nvme: In-band authentication support
Subsystems (5)
Hannes Reinecke
maintainerv15 cover nvme: In-band authentication support
Hi all,
recent updates to the NVMe spec have added definitions for in-band authentication, and seeing that it provides some real benefit especially for NVMe-TCP here's an attempt to implement it.
Thanks to Nicolai Stange the crypto DH framework has been upgraded to provide us with a FFDHE implementation; I've updated the patchset to use the ephemeral key generation provided there.
Note that this is just for in-band authentication. Secure concatenation (ie starting TLS with the negotiated parameters) requires a TLS handshake, which the in-kernel TLS implementation does not provide. This is being worked on with a different patchset which is still WIP.
The nvme-cli support has already been merged; please use the latest nvme-cli git repository to build the most recent version.
A copy of this patchset can be found atgit://git.kernel.org/pub/scm/linux/kernel/git/hare/scsi-devel
branch auth.v15
The patchset is being cut against nvme-5.19.
As usual, comments and reviews are welcome.
Changes to v14:
- Rebase to nvme-5.19
- Fixup sparse warnings as noted by Christoph
- Fixup Kconfig dependencies
- Add NVME_CONNECT_AUTHREQ definitions
Changes to v13:
- Fixup sysfs attribute visibility (reported by Chaitanya)
- Fixup SUCCESS1 handling
- Re-introduce crypto_has_shash() and crypto_has_kpp() helper
- Unse 'authenticated' flag on failure
Changes to v12:
- Replaces crypto_has_shash() and crypto_has_kpp() helper with CONFIG_ checks
- Fixed kbuild robot warning in pr_fmt()
Changes to v11:
- Fixup type for FAILURE2 message (Prashant Nayak)
- Do not sent SUCCESS2 if bi-directional authentication is not requested (Martin George)
Changes to v10:
- Fixup error return value when authentication failed
Changes to v9:
- Include review from Chaitanya
- Use sparse array for dhgroup and hash lookup
- Common function for auth_send and auth_receive
Changes to v8:
- Rebased to Nicolais crypto DH rework
- Fixed oops on non-fabrics devices
Changes to v7:
- Space out hash list and dhgroup list in nvme negotiate data to be conformant with the spec
- Update sequence number handling to start with a random value and ignore '0' as mandated by the spec
- Update nvme_auth_generate_key to return the key as suggested by Sagi
- Add nvmet_parse_fabrics_io_cmd() as suggested by hch
Changes to v6:
- Use 'u8' for DH group id and hash id
- Use 'struct nvme_dhchap_key'
- Rename variables to drop 'DHCHAP'
- Include reviews from Chaitanya
Changes to v5:
- Unify nvme_auth_generate_key()
- Unify nvme_auth_extract_key()
- Fixed bug where re-authentication with wrong controller key would not fail
- Include reviews from Sagi
Changes to v4:
- Validate against blktest suite
- Fixup base64 decoding
- Transform secret with correct hmac algorithm
Changes to v3:
- Renamed parameter to 'dhchap_ctrl_key'
- Fixed bi-directional authentication
- Included reviews from Sagi
- Fixed base64 algorithm for transport encoding
Changes to v2:
- Dropped non-standard algorithms
- Reworked base64 based on fs/crypto/fname.c
- Fixup crash with no keys
Changes to the original submission:
- Included reviews from Vladislav
- Included reviews from Sagi
- Implemented re-authentication support
- Fixed up key handling
*
** BLURB HERE ***
Hannes Reinecke (11):
crypto: add crypto_has_shash()
crypto: add crypto_has_kpp()
lib/base64: RFC4648-compliant base64 encoding
nvme: add definitions for NVMe In-Band authentication
nvme-fabrics: decode 'authentication required' connect error
nvme: Implement In-Band authentication
nvme-auth: Diffie-Hellman key exchange support
nvmet: parse fabrics commands on io queues
nvmet: Implement basic In-Band Authentication
nvmet-auth: Diffie-Hellman key exchange support
nvmet-auth: expire authentication sessions
crypto/kpp.c | 6 + crypto/shash.c | 6 + drivers/nvme/Kconfig | 1 + drivers/nvme/Makefile | 1 + drivers/nvme/common/Kconfig | 4 + drivers/nvme/common/Makefile | 7 + drivers/nvme/common/auth.c | 481 ++++++++++++ drivers/nvme/host/Kconfig | 15 + drivers/nvme/host/Makefile | 1 + drivers/nvme/host/auth.c | 1000 ++++++++++++++++++++++++ drivers/nvme/host/core.c | 143 +++- drivers/nvme/host/fabrics.c | 83 +- drivers/nvme/host/fabrics.h | 7 + drivers/nvme/host/nvme.h | 30 + drivers/nvme/host/rdma.c | 1 + drivers/nvme/host/tcp.c | 1 + drivers/nvme/host/trace.c | 32 + drivers/nvme/target/Kconfig | 15 + drivers/nvme/target/Makefile | 1 + drivers/nvme/target/admin-cmd.c | 4 +- drivers/nvme/target/auth.c | 525 +++++++++++++ drivers/nvme/target/configfs.c | 138 +++- drivers/nvme/target/core.c | 15 + drivers/nvme/target/fabrics-cmd-auth.c | 545 +++++++++++++ drivers/nvme/target/fabrics-cmd.c | 55 +- drivers/nvme/target/nvmet.h | 75 +- include/crypto/hash.h | 2 + include/crypto/kpp.h | 2 + include/linux/base64.h | 16 + include/linux/nvme-auth.h | 41 + include/linux/nvme.h | 209 ++++- lib/Makefile | 2 +- lib/base64.c | 103 +++ 33 files changed, 3552 insertions(+), 15 deletions(-) create mode 100644 drivers/nvme/common/Kconfig create mode 100644 drivers/nvme/common/Makefile create mode 100644 drivers/nvme/common/auth.c create mode 100644 drivers/nvme/host/auth.c create mode 100644 drivers/nvme/target/auth.c create mode 100644 drivers/nvme/target/fabrics-cmd-auth.c create mode 100644 include/linux/base64.h create mode 100644 include/linux/nvme-auth.h create mode 100644 lib/base64.c
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 1/11 crypto: add crypto_has_shash()
Add helper function to determine if a given synchronous hash is supported.
--- crypto/shash.c | 6 ++++++ include/crypto/hash.h | 2 ++ 2 files changed, 8 insertions(+)
crypto/shash.c +6 -0
@@ -521,6 +521,12 @@ struct crypto_shash *crypto_alloc_shash(const char *alg_name, u32 type, }EXPORT_SYMBOL_GPL (crypto_alloc_shash); +int crypto_has_shash (const char *alg_name,u32 type,u32 mask) +{ +return crypto_type_has_alg (alg_name, &crypto_shash_type, type, mask); +} +EXPORT_SYMBOL_GPL (crypto_has_shash); +static int shash_prepare_alg (struct shash_alg *alg) {struct crypto_alg *base = &alg->base;
include/crypto/hash.h +2 -0
@@ -718,6 +718,8 @@ static inline void ahash_request_set_crypt(struct ahash_request *req,struct crypto_shash *crypto_alloc_shash (const char *alg_name,u32 type,u32 mask); +int crypto_has_shash (const char *alg_name,u32 type,u32 mask); +static inline struct crypto_tfm *crypto_shash_tfm (struct crypto_shash *tfm) {return &tfm->base;
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 2/11 crypto: add crypto_has_kpp()
In reply to Hannes ReineckeAdd helper function to determine if a given key-agreement protocol primitive is supported.
--- crypto/kpp.c | 6 ++++++ include/crypto/kpp.h | 2 ++ 2 files changed, 8 insertions(+)
crypto/kpp.c +6 -0
@@ -104,6 +104,12 @@ int crypto_grab_kpp(struct crypto_kpp_spawn *spawn, }EXPORT_SYMBOL_GPL (crypto_grab_kpp); +int crypto_has_kpp (const char *alg_name,u32 type,u32 mask) +{ +return crypto_type_has_alg (alg_name, &crypto_kpp_type, type, mask); +} +EXPORT_SYMBOL_GPL (crypto_has_kpp); +static void kpp_prepare_alg (struct kpp_alg *alg) {struct crypto_alg *base = &alg->base;
include/crypto/kpp.h +2 -0
@@ -104,6 +104,8 @@ struct kpp_alg {*/ struct crypto_kpp *crypto_alloc_kpp (const char *alg_name,u32 type,u32 mask); +int crypto_has_kpp (const char *alg_name,u32 type,u32 mask); +static inline struct crypto_tfm *crypto_kpp_tfm (struct crypto_kpp *tfm) {return &tfm->base;
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 3/11 lib/base64: RFC4648-compliant base64 encoding
In reply to Hannes ReineckeAdd RFC4648-compliant base64 encoding and decoding routines, based on the base64url encoding in fs/crypto/fname.c.
--- include/linux/base64.h | 16 +++++++ lib/Makefile | 2 +- lib/base64.c | 103 +++++++++++++++++++++++++++++++++++++++++ 3 files changed, 120 insertions(+), 1 deletion(-) create mode 100644 include/linux/base64.h create mode 100644 lib/base64.c
include/linux/base64.h new file +16 -0
@@ -0,0 +1,16 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* base64 encoding, lifted from fs/crypto/fname.c. +*/ + +#ifndef _LINUX_BASE64_H +#define _LINUX_BASE64_H + +#include <linux/types.h> + +#define BASE64_CHARS (nbytes) DIV_ROUND_UP((nbytes) * 4, 3) + +int base64_encode (const u8 *src,int len,char *dst); +int base64_decode (const char *src,int len,u8 *dst); + +#endif /* _LINUX_BASE64_H */
lib/Makefile +1 -1
@@ -46,7 +46,7 @@ obj-y += bcd.o sort.o parser.o debug_locks.o random32.o \ bust_spinlocks.o kasprintf.o bitmap.o scatterlist.o \ list_sort.o uuid.o iov_iter.o clz_ctz.o \ bsearch.o find_bit.o llist.o memweight.o kfifo.o \ - percpu-refcount.o rhashtable.o \ + percpu-refcount.o rhashtable.o base64.o \ once.o refcount.o usercopy.o errseq.o bucket_locks.o \ generic-radix-tree.o obj-$(CONFIG_STRING_SELFTEST) += test_string.o
lib/base64.c new file +103 -0
103 lines not on this page · raw message
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 4/11 nvme: add definitions for NVMe In-Band authentication
In reply to Hannes ReineckeAdd new definitions for NVMe In-band authentication as defined in the NVMe Base Specification v2.0.
--- include/linux/nvme.h | 209 ++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 208 insertions(+), 1 deletion(-)
include/linux/nvme.h +208 -1
@@ -19,6 +19,7 @@#define NVMF_TRSVCID_SIZE 32#define NVMF_TRADDR_SIZE 256#define NVMF_TSAS_SIZE 256 +#define NVMF_AUTH_HASH_LEN 64#define NVME_DISC_SUBSYS_NAME "nqn.2014-08.org.nvmexpress.discovery"
@@ -1367,6 +1368,8 @@ enum nvmf_capsule_command { nvme_fabrics_type_property_set =0x00 , nvme_fabrics_type_connect =0x01 , nvme_fabrics_type_property_get =0x04 , + nvme_fabrics_type_auth_send =0x05 , + nvme_fabrics_type_auth_receive = 0x06, };#define nvme_fabrics_type_name (type) { type, #type }
@@ -1374,7 +1377,9 @@ enum nvmf_capsule_command {__print_symbolic (type, \nvme_fabrics_type_name (nvme_fabrics_type_property_set), \nvme_fabrics_type_name (nvme_fabrics_type_connect), \ -nvme_fabrics_type_name (nvme_fabrics_type_property_get)) +nvme_fabrics_type_name (nvme_fabrics_type_property_get), \ +nvme_fabrics_type_name (nvme_fabrics_type_auth_send), \ +nvme_fabrics_type_name (nvme_fabrics_type_auth_receive))/* * If not fabrics command, fctype will be ignored.
@@ -1470,6 +1475,11 @@ struct nvmf_connect_command {__u8 resv4[12 ]; }; +enum { +NVME_CONNECT_AUTHREQ_ASCR = (1 <<2 ), +NVME_CONNECT_AUTHREQ_ATR = (1 <<1 ), +}; +struct nvmf_connect_data {uuid_t hostid;__le16 cntlid;
@@ -1504,6 +1514,200 @@ struct nvmf_property_get_command {__u8 resv4[16 ]; }; +struct nvmf_auth_common_command { +__u8 opcode; +__u8 resv1; +__u16 command_id; +__u8 fctype; +__u8 resv2[19 ]; +union nvme_data_ptr dptr; +__u8 resv3; +__u8 spsp0; +__u8 spsp1; +__u8 secp; +__le32 al_tl; +__u8 resv4[16 ]; +}; + +struct nvmf_auth_send_command { +__u8 opcode; +__u8 resv1; +__u16 command_id; +__u8 fctype; +__u8 resv2[19 ]; +union nvme_data_ptr dptr; +__u8 resv3; +__u8 spsp0; +__u8 spsp1; +__u8 secp; +__le32 tl; +__u8 resv4[16 ]; +}; + +struct nvmf_auth_receive_command { +__u8 opcode; +__u8 resv1; +__u16 command_id; +__u8 fctype; +__u8 resv2[19 ]; +union nvme_data_ptr dptr; +__u8 resv3; +__u8 spsp0; +__u8 spsp1; +__u8 secp; +__le32 al; +__u8 resv4[16 ]; +}; + +/* Value for secp */ +enum { +NVME_AUTH_DHCHAP_PROTOCOL_IDENTIFIER =0xe9 , +}; + +/* Defined value for auth_type */ +enum { +NVME_AUTH_COMMON_MESSAGES =0x00 , +NVME_AUTH_DHCHAP_MESSAGES =0x01 , +}; + +/* Defined messages for auth_id */ +enum { +NVME_AUTH_DHCHAP_MESSAGE_NEGOTIATE =0x00 , +NVME_AUTH_DHCHAP_MESSAGE_CHALLENGE =0x01 , +NVME_AUTH_DHCHAP_MESSAGE_REPLY =0x02 , +NVME_AUTH_DHCHAP_MESSAGE_SUCCESS1 =0x03 , +NVME_AUTH_DHCHAP_MESSAGE_SUCCESS2 =0x04 , +NVME_AUTH_DHCHAP_MESSAGE_FAILURE2 =0xf0 , +NVME_AUTH_DHCHAP_MESSAGE_FAILURE1 =0xf1 , +}; + +struct nvmf_auth_dhchap_protocol_descriptor { +__u8 authid; +__u8 rsvd; +__u8 halen; +__u8 dhlen; +__u8 idlist[60 ]; +}; + +enum { +NVME_AUTH_DHCHAP_AUTH_ID =0x01 , +}; + +/* Defined hash functions for DH-HMAC-CHAP authentication */ +enum { +NVME_AUTH_HASH_SHA256 =0x01 , +NVME_AUTH_HASH_SHA384 =0x02 , +NVME_AUTH_HASH_SHA512 =0x03 , +NVME_AUTH_HASH_INVALID =0xff , +}; + +/* Defined Diffie-Hellman group identifiers for DH-HMAC-CHAP authentication */ +enum { +NVME_AUTH_DHGROUP_NULL =0x00 , +NVME_AUTH_DHGROUP_2048 =0x01 , +NVME_AUTH_DHGROUP_3072 =0x02 , +NVME_AUTH_DHGROUP_4096 =0x03 , +NVME_AUTH_DHGROUP_6144 =0x04 , +NVME_AUTH_DHGROUP_8192 =0x05 , +NVME_AUTH_DHGROUP_INVALID =0xff , +}; + +union nvmf_auth_protocol { +struct nvmf_auth_dhchap_protocol_descriptor dhchap; +}; + +struct nvmf_auth_dhchap_negotiate_data { +__u8 auth_type; +__u8 auth_id; +__le16 rsvd; +__le16 t_id; +__u8 sc_c; +__u8 napd; +union nvmf_auth_protocol auth_protocol[]; +}; + +struct nvmf_auth_dhchap_challenge_data { +__u8 auth_type; +__u8 auth_id; +__u16 rsvd1; +__le16 t_id; +__u8 hl; +__u8 rsvd2; +__u8 hashid; +__u8 dhgid; +__le16 dhvlen; +__le32 seqnum; +/* 'hl' bytes of challenge value */ +__u8 cval[]; +/* followed by 'dhvlen' bytes of DH value */ +}; + +struct nvmf_auth_dhchap_reply_data { +__u8 auth_type; +__u8 auth_id; +__le16 rsvd1; +__le16 t_id; +__u8 hl; +__u8 rsvd2; +__u8 cvalid; +__u8 rsvd3; +__le16 dhvlen; +__le32 seqnum; +/* 'hl' bytes of response data */ +__u8 rval[]; +/* followed by 'hl' bytes of Challenge value */ +/* followed by 'dhvlen' bytes of DH value */ +}; + +enum { +NVME_AUTH_DHCHAP_RESPONSE_VALID = (1 <<0 ), +}; + +struct nvmf_auth_dhchap_success1_data { +__u8 auth_type; +__u8 auth_id; +__le16 rsvd1; +__le16 t_id; +__u8 hl; +__u8 rsvd2; +__u8 rvalid; +__u8 rsvd3[7 ]; +/* 'hl' bytes of response value if 'rvalid' is set */ +__u8 rval[]; +}; + +struct nvmf_auth_dhchap_success2_data { +__u8 auth_type; +__u8 auth_id; +__le16 rsvd1; +__le16 t_id; +__u8 rsvd2[10 ]; +}; + +struct nvmf_auth_dhchap_failure_data { +__u8 auth_type; +__u8 auth_id; +__le16 rsvd1; +__le16 t_id; +__u8 rescode; +__u8 rescode_exp; +}; + +enum { +NVME_AUTH_DHCHAP_FAILURE_REASON_FAILED =0x01 , +}; + +enum { +NVME_AUTH_DHCHAP_FAILURE_FAILED =0x01 , +NVME_AUTH_DHCHAP_FAILURE_NOT_USABLE =0x02 , +NVME_AUTH_DHCHAP_FAILURE_CONCAT_MISMATCH =0x03 , +NVME_AUTH_DHCHAP_FAILURE_HASH_UNUSABLE =0x04 , +NVME_AUTH_DHCHAP_FAILURE_DHGROUP_UNUSABLE =0x05 , +NVME_AUTH_DHCHAP_FAILURE_INCORRECT_PAYLOAD =0x06 , +NVME_AUTH_DHCHAP_FAILURE_INCORRECT_MESSAGE =0x07 , +}; + +struct nvme_dbbuf {__u8 opcode;__u8 flags;
@@ -1547,6 +1751,9 @@ struct nvme_command {struct nvmf_connect_command connect;struct nvmf_property_set_command prop_set;struct nvmf_property_get_command prop_get; +struct nvmf_auth_common_command auth_common; +struct nvmf_auth_send_command auth_send; +struct nvmf_auth_receive_command auth_receive;struct nvme_dbbuf dbbuf;struct nvme_directive_cmd directive; };
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 5/11 nvme-fabrics: decode 'authentication required' connect error
In reply to Hannes ReineckeThe 'connect' command might fail with NVME_SC_AUTH_REQUIRED, so we should be decoding this error, too.
--- drivers/nvme/host/fabrics.c | 4 ++++ 1 file changed, 4 insertions(+)
drivers/nvme/host/fabrics.c +4 -0
@@ -331,6 +331,10 @@ static void nvmf_log_connect_error(struct nvme_ctrl *ctrl,dev_err (ctrl->device,"Connect command failed: host path error\n" );break ; +case NVME_SC_AUTH_REQUIRED : +dev_err (ctrl->device, +"Connect command failed: authentication required\n" ); +break ;default : dev_err(ctrl->device, "Connect command failed, error wo/DNR bit: %d\n",
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 6/11 nvme: Implement In-Band authentication
In reply to Hannes ReineckeImplement NVMe-oF In-Band authentication according to NVMe TPAR 8006. This patch adds two new fabric options 'dhchap_secret' to specify the pre-shared key (in ASCII respresentation according to NVMe 2.0 section 8.13.5.8 'Secret representation') and 'dhchap_ctrl_secret' to specify the pre-shared controller key for bi-directional authentication of both the host and the controller.
Re-authentication can be triggered by writing the PSK into the new controller sysfs attribute 'dhchap_secret' or 'dhchap_ctrl_secret'.
--- drivers/nvme/Kconfig | 1 + drivers/nvme/Makefile | 1 + drivers/nvme/common/Kconfig | 4 + drivers/nvme/common/Makefile | 7 + drivers/nvme/common/auth.c | 328 ++++++++++++++ drivers/nvme/host/Kconfig | 13 + drivers/nvme/host/Makefile | 1 + drivers/nvme/host/auth.c | 813 +++++++++++++++++++++++++++++++++++ drivers/nvme/host/core.c | 143 +++++- drivers/nvme/host/fabrics.c | 79 +++- drivers/nvme/host/fabrics.h | 7 + drivers/nvme/host/nvme.h | 30 ++ drivers/nvme/host/rdma.c | 1 + drivers/nvme/host/tcp.c | 1 + drivers/nvme/host/trace.c | 32 ++ include/linux/nvme-auth.h | 33 ++ 16 files changed, 1487 insertions(+), 7 deletions(-) create mode 100644 drivers/nvme/common/Kconfig create mode 100644 drivers/nvme/common/Makefile create mode 100644 drivers/nvme/common/auth.c create mode 100644 drivers/nvme/host/auth.c create mode 100644 include/linux/nvme-auth.h
drivers/nvme/Kconfig +1 -0
@@ -1,6 +1,7 @@ # SPDX-License-Identifier: GPL-2.0-only menu "NVME Support" +source "drivers/nvme/common/Kconfig" source "drivers/nvme/host/Kconfig" source "drivers/nvme/target/Kconfig"
drivers/nvme/Makefile +1 -0
@@ -1,4 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only +obj-$(CONFIG_NVME_COMMON) += common/ obj-y += host/ obj-y += target/
drivers/nvme/common/Kconfig new file +4 -0
@@ -0,0 +1,4 @@ +# SPDX-License-Identifier: GPL-2.0-only + +config NVME_COMMON + tristate
drivers/nvme/common/Makefile new file +7 -0
@@ -0,0 +1,7 @@ +# SPDX-License-Identifier: GPL-2.0 + +ccflags-y += -I$(src) + +obj-$(CONFIG_NVME_COMMON) += nvme-common.o + +nvme-common-y += auth.o
drivers/nvme/common/auth.c new file +328 -0
328 lines not on this page · raw message
drivers/nvme/host/Kconfig +13 -0
@@ -92,6 +92,19 @@ config NVME_TCP If unsure, say N. +config NVME_AUTH + bool "NVM Express over Fabrics In-Band Authentication" + depends on NVME_CORE + select NVME_COMMON + select CRYPTO + select CRYPTO_HMAC + select CRYPTO_SHA256 + select CRYPTO_SHA512 + help + This provides support for NVMe over Fabrics In-Band Authentication. + + If unsure, say N. + config NVME_APPLE tristate "Apple ANS2 NVM Express host driver" depends on OF && BLOCK
drivers/nvme/host/Makefile +1 -0
@@ -16,6 +16,7 @@ nvme-core-$(CONFIG_NVME_MULTIPATH) += multipath.o nvme-core-$(CONFIG_BLK_DEV_ZONED) += zns.o nvme-core-$(CONFIG_FAULT_INJECTION_DEBUG_FS) += fault_inject.o nvme-core-$(CONFIG_NVME_HWMON) += hwmon.o +nvme-core-$(CONFIG_NVME_AUTH) += auth.o nvme-y += pci.o
drivers/nvme/host/auth.c new file +813 -0
813 lines not on this page · raw message
drivers/nvme/host/core.c +139 -4
225 lines not on this page · raw message
drivers/nvme/host/fabrics.c +76 -3
134 lines not on this page · raw message
drivers/nvme/host/fabrics.h +7 -0
@@ -68,6 +68,8 @@ enum {NVMF_OPT_FAIL_FAST_TMO =1 <<20 ,NVMF_OPT_HOST_IFACE =1 <<21 ,NVMF_OPT_DISCOVERY =1 <<22 , +NVMF_OPT_DHCHAP_SECRET =1 <<23 , + NVMF_OPT_DHCHAP_CTRL_SECRET = 1 << 24, };/**
@@ -97,6 +99,9 @@ enum {* @max_reconnects: maximum number of allowed reconnect attempts before removing * the controller, (-1) means reconnect forever, zero means remove * immediately; +* @dhchap_secret: DH-HMAC-CHAP secret +* @dhchap_ctrl_secret: DH-HMAC-CHAP controller secret for bi-directional +* authentication * @disable_sqflow: disable controller sq flow control * @hdr_digest: generate/verify header digest (TCP) * @data_digest: generate/verify data digest (TCP)
@@ -121,6 +126,8 @@ struct nvmf_ctrl_options {unsigned int kato;struct nvmf_host *host;int max_reconnects; +char *dhchap_secret; +char *dhchap_ctrl_secret;bool disable_sqflow;bool hdr_digest;bool data_digest;
drivers/nvme/host/nvme.h +30 -0
@@ -328,6 +328,15 @@ struct nvme_ctrl {struct work_struct ana_work;#endif +#ifdef CONFIG_NVME_AUTH +struct work_struct dhchap_auth_work; +struct list_head dhchap_auth_list; +struct mutex dhchap_auth_mutex; +struct nvme_dhchap_key *host_key; +struct nvme_dhchap_key *ctrl_key; +u16 transaction; +#endif +/* Power saving configuration */ u64 ps_max_latency_us;bool apst_enabled;
@@ -991,6 +1000,27 @@ static inline bool nvme_ctrl_sgl_supported(struct nvme_ctrl *ctrl)return ctrl->sgls & ((1 <<0 ) | (1 <<1 )); } +#ifdef CONFIG_NVME_AUTH +void nvme_auth_init_ctrl (struct nvme_ctrl *ctrl); +void nvme_auth_stop (struct nvme_ctrl *ctrl); +int nvme_auth_negotiate (struct nvme_ctrl *ctrl,int qid); +int nvme_auth_wait (struct nvme_ctrl *ctrl,int qid); +void nvme_auth_reset (struct nvme_ctrl *ctrl); +void nvme_auth_free (struct nvme_ctrl *ctrl); +#else +static inline void nvme_auth_init_ctrl (struct nvme_ctrl *ctrl) {}; +static inline void nvme_auth_stop (struct nvme_ctrl *ctrl) {}; +static inline int nvme_auth_negotiate (struct nvme_ctrl *ctrl,int qid) +{ +return -EPROTONOSUPPORT ; +} +static inline int nvme_auth_wait (struct nvme_ctrl *ctrl,int qid) +{ +return NVME_SC_AUTH_REQUIRED ; +} +static inline void nvme_auth_free (struct nvme_ctrl *ctrl) {}; +#endif +u32 nvme_command_effects (struct nvme_ctrl *ctrl,struct nvme_ns *ns,u8 opcode);int nvme_execute_passthru_rq (struct request *rq);
drivers/nvme/host/rdma.c +1 -0
@@ -1197,6 +1197,7 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work)struct nvme_rdma_ctrl *ctrl =container_of (work, struct nvme_rdma_ctrl, err_work); +nvme_auth_stop (&ctrl->ctrl);nvme_stop_keep_alive (&ctrl->ctrl);flush_work (&ctrl->ctrl.async_event_work);nvme_rdma_teardown_io_queues (ctrl, false);
drivers/nvme/host/tcp.c +1 -0
@@ -2174,6 +2174,7 @@ static void nvme_tcp_error_recovery_work(struct work_struct *work) struct nvme_tcp_ctrl, err_work);struct nvme_ctrl *ctrl = &tcp_ctrl->ctrl; +nvme_auth_stop (ctrl);nvme_stop_keep_alive (ctrl);flush_work (&ctrl->async_event_work);nvme_tcp_teardown_io_queues (ctrl, false);
drivers/nvme/host/trace.c +32 -0
@@ -287,6 +287,34 @@ static const char *nvme_trace_fabrics_property_get(struct trace_seq *p, u8 *spc)return ret; } +static const char *nvme_trace_fabrics_auth_send (struct trace_seq *p,u8 *spc) +{ +const char *ret =trace_seq_buffer_ptr (p); +u8 spsp0 = spc[1 ]; +u8 spsp1 = spc[2 ]; +u8 secp = spc[3 ]; +u32 tl =get_unaligned_le32 (spc +4 ); + +trace_seq_printf (p,"spsp0=%02x, spsp1=%02x, secp=%02x, tl=%u" , + spsp0, spsp1, secp, tl); +trace_seq_putc (p,0 ); +return ret; +} + +static const char *nvme_trace_fabrics_auth_receive (struct trace_seq *p,u8 *spc) +{ +const char *ret =trace_seq_buffer_ptr (p); +u8 spsp0 = spc[1 ]; +u8 spsp1 = spc[2 ]; +u8 secp = spc[3 ]; +u32 al =get_unaligned_le32 (spc +4 ); + +trace_seq_printf (p,"spsp0=%02x, spsp1=%02x, secp=%02x, al=%u" , + spsp0, spsp1, secp, al); +trace_seq_putc (p,0 ); +return ret; +} +static const char *nvme_trace_fabrics_common (struct trace_seq *p,u8 *spc) {const char *ret =trace_seq_buffer_ptr (p);
@@ -306,6 +334,10 @@ const char *nvme_trace_parse_fabrics_cmd(struct trace_seq *p,return nvme_trace_fabrics_connect (p, spc);case nvme_fabrics_type_property_get:return nvme_trace_fabrics_property_get (p, spc); +case nvme_fabrics_type_auth_send: +return nvme_trace_fabrics_auth_send (p, spc); +case nvme_fabrics_type_auth_receive: +return nvme_trace_fabrics_auth_receive (p, spc);default :return nvme_trace_fabrics_common (p, spc); }
include/linux/nvme-auth.h new file +33 -0
@@ -0,0 +1,33 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* +* Copyright (c) 2021 Hannes Reinecke, SUSE Software Solutions +*/ + +#ifndef _NVME_AUTH_H +#define _NVME_AUTH_H + +#include <crypto/kpp.h> + +struct nvme_dhchap_key { +u8 *key; +size_t len; +u8 hash; +}; + +u32 nvme_auth_get_seqnum (void ); +const char *nvme_auth_dhgroup_name (u8 dhgroup_id); +const char *nvme_auth_dhgroup_kpp (u8 dhgroup_id); +u8 nvme_auth_dhgroup_id (const char *dhgroup_name); + +const char *nvme_auth_hmac_name (u8 hmac_id); +const char *nvme_auth_digest_name (u8 hmac_id); +size_t nvme_auth_hmac_hash_len (u8 hmac_id); +u8 nvme_auth_hmac_id (const char *hmac_name); + +struct nvme_dhchap_key *nvme_auth_extract_key (unsigned char *secret, +u8 key_hash); +void nvme_auth_free_key (struct nvme_dhchap_key *key); +u8 *nvme_auth_transform_key (struct nvme_dhchap_key *key,char *nqn); +int nvme_auth_generate_key (u8 *secret,struct nvme_dhchap_key **ret_key); + +#endif /* _NVME_AUTH_H */
Signature · 1 line
2.29.2
Sagi Grimberg
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
This one doesn't compile against nvme-5.20.
drivers/nvme/host/auth.c line 65
Quoted 3 lines
drivers/nvme/host/auth.c new file +3
-0+ ret =__nvme_submit_sync_cmd (q, &cmd,NULL , data, data_len,0 , + qid ==0 ?NVME_QID_ANY : qid, +0 , flags);
here ^^^^^^^
Hannes Reinecke
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 7 lines · Sagi Grimberg
This one doesn't compile against nvme-5.20.
Quoted 3 lines
+ ret = __nvme_submit_sync_cmd(q, &cmd, NULL, data,
data_len, 0, + qid == 0 ?NVME_QID_ANY: qid,
+ 0, flags);here ^^^^^^^
Sheesh. I've compiled it against nvme-5.19, where it works perfectly. Alright, once more unto the breach ...
Cheers,
Hannes
Sagi Grimberg
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 11 lines · Hannes Reinecke
Quoted 7 lines · Sagi Grimberg
This one doesn't compile against nvme-5.20.
Quoted 3 lines
+ ret = __nvme_submit_sync_cmd(q, &cmd, NULL, data,
data_len, 0, + qid == 0 ?NVME_QID_ANY: qid,
+ 0, flags);here ^^^^^^^
Sheesh. I've compiled it against nvme-5.19, where it works perfectly. Alright, once more unto the breach ...
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
--
[ 84.793307] Workqueue: nvme-wq __nvme_auth_work [nvme_core] [ 84.794790] RIP: 0010:nvme_auth_transform_key+0x19/0x1f0 [nvme_common] [ 84.796468] Code: bc f4 ff ff ff eb bf e8 f5 2f e1 ee 0f 1f 44 00 00 0f 1f 44 00 00 41 57 41 56 41 55 49 89 f5 41 54 55 53 48 89 fb 48 83 ec 08 <0f> b6 7f 10 e8 ce f9 ff ff 48 89 c7 0f b6 43 10 84 c0 0f 84 4c 01 [ 84.800112] RSP: 0018:ffffae5f8047bc78 EFLAGS: 00010296 [ 84.801048] RAX: ffff973b854ca0c0 RBX: 0000000000000000 RCX: 0000000000000000 [ 84.802289] RDX: ffff973b8355b000 RSI: ffff973b84f568a0 RDI: 0000000000000000 [ 84.803447] RBP: ffff973b842dd6b9 R08: 0000000000000003 R09: ffff973bbec308a8 [ 84.804638] R10: 0000000000000147 R11: 0000000000000000 R12: ffff973b842dd600 [ 84.805767] R13: ffff973b84f568a0 R14: 0000000000000000 R15: ffff973b9e94250d [ 84.806929] FS: 0000000000000000(0000) GS:ffff973bbec00000(0000) knlGS:0000000000000000 [ 84.808220] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 84.809178] CR2: 0000000000000010 CR3: 00000000061ac003 CR4: 0000000000370ef0 [ 84.810337] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 84.811432] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 84.812527] Call Trace: [ 84.812974] <TASK> [ 84.813372] nvme_auth_dhchap_setup_ctrl_response+0x4b/0x3b0 [nvme_core] [ 84.814355] ? preempt_count_add+0x68/0xa0 [ 84.815080] ? _raw_spin_unlock_irq+0x16/0x28 [ 84.815840] ? __wait_for_common+0x19f/0x1d0 [ 84.816587] ? firmware_map_remove+0x87/0x87 [ 84.817333] ? blk_mq_hctx_has_pending+0x38/0x70 [ 84.818123] ? blk_mq_run_hw_queue+0x7d/0xe0 [ 84.818784] ? __blk_mq_free_request+0x9b/0xa0 [ 84.819482] ? blk_queue_exit+0xe/0x40 [ 84.820124] ? __nvme_submit_sync_cmd+0xe8/0x160 [nvme_core] [ 84.821096] ? nvme_auth_submit+0x8f/0xd0 [nvme_core] [ 84.821970] __nvme_auth_work+0x1fb/0x480 [nvme_core] [ 84.822869] process_one_work+0x1e5/0x3b0 [ 84.823608] worker_thread+0x1c4/0x3a0 [ 84.824343] ? rescuer_thread+0x390/0x390 [ 84.825044] kthread+0xe8/0x110
--
Sagi Grimberg
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"
The dhchap_ctrl_key is the offending string...
Quoted 34 lines
Signature · 33 lines
[ 84.793307] Workqueue: nvme-wq __nvme_auth_work [nvme_core] [ 84.794790] RIP: 0010:nvme_auth_transform_key+0x19/0x1f0 [nvme_common] [ 84.796468] Code: bc f4 ff ff ff eb bf e8 f5 2f e1 ee 0f 1f 44 00 00 0f 1f 44 00 00 41 57 41 56 41 55 49 89 f5 41 54 55 53 48 89 fb 48 83 ec 08 <0f> b6 7f 10 e8 ce f9 ff ff 48 89 c7 0f b6 43 10 84 c0 0f 84 4c 01 [ 84.800112] RSP: 0018:ffffae5f8047bc78 EFLAGS: 00010296 [ 84.801048] RAX: ffff973b854ca0c0 RBX: 0000000000000000 RCX: 0000000000000000 [ 84.802289] RDX: ffff973b8355b000 RSI: ffff973b84f568a0 RDI: 0000000000000000 [ 84.803447] RBP: ffff973b842dd6b9 R08: 0000000000000003 R09: ffff973bbec308a8 [ 84.804638] R10: 0000000000000147 R11: 0000000000000000 R12: ffff973b842dd600 [ 84.805767] R13: ffff973b84f568a0 R14: 0000000000000000 R15: ffff973b9e94250d [ 84.806929] FS: 0000000000000000(0000) GS:ffff973bbec00000(0000) knlGS:0000000000000000 [ 84.808220] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 84.809178] CR2: 0000000000000010 CR3: 00000000061ac003 CR4: 0000000000370ef0 [ 84.810337] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 84.811432] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 84.812527] Call Trace: [ 84.812974] <TASK> [ 84.813372] nvme_auth_dhchap_setup_ctrl_response+0x4b/0x3b0 [nvme_core] [ 84.814355] ? preempt_count_add+0x68/0xa0 [ 84.815080] ? _raw_spin_unlock_irq+0x16/0x28 [ 84.815840] ? __wait_for_common+0x19f/0x1d0 [ 84.816587] ? firmware_map_remove+0x87/0x87 [ 84.817333] ? blk_mq_hctx_has_pending+0x38/0x70 [ 84.818123] ? blk_mq_run_hw_queue+0x7d/0xe0 [ 84.818784] ? __blk_mq_free_request+0x9b/0xa0 [ 84.819482] ? blk_queue_exit+0xe/0x40 [ 84.820124] ? __nvme_submit_sync_cmd+0xe8/0x160 [nvme_core] [ 84.821096] ? nvme_auth_submit+0x8f/0xd0 [nvme_core] [ 84.821970] __nvme_auth_work+0x1fb/0x480 [nvme_core] [ 84.822869] process_one_work+0x1e5/0x3b0 [ 84.823608] worker_thread+0x1c4/0x3a0 [ 84.824343] ? rescuer_thread+0x390/0x390 [ 84.825044] kthread+0xe8/0x110 --
Hannes Reinecke
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 9 lines · Sagi Grimberg
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"The
dhchap_ctrl_keyis the offending string...
Right. Should be fixed with the attached patch.
Can you check?
Cheers,
Hannes
Signature · 5 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), Geschäftsführer: Ivo Totev, Andrew Myers, Andrew McDonald, Martje Boudien Moerman
From 1f4d34016e7cad41f6947143f07f802b05415e26 Mon Sep 17 00:00:00 2001
From: Hannes Reinecke <[email protected]>
Date: Wed, 22 Jun 2022 07:57:06 +0200
Subject: [PATCH] nvme-auth: do not use ctrl->opts->dhchap_ctrl_secret
The user might have passed in an invalid controller secret, causing ctrl->opts->dhchap_ctrl_secret to be present, but ctrl->ctrl_key to be empty. So always use ctrl->ctrl_key when checking if a valid controller secret is present.
--- drivers/nvme/common/auth.c | 13 +++++++------ drivers/nvme/host/auth.c | 8 ++++---- 2 files changed, 11 insertions(+), 10 deletions(-)
drivers/nvme/common/auth.c +7 -6
@@ -237,21 +237,21 @@ EXPORT_SYMBOL_GPL(nvme_auth_free_key);u8 *nvme_auth_transform_key (struct nvme_dhchap_key *key,char *nqn) { -const char *hmac_name =nvme_auth_hmac_name (key->hash); +const char *hmac_name;struct crypto_shash *key_tfm;struct shash_desc *shash;u8 *transformed_key;int ret; -if (key->hash ==0 ) { - transformed_key =kmemdup (key->key, key->len,GFP_KERNEL ); -return transformed_key ? transformed_key :ERR_PTR (-ENOMEM ); - } -if (!key || !key->key) {pr_warn ("No key specified\n" );return ERR_PTR (-ENOKEY ); } +if (key->hash ==0 ) { + transformed_key =kmemdup (key->key, key->len,GFP_KERNEL ); +return transformed_key ? transformed_key :ERR_PTR (-ENOMEM ); + } + hmac_name =nvme_auth_hmac_name (key->hash);if (!hmac_name) {pr_warn ("Invalid key hash id %d\n" , key->hash);return ERR_PTR (-EINVAL );
@@ -470,6 +470,7 @@ int nvme_auth_generate_key(u8 *secret, struct nvme_dhchap_key **ret_key)/* Pass in the secret without the 'DHHC-1:XX:' prefix */ key =nvme_auth_extract_key (secret +10 , key_hash);if (IS_ERR (key)) { + *ret_key =NULL ;return PTR_ERR (key); }
drivers/nvme/host/auth.c +4 -4
@@ -314,7 +314,7 @@ static int nvme_auth_set_dhchap_reply_data(struct nvme_ctrl *ctrl, data->hl = chap->hash_len; data->dhvlen =cpu_to_le16 (chap->host_key_len);memcpy (data->rval, chap->response, chap->hash_len); -if (ctrl->opts->dhchap_ctrl_secret) { +if (ctrl->ctrl_key) {get_random_bytes (chap->c2, chap->hash_len); data->cvalid =1 ; chap->s2 =nvme_auth_get_seqnum ();
@@ -344,7 +344,7 @@ static int nvme_auth_process_dhchap_success1(struct nvme_ctrl *ctrl,struct nvmf_auth_dhchap_success1_data *data = chap->buf;size_t size =sizeof (*data); -if (ctrl->opts->dhchap_ctrl_secret) +if (ctrl->ctrl_key) size += chap->hash_len;if (chap->buf_size < size) {
@@ -791,7 +791,7 @@ static void __nvme_auth_work(struct work_struct *work)return ; } -if (ctrl->opts->dhchap_ctrl_secret) { +if (ctrl->ctrl_key) {dev_dbg (ctrl->device,"%s: qid %d controller response\n" , __func__, chap->qid);
@@ -809,7 +809,7 @@ static void __nvme_auth_work(struct work_struct *work) goto fail2; } -if (ctrl->opts->dhchap_ctrl_secret) { +if (ctrl->ctrl_key) {/* DH-HMAC-CHAP Step 5: send success2 */ dev_dbg (ctrl->device,"%s: qid %d send success2\n" , __func__, chap->qid);
Signature · 1 line
2.26.2
Sagi Grimberg
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 11 lines
Quoted 9 lines
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"The
dhchap_ctrl_keyis the offending string...Right. Should be fixed with the attached patch.
Can you check?
Yes, that works.
Hannes Reinecke
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 7 lines · Sagi Grimberg
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
The
dhchap_ctrl_keyis the offending string...Right. Should be fixed with the attached patch.
Can you check?Yes, that works.
So, to summarize: With this one and the tentative patch I've sent earlier:
drivers/nvme/host/auth.c +4 -0
@@ -842,6 +842,10 @@ int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid)dev_warn (ctrl->device,"qid %d: no key\n" , qid);return -ENOKEY ; } +if (ctrl->opts->dhchap_ctrl_secret && !ctrl->ctrl_key) { +dev_warn (ctrl->device,"qid %d: invalid ctrl key\n" , qid); +return -ENOKEY ; + }mutex_lock (&ctrl->dhchap_auth_mutex);/* Check if the context is already queued */
your issues are resolved?
Just to clarify before I sent another round of patches ...
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
Sagi Grimberg
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 35 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
The
dhchap_ctrl_keyis the offending string...Right. Should be fixed with the attached patch.
Can you check?Yes, that works.
So, to summarize: With this one and the tentative patch I've sent earlier:
drivers/nvme/host/auth.c +4
-0@@ -842,6 +842,10 @@ int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid) dev_warn(ctrl->device, "qid %d: no key\n", qid); return -ENOKEY; } + if (ctrl->opts->dhchap_ctrl_secret && !ctrl->ctrl_key) { + dev_warn(ctrl->device, "qid %d: invalid ctrl key\n", qid); + return -ENOKEY; + } mutex_lock(&ctrl->dhchap_auth_mutex);/* Check if the context is already queued */
Quoted 2 lines
your issues are resolved?
Just to clarify before I sent another round of patches ...
No. I don't understand why this patch is needed. As I noted earlier, when I pass a wrong ctrl key I fail to connect do to authentication error. Also when I pass a malformed ctrl key.
So what does this patch fix? Can you give me an example that should have failed before and only with this patch it fails?
And for the other issue, please read it again, I think it is related to nvme-cli and not this patchset.
Hannes Reinecke
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 43 lines · Sagi Grimberg
Quoted 36 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
The
dhchap_ctrl_keyis the offending string...Right. Should be fixed with the attached patch.
Can you check?Yes, that works.
So, to summarize: With this one and the tentative patch I've sent earlier:
drivers/nvme/host/auth.c +4
-0@@ -842,6 +842,10 @@ int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid) dev_warn(ctrl->device, "qid %d: no key\n", qid); return -ENOKEY; } + if (ctrl->opts->dhchap_ctrl_secret && !ctrl->ctrl_key) { + dev_warn(ctrl->device, "qid %d: invalid ctrl key\n", qid); + return -ENOKEY; + }
mutex_lock(&ctrl->dhchap_auth_mutex); /* Check if the context is already queued */Quoted 2 lines
your issues are resolved?
Just to clarify before I sent another round of patches ...No. I don't understand why this patch is needed. As I noted earlier, when I pass a wrong ctrl key I fail to connect do to authentication error. Also when I pass a malformed ctrl key.
Quoted 3 lines
So what does this patch fix? Can you give me an example that should have failed before and only with this patch it fails?
Reasoning is as follows:
When an invalid controller key is passed via 'ctrl->opts->dhchap_ctrl_secret' we'll end up with an empty 'ctrl->ctrl_key' in nvme_auth_init_ctrl() (which is what caused the original oops). But the negotiation will still start, and, seeing that 'ctrl->ctrl_key' is empty, will _not_ attempt controller authentication.
But as a controller key is passed from userland we should have attempted it, and hence an error is in order.
Quoted 2 lines
And for the other issue, please read it again, I think it is related to nvme-cli and not this patchset.
Hmm. Okay, I'll check, and possibly have to create another blktest for this ...
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
Sagi Grimberg
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 44 lines · Hannes Reinecke
Quoted 43 lines · Sagi Grimberg
Quoted 36 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
The
dhchap_ctrl_keyis the offending string...Right. Should be fixed with the attached patch.
Can you check?Yes, that works.
So, to summarize: With this one and the tentative patch I've sent earlier:
drivers/nvme/host/auth.c +4
-0@@ -842,6 +842,10 @@ int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid) dev_warn(ctrl->device, "qid %d: no key\n", qid); return -ENOKEY; } + if (ctrl->opts->dhchap_ctrl_secret && !ctrl->ctrl_key) { + dev_warn(ctrl->device, "qid %d: invalid ctrl key\n", qid); + return -ENOKEY; + }
mutex_lock(&ctrl->dhchap_auth_mutex); /* Check if the context is already queued */Quoted 2 lines
your issues are resolved?
Just to clarify before I sent another round of patches ...No. I don't understand why this patch is needed. As I noted earlier, when I pass a wrong ctrl key I fail to connect do to authentication error. Also when I pass a malformed ctrl key.
Quoted 7 lines
Quoted 3 lines
So what does this patch fix? Can you give me an example that should have failed before and only with this patch it fails?
Reasoning is as follows:
When an invalid controller key is passed via 'ctrl->opts->dhchap_ctrl_secret' we'll end up with an empty 'ctrl->ctrl_key' innvme_auth_init_ctrl()(which is what caused the original oops). But the negotiation will still start, and, seeing that 'ctrl->ctrl_key' is empty, will _not_ attempt controller authentication.
But as a controller key is passed from userland we should have attempted it, and hence an error is in order.
Yes, I see it now.
And the patch seems to work.
Quoted 4 lines
Quoted 2 lines
And for the other issue, please read it again, I think it is related to nvme-cli and not this patchset.
Hmm. Okay, I'll check, and possibly have to create another blktest for this ...
Looks like a libnvme fix makes the behavior go away:
--
src/nvme/fabrics.c +4 -1
@@ -596,8 +596,11 @@ int nvmf_add_ctrl(nvme_host_t h, nvme_ctrl_t c,nvme_ctrl_get_host_iface (c),nvme_ctrl_get_trsvcid (c), NULL); -if (fc) +if (fc) { cfg =merge_config (c,nvme_ctrl_get_config (fc)); +if (fc->dhchap_key) +nvme_ctrl_set_dhchap_key (c, fc->dhchap_key); + } }nvme_ctrl_set_discovered (c, true);
--
Looks like when we scan to see an existing controller in the config, we find it, but merge_config does not touch the ctrl dhchap_key.
Maybe it would be better to just add the dhchap_key to the ctrl cfg and have merge_config merge it as well? Or we can live with dhchap_key merged explicitly?
Hannes Reinecke
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 82 lines · Sagi Grimberg
Quoted 44 lines · Hannes Reinecke
Quoted 43 lines · Sagi Grimberg
Quoted 36 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
The
dhchap_ctrl_keyis the offending string...Right. Should be fixed with the attached patch.
Can you check?Yes, that works.
So, to summarize: With this one and the tentative patch I've sent earlier:
drivers/nvme/host/auth.c +4
-0@@ -842,6 +842,10 @@ int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid) dev_warn(ctrl->device, "qid %d: no key\n", qid); return -ENOKEY; } + if (ctrl->opts->dhchap_ctrl_secret && !ctrl->ctrl_key) { + dev_warn(ctrl->device, "qid %d: invalid ctrl key\n", qid); + return -ENOKEY; + }
mutex_lock(&ctrl->dhchap_auth_mutex); /* Check if the context is already queued */Quoted 2 lines
your issues are resolved?
Just to clarify before I sent another round of patches ...No. I don't understand why this patch is needed. As I noted earlier, when I pass a wrong ctrl key I fail to connect do to authentication error. Also when I pass a malformed ctrl key.
Quoted 7 lines
Quoted 3 lines
So what does this patch fix? Can you give me an example that should have failed before and only with this patch it fails?
Reasoning is as follows:
When an invalid controller key is passed via 'ctrl->opts->dhchap_ctrl_secret' we'll end up with an empty 'ctrl->ctrl_key' innvme_auth_init_ctrl()(which is what caused the original oops). But the negotiation will still start, and, seeing that 'ctrl->ctrl_key' is empty, will _not_ attempt controller authentication.
But as a controller key is passed from userland we should have attempted it, and hence an error is in order.Yes, I see it now.
And the patch seems to work.Quoted 4 lines
Quoted 2 lines
And for the other issue, please read it again, I think it is related to nvme-cli and not this patchset.
Hmm. Okay, I'll check, and possibly have to create another blktest for this ...
Looks like a libnvme fix makes the behavior go away:
src/nvme/fabrics.c +4
-1@@ -596,8 +596,11 @@ int nvmf_add_ctrl(nvme_host_t h, nvme_ctrl_t c, nvme_ctrl_get_host_iface(c), nvme_ctrl_get_trsvcid(c), NULL); - if (fc) + if (fc) { cfg = merge_config(c, nvme_ctrl_get_config(fc)); + if (fc->dhchap_key) + nvme_ctrl_set_dhchap_key(c, fc->dhchap_key); + } }
Quoted 9 lines
nvme_ctrl_set_discovered(c, true);Signature · 6 lines
Looks like when we scan to see an existing controller in the config, we find it, but merge_config does not touch the ctrl dhchap_key. Maybe it would be better to just add the dhchap_key to the ctrl cfg and have merge_config merge it as well? Or we can live with dhchap_key merged explicitly?
Hmm. That runs slightly afoul of the merge logic.
'merge_config' will overwrite any value _not_ being the default.
Consequently, if a non-default value is already set it won't be overwritten.
Which means that if the 'cfg' structure already contained a controller secret (which could've been used, say, for discovery) we won't overwrite it in merge_config(). Hence I fear we have to go with your suggestion.
Will be preparing a pull request.
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
Sagi Grimberg
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
Quoted 89 lines · Hannes Reinecke
Quoted 88 lines · Sagi Grimberg
Quoted 44 lines · Hannes Reinecke
Quoted 43 lines · Sagi Grimberg
Quoted 36 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
This was what I used in this:
$ nvme connect -a 192.168.123.1 -t tcp -s 8009 -n testnqn1 -S "DHHC-1:00:QpxVGpctx5J+4SeW2MClUI8rfZO3WdP1llImvsPsx7e3TK+I:" -C "DHHC-1:00:Jc/My1o0qtLCWRp+sHhAVafdfaS7YQOMYhk9zSmlatobqB8C:"Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
The
dhchap_ctrl_keyis the offending string...Right. Should be fixed with the attached patch.
Can you check?Yes, that works.
So, to summarize: With this one and the tentative patch I've sent earlier:
drivers/nvme/host/auth.c +4
-0@@ -842,6 +842,10 @@ int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid) dev_warn(ctrl->device, "qid %d: no key\n", qid); return -ENOKEY; } + if (ctrl->opts->dhchap_ctrl_secret && !ctrl->ctrl_key) { + dev_warn(ctrl->device, "qid %d: invalid ctrl key\n", qid); + return -ENOKEY; + }
mutex_lock(&ctrl->dhchap_auth_mutex); /* Check if the context is already queued */Quoted 2 lines
your issues are resolved?
Just to clarify before I sent another round of patches ...No. I don't understand why this patch is needed. As I noted earlier, when I pass a wrong ctrl key I fail to connect do to authentication error. Also when I pass a malformed ctrl key.
Quoted 7 lines
Quoted 3 lines
So what does this patch fix? Can you give me an example that should have failed before and only with this patch it fails?
Reasoning is as follows:
When an invalid controller key is passed via 'ctrl->opts->dhchap_ctrl_secret' we'll end up with an empty 'ctrl->ctrl_key' innvme_auth_init_ctrl()(which is what caused the original oops). But the negotiation will still start, and, seeing that 'ctrl->ctrl_key' is empty, will _not_ attempt controller authentication.
But as a controller key is passed from userland we should have attempted it, and hence an error is in order.Yes, I see it now.
And the patch seems to work.Quoted 4 lines
Quoted 2 lines
And for the other issue, please read it again, I think it is related to nvme-cli and not this patchset.
Hmm. Okay, I'll check, and possibly have to create another blktest for this ...
Looks like a libnvme fix makes the behavior go away:
src/nvme/fabrics.c +4
-1@@ -596,8 +596,11 @@ int nvmf_add_ctrl(nvme_host_t h, nvme_ctrl_t c, nvme_ctrl_get_host_iface(c), nvme_ctrl_get_trsvcid(c), NULL); - if (fc) + if (fc) { cfg = merge_config(c, nvme_ctrl_get_config(fc)); + if (fc->dhchap_key) + nvme_ctrl_set_dhchap_key(c, fc->dhchap_key); + } }
nvme_ctrl_set_discovered(c, true);Signature · 2 lines
Looks like when we scan to see an existing controller in the config, we find it, but merge_config does not touch the ctrl dhchap_key.
Quoted 6 lines
Quoted 3 lines
Maybe it would be better to just add the
dhchap_keyto the ctrl cfg and havemerge_configmerge it as well? Or we can live withdhchap_keymerged explicitly?Hmm. That runs slightly afoul of the merge logic. '
merge_config' will overwrite any value _not_ being the default.
There is no ctrl dhchap_key default though...
Quoted 1 line
Consequently, if a non-default value is already set it won't be overwritten.
Makes sense.
Quoted 1 line
Which means that if the 'cfg' structure already contained a controller secret (which could've been used, say, for discovery) we won't overwrite it in
merge_config().
We probably shouldn't override it this way either if it was explicitly set by the user no?
Quoted 1 line
Hence I fear we have to go with your suggestion.
Whatever you choose. Don't really care...
Quoted 3 lines
Will be preparing a pull request.
Thanks
Hannes Reinecke
maintainerRe: [PATCH 06/11] nvme: Implement In-Band authentication
In reply to Sagi GrimbergQuoted 10 lines · Sagi Grimberg
Quoted 7 lines · Hannes Reinecke
Quoted 6 lines · Sagi Grimberg
This one doesn't compile against nvme-5.20.
Quoted 3 lines
+ ret = __nvme_submit_sync_cmd(q, &cmd, NULL, data,
data_len, 0, + qid == 0 ?NVME_QID_ANY: qid,
+ 0, flags);
Quoted 41 lines
Quoted 4 lines
Quoted 1 line
here ^^^^^^^
Sheesh. I've compiled it against nvme-5.19, where it works perfectly. Alright, once more unto the breach ...
Looks like if I pass a malformed ctrl key to nvme connect I am able to crash the system:
Signature · 33 lines
[ 84.793307] Workqueue: nvme-wq __nvme_auth_work [nvme_core] [ 84.794790] RIP: 0010:nvme_auth_transform_key+0x19/0x1f0 [nvme_common] [ 84.796468] Code: bc f4 ff ff ff eb bf e8 f5 2f e1 ee 0f 1f 44 00 00 0f 1f 44 00 00 41 57 41 56 41 55 49 89 f5 41 54 55 53 48 89 fb 48 83 ec 08 <0f> b6 7f 10 e8 ce f9 ff ff 48 89 c7 0f b6 43 10 84 c0 0f 84 4c 01 [ 84.800112] RSP: 0018:ffffae5f8047bc78 EFLAGS: 00010296 [ 84.801048] RAX: ffff973b854ca0c0 RBX: 0000000000000000 RCX: 0000000000000000 [ 84.802289] RDX: ffff973b8355b000 RSI: ffff973b84f568a0 RDI: 0000000000000000 [ 84.803447] RBP: ffff973b842dd6b9 R08: 0000000000000003 R09: ffff973bbec308a8 [ 84.804638] R10: 0000000000000147 R11: 0000000000000000 R12: ffff973b842dd600 [ 84.805767] R13: ffff973b84f568a0 R14: 0000000000000000 R15: ffff973b9e94250d [ 84.806929] FS: 0000000000000000(0000) GS:ffff973bbec00000(0000) knlGS:0000000000000000 [ 84.808220] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 84.809178] CR2: 0000000000000010 CR3: 00000000061ac003 CR4: 0000000000370ef0 [ 84.810337] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 84.811432] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 84.812527] Call Trace: [ 84.812974] <TASK> [ 84.813372] nvme_auth_dhchap_setup_ctrl_response+0x4b/0x3b0 [nvme_core] [ 84.814355] ? preempt_count_add+0x68/0xa0 [ 84.815080] ? _raw_spin_unlock_irq+0x16/0x28 [ 84.815840] ? __wait_for_common+0x19f/0x1d0 [ 84.816587] ? firmware_map_remove+0x87/0x87 [ 84.817333] ? blk_mq_hctx_has_pending+0x38/0x70 [ 84.818123] ? blk_mq_run_hw_queue+0x7d/0xe0 [ 84.818784] ? __blk_mq_free_request+0x9b/0xa0 [ 84.819482] ? blk_queue_exit+0xe/0x40 [ 84.820124] ? __nvme_submit_sync_cmd+0xe8/0x160 [nvme_core] [ 84.821096] ? nvme_auth_submit+0x8f/0xd0 [nvme_core] [ 84.821970] __nvme_auth_work+0x1fb/0x480 [nvme_core] [ 84.822869] process_one_work+0x1e5/0x3b0 [ 84.823608] worker_thread+0x1c4/0x3a0 [ 84.824343] ? rescuer_thread+0x390/0x390 [ 84.825044] kthread+0xe8/0x110 --
Malformed exactly how?
We should've captured any malformed key during nvme_auth_extract_key(). Can you share an example of your malformed key?
Cheers,
Hannes
Signature · 5 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg HRB 36809 (AG Nürnberg), Geschäftsführer: Ivo Totev, Andrew Myers, Andrew McDonald, Martje Boudien Moerman
Hannes Reinecke
maintainerPATCH v15 7/11 nvme-auth: Diffie-Hellman key exchange support
In reply to Hannes ReineckeImplement Diffie-Hellman key exchange using FFDHE groups for NVMe In-Band Authentication.
--- drivers/nvme/common/auth.c | 153 ++++++++++++++++++++++++++++ drivers/nvme/host/Kconfig | 2 + drivers/nvme/host/auth.c | 199 +++++++++++++++++++++++++++++++++++-- include/linux/nvme-auth.h | 8 ++ 4 files changed, 356 insertions(+), 6 deletions(-)
drivers/nvme/common/auth.c +153 -0
@@ -301,6 +301,159 @@ u8 *nvme_auth_transform_key(struct nvme_dhchap_key *key, char *nqn) }EXPORT_SYMBOL_GPL (nvme_auth_transform_key); +static int nvme_auth_hash_skey (int hmac_id,u8 *skey,size_t skey_len,u8 *hkey) +{ +const char *digest_name; +struct crypto_shash *tfm; +int ret; + + digest_name =nvme_auth_digest_name (hmac_id); +if (!digest_name) { +pr_debug ("%s: failed to get digest for %d\n" , __func__, + hmac_id); +return -EINVAL ; + } + tfm =crypto_alloc_shash (digest_name,0 ,0 ); +if (IS_ERR (tfm)) +return -ENOMEM ; + + ret =crypto_shash_tfm_digest (tfm, skey, skey_len, hkey); +if (ret <0 ) +pr_debug ("%s: Failed to hash digest len %zu\n" , __func__, + skey_len); + +crypto_free_shash (tfm); +return ret; +} + +int nvme_auth_augmented_challenge (u8 hmac_id,u8 *skey,size_t skey_len, +u8 *challenge,u8 *aug,size_t hlen) +{ +struct crypto_shash *tfm; +struct shash_desc *desc; +u8 *hashed_key; +const char *hmac_name; +int ret; + + hashed_key =kmalloc (hlen,GFP_KERNEL ); +if (!hashed_key) +return -ENOMEM ; + + ret =nvme_auth_hash_skey (hmac_id, skey, + skey_len, hashed_key); +if (ret <0 ) + goto out_free_key; + + hmac_name =nvme_auth_hmac_name (hmac_id); +if (!hmac_name) { +pr_warn ("%s: invalid hash algoritm %d\n" , + __func__, hmac_id); + ret = -EINVAL ; + goto out_free_key; + } + + tfm =crypto_alloc_shash (hmac_name,0 ,0 ); +if (IS_ERR (tfm)) { + ret =PTR_ERR (tfm); + goto out_free_key; + } + + desc =kmalloc (sizeof (struct shash_desc ) +crypto_shash_descsize (tfm), +GFP_KERNEL ); +if (!desc) { + ret = -ENOMEM ; + goto out_free_hash; + } + desc->tfm = tfm; + + ret =crypto_shash_setkey (tfm, hashed_key, hlen); +if (ret) + goto out_free_desc; + + ret =crypto_shash_init (desc); +if (ret) + goto out_free_desc; + + ret =crypto_shash_update (desc, challenge, hlen); +if (ret) + goto out_free_desc; + + ret =crypto_shash_final (desc, aug); +out_free_desc: +kfree_sensitive (desc); +out_free_hash: +crypto_free_shash (tfm); +out_free_key: +kfree_sensitive (hashed_key); +return ret; +} +EXPORT_SYMBOL_GPL (nvme_auth_augmented_challenge); + +int nvme_auth_gen_privkey (struct crypto_kpp *dh_tfm,u8 dh_gid) +{ +int ret; + + ret =crypto_kpp_set_secret (dh_tfm,NULL ,0 ); +if (ret) +pr_debug ("failed to set private key, error %d\n" , ret); + +return ret; +} +EXPORT_SYMBOL_GPL (nvme_auth_gen_privkey); + +int nvme_auth_gen_pubkey (struct crypto_kpp *dh_tfm, +u8 *host_key,size_t host_key_len) +{ +struct kpp_request *req; +struct crypto_wait wait; +struct scatterlist dst; +int ret; + + req =kpp_request_alloc (dh_tfm,GFP_KERNEL ); +if (!req) +return -ENOMEM ; + +crypto_init_wait (&wait); +kpp_request_set_input (req,NULL ,0 ); +sg_init_one (&dst, host_key, host_key_len); +kpp_request_set_output (req, &dst, host_key_len); +kpp_request_set_callback (req,CRYPTO_TFM_REQ_MAY_BACKLOG , + crypto_req_done, &wait); + + ret =crypto_wait_req (crypto_kpp_generate_public_key (req), &wait); +kpp_request_free (req); +return ret; +} +EXPORT_SYMBOL_GPL (nvme_auth_gen_pubkey); + +int nvme_auth_gen_shared_secret (struct crypto_kpp *dh_tfm, +u8 *ctrl_key,size_t ctrl_key_len, +u8 *sess_key,size_t sess_key_len) +{ +struct kpp_request *req; +struct crypto_wait wait; +struct scatterlist src, dst; +int ret; + + req =kpp_request_alloc (dh_tfm,GFP_KERNEL ); +if (!req) +return -ENOMEM ; + +crypto_init_wait (&wait); +sg_init_one (&src, ctrl_key, ctrl_key_len); +kpp_request_set_input (req, &src, ctrl_key_len); +sg_init_one (&dst, sess_key, sess_key_len); +kpp_request_set_output (req, &dst, sess_key_len); +kpp_request_set_callback (req,CRYPTO_TFM_REQ_MAY_BACKLOG , + crypto_req_done, &wait); + + ret =crypto_wait_req (crypto_kpp_compute_shared_secret (req), &wait); + +kpp_request_free (req); +return ret; +} +EXPORT_SYMBOL_GPL (nvme_auth_gen_shared_secret); +int nvme_auth_generate_key (u8 *secret,struct nvme_dhchap_key **ret_key) {struct nvme_dhchap_key *key;
drivers/nvme/host/Kconfig +2 -0
@@ -100,6 +100,8 @@ config NVME_AUTH select CRYPTO_HMAC select CRYPTO_SHA256 select CRYPTO_SHA512 + select CRYPTO_DH + select CRYPTO_DH_RFC7919_GROUPS help This provides support for NVMe over Fabrics In-Band Authentication.
drivers/nvme/host/auth.c +193 -6
285 lines not on this page · raw message
include/linux/nvme-auth.h +8 -0
@@ -29,5 +29,13 @@ struct nvme_dhchap_key *nvme_auth_extract_key(unsigned char *secret,void nvme_auth_free_key (struct nvme_dhchap_key *key);u8 *nvme_auth_transform_key (struct nvme_dhchap_key *key,char *nqn);int nvme_auth_generate_key (u8 *secret,struct nvme_dhchap_key **ret_key); +int nvme_auth_augmented_challenge (u8 hmac_id,u8 *skey,size_t skey_len, +u8 *challenge,u8 *aug,size_t hlen); +int nvme_auth_gen_privkey (struct crypto_kpp *dh_tfm,u8 dh_gid); +int nvme_auth_gen_pubkey (struct crypto_kpp *dh_tfm, +u8 *host_key,size_t host_key_len); +int nvme_auth_gen_shared_secret (struct crypto_kpp *dh_tfm, +u8 *ctrl_key,size_t ctrl_key_len, +u8 *sess_key,size_t sess_key_len);#endif /* _NVME_AUTH_H */
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 8/11 nvmet: parse fabrics commands on io queues
In reply to Hannes ReineckeSome fabrics commands can be sent via io queues, so add a new function nvmet_parse_fabrics_io_cmd() and rename the existing nvmet_parse_fabrics_cmd() to nvmet_parse_fabrics_admin_cmd().
--- drivers/nvme/target/admin-cmd.c | 2 +- drivers/nvme/target/core.c | 4 ++++ drivers/nvme/target/fabrics-cmd.c | 17 ++++++++++++++++- drivers/nvme/target/nvmet.h | 3 ++- 4 files changed, 23 insertions(+), 3 deletions(-)
drivers/nvme/target/admin-cmd.c +1 -1
@@ -1017,7 +1017,7 @@ u16 nvmet_parse_admin_cmd(struct nvmet_req *req)u16 ret;if (nvme_is_fabrics (cmd)) -return nvmet_parse_fabrics_cmd (req); +return nvmet_parse_fabrics_admin_cmd (req);if (nvmet_is_disc_subsys (nvmet_req_subsys (req)))return nvmet_parse_discovery_cmd (req);
drivers/nvme/target/core.c +4 -0
@@ -865,8 +865,12 @@ static inline u16 nvmet_io_cmd_check_access(struct nvmet_req *req)static u16 nvmet_parse_io_cmd (struct nvmet_req *req) { +struct nvme_command *cmd = req->cmd;u16 ret; +if (nvme_is_fabrics (cmd)) +return nvmet_parse_fabrics_io_cmd (req); + ret =nvmet_check_ctrl_status (req);if (unlikely (ret))return ret;
drivers/nvme/target/fabrics-cmd.c +16 -1
@@ -82,7 +82,7 @@ static void nvmet_execute_prop_get(struct nvmet_req *req)nvmet_req_complete (req, status); } -u16 nvmet_parse_fabrics_cmd (struct nvmet_req *req) +u16 nvmet_parse_fabrics_admin_cmd (struct nvmet_req *req) {struct nvme_command *cmd = req->cmd;
@@ -103,6 +103,21 @@ u16 nvmet_parse_fabrics_cmd(struct nvmet_req *req)return 0 ; } +u16 nvmet_parse_fabrics_io_cmd (struct nvmet_req *req) +{ +struct nvme_command *cmd = req->cmd; + +switch (cmd->fabrics.fctype) { +default : +pr_debug ("received unknown capsule type 0x%x\n" , + cmd->fabrics.fctype); + req->error_loc = offsetof(struct nvmf_common_command , fctype); +return NVME_SC_INVALID_OPCODE |NVME_SC_DNR ; + } + +return 0 ; +} +static u16 nvmet_install_queue (struct nvmet_ctrl *ctrl,struct nvmet_req *req) {struct nvmf_connect_command *c = &req->cmd->connect;
drivers/nvme/target/nvmet.h +2 -1
@@ -419,7 +419,8 @@ u16 nvmet_file_parse_io_cmd(struct nvmet_req *req);u16 nvmet_bdev_zns_parse_io_cmd (struct nvmet_req *req);u16 nvmet_parse_admin_cmd (struct nvmet_req *req);u16 nvmet_parse_discovery_cmd (struct nvmet_req *req); -u16 nvmet_parse_fabrics_cmd (struct nvmet_req *req); +u16 nvmet_parse_fabrics_admin_cmd (struct nvmet_req *req); +u16 nvmet_parse_fabrics_io_cmd (struct nvmet_req *req);bool nvmet_req_init (struct nvmet_req *req,struct nvmet_cq *cq,struct nvmet_sq *sq,const struct nvmet_fabrics_ops *ops);
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 9/11 nvmet: Implement basic In-Band Authentication
In reply to Hannes ReineckeImplement NVMe-oF In-Band authentication according to NVMe TPAR 8006. This patch adds three additional configfs entries 'dhchap_key', 'dhchap_ctrl_key', and 'dhchap_hash' to the 'host' configfs directory. The 'dhchap_key' and 'dhchap_ctrl_key' entries need to be in the ASCII format as specified in NVMe Base Specification v2.0 section 8.13.5.8 'Secret representation'.
'dhchap_hash' defaults to 'hmac(sha256)', and can be written to to switch to a different HMAC algorithm.
--- drivers/nvme/target/Kconfig | 13 + drivers/nvme/target/Makefile | 1 + drivers/nvme/target/admin-cmd.c | 2 + drivers/nvme/target/auth.c | 367 ++++++++++++++++++ drivers/nvme/target/configfs.c | 107 +++++- drivers/nvme/target/core.c | 11 + drivers/nvme/target/fabrics-cmd-auth.c | 502 +++++++++++++++++++++++++ drivers/nvme/target/fabrics-cmd.c | 38 +- drivers/nvme/target/nvmet.h | 62 +++ 9 files changed, 1100 insertions(+), 3 deletions(-) create mode 100644 drivers/nvme/target/auth.c create mode 100644 drivers/nvme/target/fabrics-cmd-auth.c
drivers/nvme/target/Kconfig +13 -0
@@ -83,3 +83,16 @@ config NVME_TARGET_TCP devices over TCP. If unsure, say N. + +config NVME_TARGET_AUTH + bool "NVMe over Fabrics In-band Authentication support" + depends on NVME_TARGET + select NVME_COMMON + select CRYPTO + select CRYPTO_HMAC + select CRYPTO_SHA256 + select CRYPTO_SHA512 + help + This enables support for NVMe over Fabrics In-band Authentication + + If unsure, say N.
drivers/nvme/target/Makefile +1 -0
@@ -13,6 +13,7 @@ nvmet-y += core.o configfs.o admin-cmd.o fabrics-cmd.o \ discovery.o io-cmd-file.o io-cmd-bdev.o nvmet-$(CONFIG_NVME_TARGET_PASSTHRU) += passthru.o nvmet-$(CONFIG_BLK_DEV_ZONED) += zns.o +nvmet-$(CONFIG_NVME_TARGET_AUTH) += fabrics-cmd-auth.o auth.o nvme-loop-y += loop.o nvmet-rdma-y += rdma.o nvmet-fc-y += fc.o
drivers/nvme/target/admin-cmd.c +2 -0
@@ -1018,6 +1018,8 @@ u16 nvmet_parse_admin_cmd(struct nvmet_req *req)if (nvme_is_fabrics (cmd))return nvmet_parse_fabrics_admin_cmd (req); +if (unlikely (!nvmet_check_auth_status (req))) +return NVME_SC_AUTH_REQUIRED |NVME_SC_DNR ;if (nvmet_is_disc_subsys (nvmet_req_subsys (req)))return nvmet_parse_discovery_cmd (req);
drivers/nvme/target/auth.c new file +367 -0
367 lines not on this page · raw message
drivers/nvme/target/configfs.c +106 -1
134 lines not on this page · raw message
drivers/nvme/target/core.c +11 -0
@@ -795,6 +795,7 @@ void nvmet_sq_destroy(struct nvmet_sq *sq)wait_for_completion (&sq->confirm_done);wait_for_completion (&sq->free_done);percpu_ref_exit (&sq->ref); +nvmet_auth_sq_free (sq);if (ctrl) {/*
@@ -871,6 +872,9 @@ static u16 nvmet_parse_io_cmd(struct nvmet_req *req)if (nvme_is_fabrics (cmd))return nvmet_parse_fabrics_io_cmd (req); +if (unlikely (!nvmet_check_auth_status (req))) +return NVME_SC_AUTH_REQUIRED |NVME_SC_DNR ; + ret =nvmet_check_ctrl_status (req);if (unlikely (ret))return ret;
@@ -1275,6 +1279,11 @@ u16 nvmet_check_ctrl_status(struct nvmet_req *req) req->cmd->common.opcode, req->sq->qid);return NVME_SC_CMD_SEQ_ERROR |NVME_SC_DNR ; } + +if (unlikely (!nvmet_check_auth_status (req))) { +pr_warn ("qid %d not authenticated\n" , req->sq->qid); +return NVME_SC_AUTH_REQUIRED |NVME_SC_DNR ; + }return 0 ; }
@@ -1465,6 +1474,8 @@ static void nvmet_ctrl_free(struct kref *ref)flush_work (&ctrl->async_event_work);cancel_work_sync (&ctrl->fatal_err_work); +nvmet_destroy_auth (ctrl); +ida_free (&cntlid_ida, ctrl->cntlid);nvmet_async_events_free (ctrl);
drivers/nvme/target/fabrics-cmd-auth.c new file +502 -0
502 lines not on this page · raw message
drivers/nvme/target/fabrics-cmd.c +36 -2
@@ -93,6 +93,14 @@ u16 nvmet_parse_fabrics_admin_cmd(struct nvmet_req *req)case nvme_fabrics_type_property_get: req->execute = nvmet_execute_prop_get;break ; +#ifdef CONFIG_NVME_TARGET_AUTH +case nvme_fabrics_type_auth_send: + req->execute = nvmet_execute_auth_send; +break ; +case nvme_fabrics_type_auth_receive: + req->execute = nvmet_execute_auth_receive; +break ; +#endif default :pr_debug ("received unknown capsule type 0x%x\n" , cmd->fabrics.fctype);
@@ -108,6 +116,14 @@ u16 nvmet_parse_fabrics_io_cmd(struct nvmet_req *req)struct nvme_command *cmd = req->cmd;switch (cmd->fabrics.fctype) { +#ifdef CONFIG_NVME_TARGET_AUTH +case nvme_fabrics_type_auth_send: + req->execute = nvmet_execute_auth_send; +break ; +case nvme_fabrics_type_auth_receive: + req->execute = nvmet_execute_auth_receive; +break ; +#endif default :pr_debug ("received unknown capsule type 0x%x\n" , cmd->fabrics.fctype);
@@ -188,6 +204,7 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)struct nvmf_connect_data *d;struct nvmet_ctrl *ctrl =NULL ;u16 status =0 ; +int ret;if (!nvmet_check_transfer_len (req,sizeof (struct nvmf_connect_data )))return ;
@@ -230,18 +247,32 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)uuid_copy (&ctrl->hostid, &d->hostid); + ret =nvmet_setup_auth (ctrl); +if (ret <0 ) { +pr_err ("Failed to setup authentication, error %d\n" , ret); +nvmet_ctrl_put (ctrl); +if (ret == -EPERM ) + status = (NVME_SC_CONNECT_INVALID_HOST |NVME_SC_DNR ); +else + status =NVME_SC_INTERNAL ; + goto out; + } + status =nvmet_install_queue (ctrl, req);if (status) {nvmet_ctrl_put (ctrl); goto out; } -pr_info ("creating %s controller %d for subsystem %s for NQN %s%s.\n" , +pr_info ("creating %s controller %d for subsystem %s for NQN %s%s%s.\n" ,nvmet_is_disc_subsys (ctrl->subsys) ?"discovery" :"nvm" , ctrl->cntlid, ctrl->subsys->subsysnqn, ctrl->hostnqn, - ctrl->pi_support ?" T10-PI is enabled" :"" ); + ctrl->pi_support ?" T10-PI is enabled" :"" , +nvmet_has_auth (ctrl) ?" with DH-HMAC-CHAP" :"" ); req->cqe->result.u16 =cpu_to_le16 (ctrl->cntlid); +if (nvmet_has_auth (ctrl)) +nvmet_init_auth (ctrl, req); out:kfree (d); complete:
@@ -301,6 +332,9 @@ static void nvmet_execute_io_connect(struct nvmet_req *req) req->cqe->result.u16 =cpu_to_le16 (ctrl->cntlid);pr_debug ("adding queue %d to ctrl %d.\n" , qid, ctrl->cntlid); + req->cqe->result.u16 =cpu_to_le16 (ctrl->cntlid); +if (nvmet_has_auth (ctrl)) +nvmet_init_auth (ctrl, req); out:kfree (d);
drivers/nvme/target/nvmet.h +62 -0
84 lines not on this page · raw message
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 10/11 nvmet-auth: Diffie-Hellman key exchange support
In reply to Hannes ReineckeImplement Diffie-Hellman key exchange using FFDHE groups for NVMe In-Band Authentication.
This patch adds a new host configfs attribute 'dhchap_dhgroup' to select the FFDHE group to use.
--- drivers/nvme/target/Kconfig | 2 + drivers/nvme/target/auth.c | 157 +++++++++++++++++++++++++ drivers/nvme/target/configfs.c | 31 +++++ drivers/nvme/target/fabrics-cmd-auth.c | 41 +++++-- drivers/nvme/target/nvmet.h | 9 ++ 5 files changed, 232 insertions(+), 8 deletions(-)
drivers/nvme/target/Kconfig +2 -0
@@ -92,6 +92,8 @@ config NVME_TARGET_AUTH select CRYPTO_HMAC select CRYPTO_SHA256 select CRYPTO_SHA512 + select CRYPTO_DH + select CRYPTO_DH_GROUPS_RFC7919 help This enables support for NVMe over Fabrics In-band Authentication
drivers/nvme/target/auth.c +157 -0
190 lines not on this page · raw message
drivers/nvme/target/configfs.c +31 -0
@@ -1746,10 +1746,41 @@ static ssize_t nvmet_host_dhchap_hash_store(struct config_item *item,CONFIGFS_ATTR (nvmet_host_, dhchap_hash); +static ssize_t nvmet_host_dhchap_dhgroup_show (struct config_item *item, +char *page) +{ +struct nvmet_host *host =to_host (item); +const char *dhgroup =nvme_auth_dhgroup_name (host->dhchap_dhgroup_id); + +return sprintf (page,"%s\n" , dhgroup ? dhgroup :"none" ); +} + +static ssize_t nvmet_host_dhchap_dhgroup_store (struct config_item *item, +const char *page,size_t count) +{ +struct nvmet_host *host =to_host (item); +int dhgroup_id; + + dhgroup_id =nvme_auth_dhgroup_id (page); +if (dhgroup_id ==NVME_AUTH_DHGROUP_INVALID ) +return -EINVAL ; +if (dhgroup_id !=NVME_AUTH_DHGROUP_NULL ) { +const char *kpp =nvme_auth_dhgroup_kpp (dhgroup_id); + +if (!crypto_has_kpp (kpp,0 ,0 )) +return -EINVAL ; + } + host->dhchap_dhgroup_id = dhgroup_id; +return count; +} + +CONFIGFS_ATTR (nvmet_host_, dhchap_dhgroup); +static struct configfs_attribute *nvmet_host_attrs[] = { &nvmet_host_attr_dhchap_key, &nvmet_host_attr_dhchap_ctrl_key, &nvmet_host_attr_dhchap_hash, + &nvmet_host_attr_dhchap_dhgroup,NULL , };#endif /* CONFIG_NVME_TARGET_AUTH */
drivers/nvme/target/fabrics-cmd-auth.c +33 -8
84 lines not on this page · raw message
drivers/nvme/target/nvmet.h +9 -0
@@ -225,6 +225,10 @@ struct nvmet_ctrl {struct nvme_dhchap_key *host_key;struct nvme_dhchap_key *ctrl_key;u8 shash_id; +struct crypto_kpp *dh_tfm; +u8 dh_gid; +u8 *dh_key; +size_t dh_keysize;#endif };
@@ -701,6 +705,7 @@ int nvmet_setup_auth(struct nvmet_ctrl *ctrl);void nvmet_init_auth (struct nvmet_ctrl *ctrl,struct nvmet_req *req);void nvmet_destroy_auth (struct nvmet_ctrl *ctrl);void nvmet_auth_sq_free (struct nvmet_sq *sq); +int nvmet_setup_dhgroup (struct nvmet_ctrl *ctrl,u8 dhgroup_id);bool nvmet_check_auth_status (struct nvmet_req *req);int nvmet_auth_host_hash (struct nvmet_req *req,u8 *response,unsigned int hash_len);
@@ -710,6 +715,10 @@ static inline bool nvmet_has_auth(struct nvmet_ctrl *ctrl) {return ctrl->host_key !=NULL ; } +int nvmet_auth_ctrl_exponential (struct nvmet_req *req, +u8 *buf,int buf_size); +int nvmet_auth_ctrl_sesskey (struct nvmet_req *req, +u8 *buf,int buf_size);#else static inline int nvmet_setup_auth (struct nvmet_ctrl *ctrl) {
Signature · 1 line
2.29.2
Hannes Reinecke
maintainerPATCH v15 11/11 nvmet-auth: expire authentication sessions
In reply to Hannes ReineckeEach authentication step is required to be completed within the KATO interval (or two minutes if not set). So add a workqueue function to reset the transaction ID and the expected next protocol step; this will automatically the next authentication command referring to the terminated authentication.
--- drivers/nvme/target/auth.c | 1 + drivers/nvme/target/fabrics-cmd-auth.c | 20 +++++++++++++++++++- drivers/nvme/target/nvmet.h | 1 + 3 files changed, 21 insertions(+), 1 deletion(-)
drivers/nvme/target/auth.c +1 -0
@@ -218,6 +218,7 @@ int nvmet_setup_auth(struct nvmet_ctrl *ctrl)void nvmet_auth_sq_free (struct nvmet_sq *sq) { +cancel_delayed_work (&sq->auth_expired_work);kfree (sq->dhchap_c1); sq->dhchap_c1 =NULL ;kfree (sq->dhchap_c2);
drivers/nvme/target/fabrics-cmd-auth.c +19 -1
@@ -12,11 +12,24 @@#include <crypto/kpp.h> #include "nvmet.h" +static void nvmet_auth_expired_work (struct work_struct *work) +{ +struct nvmet_sq *sq =container_of (to_delayed_work (work), + struct nvmet_sq, auth_expired_work); + +pr_debug ("%s: ctrl %d qid %d transaction %u expired, resetting\n" , + __func__, sq->ctrl->cntlid, sq->qid, sq->dhchap_tid); + sq->dhchap_step =NVME_AUTH_DHCHAP_MESSAGE_NEGOTIATE ; + sq->dhchap_tid =-1 ; +} +void nvmet_init_auth (struct nvmet_ctrl *ctrl,struct nvmet_req *req) {u32 result =le32_to_cpu (req->cqe->result.u32);/* Initialize in-band authentication */ +INIT_DELAYED_WORK (&req->sq->auth_expired_work, + nvmet_auth_expired_work); req->sq->authenticated = false; req->sq->dhchap_step =NVME_AUTH_DHCHAP_MESSAGE_NEGOTIATE ; result |= (u32 )NVME_CONNECT_AUTHREQ_ATR <<16 ;
@@ -333,8 +346,13 @@ void nvmet_execute_auth_send(struct nvmet_req *req) req->cqe->result.u64 =0 ;nvmet_req_complete (req, status);if (req->sq->dhchap_step !=NVME_AUTH_DHCHAP_MESSAGE_SUCCESS2 && - req->sq->dhchap_step !=NVME_AUTH_DHCHAP_MESSAGE_FAILURE2 ) + req->sq->dhchap_step !=NVME_AUTH_DHCHAP_MESSAGE_FAILURE2 ) { +unsigned long auth_expire_secs = ctrl->kato ? ctrl->kato :120 ; + +mod_delayed_work (system_wq, &req->sq->auth_expired_work, + auth_expire_secs *HZ );return ; + }/* Final states, clear up variables */ nvmet_auth_sq_free (req->sq);if (req->sq->dhchap_step ==NVME_AUTH_DHCHAP_MESSAGE_FAILURE2 )
drivers/nvme/target/nvmet.h +1 -0
@@ -109,6 +109,7 @@ struct nvmet_sq {u32 sqhd;bool sqhd_disabled;#ifdef CONFIG_NVME_TARGET_AUTH +struct delayed_work auth_expired_work;bool authenticated;u16 dhchap_tid;u16 dhchap_status;
Signature · 1 line
2.29.2