VMaq7VCMgmVuha-uwQKkVEmMw@mail.gmail.com>

Hi Anup,

On Mon, Nov 16, 2026 at 06:49:50PM +0530, Anup Patel wrote:
> On Tue, Nov 10, 2026 at 3:20 PM Yu-Chien Peter Lin <peter.lin@sifive.com> wrote:
> >
> > Lock the M-mode World ID (mwid) CSR during hart re-initialization
> > to enforce immutability of the WID established by the root-of-trust.
> >
> > OpenSBI does not assign the WID value itself; it only sets MWID_LOCK
> > to freeze the value established by prior RoT stage. The MWID_LOCK bit
> > at XLEN-1 is sticky and makes the CSR read-only until reset, enforcing
> > a temporal security boundary per the RISC-V Worlds specification.
> >
> > Signed-off-by: Yu-Chien Peter Lin <peter.lin@sifive.com>
> > ---
> > Changes v4->v5:
> > - Move the MWID lock out of sbi_hart_reinit() into smwid_hart_init()
> > - Drop RB-tags from Oza and Anup
> > ---
> >  lib/sbi/sbi_hart.c | 12 ++++++++++++
> >  1 file changed, 12 insertions(+)
> >
> > diff --git a/lib/sbi/sbi_hart.c b/lib/sbi/sbi_hart.c
> > index d86abf07..a721bef4 100644
> > --- a/lib/sbi/sbi_hart.c
> > +++ b/lib/sbi/sbi_hart.c
> > @@ -469,6 +469,16 @@ static int hart_mhpm_get_allowed_bits(void)
> >         return num_bits;
> >  }
> >
> > +static void smwid_hart_init(struct sbi_scratch *scratch)
> > +{
> > +       /*
> > +        * Assume MWID is restored by root-of-trust M-mode in previous
> > +        * stage. Lock mwid so RoT-defined WID remains immutable.
> > +        */
> > +       if (sbi_hart_has_extension(scratch, SBI_HART_EXT_SMWID))
> > +               csr_set(CSR_MWID, MWID_LOCK);
> 
> If pmwid is available then we should set pmwid in MWID CSR
> before locking it. This will help platforms where OpenSBI boots
> first without any prior booting stage.
> 
> What do you think ?

Thanks for the suggestion. I'd prefer lock-only, because writing pmwid before the
lock is either a no-op or can undo an earlier stage's decision.

The spec says "On-reset, the mwid CSR is expected to show the platform-defined pmwid
value", so if OpenSBI is the first stage, mwid already equals pmwid and the write
changes nothing.

It only changes something when mwid != pmwid, which means an earlier stage has already
programmed mwid. Example with nWorlds=16, trusted WID = 15:

1. Out of reset, mwid = pmwid = 15. The RoT runs in the trusted WID.
2. The RoT configures the wgCheckers so that WID 3 can only access OpenSBI's 
   RAM/devices and cannot touch the wgChecker configuration registers.
   - The RoT writes mwid = 3 to demote runtime M-mode, but leaves the lock
     bit unset and hands over to OpenSBI.
4. OpenSBI writes pmwid (15) to mwid, then sets the lock bit.
5. OpenSBI is now back in the trusted WID, and locked there until the next reset.
   It can reconfigure the wgCheckers and the RoT's demotion is undone.

Given the case above, I think we should keep mwid untouched.

Thanks,
Peter Lin

> 
> > +}
> > +
> >  static int smrnmi_hart_init(struct sbi_scratch *scratch)
> >  {
> >         const struct sbi_platform *plat;
> > @@ -721,6 +731,8 @@ int sbi_hart_reinit(struct sbi_scratch *scratch)
> >  {
> >         int rc;
> >
> > +       smwid_hart_init(scratch);
> > +
> >         rc = smrnmi_hart_init(scratch);
> >         if (rc)
> >                 return rc;
> > --
> > 2.43.7
> >
> 
> Regards,
> Anup

-- 
opensbi mailing list
opensbi@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi

  reply	other threads:[~2026-11-18  2:01 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-10  9:50 [PATCH v5 0/9] Add RISC-V Worlds ISA support to OpenSBI Yu-Chien Peter Lin
2026-11-10  9:50 ` [PATCH v5 1/9] lib: sbi_hart: add RISC-V Worlds ISA extensions Yu-Chien Peter Lin
2026-11-10 18:48   ` Pawandeep Oza
2026-11-10  9:50 ` [PATCH v5 2/9] lib: utils: fdt_helper: parse RISC-V Worlds per-hart WID properties Yu-Chien Peter Lin
2026-11-10 18:48   ` Pawandeep Oza
2026-11-10  9:50 ` [PATCH v5 3/9] lib: fdt_domain: parse domain " Yu-Chien Peter Lin
2026-11-10 18:48   ` Pawandeep Oza
2026-11-16 13:14   ` Anup Patel
2026-11-10  9:50 ` [PATCH v5 4/9] lib: sbi_hart: lock mwid CSR for RoT immutability Yu-Chien Peter Lin
2026-11-10 18:49   ` Pawandeep Oza
2026-11-16 13:19   ` Anup Patel
2026-11-18  2:00     ` Yu-Chien Peter Lin [this message]
2026-11-18  4:55       ` Yu-Chien Peter Lin
2026-11-10  9:50 ` [PATCH v5 5/9] lib: sbi_domain_context: propagate error from sbi_hart_protection_reconfigure Yu-Chien Peter Lin
2026-11-10  9:50 ` [PATCH v5 6/9] lib: sbi_hart_worlds: Implement WID hart protection backend Yu-Chien Peter Lin
2026-11-10 18:49   ` Pawandeep Oza
2026-11-16 13:21   ` Anup Patel
2026-11-10  9:50 ` [PATCH v5 7/9] include: sbi_types: add PRIx64 format macro Yu-Chien Peter Lin
2026-11-10 18:49   ` Pawandeep Oza
2026-11-10  9:50 ` [PATCH v5 8/9] lib: sbi: display World ID configuration at boot Yu-Chien Peter Lin
2026-11-10  9:50 ` [PATCH v5 9/9] docs: document WID DT properties Yu-Chien Peter Lin
2026-11-10 18:50   ` Pawandeep Oza
2026-11-16 13:32   ` Anup Patel

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=av0GzyFfzDvasVEW@plin-1878 \
    --to=peter.lin@sifive.com \
    --cc=anup@brainfault.org \
    --cc=greentime.hu@sifive.com \
    --cc=opensbi@lists.infradead.org \
    --cc=pawandeep.oza@oss.qualcomm.com \
    --cc=raymond.mao@riscstar.com \
    --cc=rkrcmar@qti.qualcomm.com \
    --cc=robin.randhawa@sifive.com \
    --cc=samuel.holland@sifive.com \
    --cc=zong.li@sifive.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox