Re: [PATCH 0/8] tools/nolibc: add support for stack protector

[Willy Tarreau]

On Sun, Mar 26, 2023 at 08:41:29PM -0700, Paul E. McKenney wrote:
> On Sun, Mar 26, 2023 at 11:00:26AM -0700, Paul E. McKenney wrote:
> > On Sun, Mar 26, 2023 at 06:55:57PM +0200, Willy Tarreau wrote:
> > > On Sun, Mar 26, 2023 at 06:05:15PM +0200, Willy Tarreau wrote:
> > > > On Sun, Mar 26, 2023 at 06:00:45PM +0200, Willy Tarreau wrote:
> > > > > On Sun, Mar 26, 2023 at 08:45:55AM -0700, Paul E. McKenney wrote:
> > > > > > Glad I could "help"!  Timers.  Huh.  ;-)
> > > > > > 
> > > > > > Checking v6.2, though the rebase is a bit messy, so I won't be all
> > > > > > that confident in the results.
> > > > > 
> > > > > I got the same as you now. I don't know what I missed before not to
> > > > > face it, maybe it's the consequence of the rebase. I've re-applied
> > > > > the patches on top of 6.2.8 and am retesting now.
> > > > > 
> > > > > I think you don't need to waste more of your time on this for now
> > > > > since we have a reproducer. Thomas and I should take over.
> > > > 
> > > > And it's a 6.3 regression, as 6.2.8 works fine:
> > > > 
> > > >   $ make run
> > > >   (...)
> > > >   Kernel: arch/x86/boot/bzImage is ready  (#2)
> > > >   make[1]: Leaving directory '/g/public/linux/master'
> > > >   126 test(s) passed.
> > > >   $ tail  run.out 
> > > >   Errors during this test: 0
> > > >   
> > > >   Running test 'protection'
> > > >   0 -fstackprotector                                               [OK]
> > > >   Errors during this test: 0
> > > >   
> > > >   Total number of errors: 0
> > > >   Leaving init with final status: 0
> > > >   [    3.388706] ACPI: PM: Preparing to enter system sleep state S5
> > > >   [    3.389424] reboot: Power down
> > > > 
> > > > Now let's have fun bisecting it!
> > > 
> > > So I have a good news, 6.3-rc1 which dev.2023.03.20a is based on, fails,
> > > while 6.3-rc3 works. I haven't got further yet and am not sure it's useful
> > > to dig further given that it's an already fixed problem that is not related
> > > to the patches in your branch. I don't know if you usually rebase on more
> > > recent tags though.
> > 
> > Thank you for chasing this down!
> > 
> > In this case, I will at the very least merge with v6.3 before testing.
> > I have Joel Fernandes and Boqun Feng trying their hands at running
> > the RCU pull request for v6.4, so I will probably resist the urge to
> > inject confusion by rebasing onto v6.3-rc1.  ;-)
> 
> And merging the -rcu tree's "dev" branch with v6.3-rc3 got me a successful test:
> 
> 	Kernel: arch/x86/boot/bzImage is ready  (#5)
> 	make[1]: Leaving directory '/home/git/linux-build'
> 	125 test(s) passed.
> 
> So looking good.  ;-)

Perfect then! Thank you for double-checking!
Willy