Skip to contentSkip to conversation

arch/x86: efistub: Invoke EFI_RNG_PROTOCOL to seed the UEFI RNG table

RFC PATCH In review Dominik Brodowski 3 participants 11 messages Oct 5, 2019 – Oct 31, 2019

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

RFC PATCH 1/1 arch/x86: efistub: Invoke EFI_RNG_PROTOCOL to seed the UEFI RNG table

Implement the same mechanism for x86 efistub as introduced by commit 568bc4e87033 ("efi/arm*/libstub: Invoke EFI_RNG_PROTOCOL to seed the UEFI RNG table") for efi/arm*/libstub, and best described here and there as:

Invoke the EFI_RNG_PROTOCOL protocol in the context of the stub and install the Linux-specific RNG seed UEFI config table. This will be picked up by the EFI routines in the core kernel to seed the kernel entropy pool.

Cc Ard Biesheuvel <[email protected]>
Cc Hsin-Yi Wang <[email protected]>
Cc Stephen Boyd <[email protected]>
Cc Rob Herring <[email protected]>
Cc Theodore Ts'o <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc "H. Peter Anvin" <[email protected]>

---

As far as I can see, we do not yet make use of the UEFI RNG on x86 at all, but only on arm. Note that this works only when Linux is booted as an EFI stub, and that the EFI-provided randomness is not credited as entropy unless RANDOM_TRUST_BOOTLOADER is set _and_ another patch (on its way upstream; thanks Ard!) is applied, see

	https://lore.kernel.org/lkml/[email protected]/

Further note that this patch is untested, as the firmware on my old x86 laptop only has UEFI v2.31. If you want to test it, you may wish to apply

	https://lore.kernel.org/lkml/[email protected]/T/#u

first to get a clear indication in dmesg.

Thanks,
Dominik

arch/x86/boot/compressed/eboot.c +3 -0
index d6662fdef300..4b909e5ab857 100644
@@ -781,6 +781,9 @@ efi_main(struct efi_config *c, struct boot_params *boot_params)
 
 	/* Ask the firmware to clear memory on unclean shutdown */
 	efi_enable_reset_attack_mitigation(sys_table);
+
+	efi_random_get_seed(sys_table);
+
 	efi_retrieve_tpm2_eventlog(sys_table);
 
 	setup_graphics(boot_params);
drivers/firmware/efi/libstub/Makefile +3 -2
index 0460c7581220..ece24c60fc2c 100644
@@ -38,7 +38,8 @@ OBJECT_FILES_NON_STANDARD	:= y
 # Prevents link failures: __sanitizer_cov_trace_pc() is not linked in.
 KCOV_INSTRUMENT			:= n
 
-lib-y				:= efi-stub-helper.o gop.o secureboot.o tpm.o
+lib-y				:= efi-stub-helper.o gop.o secureboot.o tpm.o \
+				   random.o
 
 # include the stub's generic dependencies from lib/ when building for ARM/arm64
 arm-deps-y := fdt_rw.c fdt_ro.c fdt_wip.c fdt.c fdt_empty_tree.c fdt_sw.c
@@ -47,7 +48,7 @@ arm-deps-$(CONFIG_ARM64) += sort.c
 $(obj)/lib-%.o: $(srctree)/lib/%.c FORCE
 	$(call if_changed_rule,cc_o_c)
 
-lib-$(CONFIG_EFI_ARMSTUB)	+= arm-stub.o fdt.o string.o random.o \
+lib-$(CONFIG_EFI_ARMSTUB)	+= arm-stub.o fdt.o string.o \
 				   $(patsubst %.c,lib-%.o,$(arm-deps-y))
 
 lib-$(CONFIG_ARM)		+= arm32-stub.o
drivers/firmware/efi/libstub/efistub.h +0 -2
index 7f1556fd867d..05739ae013c8 100644
@@ -63,8 +63,6 @@ efi_status_t efi_random_alloc(efi_system_table_t *sys_table_arg,
 
 efi_status_t check_platform_features(efi_system_table_t *sys_table_arg);
 
-efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
-
 void *get_efi_config_table(efi_system_table_t *sys_table, efi_guid_t guid);
 
 /* Helper macros for the usual case of using simple C variables: */
include/linux/efi.h +2 -0
index bd3837022307..a17cc5841668 100644
@@ -1631,6 +1631,8 @@ static inline void
 efi_enable_reset_attack_mitigation(efi_system_table_t *sys_table_arg) { }
 #endif
 
+efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
+
 void efi_retrieve_tpm2_eventlog(efi_system_table_t *sys_table);
 
 /*
#1 lore raw
Quoted 23 lines · Dominik Brodowski

Implement the same mechanism for x86 efistub as introduced by commit 568bc4e87033 ("efi/arm*/libstub: Invoke EFI_RNG_PROTOCOL to seed the UEFI RNG table") for efi/arm*/libstub, and best described here and there as:

Invoke the EFI_RNG_PROTOCOL protocol in the context of the stub and install the Linux-specific RNG seed UEFI config table. This will be picked up by the EFI routines in the core kernel to seed the kernel entropy pool.

Cc Ard Biesheuvel <[email protected]>
Cc Hsin-Yi Wang <[email protected]>
Cc Stephen Boyd <[email protected]>
Cc Rob Herring <[email protected]>
Cc Theodore Ts'o <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc "H. Peter Anvin" <[email protected]>
Quoted 4 lines

---

As far as I can see, we do not yet make use of the UEFI RNG on x86 at all, but only on arm.

If you have a ChaosKey, you may be able to use this driver

http://people.linaro.org/~ard.biesheuvel/ChaosKey/ChaosKeyDxe.efi

to get UEFI to expose the EFI_RNG_PROTOCOL. (Use efibootmgr -r to install it as Driver#### or load it from the UEFI shell using 'load xxx.efi')

Quoted 77 lines

Note that this works only when Linux is booted as an EFI stub, and that the EFI-provided randomness is not credited as entropy unless RANDOM_TRUST_BOOTLOADER is set _and_ another patch (on its way upstream; thanks Ard!) is applied, see

        https://lore.kernel.org/lkml/[email protected]/

Further note that this patch is untested, as the firmware on my old x86 laptop only has UEFI v2.31. If you want to test it, you may wish to apply

        https://lore.kernel.org/lkml/[email protected]/T/#u

first to get a clear indication in dmesg.

arch/x86/boot/compressed/eboot.c +3 -0
index d6662fdef300..4b909e5ab857 100644
@@ -781,6 +781,9 @@ efi_main(struct efi_config *c, struct boot_params *boot_params)
 
        /* Ask the firmware to clear memory on unclean shutdown */
        efi_enable_reset_attack_mitigation(sys_table);
+
+       efi_random_get_seed(sys_table);
+
        efi_retrieve_tpm2_eventlog(sys_table);
 
        setup_graphics(boot_params);
drivers/firmware/efi/libstub/Makefile +3 -2
index 0460c7581220..ece24c60fc2c 100644
@@ -38,7 +38,8 @@ OBJECT_FILES_NON_STANDARD     := y
 # Prevents link failures: __sanitizer_cov_trace_pc() is not linked in.
 KCOV_INSTRUMENT                        := n
 
-lib-y                          := efi-stub-helper.o gop.o secureboot.o tpm.o
+lib-y                          := efi-stub-helper.o gop.o secureboot.o tpm.o \
+                                  random.o
 
 # include the stub's generic dependencies from lib/ when building for ARM/arm64
 arm-deps-y := fdt_rw.c fdt_ro.c fdt_wip.c fdt.c fdt_empty_tree.c fdt_sw.c
@@ -47,7 +48,7 @@ arm-deps-$(CONFIG_ARM64) += sort.c
 $(obj)/lib-%.o: $(srctree)/lib/%.c FORCE
        $(call if_changed_rule,cc_o_c)
 
-lib-$(CONFIG_EFI_ARMSTUB)      += arm-stub.o fdt.o string.o random.o \
+lib-$(CONFIG_EFI_ARMSTUB)      += arm-stub.o fdt.o string.o \
                                   $(patsubst %.c,lib-%.o,$(arm-deps-y))
 
 lib-$(CONFIG_ARM)              += arm32-stub.o
drivers/firmware/efi/libstub/efistub.h +0 -2
index 7f1556fd867d..05739ae013c8 100644
@@ -63,8 +63,6 @@ efi_status_t efi_random_alloc(efi_system_table_t *sys_table_arg,
 
 efi_status_t check_platform_features(efi_system_table_t *sys_table_arg);
 
-efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
-
 void *get_efi_config_table(efi_system_table_t *sys_table, efi_guid_t guid);
 
 /* Helper macros for the usual case of using simple C variables: */
include/linux/efi.h +2 -0
index bd3837022307..a17cc5841668 100644
@@ -1631,6 +1631,8 @@ static inline void
 efi_enable_reset_attack_mitigation(efi_system_table_t *sys_table_arg) { }
 #endif
 
+efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
+
 void efi_retrieve_tpm2_eventlog(efi_system_table_t *sys_table);
 
 /*
#2 lore raw

efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

In reply to Dominik Brodowski

Ard, Mario,

in theory, invoking EFI_RNG_PROTOCOL on a Dell Inc. Latitude 7390/09386V, BIOS 1.9.2 04/03/2019, should work fine as the system provides EFI v2.60. Using my patch from a few weeks ago[1], efi_random_get_seed() is called from arch/x86/boot/compressed/eboot.c::efi_main(). In efi_random_get_seed(), the calls to

	status = efi_call_early(locate_protocol, &rng_proto, NULL,
				(void **)&rng);

and

status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
sizeof(*seed) + EFI_RANDOM_SEED_SIZE, (void **)&seed);

succeed. However,

	status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
			      seed->bits);

returns EFI_INVALID_PARAMETER, though I can't see why one of these parameters is invalid. When trying to use the default rng algorithm (by modifying the test to "(status != EFI_SUCCESS)"),

		status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
				      seed->bits);

the call doesn't seem to return.

Any ideas?

Best,
Dominik

[1] https://lore.kernel.org/lkml/[email protected]/

#3 lore raw

Re: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

Hello Dominik,

Quoted 18 lines · Dominik Brodowski

Ard, Mario,

in theory, invoking EFI_RNG_PROTOCOL on a Dell Inc. Latitude 7390/09386V, BIOS 1.9.2 04/03/2019, should work fine as the system provides EFI v2.60. Using my patch from a few weeks ago[1], efi_random_get_seed() is called from arch/x86/boot/compressed/eboot.c::efi_main(). In efi_random_get_seed(), the calls to

status = efi_call_early(locate_protocol, &rng_proto, NULL,
(void **)&rng);

and

status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
sizeof(*seed) + EFI_RANDOM_SEED_SIZE, (void **)&seed);

Quoted 7 lines

succeed. However,

        status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
                              seed->bits);

returns EFI_INVALID_PARAMETER, though I can't see why one of these parameters is invalid.

The UEFI spec defines the conditions under which this function may return EFI_INVALID_PARAMETER as

"""
RNGValue is null or RNGValueLength is zero.

"""

Quoted 10 lines

When trying to use the default rng algorithm (by
modifying the test to "(status != EFI_SUCCESS)"),

                status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
                                      seed->bits);

the call doesn't seem to return.

Any ideas?

Try running this from the UEFI shell:

http://people.linaro.org/~ard.biesheuvel/RngTest-X64.efi

#4 lore raw

Re: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

Hello Ard,

Quoted 45 lines · Ard Biesheuvel

Hello Dominik,

Quoted 18 lines · Dominik Brodowski

Ard, Mario,

in theory, invoking EFI_RNG_PROTOCOL on a Dell Inc. Latitude 7390/09386V, BIOS 1.9.2 04/03/2019, should work fine as the system provides EFI v2.60. Using my patch from a few weeks ago[1], efi_random_get_seed() is called from arch/x86/boot/compressed/eboot.c::efi_main(). In efi_random_get_seed(), the calls to

status = efi_call_early(locate_protocol, &rng_proto, NULL,
(void **)&rng);

and

status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
sizeof(*seed) + EFI_RANDOM_SEED_SIZE, (void **)&seed);

Quoted 7 lines

succeed. However,

        status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
                              seed->bits);

returns EFI_INVALID_PARAMETER, though I can't see why one of these parameters is invalid.

The UEFI spec defines the conditions under which this function may return EFI_INVALID_PARAMETER as

"""
RNGValue is null or RNGValueLength is zero.

"""

Quoted 6 lines

When trying to use the default rng algorithm (by
modifying the test to "(status != EFI_SUCCESS)"),

                status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
                                      seed->bits);
Quoted 8 lines
Quoted 4 lines

the call doesn't seem to return.

Any ideas?

Try running this from the UEFI shell:

http://people.linaro.org/~ard.biesheuvel/RngTest-X64.efi

Interestingly, this succeeds -- with the default algorithm, SP800-90-CTR-256 and RAW. So I am more confused than before on why the call to ->get_rng() fails in efi_random_get_seed().

Best,
Dominik

#5 lore raw

Re: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

Quoted 54 lines · Dominik Brodowski

Hello Ard,

Quoted 50 lines · Ard Biesheuvel

Hello Dominik,

Quoted 18 lines · Dominik Brodowski

Ard, Mario,

in theory, invoking EFI_RNG_PROTOCOL on a Dell Inc. Latitude 7390/09386V, BIOS 1.9.2 04/03/2019, should work fine as the system provides EFI v2.60. Using my patch from a few weeks ago[1], efi_random_get_seed() is called from arch/x86/boot/compressed/eboot.c::efi_main(). In efi_random_get_seed(), the calls to

status = efi_call_early(locate_protocol, &rng_proto, NULL,
(void **)&rng);

and

status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
sizeof(*seed) + EFI_RANDOM_SEED_SIZE, (void **)&seed);

Quoted 7 lines

succeed. However,

        status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
                              seed->bits);

returns EFI_INVALID_PARAMETER, though I can't see why one of these parameters is invalid.

The UEFI spec defines the conditions under which this function may return EFI_INVALID_PARAMETER as

"""
RNGValue is null or RNGValueLength is zero.

"""

Quoted 10 lines

When trying to use the default rng algorithm (by
modifying the test to "(status != EFI_SUCCESS)"),

                status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
                                      seed->bits);

the call doesn't seem to return.

Any ideas?

Quoted 8 lines
Quoted 3 lines

Try running this from the UEFI shell:

http://people.linaro.org/~ard.biesheuvel/RngTest-X64.efi

Interestingly, this succeeds -- with the default algorithm, SP800-90-CTR-256 and RAW. So I am more confused than before on why the call to ->get_rng() fails in efi_random_get_seed().

It might be an issue with the size of the output.

The original RDRAND based driver in EDK2 contains an apparent misconception that, due to the fact that certain SP800-90-CTR-256 DRBGs require 32 bytes of raw entropy as a seed, it is only valid for the RAW algorithm to be called with an output size of 32.

So in this case, it might be that 32 is treated as a magic number too, and any other size is rejected by the RAW algorithm.

Not sure why the other one fails, though, but the fact that RAW and SP800-90-CTR-256 are the only supported algorithms suggests that your implementation is at least similar to the RDRAND based RngDxe implementation in EDK2.

#6 lore raw

Re: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

Quoted 73 lines · Ard Biesheuvel
Quoted 54 lines · Dominik Brodowski

Hello Ard,

Quoted 50 lines · Ard Biesheuvel

Hello Dominik,

Quoted 18 lines · Dominik Brodowski

Ard, Mario,

in theory, invoking EFI_RNG_PROTOCOL on a Dell Inc. Latitude 7390/09386V, BIOS 1.9.2 04/03/2019, should work fine as the system provides EFI v2.60. Using my patch from a few weeks ago[1], efi_random_get_seed() is called from arch/x86/boot/compressed/eboot.c::efi_main(). In efi_random_get_seed(), the calls to

status = efi_call_early(locate_protocol, &rng_proto, NULL,
(void **)&rng);

and

status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
sizeof(*seed) + EFI_RANDOM_SEED_SIZE, (void **)&seed);

Quoted 7 lines

succeed. However,

        status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
                              seed->bits);

returns EFI_INVALID_PARAMETER, though I can't see why one of these parameters is invalid.

The UEFI spec defines the conditions under which this function may return EFI_INVALID_PARAMETER as

"""
RNGValue is null or RNGValueLength is zero.

"""

Quoted 10 lines

When trying to use the default rng algorithm (by
modifying the test to "(status != EFI_SUCCESS)"),

                status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
                                      seed->bits);

the call doesn't seem to return.

Any ideas?

Quoted 8 lines
Quoted 3 lines

Try running this from the UEFI shell:

http://people.linaro.org/~ard.biesheuvel/RngTest-X64.efi

Interestingly, this succeeds -- with the default algorithm, SP800-90-CTR-256 and RAW. So I am more confused than before on why the call to ->get_rng() fails in efi_random_get_seed().

It might be an issue with the size of the output.

The original RDRAND based driver in EDK2 contains an apparent misconception that, due to the fact that certain SP800-90-CTR-256 DRBGs require 32 bytes of raw entropy as a seed, it is only valid for the RAW algorithm to be called with an output size of 32.

Quoted 7 lines

So in this case, it might be that 32 is treated as a magic number too, and any other size is rejected by the RAW algorithm.

Not sure why the other one fails, though, but the fact that RAW and SP800-90-CTR-256 are the only supported algorithms suggests that your implementation is at least similar to the RDRAND based RngDxe implementation in EDK2.

I've updated the RngTest-X64.efi binary with a version that invokes both the RAW and the default algorithm twice with a request for 64 bytes of entropy, like we do in the kernel. It runs fine on my Thinkpad, can you check whether it works on your Dell system as well?

#7 lore raw

Re: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

Quoted 15 lines · Ard Biesheuvel
Quoted 7 lines

It might be an issue with the size of the output.

The original RDRAND based driver in EDK2 contains an apparent misconception that, due to the fact that certain SP800-90-CTR-256 DRBGs require 32 bytes of raw entropy as a seed, it is only valid for the RAW algorithm to be called with an output size of 32.

Quoted 7 lines

So in this case, it might be that 32 is treated as a magic number too, and any other size is rejected by the RAW algorithm.

Not sure why the other one fails, though, but the fact that RAW and SP800-90-CTR-256 are the only supported algorithms suggests that your implementation is at least similar to the RDRAND based RngDxe implementation in EDK2.

Quoted 4 lines

I've updated the RngTest-X64.efi binary with a version that invokes both the RAW and the default algorithm twice with a request for 64 bytes of entropy, like we do in the kernel. It runs fine on my Thinkpad, can you check whether it works on your Dell system as well?

RngTest-X64.efi seems to run through fine.[*] Additionally, if I modify the kernel code to request only 32 bytes, it returns EFI_INVALID_PARAMETER as well. So something else seems to be missing... Just to verify: does efi_random_get_seed() work on your Thinkpad, if called from efi_main()? Have you tested that, by chance?[**]

Thanks,
Dominik

[*] As a sidenote: I say "seems" as the screen is immediately blanked once the RngTest-X64.efi binary has run through, so I have to capture the output with a camera. Maybe add a delay of some sort to the RngTest-X86.efi binary?

[**] patch with additional debug outputs (on top of Linus' tree) here:

arch/x86/boot/compressed/eboot.c +3 -0
index d6662fdef300..4b909e5ab857 100644
@@ -781,6 +781,9 @@ efi_main(struct efi_config *c, struct boot_params *boot_params)
 
 	/* Ask the firmware to clear memory on unclean shutdown */
 	efi_enable_reset_attack_mitigation(sys_table);
+
+	efi_random_get_seed(sys_table);
+
 	efi_retrieve_tpm2_eventlog(sys_table);
 
 	setup_graphics(boot_params);
drivers/char/random.c +8 -0
index de434feb873a..45572b7e0548 100644
@@ -2515,6 +2515,14 @@ EXPORT_SYMBOL_GPL(add_hwgenerator_randomness);
  */
 void add_bootloader_randomness(const void *buf, unsigned int size)
 {
+	unsigned long *data = buf;
+
+	pr_notice("random: adding %u bits of %sbootloader-provided randomness",
+		size * 8,
+		IS_ENABLED(CONFIG_RANDOM_TRUST_BOOTLOADER) ? "trusted " : "");
+
+	pr_notice("random: is this random? %lx", *data);
+
 	if (IS_ENABLED(CONFIG_RANDOM_TRUST_BOOTLOADER))
 		add_hwgenerator_randomness(buf, size, size * 8);
 	else
drivers/firmware/efi/efi.c +1 -1
index 69f00f7453a3..e98bbf8e56d9 100644
@@ -554,7 +554,7 @@ int __init efi_config_parse_tables(void *config_tables, int count, int sz,
 					      sizeof(*seed) + size);
 			if (seed != NULL) {
 				pr_notice("seeding entropy pool\n");
-				add_device_randomness(seed->bits, seed->size);
+				add_bootloader_randomness(seed->bits, seed->size);
 				early_memunmap(seed, sizeof(*seed) + size);
 			} else {
 				pr_err("Could not map UEFI random seed!\n");
drivers/firmware/efi/libstub/Makefile +3 -2
index 0460c7581220..ece24c60fc2c 100644
@@ -38,7 +38,8 @@ OBJECT_FILES_NON_STANDARD	:= y
 # Prevents link failures: __sanitizer_cov_trace_pc() is not linked in.
 KCOV_INSTRUMENT			:= n
 
-lib-y				:= efi-stub-helper.o gop.o secureboot.o tpm.o
+lib-y				:= efi-stub-helper.o gop.o secureboot.o tpm.o \
+				   random.o
 
 # include the stub's generic dependencies from lib/ when building for ARM/arm64
 arm-deps-y := fdt_rw.c fdt_ro.c fdt_wip.c fdt.c fdt_empty_tree.c fdt_sw.c
@@ -47,7 +48,7 @@ arm-deps-$(CONFIG_ARM64) += sort.c
 $(obj)/lib-%.o: $(srctree)/lib/%.c FORCE
 	$(call if_changed_rule,cc_o_c)
 
-lib-$(CONFIG_EFI_ARMSTUB)	+= arm-stub.o fdt.o string.o random.o \
+lib-$(CONFIG_EFI_ARMSTUB)	+= arm-stub.o fdt.o string.o \
 				   $(patsubst %.c,lib-%.o,$(arm-deps-y))
 
 lib-$(CONFIG_ARM)		+= arm32-stub.o
drivers/firmware/efi/libstub/efistub.h +0 -2
index 7f1556fd867d..05739ae013c8 100644
@@ -63,8 +63,6 @@ efi_status_t efi_random_alloc(efi_system_table_t *sys_table_arg,
 
 efi_status_t check_platform_features(efi_system_table_t *sys_table_arg);
 
-efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
-
 void *get_efi_config_table(efi_system_table_t *sys_table, efi_guid_t guid);
 
 /* Helper macros for the usual case of using simple C variables: */
drivers/firmware/efi/libstub/random.c +21 -5
index b4b1d1dcb5fd..70612cdd62c8 100644
@@ -152,33 +152,49 @@ efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg)
 
 	status = efi_call_early(locate_protocol, &rng_proto, NULL,
 				(void **)&rng);
-	if (status != EFI_SUCCESS)
+	if (status != EFI_SUCCESS) {
+		pr_efi_err(sys_table_arg, "random: no protocol");
 		return status;
+	}
 
 	status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
 				sizeof(*seed) + EFI_RANDOM_SEED_SIZE,
 				(void **)&seed);
-	if (status != EFI_SUCCESS)
+	if (status != EFI_SUCCESS){
+		pr_efi_err(sys_table_arg, "random: no pool");
 		return status;
+	}
 
 	status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
 			      seed->bits);
-	if (status == EFI_UNSUPPORTED)
+	if (status == EFI_DEVICE_ERROR)
+		pr_efi_err(sys_table_arg, "random: device error");
+	if (status == EFI_NOT_READY)
+		pr_efi_err(sys_table_arg, "random: not ready");
+	if (status == EFI_INVALID_PARAMETER)
+		pr_efi_err(sys_table_arg, "random: invalid parameter"); // that's the reason...
+	if (status == EFI_UNSUPPORTED) {
 		/*
 		 * Use whatever algorithm we have available if the raw algorithm
 		 * is not implemented.
 		 */
+
 		status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
 				      seed->bits);
+	}
 
-	if (status != EFI_SUCCESS)
+	if (status != EFI_SUCCESS) {
+		pr_efi_err(sys_table_arg, "random: no bytes");
 		goto err_freepool;
+	}
 
 	seed->size = EFI_RANDOM_SEED_SIZE;
 	status = efi_call_early(install_configuration_table, &rng_table_guid,
 				seed);
-	if (status != EFI_SUCCESS)
+	if (status != EFI_SUCCESS) {
+		pr_efi_err(sys_table_arg, "random: install_configuration_table");
 		goto err_freepool;
+	}
 
 	return EFI_SUCCESS;
 
include/linux/efi.h +2 -0
index bd3837022307..a17cc5841668 100644
@@ -1631,6 +1631,8 @@ static inline void
 efi_enable_reset_attack_mitigation(efi_system_table_t *sys_table_arg) { }
 #endif
 
+efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
+
 void efi_retrieve_tpm2_eventlog(efi_system_table_t *sys_table);
 
 /*
#8 lore raw

Re: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

Quoted 22 lines · Dominik Brodowski
Quoted 15 lines · Ard Biesheuvel
Quoted 7 lines

It might be an issue with the size of the output.

The original RDRAND based driver in EDK2 contains an apparent misconception that, due to the fact that certain SP800-90-CTR-256 DRBGs require 32 bytes of raw entropy as a seed, it is only valid for the RAW algorithm to be called with an output size of 32.

Quoted 7 lines

So in this case, it might be that 32 is treated as a magic number too, and any other size is rejected by the RAW algorithm.

Not sure why the other one fails, though, but the fact that RAW and SP800-90-CTR-256 are the only supported algorithms suggests that your implementation is at least similar to the RDRAND based RngDxe implementation in EDK2.

Quoted 4 lines

I've updated the RngTest-X64.efi binary with a version that invokes both the RAW and the default algorithm twice with a request for 64 bytes of entropy, like we do in the kernel. It runs fine on my Thinkpad, can you check whether it works on your Dell system as well?

Quoted 6 lines

RngTest-X64.efi seems to run through fine.[*] Additionally, if I modify the kernel code to request only 32 bytes, it returns EFI_INVALID_PARAMETER as well. So something else seems to be missing... Just to verify: does efi_random_get_seed() work on your Thinkpad, if called from efi_main()? Have you tested that, by chance?[**]

I have just tried a v5.4-rc5 kernel built with your patch applied, and it does not boot at all in EFI mode, without any output whatsoever.

Unfortunately, I don't have the bandwidth currently to dig into this, but I should be able to help out a bit more next week.

Quoted 6 lines

[*] As a sidenote: I say "seems" as the screen is immediately blanked once the RngTest-X64.efi binary has run through, so I have to capture the output with a camera. Maybe add a delay of some sort to the RngTest-X86.efi binary?

Are you running it from the UEFI shell?

Quoted 157 lines

[**] patch with additional debug outputs (on top of Linus' tree) here:

arch/x86/boot/compressed/eboot.c +3 -0
index d6662fdef300..4b909e5ab857 100644
@@ -781,6 +781,9 @@ efi_main(struct efi_config *c, struct boot_params *boot_params)
 
        /* Ask the firmware to clear memory on unclean shutdown */
        efi_enable_reset_attack_mitigation(sys_table);
+
+       efi_random_get_seed(sys_table);
+
        efi_retrieve_tpm2_eventlog(sys_table);
 
        setup_graphics(boot_params);
drivers/char/random.c +8 -0
index de434feb873a..45572b7e0548 100644
@@ -2515,6 +2515,14 @@ EXPORT_SYMBOL_GPL(add_hwgenerator_randomness);
  */
 void add_bootloader_randomness(const void *buf, unsigned int size)
 {
+       unsigned long *data = buf;
+
+       pr_notice("random: adding %u bits of %sbootloader-provided randomness",
+               size * 8,
+               IS_ENABLED(CONFIG_RANDOM_TRUST_BOOTLOADER) ? "trusted " : "");
+
+       pr_notice("random: is this random? %lx", *data);
+
        if (IS_ENABLED(CONFIG_RANDOM_TRUST_BOOTLOADER))
                add_hwgenerator_randomness(buf, size, size * 8);
        else
drivers/firmware/efi/efi.c +1 -1
index 69f00f7453a3..e98bbf8e56d9 100644
@@ -554,7 +554,7 @@ int __init efi_config_parse_tables(void *config_tables, int count, int sz,
                                              sizeof(*seed) + size);
                        if (seed != NULL) {
                                pr_notice("seeding entropy pool\n");
-                               add_device_randomness(seed->bits, seed->size);
+                               add_bootloader_randomness(seed->bits, seed->size);
                                early_memunmap(seed, sizeof(*seed) + size);
                        } else {
                                pr_err("Could not map UEFI random seed!\n");
drivers/firmware/efi/libstub/Makefile +3 -2
index 0460c7581220..ece24c60fc2c 100644
@@ -38,7 +38,8 @@ OBJECT_FILES_NON_STANDARD     := y
 # Prevents link failures: __sanitizer_cov_trace_pc() is not linked in.
 KCOV_INSTRUMENT                        := n
 
-lib-y                          := efi-stub-helper.o gop.o secureboot.o tpm.o
+lib-y                          := efi-stub-helper.o gop.o secureboot.o tpm.o \
+                                  random.o
 
 # include the stub's generic dependencies from lib/ when building for ARM/arm64
 arm-deps-y := fdt_rw.c fdt_ro.c fdt_wip.c fdt.c fdt_empty_tree.c fdt_sw.c
@@ -47,7 +48,7 @@ arm-deps-$(CONFIG_ARM64) += sort.c
 $(obj)/lib-%.o: $(srctree)/lib/%.c FORCE
        $(call if_changed_rule,cc_o_c)
 
-lib-$(CONFIG_EFI_ARMSTUB)      += arm-stub.o fdt.o string.o random.o \
+lib-$(CONFIG_EFI_ARMSTUB)      += arm-stub.o fdt.o string.o \
                                   $(patsubst %.c,lib-%.o,$(arm-deps-y))
 
 lib-$(CONFIG_ARM)              += arm32-stub.o
drivers/firmware/efi/libstub/efistub.h +0 -2
index 7f1556fd867d..05739ae013c8 100644
@@ -63,8 +63,6 @@ efi_status_t efi_random_alloc(efi_system_table_t *sys_table_arg,
 
 efi_status_t check_platform_features(efi_system_table_t *sys_table_arg);
 
-efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
-
 void *get_efi_config_table(efi_system_table_t *sys_table, efi_guid_t guid);
 
 /* Helper macros for the usual case of using simple C variables: */
drivers/firmware/efi/libstub/random.c +21 -5
index b4b1d1dcb5fd..70612cdd62c8 100644
@@ -152,33 +152,49 @@ efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg)
 
        status = efi_call_early(locate_protocol, &rng_proto, NULL,
                                (void **)&rng);
-       if (status != EFI_SUCCESS)
+       if (status != EFI_SUCCESS) {
+               pr_efi_err(sys_table_arg, "random: no protocol");
                return status;
+       }
 
        status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
                                sizeof(*seed) + EFI_RANDOM_SEED_SIZE,
                                (void **)&seed);
-       if (status != EFI_SUCCESS)
+       if (status != EFI_SUCCESS){
+               pr_efi_err(sys_table_arg, "random: no pool");
                return status;
+       }
 
        status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
                              seed->bits);
-       if (status == EFI_UNSUPPORTED)
+       if (status == EFI_DEVICE_ERROR)
+               pr_efi_err(sys_table_arg, "random: device error");
+       if (status == EFI_NOT_READY)
+               pr_efi_err(sys_table_arg, "random: not ready");
+       if (status == EFI_INVALID_PARAMETER)
+               pr_efi_err(sys_table_arg, "random: invalid parameter"); // that's the reason...
+       if (status == EFI_UNSUPPORTED) {
                /*
                 * Use whatever algorithm we have available if the raw algorithm
                 * is not implemented.
                 */
+
                status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
                                      seed->bits);
+       }
 
-       if (status != EFI_SUCCESS)
+       if (status != EFI_SUCCESS) {
+               pr_efi_err(sys_table_arg, "random: no bytes");
                goto err_freepool;
+       }
 
        seed->size = EFI_RANDOM_SEED_SIZE;
        status = efi_call_early(install_configuration_table, &rng_table_guid,
                                seed);
-       if (status != EFI_SUCCESS)
+       if (status != EFI_SUCCESS) {
+               pr_efi_err(sys_table_arg, "random: install_configuration_table");
                goto err_freepool;
+       }
 
        return EFI_SUCCESS;
 
include/linux/efi.h +2 -0
index bd3837022307..a17cc5841668 100644
@@ -1631,6 +1631,8 @@ static inline void
 efi_enable_reset_attack_mitigation(efi_system_table_t *sys_table_arg) { }
 #endif
 
+efi_status_t efi_random_get_seed(efi_system_table_t *sys_table_arg);
+
 void efi_retrieve_tpm2_eventlog(efi_system_table_t *sys_table);
 
 /*
#9 lore raw

RE: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

In reply to Ard Biesheuvel
Quoted 84 lines

-----Original Message-----
From: Ard Biesheuvel <[email protected]>
Sent: Monday, October 28, 2019 3:57 AM
To: Dominik Brodowski
Cc: Limonciello, Mario; linux-efi
Subject: Re: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

[EXTERNAL EMAIL]

Quoted 14 lines · Dominik Brodowski

Hello Ard,

Quoted 10 lines · Ard Biesheuvel

Hello Dominik,

Quoted 6 lines · Dominik Brodowski

Ard, Mario,

in theory, invoking EFI_RNG_PROTOCOL on a Dell Inc. Latitude 7390/09386V, BIOS 1.9.2 04/03/2019, should work fine as the system provides EFI v2.60. Using my patch from a few weeks ago[1], efi_random_get_seed() is called

from

Quoted 1 line
Quoted 1 line
Quoted 1 line

arch/x86/boot/compressed/eboot.c::efi_main(). In

efi_random_get_seed(), the

Quoted 39 lines
Quoted 39 lines
Quoted 11 lines

calls to

status = efi_call_early(locate_protocol, &rng_proto, NULL,
(void **)&rng);

and

status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
sizeof(*seed) + EFI_RANDOM_SEED_SIZE, (void **)&seed);

Quoted 7 lines

succeed. However,

        status = rng->get_rng(rng, &rng_algo_raw, EFI_RANDOM_SEED_SIZE,
                              seed->bits);

returns EFI_INVALID_PARAMETER, though I can't see why one of these parameters is invalid.

The UEFI spec defines the conditions under which this function may return EFI_INVALID_PARAMETER as

"""
RNGValue is null or RNGValueLength is zero.

"""

Quoted 10 lines

When trying to use the default rng algorithm (by
modifying the test to "(status != EFI_SUCCESS)"),

                status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
                                      seed->bits);

the call doesn't seem to return.

Any ideas?

Quoted 8 lines
Quoted 3 lines

Try running this from the UEFI shell:

http://people.linaro.org/~ard.biesheuvel/RngTest-X64.efi

Interestingly, this succeeds -- with the default algorithm, SP800-90-CTR-256 and RAW. So I am more confused than before on why the call to ->get_rng() fails in efi_random_get_seed().

It might be an issue with the size of the output.

The original RDRAND based driver in EDK2 contains an apparent misconception that, due to the fact that certain SP800-90-CTR-256 DRBGs require 32 bytes of raw entropy as a seed, it is only valid for the RAW algorithm to be called with an output size of 32.

Quoted 7 lines

So in this case, it might be that 32 is treated as a magic number too, and any other size is rejected by the RAW algorithm.

Not sure why the other one fails, though, but the fact that RAW and SP800-90-CTR-256 are the only supported algorithms suggests that your implementation is at least similar to the RDRAND based RngDxe implementation in EDK2.

I confirmed with our BIOS team for this platform that the implementation of
https://github.com/tianocore/edk2/blob/master/SecurityPkg/RandomNumberGenerator/RngDxe/RngDxe.c
matches EDK2 implementation as of commit https://github.com/tianocore/edk2/commit/289b714b77008aa4200c0be25c4b4e25df04955a

Thanks,

#10 lore raw

RE: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

In reply to Dominik Brodowski

Dominik,

Quoted 11 lines

-----Original Message-----
From: Dominik Brodowski <[email protected]>
Sent: Monday, October 28, 2019 2:21 AM
To: Ard Biesheuvel; Limonciello, Mario
Cc: [email protected]
Subject: efistub: EFI_RNG_PROTOCOL on Dell Inc. Latitude 7390

[EXTERNAL EMAIL]

Ard, Mario,

I had some discussion to one of our BIOS guys and wanted to point a few things out.

Quoted 6 lines

in theory, invoking EFI_RNG_PROTOCOL on a Dell Inc. Latitude 7390/09386V, BIOS 1.9.2 04/03/2019, should work fine as the system provides EFI v2.60. Using my patch from a few weeks ago[1], efi_random_get_seed() is called from arch/x86/boot/compressed/eboot.c::efi_main(). In efi_random_get_seed(), the calls to

Just providing UEFI 2.6 doesn't guarantee an implementation of particular protocols. the protocols are still "optional" but the spec prescribes how they should behave if implemented.

Quoted 20 lines
	status = efi_call_early(locate_protocol, &rng_proto, NULL,
				(void **)&rng);

and

status = efi_call_early(allocate_pool, EFI_RUNTIME_SERVICES_DATA,
sizeof(*seed) + EFI_RANDOM_SEED_SIZE, (void **)&seed);

succeed. However,

	status = rng->get_rng(rng, &rng_algo_raw,
EFI_RANDOM_SEED_SIZE,
			      seed->bits);

returns EFI_INVALID_PARAMETER, though I can't see why one of these parameters is invalid. When trying to use the default rng algorithm (by modifying the test to "(status != EFI_SUCCESS)"),

Quoted 6 lines
		status = rng->get_rng(rng, NULL, EFI_RANDOM_SEED_SIZE,
				      seed->bits);

the call doesn't seem to return.

Any ideas?

You should call RngGetInfo() to find the supported protocols instead of making assumptions on which algorithms are supported.

On the particular platform you mentioned above, these are the only two algorithms supported (which you of course also mentioned in a follow up).

o EFI_RNG_ALGORITHM_SP800_90_CTR_256_GUID
o EFI_RNG_ALGORITHM_RAW

Quoted 6 lines

Not sure why the other one fails, though, but the fact that RAW and SP800-90-CTR-256 are the only supported algorithms suggests that your implementation is at least similar to the RDRAND based RngDxe implementation in EDK2.

I would agree it's unlikely to have changed in any significant way.