2008-10-01 06:40:59

by Roland McGrath

Subject: Re: [RFC patch 0/3] signals: add rt_tgsigqueueinfo syscall

The core of this looks fine to me. Presumably this would be expressed in
userland as pthread_sigqueue.

You are missing compat_sys_rt_tgsigqueueinfo for e.g. the
arch/x86/ia32/ia32entry.S table.

The clean way to do that would be a do_rt_tgsigqueueinfo taking
the siginfo_t * (not __user). That is, just split out the copy_from_user,
so compat_sys_rt_tgsigqueueinfo does copy_siginfo_from_user32 instead.


Thanks,
Roland

2008-10-01 07:53:31

by Thomas Gleixner

Subject: Re: [RFC patch 0/3] signals: add rt_tgsigqueueinfo syscall

On Tue, 30 Sep 2008, Roland McGrath wrote:

> The core of this looks fine to me. Presumably this would be expressed in
> userland as pthread_sigqueue.

Yes, that would match the pthread_kill then.

> You are missing compat_sys_rt_tgsigqueueinfo for e.g. the
> arch/x86/ia32/ia32entry.S table.

Duh, did not think about that.

> The clean way to do that would be a do_rt_tgsigqueueinfo taking
> the siginfo_t * (not __user). That is, just split out the copy_from_user,
> so compat_sys_rt_tgsigqueueinfo does copy_siginfo_from_user32 instead.

Will do.

Thanks,

tglx