From: Paul Moore <paul@paul-moore.com> To: John Johansen <john.johansen@canonical.com> Cc: Casey Schaufler <casey@schaufler-ca.com>, casey.schaufler@intel.com, jmorris@namei.org, linux-security-module@vger.kernel.org, selinux@vger.kernel.org, linux-audit@redhat.com, keescook@chromium.org, penguin-kernel@i-love.sakura.ne.jp, stephen.smalley.work@gmail.com, linux-kernel@vger.kernel.org Subject: Re: [PATCH v35 27/29] Audit: Add record for multiple object contexts Date: Tue, 26 Apr 2022 14:57:11 -0400 [thread overview] Message-ID: <CAHC9VhTXgBTH+7ny-fcMP_HC1ojA1ass38PGHS2tJny0bCGXzA@mail.gmail.com> (raw) In-Reply-To: <f2186f22-5bcd-d962-7e49-c816fc5fbd07@canonical.com> On Mon, Apr 25, 2022 at 11:38 PM John Johansen <john.johansen@canonical.com> wrote: > On 4/18/22 07:59, Casey Schaufler wrote: > > Create a new audit record AUDIT_MAC_OBJ_CONTEXTS. > > An example of the MAC_OBJ_CONTEXTS (1421) record is: > > > > type=MAC_OBJ_CONTEXTS[1421] > > msg=audit(1601152467.009:1050): > > obj_selinux=unconfined_u:object_r:user_home_t:s0 > > > > When an audit event includes a AUDIT_MAC_OBJ_CONTEXTS record > > the "obj=" field in other records in the event will be "obj=?". > > An AUDIT_MAC_OBJ_CONTEXTS record is supplied when the system has > > multiple security modules that may make access decisions based > > on an object security context. > > > > Signed-off-by: Casey Schaufler <casey@schaufler-ca.com> > > --- > > include/linux/audit.h | 5 +++ > > include/uapi/linux/audit.h | 1 + > > kernel/audit.c | 47 +++++++++++++++++++++++ > > kernel/auditsc.c | 79 ++++++++++++-------------------------- > > 4 files changed, 77 insertions(+), 55 deletions(-) ... > > diff --git a/kernel/audit.c b/kernel/audit.c > > index 8ed2d717c217..a8c3ec6ba60b 100644 > > --- a/kernel/audit.c > > +++ b/kernel/audit.c > > @@ -2226,6 +2226,53 @@ static void audit_buffer_aux_end(struct audit_buffer *ab) > > ab->skb = skb_peek(&ab->skb_list); > > } > > > > +void audit_log_object_context(struct audit_buffer *ab, struct lsmblob *blob) > > +{ > > + int i; > > + int error; > > + struct lsmcontext context; > > + > > + if (!lsm_multiple_contexts()) { > > + error = security_secid_to_secctx(blob, &context, LSMBLOB_FIRST); > > + if (error) { > > + if (error != -EINVAL) > > + goto error_path; > > + return; > > + } > > + audit_log_format(ab, " obj=%s", context.context); > > + security_release_secctx(&context); > > + } else { > > + audit_log_format(ab, " obj=?"); > > + error = audit_buffer_aux_new(ab, AUDIT_MAC_OBJ_CONTEXTS); > > + if (error) > > + goto error_path; > > + > > + for (i = 0; i < LSMBLOB_ENTRIES; i++) { > > + if (blob->secid[i] == 0) > > + continue; > > + error = security_secid_to_secctx(blob, &context, i); > > + if (error) { > > + audit_log_format(ab, "%sobj_%s=?", > > + i ? " " : "", > > + lsm_slot_to_name(i)); > > + if (error != -EINVAL) > > + audit_panic("error in audit_log_object_context"); > > + } else { > > + audit_log_format(ab, "%sobj_%s=%s", > > + i ? " " : "", > > + lsm_slot_to_name(i), > > + context.context); > > + security_release_secctx(&context); > > + } > > + } > > + > > + audit_buffer_aux_end(ab); > > + } > > + return; > > + > > +error_path: > > + audit_panic("error in audit_log_object_context"); > > This moves the audit_panic around, so certain operations are not > done before the call. I am currently not sure of the implications. Short version: It's okay. Longer version: The audit_panic() call is either going to panic the kernel (NOT the default), do a pr_err(), or essentially be a no-op. In the case of the full blown kernel panic we don't really care, the system is going to die before there is any chance of this record in progress getting logged. In the case of a pr_err() or no-op the key part is making sure we leave the audit_buffer in a consistent state so that we preserve whatever information is already present. In the !lsm_multiple_contexts case we simply return without making any changes to the audit_buffer so we're good there; in the multiple LSM case we always end the aux record properly (using a "?" when necessary) if an aux record has been successfully created. Feel free to point out a specific scenario that you think looks wrong - I may have missed it - but I believe this code to be correct. > > diff --git a/kernel/auditsc.c b/kernel/auditsc.c > > index 557713954a69..04bf3c04ef3d 100644 > > --- a/kernel/auditsc.c > > +++ b/kernel/auditsc.c > > @@ -1420,18 +1409,10 @@ static void show_special(struct audit_context *context, int *call_panic) > > If pushing audit_panic into audit_log_object_context() is acceptable then this call_panic arg is > no longer needed. The same goes for the call_panic arg in audit_log_name(). And call_panic can > be dropped from audit_log_exit() Good catch. I suspect this is a vestige from when audit_log_end() used to do the record's skb write to userspace, meaning it was possible that you might get some of the records written to userspace before the system killed itself. Now with all of the queuing involved it's less likely that this would be the case, and even if it does happen in some cases, it's basically a toss up depending on how the system is loaded, the scheduler, etc. -- paul-moore.com
WARNING: multiple messages have this Message-ID (diff)
From: Paul Moore <paul@paul-moore.com> To: John Johansen <john.johansen@canonical.com> Cc: selinux@vger.kernel.org, jmorris@namei.org, linux-kernel@vger.kernel.org, casey.schaufler@intel.com, linux-security-module@vger.kernel.org, linux-audit@redhat.com Subject: Re: [PATCH v35 27/29] Audit: Add record for multiple object contexts Date: Tue, 26 Apr 2022 14:57:11 -0400 [thread overview] Message-ID: <CAHC9VhTXgBTH+7ny-fcMP_HC1ojA1ass38PGHS2tJny0bCGXzA@mail.gmail.com> (raw) In-Reply-To: <f2186f22-5bcd-d962-7e49-c816fc5fbd07@canonical.com> On Mon, Apr 25, 2022 at 11:38 PM John Johansen <john.johansen@canonical.com> wrote: > On 4/18/22 07:59, Casey Schaufler wrote: > > Create a new audit record AUDIT_MAC_OBJ_CONTEXTS. > > An example of the MAC_OBJ_CONTEXTS (1421) record is: > > > > type=MAC_OBJ_CONTEXTS[1421] > > msg=audit(1601152467.009:1050): > > obj_selinux=unconfined_u:object_r:user_home_t:s0 > > > > When an audit event includes a AUDIT_MAC_OBJ_CONTEXTS record > > the "obj=" field in other records in the event will be "obj=?". > > An AUDIT_MAC_OBJ_CONTEXTS record is supplied when the system has > > multiple security modules that may make access decisions based > > on an object security context. > > > > Signed-off-by: Casey Schaufler <casey@schaufler-ca.com> > > --- > > include/linux/audit.h | 5 +++ > > include/uapi/linux/audit.h | 1 + > > kernel/audit.c | 47 +++++++++++++++++++++++ > > kernel/auditsc.c | 79 ++++++++++++-------------------------- > > 4 files changed, 77 insertions(+), 55 deletions(-) ... > > diff --git a/kernel/audit.c b/kernel/audit.c > > index 8ed2d717c217..a8c3ec6ba60b 100644 > > --- a/kernel/audit.c > > +++ b/kernel/audit.c > > @@ -2226,6 +2226,53 @@ static void audit_buffer_aux_end(struct audit_buffer *ab) > > ab->skb = skb_peek(&ab->skb_list); > > } > > > > +void audit_log_object_context(struct audit_buffer *ab, struct lsmblob *blob) > > +{ > > + int i; > > + int error; > > + struct lsmcontext context; > > + > > + if (!lsm_multiple_contexts()) { > > + error = security_secid_to_secctx(blob, &context, LSMBLOB_FIRST); > > + if (error) { > > + if (error != -EINVAL) > > + goto error_path; > > + return; > > + } > > + audit_log_format(ab, " obj=%s", context.context); > > + security_release_secctx(&context); > > + } else { > > + audit_log_format(ab, " obj=?"); > > + error = audit_buffer_aux_new(ab, AUDIT_MAC_OBJ_CONTEXTS); > > + if (error) > > + goto error_path; > > + > > + for (i = 0; i < LSMBLOB_ENTRIES; i++) { > > + if (blob->secid[i] == 0) > > + continue; > > + error = security_secid_to_secctx(blob, &context, i); > > + if (error) { > > + audit_log_format(ab, "%sobj_%s=?", > > + i ? " " : "", > > + lsm_slot_to_name(i)); > > + if (error != -EINVAL) > > + audit_panic("error in audit_log_object_context"); > > + } else { > > + audit_log_format(ab, "%sobj_%s=%s", > > + i ? " " : "", > > + lsm_slot_to_name(i), > > + context.context); > > + security_release_secctx(&context); > > + } > > + } > > + > > + audit_buffer_aux_end(ab); > > + } > > + return; > > + > > +error_path: > > + audit_panic("error in audit_log_object_context"); > > This moves the audit_panic around, so certain operations are not > done before the call. I am currently not sure of the implications. Short version: It's okay. Longer version: The audit_panic() call is either going to panic the kernel (NOT the default), do a pr_err(), or essentially be a no-op. In the case of the full blown kernel panic we don't really care, the system is going to die before there is any chance of this record in progress getting logged. In the case of a pr_err() or no-op the key part is making sure we leave the audit_buffer in a consistent state so that we preserve whatever information is already present. In the !lsm_multiple_contexts case we simply return without making any changes to the audit_buffer so we're good there; in the multiple LSM case we always end the aux record properly (using a "?" when necessary) if an aux record has been successfully created. Feel free to point out a specific scenario that you think looks wrong - I may have missed it - but I believe this code to be correct. > > diff --git a/kernel/auditsc.c b/kernel/auditsc.c > > index 557713954a69..04bf3c04ef3d 100644 > > --- a/kernel/auditsc.c > > +++ b/kernel/auditsc.c > > @@ -1420,18 +1409,10 @@ static void show_special(struct audit_context *context, int *call_panic) > > If pushing audit_panic into audit_log_object_context() is acceptable then this call_panic arg is > no longer needed. The same goes for the call_panic arg in audit_log_name(). And call_panic can > be dropped from audit_log_exit() Good catch. I suspect this is a vestige from when audit_log_end() used to do the record's skb write to userspace, meaning it was possible that you might get some of the records written to userspace before the system killed itself. Now with all of the queuing involved it's less likely that this would be the case, and even if it does happen in some cases, it's basically a toss up depending on how the system is loaded, the scheduler, etc. -- paul-moore.com -- Linux-audit mailing list Linux-audit@redhat.com https://listman.redhat.com/mailman/listinfo/linux-audit
next prev parent reply other threads:[~2022-04-26 18:57 UTC|newest] Thread overview: 132+ messages / expand[flat|nested] mbox.gz Atom feed top [not found] <20220418145945.38797-1-casey.ref@schaufler-ca.com> 2022-04-18 14:59 ` [PATCH v35 00/29] LSM: Module stacking for AppArmor Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 01/29] integrity: disassociate ima_filter_rule from security_audit_rule Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-21 16:51 ` John Johansen 2022-04-21 16:51 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 02/29] LSM: Infrastructure management of the sock security Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 03/29] LSM: Add the lsmblob data structure Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-26 23:15 ` John Johansen 2022-04-26 23:15 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 04/29] LSM: provide lsm name and id slot mappings Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-21 16:50 ` John Johansen 2022-04-21 16:50 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 05/29] IMA: avoid label collisions with stacked LSMs Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-19 16:50 ` Casey Schaufler 2022-04-20 19:23 ` Mimi Zohar 2022-04-20 21:15 ` Casey Schaufler 2022-04-21 3:22 ` Mimi Zohar 2022-04-21 16:50 ` John Johansen 2022-04-21 16:50 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 06/29] LSM: Use lsmblob in security_audit_rule_match Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-21 16:49 ` John Johansen 2022-04-21 16:49 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 07/29] LSM: Use lsmblob in security_kernel_act_as Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 08/29] LSM: Use lsmblob in security_secctx_to_secid Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-27 0:38 ` John Johansen 2022-04-27 0:38 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 09/29] LSM: Use lsmblob in security_secid_to_secctx Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 10/29] LSM: Use lsmblob in security_ipc_getsecid Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 11/29] LSM: Use lsmblob in security_current_getsecid Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 12/29] LSM: Use lsmblob in security_inode_getsecid Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 13/29] LSM: Use lsmblob in security_cred_getsecid Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 18:02 ` kernel test robot 2022-04-18 18:02 ` kernel test robot 2022-04-19 0:41 ` kernel test robot 2022-04-19 0:41 ` kernel test robot 2022-04-19 0:51 ` kernel test robot 2022-04-19 0:51 ` kernel test robot 2022-04-18 14:59 ` [PATCH v35 14/29] LSM: Specify which LSM to display Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 15/29] LSM: Ensure the correct LSM context releaser Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 16/29] LSM: Use lsmcontext in security_secid_to_secctx Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 17/29] LSM: Use lsmcontext in security_inode_getsecctx Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 18/29] LSM: security_secid_to_secctx in netlink netfilter Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 19/29] NET: Store LSM netlabel data in a lsmblob Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 20/29] binder: Pass LSM identifier for confirmation Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 21/29] LSM: Extend security_secid_to_secctx to include module selection Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-25 23:32 ` John Johansen 2022-04-25 23:32 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 22/29] Audit: Keep multiple LSM data in audit_names Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-25 23:32 ` John Johansen 2022-04-25 23:32 ` John Johansen 2022-04-26 17:57 ` Paul Moore 2022-04-26 17:57 ` Paul Moore 2022-04-18 14:59 ` [PATCH v35 23/29] Audit: Create audit_stamp structure Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-25 23:31 ` John Johansen 2022-04-25 23:31 ` John Johansen 2022-04-26 18:03 ` Paul Moore 2022-04-26 18:03 ` Paul Moore 2022-04-26 18:58 ` John Johansen 2022-04-26 18:58 ` John Johansen 2022-04-26 19:18 ` Paul Moore 2022-04-26 19:18 ` Paul Moore 2022-04-27 15:49 ` Casey Schaufler 2022-04-27 15:49 ` Casey Schaufler 2022-04-27 16:02 ` Paul Moore 2022-04-27 16:02 ` Paul Moore 2022-04-27 20:55 ` Casey Schaufler 2022-04-27 20:55 ` Casey Schaufler 2022-04-18 14:59 ` [PATCH v35 24/29] LSM: Add a function to report multiple LSMs Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-22 16:26 ` Paul Moore 2022-04-22 16:26 ` Paul Moore 2022-04-25 23:33 ` John Johansen 2022-04-25 23:33 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 25/29] Audit: Allow multiple records in an audit_buffer Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-22 16:27 ` Paul Moore 2022-04-22 16:27 ` Paul Moore 2022-04-26 1:06 ` John Johansen 2022-04-26 1:06 ` John Johansen 2022-04-26 18:12 ` Paul Moore 2022-04-26 18:12 ` Paul Moore 2022-04-26 19:01 ` John Johansen 2022-04-26 19:01 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 26/29] Audit: Add record for multiple task security contexts Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-22 16:28 ` Paul Moore 2022-04-22 16:28 ` Paul Moore 2022-04-26 1:08 ` John Johansen 2022-04-26 1:08 ` John Johansen 2022-04-26 18:15 ` Paul Moore 2022-04-26 18:15 ` Paul Moore 2022-04-26 19:07 ` John Johansen 2022-04-26 19:07 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 27/29] Audit: Add record for multiple object contexts Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-22 16:29 ` Paul Moore 2022-04-22 16:29 ` Paul Moore 2022-04-26 3:37 ` John Johansen 2022-04-26 3:37 ` John Johansen 2022-04-26 18:57 ` Paul Moore [this message] 2022-04-26 18:57 ` Paul Moore 2022-04-26 19:24 ` John Johansen 2022-04-26 19:24 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 28/29] LSM: Add /proc attr entry for full LSM context Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler 2022-04-22 8:37 ` John Johansen 2022-04-22 8:37 ` John Johansen 2022-04-18 14:59 ` [PATCH v35 29/29] AppArmor: Remove the exclusive flag Casey Schaufler 2022-04-18 14:59 ` Casey Schaufler
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=CAHC9VhTXgBTH+7ny-fcMP_HC1ojA1ass38PGHS2tJny0bCGXzA@mail.gmail.com \ --to=paul@paul-moore.com \ --cc=casey.schaufler@intel.com \ --cc=casey@schaufler-ca.com \ --cc=jmorris@namei.org \ --cc=john.johansen@canonical.com \ --cc=keescook@chromium.org \ --cc=linux-audit@redhat.com \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-security-module@vger.kernel.org \ --cc=penguin-kernel@i-love.sakura.ne.jp \ --cc=selinux@vger.kernel.org \ --cc=stephen.smalley.work@gmail.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.