> -----Original Message----- > From: Alistair Francis <alistai...@gmail.com>
> Sent: Tuesday, September 19, 2023 12:42 PM > To: Alvin Che-Chia Chang(張哲嘉) <alvi...@andestech.com> > Cc: qemu-ri...@nongnu.org; qemu-devel@nongnu.org; > alistair.fran...@wdc.com; ajo...@ventanamicro.com > Subject: Re: [PATCH v3] target/riscv: update checks on writing pmpcfg for > Smepmp to version 1.0 > > On Fri, Sep 15, 2023 at 6:32 PM Alvin Chang <alvi...@andestech.com> wrote: > > > > Current checks on writing pmpcfg for Smepmp follows Smepmp version > > 0.9.1. However, Smepmp specification has already been ratified, and > > there are some differences between version 0.9.1 and 1.0. In this > > commit we update the checks of writing pmpcfg to follow Smepmp version > > 1.0. > > > > When mseccfg.MML is set, the constraints to modify PMP rules are: > > 1. Locked rules cannot be removed or modified until a PMP reset, unless > > mseccfg.RLB is set. > > 2. From Smepmp specification version 1.0, chapter 2 section 4b: > > Adding a rule with executable privileges that either is M-mode-only > > or a locked Shared-Region is not possible and such pmpcfg writes are > > ignored, leaving pmpcfg unchanged. > > > > The commit transfers the value of pmpcfg into the index of the Smepmp > > truth table, and checks the rules by aforementioned specification > > changes. > > > > Signed-off-by: Alvin Chang <alvi...@andestech.com> > > --- > > Changes from v2: Adopt switch case ranges and numerical order. > > > > Changes from v1: Convert ePMP over to Smepmp. > > Did this part get lost? > > Alistair > Sorry, do you mean that the term "ePMP" should be changed to "Smepmp" in source code? Alvin Chang > > > > target/riscv/pmp.c | 40 ++++++++++++++++++++++++++++++++-------- > > 1 file changed, 32 insertions(+), 8 deletions(-) > > > > diff --git a/target/riscv/pmp.c b/target/riscv/pmp.c index > > a08cd95658..b144932b3b 100644 > > --- a/target/riscv/pmp.c > > +++ b/target/riscv/pmp.c > > @@ -99,16 +99,40 @@ static void pmp_write_cfg(CPURISCVState *env, > uint32_t pmp_index, uint8_t val) > > locked = false; > > } > > > > - /* mseccfg.MML is set */ > > - if (MSECCFG_MML_ISSET(env)) { > > - /* not adding execute bit */ > > - if ((val & PMP_LOCK) != 0 && (val & PMP_EXEC) != > PMP_EXEC) { > > + /* > > + * mseccfg.MML is set. Locked rules cannot be removed or > modified > > + * until a PMP reset. Besides, from Smepmp specification > version 1.0 > > + * , chapter 2 section 4b says: > > + * Adding a rule with executable privileges that either is > > + * M-mode-only or a locked Shared-Region is not possible > and such > > + * pmpcfg writes are ignored, leaving pmpcfg unchanged. > > + */ > > + if (MSECCFG_MML_ISSET(env) && !pmp_is_locked(env, > pmp_index)) { > > + /* > > + * Convert the PMP permissions to match the truth > table in the > > + * ePMP spec. > > + */ > > + const uint8_t epmp_operation = > > + ((val & PMP_LOCK) >> 4) | ((val & PMP_READ) << > 2) | > > + (val & PMP_WRITE) | ((val & PMP_EXEC) >> 2); > > + > > + switch (epmp_operation) { > > + case 0 ... 8: > > locked = false; > > - } > > - /* shared region and not adding X bit */ > > - if ((val & PMP_LOCK) != PMP_LOCK && > > - (val & 0x7) != (PMP_WRITE | PMP_EXEC)) { > > + break; > > + case 9 ... 11: > > + break; > > + case 12: > > + locked = false; > > + break; > > + case 13: > > + break; > > + case 14: > > + case 15: > > locked = false; > > + break; > > + default: > > + g_assert_not_reached(); > > } > > } > > } else { > > -- > > 2.34.1 > > > >