crypto/fsl: add RNG support
cover crypto/fsl: add RNG support
First, improve the compatibility on newer Era CAAMs. These introduced new version registers. Secondly, add RNG support for the CAAM. This way we get random number generator support for EFI for free and KASLR will work with ARM64 kernels booted with bootefi.
Michael Walle (4):
crypto/fsl: make SEC%u status line consistent
crypto/fsl: export caam_get_era()
crypto/fsl: support newer SEC modules
crypto/fsl: add RNG support
drivers/crypto/fsl/Kconfig | 11 +++++ drivers/crypto/fsl/Makefile | 1 + drivers/crypto/fsl/jobdesc.c | 9 ++++ drivers/crypto/fsl/jobdesc.h | 3 ++ drivers/crypto/fsl/jr.c | 23 ++++++++-- drivers/crypto/fsl/rng.c | 87 ++++++++++++++++++++++++++++++++++++ drivers/crypto/fsl/sec.c | 2 +- include/fsl_sec.h | 57 +++++++++++++++++++---- 8 files changed, 180 insertions(+), 13 deletions(-) create mode 100644 drivers/crypto/fsl/rng.c
Signature · 1 line
2.20.1
PATCH 1/4 crypto/fsl: make SEC%u status line consistent
Align the status line with all the other output in U-Boot.
Before the change:
DDR 3.9 GiB (DDR3, 32-bit, CL=11, ECC on)
SEC0: RNG instantiated
WDT: Started with servicing (60s timeout)
After the change:
DDR 3.9 GiB (DDR3, 32-bit, CL=11, ECC on) SEC0: RNG instantiated WDT: Started with servicing (60s timeout)
--- drivers/crypto/fsl/jr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/crypto/fsl/jr.c +1 -1
@@ -657,7 +657,7 @@ int sec_init_idx(uint8_t sec_idx)printf ("SEC%u: RNG instantiation failed\n" , sec_idx); return -1; } -printf ("SEC%u: RNG instantiated\n" , sec_idx); +printf ("SEC%u: RNG instantiated\n" , sec_idx); }#endif return ret;
Signature · 1 line
2.20.1
PATCH 1/4 crypto/fsl: make SEC%u status line consistent
drivers/crypto/fsl/jr.c int sec_init_idx(uint8_t sec_idx) · line 660
Quoted 6 lines · Michael Walle
drivers/crypto/fsl/jr.c +1
-1@@ -657,7 +657,7 @@ int sec_init_idx(uint8_t sec_idx)printf ("SEC%u: RNG instantiation failed\n" , sec_idx);return -1 ; } -printf ("SEC%u: RNG instantiated\n" , sec_idx); +printf ("SEC%u: RNG instantiated\n" , sec_idx);
Shouldn't also the string printed in case of failure be updated?
Horia
PATCH 1/4 crypto/fsl: make SEC%u status line consistent
In reply to Michael WalleQuoted 22 lines · Michael Walle
Align the status line with all the other output in U-Boot.
Before the change:
DDR 3.9 GiB (DDR3, 32-bit, CL=11, ECC on)
SEC0: RNG instantiated
WDT: Started with servicing (60s timeout)After the change:
DDR 3.9 GiB (DDR3, 32-bit, CL=11, ECC on) SEC0: RNG instantiated WDT: Started with servicing (60s timeout)Signed-off-by Michael Walle <[email protected]>--- drivers/crypto/fsl/jr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/crypto/fsl/jr.c +0
-0
drivers/crypto/fsl/jr.c int sec_init_idx(uint8_t sec_idx) · line 657
Quoted 1 line
drivers/crypto/fsl/jr.c +0
-0@@ -657,7 +657,7 @@ int sec_init_idx(uint8_t sec_idx)printf ("SEC%u: RNG instantiation failed\n" , sec_idx);
And how about this line?
Quoted 8 lines
drivers/crypto/fsl/jr.c +1
-1return -1; } -printf ("SEC%u: RNG instantiated\n" , sec_idx); +printf ("SEC%u: RNG instantiated\n" , sec_idx); }#endif return ret;
PATCH 2/4 crypto/fsl: export caam_get_era()
In reply to Michael WalleWe need the era in other modules, too. For example, to get the RNG version.
--- drivers/crypto/fsl/sec.c | 2 +- include/fsl_sec.h | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-)
drivers/crypto/fsl/sec.c +1 -1
@@ -98,7 +98,7 @@ void fdt_fixup_crypto_node(void *blob, int sec_rev) fdt_strerror(err)); }#elif CONFIG_SYS_FSL_SEC_COMPAT >=4 /* SEC4 */ -static u8 caam_get_era (void ) +u8 caam_get_era (void ) { static const struct { u16 ip_id;
include/fsl_sec.h +2 -0
@@ -316,6 +316,8 @@ int blob_dek(const u8 *src, u8 *dst, u8 len);int sec_init_idx (uint8_t );#endif int sec_init (void ); + +u8 caam_get_era (void );#endif #endif /* __FSL_SEC_H */
Signature · 1 line
2.20.1
PATCH 2/4 crypto/fsl: export caam_get_era()
Quoted 4 lines · Michael Walle
We need the era in other modules, too. For example, to get the RNG version.
Signed-off-by Michael Walle <[email protected]>
Thanks,
Horia
PATCH 2/4 crypto/fsl: export caam_get_era()
In reply to Michael WalleQuoted 27 lines · Michael Walle
We need the era in other modules, too. For example, to get the RNG version.
Signed-off-by Michael Walle <[email protected]>--- drivers/crypto/fsl/sec.c | 2 +- include/fsl_sec.h | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-)drivers/crypto/fsl/sec.c +1
-1@@ -98,7 +98,7 @@ void fdt_fixup_crypto_node(void *blob, int sec_rev) fdt_strerror(err)); }#elif CONFIG_SYS_FSL_SEC_COMPAT >=4 /* SEC4 */ -static u8 caam_get_era (void ) +u8 caam_get_era (void ) { static const struct { u16 ip_id;include/fsl_sec.h +0
-0
include/fsl_sec.h int blob_dek(const u8 *src, u8 *dst, u8 len); · line 319
Quoted 4 lines
include/fsl_sec.h +1
-0@@ -316,6 +316,8 @@ int blob_dek(const u8 *src, u8 *dst, u8 len);int sec_init_idx (uint8_t );#endif int sec_init (void ); +
Exported functions should be documented. See
https://www.kernel.org/doc/html/latest/doc-guide/kernel-doc.html#function-documentation
Quoted 5 lines
include/fsl_sec.h +1
-0+u8 caam_get_era (void );#endif #endif /* __FSL_SEC_H */
PATCH 3/4 crypto/fsl: support newer SEC modules
In reply to Michael WalleSince Era 10, the version registers changed. Add the version registers and use them on newer modules.
--- drivers/crypto/fsl/jr.c | 12 ++++++++-- include/fsl_sec.h | 51 +++++++++++++++++++++++++++++++++++------ 2 files changed, 54 insertions(+), 9 deletions(-)
drivers/crypto/fsl/jr.c +10 -2
@@ -498,9 +498,17 @@ static int instantiate_rng(uint8_t sec_idx)static u8 get_rng_vid (uint8_t sec_idx) {ccsr_sec_t *sec = (void *)SEC_ADDR (sec_idx); -u32 cha_vid =sec_in32 (&sec->chavid_ls); +u8 vid; -return (cha_vid &SEC_CHAVID_RNG_LS_MASK ) >>SEC_CHAVID_LS_RNG_SHIFT ; +if (caam_get_era () <10 ) { + vid = (sec_in32 (&sec->chavid_ls) &SEC_CHAVID_RNG_LS_MASK ) + >>SEC_CHAVID_LS_RNG_SHIFT ; + }else { + vid = (sec_in32 (&sec->vreg.rng) &CHA_VER_VID_MASK ) + >>CHA_VER_VID_SHIFT ; + } + +return vid; }/*
include/fsl_sec.h +44 -7
83 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH 3/4 crypto/fsl: support newer SEC modules
Quoted 4 lines · Michael Walle
Since Era 10, the version registers changed. Add the version registers and use them on newer modules.
Signed-off-by Michael Walle <[email protected]>
Thanks,
Horia
PATCH 4/4 crypto/fsl: add RNG support
In reply to Michael WalleRegister the random number generator with the rng subsystem in u-boot. This way it can be used by EFI as well as for the 'rng' command.
--- drivers/crypto/fsl/Kconfig | 11 +++++ drivers/crypto/fsl/Makefile | 1 + drivers/crypto/fsl/jobdesc.c | 9 ++++ drivers/crypto/fsl/jobdesc.h | 3 ++ drivers/crypto/fsl/jr.c | 9 ++++ drivers/crypto/fsl/rng.c | 84 ++++++++++++++++++++++++++++++++++++ 6 files changed, 117 insertions(+) create mode 100644 drivers/crypto/fsl/rng.c
drivers/crypto/fsl/Kconfig +11 -0
@@ -45,3 +45,14 @@ config SYS_FSL_SEC_COMPAT config SYS_FSL_SEC_LE bool "Little-endian access to Freescale Secure Boot" + +if FSL_CAAM + +config FSL_CAAM_RNG + bool "Enable Random Number Generator support" + depends on DM_RNG + default y + help + Enable support for the random number generator module of the CAAM. + +endif
drivers/crypto/fsl/Makefile +1 -0
@@ -7,3 +7,4 @@ obj-$(CONFIG_FSL_CAAM) += jr.o fsl_hash.o jobdesc.o error.o obj-$(CONFIG_CMD_BLOB) += fsl_blob.o obj-$(CONFIG_CMD_DEKBLOB) += fsl_blob.o obj-$(CONFIG_RSA_FREESCALE_EXP) += fsl_rsa.o +obj-$(CONFIG_FSL_CAAM_RNG) += rng.o
drivers/crypto/fsl/jobdesc.c +9 -0
@@ -286,6 +286,15 @@ void inline_cnstr_jobdesc_rng_instantiation(uint32_t *desc, int handle) } } +void inline_cnstr_jobdesc_rng (u32 *desc,void *data_out,u32 size) +{ +dma_addr_t dma_data_out =virt_to_phys (data_out); + +init_job_desc (desc,0 ); +append_operation (desc,OP_ALG_ALGSEL_RNG |OP_TYPE_CLASS1_ALG ); +append_fifo_store (desc, dma_data_out, size,FIFOST_TYPE_RNGSTORE ); +} +/* Change key size to bytes form bits in calling function*/ void inline_cnstr_jobdesc_pkha_rsaexp(uint32_t *desc, struct pk_in_params *pkin, uint8_t *out,
drivers/crypto/fsl/jobdesc.h +3 -0
@@ -41,7 +41,10 @@ void inline_cnstr_jobdesc_blob_decap(uint32_t *desc, uint8_t *key_idnfr,void inline_cnstr_jobdesc_rng_instantiation (uint32_t *desc,int handle); +void inline_cnstr_jobdesc_rng (u32 *desc,void *data_out,u32 size); +void inline_cnstr_jobdesc_pkha_rsaexp (uint32_t *desc,struct pk_in_params *pkin,uint8_t *out,uint32_t out_siz); +#endif
drivers/crypto/fsl/jr.c +9 -0
@@ -19,6 +19,7 @@#include <asm/cache.h> #include <asm/fsl_pamu.h> #endif +#include <dm/lists.h> #define CIRC_CNT (head, tail, size) (((head) - (tail)) & (size - 1))#define CIRC_SPACE (head, tail, size) CIRC_CNT((tail), (head) + 1, (size))
@@ -665,6 +666,14 @@ int sec_init_idx(uint8_t sec_idx)printf ("SEC%u: RNG instantiation failed\n" , sec_idx); return -1; } +#ifdef CONFIG_DM_RNG +if (IS_ENABLED (CONFIG_DM_RNG )) { + ret =device_bind_driver (NULL ,"caam-rng" ,"caam-rng" , +NULL ); +if (ret) +printf ("Couldn't bind rng driver (%d)\n" , ret); + } +#endif printf ("SEC%u: RNG instantiated\n" , sec_idx); }#endif
drivers/crypto/fsl/rng.c new file +84 -0
84 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH 4/4 crypto/fsl: add RNG support
Quoted 3 lines · Michael Walle
Register the random number generator with the rng subsystem in u-boot. This way it can be used by EFI as well as for the 'rng' command.
I am trying to understand what's expected from UCLASS_RNG...
UEFI spec mentions:
<<The ?raw? algorithm, when supported, is intended to provide entropy directly from the source, without it going through
some deterministic random bit generator.>>
lib/efi_loader/efi_rng.c uses EFI_RNG_ALGORITHM_RAW and is happy with whatever UCLASS_RNG implementation it gets.
From above I'd conclude all UCLASS_RNG implementations are expected to provide "full" entropy directly from a TRNG source, without any DRBG connected in the middle (b/w TRNG and user).
Is this correct?
If so, note that using CAAM's job ring interface without prediction resistance to extract randomness does not fit the bill, since TRNG output
is connected to a DRBG (DRBG_Hash, SP800-90A compliant).
For CAAM prediction resistance (PR) details I suggest looking in the kernel:
https://lore.kernel.org/linux-crypto/VI1PR0402MB3485EF10976A4A69F90E5B0F98580 at VI1PR0402MB3485.eurprd04.prod.outlook.com 358ba762d9f1 crypto: caam - enable prediction resistance in HRWNGea53756d831a crypto: caam - limit single JD RNG output to maximum of 16 bytes
Steps required to add PR support:
-initialize ("instantiate") the RNG state handles (the DRBGs) with PR support; if already instantiated (by ROM, OP-TEE etc.), they must be re-instantiated if PR support was not enabled -use the "PR" option when extracting randomness from the DRBG; this forces a re-seed of the DRBG
-limit the data size drawn from the DRBG; this is already done (see CAAM_RNG_MAX_FIFO_STORE_SIZE)
drivers/crypto/fsl/jr.c int sec_init_idx(uint8_t sec_idx) · line 670
Quoted 6 lines
drivers/crypto/fsl/jr.c +2
-0@@ -665,6 +666,14 @@ int sec_init_idx(uint8_t sec_idx)printf ("SEC%u: RNG instantiation failed\n" , sec_idx);return -1 ; } +#ifdef CONFIG_DM_RNG +if (IS_ENABLED (CONFIG_DM_RNG )) {
Shouldn't one or the other (#ifdef / IS_ENABLED) suffice?
drivers/crypto/fsl/rng.c line 36
Quoted 11 lines
drivers/crypto/fsl/rng.c new file +11
-0+static int caam_rng_read_one (struct caam_rng_platdata *pdata) +{ +int size =CAAM_RNG_MAX_FIFO_STORE_SIZE ; +int ret; + + ret =run_descriptor_jr (pdata->desc); +if (ret <0 ) +return -EIO ; + +invalidate_dcache_range ((unsigned long )pdata->data, + (unsigned long )pdata->data + size);
Side note: this is not required on Layerscape SoCs, since CAAM is HW coherent. Most surely this needs to be handled in a separate patch set, throughout drivers/crypto/fsl/*.
Horia
PATCH 4/4 crypto/fsl: add RNG support
Quoted 14 lines · Horia Geant?
Quoted 3 lines · Michael Walle
Register the random number generator with the rng subsystem in u-boot. This way it can be used by EFI as well as for the 'rng' command.
I am trying to understand what's expected from
UCLASS_RNG...UEFI spec mentions:
<<The ?raw? algorithm, when supported, is intended to provide entropy directly from the source, without it going through
some deterministic random bit generator.>>
lib/efi_loader/efi_rng.cusesEFI_RNG_ALGORITHM_RAWand is happy with whateverUCLASS_RNGimplementation it gets.
Quoted 8 lines
Quoted 1 line
From above I'd conclude all
UCLASS_RNGimplementations are expected toprovide "full" entropy directly from a TRNG source, without any DRBG connected in the middle (b/w TRNG and user).
Is this correct?
If so, note that using CAAM's job ring interface without prediction resistance to extract randomness does not fit the bill, since TRNG output
is connected to a DRBG (DRBG_Hash, SP800-90A compliant).
I would assume that all hardware RNGs use some algorithms to even out the distribution of the random bits coming from noise source. My understanding of the UEFI spec is that EFI_RNG_ALGORITHM_RAW simply does not provide any guarantee for the distribution quality while a PRNG reseeded via a hardware ring provides some guarantees.
Should you be aware that the FSL hardware RNG does not provide a well distributed entropy it would be preferable to pass the stream through a PRNG even if provided as EFI_RNG_ALGORITHM_RAW.
Best regards
Heinrich
Quoted 40 lines
For CAAM prediction resistance (PR) details I suggest looking in the kernel:
https://lore.kernel.org/linux-crypto/VI1PR0402MB3485EF10976A4A69F90E5B0F98580 at VI1PR0402MB3485.eurprd04.prod.outlook.com358ba762d9f1crypto: caam - enable prediction resistance in HRWNGea53756d831acrypto: caam - limit single JD RNG output to maximum of 16 bytesSteps required to add PR support:
-initialize ("instantiate") the RNG state handles (the DRBGs) with PR support; if already instantiated (by ROM, OP-TEE etc.), they must be re-instantiated if PR support was not enabled -use the "PR" option when extracting randomness from the DRBG; this forces a re-seed of the DRBG
-limit the data size drawn from the DRBG; this is already done (seeCAAM_RNG_MAX_FIFO_STORE_SIZE)drivers/crypto/fsl/jr.c int sec_init_idx(uint8_t sec_idx) · line 670
Quoted 6 lines
drivers/crypto/fsl/jr.c +2
-0@@ -665,6 +666,14 @@ int sec_init_idx(uint8_t sec_idx)printf ("SEC%u: RNG instantiation failed\n" , sec_idx);return -1 ; } +#ifdef CONFIG_DM_RNG +if (IS_ENABLED (CONFIG_DM_RNG )) {Shouldn't one or the other (#ifdef /
IS_ENABLED) suffice?drivers/crypto/fsl/rng.c line 36
Quoted 11 lines
drivers/crypto/fsl/rng.c new file +11
-0+static int caam_rng_read_one (struct caam_rng_platdata *pdata) +{ +int size =CAAM_RNG_MAX_FIFO_STORE_SIZE ; +int ret; + + ret =run_descriptor_jr (pdata->desc); +if (ret <0 ) +return -EIO ; + +invalidate_dcache_range ((unsigned long )pdata->data, + (unsigned long )pdata->data + size);Side note: this is not required on Layerscape SoCs, since CAAM is HW coherent. Most surely this needs to be handled in a separate patch set, throughout
drivers/crypto/fsl/*.Horia
PATCH 4/4 crypto/fsl: add RNG support
Hi Horia, Hi Heinrich,
thanks for reviewing.
Am 2020-06-03 19:35, schrieb Heinrich Schuchardt:
Quoted 34 lines · Horia Geant?
Quoted 17 lines · Michael Walle
Quoted 4 lines
Register the random number generator with the rng subsystem in u-boot.
This way it can be used by EFI as well as for the 'rng' command.I am trying to understand what's expected from
UCLASS_RNG...UEFI spec mentions:
<<The ?raw? algorithm, when supported, is intended to provide entropy directly from the source, without it going through
some deterministic random bit generator.>>
lib/efi_loader/efi_rng.cusesEFI_RNG_ALGORITHM_RAWand is happy with whateverUCLASS_RNGimplementation it gets.Quoted 2 lines
From above I'd conclude all
UCLASS_RNGimplementations are expected toQuoted 8 lines
provide "full" entropy directly from a TRNG source, without any DRBG connected in the middle (b/w TRNG and user).
Is this correct?
If so, note that using CAAM's job ring interface without prediction resistance
to extract randomness does not fit the bill, since TRNG output is connected to a DRBG (DRBG_Hash, SP800-90A compliant).I would assume that all hardware RNGs use some algorithms to even out the distribution of the random bits coming from noise source. My understanding of the UEFI spec is that
EFI_RNG_ALGORITHM_RAWsimply does
not provide any guarantee for the distribution quality while a PRNG reseeded via a hardware ring provides some guarantees.
Quoted 10 lines
Should you be aware that the FSL hardware RNG does not provide a well distributed entropy it would be preferable to pass the stream through a PRNG even if provided as
EFI_RNG_ALGORITHM_RAW.Quoted 7 lines
For CAAM prediction resistance (PR) details I suggest looking in the
kernel:
https://lore.kernel.org/linux-crypto/VI1PR0402MB3485EF10976A4A69F90E5B0F98580 at VI1PR0402MB3485.eurprd04.prod.outlook.com358ba762d9f1crypto: caam - enable prediction resistance in HRWNGea53756d831acrypto: caam - limit single JD RNG output to maximum of
16 bytes
I noticed that PR bit, but I wanted to keep things simple. Does all SEC modules and
versions support this, i.e. the kernel checks versions of some management controller
in QorIQ SoCs.
Quoted 9 lines
Quoted 9 lines
Steps required to add PR support:
-initialize ("instantiate") the RNG state handles (the DRBGs) with PR support; if already instantiated (by ROM, OP-TEE etc.), they must be re-instantiated if PR support was not enabled -use the "PR" option when extracting randomness from the DRBG; this forces a re-seed of the DRBG
-limit the data size drawn from the DRBG; this is already done (seeCAAM_RNG_MAX_FIFO_STORE_SIZE)
Quoted 7 lines
Quoted 7 lines
drivers/crypto/fsl/jr.c int sec_init_idx(uint8_t sec_idx) · line 670
Quoted 6 lines
drivers/crypto/fsl/jr.c +2
-0@@ -665,6 +666,14 @@ int sec_init_idx(uint8_t sec_idx)printf ("SEC%u: RNG instantiation failed\n" , sec_idx);return -1 ; } +#ifdef CONFIG_DM_RNG +if (IS_ENABLED (CONFIG_DM_RNG )) {Shouldn't one or the other (#ifdef /
IS_ENABLED) suffice?
Yes, and it should have only been the if (IS_ENABLED(..)).
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
drivers/crypto/fsl/rng.c new file +7
-0+static int caam_rng_read_one (struct caam_rng_platdata *pdata) +{ +int size =CAAM_RNG_MAX_FIFO_STORE_SIZE ; +int ret; + + ret =run_descriptor_jr (pdata->desc); +if (ret <0 )
Quoted 8 lines
Quoted 8 lines
drivers/crypto/fsl/rng.c line 36
Quoted 4 lines
drivers/crypto/fsl/rng.c new file +4
-0+return -EIO ; + +invalidate_dcache_range ((unsigned long )pdata->data, + (unsigned long )pdata->data + size);Side note: this is not required on Layerscape SoCs, since CAAM is HW coherent.
Most surely this needs to be handled in a separate patch set, throughoutdrivers/crypto/fsl/*.
Is this also true for IMX and the PPC QorIQ SoCs?
-michael
PATCH 4/4 crypto/fsl: add RNG support
Quoted 43 lines · Michael Walle
Hi Horia, Hi Heinrich,
thanks for reviewing.
Am 2020-06-03 19:35, schrieb Heinrich Schuchardt:
Quoted 38 lines · Horia Geant?
Quoted 17 lines · Michael Walle
Quoted 4 lines
Register the random number generator with the rng subsystem in u-boot.
This way it can be used by EFI as well as for the 'rng' command.I am trying to understand what's expected from
UCLASS_RNG...UEFI spec mentions:
<<The ?raw? algorithm, when supported, is intended to provide entropy directly from the source, without it going through
some deterministic random bit generator.>>
lib/efi_loader/efi_rng.cusesEFI_RNG_ALGORITHM_RAWand is happy with whateverUCLASS_RNGimplementation it gets.Quoted 2 lines
From above I'd conclude all
UCLASS_RNGimplementations are expected toQuoted 8 lines
provide "full" entropy directly from a TRNG source, without any DRBG connected in the middle (b/w TRNG and user).
Is this correct?
If so, note that using CAAM's job ring interface without prediction resistance
to extract randomness does not fit the bill, since TRNG output is connected to a DRBG (DRBG_Hash, SP800-90A compliant).I would assume that all hardware RNGs use some algorithms to even out the distribution of the random bits coming from noise source. My understanding of the UEFI spec is that
EFI_RNG_ALGORITHM_RAWsimply does
not provide any guarantee for the distribution quality while a PRNG reseeded via a hardware ring provides some guarantees.Should you be aware that the FSL hardware RNG does not provide a well distributed entropy it would be preferable to pass the stream through a PRNG even if provided as
EFI_RNG_ALGORITHM_RAW.Quoted 1 line
Quoted 8 lines
Quoted 6 lines
Quoted 6 lines
For CAAM prediction resistance (PR) details I suggest looking in the
kernel:
https://lore.kernel.org/linux-crypto/VI1PR0402MB3485EF10976A4A69F90E5B0F98580 at VI1PR0402MB3485.eurprd04.prod.outlook.com358ba762d9f1crypto: caam - enable prediction resistance in HRWNGea53756d831acrypto: caam - limit single JD RNG output to maximum of
16 bytesI noticed that PR bit, but I wanted to keep things simple. Does all SEC
If UCLASS_RNG does not mandate for a TRNG,
i.e. providing randomness from a PRNG / DRBG is acceptable, then adding PR support is not needed.
Quoted 5 lines
modules and
versions support this, i.e. the kernel checks versions of some management controller
in QorIQ SoCs.
All SEC / CAAM having RNG4 block support the PR bit.
I assume the version checking you are referring to is for
the Management Complex (MC) firmware.
MC block is available on SoCs with DPAA2 architecture:
LS1088A, LS2088A, LX2160A
The reason the check was added is that up to a certain point the f/w did not configure the RNG with PR support, and kernel would have to re-initialize the RNG only for those legacy f/w versions.
Quoted 16 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
drivers/crypto/fsl/rng.c new file +7
-0+static int caam_rng_read_one (struct caam_rng_platdata *pdata) +{ +int size =CAAM_RNG_MAX_FIFO_STORE_SIZE ; +int ret; + + ret =run_descriptor_jr (pdata->desc); +if (ret <0 )Quoted 8 lines
Quoted 8 lines
drivers/crypto/fsl/rng.c line 36
Quoted 4 lines
drivers/crypto/fsl/rng.c new file +4
-0+return -EIO ; + +invalidate_dcache_range ((unsigned long )pdata->data, + (unsigned long )pdata->data + size);Side note: this is not required on Layerscape SoCs, since CAAM is HW coherent.
Most surely this needs to be handled in a separate patch set, throughoutdrivers/crypto/fsl/*.
Quoted 2 lines
Is this also true for IMX and the PPC QorIQ SoCs?
CAAM is _NOT_ HW coherent on i.MX SoCs,
while on the other PPC / ARM SoCs (QorIQ, Layerscape) it is.
Horia
PATCH 4/4 crypto/fsl: add RNG support
In reply to Michael WalleQuoted 21 lines · Michael Walle
Register the random number generator with the rng subsystem in u-boot. This way it can be used by EFI as well as for the 'rng' command.
Signed-off-by Michael Walle <[email protected]>--- drivers/crypto/fsl/Kconfig | 11 +++++ drivers/crypto/fsl/Makefile | 1 + drivers/crypto/fsl/jobdesc.c | 9 ++++ drivers/crypto/fsl/jobdesc.h | 3 ++ drivers/crypto/fsl/jr.c | 9 ++++ drivers/crypto/fsl/rng.c | 84 ++++++++++++++++++++++++++++++++++++ 6 files changed, 117 insertions(+) create mode 100644 drivers/crypto/fsl/rng.cdrivers/crypto/fsl/Kconfig +0
-0@@ -45,3 +45,14 @@ config SYS_FSL_SEC_COMPAT config SYS_FSL_SEC_LE
drivers/crypto/fsl/Kconfig config SYS_FSL_SEC_COMPAT · line 56
Quoted 10 lines
drivers/crypto/fsl/Kconfig +9
-0bool "Little-endian access to Freescale Secure Boot" + +if FSL_CAAM + +config FSL_CAAM_RNG + bool "Enable Random Number Generator support" + depends on DM_RNG + default y + help + Enable support for the random number generator module of the CAAM.
Hello Michael,
when typing CAAM into Google I got a lot of answers but "Cryptographic Accelerator and Assurance Module" was not under the first 50 hits.
If this is a hardware RNG I think we should put this into the text.
So how about:
"Enable support the hardware random number generator of Freescale SOCs using the Cryptographic Accelerator and Assurance Module (CAAM)."
Best regards
Heinrich
PATCH 4/4 crypto/fsl: add RNG support
Quoted 35 lines · Heinrich Schuchardt
Quoted 21 lines · Michael Walle
Register the random number generator with the rng subsystem in u-boot. This way it can be used by EFI as well as for the 'rng' command.
Signed-off-by Michael Walle <[email protected]>--- drivers/crypto/fsl/Kconfig | 11 +++++ drivers/crypto/fsl/Makefile | 1 + drivers/crypto/fsl/jobdesc.c | 9 ++++ drivers/crypto/fsl/jobdesc.h | 3 ++ drivers/crypto/fsl/jr.c | 9 ++++ drivers/crypto/fsl/rng.c | 84 ++++++++++++++++++++++++++++++++++++ 6 files changed, 117 insertions(+) create mode 100644 drivers/crypto/fsl/rng.cdrivers/crypto/fsl/Kconfig +0
-0@@ -45,3 +45,14 @@ config SYS_FSL_SEC_COMPAT config SYS_FSL_SEC_LEdrivers/crypto/fsl/Kconfig config SYS_FSL_SEC_COMPAT · line 56
Quoted 10 lines
drivers/crypto/fsl/Kconfig +9
-0bool "Little-endian access to Freescale Secure Boot" + +if FSL_CAAM + +config FSL_CAAM_RNG + bool "Enable Random Number Generator support" + depends on DM_RNG + default y + help + Enable support for the random number generator module of the CAAM.Hello Michael,
Quoted 5 lines
when typing CAAM into Google I got a lot of answers but "Cryptographic Accelerator and Assurance Module" was not under the first 50 hits.
If this is a hardware RNG I think we should put this into the text.
Totally agree.
Besides other cryptographic services, CAAM offers:
-a hardware RNG / TRNG
-a PRNG / DRBG (SP800-90A compliant DRBG_Hash) - which is seeded from the TRNG
Both are accessible by SW, so clarifying what the driver does would be useful (unless DM_RNG / UCLASS_RNG already implies one or the other).
From what I see, driver added by Michael is using the PRNG / DRBG and not the TRNG. Is this acceptable?
Conceptually this is similar to choosing between
RDSEED vs. RDRDAND x86 instructions:
https://software.intel.com/content/www/us/en/develop/blogs/the-difference-between-rdrand-and-rdseed.html
Quoted 5 lines
So how about:
"Enable support the hardware random number generator of Freescale SOCs using the Cryptographic Accelerator and Assurance Module (CAAM)."
The CAAM acronym is expanded at the top of the same file, under FSL_CAAM's help:
<<Enables the Freescale's Cryptographic Accelerator and Assurance Module (CAAM), also known as the SEC version 4 (SEC4). The driver uses Job Ring as interface to communicate with CAAM.>>
Horia
PATCH 4/4 crypto/fsl: add RNG support
Hi Horia, Hi Heinrich,
Am 2020-06-04 10:05, schrieb Horia Geant?:
Quoted 39 lines · Heinrich Schuchardt
Quoted 38 lines · Michael Walle
Quoted 33 lines
Register the random number generator with the rng subsystem in u-boot.
This way it can be used by EFI as well as for the 'rng' command.Signed-off-by Michael Walle <[email protected]>--- drivers/crypto/fsl/Kconfig | 11 +++++ drivers/crypto/fsl/Makefile | 1 + drivers/crypto/fsl/jobdesc.c | 9 ++++ drivers/crypto/fsl/jobdesc.h | 3 ++ drivers/crypto/fsl/jr.c | 9 ++++ drivers/crypto/fsl/rng.c | 84 ++++++++++++++++++++++++++++++++++++ 6 files changed, 117 insertions(+) create mode 100644 drivers/crypto/fsl/rng.cdrivers/crypto/fsl/Kconfig +9
-0@@ -45,3 +45,14 @@ config SYS_FSL_SEC_COMPAT config SYS_FSL_SEC_LE bool "Little-endian access to Freescale Secure Boot" + +if FSL_CAAM + +config FSL_CAAM_RNG + bool "Enable Random Number Generator support" + depends on DM_RNG + default y + help + Enable support for the random number generator module of theQuoted 1 line
CAAM.
Hello Michael,
Quoted 6 lines
Quoted 5 lines
when typing CAAM into Google I got a lot of answers but "Cryptographic Accelerator and Assurance Module" was not under the first 50 hits.
If this is a hardware RNG I think we should put this into the text.
Totally agree.
Well I was under the impression that UCLASS_RNG is just for hardware RNGs.
config DM_RNG
bool "Driver support for Random Number Generator devices"
Whatever "device" means in that context. But I can certainly add that this is a h/w rng.
Quoted 4 lines
Besides other cryptographic services, CAAM offers:
(quoted diff) +0
-2-a hardware RNG / TRNG -a PRNG / DRBG (SP800-90A compliant DRBG_Hash) - which is seededfrom the TRNG
Together with that.
Quoted 6 lines
Both are accessible by SW, so clarifying what the driver does would be useful (unless
DM_RNG/UCLASS_RNGalready implies one or the other).From what I see, driver added by Michael is using the PRNG / DRBG and not the TRNG. Is this acceptable?
Well there is no, expectation from UCLASS_RNG. EFI "blindly" uses the first RNG device.. so it is just a "better than nothing".
RNG is also used for the BLOB protocol. Will it interfere this if I instantiate the RNG with PR?
Quoted 6 lines
Conceptually this is similar to choosing between
RDSEED vs. RDRDAND x86 instructions:
https://software.intel.com/content/www/us/en/develop/blogs/the-difference-between-rdrand-and-rdseed.htmlQuoted 2 lines
So how about:
Quoted 8 lines
Quoted 3 lines
"Enable support the hardware random number generator of Freescale SOCs using the Cryptographic Accelerator and Assurance Module (CAAM)."
The CAAM acronym is expanded at the top of the same file, under FSL_CAAM's help:
<<Enables the Freescale's Cryptographic Accelerator and Assurance Module (CAAM), also known as the SEC version 4 (SEC4). The driver uses Job Ring as interface to communicate with CAAM.>>
This isn't apparent from the patch. But please note that the new kconfig option is "if FSL_CAAM", where CAAM is explained.
-michael
PATCH 4/4 crypto/fsl: add RNG support
In reply to Horia GeantăQuoted 48 lines · Horia Geant?
Quoted 35 lines · Heinrich Schuchardt
Quoted 21 lines · Michael Walle
Register the random number generator with the rng subsystem in u-boot. This way it can be used by EFI as well as for the 'rng' command.
Signed-off-by Michael Walle <[email protected]>--- drivers/crypto/fsl/Kconfig | 11 +++++ drivers/crypto/fsl/Makefile | 1 + drivers/crypto/fsl/jobdesc.c | 9 ++++ drivers/crypto/fsl/jobdesc.h | 3 ++ drivers/crypto/fsl/jr.c | 9 ++++ drivers/crypto/fsl/rng.c | 84 ++++++++++++++++++++++++++++++++++++ 6 files changed, 117 insertions(+) create mode 100644 drivers/crypto/fsl/rng.cdrivers/crypto/fsl/Kconfig +0
-0@@ -45,3 +45,14 @@ config SYS_FSL_SEC_COMPAT config SYS_FSL_SEC_LEdrivers/crypto/fsl/Kconfig config SYS_FSL_SEC_COMPAT · line 56
Quoted 10 lines
drivers/crypto/fsl/Kconfig +9
-0bool "Little-endian access to Freescale Secure Boot" + +if FSL_CAAM + +config FSL_CAAM_RNG + bool "Enable Random Number Generator support" + depends on DM_RNG + default y + help + Enable support for the random number generator module of the CAAM.Hello Michael,
Quoted 5 lines
when typing CAAM into Google I got a lot of answers but "Cryptographic Accelerator and Assurance Module" was not under the first 50 hits.
If this is a hardware RNG I think we should put this into the text.
Totally agree.
Besides other cryptographic services, CAAM offers:
(quoted diff) +0
-2-a hardware RNG / TRNG -a PRNG / DRBG (SP800-90A compliant DRBG_Hash) - which is seededfrom the TRNG
Quoted 3 lines
Both are accessible by SW, so clarifying what the driver does would be useful (unless
DM_RNG/UCLASS_RNGalready implies one or the other).
The idea of DM_RNG is to collect entropy from hardware RNGs needed for implementing the EFI_RNG_PROTOCOL. Our implementation of the EFI_RNG_PROTOCOL up to now can only supply raw entropy. The UEFI specification allows to additionally implement certain PRNG algorithms seeded by the raw entropy which the caller can choose. So in a later embodiment it may make sense to use what exists as hardware acceleration for these.
Cf. UEFI Specification Version 2.8 (Errata A) (released February 2020)
https://uefi.org/sites/default/files/resources/UEFI_Spec_2_8_A_Feb14.pdf
Quoted 4 lines
Quoted 2 lines
From what I see, driver added by Michael is using the PRNG / DRBG
and not the TRNG. Is this acceptable?
If it is only PRNG, this is not what we look for. If a PRNG/DRBG is used to ameliorate the raw entropy stream like Linux does for the /dev/random device this is fine. We need something non-deterministic.
Isn't this what Linux' drivers/crypto/caam/caamrng.c does?
Best regards
Heinrich
Quoted 17 lines
Conceptually this is similar to choosing between
RDSEED vs. RDRDAND x86 instructions:
https://software.intel.com/content/www/us/en/develop/blogs/the-difference-between-rdrand-and-rdseed.htmlQuoted 5 lines
So how about:
"Enable support the hardware random number generator of Freescale SOCs using the Cryptographic Accelerator and Assurance Module (CAAM)."
The CAAM acronym is expanded at the top of the same file, under FSL_CAAM's help:
<<Enables the Freescale's Cryptographic Accelerator and Assurance Module (CAAM), also known as the SEC version 4 (SEC4). The driver uses Job Ring as interface to communicate with CAAM.>>Horia
PATCH 4/4 crypto/fsl: add RNG support
Am 2020-06-04 14:26, schrieb Heinrich Schuchardt:
Quoted 7 lines · Horia Geant?
Quoted 5 lines · Heinrich Schuchardt
From what I see, driver added by Michael is using the PRNG / DRBG and not the TRNG. Is this acceptable?
Quoted 5 lines
If it is only PRNG, this is not what we look for. If a PRNG/DRBG is used
to ameliorate the raw entropy stream like Linux does for the/dev/random
device this is fine. We need something non-deterministic.
What do you mean by "only PRNG"?
Quoted 2 lines
Quoted 2 lines
-a PRNG / DRBG (SP800-90A compliant
DRBG_Hash) - which is seeded from the TRNG
So while it is a PRNG, it is non-deterministic because its seeded from the TRNG.
-michael
PATCH 4/4 crypto/fsl: add RNG support
Quoted 12 lines · Michael Walle
Am 2020-06-04 14:26, schrieb Heinrich Schuchardt:
Quoted 10 lines · Horia Geant?
Quoted 5 lines · Heinrich Schuchardt
From what I see, driver added by Michael is using the PRNG / DRBG and not the TRNG. Is this acceptable?
If it is only PRNG, this is not what we look for. If a PRNG/DRBG is used to ameliorate the raw entropy stream like Linux does for the
/dev/randomdevice this is fine. We need something non-deterministic.
Quoted 7 lines
What do you mean by "only PRNG"?
Quoted 2 lines
Quoted 2 lines
-a PRNG / DRBG (SP800-90A compliant
DRBG_Hash) - which is seeded from the TRNGSo while it is a PRNG, it is non-deterministic because its seeded from the TRNG.
If for every byte that your DM_RNG driver outputs at least one byte from the TRNG is consumed, it is fine. Otherwise it is not what we are looking for.
Best regards
Heinrich
PATCH 4/4 crypto/fsl: add RNG support
Am 2020-06-04 14:58, schrieb Heinrich Schuchardt:
Quoted 20 lines · Michael Walle
Quoted 19 lines
Am 2020-06-04 14:26, schrieb Heinrich Schuchardt:
Quoted 7 lines · Horia Geant?
Quoted 5 lines · Heinrich Schuchardt
From what I see, driver added by Michael is using the PRNG / DRBG and not the TRNG. Is this acceptable?
Quoted 5 lines
If it is only PRNG, this is not what we look for. If a PRNG/DRBG is used
to ameliorate the raw entropy stream like Linux does for the/dev/random
device this is fine. We need something non-deterministic.What do you mean by "only PRNG"?
Quoted 2 lines
Quoted 2 lines
-a PRNG / DRBG (SP800-90A compliant
DRBG_Hash) - which is seeded from the TRNG
Quoted 7 lines
Quoted 2 lines
So while it is a PRNG, it is non-deterministic because its seeded from the TRNG.
If for every byte that your
DM_RNGdriver outputs at least one byte from
the TRNG is consumed, it is fine. Otherwise it is not what we are looking for.
And why is that? This should really be documented somewhere.
-michael
PATCH 4/4 crypto/fsl: add RNG support
Quoted 21 lines · Michael Walle
Am 2020-06-04 14:58, schrieb Heinrich Schuchardt:
Quoted 20 lines · Michael Walle
Quoted 19 lines
Am 2020-06-04 14:26, schrieb Heinrich Schuchardt:
Quoted 7 lines · Horia Geant?
Quoted 5 lines · Heinrich Schuchardt
From what I see, driver added by Michael is using the PRNG / DRBG and not the TRNG. Is this acceptable?
Quoted 5 lines
If it is only PRNG, this is not what we look for. If a PRNG/DRBG is used
to ameliorate the raw entropy stream like Linux does for the/dev/random
device this is fine. We need something non-deterministic.What do you mean by "only PRNG"?
Quoted 2 lines
Quoted 2 lines
-a PRNG / DRBG (SP800-90A compliant
DRBG_Hash) - which is seeded from the TRNG
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
So while it is a PRNG, it is non-deterministic because its seeded from the TRNG.
If for every byte that your
DM_RNGdriver outputs at least one byte from the TRNG is consumed, it is fine. Otherwise it is not what we are looking for.And why is that? This should really be documented somewhere.
We want to provide raw entropy in the EFI_RNG_PROTOCOL. So this cannot be a deterministic sequence of bytes where you only have to know the current state of a PRNG to find the next byte.
As mentioned above you have a TRNG available. What is problematic about providing its output?
Best regards
Heinrich
PATCH 4/4 crypto/fsl: add RNG support
Am 2020-06-04 17:45, schrieb Heinrich Schuchardt:
Quoted 30 lines · Michael Walle
Quoted 29 lines
Am 2020-06-04 14:58, schrieb Heinrich Schuchardt:
Quoted 27 lines · Michael Walle
Quoted 14 lines
Am 2020-06-04 14:26, schrieb Heinrich Schuchardt:
Quoted 7 lines · Horia Geant?
Quoted 5 lines · Heinrich Schuchardt
From what I see, driver added by Michael is using the PRNG / DRBG and not the TRNG. Is this acceptable?
Quoted 5 lines
If it is only PRNG, this is not what we look for. If a PRNG/DRBG is used
to ameliorate the raw entropy stream like Linux does for the/dev/random
device this is fine. We need something non-deterministic.Quoted 7 lines
What do you mean by "only PRNG"?
Quoted 2 lines
Quoted 2 lines
-a PRNG / DRBG (SP800-90A compliant
DRBG_Hash) - which is seeded from the TRNGSo while it is a PRNG, it is non-deterministic because its seeded from the TRNG.
If for every byte that your
DM_RNGdriver outputs at least one byte from
the TRNG is consumed, it is fine. Otherwise it is not what we are looking for.
Quoted 5 lines
Quoted 1 line
And why is that? This should really be documented somewhere.
We want to provide raw entropy in the
EFI_RNG_PROTOCOL. So this cannot be a deterministic sequence of bytes where you only have to know the current state of a PRNG to find the next byte.
I wasn't aware of the fact that UCLASS_RNG was solely for EFI_RNG_PROTOCOL. And there are no requirements for the UCLASS_RNG, are there?
TBH I find this somewhat overkill for just having a random seed for KASLR. Everyone is complaining about the size of the bootloader steadily increasing, but then we throw in more and more for what use? Even the UEFI
spec states:
When a Deterministic Random Bit Generator (DRBG) is used on the output of a (raw) entropy source, its security level must be at least 256
bits.
Why does linux use ALGORITHM_RAW? What happens if that is not supported?
Quoted 2 lines
As mentioned above you have a TRNG available. What is problematic about providing its output?
See v2, it should be now be the TRNG output, or at least it it reseeded on every read and the read is limited to 16 bytes, like Horia said in its very first reply. So I conclude the PRNG is at least seeded with 16 bytes.
-michael