This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Add finit_module syscall for Linux


On Fri, Jan 25, 2013 at 03:01:37PM -0500, KOSAKI Motohiro wrote:
> On Fri, Jan 25, 2013 at 2:50 PM, Michael Kerrisk (man-pages)
> <mtk.manpages@gmail.com> wrote:
> > On Fri, Jan 25, 2013 at 8:14 PM, Kees Cook <keescook@chromium.org> wrote:
> >> On Fri, Jan 25, 2013 at 9:31 AM, KOSAKI Motohiro
> >> <kosaki.motohiro@gmail.com> wrote:
> >>>>     gettid(2)
> >>>
> >>> Very frequently uses. because linux /proc uses tid. pthrad_self()
> >>> don't help us when facing libc/kernel combination issue.
> >>> I have no seen any issue of exporting.
> >>
> >> Yeah, I've needed gettid() in multiple situations before. I think
> >> having this available in glibc would be great.
> >
> > http://sourceware.org/bugzilla/show_bug.cgi?id=6399 has been around
> > for a long time now, but didn't get much traction with Ulrich...
> 
> I completely agree with you. If we can roll back time, we should fix
> pthread_self()
> to return tid because current implementation is POSIX invalid (*).
> however it's impossible and making new interface for getting tid is
> best practical solution.
> 
> (*) pthread_kill() must return -ESRCH if target thread was already
> exited. that imply
>     pthread_self() returns pid like id number, not bare C pointer.

false. that's only required if the impl *detects* the condition, and
even then only in outdated posix 2001. posix 2008 clarified the issue.
using kernel tids for pthread_t is a very bad design.

Rich


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]