AMD Unsupported hardware watermark

Issues Related to Hardware (Sound, Video, Printers, Scanners, Webcams, Touchpads, Keyboards, Sensors, Drives, GRUB, etc.)

Moderator: Moderators

Land Owner
Posts: 3596
Joined: Thu Oct 13, 2005 23:16
Location: Italy

Re: AMD Unsupported hardware watermark

Post by lxnay » Wed Jun 10, 2009 19:45

Eventually someone pushed me this link. I am reverting the entropy version to 8.593-r10 :) thanks
Just in time for 4.2
Join us on IRC ( #sabayon or WebChat)
Submit bugs to our Bug Tracker
Follow me on Twitter
Add me on Facebook
Add me on Google+

User avatar
Sagely Hen
Posts: 8091
Joined: Sat Mar 10, 2007 5:40
Location: United Kingdom

Re: AMD Unsupported hardware watermark

Post by Fitzcarraldo » Tue Jun 23, 2009 12:13

Just for the record in case it causes any unforseen effects, today (23 June 2009) lxnay changed the contents of the ati-drivers-8.593-r10 ebuild. Therefore, although the ebuild in the SL overlay repository is called ati-drivers-8.593-r10.ebuild, its contents are not the same as they were yesterday. The corresponding package for the 2.6.29 kernel in the main SL Entropy repository is called ati-drivers-8.593-r10#2.6.29-sabayon and it is presumably built from the new version of the ati-drivers-8.593-r10 ebuild.

The contents of ati-drivers-8.593-r10.ebuild before lxnay made the change were as follows:

Code: Select all

# Copyright 1999-2009 Gentoo Foundation
# Distributed under the terms of the GNU General Public License v2
# $Header: /var/cvsroot/gentoo-x86/x11-drivers/ati-drivers/ati-drivers-8.582.ebuild,v 1.1 2009/02/26 17:14:48 je_fro Exp $

IUSE="acpi debug distribution"

inherit eutils multilib linux-mod toolchain-funcs versionator

DESCRIPTION="Ati precompiled drivers for recent chipsets"

LICENSE="AMD GPL-2 QPL-1.0 as-is"
KEYWORDS="~amd64 ~x86"

# The portage dep is for COLON_SEPARATED support in env-update.
# The eselect dep (>=1.0.9) is for COLON_SEPARATED in eselect env update.
   amd64? ( app-emulation/emul-linux-x86-xlibs )
   acpi? (





pkg_setup() {

   # Define module dir.

   #check kernel and sets up KV_OBJ
   BUILD_PARAMS="GCC_VER_MAJ=$(gcc-major-version) KVER=${KV_FULL} KDIR=${KV_DIR}"

   if ! kernel_is 2 6; then
      eerror "Need a 2.6 kernel to compile against!"
      die "Need a 2.6 kernel to compile against!"

   if kernel_is ge 2 6 25 && linux_chkconfig_present PREEMPT_RCU; then
      die "${P} is not compatible with RCU Preemption (bug #223281), please disable it"

   if kernel_is ge 2 6 26 && ! linux_chkconfig_present UNUSED_SYMBOLS; then
         ewarn "You have to Enable unused/obsolete exported symbols in Kernel hacking section of kernel config for fglrx to load"

   if kernel_is ge 2 6 24 && ! linux_chkconfig_present PCI_LEGACY; then
      eerror "${P} requires support for pci_find_slot."
      die "${P} requires support for pci_find_slot."

   if ! linux_chkconfig_present MTRR; then
      ewarn "You don't have MTRR support enabled, the direct rendering will not work."

   if linux_chkconfig_builtin DRM; then
      ewarn "You have DRM support enabled builtin, the direct rendering will not work."

   if ! linux_chkconfig_present AGP && \
      ! linux_chkconfig_present PCIEPORTBUS; then
      ewarn "You need AGP and/or PCI Express support for direct rendering to work."

   if linux_chkconfig_present PARAVIRT; then
      eerror "The current ati-drivers don't compile when having"
      eerror "paravirtualization active due to GPL symbol export"
      eerror "restrictions."
      eerror "Please disable it:"
      eerror "   CONFIG_PARAVIRT=n"
      eerror "in /usr/src/linux/.config or"
      eerror "   Processor type and features -->"
      eerror "      [ ] Paravirtualization support (EXPERIMENTAL)"
      eerror "in 'menuconfig'"
      die "CONFIG_PARAVIRT enabled"

   if ! linux_chkconfig_present MAGIC_SYSRQ; then
      eerror "You need MAGIC_SYSRQ enabled in order to build ati-drivers"
      die "CONFIG_MAGIC_SYSRQ disabled"

   # Only support xorg-server >=1.5

   if ! linux_chkconfig_present PCI_MSI; then
      eerror "You need PCI_MSI enabled in order to build ati-drivers"
      die "CONFIG_PCI_MSI disabled"

   # This is used like $(get_libdir) for paths in ati's package.
   if use amd64 ; then

src_unpack() {
   #Switching to a standard way to extract the files since otherwise no signature file
   #would be created
   local src="${DISTDIR}/${A}"
   sh "${src}" --extract "${S}"  2&>1 /dev/null
   cd "${S}"

        if kernel_is ge 2 6 29; then
                cd "${S}/common/lib/modules/fglrx/build_mod"
                epatch "${FILESDIR}"/8.600/${PN}-8.600-acpica-2.6.29.patch
                cd "${S}"

   epatch "${FILESDIR}"/${PV}/ati-drivers-xen-8.552.patch || die "epatch failed"

   # These are the userspace utilities that we also have source for.
   # We rebuild these later.
   rm \
      "${ARCH_DIR}"/usr/X11R6/bin/{fgl_glxgears,fglrx_xgamma} \
      "${ARCH_DIR}"/usr/X11R6/${PKG_LIBDIR}/libfglrx_gamma* \
      || die "bin rm failed"

   if use debug; then
      # Enable debug mode in the Source Code.
      sed -i '/^#define DRM_DEBUG_CODE/s/0/1/' \
         "${MODULE_DIR}/firegl_public.c" \
         || die "Failed to enable debug output."

   if use acpi; then
      sed -i \
         -e "s:/var/lib/xdm/authdir/authfiles/:/var/run/xauth/:" \
         -e "s:/var/lib/gdm/:/var/gdm/:" \
         "${S}/common/etc/ati/" \
         || die "sed failed."

      # Since "who" is in coreutils, we're using that one instead of "finger".
      sed -i -e 's:finger:who:' \
         "${S}/common/usr/share/doc/fglrx/examples/etc/acpi/" \
         || die "Replacing 'finger' with 'who' failed."
      # Adjust paths in the script from /usr/X11R6/bin/ to /opt/bin/ and
      # add funktion to detect default state.
      epatch "${FILESDIR}"/${PV}/ati-powermode-opt-path-2.patch

   pushd ${MODULE_DIR} >/dev/null
   ln -s "${ARCH_DIR}"/lib/modules/fglrx/build_mod/libfglrx_ip.a.GCC$(gcc-major-version) \
      || die "symlinking precompiled core failed"

   convert_to_m 2.6.x/Makefile || die "convert_to_m failed"

   # When built with ati's it defines a bunch of macros if
   # certain .config values are set, falling back to less reliable
   # detection methods if linux/autoconf.h is not available. We
   # simply use the linux/autoconf.h settings directly, bypassing the
   # detection script.
   sed -i -e 's/__SMP__/CONFIG_SMP/' *.c *h || die "SMP sed failed"
   sed -i -e 's/ifdef MODVERSIONS/ifdef CONFIG_MODVERSIONS/' *.c *.h \
      || die "MODVERSIONS sed failed"
   popd >/dev/null

   mkdir extra || die "mkdir failed"
   cd extra
   unpack ./../common/usr/src/ati/fglrx_sample_source.tgz
   sed -i -e 's:include/extensions/extutil.h:X11/extensions/extutil.h:' \
      lib/fglrx_gamma/fglrx_gamma.c || die "include fixup failed"
   # Add a category.
   mv programs/fglrx_gamma/fglrx_xgamma.{man,1} || die "man mv failed"
   cd ..

src_compile() {

   einfo "Building fgl_glxgears"
   cd "${S}"/extra/fgl_glxgears
   # These extra libs/utils either have an Imakefile that does not
   # work very well without tweaking or a Makefile ignoring CFLAGS
   # and the like. We bypass those.

   # The -DUSE_GLU is needed to compile using nvidia headers
   # according to a comment in ati-drivers-extra-8.33.6.ebuild.
   "$(tc-getCC)" -o fgl_glxgears ${CFLAGS} ${LDFLAGS} -DUSE_GLU \
      -I"${S}"/common/usr/include fgl_glxgears.c \
      -lGL -lGLU -lX11 -lm || die "fgl_glxgears build failed"

   einfo "Building fglrx_gamma lib"
   cd "${S}"/extra/lib/fglrx_gamma
   "$(tc-getCC)" -shared -fpic -o ${CFLAGS} ${LDFLAGS} \
      -DXF86MISC -Wl,-soname, fglrx_gamma.c \
      -lXext || die "fglrx_gamma lib build failed"
   ln -s || die "ln failed"
   ln -s || die "ln failed"

   einfo "Building fglrx_gamma util"
   cd "${S}"/extra/programs/fglrx_gamma
   "$(tc-getCC)" -o fglrx_xgamma ${CFLAGS} ${LDFLAGS} \
      -I../../../common/usr/X11R6/include -L../../lib/fglrx_gamma \
      fglrx_xgamma.c -lm -lfglrx_gamma -lX11 \
      || die "fglrx_gamma util build failed"


src_install() {

   # We can do two things here, and neither of them is very nice.

   # For direct rendering libGL has to be able to load one or more
   # dri modules (files ending in, like
   # Gentoo's mesa looks for these files in the location specified by
   # LIBGL_DRIVERS_PATH or LIBGL_DRIVERS_DIR, then in the hardcoded
   # location /usr/$(get_libdir)/dri. Ati's libGL does the same
   # thing, but the hardcoded location is /usr/X11R6/lib/modules/dri
   # on x86 and amd64 32bit, /usr/X11R6/lib64/modules/dri on amd64
   # 64bit. So we can either put the .so files in that (unusual,
   # compared to "normal" mesa libGL) location or set
   # LIBGL_DRIVERS_PATH. We currently do the latter. See also bug
   # 101539.

   # The problem with this approach is that LIBGL_DRIVERS_PATH
   # *overrides* the default hardcoded location, it does not extend
   # it. So if ati-drivers is merged but a non-ati libGL is selected
   # and its hardcoded path does not match our LIBGL_DRIVERS_PATH
   # (because it changed in a newer mesa or because it was compiled
   # for a different set of multilib abis than we are) stuff breaks.

   # We create one file per ABI to work with "native" multilib, see
   # below.

   echo "COLON_SEPARATED=LIBGL_DRIVERS_PATH" > "${T}/03ati-colon-sep"
   doenvd "${T}/03ati-colon-sep"

   # All libraries that we have a 32 bit and 64 bit version of on
   # amd64 are installed in src_install-libs. Everything else
   # (including libraries only available in native 64bit on amd64)
   # goes in here.

   # There used to be some code here that tried to detect running
   # under a "native multilib" portage ((precursor of)
   # I removed that, it
   # should just work (only doing some duplicate work). --marienz
   if has_multilib_profile; then
      local OABI=${ABI}
      for ABI in $(get_install_abis); do
      unset OABI

   # This is sorted by the order the files occur in the source tree.

   # X modules.
   exeinto /usr/$(get_libdir)/xorg/modules/drivers
   doexe "${MY_BASE_DIR}"/usr/X11R6/${PKG_LIBDIR}/modules/drivers/
   exeinto /usr/$(get_libdir)/xorg/modules/linux
   doexe "${MY_BASE_DIR}"/usr/X11R6/${PKG_LIBDIR}/modules/linux/
   exeinto /usr/$(get_libdir)/xorg/modules
   doexe "${MY_BASE_DIR}"/usr/X11R6/${PKG_LIBDIR}/modules/{esut.a,,}

   # Arch-specific files.
   # (s)bin.
   into /opt
   if use acpi; then
      dosbin "${ARCH_DIR}"/usr/sbin/atieventsd
   # We cleaned out the compilable stuff in src_unpack
   dobin "${ARCH_DIR}"/usr/X11R6/bin/*

   # lib.
   exeinto /usr/$(get_libdir)
   # Everything except for the installed in src_install-libs.
   doexe $(find "${ARCH_DIR}"/usr/X11R6/${PKG_LIBDIR} \
      -maxdepth 1 -type f -name '*.so*' -not -name '*')
   insinto /usr/$(get_libdir)
   doins $(find "${ARCH_DIR}"/usr/X11R6/${PKG_LIBDIR} \
      -maxdepth 1 -type f -not -name '*.so*')

   # Common files.
   # etc.
   insinto /etc/ati
   # Everything except for the script.
   doins common/etc/ati/{logo*,control,atiogl.xml,signature,amdpcsdb.default}
   if use acpi; then
      insopts -m0755
      doins common/etc/ati/

   # include.
   insinto /usr
   doins -r common/usr/include
   insinto /usr/include/X11/extensions
   doins common/usr/X11R6/include/X11/extensions/fglrx_gamma.h

   # Just the script.
   into /usr
   dosbin common/usr/sbin/*

   # data files for the control panel.
   insinto /usr/share
   doins -r common/usr/share/ati
   insinto /usr/share/pixmaps
   doins common/usr/share/icons/ccc_{large,small}.xpm
   make_desktop_entry amdcccle 'ATI Catalyst Control Center' \
      ccc_large System

   # doc.
   dohtml -r common/usr/share/doc/fglrx

   if use acpi; then
      doman common/usr/share/man/man8/atieventsd.8

      pushd common/usr/share/doc/fglrx/examples/etc/acpi >/dev/null

      exeinto /etc/acpi
      insinto /etc/acpi/events
      doins events/*

      popd >/dev/null

   # Done with the "source" tree. Install tools we rebuilt:
   dobin extra/fgl_glxgears/fgl_glxgears
   newdoc extra/fgl_glxgears/README README.fgl_glxgears

   dolib extra/lib/fglrx_gamma/*so*
   newdoc extra/lib/fglrx_gamma/README README.libfglrx_gamma

   dobin extra/programs/fglrx_gamma/fglrx_xgamma
   doman extra/programs/fglrx_gamma/fglrx_xgamma.1
   newdoc extra/programs/fglrx_gamma/README README.fglrx_gamma

   # Gentoo-specific stuff:
   if use acpi; then
      newinitd "${FILESDIR}"/atieventsd.init atieventsd \
         || die "Failed to install atieventsd.init.d"
      echo 'ATIEVENTSDOPTS=""' > "${T}"/atieventsd.conf
      newconfd "${T}"/atieventsd.conf atieventsd

        # Sabayon-specific stuff
        if use distribution && ! use x86-fbsd; then
                insinto /lib/fglrx
                doins "${WORKDIR}/common/lib/modules/fglrx/build_mod/2.6.x/fglrx.o"
                insinto /lib/fglrx
                doins "${WORKDIR}/common/lib/modules/fglrx/build_mod/2.6.x/fglrx.mod.o"


src_install-libs() {
   if [[ "${ABI}" == "amd64" ]]; then
      local EX_BASE_DIR="${BASE_DIR}_64a"
      local pkglibdir=lib64
      local MY_ARCH_DIR="${S}/arch/x86_64"
      local EX_BASE_DIR="${BASE_DIR}"
      local pkglibdir=lib
      local MY_ARCH_DIR="${S}/arch/x86"
   einfo "ati tree '${pkglibdir}' -> '$(get_libdir)' on system"

   local ATI_ROOT=/usr/$(get_libdir)/opengl/ati
   # To make sure we do not miss a spot when these change.
   local libmajor=1 libminor=2
   local libver=${libmajor}.${libminor}

   # The GLX libraries
   # (yes, this really is "lib" even on amd64/multilib --marienz)
   exeinto ${ATI_ROOT}/lib
   doexe "${MY_ARCH_DIR}"/usr/X11R6/${pkglibdir}/${libver}
   dosym${libver} ${ATI_ROOT}/lib/${libmajor}
   dosym${libver} ${ATI_ROOT}/lib/

   exeinto ${ATI_ROOT}/extensions
   doexe "${EX_BASE_DIR}"/usr/X11R6/${pkglibdir}/modules/extensions/*

   # DRI modules, installed into the path used by recent versions of mesa.
   exeinto /usr/$(get_libdir)/dri
   doexe "${MY_ARCH_DIR}"/usr/X11R6/${pkglibdir}/modules/dri/

   # AMD Cal libraries
   exeinto /usr/$(get_libdir)
   doexe "${MY_ARCH_DIR}"/usr/${pkglibdir}/*.so

   # Make up a Ati does not provide one, but mesa does. If
   # a (libtool-based) libfoo is built with present a
   # reference to it is put into, and compiling
   # (libtool-based) things that link too will complain if
   # disappears. So if we do not make up a
   # switching between mesa and ati becomes painful.
   local revision=$(printf '%d%02d%02d' $(get_version_components))
   sed -e "s:\${libmajor}:${libmajor}:g" \
      -e "s:\${libminor}:${libminor}:g" \
      -e "s:\${libdir}:$(get_libdir):g" \
      -e "s:\${revision}:${revision}:g" \
      "${FILESDIR}"/ > "${D}"/${ATI_ROOT}/lib/ \
      || die "sed failed to make"

   local envname="${T}"/04ati-dri-path
   if [[ -n ${ABI} ]]; then
   echo "LIBGL_DRIVERS_PATH=/usr/$(get_libdir)/dri" > "${envname}"
   doenvd "${envname}"

pkg_postinst() {
   /usr/bin/eselect opengl set --use-old ati

   elog "To switch to ATI OpenGL, run \"eselect opengl set ati\""
   elog "To change your xorg.conf you can use the bundled \"aticonfig\""
   elog "If you experience unexplained segmentation faults and kernel crashes"
   elog "with this driver and multi-threaded applications such as wine,"
   elog "set UseFastTLS in xorg.conf to either 0 or 1, but not 2."
   elog "You will have to source /etc/profile (or logout and back in) for dri"
   elog "to work, unless you previously had ati-drivers installed."

   # Workaroud screen corruption
   ewarn "If you experience screen corruption with this driver, try putting"
   ewarn '         Option "XAANoOffscreenPixmaps" "true"'
   ewarn "in the Device Section of /etc/X11/xorg.conf."


pkg_prerm() {
   /usr/bin/eselect opengl set xorg-x11

and the differences between the previous ati-drivers-8.593-r10 and his new ati-drivers-8.593-r10 are as follows:

Code: Select all

$ diff ati-drivers-8.593-r10.ebuild
<                 epatch "${FILESDIR}"/${PV}/${P}-acpica-2.6.29.patch
>                 epatch "${FILESDIR}"/8.600/${PN}-8.600-acpica-2.6.29.patch
<       /usr/bin/eselect opengl set --use-old xorg-x11
>       /usr/bin/eselect opengl set xorg-x11

and he added a file called ati-drivers-8.593-acpica-2.6.29.patch to the ebuild's files directory (/usr/local/portage/layman/sabayon/x11-drivers/ati-drivers/files/).

Now, the differences between the previous ati-drivers-8.593-r10 and ati-drivers-8.593-r11 are as follows:

Code: Select all

$ diff /usr/local/portage/layman/sabayon/x11-drivers/ati-drivers/ati-drivers-8.593-r11.ebuild /usr/local/portage/layman/sabayon/x11-drivers/ati-drivers/ati-drivers-8.593-r10.ebuild
<                 epatch "${FILESDIR}"/${P}-2.6.29.patch
>                 epatch "${FILESDIR}"/8.600/${PN}-8.600-acpica-2.6.29.patch
<       /usr/bin/eselect opengl set --use-old xorg-x11
>       /usr/bin/eselect opengl set xorg-x11

and the difference between the new ati-drivers-8.593-r10 and ati-drivers-8.593-r11 is:

Code: Select all

$ diff /usr/local/portage/layman/sabayon/x11-drivers/ati-drivers/ati-drivers-8.593-r11.ebuild
<                 epatch "${FILESDIR}"/${PV}/${P}-acpica-2.6.29.patch
>                 epatch "${FILESDIR}"/${P}-2.6.29.patch

EDIT (April 22, 2010 14:31 UTC): Just in case anyone arrives here by searching the forums for a solution to the 'AMD Unsupported Hardware watermark', there is a patch available: Unsupported Hardware watermark. I have successfully used this hack to remove the "AMD Unsupported hardware" watermark on a laptop with an ATI Mobility Radeon HD 5650 running SL amd64 5.2 KDE.

Post Reply