[PATCH v2 5/6] arm64: audit: Add makefile rule to create unistd_32.h for compat syscalls

[takahiro.akashi@linaro.org (AKASHI Takahiro)]

Catalin,

On 01/23/2014 11:53 PM, Catalin Marinas wrote:
> On Fri, Jan 17, 2014 at 08:13:18AM +0000, AKASHI Takahiro wrote:
>> generic compat sycall audit (lib/compat_audit.c) requires unistd_32.h
>> for __NR_xyx compat syscall numbers. This is a different file from unistd32.h
>> on arm64 and so it must be generated from unistd32.h.
>>
>> Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
>> ---
>>   arch/arm64/Makefile                 |    4 ++++
>>   arch/arm64/kernel/syscalls/Makefile |   20 ++++++++++++++++++++
>>   2 files changed, 24 insertions(+)
>>   create mode 100644 arch/arm64/kernel/syscalls/Makefile
>>
>> diff --git a/arch/arm64/Makefile b/arch/arm64/Makefile
>> index 2fceb71..6d24f92 100644
>> --- a/arch/arm64/Makefile
>> +++ b/arch/arm64/Makefile
>> @@ -72,6 +72,10 @@ PHONY += vdso_install
>>   vdso_install:
>>   	$(Q)$(MAKE) $(build)=arch/arm64/kernel/vdso $@
>>
>> +# Compat syscall header generation
>> +archheaders:
>> +	$(Q)$(MAKE) $(build)=arch/arm64/kernel/syscalls $@
>
> See my other post to the lib/compat_audit.c file. I think that's too
> complex for what you need.

Generation script is getting more complexed than I assumed at first
because some of system call names are a bit inconsistent with native 32-bit
system calls, for example, fchown16 vs. fchown, fchown vs. fchown32.

Now my tentative sed script looks like:
+quiet_cmd_syshdr = SYSHDR $@
+        cmd_syshdr = cat $< | sed -r \
+        -e 's/compat_//' \
+        -e 's/_wrapper//' \
+        -e 's/(sys_[fl]?chown)(\))/\132\)/' \
+        -e 's/(sys_[gs]et)(|e|fs|re|res)(uid\))/\1\2uid32\)/' \
+        -e 's/(sys_[gs]et)(|e|fs|re|res)(gid\))/\1\2gid32\)/' \
+        -e 's/(sys_[gs]etgroups)(\))/\132\)/' \
+        -e 's/(sys_new)(.*)/sys_\2/' \
+        -e 's/sys_mmap_pgoff/sys_mmap2/' \
+        -e 's/(sys_[_a-z]*)16(.*)/\1\2/' \
+        -e 's/^__SYSCALL\((.*),[ ]*sys_([^)].*)\).*/\#define __NR_\2 \1/p;d' \
+        | grep -v __NR_ni_syscall > $@

So, yeah, I agree with you now.

-Takahiro AKASHI