Re: [PATCH 12/14] bsd-user/sysarch: Provide a per-arch framework for sysarch syscall

From: Warner Losh <imp@bsdimp.com>
To: "Philippe Mathieu-Daudé" <f4bug@amsat.org>
Cc: Kyle Evans <kevans@freebsd.org>,
	Richard Henderson <richard.henderson@linaro.org>,
	QEMU Developers <qemu-devel@nongnu.org>,
	Stacey Son <sson@freebsd.org>
Subject: Re: [PATCH 12/14] bsd-user/sysarch: Provide a per-arch framework for sysarch syscall
Date: Sun, 26 Sep 2021 11:21:18 -0600	[thread overview]
Message-ID: <CANCZdfprC16ezJQCWJmYEApX6eym9nxSOqAtBAGr+cziS4r2qw@mail.gmail.com> (raw)
In-Reply-To: <64e8f0f5-1239-fc04-1c58-189421212f2c@amsat.org>

[-- Attachment #1: Type: text/plain, Size: 3083 bytes --]

On Sat, Sep 25, 2021 at 4:46 AM Philippe Mathieu-Daudé <f4bug@amsat.org>
wrote:

> Hi Warner,
>
> On 9/22/21 08:14, Warner Losh wrote:
> > Add the missing glue to pull in do_freebsd_sysarch to call
> > do_freebsd_arch_sysarch. Put it in os-sys.c, which will be used for
> > sysctl and sysarch system calls because they are mostly arch specific.
> >
> > Signed-off-by: Stacey Son <sson@FreeBSD.org>
> > Signed-off-by: Warner Losh <imp@bsdimp.com>
> > ---
> >   bsd-user/freebsd/meson.build |  3 +++
> >   bsd-user/freebsd/os-sys.c    | 28 ++++++++++++++++++++++++++++
> >   bsd-user/meson.build         |  6 ++++++
> >   bsd-user/qemu.h              |  3 +++
> >   bsd-user/syscall.c           |  7 -------
> >   5 files changed, 40 insertions(+), 7 deletions(-)
> >   create mode 100644 bsd-user/freebsd/meson.build
> >   create mode 100644 bsd-user/freebsd/os-sys.c
>
> > +/* sysarch() is architecture dependent. */
> > +abi_long do_freebsd_sysarch(void *cpu_env, abi_long arg1, abi_long arg2)
> > +{
> > +
> > +    return do_freebsd_arch_sysarch(cpu_env, arg1, arg2);
> > +}
> > diff --git a/bsd-user/meson.build b/bsd-user/meson.build
> > index 0369549340..561913de05 100644
> > --- a/bsd-user/meson.build
> > +++ b/bsd-user/meson.build
> > @@ -8,3 +8,9 @@ bsd_user_ss.add(files(
> >     'syscall.c',
> >     'uaccess.c',
> >   ))
> > +
> > +# Pull in the OS-specific build glue, if any
> > +if fs.exists(targetos)
> > +   subdir(targetos)
>
> I am a bit confused here, we have an optional implementation ...
>
> > +endif
> > +
> > diff --git a/bsd-user/qemu.h b/bsd-user/qemu.h
> > index 4ee57b91f0..3dde381d5d 100644
> > --- a/bsd-user/qemu.h
> > +++ b/bsd-user/qemu.h
> > @@ -239,6 +239,9 @@ extern unsigned long target_sgrowsiz;
> >   abi_long get_errno(abi_long ret);
> >   int is_error(abi_long ret);
> >
> > +/* os-sys.c */
> > +abi_long do_freebsd_sysarch(void *cpu_env, abi_long arg1, abi_long
> arg2);
>
> ... that is declared as non-optional.
>
> In fact it is called by do_freebsd_syscall(), not restricted to
> TARGET_I386.
>
> This shouldn't be (meson) optional IMO.
>

I made it optional because we descend into this directory even
for linux targets when building linux-user on at least ubuntu (and
I think all linux targets).

Your comment suggests that I need to have a different meson fix
for this situation...

Warner

> > +
> >   /* user access */
> >
> >   #define VERIFY_READ  PAGE_READ
> > diff --git a/bsd-user/syscall.c b/bsd-user/syscall.c
> > index 9bc72501b2..9f51563abd 100644
> > --- a/bsd-user/syscall.c
> > +++ b/bsd-user/syscall.c
> > @@ -88,13 +88,6 @@ static abi_long do_obreak(abi_ulong new_brk)
> >       return 0;
> >   }
> >
> > -#if defined(TARGET_I386)
> > -static abi_long do_freebsd_sysarch(CPUX86State *env, int op, abi_ulong
> parms)
> > -{
> > -    do_freebsd_arch_sysarch(env, op, parms);
> > -}
> > -#endif
> > -
> >   #ifdef __FreeBSD__
> >   /*
> >    * XXX this uses the undocumented oidfmt interface to find the kind of
> >
>

[-- Attachment #2: Type: text/html, Size: 4152 bytes --]