integrity: IMA as an integrity service provider
Subsystems (4)
RFC PATCH 5/5 integrity: IMA as an integrity service provider
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM must_measure(), collect_measurement(), store_measurement(), and display_template() API calls. The store_measurement() call supports two types of data, IMA (i.e. file data) and generic template data.
When store_measurement() is called for the IMA type of data, the file measurement and the file name hint are used to form an IMA template. IMA then calculates the IMA template measurement(hash) and submits it to the TPM chip for inclusion in one of the chip's Platform Configuration Registers (PCR).
When store_measurement() is called for generic template data, IMA calculates the measurement(hash) of the template data, and submits the template measurement to the TPM chip for inclusion in one of the chip's Platform Configuration Registers(PCR).
In order to view the contents of template data through securityfs, the template_display() function must be defined in the registered template_operations. In the case of the IMA template, the list of file names and files hashes submitted can be viewed through securityfs.
IMA can be included or excluded in the kernel configuration. If included in the kernel and the IMA_BOOTPARAM is selected, IMA can also be enabled/disabled on the kernel command line with 'ima='.
---
security/integrity/ima/Kconfig new file +68 -0
@@ -0,0 +1,68 @@ +# +# IBM Integrity Measurement Architecture +# + +config IMA_MEASURE + bool "TCG run-time Integrity Measurement Architecture(IMA)" + depends on INTEGRITY + depends on ACPI + select CRYPTO + select CRYPTO_HMAC + select CRYPTO_MD5 + select CRYPTO_SHA1 + select TCG_TPM + help + IMA maintains a list of hash values of executables and + other sensitive system files loaded into the run-time + of this system. If your system has a TPM chip, then IMA + also maintains an aggregate integrity value over this + list inside the TPM hardware. These measurements and + the aggregate (signed inside the TPM) can be retrieved + and presented to remote parties to establish system + properties. If unsure, say N. + +config IMA_BOOTPARAM + bool "IMA boot parameter" + depends on IMA_MEASURE + default n + help + This option adds a kernel parameter 'ima', which allows IMA + to be disabled at boot. If this option is selected, IMA + functionality can be disabled with ima=0 on the kernel + command line. The purpose of this option is to allow a single + kernel image to be distributed with IMA built in, but not + necessarily enabled. + + If you are unsure how to answer this question, answer N. + +config IMA_BOOTPARAM_VALUE + int "IMA boot parameter default value" + depends on IMA_BOOTPARAM + range 0 1 + default 0 + help + This option sets the default value for the kernel parameter + 'ima=', which allows IMA to be disabled at boot. If this + option is set to 0 (zero), the IMA kernel parameter will + default to 0, disabling IMA at bootup. If this option is + set to 1 (one), the IMA kernel parameter will default to 1, + enabling IMA at bootup. + + If you are unsure how to answer this question, answer 0. + +config IMA_MEASURE_PCR_IDX + int "PCR for Aggregate (8<= Index <= 14)" + depends on IMA_MEASURE + range 8 14 + default 10 + help + IMA_MEASURE_PCR_IDX determines the TPM PCR register index + that IMA uses to maintain the integrity aggregate of the + measurement list. If unsure, use the default 10. + +config IMA_BASE_HOOKS + bool "IMA base hooks" + depends on IMA_MEASURE + default n + help + Enable this option to allow the LSM module to enforce integrity.
security/integrity/ima/Makefile new file +6 -0
@@ -0,0 +1,6 @@ + +obj-$(CONFIG_IMA_MEASURE) += ima.o + +ifeq ($(CONFIG_IMA_MEASURE), y) +ima-y := ima_fs.o ima_queue.o ima_init.o ima_main.o ima_crypto.o ima_api.o +endif
security/integrity/ima/ima.h new file +140 -0
140 lines not on this page · raw message
security/integrity/ima/ima_crypto.c new file +152 -0
152 lines not on this page · raw message
security/integrity/ima/ima_fs.c new file +437 -0
437 lines not on this page · raw message
security/integrity/ima/ima_init.c new file +103 -0
103 lines not on this page · raw message
security/integrity/ima/ima_main.c new file +415 -0
415 lines not on this page · raw message
security/integrity/ima/ima_queue.c new file +125 -0
125 lines not on this page · raw message
Documentation/kernel-parameters.txt +12 -0
@@ -44,6 +44,7 @@ parameter is applicable: FB The frame buffer device is enabled. HW Appropriate hardware is enabled. IA-64 IA-64 architecture is enabled. + IMA Integrity measurement architecture is enabled. IOSCHED More than one I/O scheduler is enabled. IP_PNP IP DHCP, BOOTP, or RARP is enabled. ISAPNP ISA PnP code is enabled.
@@ -804,6 +805,17 @@ and is between 256 and 4096 characters. ihash_entries= [KNL] Set number of hash buckets for inode cache. + ima= [IMA] Disable or enable IMA at boot time. + Format: { "0" | "1" } + See security/ima/Kconfig help text. + 0 -- disable. + 1 -- enable. + Default value is set via kernel config option. + + ima_hash= [IMA] runtime ability to define hash crypto alg. + Format: { "MD5" | "SHA1" } + Default is "SHA1". + in2000= [HW,SCSI] See header of drivers/scsi/in2000.c.
security/integrity/Kconfig +2 -3
@@ -2,8 +2,6 @@ # Integrity configuration # -menu "Integrity options" - config INTEGRITY bool "Enable different integrity models" help
@@ -21,4 +19,5 @@ config INTEGRITY_AUDIT allows integrity auditing to be disabled at boot. If this option is selected, integrity auditing can be disabled with 'integrity_audit=0' on the kernel command line. -endmenu + +source security/integrity/ima/Kconfig
security/integrity/Makefile +2 -0
@@ -5,3 +5,5 @@ # Object file lists obj-$(CONFIG_INTEGRITY) += integrity.o integrity_dummy.o \ integrity_audit.o integrity_policy.o + +obj-$(CONFIG_IMA_MEASURE) += ima/
include/linux/ima.h new file +48 -0
48 lines not on this page · raw message
security/integrity/ima/ima_api.c new file +362 -0
362 lines not on this page · raw message
security/integrity/ima/Kconfig line 6
Quoted 12 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar wrote:
---
Index: linux-2.6.26-rc3-git2/security/integrity/ima/Kconfig
===================================================================security/integrity/ima/Kconfig new file +6
-0@@ -0,0 +1,68 @@ +# +# IBM Integrity Measurement Architecture +# + +config IMA_MEASURE + bool "TCG run-time Integrity Measurement Architecture(IMA)"
What is TCG? It seems to be missing from all of those other TLAs.
Quoted 67 lines
security/integrity/ima/Kconfig new file +62
-0+ depends on INTEGRITY + depends on ACPI + select CRYPTO + select CRYPTO_HMAC + select CRYPTO_MD5 + select CRYPTO_SHA1 + select TCG_TPM + help + IMA maintains a list of hash values of executables and + other sensitive system files loaded into the run-time + of this system. If your system has a TPM chip, then IMA + also maintains an aggregate integrity value over this + list inside the TPM hardware. These measurements and + the aggregate (signed inside the TPM) can be retrieved + and presented to remote parties to establish system + properties. If unsure, say N. + +config IMA_BOOTPARAM + bool "IMA boot parameter" + depends on IMA_MEASURE + default n + help + This option adds a kernel parameter 'ima', which allows IMA + to be disabled at boot. If this option is selected, IMA + functionality can be disabled with ima=0 on the kernel + command line. The purpose of this option is to allow a single + kernel image to be distributed with IMA built in, but not + necessarily enabled. + + If you are unsure how to answer this question, answer N. + +config IMA_BOOTPARAM_VALUE + int "IMA boot parameter default value" + depends on IMA_BOOTPARAM + range 0 1 + default 0 + help + This option sets the default value for the kernel parameter + 'ima=', which allows IMA to be disabled at boot. If this + option is set to 0 (zero), the IMA kernel parameter will + default to 0, disabling IMA at bootup. If this option is + set to 1 (one), the IMA kernel parameter will default to 1, + enabling IMA at bootup. + + If you are unsure how to answer this question, answer 0. + +config IMA_MEASURE_PCR_IDX + int "PCR for Aggregate (8<= Index <= 14)" + depends on IMA_MEASURE + range 8 14 + default 10 + help + IMA_MEASURE_PCR_IDX determines the TPM PCR register index + that IMA uses to maintain the integrity aggregate of the + measurement list. If unsure, use the default 10. + +config IMA_BASE_HOOKS + bool "IMA base hooks" + depends on IMA_MEASURE + default n + help + Enable this option to allow the LSM module to enforce integrity.Index: linux-2.6.26-rc3-git2/security/integrity/ima/Makefile
===================================================================
security/integrity/ima/Makefile line 6
Quoted 6 lines
security/integrity/ima/Makefile new file +6
-0@@ -0,0 +1,6 @@ + +obj-$(CONFIG_IMA_MEASURE) += ima.o + +ifeq ($(CONFIG_IMA_MEASURE), y) +ima-y := ima_fs.o ima_queue.o ima_init.o ima_main.o ima_crypto.o ima_api.o +endif
Why are the ifeq and endif lines needed?
Quoted 5 lines
Index:
linux-2.6.26-rc3-git2/security/integrity/ima/ima_main.c
===================================================================security/integrity/ima/ima_main.c new file +0
-0@@ -0,0 +1,415 @@
...
Quoted 7 lines
+/**
+ *ima_alloc_integrity- allocate and attach an integrity structure
+ * associated with the inode.
+ * @inode: the inode structure
+ *
+ * Returns 0 on success,-ENOMEMon failure
+ */
kernel-doc notation first line (short description) cannot be continued on the following line(s).
Quoted 8 lines
security/integrity/ima/ima_main.c new file +8
-0+static int ima_inode_alloc_integrity (struct inode *inode) +{ +struct ima_iint_cache *iint; + + iint =ima_alloc_integrity (); +if (!iint) +return -ENOMEM ; +
security/integrity/ima/ima_main.c line 189
Quoted 10 lines
security/integrity/ima/ima_main.c new file +10
-0+ inode->i_integrity = iint; +timespec_set (&iint->mtime, &inode->i_mtime); +return 0 ; +} + +/** +* ima_inode_free_integrity - free the integrity structure associated with +* the inode. +* @inode: the inode structure +*/
ditto.
security/integrity/ima/ima_main.c line 201
Quoted 12 lines
security/integrity/ima/ima_main.c new file +12
-0+static void ima_inode_free_integrity (struct inode *inode) +{ +struct ima_iint_cache *iint = inode->i_integrity; + +if (iint) { + inode->i_integrity =NULL ; +kfree (iint); + } +} + +/** +* ima_inode_permission
Missing short function description.
Quoted 62 lines
+ * @inode: pointer to the inode to be measured
+ * @mask: containsMAY_READ,MAY_WRITE,MAY_APPENDorMAY_EXECUTE
+ * @nd: pointer to a nameidata
+ *
+ * Measure the file associated with the inode, if the
+ * file is open for read and the results of the call to
+ *ima_must_measure()require the file to be measured.
+ *
+ * Invalidate the PCR:
+ * - Opening a file for write when already open for read,
+ * results in a time of measure, time of use (ToMToU) error.
+ * - Opening a file for read when already open for write,
+ * could result in a file measurement error.
+ */security/integrity/ima/ima_main.c new file +48
-0+static int ima_inode_permission (struct inode *inode,int mask, +struct nameidata *nd) +{ +struct ima_data idata; +struct ima_args_data *data = &idata.data.args; + +memset (&idata,0 ,sizeof idata); +ima_fixup_argsdata (data, inode,NULL ,NULL , nd, mask,INODE_PERMISSION ); + +/* The file name is not required, but only a hint. */ +if (nd) + data->filename = (!nd->path.dentry->d_name.name) ? + (char *)nd->path.dentry->d_iname : + (char *)nd->path.dentry->d_name.name; + +/* Invalidate PCR, if a measured file is already open for read */ +if ((mask ==MAY_WRITE ) || (mask ==MAY_APPEND )) { +int mask_sav = data->mask; +int rc; + + data->mask =MAY_READ ; + rc =ima_must_measure (&idata); +if (!rc) { +if (atomic_read (&(data->dentry->d_count)) -1 > +atomic_read (&(inode->i_writecount))) +ima_add_violation (inode, data->filename, +"invalid_pcr" ,"ToMToU" ); + } + data->mask = mask_sav; + goto out; + } + +/* measure executables later */ +if (mask &MAY_READ ) { +int rc; + + rc =ima_must_measure (&idata); +if (!rc) { +/* Invalidate PCR, if a measured file is +* already open for write. +*/ +if (atomic_read (&(inode->i_writecount)) >0 ) +ima_add_violation (inode, data->filename, +"invalid_pcr" , +"open_writers" ); + + idata.type =IMA_DATA ; + rc =ima_collect_measurement (&idata);
security/integrity/ima/ima_main.c line 273
Quoted 10 lines
security/integrity/ima/ima_main.c new file +10
-0+if (!rc) +ima_store_measurement (&idata); + } + } +out: +return 0 ; +} + +/** +* ima_file_mmap
ditto.
Quoted 27 lines
+ * @inode: pointer to the inode to be measured
+ * @mask: containsMAY_READ,MAY_WRITE,MAY_APPENDorMAY_EXECUTE
+ * @nd: pointer to a nameidata
+ *
+ * Measure files being mmapped executable based on theima_must_measure()
+ * policy decision.
+ */security/integrity/ima/ima_main.c new file +20
-0+static int ima_file_mmap (struct file *file,unsigned long reqprot, +unsigned long prot,unsigned long flags, +unsigned long addr,unsigned long addr_only) +{ +struct ima_data idata; +struct ima_args_data *data = &idata.data.args; +int rc =0 ; + +if (!file || !file->f_dentry) +return rc; +if (!(prot &VM_EXEC )) +return rc; + +ima_fixup_argsdata (data,NULL ,NULL , file,NULL ,MAY_EXEC ,FILE_MMAP ); + data->filename = (file->f_dentry->d_name.name) ? + (char *)file->f_dentry->d_iname : + (char *)file->f_dentry->d_name.name; + + rc =ima_must_measure (&idata); +if (!rc) {
security/integrity/ima/ima_main.c line 310
Quoted 10 lines
security/integrity/ima/ima_main.c new file +10
-0+ idata.type =IMA_DATA ; + rc =ima_collect_measurement (&idata); +if (!rc) +ima_store_measurement (&idata); + } +return 0 ; +} + +/** +* ima_bprm_check_integrity
ditto.
Quoted 28 lines
+ * @bprm: contains the
linux_binprmstructure
+ *
+ * The OS protects against an executable file, already open for write,
+ * from being executed indeny_write_access()and an executable file,
+ * already open for execute, from being modified inget_write_access().
+ * So we can be certain that what we verify and measure here is actually
+ * what is being executed.
+ */security/integrity/ima/ima_main.c new file +20
-0+static int ima_bprm_check_integrity (struct linux_binprm *bprm) +{ +struct ima_data idata; +struct ima_args_data *data = &idata.data.args; +int rc =0 ; + +ima_fixup_argsdata (data,NULL ,NULL , bprm->file,NULL ,MAY_EXEC , +BPRM_CHECK ); + data->filename = bprm->filename; + + rc =ima_must_measure (&idata); +if (!rc) { + idata.type =IMA_DATA ; + rc =ima_collect_measurement (&idata); +if (!rc) +ima_store_measurement (&idata); + } +return 0 ; +} +
...
Quoted 35 lines
Index:
linux-2.6.26-rc3-git2/Documentation/kernel-parameters.txt
===================================================================Documentation/kernel-parameters.txt +12
-0@@ -44,6 +44,7 @@ parameter is applicable: FB The frame buffer device is enabled. HW Appropriate hardware is enabled. IA-64 IA-64 architecture is enabled. + IMA Integrity measurement architecture is enabled. IOSCHED More than one I/O scheduler is enabled. IP_PNP IP DHCP, BOOTP, or RARP is enabled. ISAPNP ISA PnP code is enabled.@@ -804,6 +805,17 @@ and is between 256 and 4096 characters. ihash_entries= [KNL] Set number of hash buckets for inode cache. + ima= [IMA] Disable or enable IMA at boot time. + Format: { "0" | "1" } + See security/ima/Kconfig help text. + 0 -- disable. + 1 -- enable. + Default value is set via kernel config option. + + ima_hash= [IMA] runtime ability to define hash crypto alg. + Format: { "MD5" | "SHA1" } + Default is "SHA1". + in2000= [HW,SCSI] See header of drivers/scsi/in2000.c.Index:
linux-2.6.26-rc3-git2/security/integrity/ima/ima_api.c
===================================================================security/integrity/ima/ima_api.c new file +0
-0@@ -0,0 +1,362 @@
...
Quoted 5 lines
+/**
+ *ima_store_inode_measure- create and store an inode template measurement
+ * @name:ascii file name associated with the measurement hash
+ * @hash_len:length of hash value in bytes (16 for MD5, 20 for SHA1)
+ * @hash:actual hash value pre-calculated
Please put a space after the ':'s above.
Quoted 24 lines
+ *
+ * Returns 0 on success, error code otherwise
+ */security/integrity/ima/ima_api.c new file +21
-0+static int ima_store_inode_measure (struct inode *inode, +const unsigned char *name, +int hash_len,char *hash,int violation) +{ +struct ima_inode_measure_entry measure_entry, *entry = &measure_entry; +int result; +int namelen; +char *op ="add_measure" ; +char *cause =" " ; + +memset (entry,0 ,sizeof *entry); +if (!violation) +memcpy (entry->digest, hash, hash_len >IMA_DIGEST_SIZE ? +IMA_DIGEST_SIZE : hash_len); +if (name) { + namelen =strlen (name); +memcpy (entry->file_name, name, namelen >IMA_EVENT_NAME_LEN_MAX + ?IMA_EVENT_NAME_LEN_MAX : namelen); + entry->file_name[namelen] ='\0' ; + } + result = ima_store_template_measure("ima", sizeof *entry, (char *)entry,
security/integrity/ima/ima_api.c line 211
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+ violation, &cause); +if (result <0 ) +integrity_audit_msg (AUDIT_INTEGRITY_PCR , inode, + name, op, cause, result); +return result; +} + +/** +* ima_add_violation - violations are flagged in the measurement list +* with zero hash values.
Short description must be on one line.
Quoted 14 lines
+ * @inode: inode associated with the violation
+ * @fname: name associated with the inode
+ * @op: string pointer to audit operation (i.e. "invalid_pcr", "add_measure")
+ * @cause: string pointer to reason for violation (i.e. "ToMToU")
+ *
+ * By extending the PCR with 0xFF's instead of with zeroes, the PCR
+ * value is invalidated.
+ */security/integrity/ima/ima_api.c new file +6
-0+void ima_add_violation (struct inode *inode,const unsigned char *fname, +char *op,char *cause) +{ +int result; + +/* can overflow, only indicator */
security/integrity/ima/ima_api.c line 235
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+atomic_inc (&ima_htable.violations); + + result =ima_store_inode_measure (inode, fname,0 ,NULL ,1 ); +integrity_audit_msg (AUDIT_INTEGRITY_PCR , inode, fname, op, + cause, result); +} + +/** +* skip_measurement - quick sanity check to make sure that only regular +* files opened for read-only or execute are measured.
Ditto.
Quoted 4 lines
+ * @inode: inode being measured
+ * @mask: contains the permission mask
+ *
+ * Return 1 to skip measure, 0 to measure
You can put as much text here (following the parameters) as you like/need.
Quoted 13 lines
+ */
security/integrity/ima/ima_api.c new file +12
-0+static int skip_measurement (struct inode *inode,int mask) +{ +/* skip pseudo/virtual devices */ +if ((inode->i_sb->s_magic ==PROC_SUPER_MAGIC ) + || (inode->i_sb->s_magic ==SYSFS_MAGIC ) + || (inode->i_sb->s_magic ==DEBUGFS_MAGIC ) + || (inode->i_sb->s_magic ==TMPFS_MAGIC ) + || (inode->i_sb->s_magic ==SECURITYFS_MAGIC ) + ||S_ISCHR (inode->i_mode) ||S_ISBLK (inode->i_mode)) +return 1 ;/* can't measure */ + +if (special_file (inode->i_mode) ||S_ISLNK (inode->i_mode))
security/integrity/ima/ima_api.c line 262
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+return 1 ;/* don't measure */ + +if (S_ISREG (inode->i_mode)) +return 0 ;/* measure */ +return 1 ;/* don't measure */ +} + +/** +* ima_must_measure - measure decision based on policy. +* @d - pointer to struct ima_data containing ima_args_data
* @d:
Quoted 17 lines
+ *
+ * The policy is defined in terms of keypairs: subj=, obj=, func=, mask=
+ * subj and obj: are LSM specific.
+ * func:INODE_PERMISSION|BPRM_CHECK|FILE_MMAP
+ * mask: contains the permission mask
+ *
+ * Return 0 to measure, error code otherwisesecurity/integrity/ima/ima_api.c new file +10
-0+*/ +int ima_must_measure (void *d) +{ +struct ima_data *idata = (struct ima_data *)d; +struct ima_args_data *data = &idata->data.args; + +if ((data->mask &MAY_WRITE ) || (data->mask &MAY_APPEND )) +return -EPERM ; + +if (skip_measurement (data->inode, data->mask))
security/integrity/ima/ima_api.c line 289
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+return -EPERM ; + +if (integrity_measure_policy (data->inode, data->function, data->mask)) +return 0 ; +return -EACCES ; +} + +/** +* ima_collect_measurement - collect an IMA measurement and store results +* in the inode
Short description on 1 line.
Quoted 1 line
+ * @d - pointer to struct
ima_data, containingima_args_data
* @d:
Quoted 24 lines
+ *
+ * Return 0 on success, error code otherwise
+ */security/integrity/ima/ima_api.c new file +21
-0+int ima_collect_measurement (void *d) +{ +struct ima_iint_cache *iint; +struct ima_data *idata = (struct ima_data *)d; +struct ima_args_data *data = &idata->data.args; +struct inode *inode = data->inode; +struct dentry *dentry = data->dentry; +struct nameidata *nd = data->nd; +struct file *file = data->file; +int result =0 ; + +if (!ima_enabled || idata->type !=IMA_DATA ) +return -EPERM ; + +if (!inode || !dentry) +return -EINVAL ; + + iint = inode->i_integrity; +mutex_lock (&iint->mutex); +if (!iint->measured) { +memset (iint->digest,0 ,IMA_DIGEST_SIZE );
security/integrity/ima/ima_api.c line 324
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+ result =ima_calc_hash (dentry, file, nd, iint->digest); + }else + result = -EEXIST ; +mutex_unlock (&iint->mutex); +return result; +} + +/** +* ima_store_measurement - either create and store an IMA template, +* or just store some other type of template measurement
ditto..
Quoted 1 line
+ * @d - pointer to struct
ima_data, containing eitherima_args_data, used
ditto...
Quoted 27 lines
+ * to create an IMA template, or a template.
+ */security/integrity/ima/ima_api.c new file +25
-0+void ima_store_measurement (void *d) +{ +struct ima_data *idata = (struct ima_data *)d; +int result; +char *op ="add_template_measure" ; +char *cause ="" ; + +if (idata->type ==IMA_DATA ) { +struct ima_args_data *data = &idata->data.args; +struct ima_iint_cache *iint; + + iint = data->inode->i_integrity; +mutex_lock (&iint->mutex); +if (iint->measured) { +mutex_unlock (&iint->mutex); +return ; + } + result =ima_store_inode_measure (data->inode, data->filename, +IMA_DIGEST_SIZE , iint->digest, +0 ); +if (!result || result == -EEXIST ) + iint->measured =1 ; +mutex_unlock (&iint->mutex); + }else if (idata->type ==IMA_TEMPLATE ) { + struct ima_store_data *template = (struct ima_store_data *)
security/integrity/ima/ima_api.c line 362
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+ &idata->data.template; + + result =ima_store_template_measure (template->name, + template->len, + template->data,0 , &cause); +if (result <0 ) +integrity_audit_msg (AUDIT_INTEGRITY_PCR ,NULL , + template->name, op, cause, result); + } +}
---
~Randy
Quoted 17 lines · Randy Dunlap
security/integrity/ima/Kconfig line 6
Quoted 12 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar wrote:
---
Index: linux-2.6.26-rc3-git2/security/integrity/ima/Kconfig
===================================================================security/integrity/ima/Kconfig new file +6
-0@@ -0,0 +1,68 @@ +# +# IBM Integrity Measurement Architecture +# + +config IMA_MEASURE + bool "TCG run-time Integrity Measurement Architecture(IMA)"What is TCG? It seems to be missing from all of those other TLAs.
TCG stands for Trusted Computing Group.
(https://www.trustedcomputinggroup.org/faq/)
Thank you for reviewing the code. I'll make the kernel doc, Kconfig and Makefile changes and repost.
Mimi
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM must_measure(), collect_measurement(), store_measurement(), and display_template() API calls. The store_measurement() call supports two types of data, IMA (i.e. file data) and generic template data.
When store_measurement() is called for the IMA type of data, the file measurement and the file name hint are used to form an IMA template. IMA then calculates the IMA template measurement(hash) and submits it to the TPM chip for inclusion in one of the chip's Platform Configuration Registers (PCR).
When store_measurement() is called for generic template data, IMA calculates the measurement(hash) of the template data, and submits the template measurement to the TPM chip for inclusion in one of the chip's Platform Configuration Registers(PCR).
In order to view the contents of template data through securityfs, the template_display() function must be defined in the registered template_operations. In the case of the IMA template, the list of file names and files hashes submitted can be viewed through securityfs.
IMA can be included or excluded in the kernel configuration. If included in the kernel and the IMA_BOOTPARAM is selected, IMA can also be enabled/disabled on the kernel command line with 'ima='.
---
security/integrity/ima/Kconfig new file +69 -0
@@ -0,0 +1,69 @@ +# +# IBM Integrity Measurement Architecture +# + +config IMA_MEASURE + bool "Integrity Measurement Architecture(IMA)" + depends on INTEGRITY + depends on ACPI + select CRYPTO + select CRYPTO_HMAC + select CRYPTO_MD5 + select CRYPTO_SHA1 + select TCG_TPM + help + The Trusted Computing Group(TCG) runtime Integrity + Measurement Architecture(IMA) maintains a list of hash + values of executables and other sensitive system files + loaded into the run-time of this system. If your system + has a TPM chip, then IMA also maintains an aggregate + integrity value over this list inside the TPM hardware. + These measurements and the aggregate (signed inside the + TPM) can be retrieved and presented to remote parties to + establish system properties. If unsure, say N. + +config IMA_BOOTPARAM + bool "IMA boot parameter" + depends on IMA_MEASURE + default n + help + This option adds a kernel parameter 'ima', which allows IMA + to be disabled at boot. If this option is selected, IMA + functionality can be disabled with ima=0 on the kernel + command line. The purpose of this option is to allow a single + kernel image to be distributed with IMA built in, but not + necessarily enabled. + + If you are unsure how to answer this question, answer N. + +config IMA_BOOTPARAM_VALUE + int "IMA boot parameter default value" + depends on IMA_BOOTPARAM + range 0 1 + default 0 + help + This option sets the default value for the kernel parameter + 'ima=', which allows IMA to be disabled at boot. If this + option is set to 0 (zero), the IMA kernel parameter will + default to 0, disabling IMA at bootup. If this option is + set to 1 (one), the IMA kernel parameter will default to 1, + enabling IMA at bootup. + + If you are unsure how to answer this question, answer 0. + +config IMA_MEASURE_PCR_IDX + int "PCR for Aggregate (8<= Index <= 14)" + depends on IMA_MEASURE + range 8 14 + default 10 + help + IMA_MEASURE_PCR_IDX determines the TPM PCR register index + that IMA uses to maintain the integrity aggregate of the + measurement list. If unsure, use the default 10. + +config IMA_BASE_HOOKS + bool "IMA base hooks" + depends on IMA_MEASURE + default n + help + Enable this option to allow the LSM module to enforce integrity.
security/integrity/ima/Makefile new file +8 -0
@@ -0,0 +1,8 @@ +# +# Makefile for building Trusted Computing Group's(TCG) runtime Integrity +# Measurement Architecture(IMA). +# + +obj-$(CONFIG_IMA_MEASURE) += ima.o + +ima-y := ima_fs.o ima_queue.o ima_init.o ima_main.o ima_crypto.o ima_api.o
security/integrity/ima/ima.h new file +140 -0
140 lines not on this page · raw message
security/integrity/ima/ima_crypto.c new file +152 -0
152 lines not on this page · raw message
security/integrity/ima/ima_fs.c new file +437 -0
437 lines not on this page · raw message
security/integrity/ima/ima_init.c new file +103 -0
103 lines not on this page · raw message
security/integrity/ima/ima_main.c new file +422 -0
422 lines not on this page · raw message
security/integrity/ima/ima_queue.c new file +125 -0
125 lines not on this page · raw message
Documentation/kernel-parameters.txt +12 -0
@@ -44,6 +44,7 @@ parameter is applicable: FB The frame buffer device is enabled. HW Appropriate hardware is enabled. IA-64 IA-64 architecture is enabled. + IMA Integrity measurement architecture is enabled. INTEGRITY Integrity support is enabled. IOSCHED More than one I/O scheduler is enabled. IP_PNP IP DHCP, BOOTP, or RARP is enabled.
@@ -805,6 +806,17 @@ and is between 256 and 4096 characters. ihash_entries= [KNL] Set number of hash buckets for inode cache. + ima= [IMA] Disable or enable IMA at boot time. + Format: { "0" | "1" } + See security/ima/Kconfig help text. + 0 -- disable. + 1 -- enable. + Default value is set via kernel config option. + + ima_hash= [IMA] runtime ability to define hash crypto alg. + Format: { "MD5" | "SHA1" } + Default is "SHA1". + in2000= [HW,SCSI] See header of drivers/scsi/in2000.c.
security/integrity/Kconfig +2 -3
@@ -2,8 +2,6 @@ # Integrity configuration # -menu "Integrity options" - config INTEGRITY bool "Enable different integrity models" help
@@ -21,4 +19,5 @@ config INTEGRITY_AUDIT allows integrity auditing to be disabled at boot. If this option is selected, integrity auditing can be disabled with 'integrity_audit=0' on the kernel command line. -endmenu + +source security/integrity/ima/Kconfig
security/integrity/Makefile +2 -0
@@ -5,3 +5,5 @@ # Object file lists obj-$(CONFIG_INTEGRITY) += integrity.o integrity_dummy.o \ integrity_audit.o integrity_policy.o + +obj-$(CONFIG_IMA_MEASURE) += ima/
include/linux/ima.h new file +48 -0
48 lines not on this page · raw message
security/integrity/ima/ima_api.c new file +365 -0
365 lines not on this page · raw message
Quoted 64 lines · On Tue, 27 May 2008 10:36:03 -0400 Mimi Zohar wrote:
Index:
linux-2.6.26-rc3-git2/security/integrity/ima/ima_api.c
===================================================================security/integrity/ima/ima_api.c new file +59
-0@@ -0,0 +1,365 @@ +/* +* Copyright (C) 2008 IBM Corporation +* +* Authors: +* Mimi Zohar <[email protected]> +* +* This program is free software; you can redistribute it and/or +* modify it under the terms of the GNU General Public License as +* published by the Free Software Foundation, version 2 of the +* License. +* +* File: ima_api.c +* - implements the LIM API +*/ +#include <linux/module.h> +#include <linux/integrity.h> +#include <linux/magic.h> +#include <linux/writeback.h> +#include <linux/string.h> +#include <linux/list.h> +#include <linux/audit.h> +#include <linux/ima.h> + +#include "ima.h" + +struct template_operations ima_template_ops = { + .must_measure = ima_must_measure, + .collect_measurement = ima_collect_measurement, + .store_measurement = ima_store_measurement, + .display_template = ima_template_show +}; + +#ifdef CONFIG_IMA_BOOTPARAM +static int ima_enabled =CONFIG_IMA_BOOTPARAM_VALUE ; + +static int __init ima_enabled_setup (char *str) +{ + + ima_enabled =simple_strtol (str,NULL ,0 ); +return 1 ; +} + +__setup ("ima=" , ima_enabled_setup); +#else +static int ima_enabled =1 ; +#endif + +/** +* mode_setup - for compatability with non-template IMA versions +* @str - is pointer to a string +*/ +int ima_template_mode =1 ; +static int __init mode_setup (char *str) +{ +if (strncmp (str,"ima" ,3 ) ==0 ) + ima_template_mode =0 ; +if (strncmp (str,"template" ,7 ) ==0 ) + ima_template_mode =1 ; + printk(KERN_INFO "%s: template_mode %s \n", __func__,
security/integrity/ima/ima_api.c line 69
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+ ima_template_mode ? "template" : "ima"); +return 1 ; +} + +__setup ("ima_mode=" , mode_setup); + +/** +* ima_digest_cpy - copy the hash in the IMA template structure to a digest +* @template_name: string containing the name of the template (i.e. "ima") +* @template - pointer to template structure
* @template:
security/integrity/ima/ima_api.c line 70
Quoted 1 line
security/integrity/ima/ima_api.c new file +1
-0+* @digest - pointer to the digest
* @digest:
Quoted 7 lines
security/integrity/ima/ima_api.c new file +7
-0+* +* Returns 0 on success, error code otherwise +*/ +static int ima_digest_cpy (char *template_name,void *template,u8 *digest) +{ +int rc, result =0 ; + struct ima_inode_measure_entry *inode_template =
security/integrity/ima/ima_api.c line 87
Quoted 10 lines
security/integrity/ima/ima_api.c new file +10
-0+ (struct ima_inode_measure_entry *)template; + + rc =strcmp (template_name,"ima" ); +if (rc ==0 ) +memcpy (digest, inode_template->digest, +sizeof inode_template->digest); +else + result = -ENODATA ; +return result; +}
Thanks.
---
~Randy
Quoted 17 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar <[email protected]> wrote:
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM
must_measure(),collect_measurement(),store_measurement(), anddisplay_template()API calls. Thestore_measurement()call supports two types of data, IMA (i.e. file data) and generic template data.When
store_measurement()is called for the IMA type of data, the file measurement and the file name hint are used to form an IMA template. IMA then calculates the IMA template measurement(hash) and submits it to the TPM chip for inclusion in one of the chip's Platform Configuration Registers (PCR).When
store_measurement()is called for generic template data, IMA calculates the measurement(hash) of the template data, and submits the template measurement to the TPM chip for inclusion in one of the chip's Platform Configuration Registers(PCR).
Quoted 9 lines
In order to view the contents of template data through securityfs, the
template_display()function must be defined in the registeredtemplate_operations. In the case of the IMA template, the list of file names and files hashes submitted can be viewed through securityfs.IMA can be included or excluded in the kernel configuration. If included in the kernel and the
IMA_BOOTPARAMis selected, IMA can also be enabled/disabled on the kernel command line with 'ima='.
- I see lots of user file I/O being done from within the kernel. This makes eyebrows raise. Also some other eyebrow-raising file-related things in there.
- A complicated-looking in-kernel string parser which is implementing an new and apparently-undocumented user->kernel ABI.
- Some GFP_ATOMICs which can hopefully become GFP_KERNEL.
- timespec_set() is unneeeded - just use struct assignment (ie: "=")
- timespec_recent() looks a bit hacky. The problems which are being solved here should be described in the changelog. Perhaps we can think of a better way, but first we have to know about it.
- shouldn't ima_inode_init() initialise tv_usec too?
- All the games with mtimes should be described in the changelog too.
- All the `static struct integrity_operations' instances could be made const. And lots of other foo_operations too, I expect.
That will lead to a constification chase all over the place, but it's probably for the best. This is after all a "security" feature and there is perhaps some benefit in getting your eminently hijackable function pointers into read-only memory.
- ima_fixup_inodes looks like it will race and crash against a well-timed unmount. I expect you will need to bump s_count before dropping sb_lock. See writeback_inodes() for an example.
- bug: ima_fixup_inodes() does a GFP_KERNEL allocation inside inode->i_lock. This bug shouldn't have got this far. Please always enable all kernel debugging features when testing code. Documentation/SubmitChecklist has useful things.
- inode.i_lock is defined as an innermost lock which is used for protecting data internal to the inode. You appear to be using it for way too much stuff in here.
- It would be useful to add a comment explaining why late_initcall(init_ima) is using late_initcall() rather than plain old module_init(). Because it is impossible for the reader to determine this information from the implementation.
- mutex_init(&ima_extend_list_mutex) is unneeded.
- Does ima_add_digest_entry() need to use the unreliable GFP_ATOMIC?
This matters. This is a security feature and if that kmalloc(GFP_ATOMIC) fails (as it easily can do) then I expect the system will either be insecure or will outright malfunction.
- Why does CONFIG_IMA_BOOTPARAM exist, and can it be removed (ie:
made unconditional)?
- Similarly CONFIG_IMA_BOOTPARAM_VALUE. Let's be decisive here - distributors only get one shot at setting these things.
- mode_setup() will identify itself as "mode_setup" in its KERN_INFO printk. That's a bit unhelpful. I'd suggest that all/most printks here be prefixed with "integrity:".
- GFP_ATOMICs everywhere :(
- As ima_htable.violations "can overflow", atomic_long_t might be a better choice of type.
- skip_measurement(): the hard-coded test for PROC_SUPER_MAGIC, SYSFS_MAGIC etc is quite unpleasant. Surely there is a better way.
-
+/**
+ * ima_must_measure - measure decision based on policy.
+ * @d - pointer to struct ima_data containing ima_args_data
So if we know the type of d, did we _have_ to make it void*? It's much better to use the C yype system if at all possible.
- ditto ima_collect_measurement()
Generally: the code is all moderately intrusive into the VFS and this sort of thing does need careful explanation and justification, please. Once we have some understanding of what you're trying to achieve here we will inevitably ask "can't that be done in userspace". So it would be best if your description were to preemptively answer all that.
Quoted 24 lines · Andrew Morton
Quoted 22 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar <[email protected]> wrote:
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM
must_measure(),collect_measurement(),store_measurement(), anddisplay_template()API calls. Thestore_measurement()call supports two types of data, IMA (i.e. file data) and generic template data.When
store_measurement()is called for the IMA type of data, the file measurement and the file name hint are used to form an IMA template. IMA then calculates the IMA template measurement(hash) and submits it to the TPM chip for inclusion in one of the chip's Platform Configuration Registers (PCR).When
store_measurement()is called for generic template data, IMA calculates the measurement(hash) of the template data, and submits the template measurement to the TPM chip for inclusion in one of the chip's Platform Configuration Registers(PCR).In order to view the contents of template data through securityfs, the
template_display()function must be defined in the registeredtemplate_operations. In the case of the IMA template, the list of file names and files hashes submitted can be viewed through securityfs.
Quoted 8 lines
Quoted 4 lines
IMA can be included or excluded in the kernel configuration. If included in the kernel and the
IMA_BOOTPARAMis selected, IMA can also be enabled/disabled on the kernel command line with 'ima='.- I see lots of user file I/O being done from within the kernel. This makes eyebrows raise. Also some other eyebrow-raising file-related things in there.
The amount of I/O is dependent on the number of files being measured. The default policy measures a whole lot. An LSM specific integrity policy would cut down on the number of files being measured. For now, either remove the third rule in default_rules or replace the default rules with a new policy. To load a new policy execute:
./integrity_load < policy
policy:
#
# Integrity measure policy
#
func=BPRM_CHECK
func=FILE_MMAP mask=MAY_EXEC
#func=INODE_PERMISSION mask=MAY_READ
integrity_load.c:
/*
* integrity_load.c
*
* Strip comments from integrity measurement policy file and load
* policy rules into the kernel by writing to security/ima/policy. *
*/
#include <sys/types.h>
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <unistd.h>
int main(int argc, char *argv[])
{
int fd;
char *policyname = "/sys/kernel/security/ima/policy";
char *line = NULL;
ssize_t len = 0, readcnt;
int rc = 0;
if (argc == 2)
policyname = argv[1];
fd = open(policyname, O_WRONLY);
if (fd == -1) {
perror("opening policy");
exit(EXIT_FAILURE);
}
while ((readcnt = getline(&line, &len, stdin)) != -1) {
if (*line == '#')
continue;
printf("%s", line);
rc = write(fd, line, readcnt);
if (rc == -1) {
perror("write");
break;
}
}
close(fd);
if (line)
free(line);
return (rc == -1) ? EXIT_FAILURE : EXIT_SUCCESS;
}Quoted 2 lines
- A complicated-looking in-kernel string parser which is implementing an new and apparently-undocumented
user->kernelABI.
oops, will document.
Quoted 1 line
- Some
GFP_ATOMICswhich can hopefully becomeGFP_KERNEL.
Ok.
Quoted 1 line
-
timespec_set()is unneeeded - just use struct assignment (ie: "=")
Am confused. timespec_set is doing an assignment. Should I replace timespec_set with a memcpy?
Quoted 3 lines
-
timespec_recent()looks a bit hacky. The problems which are being solved here should be described in the changelog. Perhaps we can think of a better way, but first we have to know about it.
Ok.
Quoted 1 line
- shouldn't
ima_inode_init()initialisetv_usectoo?
I don't think it matters. timespec_equal checks that the mtime in the iint and inode are the same. If they are the same then we don't need to recalculate the hash, unless it's been updated in really quick succession.
Quoted 1 line
- All the games with mtimes should be described in the changelog too.
Ok. The timespec_recent and mtime issues are part of the same problem of detecting when a file has been modified.
Quoted 7 lines
- All the `static struct
integrity_operations' instances could be made const. And lots of otherfoo_operationstoo, I expect.That will lead to a constification chase all over the place, but it's probably for the best. This is after all a "security" feature and there is perhaps some benefit in getting your eminently hijackable function pointers into read-only memory.
Ok.
Quoted 3 lines
-
ima_fixup_inodeslooks like it will race and crash against a well-timed unmount. I expect you will need to bumps_countbefore droppingsb_lock. Seewriteback_inodes()for an example.
ima_fixup_inodes() is called once at initialization.
Quoted 4 lines
- bug:
ima_fixup_inodes()does aGFP_KERNELallocation insideinode->i_lock. This bug shouldn't have got this far. Please always enable all kernel debugging features when testing code.Documentation/SubmitChecklisthas useful things.
sorry.
Quoted 10 lines
- inode.i_lock is defined as an innermost lock which is used for protecting data internal to the inode. You appear to be using it for way too much stuff in here.
- It would be useful to add a comment explaining why
late_initcall(init_ima)is usinglate_initcall()rather than plain oldmodule_init(). Because it is impossible for the reader to determine this information from the implementation.-
mutex_init(&ima_extend_list_mutex)is unneeded.
Ok.
Quoted 5 lines
- Does
ima_add_digest_entry()need to use the unreliableGFP_ATOMIC?This matters. This is a security feature and if that
kmalloc(GFP_ATOMIC)fails (as it easily can do) then I expect the system will either be insecure or will outright malfunction.
Ok.
Quoted 5 lines
- Why does
CONFIG_IMA_BOOTPARAMexist, and can it be removed (ie:
made unconditional)?- Similarly
CONFIG_IMA_BOOTPARAM_VALUE. Let's be decisive here - distributors only get one shot at setting these things.
Fine with me.
Quoted 3 lines
-
mode_setup()will identify itself as "mode_setup" in itsKERN_INFOprintk. That's a bit unhelpful. I'd suggest that all/most printks here be prefixed with "integrity:".
Yes, will do throughout.
Quoted 1 line
-
GFP_ATOMICseverywhere :(
Will address throughout.
Quoted 5 lines
- As ima_htable.violations "can overflow",
atomic_long_tmight be a better choice of type.-
skip_measurement(): the hard-coded test forPROC_SUPER_MAGIC,SYSFS_MAGICetc is quite unpleasant. Surely there is a better way.
Yes, one solution is to add policy support so that rules could be defined instead of actually hard coding each and every fs magic number here in skip_measurement. This is on my list of todo's.
Quoted 5 lines
Generally: the code is all moderately intrusive into the VFS and this sort of thing does need careful explanation and justification, please. Once we have some understanding of what you're trying to achieve here we will inevitably ask "can't that be done in userspace". So it would be best if your description were to preemptively answer all that.
Ok.
Quoted 34 lines · On Wed, 28 May 2008 23:17:26 -0400 Mimi Zohar <[email protected]> wrote:
Quoted 24 lines · Andrew Morton
Quoted 22 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar <[email protected]> wrote:
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM
must_measure(),collect_measurement(),store_measurement(), anddisplay_template()API calls. Thestore_measurement()call supports two types of data, IMA (i.e. file data) and generic template data.When
store_measurement()is called for the IMA type of data, the file measurement and the file name hint are used to form an IMA template. IMA then calculates the IMA template measurement(hash) and submits it to the TPM chip for inclusion in one of the chip's Platform Configuration Registers (PCR).When
store_measurement()is called for generic template data, IMA calculates the measurement(hash) of the template data, and submits the template measurement to the TPM chip for inclusion in one of the chip's Platform Configuration Registers(PCR).In order to view the contents of template data through securityfs, the
template_display()function must be defined in the registeredtemplate_operations. In the case of the IMA template, the list of file names and files hashes submitted can be viewed through securityfs.Quoted 8 lines
Quoted 4 lines
IMA can be included or excluded in the kernel configuration. If included in the kernel and the
IMA_BOOTPARAMis selected, IMA can also be enabled/disabled on the kernel command line with 'ima='.- I see lots of user file I/O being done from within the kernel. This makes eyebrows raise. Also some other eyebrow-raising file-related things in there.
Quoted 6 lines
The amount of I/O is dependent on the number of files being measured. The default policy measures a whole lot. An LSM specific integrity policy would cut down on the number of files being measured. For now, either remove the third rule in
default_rulesor replace the default rules with a new policy. To load a new policy execute:
./integrity_load < policy
The problem is that the code is doing in-kernel user file I/O *at all*. It's a red flag.
Look who else is using kernel_read(): just the exec code. Plus something in v9fs which I'd better not look at.
Quoted 7 lines
...
Quoted 1 line
-
timespec_set()is unneeeded - just use struct assignment (ie: "=")Am confused.
timespec_setis doing an assignment. Should I replacetimespec_setwith a memcpy?
struct timespec a, b; a = b;
Quoted 5 lines
Quoted 2 lines
- All the games with mtimes should be described in the changelog too.
Ok. The
timespec_recentand mtime issues are part of the same problem of detecting when a file has been modified.
Can't use inode.i_version?
Quoted 6 lines
Quoted 4 lines
-
ima_fixup_inodeslooks like it will race and crash against a well-timed unmount. I expect you will need to bumps_countbefore droppingsb_lock. Seewriteback_inodes()for an example.
ima_fixup_inodes()is called once at initialization.
What is "initialisation"? During initcalls? Are there even any files in cache at that time? I bet we can arrange for the answer to become "no".
Quoted 8 lines · Andrew Morton
Quoted 6 lines
...
Quoted 1 line
- All the games with mtimes should be described in the changelog too.
Ok. The
timespec_recentand mtime issues are part of the same problem of detecting when a file has been modified.Can't use inode.i_version?
i_version is now working on my system. It looks good. Is there anything that I need to be concerned about, such as limited filesystem support or the i_version is not updated by file_close for mmaped files?
Thanks!
Mimi
Quoted 5 lines · On Thu, 29 May 2008 17:50:34 -0400 Mimi Zohar <[email protected]> wrote:
Quoted 2 lines · Andrew Morton
Quoted 2 lines
...
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Quoted 1 line
- All the games with mtimes should be described in the changelog too.
Ok. The
timespec_recentand mtime issues are part of the same problem of detecting when a file has been modified.Can't use inode.i_version?
i_versionis now working on my system.
Oh good.
Quoted 4 lines
It looks good. Is
there anything that I need to be concerned about, such as limited filesystem support or thei_versionis not updated byfile_closefor mmaped files?
erk, I'm not an i_version person. It seems that it's only used on directories (to patch up readdir coherency problems) so I guess I misled you there.
There's file_struct.f_version, which is no good.
i_generation is no good either.
We've documented these things so wonderfully!
i_writecount looks good?
Quoted 19 lines · Andrew Morton
Quoted 5 lines · On Thu, 29 May 2008 17:50:34 -0400 Mimi Zohar <[email protected]> wrote:
Quoted 2 lines · Andrew Morton
Quoted 2 lines
...
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Quoted 1 line
- All the games with mtimes should be described in the changelog too.
Ok. The
timespec_recentand mtime issues are part of the same problem of detecting when a file has been modified.Can't use inode.i_version?
i_versionis now working on my system.Oh good.
Quoted 8 lines
Quoted 4 lines
It looks good. Is
there anything that I need to be concerned about, such as limited filesystem support or thei_versionis not updated byfile_closefor mmaped files?erk, I'm not an
i_versionperson. It seems that it's only used on directories (to patch up readdir coherency problems) so I guess I misled you there.
No, no. Initially, that's what I thought. I finally found file_update_time() calls inode_inc_iversion(), which updates the i_version. So, it does work. The question is whether or not it works all the time. :-)
Quoted 7 lines
There's file_struct.f_version, which is no good.
i_generationis no good either.We've documented these things so wonderfully!
i_writecountlooks good?
Quoted 8 lines · On Thu, 29 May 2008 21:58:08 -0400 Mimi Zohar <[email protected]> wrote:
Quoted 3 lines
erk, I'm not an
i_versionperson. It seems that it's only used on directories (to patch up readdir coherency problems) so I guess I misled you there.No, no. Initially, that's what I thought. I finally found
file_update_time()callsinode_inc_iversion(), which updates thei_version. So, it does work. The question is whether or not it works all the time. :-)
OK.
Before 2.6.17 it wouldn't have worked much at all on MAP_SHARED modifications.
After 2.6.17 things will be better - we update the mtime on the clean->dirty transitions of a page. So the first modification after an mmap will update the time.
Subsequent modifications via the mmap will not update the file time. Until something (usually pdflush) writes the page out. Then the next modification via mmap will cause another clean->dirty transition on the page, hence another mtime update.
So there's a by-default 30-odd second uncertainty with MAP_SHARED alterations.
Quoted 9 lines · Andrew Morton
Quoted 5 lines
Quoted 3 lines
-
ima_fixup_inodeslooks like it will race and crash against a well-timed unmount. I expect you will need to bumps_countbefore droppingsb_lock. Seewriteback_inodes()for an example.
ima_fixup_inodes()is called once at initialization.What is "initialisation"? During initcalls? Are there even any files in cache at that time? I bet we can arrange for the answer to become "no".
Yes, during the initcalls. Is this possible even when using late_initcall(). IMA is dependent on the TPM being available, if it is being used.
Mimi
Quoted 6 lines · Andrew Morton
+/**
+ *ima_must_measure- measure decision based on policy.
+ * @d - pointer to structima_datacontainingima_args_dataSo if we know the type of d, did we _have_ to make it void*? It's much better to use the C yype system if at all possible.
This is one of the five integrity API calls. Each integrity template will define it differently, using a different data structure.
Quoted 1 line
- ditto
ima_collect_measurement()
ima_collect_measurement is also one of the five integrity API calls.
Here is a sample template kernel module that measures kernel memory. Of the five integrity API calls, it implements integrity_collect_measurement(), integrity_store_measurement(), and integrity_display_measurement(). It collects and stores measurements based on data read from security/kmem-template. The format is "name length address". The name can be any string identifier such as "proc_root"; the length is the number of bytes to measure; and address is a kernel memory address, which can be looked up in /proc/kallsyms. One caveat, the sample program currently does not validate the address. A userspace application triggers the measurement by writing to security/kmem-template.
/*
* Copyright (C) 2008 IBM Corporation
* Author: Mimi Zohar <[email protected]>
*
* This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation, version 2 of the License.
*
* kmem-template.c
* - defines a kernel memory template
* - reads from security/kmem-template "name length address"
* - collects and stores measurement from address for length bytes
*/
#include <asm/uaccess.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/kernel.h>
#include <linux/fs.h>
#include <linux/crypto.h>
#include <linux/scatterlist.h>
#include <linux/notifier.h>
#include <linux/security.h>
#include <linux/debugfs.h>
#include <linux/seq_file.h>
#include <linux/string.h>
#include <linux/proc_fs.h>
#include <linux/security.h>
#include <linux/integrity.h>
#include <linux/ima.h>
#define MY_NAME THIS_MODULE->name
#define IMA_DIGEST_SIZE 20
static int __init init_kmem_template(void);
static void __exit cleanup_kmem_template(void);
struct kmem_data {
char name[25];
char *buf;
int buflen;
u8 digest[IMA_DIGEST_SIZE];
};static void hexdump(unsigned char *buf, unsigned int len)
{
print_hex_dump(KERN_CONT, "", DUMP_PREFIX_OFFSET,
16, 1,
buf, len, false);
}int calc_hash(int buflen, char *buf, char *digest)
{
struct crypto_hash *tfm;
struct hash_desc desc;
struct scatterlist sg[1];
int error, result = 0;
tfm = crypto_alloc_hash("sha1", 0, CRYPTO_ALG_ASYNC);
if (IS_ERR(tfm)) {
printk(KERN_INFO "%s: failed to load %s transform: %ld\n",
__func__, "sha1", PTR_ERR(tfm));
return -ENOSYS;
}
desc.tfm = tfm;
desc.flags = 0;
error = crypto_hash_init(&desc);
if (error) {
result = -EINVAL;
goto out;
}
sg_set_buf(sg, buf, buflen);
result = crypto_hash_update(&desc, sg, buflen);
if (!result) {
error = crypto_hash_final(&desc, digest);
if (error)
result = -EINVAL;
}out:
crypto_free_hash(tfm); return result; }
static int kmem_collect_measurement(void *d)
{
struct kmem_data *data = (struct kmem_data *)d;
memset(data->digest, 0, sizeof data->digest);
calc_hash(data->buflen, data->buf, data->digest);
hexdump(data->buf, data->buflen);
return 0;
}/* Transform local kmem data to store data */
void kmem_store_measurement(void *d)
{
struct kmem_data *data = (struct kmem_data *)d;
struct ima_data idata;
struct ima_store_data *template = &idata.data.template;
idata.type = IMA_TEMPLATE;
template->name = "kmem";
template->len = sizeof *data;
template->data = (char *)data;
template->violation = 0;
integrity_store_measurement("ima", (void *)&idata);
return;
}static void kmem_template_show(struct seq_file *m, void *e,
enum integrity_show_type show)
{
struct kmem_data *data = (struct kmem_data *)e;
int filename_len;
char len[4];
int i;
for (i = 0; i < 20; i++) {
switch (show) {
case INTEGRITY_SHOW_ASCII:
seq_printf(m, "%02x", data->digest[i]);
break;
case INTEGRITY_SHOW_BINARY:
seq_putc(m, data->digest[i]);
default:
break;
}
}
switch (show) {
case INTEGRITY_SHOW_ASCII:
seq_printf(m, " %s %d \n", data->name, data->buflen);
break;
case INTEGRITY_SHOW_BINARY:
filename_len = strlen(data->name);
memcpy(len, &filename_len, 4);
for (i = 0; i < 4; i++)
seq_putc(m, len[i]);
for (i = 0; i < strlen(data->name); i++)
seq_putc(m, data->name[i]);
default:
break;
}
}static struct template_operations kmem_ops = {
.collect_measurement = kmem_collect_measurement, .store_measurement = kmem_store_measurement, .display_template = kmem_template_show
};
static int kmem_add_measure(char *name, unsigned int buflen,
unsigned int addr)
{
struct kmem_data data;
int rc;
strncpy(data.name, name, sizeof data.name);
data.buflen = buflen;
data.buf = (char *)addr;
rc = integrity_collect_measurement("kmem", &data);
if (!rc)
integrity_store_measurement("kmem", &data);
return rc;
} static ssize_t kmem_write_template(struct file *file, const char __user *buf,
size_t buflen, loff_t *ppos)
{
char *data;
char name[26];
size_t result = 0, datalen;
int rc;
unsigned int addr, len;
datalen = buflen > 256 ? 256 : buflen;
data = kzalloc(datalen + 1, GFP_KERNEL);
if (!data)
result = -ENOMEM;
if (copy_from_user(data, buf, datalen)) {
result = -EFAULT;
goto out;
}
result = datalen;
rc = sscanf(data, "%25s %d %x ", name, &len, &addr);
if (rc == 3)
kmem_add_measure(name, len, addr);
else {
printk(KERN_INFO "%s: rc = %d\n", __func__, rc);
result = -EINVAL;
}
out:
if (!data)
kfree(data);
return result;
}static struct file_operations kmem_template_ops = {
.write = kmem_write_template
};
static struct dentry *kmem_template;
static int __init init_kmem_template(void)
{
printk(KERN_INFO "%s: \n", __func__);
register_template("kmem", &kmem_ops);
kmem_template = securityfs_create_file("kmem-template",
S_IRUSR | S_IRGRP | S_IWUSR,
NULL, NULL, &kmem_template_ops);
return 0;
}static void __exit cleanup_kmem_template(void)
{
printk(KERN_INFO "%s\n", __FUNCTION__);
unregister_template("kmem");
securityfs_remove(kmem_template);
}module_init(init_kmem_template);module_exit(cleanup_kmem_template);
MODULE_LICENSE("GPL");
Quoted 7 lines · Andrew Morton
Quoted 5 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar <[email protected]> wrote:
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM
must_measure(),collect_measurement(),store_measurement(), anddisplay_template()API calls. Thestore_measurement()call supports two types of data, IMA (i.e. file data) and generic template data.
...
Quoted 5 lines
Generally: the code is all moderately intrusive into the VFS and this sort of thing does need careful explanation and justification, please. Once we have some understanding of what you're trying to achieve here we will inevitably ask "can't that be done in userspace". So it would be best if your description were to preemptively answer all that.
...also, it would be nice to see explanation 'what is this good for'.
Closest explanation I remember was 'it will protect you by making system unbootable if someone stole disk with your /usr filesystem -- but not / filesystem -- added some rootkit, and then stealthily returned it'. That seems a) very unlikely scenario and b) probably better solved by encrypting /usr.
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Quoted 11 lines · Pavel Machek
Quoted 7 lines · Andrew Morton
Quoted 5 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar <[email protected]> wrote:
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM
must_measure(),collect_measurement(),store_measurement(), anddisplay_template()API calls. Thestore_measurement()call supports two types of data, IMA (i.e. file data) and generic template data....
...also, it would be nice to see explanation 'what is this good for'.
Quoted 6 lines
Closest explanation I remember was 'it will protect you by making system unbootable if someone stole disk with your /usr filesystem -- but not / filesystem -- added some rootkit, and then stealthily returned it'. That seems a) very unlikely scenario and b) probably better solved by encrypting /usr.
Pavel
Sorry about this delayed response - we are about to repost for RFC, and noticed we missed responding to this.
You are thinking about a related project, EVM, which HMAC's a file's metadata, to protect against off-line attacks, (which admittedly many users are not concerned about.)
This submission, IMA, provides hardware (TPM) based measurement and attestation, which measures all files before they are accessed in any way (on the inode_permission, bprm and mmap hooks), and commits the measurements to the TPM. The TPM can sign these measurement lists, and thus the system can prove to itself and to a third party these measurements in a way that cannot be circumvented by malicious or compromised software. IMA is just one part of integrity detection, as it does not detect purely in-memory attacks, such as worms.
dave safford
Quoted 26 lines · david safford
Quoted 11 lines · Pavel Machek
Quoted 7 lines · Andrew Morton
Quoted 5 lines · On Fri, 23 May 2008 11:05:45 -0400 Mimi Zohar <[email protected]> wrote:
This is a re-release of Integrity Measurement Architecture(IMA) as an independent Linunx Integrity Module(LIM) service provider, which implements the new LIM
must_measure(),collect_measurement(),store_measurement(), anddisplay_template()API calls. Thestore_measurement()call supports two types of data, IMA (i.e. file data) and generic template data....
...also, it would be nice to see explanation 'what is this good for'.Quoted 6 lines
Closest explanation I remember was 'it will protect you by making system unbootable if someone stole disk with your /usr filesystem -- but not / filesystem -- added some rootkit, and then stealthily returned it'. That seems a) very unlikely scenario and b) probably better solved by encrypting /usr.
PavelSorry about this delayed response - we are about to repost for RFC, and noticed we missed responding to this.
You are thinking about a related project, EVM, which HMAC's a file's metadata, to protect against off-line attacks, (which admittedly many users are not concerned about.)
Quoted 5 lines
This submission, IMA, provides hardware (TPM) based measurement and attestation, which measures all files before they are accessed in any way (on the
inode_permission, bprm and mmap hooks), and commits the measurements to the TPM. The TPM can sign these measurement lists, and thus the system can prove to itself and
System can never proof to itself.
Quoted 4 lines
to a third party these measurements in a way that cannot be circumvented by malicious or compromised software. IMA is just one part of integrity detection, as it does not detect purely in-memory attacks, such as worms.
And proofing to third party is useful for what....? Given that it can be worked around by modifying files in memory, or by special hardware...? Disney?
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Quoted 11 lines · Andrew Morton
- I see lots of user file I/O being done from within the kernel. This makes eyebrows raise. Also some other eyebrow-raising file-related things in there.
Generally: the code is all moderately intrusive into the VFS and this sort of thing does need careful explanation and justification, please. Once we have some understanding of what you're trying to achieve here we will inevitably ask "can't that be done in userspace". So it would be best if your description were to preemptively answer all that.
Sorry about this delayed response - we are about to repost for RFC, and noticed we missed responding to this.
The Trusted Computing (TPM) model requires that all files be measured, (hashed) and the measurement committed to the hardware TPM before any data of the file is accessed in any way. In addition, if the measurement is incorrect, all access to the file must be denied.
This requirement parallels the LSM mandatory access control decisions in the inode_permission, bprm, and mmap hooks, and naturally leads to IMA hooks in the same locations, with similar functionality, but with the addition of hashing the data. The code would have to significantly more complex to do the hashing at these points through userspace.
In addition, doing the hashing in userspace gives significantly poorer performance. With in-kernel hashing, at boot time, we typically measure some six thousand files with less than 10% (5 seconds) overhead, which is acceptable to most users. Anything much slower can be annoying enough that users will turn the measurement off.
dave safford
Quoted 13 lines · david safford
Quoted 11 lines · Andrew Morton
- I see lots of user file I/O being done from within the kernel. This makes eyebrows raise. Also some other eyebrow-raising file-related things in there.
Generally: the code is all moderately intrusive into the VFS and this sort of thing does need careful explanation and justification, please. Once we have some understanding of what you're trying to achieve here we will inevitably ask "can't that be done in userspace". So it would be best if your description were to preemptively answer all that.
Quoted 7 lines
Sorry about this delayed response - we are about to repost for RFC, and noticed we missed responding to this.
The Trusted Computing (TPM) model requires that all files be measured, (hashed) and the measurement committed to the hardware TPM before any data of the file is accessed in any way. In addition, if the measurement is incorrect, all access to the file must be denied.
TPM model may require this, but what is the benefit to the user/owner of the machine?
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html