Re: + kallsyms-increase-marker-density-to-16-1-to-accelerate-lookups-fix.patch added to mm-nonmm-unstable branch

[David Laight <david.laight.linux@gmail.com>]

On Thu, 15 Oct 2026 14:11:28 +0200
Nicolas Schier <n.schier@fritz.com> wrote:

> On Sun, Oct 11, 2026 at 03:44:01PM +0100, David Laight wrote:
> > On Sat, 10 Oct 2026 17:23:40 -0700
> > Andrew Morton <akpm@linux-foundation.org> wrote:
> >   
> > > The patch titled
> > >      Subject: kallsyms: copy constants from __KERNEL__ to user, drop the inclusion
> > > has been added to the -mm mm-nonmm-unstable branch.  Its filename is
> > >      kallsyms-increase-marker-density-to-16-1-to-accelerate-lookups-fix.patch  
> > 
> > There is a problem with the way this change in implemented.
> > Basically the kernel Makefile (etc) only builds scripts/kallsyms if it doesn't exist.  
> 
> Why do you think so?  Kbuild generates 
> $(objtree)/scripts/.kallsyms.o.cmd that should list all included header 
> files and other sources.  If any of them is touched, a rebuild will 
> happen next time the kernel build wants kallsyms binary.  It seems to 
> work exactly that way on my machine.  Or have I overlooked something 
> here?
> 

I've been doing a lot of test build through the kernel makefile.
They definitely weren't doing the correct rebuilds unless I deleted
both vmlinux.o and scripts/kallsyms, I didn't look why.

I wasn't booting the kernel, I just wanted the .S file to link with
a user program that included (most of) kernel/kallsyms.c and
kernel/kallsyms_seltests.c
(I've changed testing to do it differently most of the time.)

I'm nearly ready to submit some patches that speed the 'by name' lookup
by a factor of 12 with 256:1 markers and by 18 with 64:1.
They also reduce the compressed symbol table by about 10k for my 'normal'
kernel.
I've also run the selftests on the .S file from an x86-64 allyesconfig
kernel - don't try that with the current code regardless of the marker
density.

David