svn commit: r357695 - in head: sys/kern sys/sys usr.bin/procstat

[Konstantin Belousov]

On Mon, Feb 10, 2020 at 10:11:43AM -0800, John Baldwin wrote:
> On 2/9/20 4:10 AM, Konstantin Belousov wrote:
> > Author: kib
> > Date: Sun Feb  9 12:10:37 2020
> > New Revision: 357695
> > URL: https://svnweb.freebsd.org/changeset/base/357695
> > 
> > Log:
> >   Add AT_BSDFLAGS auxv entry.
> >   
> >   The intent is to provide bsd-specific flags relevant to interpreter
> >   and C runtime.  I did not want to reuse AT_FLAGS which is common ELF
> >   auxv entry.
> >   
> >   Use bsdflags to report kernel support for sigfastblock(2).  This
> >   allows rtld and libthr to safely infer the syscall presence without
> >   SIGSYS.  The tunable kern.elf{32,64}.sigfastblock blocks reporting.
> >   
> >   Tested by:	pho
> >   Disscussed with:	cem, emaste, jilles
> >   Sponsored by:	The FreeBSD Foundation
> >   Differential revision:	https://reviews.freebsd.org/D12773
> 
> I find adding a new auxv type curious.  The MIPS ABI doc says that
> "bits under the 0xff000000 mask are reserved for system semantics".
> The powerpc and x86-64 docs don't define any bits at all.  In
> practice I think we are free to use AT_FLAGS however we wish as no
> use cases of "standard" bits have arisen since AT_FLAGS was first
> defined.

So you would prefer to have me used AT_FLAGS for sigfastblock indicator ?
I am feeling uncomfortable doing that.

My reasoning, to reformulate it from what I wrote in the commit message,
is to not pollute neither compilation nor ABI namespace for bsd-specific
flags.  AT_FLAGS was not touched by anybody and I do not want to open
it for use, since ABI group my finally find some use for it.