Re: port-mips/59064 (jemalloc switch to 5.3 broke userland)

[Rin Okuyama <rokuyama.rk%gmail.com@localhost>]

On 2025/04/12 21:52, Christos Zoulas wrote:
> This means that:
>
> Index: include/jemalloc/internal/jemalloc_internal_defs.h
> ===================================================================
> RCS file: /cvsroot/src/external/bsd/jemalloc/include/jemalloc/internal/jemalloc_internal_defs.h,v
> retrieving revision 1.19
> diff -u -p -u -r1.19 jemalloc_internal_defs.h
> --- include/jemalloc/internal/jemalloc_internal_defs.h  1 Feb 2025 16:26:58 -0000       1.19
> +++ include/jemalloc/internal/jemalloc_internal_defs.h  12 Apr 2025 12:50:39 -0000
> @@ -157,7 +157,9 @@
>   /* #undef JEMALLOC_MUTEX_INIT_CB */
>   
>   /* Non-empty if the tls_model attribute is supported. */
> +#ifndef __mips__
>   #define JEMALLOC_TLS_MODEL __attribute__((tls_model("initial-exec")))
> +#endif
>   
>   /*
>    * JEMALLOC_DEBUG enables assertions and other sanity checks, and disables
> Exit 1
>
> will probably fix it.

Thanks, it works fine! (We need empty JEMALLOC_TLS_MODEL define
even for mips, although.)

> JEMALLOC_TLS_MODEL is set by autoconf. Can you run configure
> on a mips machine and see what it does.

It is set even for mips, as configure script checks only
whether compiler supports it or not:

https://github.com/jemalloc/jemalloc/blob/dev/configure.ac#L956-L967

> My edgerouter died. I should probably get another.

Octeon processors easily heat up too much... I have three ERLite-3,
two of that are inherited from ryo@. Can I send one for you? ;)

Thanks,
rin

> christos
>
> > On Apr 12, 2025, at 5:40 AM, Rin Okuyama via gnats <gnats-admin%NetBSD.org@localhost> wrote:
> >
> > The following reply was made to PR port-mips/59064; it has been noted by GNATS.
> >
> > From: Rin Okuyama <rokuyama.rk%gmail.com@localhost>
> > To: gnats-bugs%NetBSD.org@localhost, Martin Husemann <martin%duskware.de@localhost>
> > Cc:
> > Subject: Re: port-mips/59064 (jemalloc switch to 5.3 broke userland)
> > Date: Sat, 12 Apr 2025 18:38:08 +0900
> >
> > Hi,
> >
> > I've carried out bisectioning for upstream repo to find out
> > the first bad commit:
> >
> > https://github.com/jemalloc/jemalloc/commit/1aabab5f
> >
> > By reverting this commit from our in-tree jemalloc, both n64 and
> > n32 userlands successfully boot up into multi-user mode on ERLite-3.
> >
> > However, now, statically-linked binaries crash via:
> > abort() <-- __tls_get_addr() <-- malloc_init_hard() <-- _initarray()
> >
> > This can be a hint for what going on with and without the
> > commit in problem.
> >
> > Just a quick report...
> >
> > Thanks,
> > rin