struct sigaction sa_restorer field

struct sigaction sa_restorer field

[John David Anglin]

The sa_restorer field is obsolete and not in the hppa struct  
sigaction, but we still
define SA_RESTORER.  What should be done?

Dave
--
John David Anglin	dave.anglin@bell.net

Re: struct sigaction sa_restorer field

[Helge Deller]

Hi Dave,

> The sa_restorer field is obsolete and not in the hppa struct  
> sigaction, but we still define SA_RESTORER.  What should be done?

Basically to not break userspace we can't (and shouldn't) add the sa_restorer struct member.
So, the only option would be to drop the SA_RESTORER #define, right?
But I have no idea how and if this will break some builds, since the major arches seem to define SA_RESTORER...

Helge

Re: struct sigaction sa_restorer field

[John David Anglin]

On 7/21/2014 4:25 AM, Helge Deller wrote:
> Hi Dave,
>
>> The sa_restorer field is obsolete and not in the hppa struct
>> sigaction, but we still define SA_RESTORER.  What should be done?
> Basically to not break userspace we can't (and shouldn't) add the sa_restorer struct member.
> So, the only option would be to drop the SA_RESTORER #define, right?
> But I have no idea how and if this will break some builds, since the major arches seem to define SA_RESTORER...
>
I asked the question when I saw the crashme kernel test package didn't 
build due to the missing field.

In the generic sigaction struct, the field is conditional on the 
SA_RESTORER define.  I imagine that
the major arches also have sa_restorer struct member.

I agree it seems best to remove the define.

Dave

-- 
John David Anglin    dave.anglin@bell.net

Aw: Re: struct sigaction sa_restorer field

[Helge Deller]

> >> The sa_restorer field is obsolete and not in the hppa struct
> >> sigaction, but we still define SA_RESTORER.  What should be done?
> > Basically to not break userspace we can't (and shouldn't) add the sa_restorer struct member.
> > So, the only option would be to drop the SA_RESTORER #define, right?
> > But I have no idea how and if this will break some builds, since the major arches seem to define SA_RESTORER...
> >
> I asked the question when I saw the crashme kernel test package didn't 
> build due to the missing field.
> 
> In the generic sigaction struct, the field is conditional on the 
> SA_RESTORER define.  I imagine that
> the major arches also have sa_restorer struct member.
> 
> I agree it seems best to remove the define.

Ok, then can you send a patch ?
Should we request to backport it into stable kernels? I think so...

Helge

Re: struct sigaction sa_restorer field

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 852 bytes --]

On Mon 21 Jul 2014 10:25:46 Helge Deller wrote:
> > The sa_restorer field is obsolete and not in the hppa struct
> > sigaction, but we still define SA_RESTORER.  What should be done?
> 
> Basically to not break userspace we can't (and shouldn't) add the
> sa_restorer struct member. So, the only option would be to drop the
> SA_RESTORER #define, right? But I have no idea how and if this will break
> some builds, since the major arches seem to define SA_RESTORER...

would be nice to drop.  i had to update strace to undef it for hppa since the 
field didn't actually exist -- it uses the define to see whether the field 
should be decoded (which is why strace decoding of this syscall hasn't worked 
in the past on parisc).  not all arches define it, so i think there's already 
cover for code needing to check for its existence.
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 819 bytes --]