Re: [Qemu-devel] [PATCH V2 4/9] Add tpm_tis driver to build process

[Blue Swirl]

On Fri, Apr 1, 2011 at 10:57 PM, Stefan Berger
<address@hidden> wrote:
> On 04/01/2011 02:14 PM, Blue Swirl wrote:
>>
>> On Wed, Mar 30, 2011 at 10:42 PM, Stefan Berger
>> <address@hidden>  wrote:
>>>
>>> The TPM interface (tpm_tis) needs to be explicitly enabled via
>>> ./configure --enable-tpm. This restricts the building of the
>>> TPM support to i386 and x86_64 targets since both backends I know
>>> of, the Xen backend and the libtpms-based backend, will likely only
>>> be available for these targets, at least initially. The list can be
>>> easily extend. This measure prevents that one will end up with support
>>> for a frontend but no available backend.
>>>
>>> Signed-off-by: Stefan Berger<address@hidden>
>>>
>>> Index:qemu/Makefile.target
>>> ===================================================================
>>> ---
>>>  Makefile.target |    7 +++++++
>>>  configure       |   20 ++++++++++++++++++++
>>>  2 files changed, 27 insertions(+)
>>>
>>> Index: qemu-git/Makefile.target
>>> ===================================================================
>>> --- qemu-git.orig/Makefile.target
>>> +++ qemu-git/Makefile.target
>>> @@ -303,6 +303,13 @@ obj-sparc-y += cs4231.o eccmemctl.o sbi.
>>>
>>>  # GRLIB
>>>  obj-sparc-y += grlib_gptimer.o grlib_irqmp.o grlib_apbuart.o
>>> +
>>> +ifeq ($(TARGET_ARCH),$(filter $(TARGET_ARCH),i386 x86_64))
>>
>> The ifeq above is not needed since the line below uses obj-i386.
>>
> Will remove.
>>>
>>> +
>>> +obj-i386-$(CONFIG_TPM) += tpm_tis.o
>>> +
>>> +endif
>>> +
>>>  endif
>>>
>>>  obj-arm-y = integratorcp.o versatilepb.o arm_pic.o arm_timer.o
>>> Index: qemu-git/configure
>>> ===================================================================
>>> --- qemu-git.orig/configure
>>> +++ qemu-git/configure
>>> @@ -175,6 +175,7 @@ trace_backend="nop"
>>>  trace_file="trace"
>>>  spice=""
>>>  rbd=""
>>> +tpm="no"
>>>
>>>  # parse CC options first
>>>  for opt do
>>> @@ -708,6 +709,8 @@ for opt do
>>>   ;;
>>>   --kerneldir=*) kerneldir="$optarg"
>>>   ;;
>>> +  --enable-tpm) tpm="yes"
>>> +  ;;
>>>   --with-pkgversion=*) pkgversion=" ($optarg)"
>>>   ;;
>>>   --disable-docs) docs="no"
>>> @@ -921,6 +924,7 @@ echo "                           Default
>>>  echo "  --disable-spice          disable spice"
>>>  echo "  --enable-spice           enable spice"
>>>  echo "  --enable-rbd             enable building the rados block device
>>> (rbd)"
>>> +echo "  --enable-tpm             enables an emulated TPM"
>>>  echo ""
>>>  echo "NOTE: The object files are built at the place where configure is
>>> launched"
>>>  exit 1
>>> @@ -2540,6 +2544,7 @@ echo "Trace output file $trace_file-<pid
>>>  echo "spice support     $spice"
>>>  echo "rbd support       $rbd"
>>>  echo "xfsctl support    $xfs"
>>> +echo "TPM support       $tpm"
>>>
>>>  if test $sdl_too_old = "yes"; then
>>>  echo "->  Your SDL version is too old - please upgrade to have SDL
>>> support"
>>> @@ -3324,6 +3329,21 @@ if test "$gprof" = "yes" ; then
>>>   fi
>>>  fi
>>>
>>> +if test "$linux" = "yes"&&  test "$tpm" = "yes"; then
>>> +  has_tpm=0
>>> +  if test "$target_softmmu" = "yes" ; then
>>> +    case "$TARGET_BASE_ARCH" in
>>> +    i386)
>>> +      has_tpm=1
>>
>> This is not a real compile test, please try to build a dummy program
>> that depends on TPM headers like for example preadv probe. Then it is
>> not needed to check for Linux or i386 explicitly.
>
> At this point there is no compile test needed since all code is 'there'.
> It's merely adding the front-end,i.e., the TPM TIS emulation to be compiled.

If the basic device (without the tpms-devel library) can be built on
any OS, the flag should go to default-configs/*86*-softmmu.mak.

> The (libtpms-based) backend is then added later in patch [9/9]. There you
> then find this here:
>
>   if test "$has_tpm" = "1"; then
> +      if test -r /usr/include/libtpms/tpm_library.h ; then

Here you make assumptions on the header file location, but it could be
in /usr/local,  /opt or somewhere where the cross compiler happens to
find it. Please just do the compile test.

> +          echo "CONFIG_TPM_BUILTIN=y">>  $config_target_mak
> +      fi
>       echo "CONFIG_TPM=y">>  $config_host_mak
>   fi
>  fi
>
>
> So this is then trying to test for the libtpms-devel package, and if found,
> adds the tpm_builtin.c to the build.
> Now did it look wrong for patch 4 and it does make sense in combination with
> patch 9?

The test should be added when the code which uses the library is added.