deprecation: don't enable TCG plugins by default on 32 bit hosts

[RFC PATCH] deprecation: don't enable TCG plugins by default on 32 bit hosts

Posted by Alex Bennée 2 months, 2 weeks ago

The existing plugins already liberally use host pointer stuffing for
passing user data which will fail when doing 64 bit guests on 32 bit
hosts. We should discourage this by officially deprecating support and
adding another nail to the 32 bit host coffin.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
---
 docs/about/deprecated.rst | 11 +++++++++++
 configure                 | 11 ++++++++++-
 2 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/docs/about/deprecated.rst b/docs/about/deprecated.rst
index 88f0f03786..8a4e249717 100644
--- a/docs/about/deprecated.rst
+++ b/docs/about/deprecated.rst
@@ -184,6 +184,17 @@ be an effective use of its limited resources, and thus intends to discontinue
 it. Since all recent x86 hardware from the past >10 years is capable of the
 64-bit x86 extensions, a corresponding 64-bit OS should be used instead.
 
+TCG Plugin support not enabled by default on 32-bit hosts (since 9.2)
+'''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''
+
+While it is still possible to enable TCG plugin support for 32-bit
+hosts there are a number of potential pitfalls when instrumenting
+64-bit guests. The plugin APIs typically pass most addresses as
+uint64_t but practices like encoding that address in a host pointer
+for passing as user-data will loose data. As most software analysis
+benefits from having plenty of host memory it seems reasonable to
+encourage users to use 64 bit builds of QEMU for analysis work
+whatever targets they are instrumenting.
 
 System emulator CPUs
 --------------------
diff --git a/configure b/configure
index d08b71f14b..8acb311527 100755
--- a/configure
+++ b/configure
@@ -424,6 +424,7 @@ fi
 # Note that this case should only have supported host CPUs, not guests.
 # Please keep it sorted and synchronized with meson.build's host_arch.
 host_arch=
+host_bits=64
 linux_arch=
 case "$cpu" in
   aarch64)
@@ -434,12 +435,14 @@ case "$cpu" in
   armv*b|armv*l|arm)
     cpu=arm
     host_arch=arm
+    host_bits=32
     linux_arch=arm
     ;;
 
   i386|i486|i586|i686)
     cpu="i386"
     host_arch=i386
+    host_bits=32
     linux_arch=x86
     CPU_CFLAGS="-m32"
     ;;
@@ -458,11 +461,13 @@ case "$cpu" in
   mips*)
     cpu=mips
     host_arch=mips
+    host_bits=32
     linux_arch=mips
     ;;
 
   ppc)
     host_arch=ppc
+    host_bits=32
     linux_arch=powerpc
     CPU_CFLAGS="-m32"
     ;;
@@ -480,11 +485,13 @@ case "$cpu" in
 
   riscv32 | riscv64)
     host_arch=riscv
+    host_bits=32
     linux_arch=riscv
     ;;
 
   s390)
     linux_arch=s390
+    host_bits=32
     CPU_CFLAGS="-m31"
     ;;
   s390x)
@@ -495,6 +502,7 @@ case "$cpu" in
 
   sparc|sun4[cdmuv])
     cpu=sparc
+    host_bits=32
     CPU_CFLAGS="-m32 -mv8plus -mcpu=ultrasparc"
     ;;
   sparc64)
@@ -505,6 +513,7 @@ case "$cpu" in
   x32)
     cpu="x86_64"
     host_arch=x86_64
+    host_bits=32
     linux_arch=x86
     CPU_CFLAGS="-mx32"
     ;;
@@ -1028,7 +1037,7 @@ if test "$static" = "yes" ; then
   fi
   plugins="no"
 fi
-if test "$plugins" != "no"; then
+if test "$plugins" != "no" && test host_bits = 64; then
   plugins=yes
   subdirs="$subdirs contrib/plugins"
 fi
-- 
2.39.2

Re: [RFC PATCH] deprecation: don't enable TCG plugins by default on 32 bit hosts

Posted by Philippe Mathieu-Daudé 2 months, 2 weeks ago

On 5/9/24 13:11, Alex Bennée wrote:
> The existing plugins already liberally use host pointer stuffing for
> passing user data which will fail when doing 64 bit guests on 32 bit
> hosts. We should discourage this by officially deprecating support and
> adding another nail to the 32 bit host coffin.
> 
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> ---
>   docs/about/deprecated.rst | 11 +++++++++++
>   configure                 | 11 ++++++++++-
>   2 files changed, 21 insertions(+), 1 deletion(-)


> @@ -480,11 +485,13 @@ case "$cpu" in
>   
>     riscv32 | riscv64)

We need to split the riscv64 case.

>       host_arch=riscv
> +    host_bits=32
>       linux_arch=riscv
>       ;;

Re: [RFC PATCH] deprecation: don't enable TCG plugins by default on 32 bit hosts

Posted by Peter Maydell 2 months, 2 weeks ago

On Thu, 5 Sept 2024 at 12:13, Alex Bennée <alex.bennee@linaro.org> wrote:
>
> The existing plugins already liberally use host pointer stuffing for
> passing user data which will fail when doing 64 bit guests on 32 bit
> hosts. We should discourage this by officially deprecating support and
> adding another nail to the 32 bit host coffin.
>
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> ---
>  docs/about/deprecated.rst | 11 +++++++++++
>  configure                 | 11 ++++++++++-
>  2 files changed, 21 insertions(+), 1 deletion(-)
>
> diff --git a/docs/about/deprecated.rst b/docs/about/deprecated.rst
> index 88f0f03786..8a4e249717 100644
> --- a/docs/about/deprecated.rst
> +++ b/docs/about/deprecated.rst
> @@ -184,6 +184,17 @@ be an effective use of its limited resources, and thus intends to discontinue
>  it. Since all recent x86 hardware from the past >10 years is capable of the
>  64-bit x86 extensions, a corresponding 64-bit OS should be used instead.
>
> +TCG Plugin support not enabled by default on 32-bit hosts (since 9.2)
> +'''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''
> +
> +While it is still possible to enable TCG plugin support for 32-bit
> +hosts there are a number of potential pitfalls when instrumenting
> +64-bit guests. The plugin APIs typically pass most addresses as
> +uint64_t but practices like encoding that address in a host pointer
> +for passing as user-data will loose data. As most software analysis

"lose"

> +benefits from having plenty of host memory it seems reasonable to
> +encourage users to use 64 bit builds of QEMU for analysis work
> +whatever targets they are instrumenting.


> diff --git a/configure b/configure
> index d08b71f14b..8acb311527 100755
> --- a/configure
> +++ b/configure
> @@ -424,6 +424,7 @@ fi
>  # Note that this case should only have supported host CPUs, not guests.
>  # Please keep it sorted and synchronized with meson.build's host_arch.
>  host_arch=
> +host_bits=64
>  linux_arch=
>  case "$cpu" in
>    aarch64)
> @@ -434,12 +435,14 @@ case "$cpu" in
>    armv*b|armv*l|arm)
>      cpu=arm
>      host_arch=arm
> +    host_bits=32
>      linux_arch=arm
>      ;;
>
>    i386|i486|i586|i686)
>      cpu="i386"
>      host_arch=i386
> +    host_bits=32
>      linux_arch=x86
>      CPU_CFLAGS="-m32"
>      ;;

This is pretty awkward. We should only put stuff into this
"switch per CPU architecture" where we absolutely cannot
automatically determine it. Host bitness can be automatically
determined (see what the compiler has set __SIZEOF_POINTER__
to), so we should do that.

thanks
-- PMM