1. Patchew
  2. linux
  3. View series

[PATCH v3] arm64/kdump: pass dm-crypt keys to kdump kernel

Coiby Xu posted 1 patch 2 weeks, 1 day ago 
arch/arm64/kernel/machine_kexec_file.c |  9 +++++++++
drivers/of/fdt.c                       | 21 +++++++++++++++++++++
drivers/of/kexec.c                     | 19 +++++++++++++++++++
3 files changed, 49 insertions(+)
[PATCH v3] arm64/kdump: pass dm-crypt keys to kdump kernel
Posted by Coiby Xu 2 weeks, 1 day ago 
CONFIG_CRASH_DM_CRYPT has been introduced to support LUKS-encrypted
device dump target by addressing two challenges [1],
 - Kdump kernel may not be able to decrypt the LUKS partition. For some
   machines, a system administrator may not have a chance to enter the
   password to decrypt the device in kdump initramfs after the 1st kernel
   crashes

 - LUKS2 by default use the memory-hard Argon2 key derivation function
   which is quite memory-consuming compared to the limited memory reserved
   for kdump.

To also enable this feature for ARM64, we only need to add device tree
property dmcryptkeys [2] as similar to elfcorehdr to pass the memory
address of the stored info of dm-crypt keys to the kdump kernel. Since
this property is only needed by the kdump kenrel, it won't be exposed to
user space.

[1] https://lore.kernel.org/all/20250502011246.99238-1-coxu@redhat.com/
[2] https://github.com/devicetree-org/dt-schema/pull/181

Cc: Arnaud Lefebvre <arnaud.lefebvre@clever-cloud.com>
Cc: Baoquan he <bhe@redhat.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Kairui Song <ryncsn@gmail.com>
Cc: Pingfan Liu <kernelfans@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Krzysztof Kozlowski <krzk@kernel.org>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: Coiby Xu <coxu@redhat.com>
---
v3
- Delete the property after reading it [Rob Herring]

v2
- Krzysztof
  - Use imperative mood for commit message
  - Add dt-schema ABI Documentation 
    https://github.com/devicetree-org/dt-schema/pull/181
- Don't print dm-crypt keys address via pr_debug


 arch/arm64/kernel/machine_kexec_file.c |  9 +++++++++
 drivers/of/fdt.c                       | 21 +++++++++++++++++++++
 drivers/of/kexec.c                     | 19 +++++++++++++++++++
 3 files changed, 49 insertions(+)

diff --git a/arch/arm64/kernel/machine_kexec_file.c b/arch/arm64/kernel/machine_kexec_file.c
index 410060ebd86d..5f3bad8ca96d 100644
--- a/arch/arm64/kernel/machine_kexec_file.c
+++ b/arch/arm64/kernel/machine_kexec_file.c
@@ -134,6 +134,15 @@ int load_other_segments(struct kimage *image,
 
 		kexec_dprintk("Loaded elf core header at 0x%lx bufsz=0x%lx memsz=0x%lx\n",
 			      image->elf_load_addr, kbuf.bufsz, kbuf.memsz);
+
+		ret = crash_load_dm_crypt_keys(image);
+
+		if (ret == -ENOENT) {
+			kexec_dprintk("No dm crypt key to load\n");
+		} else if (ret) {
+			pr_err("Failed to load dm crypt keys\n");
+			goto out_err;
+		}
 	}
 #endif
 
diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
index 331646d667b9..2967e4aff807 100644
--- a/drivers/of/fdt.c
+++ b/drivers/of/fdt.c
@@ -866,6 +866,26 @@ static void __init early_init_dt_check_for_elfcorehdr(unsigned long node)
 		 elfcorehdr_addr, elfcorehdr_size);
 }
 
+static void __init early_init_dt_check_for_dmcryptkeys(unsigned long node)
+{
+	const char *prop_name = "linux,dmcryptkeys";
+	const __be32 *prop;
+
+	if (!IS_ENABLED(CONFIG_CRASH_DM_CRYPT))
+		return;
+
+	pr_debug("Looking for dmcryptkeys property... ");
+
+	prop = of_get_flat_dt_prop(node, prop_name, NULL);
+	if (!prop)
+		return;
+
+	dm_crypt_keys_addr = dt_mem_next_cell(dt_root_addr_cells, &prop);
+
+	/* Property only accessible to crash dump kernel */
+	fdt_delprop(initial_boot_params, node, prop_name);
+}
+
 static unsigned long chosen_node_offset = -FDT_ERR_NOTFOUND;
 
 /*
@@ -1097,6 +1117,7 @@ int __init early_init_dt_scan_chosen(char *cmdline)
 
 	early_init_dt_check_for_initrd(node);
 	early_init_dt_check_for_elfcorehdr(node);
+	early_init_dt_check_for_dmcryptkeys(node);
 
 	rng_seed = of_get_flat_dt_prop(node, "rng-seed", &l);
 	if (rng_seed && l > 0) {
diff --git a/drivers/of/kexec.c b/drivers/of/kexec.c
index 1ee2d31816ae..4bfb1ea5744e 100644
--- a/drivers/of/kexec.c
+++ b/drivers/of/kexec.c
@@ -432,6 +432,25 @@ void *of_kexec_alloc_and_setup_fdt(const struct kimage *image,
 		if (ret)
 			goto out;
 
+		if (image->dm_crypt_keys_addr != 0) {
+			ret = fdt_appendprop_addrrange(fdt, 0, chosen_node,
+						       "linux,dmcryptkeys",
+						       image->dm_crypt_keys_addr,
+						       image->dm_crypt_keys_sz);
+
+			if (ret)
+				goto out;
+
+			/*
+			 * Avoid dmcryptkeys from being stomped on in kdump kernel by
+			 * setting up memory reserve map.
+			 */
+			ret = fdt_add_mem_rsv(fdt, image->dm_crypt_keys_addr,
+					      image->dm_crypt_keys_sz);
+			if (ret)
+				goto out;
+		}
+
 #ifdef CONFIG_CRASH_DUMP
 		/* add linux,usable-memory-range */
 		ret = fdt_appendprop_addrrange(fdt, 0, chosen_node,

base-commit: c072629f05d7bca1148ab17690d7922a31423984
-- 
2.52.0
Re: [PATCH v3] arm64/kdump: pass dm-crypt keys to kdump kernel
Posted by Will Deacon 1 week, 4 days ago 
On Fri, Jan 23, 2026 at 04:13:25PM +0800, Coiby Xu wrote:
> CONFIG_CRASH_DM_CRYPT has been introduced to support LUKS-encrypted
> device dump target by addressing two challenges [1],
>  - Kdump kernel may not be able to decrypt the LUKS partition. For some
>    machines, a system administrator may not have a chance to enter the
>    password to decrypt the device in kdump initramfs after the 1st kernel
>    crashes
> 
>  - LUKS2 by default use the memory-hard Argon2 key derivation function
>    which is quite memory-consuming compared to the limited memory reserved
>    for kdump.
> 
> To also enable this feature for ARM64, we only need to add device tree
> property dmcryptkeys [2] as similar to elfcorehdr to pass the memory
> address of the stored info of dm-crypt keys to the kdump kernel. Since
> this property is only needed by the kdump kenrel, it won't be exposed to
> user space.
> 
> [1] https://lore.kernel.org/all/20250502011246.99238-1-coxu@redhat.com/
> [2] https://github.com/devicetree-org/dt-schema/pull/181
> 
> Cc: Arnaud Lefebvre <arnaud.lefebvre@clever-cloud.com>
> Cc: Baoquan he <bhe@redhat.com>
> Cc: Dave Young <dyoung@redhat.com>
> Cc: Kairui Song <ryncsn@gmail.com>
> Cc: Pingfan Liu <kernelfans@gmail.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Krzysztof Kozlowski <krzk@kernel.org>
> Cc: Rob Herring <robh@kernel.org>
> Signed-off-by: Coiby Xu <coxu@redhat.com>
> ---
> v3
> - Delete the property after reading it [Rob Herring]
> 
> v2
> - Krzysztof
>   - Use imperative mood for commit message
>   - Add dt-schema ABI Documentation 
>     https://github.com/devicetree-org/dt-schema/pull/181
> - Don't print dm-crypt keys address via pr_debug
> 
> 
>  arch/arm64/kernel/machine_kexec_file.c |  9 +++++++++
>  drivers/of/fdt.c                       | 21 +++++++++++++++++++++
>  drivers/of/kexec.c                     | 19 +++++++++++++++++++
>  3 files changed, 49 insertions(+)
> 
> diff --git a/arch/arm64/kernel/machine_kexec_file.c b/arch/arm64/kernel/machine_kexec_file.c
> index 410060ebd86d..5f3bad8ca96d 100644
> --- a/arch/arm64/kernel/machine_kexec_file.c
> +++ b/arch/arm64/kernel/machine_kexec_file.c
> @@ -134,6 +134,15 @@ int load_other_segments(struct kimage *image,
>  
>  		kexec_dprintk("Loaded elf core header at 0x%lx bufsz=0x%lx memsz=0x%lx\n",
>  			      image->elf_load_addr, kbuf.bufsz, kbuf.memsz);
> +
> +		ret = crash_load_dm_crypt_keys(image);
> +
> +		if (ret == -ENOENT) {
> +			kexec_dprintk("No dm crypt key to load\n");
> +		} else if (ret) {
> +			pr_err("Failed to load dm crypt keys\n");
> +			goto out_err;
> +		}

This looks like an unusual mixture of kexec_dprintk() and pr_err().

Stepping back a second, why do we need to print anything from the arch
code at all? It looks like crash_load_dm_crypt_keys() already prints for
the -ENOENT case so I'd be inclined just to do:

	ret = crash_load_dm_crypt_keys(image);
	if (ret)
		goto out_err;

Will
Re: [PATCH v3] arm64/kdump: pass dm-crypt keys to kdump kernel
Posted by Coiby Xu 1 week, 1 day ago 
On Mon, Jan 26, 2026 at 02:36:08PM +0000, Will Deacon wrote:
>On Fri, Jan 23, 2026 at 04:13:25PM +0800, Coiby Xu wrote:
>> CONFIG_CRASH_DM_CRYPT has been introduced to support LUKS-encrypted
>> device dump target by addressing two challenges [1],
>>  - Kdump kernel may not be able to decrypt the LUKS partition. For some
>>    machines, a system administrator may not have a chance to enter the
>>    password to decrypt the device in kdump initramfs after the 1st kernel
>>    crashes
>>
>>  - LUKS2 by default use the memory-hard Argon2 key derivation function
>>    which is quite memory-consuming compared to the limited memory reserved
>>    for kdump.
>>
>> To also enable this feature for ARM64, we only need to add device tree
>> property dmcryptkeys [2] as similar to elfcorehdr to pass the memory
>> address of the stored info of dm-crypt keys to the kdump kernel. Since
>> this property is only needed by the kdump kenrel, it won't be exposed to
>> user space.
>>
>> [1] https://lore.kernel.org/all/20250502011246.99238-1-coxu@redhat.com/
>> [2] https://github.com/devicetree-org/dt-schema/pull/181
>>
>> Cc: Arnaud Lefebvre <arnaud.lefebvre@clever-cloud.com>
>> Cc: Baoquan he <bhe@redhat.com>
>> Cc: Dave Young <dyoung@redhat.com>
>> Cc: Kairui Song <ryncsn@gmail.com>
>> Cc: Pingfan Liu <kernelfans@gmail.com>
>> Cc: Andrew Morton <akpm@linux-foundation.org>
>> Cc: Krzysztof Kozlowski <krzk@kernel.org>
>> Cc: Rob Herring <robh@kernel.org>
>> Signed-off-by: Coiby Xu <coxu@redhat.com>
>> ---
>> v3
>> - Delete the property after reading it [Rob Herring]
>>
>> v2
>> - Krzysztof
>>   - Use imperative mood for commit message
>>   - Add dt-schema ABI Documentation
>>     https://github.com/devicetree-org/dt-schema/pull/181
>> - Don't print dm-crypt keys address via pr_debug
>>
>>
>>  arch/arm64/kernel/machine_kexec_file.c |  9 +++++++++
>>  drivers/of/fdt.c                       | 21 +++++++++++++++++++++
>>  drivers/of/kexec.c                     | 19 +++++++++++++++++++
>>  3 files changed, 49 insertions(+)
>>
>> diff --git a/arch/arm64/kernel/machine_kexec_file.c b/arch/arm64/kernel/machine_kexec_file.c
>> index 410060ebd86d..5f3bad8ca96d 100644
>> --- a/arch/arm64/kernel/machine_kexec_file.c
>> +++ b/arch/arm64/kernel/machine_kexec_file.c
>> @@ -134,6 +134,15 @@ int load_other_segments(struct kimage *image,
>>
>>  		kexec_dprintk("Loaded elf core header at 0x%lx bufsz=0x%lx memsz=0x%lx\n",
>>  			      image->elf_load_addr, kbuf.bufsz, kbuf.memsz);
>> +
>> +		ret = crash_load_dm_crypt_keys(image);
>> +
>> +		if (ret == -ENOENT) {
>> +			kexec_dprintk("No dm crypt key to load\n");
>> +		} else if (ret) {
>> +			pr_err("Failed to load dm crypt keys\n");
>> +			goto out_err;
>> +		}
>
>This looks like an unusual mixture of kexec_dprintk() and pr_err().
>
>Stepping back a second, why do we need to print anything from the arch
>code at all? It looks like crash_load_dm_crypt_keys() already prints for
>the -ENOENT case so I'd be inclined just to do:
>
>	ret = crash_load_dm_crypt_keys(image);
>	if (ret)
>		goto out_err;
>
>Will
>

Hi Will,

Thanks for carefully reviewing the patch! Yeah, crash_load_dm_crypt_keys
already prints for the -ENOTENT case, good catch! And it's also a good
idea to not let the arch code print anything since
crash_load_dm_crypt_keys is a better place. I'll make
crash_load_dm_crypt_keys print more logs and also return 0 for the case
of -ENOENT. This can make the arch code more succinct.


-- 
Best regards,
Coiby

Patchew 2.1-devel-b67e4c2

© 2016 - 2026 Red Hat, Inc.