On 07/11/16 16:05, Topi Miettinen wrote:
> On 07/11/16 15:25, Serge E. Hallyn wrote:
>> Quoting Topi Miettinen (toiwoton@xxxxxxxxx):
>>> There are many basic ways to control processes, including capabilities,
>>> cgroups and resource limits. However, there are far fewer ways to find
>>> out useful values for the limits, except blind trial and error.
>>>
>>> Currently, there is no way to know which capabilities are actually used.
>>> Even the source code is only implicit, in-depth knowledge of each
>>> capability must be used when analyzing a program to judge which
>>> capabilities the program will exercise.
>>>
>>> Generate an audit message at system call exit, when capabilities are used.
>>> This can then be used to configure capability sets for services by a
>>> software developer, maintainer or system administrator.
>>>
>>> Test case demonstrating basic capability monitoring with the new
>>> message types 1330 and 1331 and how the cgroups are displayed (boot to
>>> rdshell):
>>
>> Thanks, Topi, I'll find time this week to look this over in detail.
>>
>> How much chattier does this make the syslog/journald during a regular
>> boot? I was thinking "this is audit, we can choose what messages
>> will show up", but I guess that' sonly what auditd actually listens to,
>> not what kernel emits? (sorry i've not looked at audit in a long
>> time). Drat, that makes it seem like tracepoints would be better
>> after all. But let's see how much it addes to the noise.
>
> For example "loadkeys" causes thousands of entries. :-( I'm checking how
> to avoid audit message rate limiting, now some messages are lost.
>
> It's still too easy to drown the logs with noise. That could be limited
> a lot by emitting a message only when the capability is used for the
> first time. But the question is how to define where to start counting
> (fork, exec, and/or setpcap?). I'm also not sure if that is the right
> way to log, since the first use of a capability could be expected and an
> innocent one, but then the 100th one could be malicious.
>
> It's also very complex and error-prone to collect a capability mask from
> audit logs, which was my original goal.
What if only a summary of capabilities was logged at task exit? That
should make the log volume reasonable.
-Topi
>
> -Topi
>
>>
>>> BusyBox v1.22.1 (Debian 1:1.22.0-19) built-in shell (ash)
>>> Enter 'help' for a list of built-in commands.
>>>
>>> (initramfs) cd /sys/fs
>>> (initramfs) mount -t cgroup2 cgroup cgroup
>>> [ 12.343152] audit_printk_skb: 5886 callbacks suppressed
>>> [ 12.355214] audit: type=1300 audit(1468234317.100:518): arch=c000003e syscall=165 success=yes exit=0 a0=7fffe1e9ae2d a1=7fffe1e9ae34 a2=7fffe1e9ae25 a3=8000 items=0 ppid=469 pid=470 auid=4294967295 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=ttyS0 ses=4294967295 comm="mount" exe="/bin/mount" key=(null)
>>> [ 12.414853] audit: type=1327 audit(1468234317.100:518): proctitle=6D6F756E74002D74006367726F757032006367726F7570006367726F7570
>>> [ 12.438338] audit: type=1330 audit(1468234317.100:518): cap_used=0000000000200000
>>> [ 12.453893] audit: type=1331 audit(1468234317.100:518): cgroups=:/;
>>> (initramfs) cd cgroup
>>> (initramfs) mkdir test; cd test
>>> [ 17.335625] audit: type=1300 audit(1468234322.092:519): arch=c000003e syscall=83 success=yes exit=0 a0=7ffddfd75e29 a1=1ff a2=0 a3=1e2 items=0 ppid=469 pid=471 auid=4294967295 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=ttyS0 ses=4294967295 comm="mkdir" exe="/bin/mkdir" key=(null)
>>> [ 17.392686] audit: type=1327 audit(1468234322.092:519): proctitle=6D6B6469720074657374
>>> [ 17.409404] audit: type=1330 audit(1468234322.092:519): cap_used=0000000000000002
>>> [ 17.425404] audit: type=1331 audit(1468234322.092:519): cgroups=:/;
>>> (initramfs) echo $$ >cgroup.procs
>>> (initramfs) mknod /dev/z_$$ c 1 2
>>> [ 28.385681] audit: type=1300 audit(1468234333.144:520): arch=c000003e syscall=133 success=yes exit=0 a0=7ffe16324e11 a1=21b6 a2=102 a3=5c9 items=0 ppid=469 pid=472 auid=4294967295 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=ttyS0 ses=4294967295 comm="mknod" exe="/bin/mknod" key=(null)
>>> [ 28.443674] audit: type=1327 audit(1468234333.144:520): proctitle=6D6B6E6F64002F6465762F7A5F343639006300310032
>>> [ 28.465888] audit: type=1330 audit(1468234333.144:520): cap_used=0000000008000000
>>> [ 28.482080] audit: type=1331 audit(1468234333.144:520): cgroups=:/test;
>>> (initramfs) chown 1234 /dev/z_*
>>> [ 34.772992] audit: type=1300 audit(1468234339.532:521): arch=c000003e syscall=92 success=yes exit=0 a0=7ffd0b563e17 a1=4d2 a2=0 a3=60a items=0 ppid=469 pid=473 auid=4294967295 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=ttyS0 ses=4294967295 comm="chown" exe="/bin/chown" key=(null)
>>> [ 34.828569] audit: type=1327 audit(1468234339.532:521): proctitle=63686F776E0031323334002F6465762F7A5F343639
>>> [ 34.848747] audit: type=1330 audit(1468234339.532:521): cap_used=0000000000000001
>>> [ 34.864404] audit: type=1331 audit(1468234339.532:521): cgroups=:/test;
>>>
>>> Signed-off-by: Topi Miettinen <toiwoton@xxxxxxxxx>
>>> ---
>>> include/linux/audit.h | 4 +++
>>> include/linux/cgroup.h | 2 ++
>>> include/uapi/linux/audit.h | 2 ++
>>> kernel/audit.c | 7 +++---
>>> kernel/audit.h | 1 +
>>> kernel/auditsc.c | 28 ++++++++++++++++++++-
>>> kernel/capability.c | 5 ++--
>>> kernel/cgroup.c | 62 ++++++++++++++++++++++++++++++++++++++++++++++
>>> 8 files changed, 105 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/include/linux/audit.h b/include/linux/audit.h
>>> index e38e3fc..971cb2e 100644
>>> --- a/include/linux/audit.h
>>> +++ b/include/linux/audit.h
>>> @@ -438,6 +438,8 @@ static inline void audit_mmap_fd(int fd, int flags)
>>> __audit_mmap_fd(fd, flags);
>>> }
>>>
>>> +extern void audit_log_cap_use(int cap);
>>> +
>>> extern int audit_n_rules;
>>> extern int audit_signals;
>>> #else /* CONFIG_AUDITSYSCALL */
>>> @@ -545,6 +547,8 @@ static inline void audit_mmap_fd(int fd, int flags)
>>> { }
>>> static inline void audit_ptrace(struct task_struct *t)
>>> { }
>>> +static inline void audit_log_cap_use(int cap)
>>> +{ }
>>> #define audit_n_rules 0
>>> #define audit_signals 0
>>> #endif /* CONFIG_AUDITSYSCALL */
>>> diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
>>> index a20320c..b5dc8aa 100644
>>> --- a/include/linux/cgroup.h
>>> +++ b/include/linux/cgroup.h
>>> @@ -100,6 +100,8 @@ char *task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
>>> int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
>>> int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
>>> struct pid *pid, struct task_struct *tsk);
>>> +struct audit_buffer;
>>> +void audit_cgroup_list(struct audit_buffer *ab);
>>>
>>> void cgroup_fork(struct task_struct *p);
>>> extern int cgroup_can_fork(struct task_struct *p);
>>> diff --git a/include/uapi/linux/audit.h b/include/uapi/linux/audit.h
>>> index d820aa9..c1ae016 100644
>>> --- a/include/uapi/linux/audit.h
>>> +++ b/include/uapi/linux/audit.h
>>> @@ -111,6 +111,8 @@
>>> #define AUDIT_PROCTITLE 1327 /* Proctitle emit event */
>>> #define AUDIT_FEATURE_CHANGE 1328 /* audit log listing feature changes */
>>> #define AUDIT_REPLACE 1329 /* Replace auditd if this packet unanswerd */
>>> +#define AUDIT_CAPABILITY 1330 /* Record showing capability use */
>>> +#define AUDIT_CGROUP 1331 /* Record showing cgroups */
>>>
>>> #define AUDIT_AVC 1400 /* SE Linux avc denial or grant */
>>> #define AUDIT_SELINUX_ERR 1401 /* Internal SE Linux Errors */
>>> diff --git a/kernel/audit.c b/kernel/audit.c
>>> index 8d528f9..98dd920 100644
>>> --- a/kernel/audit.c
>>> +++ b/kernel/audit.c
>>> @@ -54,6 +54,7 @@
>>> #include <linux/kthread.h>
>>> #include <linux/kernel.h>
>>> #include <linux/syscalls.h>
>>> +#include <linux/cgroup.h>
>>>
>>> #include <linux/audit.h>
>>>
>>> @@ -1682,7 +1683,7 @@ void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
>>> {
>>> int i;
>>>
>>> - audit_log_format(ab, " %s=", prefix);
>>> + audit_log_format(ab, "%s=", prefix);
>>> CAP_FOR_EACH_U32(i) {
>>> audit_log_format(ab, "%08x",
>>> cap->cap[CAP_LAST_U32 - i]);
>>> @@ -1696,11 +1697,11 @@ static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
>>> int log = 0;
>>>
>>> if (!cap_isclear(*perm)) {
>>> - audit_log_cap(ab, "cap_fp", perm);
>>> + audit_log_cap(ab, " cap_fp", perm);
>>> log = 1;
>>> }
>>> if (!cap_isclear(*inh)) {
>>> - audit_log_cap(ab, "cap_fi", inh);
>>> + audit_log_cap(ab, " cap_fi", inh);
>>> log = 1;
>>> }
>>>
>>> diff --git a/kernel/audit.h b/kernel/audit.h
>>> index a492f4c..680e8b5 100644
>>> --- a/kernel/audit.h
>>> +++ b/kernel/audit.h
>>> @@ -202,6 +202,7 @@ struct audit_context {
>>> };
>>> int fds[2];
>>> struct audit_proctitle proctitle;
>>> + kernel_cap_t cap_used;
>>> };
>>>
>>> extern u32 audit_ever_enabled;
>>> diff --git a/kernel/auditsc.c b/kernel/auditsc.c
>>> index 2672d10..32c3813 100644
>>> --- a/kernel/auditsc.c
>>> +++ b/kernel/auditsc.c
>>> @@ -197,7 +197,6 @@ static int audit_match_filetype(struct audit_context *ctx, int val)
>>> * References in it _are_ dropped - at the same time we free/drop aux stuff.
>>> */
>>>
>>> -#ifdef CONFIG_AUDIT_TREE
>>> static void audit_set_auditable(struct audit_context *ctx)
>>> {
>>> if (!ctx->prio) {
>>> @@ -206,6 +205,7 @@ static void audit_set_auditable(struct audit_context *ctx)
>>> }
>>> }
>>>
>>> +#ifdef CONFIG_AUDIT_TREE
>>> static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
>>> {
>>> struct audit_tree_refs *p = ctx->trees;
>>> @@ -1439,6 +1439,18 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts
>>>
>>> audit_log_proctitle(tsk, context);
>>>
>>> + ab = audit_log_start(context, GFP_KERNEL, AUDIT_CAPABILITY);
>>> + if (ab) {
>>> + audit_log_cap(ab, "cap_used", &context->cap_used);
>>> + audit_log_end(ab);
>>> + }
>>> + ab = audit_log_start(context, GFP_KERNEL, AUDIT_CGROUP);
>>> + if (ab) {
>>> + audit_log_format(ab, "cgroups=");
>>> + audit_cgroup_list(ab);
>>> + audit_log_end(ab);
>>> + }
>>> +
>>> /* Send end of event record to help user space know we are finished */
>>> ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
>>> if (ab)
>>> @@ -2428,3 +2440,17 @@ struct list_head *audit_killed_trees(void)
>>> return NULL;
>>> return &ctx->killed_trees;
>>> }
>>> +
>>> +void audit_log_cap_use(int cap)
>>> +{
>>> + struct audit_context *context = current->audit_context;
>>> +
>>> + if (context) {
>>> + cap_raise(context->cap_used, cap);
>>> + audit_set_auditable(context);
>>> + } else {
>>> + audit_log(NULL, GFP_NOFS, AUDIT_CAPABILITY,
>>> + "cap_used=%d pid=%d no audit_context",
>>> + cap, task_pid_nr(current));
>>> + }
>>> +}
>>> diff --git a/kernel/capability.c b/kernel/capability.c
>>> index 45432b5..d45d5b1 100644
>>> --- a/kernel/capability.c
>>> +++ b/kernel/capability.c
>>> @@ -366,8 +366,8 @@ bool has_capability_noaudit(struct task_struct *t, int cap)
>>> * @ns: The usernamespace we want the capability in
>>> * @cap: The capability to be tested for
>>> *
>>> - * Return true if the current task has the given superior capability currently
>>> - * available for use, false if not.
>>> + * Return true if the current task has the given superior capability
>>> + * currently available for use, false if not. Write an audit message.
>>> *
>>> * This sets PF_SUPERPRIV on the task if the capability is available on the
>>> * assumption that it's about to be used.
>>> @@ -380,6 +380,7 @@ bool ns_capable(struct user_namespace *ns, int cap)
>>> }
>>>
>>> if (security_capable(current_cred(), ns, cap) == 0) {
>>> + audit_log_cap_use(cap);
>>> current->flags |= PF_SUPERPRIV;
>>> return true;
>>> }
>>> diff --git a/kernel/cgroup.c b/kernel/cgroup.c
>>> index 75c0ff0..1931679 100644
>>> --- a/kernel/cgroup.c
>>> +++ b/kernel/cgroup.c
>>> @@ -63,6 +63,7 @@
>>> #include <linux/nsproxy.h>
>>> #include <linux/proc_ns.h>
>>> #include <net/sock.h>
>>> +#include <linux/audit.h>
>>>
>>> /*
>>> * pidlists linger the following amount before being destroyed. The goal
>>> @@ -5789,6 +5790,67 @@ out:
>>> return retval;
>>> }
>>>
>>> +/*
>>> + * audit_cgroup_list()
>>> + * - Print task's cgroup paths with audit_log_format()
>>> + * - Used for capability audit logging
>>> + * - Otherwise very similar to proc_cgroup_show().
>>> + */
>>> +void audit_cgroup_list(struct audit_buffer *ab)
>>> +{
>>> + char *buf, *path;
>>> + struct cgroup_root *root;
>>> +
>>> + buf = kmalloc(PATH_MAX, GFP_NOFS);
>>> + if (!buf)
>>> + return;
>>> +
>>> + mutex_lock(&cgroup_mutex);
>>> + spin_lock_irq(&css_set_lock);
>>> +
>>> + for_each_root(root) {
>>> + struct cgroup_subsys *ss;
>>> + struct cgroup *cgrp;
>>> + int ssid, count = 0;
>>> +
>>> + if (root == &cgrp_dfl_root && !cgrp_dfl_visible)
>>> + continue;
>>> +
>>> + if (root != &cgrp_dfl_root)
>>> + for_each_subsys(ss, ssid)
>>> + if (root->subsys_mask & (1 << ssid))
>>> + audit_log_format(ab, "%s%s",
>>> + count++ ? "," : "",
>>> + ss->legacy_name);
>>> + if (strlen(root->name))
>>> + audit_log_format(ab, "%sname=%s", count ? "," : "",
>>> + root->name);
>>> + audit_log_format(ab, ":");
>>> +
>>> + cgrp = task_cgroup_from_root(current, root);
>>> +
>>> + if (cgroup_on_dfl(cgrp) || !(current->flags & PF_EXITING)) {
>>> + path = cgroup_path_ns_locked(cgrp, buf, PATH_MAX,
>>> + current->nsproxy->cgroup_ns);
>>> + if (!path)
>>> + goto out_unlock;
>>> + } else
>>> + path = "/";
>>> +
>>> + audit_log_format(ab, "%s", path);
>>> +
>>> + if (cgroup_on_dfl(cgrp) && cgroup_is_dead(cgrp))
>>> + audit_log_format(ab, " (deleted);");
>>> + else
>>> + audit_log_format(ab, ";");
>>> + }
>>> +
>>> +out_unlock:
>>> + spin_unlock_irq(&css_set_lock);
>>> + mutex_unlock(&cgroup_mutex);
>>> + kfree(buf);
>>> +}
>>> +
>>> /* Display information about each subsystem and each hierarchy */
>>> static int proc_cgroupstats_show(struct seq_file *m, void *v)
>>> {
>>> --
>>> 2.8.1
>
--
To unsubscribe from this list: send the line "unsubscribe cgroups" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
