Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230711062202.3542367-1-CoelacanthusHex@gmail.com>
 <CAJF2gTRgXbEfQWiWtZtVNBDhpY2miG_ByHrVZF2fbfu60hCHcQ@mail.gmail.com> <37a22693-e33c-f64a-e3d7-febed8e9f195@gmail.com>
In-Reply-To: <37a22693-e33c-f64a-e3d7-febed8e9f195@gmail.com>
From:   Guo Ren <guoren@kernel.org>
Date:   Thu, 13 Jul 2023 12:59:40 -0400
Message-ID: <CAJF2gTTmKPR-jwEpKFT4qzNqZLQV--RMzSH5LA-EPw8eSVi56g@mail.gmail.com>
Subject: Re: [PATCH] riscv: entry: set a0 prior to syscall_handler
To:     Celeste Liu <coelacanthushex@gmail.com>
Cc:     Palmer Dabbelt <palmer@rivosinc.com>,
        Paul Walmsley <paul.walmsley@sifive.com>,
        Albert Ou <aou@eecs.berkeley.edu>,
        =?UTF-8?B?QmrDtnJuIFTDtnBlbA==?= <bjorn@rivosinc.com>,
        Conor Dooley <conor.dooley@microchip.com>,
        linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org,
        Felix Yan <felixonmars@archlinux.org>,
        Ruizhe Pan <c141028@gmail.com>,
        Shiqi Zhang <shiqi@isrc.iscas.ac.cn>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Jul 13, 2023 at 11:28=E2=80=AFAM Celeste Liu <coelacanthushex@gmail=
.com> wrote:
>
>
> On 2023/7/13 08:00, Guo Ren wrote:
> > On Tue, Jul 11, 2023 at 2:22=E2=80=AFAM Celeste Liu <coelacanthushex@gm=
ail.com> wrote:
> >>
> >> When we test seccomp with 6.4 kernel, we found errno has wrong value.
> >> If we deny NETLINK_AUDIT with EAFNOSUPPORT, after f0bddf50586d, we wil=
l
> >> get ENOSYS. We got same result with 9c2598d43510 ("riscv: entry: Save =
a0
> >> prior syscall_enter_from_user_mode()").
> >>
> >> Compared with x86 and loongarch's implementation of this part of the
> >> function, we think that regs->a0 =3D -ENOSYS should be advanced before
> >> syscall_handler to fix this problem. We have written the following pat=
ch,
> >> which can fix this problem after testing. But we don't know enough abo=
ut
> >> this part of the code to explain the root cause. Hope someone can find
> >> a reasonable explanation. And we'd like to reword this commit message
> >> according to the explanation in v2
> >>
> >> Fixes: f0bddf50586d ("riscv: entry: Convert to generic entry")
> >> Reported-by: Felix Yan <felixonmars@archlinux.org>
> >> Co-developed-by: Ruizhe Pan <c141028@gmail.com>
> >> Signed-off-by: Ruizhe Pan <c141028@gmail.com>
> >> Co-developed-by: Shiqi Zhang <shiqi@isrc.iscas.ac.cn>
> >> Signed-off-by: Shiqi Zhang <shiqi@isrc.iscas.ac.cn>
> >> Signed-off-by: Celeste Liu <CoelacanthusHex@gmail.com>
> >> Tested-by: Felix Yan <felixonmars@archlinux.org>
> >> ---
> >>  arch/riscv/kernel/traps.c | 3 +--
> >>  1 file changed, 1 insertion(+), 2 deletions(-)
> >>
> >> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
> >> index f910dfccbf5d2..ccadb5ffd063c 100644
> >> --- a/arch/riscv/kernel/traps.c
> >> +++ b/arch/riscv/kernel/traps.c
> >> @@ -301,6 +301,7 @@ asmlinkage __visible __trap_section void do_trap_e=
call_u(struct pt_regs *regs)
> >>
> >>                 regs->epc +=3D 4;
> >>                 regs->orig_a0 =3D regs->a0;
> >> +               regs->a0 =3D -ENOSYS;
> > Oh, no. You destroyed the a0 for syscall_handler, right? It's not
> > reasonable. Let's see which syscall_handler needs a0=3D-ENOSYS.
>
> syscall_handler always use orig_a0, not a0.
> And I have a mistake in original email, corret one is
> syscall_enter_from_user_mode not syscall_handler.
I misunderstood. Yes, a0 would be replaced by orig_a0:
syscall_enter_from_user_mode_work ->  syscall_rollback

If the syscall was denied by syscall_enter_from_user_mode(), the
return number is forced to be -ENOSYS. Maybe regs->a0 has already been
updated by SYSCALL_WORK_SECCOMP. eg:

__seccomp_filter() {
...
        case SECCOMP_RET_TRAP:
                /* Show the handler the original registers. */
                syscall_rollback(current, current_pt_regs());
                /* Let the filter pass back 16 bits of data. */
                force_sig_seccomp(this_syscall, data, false);
                goto skip;

>
> > Could you give out more detail on your test case?
> >
> >>
> >>                 riscv_v_vstate_discard(regs);
> >>
> >> @@ -308,8 +309,6 @@ asmlinkage __visible __trap_section void do_trap_e=
call_u(struct pt_regs *regs)
> >>
> >>                 if (syscall < NR_syscalls)
> >>                         syscall_handler(regs, syscall);
> >> -               else
> >> -                       regs->a0 =3D -ENOSYS;
> >>
> >>                 syscall_exit_to_user_mode(regs);
> >>         } else {
> >> --
> >> 2.41.0
> >>
> >
> >
>


--=20
Best Regards
 Guo Ren