[PATCH v21 00/23] LSM: Module stacking for AppArmor

This patchset provides the changes required for the AppArmor security module to stack safely with any other. v21: Rebase to 5.9-rc4 Incorporate feedback from v20 - Further revert UDS SO_PEERSEC to use scaffolding around the interfaces that use lsmblobs and store only a single secid. The possibility of multiple security modules requiring data here is still a future problem. - Incorporate Richard Guy Briggs' non-syscall auxiliary records patch (patch 0019-0021) in place of my "supplimental" records implementation. [I'm not sure I've given proper attestation. I will correct as appropriate] v20: Rebase to 5.9-rc1 Change the BPF security module to use the lsmblob data. (patch 0002) Repair length logic in subject label processing (patch 0015) Handle -EINVAL from the empty BPF setprocattr hook (patch 0020) Correct length processing in append_ctx() (patch 0022) v19: Rebase to 5.8-rc6 Incorporate feedback from v18 - Revert UDS SO_PEERSEC implementation to use lsmblobs directly, rather than allocating as needed. The correct treatment of out-of-memory conditions in the later case is difficult to define. (patch 0005) - Use a size_t in append_ctx() (patch 0021) - Fix a memory leak when creating compound contexts. (patch 0021) Fix build error when CONFIG_SECURITY isn't set (patch 0013) Fix build error when CONFIG_SECURITY isn't set (patch 0020) Fix build error when CONFIG_SECURITY isn't set (patch 0021) v18: Rebase to 5.8-rc3 Incorporate feedback from v17 - Null pointer checking in UDS (patch 0005) Match changes in IMA code (patch 0012) Fix the behavior of LSM context supplimental audit records so that there's always exactly one when it's appropriate for there to be one. This is a substantial change that requires extention of the audit_context beyond syscall events. (patch 0020) v17: Rebase to 5.7-rc4 v16: Rebase to 5.6 Incorporate feedback from v15 - Thanks Stephen, Mimi and Paul - Generally improve commit messages WRT scaffolding - Comment ima_lsm_isset() (patch 0002) - Some question may remain on IMA warning (patch 0002) - Mark lsm_slot as __lsm_ro_after_init not __init_data (patch 0002) - Change name of lsmblob variable in ima_match_rules() (patch 0003) - Instead of putting a struct lsmblob into the unix_skb_parms structure put a pointer to an allocated instance. There is currently only space for 5 u32's in unix_skb_parms and it is likely to get even tighter. Fortunately, the lifecycle management of the allocated lsmblob is simple. (patch 0005) - Dropped Acks due to the above change (patch 0005) - Improved commentary on secmark labeling scaffolding. (patch 0006) - Reduced secmark related labeling scaffolding. (patch 0006) - Replace use of the zeroth entry of an lsmblob in scaffolding with a function lsmblob_value() to hopefully make it less obscure. (patch 0006) - Convert security_secmark_relabel_packet to use lsmblob as this reduces much of the most contentious scaffolding. (patch 0006) - Dropped Acks due to the above change (patch 0006) - Added BUILD_BUG_ON() for CIPSO tag 6. (patch 0018) - Reworked audit subject information. Instead of adding fields in the middle of existing records add a new record to the event. When a separate record is required use subj="?". (patch 0020) - Dropped Acks due to the above change (patch 0020) - Reworked audit object information. Instead of adding fields in the middle of existing records add a new record to the event. When a separate record is required use obj="?". (patch 0021) - Dropped Acks due to the above change (patch 0021) - Enhanced documentation (patch 0022) - Removed unnecessary error code check in security_getprocattr() (patch 0021) v15: Rebase to 5.6-rc1 - Revise IMA data use (patch 0002) Incorporate feedback from v14 - Fix lockdown module registration naming (patch 0002) - Revise how /proc/self/attr/context is gathered. (patch 0022) - Revise access modes on /proc/self/attr/context. (patch 0022) - Revise documentation on LSM external interfaces. (patch 0022) v14: Rebase to 5.5-rc5 Incorporate feedback from v13 - Use an array of audit rules (patch 0002) - Significant change, removed Acks (patch 0002) - Remove unneeded include (patch 0013) - Use context.len correctly (patch 0015) - Reorder code to be more sensible (patch 0016) - Drop SO_PEERCONTEXT as it's not needed yet (patch 0023) v13: Rebase to 5.5-rc2 Incorporate feedback from v12 - Print lsmblob size with %z (Patch 0002) - Convert lockdown LSM initialization. (Patch 0002) - Restore error check in nft_secmark_compute_secid (Patch 0006) - Correct blob scaffolding in ima_must_appraise() (Patch 0009) - Make security_setprocattr() clearer (Patch 0013) - Use lsm_task_display more widely (Patch 0013) - Use passed size in lsmcontext_init() (Patch 0014) - Don't add a smack_release_secctx() hook (Patch 0014) - Don't print warning in security_release_secctx() (Patch 0014) - Don't duplicate the label in nfs4_label_init_security() (Patch 0016) - Remove reviewed-by as code has significant change (Patch 0016) - Send the entire lsmblob for Tag 6 (Patch 0019) - Fix description of socket_getpeersec_stream parameters (Patch 0023) - Retain LSMBLOB_FIRST. What was I thinking? (Patch 0023) - Add compound context to LSM documentation (Patch 0023) v12: Rebase to 5.5-rc1 Fixed a couple of incorrect contractions in the text. v11: Rebase to 5.4-rc6 Incorporate feedback from v10 - Disambiguate reading /proc/.../attr/display by restricting all use of the interface to the current process. - Fix a merge error in AppArmor's display attribute check v10: Ask the security modules if the display can be changed. v9: There is no version 9 v8: Incorporate feedback from v7 - Minor clean-up in display value management - refactor "compound" context creation to use a common append_ctx() function. v7: Incorporate feedback from v6 - Make setting the display a privileged operation. The availability of compound contexts reduces the need for setting the display. v6: Incorporate feedback from v5 - Add subj_<lsm>= and obj_<lsm>= fields to audit records - Add /proc/.../attr/context to get the full context in lsmname\0value\0... format as suggested by Simon McVittie - Add SO_PEERCONTEXT for getsockopt() to get the full context in the same format, also suggested by Simon McVittie. - Add /sys/kernel/security/lsm_display_default to provide the display default value. v5: Incorporate feedback from v4 - Initialize the lsmcontext in security_secid_to_secctx() - Clear the lsmcontext in all security_release_secctx() cases - Don't use the "display" on strictly internal context interfaces. - The SELinux binder hooks check for cases where the context "display" isn't compatible with SELinux. v4: Incorporate feedback from v3 - Mark new lsm_<blob>_alloc functions static - Replace the lsm and slot fields of the security_hook_list with a pointer to a LSM allocated lsm_id structure. The LSM identifies if it needs a slot explicitly. Use the lsm_id rather than make security_add_hooks return the slot value. - Validate slot values used in security.c - Reworked the "display" process attribute handling so that it works right and doesn't use goofy list processing. - fix display value check in dentry_init_security - Replace audit_log of secids with '?' instead of deleting the audit log v3: Incorporate feedback from v2 - Make lsmblob parameter and variable names more meaningful, changing "le" and "l" to "blob". - Improve consistency of constant naming. - Do more sanity checking during LSM initialization. - Be a bit clearer about what is temporary scaffolding. - Rather than clutter security_getpeersec_dgram with otherwise unnecessary checks remove the apparmor stub, which does nothing useful. Patch 0001 moves management of the sock security blob from the individual modules to the infrastructure. Patches 0002-0011 replace system use of a "secid" with a structure "lsmblob" containing information from the security modules to be held and reused later. At this point lsmblob contains an array of u32 secids, one "slot" for each of the security modules compiled into the kernel that used secids. A "slot" is allocated when a security module requests one. The infrastructure is changed to use the slot number to pass the correct secid to or from the security module hooks. It is important that the lsmblob be a fixed size entity that does not have to be allocated. Several of the places where it is used would have performance and/or locking issues with dynamic allocation. Patch 0012 provides a mechanism for a process to identify which security module's hooks should be used when displaying or converting a security context string. A new interface /proc/self/attr/display contains the name of the security module to show. Reading from this file will present the name of the module, while writing to it will set the value. Only names of active security modules are accepted. Internally, the name is translated to the appropriate "slot" number for the module which is then stored in the task security blob. Setting the display requires that all modules using the /proc interfaces allow the transition. The "display" of other processess can be neither read nor written. All suggested cases for reading the display of a different process have race conditions. Patch 0013 Starts the process of changing how a security context is represented. Since it is possible for a security context to have been generated by more than one security module it is now necessary to note which module created a security context so that the correct "release" hook can be called. There are several places where the module that created a security context cannot be inferred. This is achieved by introducing a "lsmcontext" structure which contains the context string, its length and the "slot" number of the security module that created it. The security_release_secctx() interface is changed, replacing the (string,len) pointer pair with a lsmcontext pointer. Patches 0014-0016 convert the security interfaces from (string,len) pointer pairs to a lsmcontext pointer. The slot number identifying the creating module is added by the infrastructure. Where the security context is stored for extended periods the data type is changed. The Netlabel code is converted to save lsmblob structures instead of secids in Patch 0017. This is not strictly necessary as there can only be one security module that uses Netlabel at this point. Using a lsmblob is much cleaner, as the interfaces that use the data have all been converted. Patch 0018 adds checks to the binder hooks which verify that both ends of a transaction use the same "display". Patches 0019-0021 add addition audit records for subject and object LSM data when there are multiple security modules with such data. The AUDIT_MAC_TASK_CONTEXTS record is used in conjuction with a "subj=?" field to identify the subject data. The AUDIT_MAC_OBJ_CONTEXTS record is used in conjuction with a "obj=?" field to identify the object data. The AUDIT_MAC_TASK_CONTEXTS record identifies the security module with the data: "subj_selinux=xyz_t subj_apparmor=abc". The AUDIT_MAC_OBJ_CONTEXTS record identifies the security module with the data: "obj_selinux=xyz_t obj_apparmor=abc". While AUDIT_MAC_TASK_CONTEXTS records will always contain an entry for each possible security modules, AUDIT_MAC_OBJ_CONTEXTS records will only contain entries for security modules for which the object in question has data. An example of the MAC_TASK_CONTEXTS (1420) record is: type=UNKNOWN[1420] msg=audit(1600880931.832:113) subj_apparmor==unconfined subj_smack=_ An example of the MAC_OBJ_CONTEXTS (1421) record is: type=UNKNOWN[1421] msg=audit(1601152467.009:1050): obj_selinux=unconfined_u:object_r:user_home_t:s0 Patch 0022 adds a new interface for getting the compound security contexts, /proc/self/attr/context. An example of the content of this file is: selinux\0one_u:one_r:one_t:s0-s0:c0.c1023\0apparmor\0unconfined\0 Finally, with all interference on the AppArmor hooks removed, Patch 0023 removes the exclusive bit from AppArmor. An unnecessary stub hook was also removed. The Ubuntu project is using an earlier version of this patchset in their distribution to enable stacking for containers. Performance measurements to date have the change within the "noise". The sockperf and dbench results are on the order of 0.2% to 0.8% difference, with better performance being as common as worse. The benchmarks were run with AppArmor and Smack on Ubuntu. https://github.com/cschaufler/lsm-stacking.git#stack-5.9-rc4-v21 Signed-off-by: Casey Schaufler <casey(a)schaufler-ca.com&gt; ---