On Thu, Jan 22, 2026 at 04:04:55AM -0800, Breno Leitao wrote:
> Hello Mike,
>
> On Thu, Jan 22, 2026 at 12:57:50PM +0200, Mike Rapoport wrote:
> > > +/**
> > > + * DOC: Kexec Metadata ABI
> > > + *
> >
> > It would be nice to link it from Documentation/ as well ;-)
>
> Ack! I am planning something as:
>
> commit 90e098ca0d611b44594f08e50ba1cff3c932dd2b
> Author: Breno Leitao <[email protected]>
> Date: Thu Jan 22 03:47:23 2026 -0800
>
> kho: document kexec-metadata tracking feature
>
> Add documentation for the kexec-metadata feature that tracks the
> previous kernel version and kexec boot count across kexec reboots.
> This helps diagnose bugs that only reproduce when kexecing from
> specific kernel versions.
>
> Suggested-by: Mike Rapoport <[email protected]>
> Signed-off-by: Breno Leitao <[email protected]>
>
> diff --git a/Documentation/admin-guide/mm/kho.rst
> b/Documentation/admin-guide/mm/kho.rst
> index 6dc18ed4b8861..1faf2c3ba4620 100644
> --- a/Documentation/admin-guide/mm/kho.rst
> +++ b/Documentation/admin-guide/mm/kho.rst
> @@ -113,3 +113,42 @@ stabilized.
> ``/sys/kernel/debug/kho/in/sub_fdts/``
> Similar to ``kho/out/sub_fdts/``, but contains sub FDT blobs
> of KHO producers passed from the old kernel.
> +
> +Kexec Metadata
> +==============
I'd move this section before "debugfs Interfaces", other than that LGTM.
> +
> +KHO automatically tracks metadata about the kexec chain, passing
> information
> +about the previous kernel to the next kernel. This feature helps
> diagnose
> +bugs that only reproduce when kexecing from specific kernel versions.
...
> > > static __init int kho_init(void)
> > > {
> > > const void *fdt = kho_get_fdt();
> > > @@ -1357,6 +1413,15 @@ static __init int kho_init(void)
> > > if (err)
> > > goto err_free_fdt;
> > >
> > > + if (fdt)
> > > + kho_process_kexec_metadata();
> >
> > Can't we move it into the existing if (fdt) below?
>
> Unfortunately, that won't work due to a data dependency between the two
> functions.
>
> kho_process_kexec_metadata() reads from the FDT subtree and populates kho_in:
>
> Basically:
>
> kho_in.kexec_count = metadata->kexec_count;
>
> While kho_populate_kexec_metadata() increments metadata->kexec_count:
>
> /* kho_in.kexec_count is set to 0 on cold boot */
> metadata->kexec_count = kho_in.kexec_count + 1;
>
> If kho_process_kexec_metadata() is moved after kho_populate_kexec_metadata(),
> the count would always increment from 0 to 1, ignoring whatever was passed in
> the FDT.
>
> Restructuring to call kho_in_debugfs_init() earlier also doesn't work:
>
>
> if (fdt) {
> kho_in_debugfs_init(&kho_in.dbg, fdt);
> kho_process_kexec_metadata();
> return 0;
> }
>
> /* Populate kexec metadata for the possible next kexec */
> err = kho_populate_kexec_metadata();
> if (err)
> pr_warn("failed to initialize kexec-metadata subtree: %d\n",
> err);
>
> This would return early without populating the kexec metadata for the next
> kexec, breaking the chain on KHO boots.
How about we rename kho_process_kexec_metadata() to
kho_retreive_kexec_metadata() and add kho_process_kexec_metadata() that
will first call _retrieve and then _populate? Something like
static int __init kho_process_kexec_metadata(const void *fdt)
{
int err;
if (fdt)
kho_retrieve_kexec_metadata();
/* Populate kexec metadata for the possible next kexec */
err = kho_populate_kexec_metadata();
if (err)
pr_warn("failed to initialize kexec-metadata subtree: %d\n",
err);
return err;
}
> --breno
--
Sincerely yours,
Mike.
