Skip to contentSkip to conversation

ARM: imx: hab: panic on authentication failure

PATCH New Marek Vasut 4 participants 10 messages May 30, 2020 – Aug 4, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>
---
 arch/arm/mach-imx/spl.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
arch/arm/mach-imx/spl.c +2 -4
index 1a231c67f5..76a5f7aca6 100644
@@ -293,8 +293,7 @@ __weak void __noreturn jump_to_image_no_args(struct spl_image_info *spl_image)
 						CSF_PAD_SIZE, offset)) {
 			image_entry();
 		} else {
-			puts("spl: ERROR:  image authentication fail\n");
-			hang();
+			panic("spl: ERROR:  image authentication fail\n");
 		}
 	}
 }
@@ -320,8 +319,7 @@ void board_spl_fit_post_load(ulong load_addr, size_t length)
 	if (imx_hab_authenticate_image(load_addr,
 				       offset + IVT_SIZE + CSF_PAD_SIZE,
 				       offset)) {
-		puts("spl: ERROR:  image authentication unsuccessful\n");
-		hang();
+		panic("spl: ERROR:  image authentication unsuccessful\n");
 	}
 }
 #endif
Signature · 1 line
2.25.1
reviewer
#1 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Hi Marek,

[Adding Breno]

Quoted 11 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 5 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>

This is a better behavior indeed:

#2 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Quoted 22 lines · Fabio Estevam

Hi Marek,

[Adding Breno]

Quoted 11 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 5 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>
Quoted 3 lines

This is a better behavior indeed:

What about this? Have you ignored this patch for a reason? :/

https://marc.info/?l=u-boot&m=159069441005730&w=2

#3 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Quoted 23 lines · Patrick Wildt
Quoted 22 lines · Fabio Estevam

Hi Marek,

[Adding Breno]

Quoted 16 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>
Quoted 7 lines
Quoted 3 lines

This is a better behavior indeed:

What about this? Have you ignored this patch for a reason? :/

https://marc.info/?l=u-boot&m=159069441005730&w=2

Yes, and the reason is I was not even aware of your patch, sorry. The CC list in this mail should cover all the interested parties, so use it when sending V2, or use patman.

The patch looks fine, one nit is that you should return errno.h return value and another is that it changes the current behavior. Now that I look at this imx code, board_spl_fit_post_load() should not even be in arch/ , sigh, but that's for separate patch either way.

So I think if you want to support this sort of fallback, you should make the board_spl_fit_post_load() be in board/ files, with default __weak implementation calling some arch_hab_authenticate...() which implements current content of board_spl_fit_post_load(), and let boards decide how to handle the fallback if it needs to be altered.

Would that work ?

#4 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Quoted 28 lines · Marek Vasut
Quoted 27 lines · Patrick Wildt
Quoted 25 lines · Fabio Estevam

Hi Marek,

[Adding Breno]

Quoted 11 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 5 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>

This is a better behavior indeed:

Quoted 7 lines
Quoted 3 lines

What about this? Have you ignored this patch for a reason? :/

https://marc.info/?l=u-boot&m=159069441005730&w=2

Yes, and the reason is I was not even aware of your patch, sorry. The CC list in this mail should cover all the interested parties, so use it when sending V2, or use patman.

I already had 11 people on CC, but apparently I missed you.

Quoted 5 lines

The patch looks fine, one nit is that you should return errno.h return value and another is that it changes the current behavior. Now that I look at this imx code, board_spl_fit_post_load() should not even be in arch/ , sigh, but that's for separate patch either way.

Quoted 7 lines

So I think if you want to support this sort of fallback, you should make the board_spl_fit_post_load() be in board/ files, with default __weak implementation calling some arch_hab_authenticate...() which implements current content of board_spl_fit_post_load(), and let boards decide how to handle the fallback if it needs to be altered.

Would that work ?

I'm not sure. In comparison to the people from NXP who are paid to upstream their code and still don't do it correctly, I'm doing this in my spare time and I'm not sure I want to bikeshed all day long.

I can send a V3 that replaces the -1 with EINVAL, EACCESS, EPERM or something the like. If you want to clean up after NXP, feel free to.

Which errno would you like to see?

Best regards,
Patrick

#5 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Quoted 50 lines · Patrick Wildt
Quoted 28 lines · Marek Vasut
Quoted 27 lines · Patrick Wildt
Quoted 25 lines · Fabio Estevam

Hi Marek,

[Adding Breno]

Quoted 11 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 5 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>

This is a better behavior indeed:

Quoted 7 lines
Quoted 3 lines

What about this? Have you ignored this patch for a reason? :/

https://marc.info/?l=u-boot&m=159069441005730&w=2

Yes, and the reason is I was not even aware of your patch, sorry. The CC list in this mail should cover all the interested parties, so use it when sending V2, or use patman.

I already had 11 people on CC, but apparently I missed you.

Quoted 11 lines

The patch looks fine, one nit is that you should return errno.h return value and another is that it changes the current behavior. Now that I look at this imx code, board_spl_fit_post_load() should not even be in arch/ , sigh, but that's for separate patch either way.

So I think if you want to support this sort of fallback, you should make the board_spl_fit_post_load() be in board/ files, with default __weak implementation calling some arch_hab_authenticate...() which implements current content of board_spl_fit_post_load(), and let boards decide how to handle the fallback if it needs to be altered.

Quoted 8 lines
Quoted 1 line

Would that work ?

I'm not sure. In comparison to the people from NXP who are paid to upstream their code and still don't do it correctly, I'm doing this in my spare time and I'm not sure I want to bikeshed all day long.

I can send a V3 that replaces the -1 with EINVAL, EACCESS, EPERM or something the like. If you want to clean up after NXP, feel free to.

In fact, what is it that you're trying to achieve with this fallback ? What are you falling back to , another fallback fitImage ?

#6 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Quoted 57 lines · Marek Vasut
Quoted 56 lines · Patrick Wildt
Quoted 28 lines · Marek Vasut
Quoted 27 lines · Patrick Wildt
Quoted 25 lines · Fabio Estevam

Hi Marek,

[Adding Breno]

Quoted 11 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 5 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>

This is a better behavior indeed:

Quoted 7 lines
Quoted 3 lines

What about this? Have you ignored this patch for a reason? :/

https://marc.info/?l=u-boot&m=159069441005730&w=2

Yes, and the reason is I was not even aware of your patch, sorry. The CC list in this mail should cover all the interested parties, so use it when sending V2, or use patman.

I already had 11 people on CC, but apparently I missed you.

Quoted 5 lines

The patch looks fine, one nit is that you should return errno.h return value and another is that it changes the current behavior. Now that I look at this imx code, board_spl_fit_post_load() should not even be in arch/ , sigh, but that's for separate patch either way.

Quoted 7 lines

So I think if you want to support this sort of fallback, you should make the board_spl_fit_post_load() be in board/ files, with default __weak implementation calling some arch_hab_authenticate...() which implements current content of board_spl_fit_post_load(), and let boards decide how to handle the fallback if it needs to be altered.

Would that work ?

I'm not sure. In comparison to the people from NXP who are paid to upstream their code and still don't do it correctly, I'm doing this in my spare time and I'm not sure I want to bikeshed all day long.

Quoted 5 lines
Quoted 2 lines

I can send a V3 that replaces the -1 with EINVAL, EACCESS, EPERM or something the like. If you want to clean up after NXP, feel free to.

In fact, what is it that you're trying to achieve with this fallback ? What are you falling back to , another fallback fitImage ?

Exactly.

I have a device with a glued enclosure, with no external sources, apart from a serial console. If the SPL fails to load the U-Boot main image from eMMC, the only way to fix it is to destroy the case, open up the enclosure and short some lines. That's not really nice, since we'd have to get a new enclosure, new serial number label,... it's a hassle.

If the SPL was gone as well, the BootROM would fall back to other sources. But with only one half of U-Boot missing, it would just hang. I'm sure that's also why you replace the hang() with a panic().

I thought that if the SPL is still fine, but only the U-Boot image was gone, why not make use of the spl_boot_list to try and boot from another source? Like yModem. For that I sent the following fix, also with many people CCd:

https://marc.info/?l=u-boot&m=158893200030861&w=2

Now the board spl boot order can have eMMC as first, and yModem as second. If eMMC fails, it falls back to yModem. If none work, it though hang()s, doing

	puts(SPL_TPL_PROMPT "failed to boot from all boot devices\n");
	hang();

Maybe you want this as panic instead?

Anyway, I think this is nicer option for recovery, instead of simply hanging or rebooting.

Patrick

#7 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Quoted 64 lines · Patrick Wildt
Quoted 57 lines · Marek Vasut
Quoted 56 lines · Patrick Wildt
Quoted 28 lines · Marek Vasut
Quoted 27 lines · Patrick Wildt
Quoted 25 lines · Fabio Estevam

Hi Marek,

[Adding Breno]

Quoted 11 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 5 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>

This is a better behavior indeed:

Quoted 7 lines
Quoted 3 lines

What about this? Have you ignored this patch for a reason? :/

https://marc.info/?l=u-boot&m=159069441005730&w=2

Yes, and the reason is I was not even aware of your patch, sorry. The CC list in this mail should cover all the interested parties, so use it when sending V2, or use patman.

I already had 11 people on CC, but apparently I missed you.

Quoted 5 lines

The patch looks fine, one nit is that you should return errno.h return value and another is that it changes the current behavior. Now that I look at this imx code, board_spl_fit_post_load() should not even be in arch/ , sigh, but that's for separate patch either way.

Quoted 7 lines

So I think if you want to support this sort of fallback, you should make the board_spl_fit_post_load() be in board/ files, with default __weak implementation calling some arch_hab_authenticate...() which implements current content of board_spl_fit_post_load(), and let boards decide how to handle the fallback if it needs to be altered.

Would that work ?

I'm not sure. In comparison to the people from NXP who are paid to upstream their code and still don't do it correctly, I'm doing this in my spare time and I'm not sure I want to bikeshed all day long.

Quoted 5 lines
Quoted 2 lines

I can send a V3 that replaces the -1 with EINVAL, EACCESS, EPERM or something the like. If you want to clean up after NXP, feel free to.

In fact, what is it that you're trying to achieve with this fallback ? What are you falling back to , another fallback fitImage ?

Quoted 7 lines

Exactly.

I have a device with a glued enclosure, with no external sources, apart from a serial console. If the SPL fails to load the U-Boot main image from eMMC, the only way to fix it is to destroy the case, open up the enclosure and short some lines. That's not really nice, since we'd have to get a new enclosure, new serial number label,... it's a hassle.

Look for SRC PERSIST_SECONDARY_BOOT in your datasheet then. This will let you use two copies of SPL, two copies of U-Boot, etc.

Quoted 15 lines

If the SPL was gone as well, the BootROM would fall back to other sources. But with only one half of U-Boot missing, it would just hang. I'm sure that's also why you replace the hang() with a panic().

I thought that if the SPL is still fine, but only the U-Boot image was gone, why not make use of the spl_boot_list to try and boot from another source? Like yModem. For that I sent the following fix, also with many people CCd:

https://marc.info/?l=u-boot&m=158893200030861&w=2

Now the board spl boot order can have eMMC as first, and yModem as second. If eMMC fails, it falls back to yModem. If none work, it though hang()s, doing

Quoted 7 lines
	puts(SPL_TPL_PROMPT "failed to boot from all boot devices\n");
	hang();

Maybe you want this as panic instead?

Anyway, I think this is nicer option for recovery, instead of simply hanging or rebooting.

The problem is, this only works for fitImage and not for raw uImage. But in your case, that is probably OK.

So if you can just fix the errno return value to some -EINVAL (?) and send a V3, I think that would be good.

#8 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

Quoted 67 lines · Marek Vasut
Quoted 66 lines · Patrick Wildt
Quoted 57 lines · Marek Vasut
Quoted 56 lines · Patrick Wildt
Quoted 28 lines · Marek Vasut
Quoted 27 lines · Patrick Wildt
Quoted 25 lines · Fabio Estevam

Hi Marek,

[Adding Breno]

Quoted 11 lines · Marek Vasut

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.

This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 5 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>

This is a better behavior indeed:

Quoted 7 lines
Quoted 3 lines

What about this? Have you ignored this patch for a reason? :/

https://marc.info/?l=u-boot&m=159069441005730&w=2

Yes, and the reason is I was not even aware of your patch, sorry. The CC list in this mail should cover all the interested parties, so use it when sending V2, or use patman.

I already had 11 people on CC, but apparently I missed you.

Quoted 5 lines

The patch looks fine, one nit is that you should return errno.h return value and another is that it changes the current behavior. Now that I look at this imx code, board_spl_fit_post_load() should not even be in arch/ , sigh, but that's for separate patch either way.

Quoted 7 lines

So I think if you want to support this sort of fallback, you should make the board_spl_fit_post_load() be in board/ files, with default __weak implementation calling some arch_hab_authenticate...() which implements current content of board_spl_fit_post_load(), and let boards decide how to handle the fallback if it needs to be altered.

Would that work ?

I'm not sure. In comparison to the people from NXP who are paid to upstream their code and still don't do it correctly, I'm doing this in my spare time and I'm not sure I want to bikeshed all day long.

Quoted 5 lines
Quoted 2 lines

I can send a V3 that replaces the -1 with EINVAL, EACCESS, EPERM or something the like. If you want to clean up after NXP, feel free to.

In fact, what is it that you're trying to achieve with this fallback ? What are you falling back to , another fallback fitImage ?

Exactly.

Quoted 8 lines
Quoted 5 lines

I have a device with a glued enclosure, with no external sources, apart from a serial console. If the SPL fails to load the U-Boot main image from eMMC, the only way to fix it is to destroy the case, open up the enclosure and short some lines. That's not really nice, since we'd have to get a new enclosure, new serial number label,... it's a hassle.

Look for SRC PERSIST_SECONDARY_BOOT in your datasheet then. This will let you use two copies of SPL, two copies of U-Boot, etc.

I'll have a look!

Quoted 20 lines
Quoted 20 lines

If the SPL was gone as well, the BootROM would fall back to other sources. But with only one half of U-Boot missing, it would just hang. I'm sure that's also why you replace the hang() with a panic().

I thought that if the SPL is still fine, but only the U-Boot image was gone, why not make use of the spl_boot_list to try and boot from another source? Like yModem. For that I sent the following fix, also with many people CCd:

https://marc.info/?l=u-boot&m=158893200030861&w=2

Now the board spl boot order can have eMMC as first, and yModem as second. If eMMC fails, it falls back to yModem. If none work, it though hang()s, doing

	puts(SPL_TPL_PROMPT "failed to boot from all boot devices\n");
	hang();

Maybe you want this as panic instead?

Quoted 8 lines
Quoted 2 lines

Anyway, I think this is nicer option for recovery, instead of simply hanging or rebooting.

The problem is, this only works for fitImage and not for raw uImage. But in your case, that is probably OK.

So if you can just fix the errno return value to some -EINVAL (?) and send a V3, I think that would be good.

Ok, will do, thanks!

#9 lore raw

PATCH 1/1 ARM: imx: hab: panic on authentication failure

In reply to Marek Vasut
Quoted 8 lines

Instead of hang()ing the system and thus disallowing any automated recovery possibility from a HAB authentication failure, panic() . The panic() function can be configured to hang() the system after printing an error message, however the default is to reset the system instead.
This allows redundant boot to work correctly. In case the primary or secondary image cannot be authenticated, the system reboots and bootrom can try to start the other one.

Quoted 6 lines
Cc Fabio Estevam <[email protected]>
Cc NXP i.MX U-Boot Team <[email protected]>
Cc Peng Fan <[email protected]>
Cc Stefano Babic <[email protected]>

Applied to u-boot-imx, master, thanks !

Best regards,
Stefano Babic

Signature · 5 lines
=====================================================================
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sbabic at denx.de
=====================================================================