Add SEV_INIT_EX support
Subsystems (5)
v2 cover Add SEV_INIT_EX support
SEV_INIT requires users to unlock their SPI bus for the PSP's non volatile (NV) storage. Users may wish to lock their SPI bus for numerous reasons, to support this the PSP firmware supports SEV_INIT_EX. INIT_EX allows the firmware to use a region of memory for its NV storage leaving the kernel responsible for actually storing the data in a persistent way. This series adds a new module parameter to ccp allowing users to specify a path to a file for use as the PSP's NV storage. The ccp driver then reads the file into memory for the PSP to use and is responsible for writing the file whenever the PSP modifies the memory region.
David Rientjes (1):
crypto: ccp - Add SEV_INIT_EX support
Peter Gonda (3):
crypto: ccp - Fix SEV_INIT error logging on init
crypto: ccp - Move SEV_INIT retry for corrupted data
crypto: ccp - Refactor out sev_fw_alloc()
.../virt/kvm/amd-memory-encryption.rst | 4 + drivers/crypto/ccp/sev-dev.c | 230 +++++++++++++++--- include/linux/psp-sev.h | 21 ++ 3 files changed, 222 insertions(+), 33 deletions(-)
Signature · 1 line
2.33.1.1089.g2158813163f-goog
PATCH v2 1/4 crypto: ccp - Fix SEV_INIT error logging on init
Currently only the firmware error code is printed. This is incomplete and also incorrect as error cases exists where the firmware is never called and therefore does not set an error code. This change zeros the firmware error code in case the call does not get that far and prints the return code for non firmware errors.
--- drivers/crypto/ccp/sev-dev.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/crypto/ccp/sev-dev.c +3 -2
@@ -1065,7 +1065,7 @@ void sev_pci_init(void) {struct sev_device *sev = psp_master->sev_data;struct page *tmr_page; -int error, rc; +int error =0 , rc;if (!sev)return ;
@@ -1104,7 +1104,8 @@ void sev_pci_init(void) }if (rc) { -dev_err (sev->dev,"SEV: failed to INIT error %#x\n" , error); +dev_err (sev->dev,"SEV: failed to INIT error %#x, rc %d\n" , + error, rc);return ; }
Signature · 1 line
2.33.1.1089.g2158813163f-goog
PATCH v2 2/4 crypto: ccp - Move SEV_INIT retry for corrupted data
In reply to Peter GondaThis change moves the data corrupted retry of SEV_INIT into the __sev_platform_init_locked() function. This is for upcoming INIT_EX support as well as helping direct callers of __sev_platform_init_locked() which currently do not support the retry.
--- drivers/crypto/ccp/sev-dev.c | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-)
drivers/crypto/ccp/sev-dev.c +12 -12
@@ -267,6 +267,18 @@ static int __sev_platform_init_locked(int *error) } rc =__sev_do_cmd_locked (SEV_CMD_INIT , &data, error); +if (rc && *error ==SEV_RET_SECURE_DATA_INVALID ) { +/* +* INIT command returned an integrity check failure +* status code, meaning that firmware load and +* validation of SEV related persistent data has +* failed and persistent state has been erased. +* Retrying INIT command here should succeed. +*/ +dev_dbg (sev->dev,"SEV: retrying INIT command" ); + rc =__sev_do_cmd_locked (SEV_CMD_INIT , &data, error); + } +if (rc)return rc;
@@ -1091,18 +1103,6 @@ void sev_pci_init(void)/* Initialize the platform */ rc =sev_platform_init (&error); -if (rc && (error ==SEV_RET_SECURE_DATA_INVALID )) { -/* -* INIT command returned an integrity check failure -* status code, meaning that firmware load and -* validation of SEV related persistent data has -* failed and persistent state has been erased. -* Retrying INIT command here should succeed. -*/ -dev_dbg (sev->dev,"SEV: retrying INIT command" ); - rc =sev_platform_init (&error); - } -if (rc) {dev_err (sev->dev,"SEV: failed to INIT error %#x, rc %d\n" , error, rc);
Signature · 1 line
2.33.1.1089.g2158813163f-goog
PATCH v2 3/4 crypto: ccp - Refactor out sev_fw_alloc()
In reply to Peter GondaCreates a helper function sev_fw_alloc() which can be used to allocate aligned memory regions for use by the PSP firmware. Currently only used for the SEV-ES TMR region but will be used for the SEV_INIT_EX NV memory region.
--- drivers/crypto/ccp/sev-dev.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-)
drivers/crypto/ccp/sev-dev.c +13 -7
@@ -141,6 +141,17 @@ static int sev_cmd_buffer_len(int cmd)return 0 ; } +static void *sev_fw_alloc (unsigned long len) +{ +struct page *page; + + page =alloc_pages (GFP_KERNEL ,get_order (len)); +if (!page) +return NULL ; + +return page_address (page); +} +static int __sev_do_cmd_locked (int cmd,void *data,int *psp_ret) {struct psp_device *psp = psp_master;
@@ -1076,7 +1087,6 @@ EXPORT_SYMBOL_GPL(sev_issue_cmd_external_user);void sev_pci_init (void ) {struct sev_device *sev = psp_master->sev_data; -struct page *tmr_page;int error =0 , rc;if (!sev)
@@ -1092,14 +1102,10 @@ void sev_pci_init(void)sev_get_api_version ();/* Obtain the TMR memory area for SEV-ES use */ - tmr_page =alloc_pages (GFP_KERNEL ,get_order (SEV_ES_TMR_SIZE )); -if (tmr_page) { - sev_es_tmr =page_address (tmr_page); - }else { - sev_es_tmr =NULL ; + sev_es_tmr =sev_fw_alloc (SEV_ES_TMR_SIZE ); +if (!sev_es_tmr)dev_warn (sev->dev,"SEV: TMR allocation failed, SEV-ES support unavailable\n" ); - }/* Initialize the platform */ rc =sev_platform_init (&error);
Signature · 1 line
2.33.1.1089.g2158813163f-goog
PATCH v2 4/4 crypto: ccp - Add SEV_INIT_EX support
In reply to Peter GondaFrom: David Rientjes <[email protected]>
Add new module parameter to allow users to use SEV_INIT_EX instead of SEV_INIT. This helps users who lock their SPI bus to use the PSP for SEV functionality. The 'init_ex_path' parameter defaults to NULL which means the kernel will use SEV_INIT, if a path is specified SEV_INIT_EX will be used with the data found at the path. On certain PSP commands this file is written to as the PSP updates the NV memory region. Depending on file system initialization this file open may fail during module init but the CCP driver for SEV already has sufficient retries for platform initialization. During normal operation of PSP system and SEV commands if the PSP has not been initialized it is at run time. If the file at 'init_ex_path' does not exist the PSP will not be initialized. The user must create the file prior to use with 32Kb of 0xFFs per spec.
--- .../virt/kvm/amd-memory-encryption.rst | 4 + drivers/crypto/ccp/sev-dev.c | 185 ++++++++++++++++-- include/linux/psp-sev.h | 21 ++ 3 files changed, 196 insertions(+), 14 deletions(-)
Documentation/virt/kvm/amd-memory-encryption.rst +4 -0
@@ -84,6 +84,10 @@ guests, such as launching, running, snapshotting, migrating and decommissioning. The KVM_SEV_INIT command is used by the hypervisor to initialize the SEV platform context. In a typical workflow, this command should be the first command issued. +The AMD-SP can be initialized either by using its own non-volatile storage or +the system can manage the NV storage for the AMD-SP using the module parameter +``init_ex_path``. This file must exist, to create a new NV storage file allocate +a the file with 32Kb of 0xFF as required by the SEV FW spec. Returns: 0 on success, -negative on error
drivers/crypto/ccp/sev-dev.c +171 -14
268 lines not on this page · raw message
include/linux/psp-sev.h +21 -0
@@ -52,6 +52,7 @@ enum sev_cmd {SEV_CMD_DF_FLUSH =0x00A ,SEV_CMD_DOWNLOAD_FIRMWARE =0x00B ,SEV_CMD_GET_ID =0x00C , +SEV_CMD_INIT_EX =0x00D ,/* Guest commands */ SEV_CMD_DECOMMISSION = 0x020,
@@ -102,6 +103,26 @@ struct sev_data_init {u32 tmr_len;/* In */ } __packed; +/** +* struct sev_data_init_ex - INIT_EX command parameters +* +* @length: len of the command buffer read by the PSP +* @flags: processing flags +* @tmr_address: system physical address used for SEV-ES +* @tmr_len: len of tmr_address +* @nv_address: system physical address used for PSP NV storage +* @nv_len: len of nv_address +*/ +struct sev_data_init_ex { +u32 length;/* In */ +u32 flags;/* In */ +u64 tmr_address;/* In */ +u32 tmr_len;/* In */ +u32 reserved;/* In */ +u64 nv_address;/* In/Out */ +u32 nv_len;/* In */ +} __packed; +#define SEV_INIT_FLAGS_SEV_ES 0x01/**
Signature · 1 line
2.33.1.1089.g2158813163f-goog
Tom Lendacky
maintainerRe: [PATCH V2 4/4] crypto: ccp - Add SEV_INIT_EX support
Quoted 40 lines · Peter Gonda
From: David Rientjes <[email protected]>
Add new module parameter to allow users to use
SEV_INIT_EXinstead ofSEV_INIT. This helps users who lock their SPI bus to use the PSP for SEV functionality. The 'init_ex_path' parameter defaults to NULL which means the kernel will useSEV_INIT, if a path is specifiedSEV_INIT_EXwill be used with the data found at the path. On certain PSP commands this file is written to as the PSP updates the NV memory region. Depending on file system initialization this file open may fail during module init but the CCP driver for SEV already has sufficient retries for platform initialization. During normal operation of PSP system and SEV commands if the PSP has not been initialized it is at run time. If the file at 'init_ex_path' does not exist the PSP will not be initialized. The user must create the file prior to use with 32Kb of 0xFFs per spec.Signed-off-by David Rientjes <[email protected]>Co-developed-by Peter Gonda <[email protected]>Signed-off-by Peter Gonda <[email protected]>Reviewed-by Marc Orr <[email protected]>Cc Tom Lendacky <[email protected]>Cc Brijesh Singh <[email protected]>Cc Marc Orr <[email protected]>Cc Joerg Roedel <[email protected]>Cc Herbert Xu <[email protected]>Cc David Rientjes <[email protected]>Cc John Allen <[email protected]>Cc "David S. Miller" <[email protected]>--- .../virt/kvm/amd-memory-encryption.rst | 4 + drivers/crypto/ccp/sev-dev.c | 185 ++++++++++++++++-- include/linux/psp-sev.h | 21 ++ 3 files changed, 196 insertions(+), 14 deletions(-)Documentation/virt/kvm/amd-memory-encryption.rst +0
-0
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 86
Quoted 3 lines
Documentation/virt/kvm/amd-memory-encryption.rst +0
-0@@ -84,6 +84,10 @@ guests, such as launching, running, snapshotting, migrating and decommissioning. The KVM_SEV_INIT command is used by the hypervisor to initialize the SEV platform context. In a typical workflow, this command should be the first command issued.
Add a blank line here.
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 87
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+The AMD-SP can be initialized either by using its own non-volatile storage or
Just to be consistent within the document, s/AMD-SP/firmware/
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 88
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+the system can manage the NV storage for the AMD-SP using the module parameter
s/system/OS/
s/AMD-SP/firmware/
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 89
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+``init_ex_path``. This file must exist, to create a new NV storage file allocate
s/This file must exist/The file specified by ``init_ex_path`` must exist./ s/, to create/ To create/
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 90
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+a the file with 32Kb of 0xFF as required by the SEV FW spec.
s/a the/the/
s/32Kb/32KB bytes/
Just to be consistent within the document, s/SEV FW spec/SEV spec/
Quoted 54 lines
Documentation/virt/kvm/amd-memory-encryption.rst +0
-0Returns: 0 on success, -negative on errordrivers/crypto/ccp/sev-dev.c +15
-0@@ -22,6 +22,7 @@#include <linux/firmware.h> #include <linux/gfp.h> #include <linux/cpufeature.h> +#include <linux/fs.h> #include <asm/smp.h> @@ -43,6 +44,10 @@ static int psp_probe_timeout = 5;module_param (psp_probe_timeout, int,0644 );MODULE_PARM_DESC (psp_probe_timeout," default timeout value, in seconds, during PSP device probe" ); +static char *init_ex_path; +module_param (init_ex_path, charp,0660 ); +MODULE_PARM_DESC (init_ex_path," Path for INIT_EX data; if set try INIT_EX" ); +MODULE_FIRMWARE ("amd/amd_sev_fam17h_model0xh.sbin" );/* 1st gen EPYC */ MODULE_FIRMWARE ("amd/amd_sev_fam17h_model3xh.sbin" );/* 2nd gen EPYC */ MODULE_FIRMWARE ("amd/amd_sev_fam19h_model0xh.sbin" );/* 3rd gen EPYC */ @@ -58,6 +63,14 @@ static int psp_timeout;#define SEV_ES_TMR_SIZE (1024 * 1024)static void *sev_es_tmr; +/* INIT_EX NV Storage: +* The NV Storage is a 32Kb area and must be 4Kb page aligned. Use the page +* allocator to allocate the memory, which will return aligned memory for the +* specified allocation order. +*/ +#define NV_LENGTH (32 * 1024) +static void *sev_init_ex_nv_address; +static inline bool sev_version_greater_or_equal (u8 maj,u8 min) {struct sev_device *sev = psp_master->sev_data;@@ -107,6 +120,7 @@ static int sev_cmd_buffer_len(int cmd) {switch (cmd) {case SEV_CMD_INIT :return sizeof (struct sev_data_init ); +case SEV_CMD_INIT_EX :return sizeof (struct sev_data_init_ex );case SEV_CMD_PLATFORM_STATUS :return sizeof (struct sev_user_data_status );case SEV_CMD_PEK_CSR :return sizeof (struct sev_data_pek_csr );case SEV_CMD_PEK_CERT_IMPORT :return sizeof (struct sev_data_pek_cert_import );@@ -152,6 +166,89 @@ static void *sev_fw_alloc(unsigned long len)return page_address (page); } +static int sev_read_nv_memory (void )
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 179
Quoted 10 lines
drivers/crypto/ccp/sev-dev.c +10
-0+{ +struct file *fp; +ssize_t nread; + +if (!sev_init_ex_nv_address) +return -EOPNOTSUPP ; + + fp =filp_open (init_ex_path,O_RDONLY ,0 ); +if (IS_ERR (fp)) { +const int ret =PTR_ERR (fp);
I don't think you need the "const" here.
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 183
Quoted 4 lines
drivers/crypto/ccp/sev-dev.c +4
-0+ +dev_err (psp_master->dev, +"sev could not open file for read, error %d\n" , + ret);
Maybe "SEV: could not open %s for read, ret=%d\n"
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 187
Quoted 4 lines
drivers/crypto/ccp/sev-dev.c +4
-0+return ret; + } + + nread =kernel_read (fp, sev_init_ex_nv_address,NV_LENGTH ,NULL );
kernel_read can return an error, you should check nread before continuing.
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 188
Quoted 1 line
drivers/crypto/ccp/sev-dev.c +1
-0+dev_dbg (psp_master->dev,"sev NV read %d bytes\n" , nread);
"SEV: read %ld bytes from NV file\n"
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 194
Quoted 6 lines
drivers/crypto/ccp/sev-dev.c +6
-0+filp_close (fp,NULL ); + +return 0 ; +} + +static int sev_write_nv_memory (void )
The return code is never checked by the caller. Is it expected that any error is not supposed to stop SEV for the current boot? Should you make this void, then?
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 206
Quoted 12 lines
drivers/crypto/ccp/sev-dev.c +12
-0+{ +struct sev_device *sev = psp_master->sev_data; +struct file *fp; +loff_t offset =0 ; +int ret; + +if (!sev_init_ex_nv_address) +return -EOPNOTSUPP ; + + fp =filp_open (init_ex_path,O_CREAT |O_WRONLY ,0600 ); +if (IS_ERR (fp)) { +dev_err (sev->dev,"sev NV data could not be created\n" );
You should output a message similar to what is in sev_read_nv_memory().
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 210
Quoted 4 lines
drivers/crypto/ccp/sev-dev.c +4
-0+return PTR_ERR (fp); + } + + ret =kernel_write (fp, sev_init_ex_nv_address,NV_LENGTH , &offset);
kernel_write returns a ssize_t, but ret is an int. and maybe use nwrite similar to nread in sev_read_nv_memory()?
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 216
Quoted 6 lines
drivers/crypto/ccp/sev-dev.c +6
-0+vfs_fsync (fp,0 ); +filp_close (fp,NULL ); + +if (ret !=NV_LENGTH ) { + dev_err(sev->dev, + "failed to write %d bytes to non volatile memory area, ret=%lu\n",
"SEV: failed to write %u bytes to NV file, ret=%ld\n"
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 223
Quoted 6 lines
drivers/crypto/ccp/sev-dev.c +6
-0+ NV_LENGTH, ret); > + if (ret >= 0) + return -EIO; +return ret; + } + +dev_dbg (sev->dev,"wrote to non volatile memory area\n" );
"SEV: write successful to NV file\n"
Quoted 156 lines
drivers/crypto/ccp/sev-dev.c +89
-13+ +return 0 ; +} + +static void sev_write_nv_memory_if_required (int cmd_id) +{ +if (!sev_init_ex_nv_address) +return ; + +/* +* Only a few platform commands modify the SPI/NV area, but none of the +* non-platform commands do. Only INIT(_EX), PLATFORM_RESET, PEK_GEN, +* PEK_CERT_IMPORT, and PDH_GEN do. +*/ +switch (cmd_id) { +case SEV_CMD_FACTORY_RESET : +case SEV_CMD_INIT_EX : +case SEV_CMD_PDH_GEN : +case SEV_CMD_PEK_CERT_IMPORT : +case SEV_CMD_PEK_GEN : +break ; +default : +return ; + }; + +sev_write_nv_memory (); +} +static int __sev_do_cmd_locked (int cmd,void *data,int *psp_ret) {struct psp_device *psp = psp_master;@@ -221,6 +318,8 @@ static int __sev_do_cmd_locked(int cmd, void *data, int *psp_ret)dev_dbg (sev->dev,"sev command %#x failed (%#010x)\n" , cmd, reg &PSP_CMDRESP_ERR_MASK ); ret = -EIO ; + }else { +sev_write_nv_memory_if_required (cmd); } print_hex_dump_debug("(out): ", DUMP_PREFIX_OFFSET, 16, 2, data,@@ -247,22 +346,42 @@ static int sev_do_cmd(int cmd, void *data, int *psp_ret)return rc; } -static int __sev_platform_init_locked (int *error) +static int __sev_init_locked (int *error) { -struct psp_device *psp = psp_master;struct sev_data_init data; -struct sev_device *sev; -int rc =0 ; -if (!psp || !psp->sev_data) -return -ENODEV ; +memset (&data,0 ,sizeof (data)); +if (sev_es_tmr) { +u64 tmr_pa; - sev = psp->sev_data; +/* +* Do not include the encryption mask on the physical +* address of the TMR (firmware should clear it anyway). +*/ + tmr_pa =__pa (sev_es_tmr); -if (sev->state ==SEV_STATE_INIT ) -return 0 ; + data.flags |=SEV_INIT_FLAGS_SEV_ES ; + data.tmr_address = tmr_pa; + data.tmr_len =SEV_ES_TMR_SIZE ; + } + +return __sev_do_cmd_locked (SEV_CMD_INIT , &data, error); +} + +static int __sev_init_ex_locked (int *error) +{ +struct sev_data_init_ex data; +int ret;memset (&data,0 ,sizeof (data)); + data.length =sizeof (data); + data.nv_address =__psp_pa (sev_init_ex_nv_address); + data.nv_len =NV_LENGTH ; + + ret =sev_read_nv_memory (); +if (ret) +return ret; +if (sev_es_tmr) {u64 tmr_pa;@@ -277,7 +396,27 @@ static int __sev_platform_init_locked(int *error) data.tmr_len =SEV_ES_TMR_SIZE ; } - rc =__sev_do_cmd_locked (SEV_CMD_INIT , &data, error); +return __sev_do_cmd_locked (SEV_CMD_INIT_EX , &data, error); +} + +static int __sev_platform_init_locked (int *error) +{ +struct psp_device *psp = psp_master; +struct sev_device *sev; +int rc; +int (*init_function)(int *error); + +if (!psp || !psp->sev_data) +return -ENODEV ; + + sev = psp->sev_data; + +if (sev->state ==SEV_STATE_INIT ) +return 0 ; + + init_function = sev_init_ex_nv_address ? __sev_init_ex_locked : + __sev_init_locked; + rc =init_function (error);if (rc && *error ==SEV_RET_SECURE_DATA_INVALID ) {/* * INIT command returned an integrity check failure @@ -286,8 +425,8 @@ static int __sev_platform_init_locked(int *error)* failed and persistent state has been erased. * Retrying INIT command here should succeed. */ -dev_dbg (sev->dev,"SEV: retrying INIT command" ); - rc =__sev_do_cmd_locked (SEV_CMD_INIT , &data, error); +dev_notice (sev->dev,"SEV: retrying INIT command" ); + rc =init_function (error); }if (rc)@@ -303,7 +442,7 @@ static int __sev_platform_init_locked(int *error)dev_dbg (sev->dev,"SEV firmware initialized\n" ); -return rc; +return 0 ; }int sev_platform_init (int *error)@@ -1057,6 +1196,12 @@ static void sev_firmware_shutdown(struct sev_device *sev) get_order(SEV_ES_TMR_SIZE)); sev_es_tmr =NULL ; } + +if (sev_init_ex_nv_address) { +free_pages ((unsigned long )sev_init_ex_nv_address, +get_order (NV_LENGTH )); + sev_init_ex_nv_address =NULL ; + } }void sev_dev_destroy (struct psp_device *psp)@@ -1101,11 +1246,23 @@ void sev_pci_init(void) sev_update_firmware(sev->dev) == 0)
drivers/crypto/ccp/sev-dev.c void sev_pci_init(void) · line 1256
Quoted 10 lines
drivers/crypto/ccp/sev-dev.c +8
-0sev_get_api_version (); +/* If an init_ex_path is provided rely on INIT_EX for PSP initialization +* instead of INIT. +*/ +if (init_ex_path) { + sev_init_ex_nv_address =sev_fw_alloc (NV_LENGTH ); +if (!sev_init_ex_nv_address) { +dev_err (sev->dev, +"SEV: INIT_EX NV storage allocation failed, INIT-EX support unavailable\n" );
Since this is a file, maybe s/storage/memory/ ?
At this point, SEV will fail to initialize, it won't fall back to the INIT support. So I think the ", INIT-EX support unavailable" can be removed from the message.
drivers/crypto/ccp/sev-dev.c void sev_pci_init(void) · line 1265
Quoted 10 lines
drivers/crypto/ccp/sev-dev.c +5
-1+ goto err; + } + } +/* Obtain the TMR memory area for SEV-ES use */ sev_es_tmr =sev_fw_alloc (SEV_ES_TMR_SIZE );if (!sev_es_tmr)dev_warn (sev->dev, -"SEV: TMR allocation failed, SEV-ES support unavailable\n" ); +"SEV: TMR allocation failed\n" );
This message shouldn't be changed.
I should have made some of these comments on the first version, sorry about that.
Thanks,
Tom
Quoted 43 lines
drivers/crypto/ccp/sev-dev.c +0
-0/* Initialize the platform */ rc =sev_platform_init (&error);include/linux/psp-sev.h +21
-0@@ -52,6 +52,7 @@ enum sev_cmd {SEV_CMD_DF_FLUSH =0x00A ,SEV_CMD_DOWNLOAD_FIRMWARE =0x00B ,SEV_CMD_GET_ID =0x00C , +SEV_CMD_INIT_EX =0x00D ,/* Guest commands */ SEV_CMD_DECOMMISSION = 0x020,@@ -102,6 +103,26 @@ struct sev_data_init {u32 tmr_len;/* In */ } __packed; +/** +* struct sev_data_init_ex - INIT_EX command parameters +* +* @length: len of the command buffer read by the PSP +* @flags: processing flags +* @tmr_address: system physical address used for SEV-ES +* @tmr_len: len of tmr_address +* @nv_address: system physical address used for PSP NV storage +* @nv_len: len of nv_address +*/ +struct sev_data_init_ex { +u32 length;/* In */ +u32 flags;/* In */ +u64 tmr_address;/* In */ +u32 tmr_len;/* In */ +u32 reserved;/* In */ +u64 nv_address;/* In/Out */ +u32 nv_len;/* In */ +} __packed; +#define SEV_INIT_FLAGS_SEV_ES 0x01/**
Re: [PATCH V2 4/4] crypto: ccp - Add SEV_INIT_EX support
Quoted 62 lines · Tom Lendacky
Quoted 40 lines · Peter Gonda
From: David Rientjes <[email protected]>
Add new module parameter to allow users to use
SEV_INIT_EXinstead ofSEV_INIT. This helps users who lock their SPI bus to use the PSP for SEV functionality. The 'init_ex_path' parameter defaults to NULL which means the kernel will useSEV_INIT, if a path is specifiedSEV_INIT_EXwill be used with the data found at the path. On certain PSP commands this file is written to as the PSP updates the NV memory region. Depending on file system initialization this file open may fail during module init but the CCP driver for SEV already has sufficient retries for platform initialization. During normal operation of PSP system and SEV commands if the PSP has not been initialized it is at run time. If the file at 'init_ex_path' does not exist the PSP will not be initialized. The user must create the file prior to use with 32Kb of 0xFFs per spec.Signed-off-by David Rientjes <[email protected]>Co-developed-by Peter Gonda <[email protected]>Signed-off-by Peter Gonda <[email protected]>Reviewed-by Marc Orr <[email protected]>Cc Tom Lendacky <[email protected]>Cc Brijesh Singh <[email protected]>Cc Marc Orr <[email protected]>Cc Joerg Roedel <[email protected]>Cc Herbert Xu <[email protected]>Cc David Rientjes <[email protected]>Cc John Allen <[email protected]>Cc "David S. Miller" <[email protected]>--- .../virt/kvm/amd-memory-encryption.rst | 4 + drivers/crypto/ccp/sev-dev.c | 185 ++++++++++++++++-- include/linux/psp-sev.h | 21 ++ 3 files changed, 196 insertions(+), 14 deletions(-)Documentation/virt/kvm/amd-memory-encryption.rst +0
-0Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 86
Quoted 3 lines
Documentation/virt/kvm/amd-memory-encryption.rst +0
-0@@ -84,6 +84,10 @@ guests, such as launching, running, snapshotting, migrating and decommissioning. The KVM_SEV_INIT command is used by the hypervisor to initialize the SEV platform context. In a typical workflow, this command should be the first command issued.Add a blank line here.
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 87
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+The AMD-SP can be initialized either by using its own non-volatile storage orJust to be consistent within the document, s/AMD-SP/firmware/
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 88
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+the system can manage the NV storage for the AMD-SP using the module parameters/system/OS/
s/AMD-SP/firmware/Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 89
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+``init_ex_path``. This file must exist, to create a new NV storage file allocates/This file must exist/The file specified by ``init_ex_path`` must exist./ s/, to create/ To create/
Quoted 6 lines
Documentation/virt/kvm/amd-memory-encryption.rst guests, such as launching, running, snapshotting, migrating and decommissioning. · line 90
Quoted 1 line
Documentation/virt/kvm/amd-memory-encryption.rst +1
-0+a the file with 32Kb of 0xFF as required by the SEV FW spec.s/a the/the/
s/32Kb/32KB bytes/Just to be consistent within the document, s/SEV FW spec/SEV spec/
Updated documentation.
Quoted 54 lines
Quoted 54 lines
Documentation/virt/kvm/amd-memory-encryption.rst +0
-0Returns: 0 on success, -negative on errordrivers/crypto/ccp/sev-dev.c +14
-0@@ -22,6 +22,7 @@#include <linux/firmware.h> #include <linux/gfp.h> #include <linux/cpufeature.h> +#include <linux/fs.h> #include <asm/smp.h> @@ -43,6 +44,10 @@ static int psp_probe_timeout = 5;module_param (psp_probe_timeout, int,0644 );MODULE_PARM_DESC (psp_probe_timeout," default timeout value, in seconds, during PSP device probe" ); +static char *init_ex_path; +module_param (init_ex_path, charp,0660 ); +MODULE_PARM_DESC (init_ex_path," Path for INIT_EX data; if set try INIT_EX" ); +MODULE_FIRMWARE ("amd/amd_sev_fam17h_model0xh.sbin" );/* 1st gen EPYC */ MODULE_FIRMWARE ("amd/amd_sev_fam17h_model3xh.sbin" );/* 2nd gen EPYC */ MODULE_FIRMWARE ("amd/amd_sev_fam19h_model0xh.sbin" );/* 3rd gen EPYC */ @@ -58,6 +63,14 @@ static int psp_timeout;#define SEV_ES_TMR_SIZE (1024 * 1024)static void *sev_es_tmr; +/* INIT_EX NV Storage: +* The NV Storage is a 32Kb area and must be 4Kb page aligned. Use the page +* allocator to allocate the memory, which will return aligned memory for the +* specified allocation order. +*/ +#define NV_LENGTH (32 * 1024) +static void *sev_init_ex_nv_address; +static inline bool sev_version_greater_or_equal (u8 maj,u8 min) {struct sev_device *sev = psp_master->sev_data;@@ -107,6 +120,7 @@ static int sev_cmd_buffer_len(int cmd) {switch (cmd) {case SEV_CMD_INIT :return sizeof (struct sev_data_init ); +case SEV_CMD_INIT_EX :return sizeof (struct sev_data_init_ex );case SEV_CMD_PLATFORM_STATUS :return sizeof (struct sev_user_data_status );case SEV_CMD_PEK_CSR :return sizeof (struct sev_data_pek_csr );case SEV_CMD_PEK_CERT_IMPORT :return sizeof (struct sev_data_pek_cert_import );@@ -152,6 +166,89 @@ static void *sev_fw_alloc(unsigned long len)return page_address (page); }
Quoted 13 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 179
Quoted 11 lines
drivers/crypto/ccp/sev-dev.c +11
-0+static int sev_read_nv_memory (void ) +{ +struct file *fp; +ssize_t nread; + +if (!sev_init_ex_nv_address) +return -EOPNOTSUPP ; + + fp =filp_open (init_ex_path,O_RDONLY ,0 ); +if (IS_ERR (fp)) { +const int ret =PTR_ERR (fp);I don't think you need the "const" here.
Sounds good, removed. I normally default to consting a variable if I don't expect/want it to change. What guidance should I be following here?
Quoted 19 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 183
Quoted 5 lines
drivers/crypto/ccp/sev-dev.c +4
-0+ +dev_err (psp_master->dev, +"sev could not open file for read, error %d\n" , + ret);Maybe "SEV: could not open %s for read, ret=%d\n"
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 187
Quoted 4 lines
drivers/crypto/ccp/sev-dev.c +4
-0+return ret; + } + + nread =kernel_read (fp, sev_init_ex_nv_address,NV_LENGTH ,NULL );
kernel_readcan return an error, you should check nread before continuing.drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 188
Quoted 1 line
drivers/crypto/ccp/sev-dev.c +1
-0+dev_dbg (psp_master->dev,"sev NV read %d bytes\n" , nread);"SEV: read %ld bytes from NV file\n"
Quoted 10 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 194
Quoted 6 lines
drivers/crypto/ccp/sev-dev.c +6
-0+filp_close (fp,NULL ); + +return 0 ; +} + +static int sev_write_nv_memory (void )The return code is never checked by the caller. Is it expected that any error is not supposed to stop SEV for the current boot? Should you make this void, then?
Changed to void.
Quoted 16 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 206
Quoted 13 lines
drivers/crypto/ccp/sev-dev.c +12
-0+{ +struct sev_device *sev = psp_master->sev_data; +struct file *fp; +loff_t offset =0 ; +int ret; + +if (!sev_init_ex_nv_address) +return -EOPNOTSUPP ; + + fp =filp_open (init_ex_path,O_CREAT |O_WRONLY ,0600 ); +if (IS_ERR (fp)) { +dev_err (sev->dev,"sev NV data could not be created\n" );You should output a message similar to what is in
sev_read_nv_memory().
Quoted 7 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 210
Quoted 4 lines
drivers/crypto/ccp/sev-dev.c +4
-0+return PTR_ERR (fp); + } + + ret =kernel_write (fp, sev_init_ex_nv_address,NV_LENGTH , &offset);
kernel_writereturns assize_t, but ret is an int. and maybe use nwrite similar to nread insev_read_nv_memory()?
Done.
Quoted 10 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 216
Quoted 7 lines
drivers/crypto/ccp/sev-dev.c +6
-0+vfs_fsync (fp,0 ); +filp_close (fp,NULL ); + +if (ret !=NV_LENGTH ) { + dev_err(sev->dev, + "failed to write %d bytes to non volatile memory area, ret=%lu\n","SEV: failed to write %u bytes to NV file, ret=%ld\n"
Quoted 8 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 223
Quoted 6 lines
drivers/crypto/ccp/sev-dev.c +6
-0+ NV_LENGTH, ret); > + if (ret >= 0) + return -EIO; +return ret; + } + +dev_dbg (sev->dev,"wrote to non volatile memory area\n" );"SEV: write successful to NV file\n"
Updated all messages. Should have noted the "SEV: .." format.
Quoted 159 lines
Quoted 159 lines
drivers/crypto/ccp/sev-dev.c +89
-13+ +return 0 ; +} + +static void sev_write_nv_memory_if_required (int cmd_id) +{ +if (!sev_init_ex_nv_address) +return ; + +/* +* Only a few platform commands modify the SPI/NV area, but none of the +* non-platform commands do. Only INIT(_EX), PLATFORM_RESET, PEK_GEN, +* PEK_CERT_IMPORT, and PDH_GEN do. +*/ +switch (cmd_id) { +case SEV_CMD_FACTORY_RESET : +case SEV_CMD_INIT_EX : +case SEV_CMD_PDH_GEN : +case SEV_CMD_PEK_CERT_IMPORT : +case SEV_CMD_PEK_GEN : +break ; +default : +return ; + }; + +sev_write_nv_memory (); +} +static int __sev_do_cmd_locked (int cmd,void *data,int *psp_ret) {struct psp_device *psp = psp_master;@@ -221,6 +318,8 @@ static int __sev_do_cmd_locked(int cmd, void *data, int *psp_ret)dev_dbg (sev->dev,"sev command %#x failed (%#010x)\n" , cmd, reg &PSP_CMDRESP_ERR_MASK ); ret = -EIO ; + }else { +sev_write_nv_memory_if_required (cmd); } print_hex_dump_debug("(out): ", DUMP_PREFIX_OFFSET, 16, 2, data,@@ -247,22 +346,42 @@ static int sev_do_cmd(int cmd, void *data, int *psp_ret)return rc; } -static int __sev_platform_init_locked (int *error) +static int __sev_init_locked (int *error) { -struct psp_device *psp = psp_master;struct sev_data_init data; -struct sev_device *sev; -int rc =0 ; -if (!psp || !psp->sev_data) -return -ENODEV ; +memset (&data,0 ,sizeof (data)); +if (sev_es_tmr) { +u64 tmr_pa; - sev = psp->sev_data; +/* +* Do not include the encryption mask on the physical +* address of the TMR (firmware should clear it anyway). +*/ + tmr_pa =__pa (sev_es_tmr); -if (sev->state ==SEV_STATE_INIT ) -return 0 ; + data.flags |=SEV_INIT_FLAGS_SEV_ES ; + data.tmr_address = tmr_pa; + data.tmr_len =SEV_ES_TMR_SIZE ; + } + +return __sev_do_cmd_locked (SEV_CMD_INIT , &data, error); +} + +static int __sev_init_ex_locked (int *error) +{ +struct sev_data_init_ex data; +int ret;memset (&data,0 ,sizeof (data)); + data.length =sizeof (data); + data.nv_address =__psp_pa (sev_init_ex_nv_address); + data.nv_len =NV_LENGTH ; + + ret =sev_read_nv_memory (); +if (ret) +return ret; +if (sev_es_tmr) {u64 tmr_pa;@@ -277,7 +396,27 @@ static int __sev_platform_init_locked(int *error) data.tmr_len =SEV_ES_TMR_SIZE ; } - rc =__sev_do_cmd_locked (SEV_CMD_INIT , &data, error); +return __sev_do_cmd_locked (SEV_CMD_INIT_EX , &data, error); +} + +static int __sev_platform_init_locked (int *error) +{ +struct psp_device *psp = psp_master; +struct sev_device *sev; +int rc; +int (*init_function)(int *error); + +if (!psp || !psp->sev_data) +return -ENODEV ; + + sev = psp->sev_data; + +if (sev->state ==SEV_STATE_INIT ) +return 0 ; + + init_function = sev_init_ex_nv_address ? __sev_init_ex_locked : + __sev_init_locked; + rc =init_function (error);if (rc && *error ==SEV_RET_SECURE_DATA_INVALID ) {/* * INIT command returned an integrity check failure @@ -286,8 +425,8 @@ static int __sev_platform_init_locked(int *error)* failed and persistent state has been erased. * Retrying INIT command here should succeed. */ -dev_dbg (sev->dev,"SEV: retrying INIT command" ); - rc =__sev_do_cmd_locked (SEV_CMD_INIT , &data, error); +dev_notice (sev->dev,"SEV: retrying INIT command" ); + rc =init_function (error); }if (rc)@@ -303,7 +442,7 @@ static int __sev_platform_init_locked(int *error)dev_dbg (sev->dev,"SEV firmware initialized\n" ); -return rc; +return 0 ; }int sev_platform_init (int *error)@@ -1057,6 +1196,12 @@ static void sev_firmware_shutdown(struct sev_device *sev) get_order(SEV_ES_TMR_SIZE)); sev_es_tmr =NULL ; } + +if (sev_init_ex_nv_address) { +free_pages ((unsigned long )sev_init_ex_nv_address, +get_order (NV_LENGTH )); + sev_init_ex_nv_address =NULL ; + } }void sev_dev_destroy (struct psp_device *psp)@@ -1101,11 +1246,23 @@ void sev_pci_init(void) sev_update_firmware(sev->dev) == 0)sev_get_api_version ();
Quoted 10 lines
drivers/crypto/ccp/sev-dev.c void sev_pci_init(void) · line 1256
Quoted 8 lines
drivers/crypto/ccp/sev-dev.c +8
-0+/* If an init_ex_path is provided rely on INIT_EX for PSP initialization +* instead of INIT. +*/ +if (init_ex_path) { + sev_init_ex_nv_address =sev_fw_alloc (NV_LENGTH ); +if (!sev_init_ex_nv_address) { +dev_err (sev->dev, +"SEV: INIT_EX NV storage allocation failed, INIT-EX support unavailable\n" );Since this is a file, maybe s/storage/memory/ ?
Done.
Quoted 16 lines
At this point, SEV will fail to initialize, it won't fall back to the INIT support. So I think the ", INIT-EX support unavailable" can be removed from the message.
drivers/crypto/ccp/sev-dev.c void sev_pci_init(void) · line 1265
Quoted 10 lines
drivers/crypto/ccp/sev-dev.c +5
-1+ goto err; + } + } +/* Obtain the TMR memory area for SEV-ES use */ sev_es_tmr =sev_fw_alloc (SEV_ES_TMR_SIZE );if (!sev_es_tmr)dev_warn (sev->dev, -"SEV: TMR allocation failed, SEV-ES support unavailable\n" ); +"SEV: TMR allocation failed\n" );
Quoted 1 line
This message shouldn't be changed.
Oops I deleted the wrong "... support unavailable". Fixed.
Quoted 4 lines
I should have made some of these comments on the first version, sorry about that.
No worries! I appreciate the reviews.
Quoted 47 lines
Thanks,
TomQuoted 43 lines
drivers/crypto/ccp/sev-dev.c +0
-0/* Initialize the platform */ rc =sev_platform_init (&error);include/linux/psp-sev.h +21
-0@@ -52,6 +52,7 @@ enum sev_cmd {SEV_CMD_DF_FLUSH =0x00A ,SEV_CMD_DOWNLOAD_FIRMWARE =0x00B ,SEV_CMD_GET_ID =0x00C , +SEV_CMD_INIT_EX =0x00D ,/* Guest commands */ SEV_CMD_DECOMMISSION = 0x020,@@ -102,6 +103,26 @@ struct sev_data_init {u32 tmr_len;/* In */ } __packed; +/** +* struct sev_data_init_ex - INIT_EX command parameters +* +* @length: len of the command buffer read by the PSP +* @flags: processing flags +* @tmr_address: system physical address used for SEV-ES +* @tmr_len: len of tmr_address +* @nv_address: system physical address used for PSP NV storage +* @nv_len: len of nv_address +*/ +struct sev_data_init_ex { +u32 length;/* In */ +u32 flags;/* In */ +u64 tmr_address;/* In */ +u32 tmr_len;/* In */ +u32 reserved;/* In */ +u64 nv_address;/* In/Out */ +u32 nv_len;/* In */ +} __packed; +#define SEV_INIT_FLAGS_SEV_ES 0x01/**
Tom Lendacky
maintainerRe: [PATCH V2 4/4] crypto: ccp - Add SEV_INIT_EX support
Quoted 2 lines · Peter Gonda
Quoted 1 line · Tom Lendacky
On 11/1/21 12:21 PM, Peter Gonda wrote:
...
Quoted 10 lines
Quoted 6 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 179
Quoted 4 lines
drivers/crypto/ccp/sev-dev.c +4
-0+ + fp =filp_open (init_ex_path,O_RDONLY ,0 ); +if (IS_ERR (fp)) { +const int ret =PTR_ERR (fp);I don't think you need the "const" here.
Sounds good, removed. I normally default to consting a variable if I don't expect/want it to change. What guidance should I be following here?
Heh, I don't know... I guess since its in such a short scope it just read easier to me. But, you're correct, const is appropriate here, so I guess feel free to leave it in if you want.
Quoted 2 lines
Quoted 2 lines
...
Quoted 3 lines
Quoted 1 line
"SEV: write successful to NV file\n"
Updated all messages. Should have noted the "SEV: .." format.
It's not like we were very consistent originally, but it would be nice to have the new messages start to maintain a consistency.
Thanks,
Tom
Quoted 1 line
Re: [PATCH V2 4/4] crypto: ccp - Add SEV_INIT_EX support
Quoted 14 lines · Tom Lendacky
Quoted 2 lines · Peter Gonda
Quoted 1 line · Tom Lendacky
On 11/1/21 12:21 PM, Peter Gonda wrote:
...
Quoted 7 lines
Quoted 6 lines
drivers/crypto/ccp/sev-dev.c static void *sev_fw_alloc(unsigned long len) · line 179
Quoted 4 lines
drivers/crypto/ccp/sev-dev.c +4
-0+ + fp =filp_open (init_ex_path,O_RDONLY ,0 ); +if (IS_ERR (fp)) { +const int ret =PTR_ERR (fp);I don't think you need the "const" here.
Quoted 7 lines
Quoted 3 lines
Sounds good, removed. I normally default to consting a variable if I don't expect/want it to change. What guidance should I be following here?
Heh, I don't know... I guess since its in such a short scope it just read easier to me. But, you're correct, const is appropriate here, so I guess feel free to leave it in if you want.
Thanks! I'll remove it here to be more consistent with this file.
Quoted 17 lines
Quoted 3 lines
Quoted 3 lines
...
Quoted 3 lines
Quoted 1 line
"SEV: write successful to NV file\n"
Updated all messages. Should have noted the "SEV: .." format.
It's not like we were very consistent originally, but it would be nice to have the new messages start to maintain a consistency.
Thanks,
TomQuoted 1 line