Date: Sun, 30 Jan 2011 12:33:20 +0200 From: Kostik Belousov <kostikbel@gmail.com> To: Juergen Lock <nox@jelal.kn-bremen.de> Cc: freebsd-hackers@freebsd.org, freebsd-emulation@freebsd.org Subject: Re: Can vm_mmap()/vm_map_remove() be called with giant held? (linuxolator dvb patches) Message-ID: <20110130103320.GO2518@deviant.kiev.zoral.com.ua> In-Reply-To: <20110129235448.GA15788@triton8.kn-bremen.de> References: <20110129201000.GA10774@triton8.kn-bremen.de> <20110129205105.GI2518@deviant.kiev.zoral.com.ua> <20110129235448.GA15788@triton8.kn-bremen.de>
next in thread | previous in thread | raw e-mail | index | archive | help
--5XswgcLA9HHGPOFO
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Sun, Jan 30, 2011 at 12:54:48AM +0100, Juergen Lock wrote:
> On Sat, Jan 29, 2011 at 10:51:05PM +0200, Kostik Belousov wrote:
> > On Sat, Jan 29, 2011 at 09:10:00PM +0100, Juergen Lock wrote:
> > > Hi!
> > >=20
> > > I was kinda hoping to be able to post a correct patch in public but
> > > getting an answer to ${Subject} seems to be more difficult than I
> > > thought... :) So, does anyone here know? copyout_map() and
> > You do not need Giant locked for vm_map* functions.
> >=20
> The question was more do I need to drop it first before calling them...
No, you do not need to drop Giant.
[snip]
> > > + error =3D ENOMEM;
> > > + goto out2;
> > > + }
> > > + error =3D copyin((void *) vps.props, l_vp, l_propsiz);
> > > + if (error)
> > > + goto out2;
> > > + for (i =3D vps.num, l_p =3D l_vp, p =3D vp; i--; ++l_p, ++p)
> > > + linux_to_bsd_dtv_property(l_p, p);
> > > +
> > > + error =3D copyout_map(td, &uvp, propsiz);
> > > + if (error)
> > > + goto out2;
> > > + copyout(vp, (void *) uvp, propsiz);
> > > +
> > > + if ((error =3D fget(td, args->fd, &fp)) !=3D 0) {
> > > + (void) copyout_unmap(td, uvp, propsiz);
> > > + goto out2;
> > > + }
> > > + vps.props =3D (void *) uvp;
> > > + if ((args->cmd & 0xffff) =3D=3D LINUX_FE_SET_PROPERTY)
> > > + error =3D fo_ioctl(fp, FE_SET_PROPERTY, &vps, td->td_ucred, td);
> > > + else
> > > + error =3D fo_ioctl(fp, FE_GET_PROPERTY, &vps, td->td_ucred, td);
> > > + if (error) {
> > > + (void) copyout_unmap(td, uvp, propsiz);
> > > + goto out;
> > > + }
> > > + error =3D copyin((void *) uvp, vp, propsiz);
> > > + (void) copyout_unmap(td, uvp, propsiz);
> > No need in space between cast and expression. Bigger issue is that I
> > do not understand this fragment. You do copyout_map(), and then
> > immediately call copyout_unmap() for the address range returned
> > by copyout_map(), or am I missing something ?
> >=20
> The vm allocated by copyout_map() is only needed for the fo_ioctl()
> call because the struct passed to FE_[GS]ET_PROPERTY describes an
> array that the drivers then copyin() and (possibly) copyout()
> themselves. So that array needs to be translated from/to the 32bit
> Linux version to (possibly) 64bit on the host (linux_to_bsd_dtv_property),
> and the 64bit version is larger so it doesn't fit in the original
> place in the userland vm.
And am I right that the drivers can only take this array from the usermode ?
How is the compatibility for 32/64 bit mode is handled by native
FE_SET_PROPERTY handlers ?
I could only say that the hack is atrocious. Might be, you indeed have
no choice there.
>=20
> (One optimization here would be to only do this when the sizes
> differ i.e. in the 32bit-on-64 case...)
>=20
> Thanx!
> Juergen
--5XswgcLA9HHGPOFO
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (FreeBSD)
iEYEARECAAYFAk1FPnAACgkQC3+MBN1Mb4gfFwCg84gjU/HrRn1o1TBh/nVWiNfb
BBsAn3UlNnalAXADys/Lfz0/Qmkwtxdv
=wnoG
-----END PGP SIGNATURE-----
--5XswgcLA9HHGPOFO--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20110130103320.GO2518>