Re: [PATCH 4/5] linux: Optimize posix_spawn spurious sigaction calls

[Joseph Myers <joseph at codesourcery dot com>]

On Mon, 7 Oct 2019, Adhemerval Zanella wrote:

> However, glibc supports older kernels as old as v3.2 and it will take
> some years and releases to make v5.3 or new the minimum support kernel.
> And I think it would be nice to have this optimization even for older
> kernels.

I should note there are two reasons I haven't proposed an increase from 
3.2 to a more recent minimum kernel version, given that 3.2 is no longer a 
maintained Linux kernel release series.

1. An increase to 3.16, the current oldest longterm kernel at 
<https://www.kernel.org/category/releases.html>, doesn't really allow much 
in the way of cleanups.  The next minimum kernel version that would allow 
many cleanups is 4.4 (most separate socket syscalls available on all 
socketcall architectures, so most socketcall support can be removed) - I 
think that's the next update that brings enough benefits to be worthwhile.  
(And after that, the next such update would be to require 64-bit time 
syscalls on 32-bit architectures once all older kernel series stop being 
maintained - in 2024 based on the dates on that page.)

2. I was hoping that before we next increase the minimum kernel version we 
can have the changes Carlos was proposing to stop giving an error at 
startup for a too-old kernel, instead allowing code to run with syscalls 
possibly failing, in order to work better in the case of (new userspace in 
a container running under) older kernel version numbers with features 
possibly backported or no relevant new features actually required by 
glibc.

-- 
Joseph S. Myers
joseph@codesourcery.com