Skip to contentSkip to conversation

efi: x86: Make the deprecated EFI handover protocol optional

PATCH In review Ard Biesheuvel 2 participants 6 messages Oct 7, 2022 – Oct 10, 2022

Subsystems (2)
lore
Share & tools
lore

Ard Biesheuvel

author maintainer
#0 lore raw

PATCH 1/1 efi: x86: Make the deprecated EFI handover protocol optional

The EFI handover protocol permits a bootloader to invoke the kernel as a EFI PE/COFF application, while passing a bootparams struct as a third argument to the entrypoint function call.

This has no basis in the UEFI specification, and there are better ways to pass additional data to a UEFI application (UEFI configuration tables, UEFI variables, UEFI protocols) than going around the StartImage() boot service and jumping to a fixed offset in the loaded image, just to call a different function that takes a third parameter.

The reason for handling struct bootparams in the bootloader was that the EFI stub could only load initrd images from the EFI system partition, and so passing it via struct bootparams was needed for loaders like GRUB, which pass the initrd in memory, and may load it from anywhere, including from the network. Another motivation was EFI mixed mode, which could not use the initrd loader in the EFI stub at all due to 32/64 bit incompatibilities (which will be fixed shortly [0]), and could not invoke the ordinary PE/COFF entry point either, for the same reasons.

Given that loaders such as GRUB already carried the bootparams handling in order to implement non-EFI boot, retaining that code and just passing bootparams to the EFI stub was a reasonable choice (although defining an alternate entrypoint could have been avoided.) However, the GRUB side changes never made it upstream, and are only shipped by some of the distros in their downstream versions.

In the meantime, EFI support has been added to other Linux architecture ports, as well as to U-boot and systemd, including arch-agnostic methods for passing initrd images in memory [1], and for doing mixed mode boot [2], none of them requiring anything like the EFI handover protocol. So given that only out-of-tree distro GRUB relies on this, let's permit it to be omitted from the build, in preparation for retiring it completely at a later date. (Note that systemd-boot does have an implementation as well, but only uses it as a fallback for booting images that do not implement the LoadFile2 based initrd loading method, i.e., v5.8 or older)

[0] https://lore.kernel.org/all/[email protected]/
[1] ec93fc371f01 ("efi/libstub: Add support for loading the initrd ...")
[2] 97aa276579b2 ("efi/x86: Add true mixed mode entry point into ...")

---
 arch/x86/Kconfig                   | 12 ++++++++++++
 arch/x86/boot/compressed/head_64.S |  4 +++-
 arch/x86/boot/header.S             |  2 +-
 arch/x86/boot/tools/build.c        |  2 ++
 4 files changed, 18 insertions(+), 2 deletions(-)
arch/x86/Kconfig +12 -0
index f9920f1341c8..0c8fcb090232 100644
@@ -1964,6 +1964,18 @@ config EFI_STUB
 
 	  See Documentation/admin-guide/efi-stub.rst for more information.
 
+config EFI_HANDOVER_PROTOCOL
+	bool "EFI handover protocol (DEPRECATED)"
+	depends on EFI_STUB
+	default y
+	help
+	  Whether to include support for the deprecated EFI handover protocol,
+	  which defines alternative entry points into the EFI stub. This is a
+	  practice that has no basis in the UEFI specification, and requires
+	  a priori knowledge on the part of the bootloader about Linux/x86
+	  specific ways of passing the command line and initrd, and where in
+	  memory those assets may be loaded.
+
 config EFI_MIXED
 	bool "EFI mixed-mode support"
 	depends on EFI_STUB && X86_64
arch/x86/boot/compressed/head_64.S +3 -1
index 6ba2c2142c33..7bcc50c6cdcc 100644
@@ -286,7 +286,7 @@ SYM_FUNC_START(startup_32)
 	lret
 SYM_FUNC_END(startup_32)
 
-#ifdef CONFIG_EFI_MIXED
+#if defined(CONFIG_EFI_MIXED) && defined(CONFIG_EFI_HANDOVER_PROTOCOL)
 	.org 0x190
 SYM_FUNC_START(efi32_stub_entry)
 	add	$0x4, %esp		/* Discard return address */
@@ -516,7 +516,9 @@ trampoline_return:
 SYM_CODE_END(startup_64)
 
 #ifdef CONFIG_EFI_STUB
+#ifdef CONFIG_EFI_HANDOVER_PROTOCOL
 	.org 0x390
+#endif
 SYM_FUNC_START(efi64_stub_entry)
 	and	$~0xf, %rsp			/* realign the stack */
 	movq	%rdx, %rbx			/* save boot_params pointer */
arch/x86/boot/header.S +1 -1
index f912d7770130..d31982509654 100644
@@ -406,7 +406,7 @@ xloadflags:
 # define XLF1 0
 #endif
 
-#ifdef CONFIG_EFI_STUB
+#ifdef CONFIG_EFI_HANDOVER_PROTOCOL
 # ifdef CONFIG_EFI_MIXED
 #  define XLF23 (XLF_EFI_HANDOVER_32|XLF_EFI_HANDOVER_64)
 # else
arch/x86/boot/tools/build.c +2 -0
index a3725ad46c5a..bd247692b701 100644
@@ -290,6 +290,7 @@ static void efi_stub_entry_update(void)
 {
 	unsigned long addr = efi32_stub_entry;
 
+#ifdef CONFIG_EFI_HANDOVER_PROTOCOL
 #ifdef CONFIG_X86_64
 	/* Yes, this is really how we defined it :( */
 	addr = efi64_stub_entry - 0x200;
@@ -298,6 +299,7 @@ static void efi_stub_entry_update(void)
 #ifdef CONFIG_EFI_MIXED
 	if (efi32_stub_entry != addr)
 		die("32-bit and 64-bit EFI entry points do not match\n");
+#endif
 #endif
 	put_unaligned_le32(addr, &buf[0x264]);
 }
Signature · 1 line
2.35.1
maintainer
#1 lore raw
Quoted 5 lines · Ard Biesheuvel

Given that loaders such as GRUB already carried the bootparams handling in order to implement non-EFI boot, retaining that code and just passing bootparams to the EFI stub was a reasonable choice (although defining an alternate entrypoint could have been avoided.) However, the GRUB side changes never made it upstream, and are only shipped by some of the

	  ^^^^^^^^^^^^^^^^^^^^^^^
Quoted 1 line

distros in their downstream versions.

Gotta love that bit particularly. :-(

Quoted 27 lines

In the meantime, EFI support has been added to other Linux architecture ports, as well as to U-boot and systemd, including arch-agnostic methods for passing initrd images in memory [1], and for doing mixed mode boot [2], none of them requiring anything like the EFI handover protocol. So given that only out-of-tree distro GRUB relies on this, let's permit it to be omitted from the build, in preparation for retiring it completely at a later date. (Note that systemd-boot does have an implementation as well, but only uses it as a fallback for booting images that do not implement the LoadFile2 based initrd loading method, i.e., v5.8 or older)

[0] https://lore.kernel.org/all/[email protected]/
[1] ec93fc371f01 ("efi/libstub: Add support for loading the initrd ...")
[2] 97aa276579b2 ("efi/x86: Add true mixed mode entry point into ...")

---
 arch/x86/Kconfig                   | 12 ++++++++++++
 arch/x86/boot/compressed/head_64.S |  4 +++-
 arch/x86/boot/header.S             |  2 +-
 arch/x86/boot/tools/build.c        |  2 ++
 4 files changed, 18 insertions(+), 2 deletions(-)
arch/x86/Kconfig +0 -0
index f9920f1341c8..0c8fcb090232 100644

arch/x86/Kconfig config EFI_STUB · line 1970

Quoted 7 lines
arch/x86/Kconfig +4 -0
@@ -1964,6 +1964,18 @@ config EFI_STUB
 
 	  See Documentation/admin-guide/efi-stub.rst for more information.
 
+config EFI_HANDOVER_PROTOCOL
+	bool "EFI handover protocol (DEPRECATED)"
+	depends on EFI_STUB
+	default y

I'd say "default n" here.

arch/x86/Kconfig config EFI_STUB · line 1972

Quoted 2 lines
arch/x86/Kconfig +2 -0
+	help
+	  Whether to include support for the deprecated EFI handover protocol,

"Select this in order to include..."

arch/x86/Kconfig config EFI_STUB · line 1977

Quoted 5 lines
arch/x86/Kconfig +5 -0
+	  which defines alternative entry points into the EFI stub. This is a
+	  practice that has no basis in the UEFI specification, and requires
+	  a priori knowledge on the part of the bootloader about Linux/x86
+	  specific ways of passing the command line and initrd, and where in
+	  memory those assets may be loaded.

"If in doubt, say N. This option and accompanying code will disappear in some future kernel as the corresponding GRUB support is not even present in upstream GRUB but only in some distros' versions."

Quoted 9 lines
arch/x86/Kconfig +1 -0
+
 config EFI_MIXED
 	bool "EFI mixed-mode support"
 	depends on EFI_STUB && X86_64
arch/x86/boot/compressed/head_64.S +0 -0
index 6ba2c2142c33..7bcc50c6cdcc 100644

arch/x86/boot/compressed/head_64.S SYM_FUNC_START(startup_32) · line 289

Quoted 5 lines
arch/x86/boot/compressed/head_64.S +1 -1
@@ -286,7 +286,7 @@ SYM_FUNC_START(startup_32)
 	lret
 SYM_FUNC_END(startup_32)
 
-#ifdef CONFIG_EFI_MIXED
+#if defined(CONFIG_EFI_MIXED) && defined(CONFIG_EFI_HANDOVER_PROTOCOL)

...

Can we use IS_ENABLED() in all that instead, in order to improve readability?

In any case, looks good. I'm thinking I'll take it into tip after -rc1 and see who cries and why...

Thx.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
author maintainer
#2 lore raw
Quoted 7 lines · Borislav Petkov
Quoted 5 lines · Ard Biesheuvel

Given that loaders such as GRUB already carried the bootparams handling in order to implement non-EFI boot, retaining that code and just passing bootparams to the EFI stub was a reasonable choice (although defining an alternate entrypoint could have been avoided.) However, the GRUB side changes never made it upstream, and are only shipped by some of the

Quoted 6 lines
          ^^^^^^^^^^^^^^^^^^^^^^^
Quoted 1 line

distros in their downstream versions.

Gotta love that bit particularly. :-(

Yeah most distros have ~100 ore more patches against GRUB, but this isn't actually their fault. GRUB maintainership was defunct for a number of years, which is why we were stuck on GRUB version 2.02-beta3 for such a long time. But in recent years, things have been getting better, and there is an agreement with the current maintainer not to merge the EFI handover protocol, and merge the new EFI protocol based initrd loading method instead, which works on all architectures instead of only on x86.

Quoted 28 lines
Quoted 28 lines

In the meantime, EFI support has been added to other Linux architecture ports, as well as to U-boot and systemd, including arch-agnostic methods for passing initrd images in memory [1], and for doing mixed mode boot [2], none of them requiring anything like the EFI handover protocol. So given that only out-of-tree distro GRUB relies on this, let's permit it to be omitted from the build, in preparation for retiring it completely at a later date. (Note that systemd-boot does have an implementation as well, but only uses it as a fallback for booting images that do not implement the LoadFile2 based initrd loading method, i.e., v5.8 or older)

[0] https://lore.kernel.org/all/[email protected]/
[1] ec93fc371f01 ("efi/libstub: Add support for loading the initrd ...")
[2] 97aa276579b2 ("efi/x86: Add true mixed mode entry point into ...")

---
 arch/x86/Kconfig                   | 12 ++++++++++++
 arch/x86/boot/compressed/head_64.S |  4 +++-
 arch/x86/boot/header.S             |  2 +-
 arch/x86/boot/tools/build.c        |  2 ++
 4 files changed, 18 insertions(+), 2 deletions(-)
arch/x86/Kconfig +0 -0
index f9920f1341c8..0c8fcb090232 100644
@@ -1964,6 +1964,18 @@ config EFI_STUB
 
Quoted 9 lines

arch/x86/Kconfig config EFI_STUB · line 1970

Quoted 6 lines
arch/x86/Kconfig +4 -0
        See Documentation/admin-guide/efi-stub.rst for more information.
 
+config EFI_HANDOVER_PROTOCOL
+     bool "EFI handover protocol (DEPRECATED)"
+     depends on EFI_STUB
+     default y

I'd say "default n" here.

I'd rather not - see below.

Quoted 5 lines

arch/x86/Kconfig config EFI_STUB · line 1972

Quoted 2 lines
arch/x86/Kconfig +2 -0
+     help
+       Whether to include support for the deprecated EFI handover protocol,

"Select this in order to include..."

OK

Quoted 10 lines

arch/x86/Kconfig config EFI_STUB · line 1977

Quoted 5 lines
arch/x86/Kconfig +5 -0
+       which defines alternative entry points into the EFI stub. This is a
+       practice that has no basis in the UEFI specification, and requires
+       a priori knowledge on the part of the bootloader about Linux/x86
+       specific ways of passing the command line and initrd, and where in
+       memory those assets may be loaded.

"If in doubt, say N. This option and accompanying code will disappear in some future kernel as the corresponding GRUB support is not even present in upstream GRUB but only in some distros' versions."

I'll adopt some of that thanks.

Quoted 12 lines
Quoted 12 lines
arch/x86/Kconfig +1 -0
+
 config EFI_MIXED
      bool "EFI mixed-mode support"
      depends on EFI_STUB && X86_64
arch/x86/boot/compressed/head_64.S +0 -0
index 6ba2c2142c33..7bcc50c6cdcc 100644
@@ -286,7 +286,7 @@ SYM_FUNC_START(startup_32)
      lret
 SYM_FUNC_END(startup_32)
 
Quoted 8 lines

arch/x86/boot/compressed/head_64.S SYM_FUNC_START(startup_32) · line 289

Quoted 2 lines
arch/x86/boot/compressed/head_64.S +1 -1
-#ifdef CONFIG_EFI_MIXED
+#if defined(CONFIG_EFI_MIXED) && defined(CONFIG_EFI_HANDOVER_PROTOCOL)

...

Can we use IS_ENABLED() in all that instead, in order to improve readability?

Never tried that in .S files but I guess it should just work.

Quoted 3 lines

In any case, looks good. I'm thinking I'll take it into tip after -rc1 and see who cries and why...

I'd venture a guess that this will break the boot even your own x86 boxes, given that almost nobody uses plain upstream GRUB..

I can work with the distros directly to start disabling this in their downstream configs once their GRUB builds are up to date with the new changes, so we can phase this out in a controlled manner. But disabling tthis right now by default is going to affect everyone who builds their own kernels and runs them on a distro Linux install.

maintainer
#3 lore raw
Quoted 8 lines · Ard Biesheuvel

Yeah most distros have ~100 ore more patches against GRUB, but this isn't actually their fault. GRUB maintainership was defunct for a number of years, which is why we were stuck on GRUB version 2.02-beta3 for such a long time. But in recent years, things have been getting better, and there is an agreement with the current maintainer not to merge the EFI handover protocol, and merge the new EFI protocol based initrd loading method instead, which works on all architectures instead of only on x86.

Aha, ok.

Quoted 1 line

Never tried that in .S files but I guess it should just work.

If not, at least in the .c files.

Quoted 6 lines

I'd venture a guess that this will break the boot even your own x86 boxes, given that almost nobody uses plain upstream GRUB..

I can work with the distros directly to start disabling this in their downstream configs once their GRUB builds are up to date with the new changes, so we can phase this out in a controlled manner.

Hm, that might turn out to be a multi-year effort considering how the enterprise distros' kernels are moving. Yeah, yeah, they have good reasons and so on.

Quoted 2 lines

But disabling tthis right now by default is going to affect everyone who builds their own kernels and runs them on a distro Linux install.

Ok, I can try it on my SUSE and Debian partitions and see what happens.

Thx.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
author maintainer
#4 lore raw
Quoted 20 lines · Borislav Petkov
Quoted 8 lines · Ard Biesheuvel

Yeah most distros have ~100 ore more patches against GRUB, but this isn't actually their fault. GRUB maintainership was defunct for a number of years, which is why we were stuck on GRUB version 2.02-beta3 for such a long time. But in recent years, things have been getting better, and there is an agreement with the current maintainer not to merge the EFI handover protocol, and merge the new EFI protocol based initrd loading method instead, which works on all architectures instead of only on x86.

Aha, ok.

Quoted 1 line

Never tried that in .S files but I guess it should just work.

If not, at least in the .c files.

Quoted 3 lines

I'd venture a guess that this will break the boot even your own x86 boxes, given that almost nobody uses plain upstream GRUB..

Quoted 8 lines
Quoted 3 lines

I can work with the distros directly to start disabling this in their downstream configs once their GRUB builds are up to date with the new changes, so we can phase this out in a controlled manner.

Hm, that might turn out to be a multi-year effort considering how the enterprise distros' kernels are moving. Yeah, yeah, they have good reasons and so on.

Yes, this is going to take time. But we simply cannot get rid of it today, so the choice we have is between doing nothing at all, or taking the next step in phasing out this stuff.

Signature · 1 line
Ard.
maintainer
#5 lore raw
Quoted 3 lines · Ard Biesheuvel

Yes, this is going to take time. But we simply cannot get rid of it today, so the choice we have is between doing nothing at all, or taking the next step in phasing out this stuff.

Yes, next step ofc. This is simply the next thing we're deprecating.

Thx.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette