Skip to contentSkip to conversation

habv4: add the possibility to changing the signing area from Kconfig

PATCH In review Maik Otto 2 participants 6 messages Dec 10, 2019 – Dec 13, 2019

Subsystems (3)
lore
Share & tools
lore

Maik Otto

author
#0 lore raw

PATCH 1/1 habv4: add the possibility to changing the signing area from Kconfig

the whole barebox with mbr and partition table will be signed by default add the possibility in the Kconfig to change from full signing to skip-mbr and from-dcdofs

---
 arch/arm/mach-imx/Kconfig                          | 26 ++++++++++++++++++++++
 arch/arm/mach-imx/include/mach/habv4-imx6-gencsf.h |  2 +-
 scripts/imx/imx.c                                  |  6 ++---
 3 files changed, 30 insertions(+), 4 deletions(-)
arch/arm/mach-imx/Kconfig +26 -0
index 6e98e95..ca1e6f3 100644
@@ -859,6 +859,32 @@ config HABV4_IMG_CRT_PEM
 
 endif
 
+if HABV4
+choice
+	prompt "Select type for signing area"
+	help
+	  select the signing area, which is checked from the HAB modul
+	  of the i.mx controller
+
+config HABV4_BLOCKS_TYPE_FULL
+	bool "complete barebox with mbr and partition table"
+
+config HABV4_BLOCKS_TYPE_FROM_DCDOFS
+	bool "signing area begins from dcdofs address"
+
+config HABV4_BLOCKS_TYPE_SKIP_MBR
+	bool "signing area begins after MBR"
+
+endchoice
+
+endif
+
+config HABV4_BLOCKS_TYPE
+	string
+	default skip-mbr if HABV4_BLOCKS_TYPE_SKIP_MBR
+	default from-dcdofs if HABV4_BLOCKS_TYPE_FROM_DCDOFS
+	default full
+
 config HABV3
 	tristate "HABv3 support"
 	select HAB
arch/arm/mach-imx/include/mach/habv4-imx6-gencsf.h +1 -1
index 5818879..60d3baf 100644
@@ -42,7 +42,7 @@ hab [Authenticate Data]
 /* verification key index in key store (2...4) */
 hab Verification index = 2
 
-hab_blocks
+hab_blocks CONFIG_HABV4_BLOCKS_TYPE
 
 hab_encrypt [Install Secret Key]
 hab_encrypt Verification index = 0
scripts/imx/imx.c +3 -3
index b3e8d62..4a96d8d 100644
@@ -365,17 +365,17 @@ static int do_hab_blocks(struct config_data *data, int argc, char *argv[])
 			offset += PLUGIN_HDMI_SIZE;
 	}
 
-	if (!strcmp(type, "full")) {
+	if (strstr(type, "full")) {
 		ret = asprintf(&str, "Blocks = 0x%08x 0x%08x 0x%08x \"%s\"\n",
 			       data->image_load_addr, offset, signed_size,
 			       data->outfile);
-	} else if (!strcmp(type, "from-dcdofs")) {
+	} else if (strstr(type, "from-dcdofs")) {
 		ret = asprintf(&str, "Blocks = 0x%08x 0x%x %d \"%s\"\n",
 			       data->image_load_addr + data->image_dcd_offset,
 			       data->image_dcd_offset,
 			       signed_size - data->image_dcd_offset,
 			       data->outfile);
-	} else if (!strcmp(type, "skip-mbr")) {
+	} else if (strstr(type, "skip-mbr")) {
 		ret = asprintf(&str,
 			       "Blocks = 0x%08x 0 440 \"%s\", \\\n"
 			       "         0x%08x 512 %d \"%s\"\n",
Signature · 7 lines
2.7.4


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
maintainer
#1 lore raw
Quoted 3 lines · Maik Otto

the whole barebox with mbr and partition table will be signed by default add the possibility in the Kconfig to change from full signing to skip-mbr and from-dcdofs

Not signing the MBR seems the right thing to do. Do we need it configurable at all or would it be better to just always skip the first KiB?

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
author
#2 lore raw

Hi Sascha,

in my opinion it is better to have it configurable, because ther are different use cases and security requirements.
i found the problem by creating  a sd-card \emmc image with wic.  The mbr, the partition table and bootloader became be signed at barebox build and wic changes
the partition table at the end of the build process. Then the sd card image could not boot , because the signature was wrong. yeah secure boot works :-)
the highest protection you have, when mbr and partition table is signed with the bootloader, but it is not always necessary.

there was implemented skip-mbr, dcdofs and full, but full was by default in the code. at the moment i think , it is a good and easy choice.

Best regards

Maik

Am 10.12.2019 um 16:21 schrieb Sascha Hauer:

Quoted 10 lines · Maik Otto
Quoted 3 lines

the whole barebox with mbr and partition table will be signed by default add the possibility in the Kconfig to change from full signing to skip-mbr and from-dcdofs

Not signing the MBR seems the right thing to do. Do we need it configurable at all or would it be better to just always skip the first KiB?

Sascha

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

maintainer
#3 lore raw
Quoted 12 lines · Maik Otto

Hi Sascha,

in my opinion it is better to have it configurable, because ther are different use cases and security requirements.
i found the problem by creating  a sd-card \emmc image with wic.  The mbr, the partition table and bootloader became be signed at barebox build and wic changes
the partition table at the end of the build process. Then the sd card image could not boot , because the signature was wrong. yeah secure boot works :-)
the highest protection you have, when mbr and partition table is signed with the bootloader, but it is not always necessary.

But in which cases is it really necessary? I can't think of any. The mbr and partition table are not evaluated by the ROM code, hence they do not need to be signed for HAB.

The images generated by the build system all do not have a partition table included, so basically we are currently enforcing no partition table at all which is just not useful.

I think the current way of including the first KiB in signed area comes from the fact that we started doing HAB on a NAND device which doesn't have a partition table. Other projects we are currently doing use eMMC where we use the boot partitions, again no MBR or partition table.

If we had started on SD cards, we wouldn't have included the partition table in the signature and also would never have thought it would be a good idea to do so.

Regards
Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
author
#4 lore raw

Hi Sascha,

so do you think we should always start from-dcdofs instead of full? at the moment i use this configuration with from-dcdofs and i think you have right, there is
not really a good case to sign the area between 0x00 and dcdofs in the barebox build
What is the best solution in your opinion?
change default from full to dcdofs in the scripts/imx/imx.c ? additional delete full and skip-mbr ?

Best regards

Maik

Am 11.12.2019 um 09:15 schrieb Sascha Hauer:

Quoted 26 lines · Maik Otto
Quoted 12 lines

Hi Sascha,

in my opinion it is better to have it configurable, because ther are different use cases and security requirements.
i found the problem by creating  a sd-card \emmc image with wic.  The mbr, the partition table and bootloader became be signed at barebox build and wic changes
the partition table at the end of the build process. Then the sd card image could not boot , because the signature was wrong. yeah secure boot works :-)
the highest protection you have, when mbr and partition table is signed with the bootloader, but it is not always necessary.

But in which cases is it really necessary? I can't think of any. The mbr and partition table are not evaluated by the ROM code, hence they do not need to be signed for HAB.

The images generated by the build system all do not have a partition table included, so basically we are currently enforcing no partition table at all which is just not useful.

I think the current way of including the first KiB in signed area comes from the fact that we started doing HAB on a NAND device which doesn't have a partition table. Other projects we are currently doing use eMMC where we use the boot partitions, again no MBR or partition table.

Quoted 7 lines

If we had started on SD cards, we wouldn't have included the partition table in the signature and also would never have thought it would be a good idea to do so.

Regards
Sascha

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

maintainer
#5 lore raw
Quoted 10 lines · Maik Otto

Hi Sascha,

so do you think we should always start from-dcdofs instead of full? at the moment i use this configuration with from-dcdofs and i think you have right, there is
not really a good case to sign the area between 0x00 and dcdofs in the barebox build
What is the best solution in your opinion?
change default from full to dcdofs in the scripts/imx/imx.c ? additional delete full and skip-mbr ?

Yes. Always only check the executed image without the 1KiB padding at the beginning. Note that on i.MX8 it's 32KiB padding until the image start.

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox