Re: [Qemu-devel] [PATCH 18/18] linux-user: Special-case ERESTARTSYS in target_strerror()

[Peter Maydell]

On 7 June 2016 at 20:53, Laurent Vivier <address@hidden> wrote:
>
>
> Le 06/06/2016 à 20:58, Peter Maydell a écrit :
>> Since TARGET_ERESTARTSYS and TARGET_ESIGRETURN are internal-to-QEMU
>> error numbers, handle them specially in target_strerror(), to avoid
>> confusing strace output like:
>>
>> 9521 rt_sigreturn(14,8,274886297808,8,0,268435456) = -1 errno=513 (Unknown 
>> error 513)
>>
>> Signed-off-by: Peter Maydell <address@hidden>
>> ---
>>  linux-user/syscall.c | 7 +++++++
>>  1 file changed, 7 insertions(+)
>>
>> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
>> index bcee02d..782d475 100644
>> --- a/linux-user/syscall.c
>> +++ b/linux-user/syscall.c
>> @@ -619,6 +619,13 @@ static inline int is_error(abi_long ret)
>>
>>  const char *target_strerror(int err)
>>  {
>> +    if (err == TARGET_ERESTARTSYS) {
>> +        return "To be restarted";
>> +    }
>> +    if (err == TARGET_QEMU_ESIGRETURN) {
>> +        return "Successful exit from sigreturn";
>> +    }
>> +
>>      if ((err >= ERRNO_TABLE_SIZE) || (err < 0)) {
>>          return NULL;
>>      }
>
> This is not the aim of this patch, but target_to_host_errno() has now
> these checks, perhaps we can remove this while we are here...

I think that would break the callers, which assume they can
pass in any number as a potential errno, and get
back NULL if it wasn't actually an errno. If we passed
them through to target_to_host_errno() it would pass
them on unchanged and the host strerror() would generate
a string "Unknown errno 134134234" or whatever.

thanks
-- PMM