add sysfs exports for TPM 2 PCR registers
Subsystems (2)
James Bottomley
maintainerv2 cover add sysfs exports for TPM 2 PCR registers
v2 to fix the failure on TPM 1.2
At last year's plumbers conference it was agreed in principle to export TPM 2 PCRs via sysfs. We also agreed we should conform to sysfs rules of one value per file, which rules out the "pcrs" file format of TPM 1.2 which has every PCR value in the same file.
I added these files using device groups, so one group per bank hash of the TPM. Using an emulator which supports a variety of hashes, you can see the structure of the group files:
root@testdeb:~# ls -F /sys/class/tpm/tpm0/ dev pcr-sha1/ pcr-sha384/ power/ tpm_version_major device@ pcr-sha256/ pcr-sha512/ subsystem@ uevent
As a future enhancement, we could use the group is_visible function to remove files corresponding to PCRs which don't exist. The reason this isn't present is that so far I've never seen a TPM with a missing PCR.
James
---
James Bottomley (1):
tpm: add sysfs exports for all banks of PCR registers
drivers/char/tpm/tpm-sysfs.c | 180 +++++++++++++++++++++++++++++++++++ include/linux/tpm.h | 8 +- 2 files changed, 187 insertions(+), 1 deletion(-)
Signature · 1 line
2.26.2
James Bottomley
maintainerPATCH v2 1/1 tpm: add sysfs exports for all banks of PCR registers
use macro magic to create sysfs per hash groups with 24 PCR files in them one for each possible agile hash of the TPM. The files are plugged in to a read function which is TPM version agnostic, so this works also for TPM 1.2 although the hash is only sha1 in that case. For every hash the TPM supports, a group named pcr-<hash> is created and each of the PCR read files placed under it.
---
v2: fix TPM 1.2 legacy links failure
--- drivers/char/tpm/tpm-sysfs.c | 180 +++++++++++++++++++++++++++++++++++ include/linux/tpm.h | 8 +- 2 files changed, 187 insertions(+), 1 deletion(-)
drivers/char/tpm/tpm-sysfs.c +180 -0
@@ -348,11 +348,191 @@ static const struct attribute_group tpm2_dev_group = { .attrs = tpm2_dev_attrs, }; +struct tpm_pcr_attr { +int alg_id; +int pcr; +struct device_attribute attr; +}; + +#define to_tpm_pcr_attr (a) container_of(a, struct tpm_pcr_attr, attr) + +static ssize_t pcr_value_show (struct device *dev, +struct device_attribute *attr, +char *buf) +{ +struct tpm_pcr_attr *ha =to_tpm_pcr_attr (attr); +struct tpm_chip *chip =to_tpm_chip (dev); +struct tpm_digest digest; +int i; +int digest_size =0 ; +int rc; +char *str = buf; + +for (i =0 ; i < chip->nr_allocated_banks; i++) +if (ha->alg_id == chip->allocated_banks[i].alg_id) + digest_size = chip->allocated_banks[i].digest_size; +/* should never happen */ +if (!digest_size) +return -EINVAL ; + + digest.alg_id = ha->alg_id; + rc =tpm_pcr_read (chip, ha->pcr, &digest); +if (rc) +return rc; +for (i =0 ; i < digest_size; i++) + str +=sprintf (str,"%02X" , digest.digest[i]); + str +=sprintf (str,"\n" ); + +return str - buf; +} + +/* +* The following set of defines represents all the magic to build +* the per hash attribute groups for displaying each bank of PCRs. +* The only slight problem with this approach is that every PCR is +* hard coded to be present, so you don't know if an PCR is missing +* until a cat of the file returns -EINVAL +* +* Also note you must ignore checkpatch warnings in this macro +* code. This is deep macro magic that checkpatch.pl doesn't +* understand. +*/ + +/* Note, this must match TPM2_PLATFORM_PCR which is fixed at 24. */ +#define _TPM_HELPER (_alg, _hash,F ) \ + F(_alg, _hash, 0) \ + F(_alg, _hash, 1) \ + F(_alg, _hash, 2) \ + F(_alg, _hash, 3) \ + F(_alg, _hash, 4) \ + F(_alg, _hash, 5) \ + F(_alg, _hash, 6) \ + F(_alg, _hash, 7) \ + F(_alg, _hash, 8) \ + F(_alg, _hash, 9) \ + F(_alg, _hash, 10) \ + F(_alg, _hash, 11) \ + F(_alg, _hash, 12) \ + F(_alg, _hash, 13) \ + F(_alg, _hash, 14) \ + F(_alg, _hash, 15) \ + F(_alg, _hash, 16) \ + F(_alg, _hash, 17) \ + F(_alg, _hash, 18) \ + F(_alg, _hash, 19) \ + F(_alg, _hash, 20) \ + F(_alg, _hash, 21) \ + F(_alg, _hash, 22) \ + F(_alg, _hash, 23) + +/* ignore checkpatch warning about trailing ; in macro. */ +#define PCR_ATTR (_alg, _hash, _pcr) \ + static struct tpm_pcr_attr dev_attr_pcr_##_hash##_##_pcr = { \ + .alg_id = _alg, \ + .pcr = _pcr, \ + .attr = { \ + .attr = { \ + .name = __stringify(_pcr), \ + .mode = 0444 \ + }, \ + .show = pcr_value_show \ + } \ + }; + +#define PCR_ATTRS (_alg, _hash) \ + _TPM_HELPER(_alg, _hash, PCR_ATTR) + +/* ignore checkpatch warning about trailing , in macro. */ +#define PCR_ATTR_VAL (_alg, _hash, _pcr) \ + &dev_attr_pcr_##_hash##_##_pcr.attr.attr, + +#define PCR_ATTR_GROUP_ARRAY (_alg, _hash) \ + static struct attribute *pcr_group_attrs_##_hash[] = { \ + _TPM_HELPER(_alg, _hash, PCR_ATTR_VAL) \ + NULL \ + } + +#define PCR_ATTR_GROUP (_alg, _hash) \ + static struct attribute_group pcr_group_##_hash = { \ + .name = "pcr-" __stringify(_hash), \ + .attrs = pcr_group_attrs_##_hash \ + } + +#define PCR_ATTR_BUILD (_alg, _hash) \ + PCR_ATTRS(_alg, _hash) \ + PCR_ATTR_GROUP_ARRAY(_alg, _hash); \ + PCR_ATTR_GROUP(_alg, _hash) +/* +* End of macro structure to build an attribute group containing 24 +* PCR value files for each supported hash algorithm +*/ + +/* +* The next set of macros implements the cleverness for each hash to +* build a static attribute group called pcr_group_<hash> which can be +* added to chip->groups[]. +* +* The first argument is the TPM algorithm id and the second is the +* hash used as both the suffix and the group name. Note: the group +* name is a directory in the top level tpm class with the name +* pcr-<hash>, so it must not clash with any other names already +* in the sysfs directory. +*/ +PCR_ATTR_BUILD (TPM_ALG_SHA1 , sha1); +PCR_ATTR_BUILD (TPM_ALG_SHA256 , sha256); +PCR_ATTR_BUILD (TPM_ALG_SHA384 , sha384); +PCR_ATTR_BUILD (TPM_ALG_SHA512 , sha512); +PCR_ATTR_BUILD (TPM_ALG_SM3_256 , sm3); + +void tpm_sysfs_add_device (struct tpm_chip *chip) { +int i; +WARN_ON (chip->groups_cnt !=0 ); +if (chip->flags &TPM_CHIP_FLAG_TPM2 ) chip->groups[chip->groups_cnt++] = &tpm2_dev_group;else chip->groups[chip->groups_cnt++] = &tpm1_dev_group; + +/* add one group for each bank hash */ +for (i =0 ; i < chip->nr_allocated_banks; i++) { +switch (chip->allocated_banks[i].alg_id) { +case TPM_ALG_SHA1 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha1; +break ; +case TPM_ALG_SHA256 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha256; +break ; +case TPM_ALG_SHA384 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha384; +break ; +case TPM_ALG_SHA512 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha512; +break ; +case TPM_ALG_SM3_256 : + chip->groups[chip->groups_cnt++] = &pcr_group_sm3; +break ; +default : +/* +* If this warning triggers, send a patch to +* add both a PCR_ATTR_BUILD() macro above for +* the missing algorithm as well as an +* additional case in this switch statement. +*/ +WARN (1 ,"TPM with unsupported bank algorthm 0x%04x" , + chip->allocated_banks[i].alg_id); +break ; + } + } + +/* +* This will only trigger if someone has added an additional +* hash to the tpm_algorithms enum without incrementing +* TPM_MAX_HASHES. This has to be a BUG_ON because under this +* condition, the chip->groups array will overflow corrupting +* the chips structure. +*/ +BUG_ON (chip->groups_cnt >TPM_MAX_HASHES ); }
include/linux/tpm.h +7 -1
@@ -42,6 +42,12 @@ enum tpm_algorithms { TPM_ALG_SM3_256 = 0x0012, }; +/* +* maximum number of hashing algorithms a TPM can have. This is +* basically a count of every hash in tpm_algorithms above +*/ +#define TPM_MAX_HASHES 5 +struct tpm_digest {u16 alg_id;u8 digest[TPM_MAX_DIGEST_SIZE ];
@@ -145,7 +151,7 @@ struct tpm_chip {struct dentry *bios_dir[TPM_NUM_EVENT_LOG_FILES ]; -const struct attribute_group *groups[3 ]; +const struct attribute_group *groups[3 +TPM_MAX_HASHES ];unsigned int groups_cnt;u32 nr_allocated_banks;
Signature · 1 line
2.26.2
Mimi Zohar
maintainerRe: [PATCH v2 1/1] tpm: add sysfs exports for all banks of PCR registers
Quoted 7 lines · James Bottomley
use macro magic to create sysfs per hash groups with 24 PCR files in them one for each possible agile hash of the TPM. The files are plugged in to a read function which is TPM version agnostic, so this works also for TPM 1.2 although the hash is only sha1 in that case. For every hash the TPM supports, a group named pcr-<hash> is created and each of the PCR read files placed under it.
Quoted 5 lines
Signed-off-by James Bottomley <[email protected]>---
v2: fix TPM 1.2 legacy links failure
Thanks, this version works with TPM 1.2.
Mimi
Re: [PATCH v2 1/1] tpm: add sysfs exports for all banks of PCR registers
In reply to James BottomleyJames Bottomley @ 2020-07-21 08:56 MST:
Quoted 202 lines
use macro magic to create sysfs per hash groups with 24 PCR files in them one for each possible agile hash of the TPM. The files are plugged in to a read function which is TPM version agnostic, so this works also for TPM 1.2 although the hash is only sha1 in that case. For every hash the TPM supports, a group named pcr-<hash> is created and each of the PCR read files placed under it.
Signed-off-by James Bottomley <[email protected]>---
v2: fix TPM 1.2 legacy links failure
--- drivers/char/tpm/tpm-sysfs.c | 180 +++++++++++++++++++++++++++++++++++ include/linux/tpm.h | 8 +- 2 files changed, 187 insertions(+), 1 deletion(-)drivers/char/tpm/tpm-sysfs.c +170
-0@@ -348,11 +348,191 @@ static const struct attribute_group tpm2_dev_group = { .attrs = tpm2_dev_attrs, }; +struct tpm_pcr_attr { +int alg_id; +int pcr; +struct device_attribute attr; +}; + +#define to_tpm_pcr_attr (a) container_of(a, struct tpm_pcr_attr, attr) + +static ssize_t pcr_value_show (struct device *dev, +struct device_attribute *attr, +char *buf) +{ +struct tpm_pcr_attr *ha =to_tpm_pcr_attr (attr); +struct tpm_chip *chip =to_tpm_chip (dev); +struct tpm_digest digest; +int i; +int digest_size =0 ; +int rc; +char *str = buf; + +for (i =0 ; i < chip->nr_allocated_banks; i++) +if (ha->alg_id == chip->allocated_banks[i].alg_id) + digest_size = chip->allocated_banks[i].digest_size; +/* should never happen */ +if (!digest_size) +return -EINVAL ; + + digest.alg_id = ha->alg_id; + rc =tpm_pcr_read (chip, ha->pcr, &digest); +if (rc) +return rc; +for (i =0 ; i < digest_size; i++) + str +=sprintf (str,"%02X" , digest.digest[i]); + str +=sprintf (str,"\n" ); + +return str - buf; +} + +/* +* The following set of defines represents all the magic to build +* the per hash attribute groups for displaying each bank of PCRs. +* The only slight problem with this approach is that every PCR is +* hard coded to be present, so you don't know if an PCR is missing +* until a cat of the file returns -EINVAL +* +* Also note you must ignore checkpatch warnings in this macro +* code. This is deep macro magic that checkpatch.pl doesn't +* understand. +*/ + +/* Note, this must match TPM2_PLATFORM_PCR which is fixed at 24. */ +#define _TPM_HELPER (_alg, _hash,F ) \ + F(_alg, _hash, 0) \ + F(_alg, _hash, 1) \ + F(_alg, _hash, 2) \ + F(_alg, _hash, 3) \ + F(_alg, _hash, 4) \ + F(_alg, _hash, 5) \ + F(_alg, _hash, 6) \ + F(_alg, _hash, 7) \ + F(_alg, _hash, 8) \ + F(_alg, _hash, 9) \ + F(_alg, _hash, 10) \ + F(_alg, _hash, 11) \ + F(_alg, _hash, 12) \ + F(_alg, _hash, 13) \ + F(_alg, _hash, 14) \ + F(_alg, _hash, 15) \ + F(_alg, _hash, 16) \ + F(_alg, _hash, 17) \ + F(_alg, _hash, 18) \ + F(_alg, _hash, 19) \ + F(_alg, _hash, 20) \ + F(_alg, _hash, 21) \ + F(_alg, _hash, 22) \ + F(_alg, _hash, 23) + +/* ignore checkpatch warning about trailing ; in macro. */ +#define PCR_ATTR (_alg, _hash, _pcr) \ + static struct tpm_pcr_attr dev_attr_pcr_##_hash##_##_pcr = { \ + .alg_id = _alg, \ + .pcr = _pcr, \ + .attr = { \ + .attr = { \ + .name = __stringify(_pcr), \ + .mode = 0444 \ + }, \ + .show = pcr_value_show \ + } \ + }; + +#define PCR_ATTRS (_alg, _hash) \ + _TPM_HELPER(_alg, _hash, PCR_ATTR) + +/* ignore checkpatch warning about trailing , in macro. */ +#define PCR_ATTR_VAL (_alg, _hash, _pcr) \ + &dev_attr_pcr_##_hash##_##_pcr.attr.attr, + +#define PCR_ATTR_GROUP_ARRAY (_alg, _hash) \ + static struct attribute *pcr_group_attrs_##_hash[] = { \ + _TPM_HELPER(_alg, _hash, PCR_ATTR_VAL) \ + NULL \ + } + +#define PCR_ATTR_GROUP (_alg, _hash) \ + static struct attribute_group pcr_group_##_hash = { \ + .name = "pcr-" __stringify(_hash), \ + .attrs = pcr_group_attrs_##_hash \ + } + +#define PCR_ATTR_BUILD (_alg, _hash) \ + PCR_ATTRS(_alg, _hash) \ + PCR_ATTR_GROUP_ARRAY(_alg, _hash); \ + PCR_ATTR_GROUP(_alg, _hash) +/* +* End of macro structure to build an attribute group containing 24 +* PCR value files for each supported hash algorithm +*/ + +/* +* The next set of macros implements the cleverness for each hash to +* build a static attribute group called pcr_group_<hash> which can be +* added to chip->groups[]. +* +* The first argument is the TPM algorithm id and the second is the +* hash used as both the suffix and the group name. Note: the group +* name is a directory in the top level tpm class with the name +* pcr-<hash>, so it must not clash with any other names already +* in the sysfs directory. +*/ +PCR_ATTR_BUILD (TPM_ALG_SHA1 , sha1); +PCR_ATTR_BUILD (TPM_ALG_SHA256 , sha256); +PCR_ATTR_BUILD (TPM_ALG_SHA384 , sha384); +PCR_ATTR_BUILD (TPM_ALG_SHA512 , sha512); +PCR_ATTR_BUILD (TPM_ALG_SM3_256 , sm3); + +void tpm_sysfs_add_device (struct tpm_chip *chip) { +int i; +WARN_ON (chip->groups_cnt !=0 ); +if (chip->flags &TPM_CHIP_FLAG_TPM2 ) chip->groups[chip->groups_cnt++] = &tpm2_dev_group;else chip->groups[chip->groups_cnt++] = &tpm1_dev_group; + +/* add one group for each bank hash */ +for (i =0 ; i < chip->nr_allocated_banks; i++) { +switch (chip->allocated_banks[i].alg_id) { +case TPM_ALG_SHA1 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha1; +break ; +case TPM_ALG_SHA256 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha256; +break ; +case TPM_ALG_SHA384 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha384; +break ; +case TPM_ALG_SHA512 : + chip->groups[chip->groups_cnt++] = &pcr_group_sha512; +break ; +case TPM_ALG_SM3_256 : + chip->groups[chip->groups_cnt++] = &pcr_group_sm3; +break ; +default : +/* +* If this warning triggers, send a patch to +* add both a PCR_ATTR_BUILD() macro above for +* the missing algorithm as well as an +* additional case in this switch statement. +*/ +WARN (1 ,"TPM with unsupported bank algorthm 0x%04x" , + chip->allocated_banks[i].alg_id); +break ; + }
drivers/char/tpm/tpm-sysfs.c static const struct attribute_group tpm2_dev_group = { · line 537
Quoted 10 lines
drivers/char/tpm/tpm-sysfs.c +10
-0+ } + +/* +* This will only trigger if someone has added an additional +* hash to the tpm_algorithms enum without incrementing +* TPM_MAX_HASHES. This has to be a BUG_ON because under this +* condition, the chip->groups array will overflow corrupting +* the chips structure. +*/ +BUG_ON (chip->groups_cnt >TPM_MAX_HASHES );
Should this check be 3 + TPM_MAX_HASHES like below?
Quoted 27 lines
drivers/char/tpm/tpm-sysfs.c +0
-0}include/linux/tpm.h +7
-1@@ -42,6 +42,12 @@ enum tpm_algorithms { TPM_ALG_SM3_256 = 0x0012, }; +/* +* maximum number of hashing algorithms a TPM can have. This is +* basically a count of every hash in tpm_algorithms above +*/ +#define TPM_MAX_HASHES 5 +struct tpm_digest {u16 alg_id;u8 digest[TPM_MAX_DIGEST_SIZE ];@@ -145,7 +151,7 @@ struct tpm_chip {struct dentry *bios_dir[TPM_NUM_EVENT_LOG_FILES ]; -const struct attribute_group *groups[3 ]; +const struct attribute_group *groups[3 +TPM_MAX_HASHES ];unsigned int groups_cnt;u32 nr_allocated_banks;
James Bottomley
maintainerRe: [PATCH v2 1/1] tpm: add sysfs exports for all banks of PCR registers
Quoted 1 line · Jerry Snitselaar
James Bottomley @ 2020-07-21 08:56 MST:
[...]
Quoted 5 lines
Quoted 5 lines
(quoted diff) +2
-0+ /* + * This will only trigger if someone has added anadditional
(quoted diff) +2
-0+ * hash to the tpm_algorithms enum without incrementing + * TPM_MAX_HASHES. This has to be a BUG_ON because under
Quoted 8 lines
drivers/char/tpm/tpm-sysfs.c static const struct attribute_group tpm2_dev_group = { · line 537
Quoted 6 lines
this
(quoted diff) +1
-0+ * condition, the chip->groups array will overflowcorrupting
drivers/char/tpm/tpm-sysfs.c +3
-0+* the chips structure. +*/ +BUG_ON (chip->groups_cnt >TPM_MAX_HASHES );Should this check be 3 +
TPM_MAX_HASHESlike below?
No, because at this point only a single additional group has been addedin addition to the hashes groups. The first line of tpm_sysfs_add_device is
WARN_ON(chip->groups_cnt != 0);
And then we add the unnamed group. This loop over the banks follows it, so chip->groups_cnt should be nr_banks_allocated by the end (it's the index, which is one fewer than the number of entries in chip-
Quoted 1 line
groups[]). We have a problem if
nr_banks_allocated>TPM_MAX_HASHES
which is what the BUG_ON checks.
James
Re: [PATCH v2 1/1] tpm: add sysfs exports for all banks of PCR registers
James Bottomley @ 2020-07-21 16:37 MST:
Quoted 23 lines · Jerry Snitselaar
Quoted 1 line
James Bottomley @ 2020-07-21 08:56 MST:
[...]
Quoted 5 lines
Quoted 5 lines
(quoted diff) +2
-0+ /* + * This will only trigger if someone has added anadditional
(quoted diff) +2
-0+ * hash to the tpm_algorithms enum without incrementing + * TPM_MAX_HASHES. This has to be a BUG_ON because underQuoted 8 lines
drivers/char/tpm/tpm-sysfs.c static const struct attribute_group tpm2_dev_group = { · line 537
Quoted 6 lines
this
(quoted diff) +1
-0+ * condition, the chip->groups array will overflowcorrupting
drivers/char/tpm/tpm-sysfs.c +3
-0+* the chips structure. +*/ +BUG_ON (chip->groups_cnt >TPM_MAX_HASHES );Should this check be 3 +
TPM_MAX_HASHESlike below?No, because at this point only a single additional group has been addedin addition to the hashes groups. The first line of
tpm_sysfs_add_deviceisWARN_ON(chip->groups_cnt != 0);
Quoted 7 lines
And then we add the unnamed group. This loop over the banks follows it, so
chip->groups_cntshould benr_banks_allocatedby the end (it's the index, which is one fewer than the number of entries in chip-Quoted 1 line
groups[]). We have a problem if
nr_banks_allocated>TPM_MAX_HASHESwhich is what the
BUG_ONchecks.James
If the chip supported all 5 listed cases wouldn't groups_cnt be 6 at this point?
James Bottomley
maintainerRe: [PATCH v2 1/1] tpm: add sysfs exports for all banks of PCR registers
Quoted 34 lines · Jerry Snitselaar
James Bottomley @ 2020-07-21 16:37 MST:
Quoted 32 lines · Jerry Snitselaar
Quoted 1 line
James Bottomley @ 2020-07-21 08:56 MST:
[...]
Quoted 7 lines
Quoted 7 lines
(quoted diff) +2
-0+ /* + * This will only trigger if someone has added anadditional
(quoted diff) +1
-0+ * hash to the tpm_algorithms enum withoutincrementing
(quoted diff) +1
-0+ * TPM_MAX_HASHES. This has to be a BUG_ON becauseunder
Quoted 8 lines
drivers/char/tpm/tpm-sysfs.c static const struct attribute_group tpm2_dev_group = { · line 537
Quoted 6 lines
this
(quoted diff) +1
-0+ * condition, the chip->groups array will overflowcorrupting
drivers/char/tpm/tpm-sysfs.c +3
-0+* the chips structure. +*/ +BUG_ON (chip->groups_cnt >TPM_MAX_HASHES );Should this check be 3 +
TPM_MAX_HASHESlike below?No, because at this point only a single additional group has been addedin addition to the hashes groups. The first line of
tpm_sysfs_add_deviceisWARN_ON(chip->groups_cnt != 0);And then we add the unnamed group. This loop over the banks follows it, so
chip->groups_cntshould benr_banks_allocatedby the end (it's the index, which is one fewer than the number of entries in chip->groups[]). We have a problem ifnr_banks_allocated>TPM_MAX_HASHES
Quoted 6 lines
Quoted 3 lines
which is what the
BUG_ONchecks.James
If the chip supported all 5 listed cases wouldn't
groups_cntbe 6 at this point?
Actually, yes, I think it would be because it's pointing at the next free index not the current one. So it should be BUG_ON (chip-
Quoted 1 line
groups_cnt>TPM_MAX_HASHES+ 1)
James
Jerry Snitselaar
Reviewed-byRe: [PATCH v2 1/1] tpm: add sysfs exports for all banks of PCR registers
James Bottomley @ 2020-07-21 17:39 MST:
Quoted 39 lines · Jerry Snitselaar
Quoted 38 lines
James Bottomley @ 2020-07-21 16:37 MST:
Quoted 35 lines · Jerry Snitselaar
Quoted 1 line
James Bottomley @ 2020-07-21 08:56 MST:
[...]
Quoted 7 lines
Quoted 7 lines
(quoted diff) +2
-0+ /* + * This will only trigger if someone has added anadditional
(quoted diff) +1
-0+ * hash to the tpm_algorithms enum withoutincrementing
(quoted diff) +1
-0+ * TPM_MAX_HASHES. This has to be a BUG_ON becauseunder
Quoted 8 lines
drivers/char/tpm/tpm-sysfs.c static const struct attribute_group tpm2_dev_group = { · line 537
Quoted 6 lines
this
(quoted diff) +1
-0+ * condition, the chip->groups array will overflowcorrupting
drivers/char/tpm/tpm-sysfs.c +3
-0+* the chips structure. +*/ +BUG_ON (chip->groups_cnt >TPM_MAX_HASHES );Should this check be 3 +
TPM_MAX_HASHESlike below?No, because at this point only a single additional group has been addedin addition to the hashes groups. The first line of
tpm_sysfs_add_deviceisWARN_ON(chip->groups_cnt != 0);And then we add the unnamed group. This loop over the banks follows it, so
chip->groups_cntshould benr_banks_allocatedby the end (it's the index, which is one fewer than the number of entries in chip->groups[]). We have a problem ifnr_banks_allocated>TPM_MAX_HASHESwhich is what the
BUG_ONchecks.James
Quoted 8 lines
Quoted 2 lines
If the chip supported all 5 listed cases wouldn't
groups_cntbe 6 at this point?Actually, yes, I think it would be because it's pointing at the next free index not the current one. So it should be
BUG_ON(chip-Quoted 1 line
groups_cnt>TPM_MAX_HASHES+ 1)James
One other thought, should a note be added above tpm_algorithms to note that when that is changed TPM_MAX_HASHES should be changed as well?
With the above change to the BUG_ON you can add to v3:
James Bottomley
maintainerRe: [PATCH v2 1/1] tpm: add sysfs exports for all banks of PCR registers
Quoted 47 lines · Jerry Snitselaar
James Bottomley @ 2020-07-21 17:39 MST:
Quoted 45 lines · Jerry Snitselaar
Quoted 35 lines
James Bottomley @ 2020-07-21 16:37 MST:
Quoted 33 lines · Jerry Snitselaar
Quoted 1 line
James Bottomley @ 2020-07-21 08:56 MST:
[...]
Quoted 8 lines
Quoted 8 lines
(quoted diff) +2
-0+ /* + * This will only trigger if someone has added anadditional
(quoted diff) +1
-0+ * hash to the tpm_algorithms enum withoutincrementing
(quoted diff) +1
-0+ * TPM_MAX_HASHES. This has to be a BUG_ONbecause
underQuoted 8 lines
drivers/char/tpm/tpm-sysfs.c static const struct attribute_group tpm2_dev_group = { · line 537
Quoted 6 lines
this
(quoted diff) +1
-0+ * condition, the chip->groups array will overflowcorrupting
drivers/char/tpm/tpm-sysfs.c +3
-0+* the chips structure. +*/ +BUG_ON (chip->groups_cnt >TPM_MAX_HASHES );Should this check be 3 +
TPM_MAX_HASHESlike below?No, because at this point only a single additional group has been addedin addition to the hashes groups. The first line of
tpm_sysfs_add_deviceisWARN_ON(chip->groups_cnt != 0);And then we add the unnamed group. This loop over the banks follows it, so
chip->groups_cntshould benr_banks_allocatedby the end (it's the index, which is one fewer than the number of entries in chip->groups[]). We have a problem ifnr_banks_allocated>TPM_MAX_HASHESQuoted 6 lines
Quoted 3 lines
which is what the
BUG_ONchecks.James
If the chip supported all 5 listed cases wouldn't
groups_cntbe 6 at this point?Actually, yes, I think it would be because it's pointing at the next free index not the current one. So it should be
BUG_ON(chip-
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
groups_cnt>TPM_MAX_HASHES+ 1)James
One other thought, should a note be added above
tpm_algorithmsto note that when that is changedTPM_MAX_HASHESshould be changed as well?
I certainly can ... it's free.
Quoted 1 line
With the above change to the
BUG_ONyou can add to v3:
OK, I also changed the BUG_ON back to a WARN_ON to match the initial one (if that one ever tripped, we'd get an overflow in the chip-
Quoted 1 line
groups[] as well, so it seems reasonable to keep them matching).
James
Quoted 2 lines
Reviewed-by Jerry Snitselaar <[email protected]>