Re: [PATCH v3 2/2] efi: Support booting with kexec handover (KHO)

[Pratyush Yadav <pratyush@kernel.org>]

On Thu, Aug 21 2025, Evangelos Petrongonas wrote:

> When KHO (Kexec HandOver) is enabled, it sets up scratch memory regions
> early during device tree scanning. After kexec, the new kernel
> exclusively uses this region for memory allocations during boot up to
> the initialization of the page allocator
>
> However, when booting with EFI, EFI's reserve_regions() uses
> memblock_remove(0, PHYS_ADDR_MAX) to clear all memory regions before
> rebuilding them from EFI data. This destroys KHO scratch regions and
> their flags, thus causing a kernel panic, as there are no scratch
> memory regions.
>
> Instead of wholesale removal, iterate through memory regions and only
> remove non-KHO ones. This preserves KHO scratch regions, which are
> good known memory, while still allowing EFI to rebuild its memory map.
>
> Acked-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
> Signed-off-by: Evangelos Petrongonas <epetron@amazon.de>
> ---
> Changes in v3:
> 	- Improve the code comments, by stating that the scratch regions are
> 	good known memory
>
> Changes in v2:
> 	- Replace the for loop with for_each_mem_region
> 	- Fix comment indentation
> 	- Amend commit message to specify that scratch regions
> 	are known good regions
>
>  drivers/firmware/efi/efi-init.c | 29 +++++++++++++++++++++++++----
>  1 file changed, 25 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/firmware/efi/efi-init.c b/drivers/firmware/efi/efi-init.c
> index a00e07b853f2..a65c2d5b9e7b 100644
> --- a/drivers/firmware/efi/efi-init.c
> +++ b/drivers/firmware/efi/efi-init.c
> @@ -12,6 +12,7 @@
>  #include <linux/efi.h>
>  #include <linux/fwnode.h>
>  #include <linux/init.h>
> +#include <linux/kexec_handover.h>
>  #include <linux/memblock.h>
>  #include <linux/mm_types.h>
>  #include <linux/of.h>
> @@ -164,12 +165,32 @@ static __init void reserve_regions(void)
>  		pr_info("Processing EFI memory map:\n");
>  
>  	/*
> -	 * Discard memblocks discovered so far: if there are any at this
> -	 * point, they originate from memory nodes in the DT, and UEFI
> -	 * uses its own memory map instead.
> +	 * Discard memblocks discovered so far except for KHO scratch
> +	 * regions. Most memblocks at this point originate from memory nodes
> +	 * in the DT and UEFI uses its own memory map instead. However, if
> +	 * KHO is enabled, scratch regions, which are good known memory
> +	 * must be preserved.
>  	 */
>  	memblock_dump_all();
> -	memblock_remove(0, PHYS_ADDR_MAX);
> +
> +	if (is_kho_boot()) {
> +		struct memblock_region *r;
> +
> +		/* Remove all non-KHO regions */
> +		for_each_mem_region(r) {
> +			if (!memblock_is_kho_scratch(r)) {
> +				memblock_remove(r->base, r->size);
> +				r--;

Hmm, this caught me off-guard. I had to do a double take to realize that
memblock_remove() would decrease memblock.memory.cnt and move the whole
regions array back. A comment would have been nice here.

But then, I wouldn't want you to do a full resend of the series for this
minor nitpick. So perhaps whoever is taking this patch can add one when
applying? Either way is fine though...

Acked-by: Pratyush Yadav <pratyush@kernel.org>

> +			}
> +		}
> +	} else {
> +		/*
> +		 * KHO is disabled. Discard memblocks discovered so far:
> +		 * if there are any at this point, they originate from memory
> +		 * nodes in the DT, and UEFI uses its own memory map instead.
> +		 */
> +		memblock_remove(0, PHYS_ADDR_MAX);
> +	}
>  
>  	for_each_efi_memory_desc(md) {
>  		paddr = md->phys_addr;

-- 
Regards,
Pratyush Yadav