Re: [PATCH 5/6] kho: Relocate vmalloc preservation structure to KHO ABI header

[Mike Rapoport <rppt@kernel.org>]

On Tue, Jan 20, 2026 at 04:26:51PM +0000, Pratyush Yadav wrote:
> On Mon, Jan 05 2026, Mike Rapoport wrote:
> 
> > From: Jason Miu <jasonmiu@google.com>
> >
> > The `struct kho_vmalloc` defines the in-memory layout for preserving
> > vmalloc regions across kexec. This layout is a contract between kernels
> > and part of the KHO ABI.
> >
> > To reflect this relationship, the related structs and helper macros are
> > relocated to the ABI header, `include/linux/kho/abi/kexec_handover.h`.
> > This move places the structure's definition under the protection of the
> > KHO_FDT_COMPATIBLE version string.
> >
> > The structure and its components are now also documented within the
> > ABI header to describe the contract and prevent ABI breaks.
> >
> > Signed-off-by: Jason Miu <jasonmiu@google.com>
> > Co-developed-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
> > Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
> [...]
> > +/* Helper macro to define a union for a serializable pointer. */
> > +#define DECLARE_KHOSER_PTR(name, type)	\
> > +	union {                        \
> > +		u64 phys;              \
> > +		type ptr;              \
> > +	} name
> > +
> > +/* Stores the physical address of a serializable pointer. */
> > +#define KHOSER_STORE_PTR(dest, val)               \
> > +	({                                        \
> > +		typeof(val) v = val;              \
> > +		typecheck(typeof((dest).ptr), v); \
> > +		(dest).phys = virt_to_phys(v);    \
> > +	})
> > +
> > +/* Loads the stored physical address back to a pointer. */
> > +#define KHOSER_LOAD_PTR(src)						\
> > +	({                                                                   \
> > +		typeof(src) s = src;                                         \
> > +		(typeof((s).ptr))((s).phys ? phys_to_virt((s).phys) : NULL); \
> > +	})
> 
> Nit: not a fan of exposing code internals to the ABI header. But without
> this the definition of kho_vmalloc_hdr won't make any sense to someone
> reading the doc without looking at the code. Dunno if we can do anything
> better though...

These might be actually useful for other KHO users. 

> > +/*
> > + * This header is embedded at the beginning of each `kho_vmalloc_chunk`
> > + * and contains a pointer to the next chunk in the linked list,
> > + * stored as a physical address for handover.
> > + */
> > +struct kho_vmalloc_hdr {
> > +	DECLARE_KHOSER_PTR(next, struct kho_vmalloc_chunk *);
> > +};
> > +
> > +#define KHO_VMALLOC_SIZE				\
> > +	((PAGE_SIZE - sizeof(struct kho_vmalloc_hdr)) / \
> > +	 sizeof(u64))
> > +
> > +/*
> > + * Each chunk is a single page and is part of a linked list that describes
> > + * a preserved vmalloc area. It contains the header with the link to the next
> > + * chunk and an array of physical addresses of the pages that make up the
> > + * preserved vmalloc area.
> 
> Perhaps also mention that the array is 0-terminated?

Yeah, sure.

@Andrew, can you please add this as a fixup:

diff --git a/include/linux/kho/abi/kexec_handover.h b/include/linux/kho/abi/kexec_handover.h
index 285eda8a36e4..2201a0d2c159 100644
--- a/include/linux/kho/abi/kexec_handover.h
+++ b/include/linux/kho/abi/kexec_handover.h
@@ -138,8 +138,8 @@ struct kho_vmalloc_hdr {
 /*
  * Each chunk is a single page and is part of a linked list that describes
  * a preserved vmalloc area. It contains the header with the link to the next
- * chunk and an array of physical addresses of the pages that make up the
- * preserved vmalloc area.
+ * chunk and a zero terminated array of physical addresses of the pages that
+ * make up the preserved vmalloc area.
  */
 struct kho_vmalloc_chunk {
 	struct kho_vmalloc_hdr hdr;


> Looks good otherwise.
> -- 
> Regards,
> Pratyush Yadav

-- 
Sincerely yours,
Mike.