[PATCH v2 for-4.15 0/2] firmware: fix build on Alpine

[PATCH v2 for-4.15 0/2] firmware: fix build on Alpine

[Roger Pau Monne]

Hello,

While the series started as a build fix for Alpine I think they are
interesting on their own for other OSes/distros.

The main risk for patch 1 is breaking the build in some obscure
distro/OS and toolchain combination. We aim to have this mostly covered
by gitlab CI. Patch 2 main risk is breaking the Alpine containers in
gitlab CI, but they are already failing.

Wanted to send this yesterday but was waiting on gitlab CI output, it's now
all green:

https://gitlab.com/xen-project/people/royger/xen/-/pipelines/264321420

Thanks, Roger.

Roger Pau Monne (2):
  firmware: provide a stand alone set of headers
  automation: enable rombios build on Alpine

 automation/scripts/build                      |  5 +--
 tools/firmware/Rules.mk                       | 13 +++++++
 tools/firmware/include/stdarg.h               | 10 +++++
 tools/firmware/include/stdbool.h              |  9 +++++
 tools/firmware/include/stddef.h               | 10 +++++
 tools/firmware/include/stdint.h               | 39 +++++++++++++++++++
 tools/firmware/rombios/32bit/rombios_compat.h |  4 +-
 7 files changed, 84 insertions(+), 6 deletions(-)
 create mode 100644 tools/firmware/include/stdarg.h
 create mode 100644 tools/firmware/include/stdbool.h
 create mode 100644 tools/firmware/include/stddef.h
 create mode 100644 tools/firmware/include/stdint.h

-- 
2.30.1

[PATCH v2 for-4.15 1/2] firmware: provide a stand alone set of headers

[Roger Pau Monne]

The current build of the firmware relies on having 32bit compatible
headers installed in order to build some of the 32bit firmware.
Usually this can be solved by using the -ffreestanding compiler option
which drops the usage of the system headers in favor of a private set
of freestanding headers provided by the compiler itself that are not
tied to libc.

However such option is broken at least in the gcc compiler provided in
Alpine Linux, as the system include path (ie: /usr/include) takes
precedence over the gcc private include path:

#include <...> search starts here:
 /usr/include
 /usr/lib/gcc/x86_64-alpine-linux-musl/10.2.1/include

And the headers in /usr/include are exclusively 64bit.

Since -ffreestanding is currently broken on at least that distro, and
for resilience against future compilers also having the option broken
provide a set of stand alone 32bit headers required for the firmware
build.

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Release-Acked-by: Ian Jackson <iwj@xenproject.org>
---
Using -ffreestanding alone seems fragile, as it's broken on some
distros. Compensate for this by providing our own set of stand alone
headers for the firmware bits. Having the include paths wrongly sorted
can easily make the system headers being picked up instead of the gcc
ones, and then building can randomly fail because the system headers
could be amd64 only (like the musl ones).

I've also seen clang-9 on Debian with the following include paths:

#include "..." search starts here:
#include <...> search starts here:
 /usr/local/include
 /usr/lib/llvm-9/lib/clang/9.0.1/include
 /usr/include/x86_64-linux-gnu
 /usr/include

Which also seems slightly dangerous as local comes before the compiler
private path.

IMO using -ffreestanding and our own set of stand alone headers is
more resilient.

Regarding the release risks, the main one would be breaking the build
(as it's currently broken on Alpine). I think there's a very low risk
of this change successfully producing a binary image that's broken,
and hence with enough build testing it should be safe to merge.
---
Changes since v2:
 - Add a __P64__ check to stdint.h.
 - Reword the comment in Rules.mk.
---
 tools/firmware/Rules.mk                       | 13 +++++++
 tools/firmware/include/stdarg.h               | 10 +++++
 tools/firmware/include/stdbool.h              |  9 +++++
 tools/firmware/include/stddef.h               | 10 +++++
 tools/firmware/include/stdint.h               | 39 +++++++++++++++++++
 tools/firmware/rombios/32bit/rombios_compat.h |  4 +-
 6 files changed, 82 insertions(+), 3 deletions(-)
 create mode 100644 tools/firmware/include/stdarg.h
 create mode 100644 tools/firmware/include/stdbool.h
 create mode 100644 tools/firmware/include/stddef.h
 create mode 100644 tools/firmware/include/stdint.h

diff --git a/tools/firmware/Rules.mk b/tools/firmware/Rules.mk
index dc372a86d8..9f78a7dec9 100644
--- a/tools/firmware/Rules.mk
+++ b/tools/firmware/Rules.mk
@@ -17,3 +17,16 @@ $(call cc-options-add,CFLAGS,CC,$(EMBEDDED_EXTRA_CFLAGS))
 
 # Extra CFLAGS suitable for an embedded type of environment.
 CFLAGS += -ffreestanding -msoft-float
+
+# Use our own set of stand alone headers to build firmware.
+#
+# Ideally using -ffreestanding should be enough, but that relies on the
+# compiler having the right order for include paths (ie: compiler private
+# headers before system ones) or the libc headers having proper arch-agnostic
+# freestanding support. This is not the case in Alpine at least which searches
+# system headers before compiler ones and has arch-specific libc headers. This
+# has been reported upstream:
+# https://gitlab.alpinelinux.org/alpine/aports/-/issues/12477
+# In the meantime (and for resilience against broken systems) use our own set
+# of headers that provide what's needed for the firmware build.
+CFLAGS += -nostdinc -I$(XEN_ROOT)/tools/firmware/include
diff --git a/tools/firmware/include/stdarg.h b/tools/firmware/include/stdarg.h
new file mode 100644
index 0000000000..c5e3761cd2
--- /dev/null
+++ b/tools/firmware/include/stdarg.h
@@ -0,0 +1,10 @@
+#ifndef _STDARG_H_
+#define _STDARG_H_
+
+typedef __builtin_va_list va_list;
+#define va_copy(dest, src) __builtin_va_copy(dest, src)
+#define va_start(ap, last) __builtin_va_start(ap, last)
+#define va_end(ap) __builtin_va_end(ap)
+#define va_arg __builtin_va_arg
+
+#endif
diff --git a/tools/firmware/include/stdbool.h b/tools/firmware/include/stdbool.h
new file mode 100644
index 0000000000..0cf76b106c
--- /dev/null
+++ b/tools/firmware/include/stdbool.h
@@ -0,0 +1,9 @@
+#ifndef _STDBOOL_H_
+#define _STDBOOL_H_
+
+#define bool _Bool
+#define true 1
+#define false 0
+#define __bool_true_false_are_defined 1
+
+#endif
diff --git a/tools/firmware/include/stddef.h b/tools/firmware/include/stddef.h
new file mode 100644
index 0000000000..c7f974608a
--- /dev/null
+++ b/tools/firmware/include/stddef.h
@@ -0,0 +1,10 @@
+#ifndef _STDDEF_H_
+#define _STDDEF_H_
+
+typedef __SIZE_TYPE__ size_t;
+
+#define NULL ((void*)0)
+
+#define offsetof(t, m) __builtin_offsetof(t, m)
+
+#endif
diff --git a/tools/firmware/include/stdint.h b/tools/firmware/include/stdint.h
new file mode 100644
index 0000000000..16a0b6de19
--- /dev/null
+++ b/tools/firmware/include/stdint.h
@@ -0,0 +1,39 @@
+#ifndef _STDINT_H_
+#define _STDINT_H_
+
+#if defined(__LP64__) || defined(__P64__)
+#error "32bit only header"
+#endif
+
+typedef unsigned char uint8_t;
+typedef signed char int8_t;
+
+typedef unsigned short uint16_t;
+typedef signed short int16_t;
+
+typedef unsigned int uint32_t;
+typedef signed int int32_t;
+
+typedef unsigned long long uint64_t;
+typedef signed long long int64_t;
+
+#define INT8_MIN        (-0x7f-1)
+#define INT16_MIN       (-0x7fff-1)
+#define INT32_MIN       (-0x7fffffff-1)
+#define INT64_MIN       (-0x7fffffffffffffffll-1)
+
+#define INT8_MAX        0x7f
+#define INT16_MAX       0x7fff
+#define INT32_MAX       0x7fffffff
+#define INT64_MAX       0x7fffffffffffffffll
+
+#define UINT8_MAX       0xff
+#define UINT16_MAX      0xffff
+#define UINT32_MAX      0xffffffffu
+#define UINT64_MAX      0xffffffffffffffffull
+
+typedef uint32_t uintptr_t;
+
+#define UINTPTR_MAX     UINT32_MAX
+
+#endif
diff --git a/tools/firmware/rombios/32bit/rombios_compat.h b/tools/firmware/rombios/32bit/rombios_compat.h
index 3fe7d67721..8ba4c17ffd 100644
--- a/tools/firmware/rombios/32bit/rombios_compat.h
+++ b/tools/firmware/rombios/32bit/rombios_compat.h
@@ -8,9 +8,7 @@
 
 #define ADDR_FROM_SEG_OFF(seg, off)  (void *)((((uint32_t)(seg)) << 4) + (off))
 
-typedef unsigned char uint8_t;
-typedef unsigned short int uint16_t;
-typedef unsigned int uint32_t;
+#include <stdint.h>
 
 typedef uint8_t  Bit8u;
 typedef uint16_t Bit16u;
-- 
2.30.1

[PATCH v2 for-4.15 2/2] automation: enable rombios build on Alpine

[Roger Pau Monne]

It's now safe to enable the build of rombios on Alpine systems, as
hvmloader already builds fine there.

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
---
 automation/scripts/build | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/automation/scripts/build b/automation/scripts/build
index d8990c3bf4..87e44bb940 100755
--- a/automation/scripts/build
+++ b/automation/scripts/build
@@ -31,9 +31,8 @@ fi
 if ! test -z "$(ldd /bin/ls|grep musl|head -1)"; then
     # disable --disable-werror for QEMUU when building with MUSL
     cfgargs+=("--with-extra-qemuu-configure-args=\"--disable-werror\"")
-    # hvmloader doesn't build on MUSL systems
-    cfgargs+=("--disable-seabios")
-    cfgargs+=("--disable-rombios")
+    # SeaBIOS doesn't build on MUSL systems
+    cfgargs+=("--with-system-seabios=/bin/false")
 fi
 
 # Qemu requires Python 3.5 or later
-- 
2.30.1

Re: [PATCH v2 for-4.15 2/2] automation: enable rombios build on Alpine

[Ian Jackson]

Roger Pau Monne writes ("[PATCH v2 for-4.15 2/2] automation: enable rombios build on Alpine"):
> It's now safe to enable the build of rombios on Alpine systems, as
> hvmloader already builds fine there.

Release-Acked-by: Ian Jackson <iwj@xenproject.org>