2011-12-22 04:17:04 +08:00
|
|
|
/*
|
|
|
|
* Yama Linux Security Module
|
|
|
|
*
|
|
|
|
* Author: Kees Cook <keescook@chromium.org>
|
|
|
|
*
|
|
|
|
* Copyright (C) 2010 Canonical, Ltd.
|
|
|
|
* Copyright (C) 2011 The Chromium OS Authors.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or modify
|
|
|
|
* it under the terms of the GNU General Public License version 2, as
|
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
2015-05-03 06:10:46 +08:00
|
|
|
#include <linux/lsm_hooks.h>
|
2011-12-22 04:17:04 +08:00
|
|
|
#include <linux/sysctl.h>
|
|
|
|
#include <linux/ptrace.h>
|
|
|
|
#include <linux/prctl.h>
|
|
|
|
#include <linux/ratelimit.h>
|
2012-11-20 07:21:26 +08:00
|
|
|
#include <linux/workqueue.h>
|
2016-04-21 06:46:26 +08:00
|
|
|
#include <linux/string_helpers.h>
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
#include <linux/task_work.h>
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/spinlock.h>
|
2011-12-22 04:17:04 +08:00
|
|
|
|
2012-04-17 02:56:45 +08:00
|
|
|
#define YAMA_SCOPE_DISABLED 0
|
|
|
|
#define YAMA_SCOPE_RELATIONAL 1
|
|
|
|
#define YAMA_SCOPE_CAPABILITY 2
|
|
|
|
#define YAMA_SCOPE_NO_ATTACH 3
|
|
|
|
|
|
|
|
static int ptrace_scope = YAMA_SCOPE_RELATIONAL;
|
2011-12-22 04:17:04 +08:00
|
|
|
|
|
|
|
/* describe a ptrace relationship for potential exception */
|
|
|
|
struct ptrace_relation {
|
|
|
|
struct task_struct *tracer;
|
|
|
|
struct task_struct *tracee;
|
2012-11-20 07:21:26 +08:00
|
|
|
bool invalid;
|
2011-12-22 04:17:04 +08:00
|
|
|
struct list_head node;
|
2012-10-19 05:53:58 +08:00
|
|
|
struct rcu_head rcu;
|
2011-12-22 04:17:04 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
static LIST_HEAD(ptracer_relations);
|
|
|
|
static DEFINE_SPINLOCK(ptracer_relations_lock);
|
|
|
|
|
2012-11-20 07:21:26 +08:00
|
|
|
static void yama_relation_cleanup(struct work_struct *work);
|
|
|
|
static DECLARE_WORK(yama_relation_work, yama_relation_cleanup);
|
|
|
|
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
struct access_report_info {
|
|
|
|
struct callback_head work;
|
|
|
|
const char *access;
|
|
|
|
struct task_struct *target;
|
|
|
|
struct task_struct *agent;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void __report_access(struct callback_head *work)
|
2016-04-21 06:46:26 +08:00
|
|
|
{
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
struct access_report_info *info =
|
|
|
|
container_of(work, struct access_report_info, work);
|
2016-04-21 06:46:26 +08:00
|
|
|
char *target_cmd, *agent_cmd;
|
|
|
|
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
target_cmd = kstrdup_quotable_cmdline(info->target, GFP_KERNEL);
|
|
|
|
agent_cmd = kstrdup_quotable_cmdline(info->agent, GFP_KERNEL);
|
2016-04-21 06:46:26 +08:00
|
|
|
|
|
|
|
pr_notice_ratelimited(
|
|
|
|
"ptrace %s of \"%s\"[%d] was attempted by \"%s\"[%d]\n",
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
info->access, target_cmd, info->target->pid, agent_cmd,
|
|
|
|
info->agent->pid);
|
2016-04-21 06:46:26 +08:00
|
|
|
|
|
|
|
kfree(agent_cmd);
|
|
|
|
kfree(target_cmd);
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
|
|
|
|
put_task_struct(info->agent);
|
|
|
|
put_task_struct(info->target);
|
|
|
|
kfree(info);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* defers execution because cmdline access can sleep */
|
|
|
|
static void report_access(const char *access, struct task_struct *target,
|
|
|
|
struct task_struct *agent)
|
|
|
|
{
|
|
|
|
struct access_report_info *info;
|
|
|
|
char agent_comm[sizeof(agent->comm)];
|
|
|
|
|
|
|
|
assert_spin_locked(&target->alloc_lock); /* for target->comm */
|
|
|
|
|
|
|
|
if (current->flags & PF_KTHREAD) {
|
|
|
|
/* I don't think kthreads call task_work_run() before exiting.
|
|
|
|
* Imagine angry ranting about procfs here.
|
|
|
|
*/
|
|
|
|
pr_notice_ratelimited(
|
|
|
|
"ptrace %s of \"%s\"[%d] was attempted by \"%s\"[%d]\n",
|
|
|
|
access, target->comm, target->pid,
|
|
|
|
get_task_comm(agent_comm, agent), agent->pid);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
info = kmalloc(sizeof(*info), GFP_ATOMIC);
|
|
|
|
if (!info)
|
|
|
|
return;
|
|
|
|
init_task_work(&info->work, __report_access);
|
|
|
|
get_task_struct(target);
|
|
|
|
get_task_struct(agent);
|
|
|
|
info->access = access;
|
|
|
|
info->target = target;
|
|
|
|
info->agent = agent;
|
|
|
|
if (task_work_add(current, &info->work, true) == 0)
|
|
|
|
return; /* success */
|
|
|
|
|
|
|
|
WARN(1, "report_access called from exiting task");
|
|
|
|
put_task_struct(target);
|
|
|
|
put_task_struct(agent);
|
|
|
|
kfree(info);
|
2016-04-21 06:46:26 +08:00
|
|
|
}
|
|
|
|
|
2012-11-20 07:21:26 +08:00
|
|
|
/**
|
|
|
|
* yama_relation_cleanup - remove invalid entries from the relation list
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
static void yama_relation_cleanup(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct ptrace_relation *relation;
|
|
|
|
|
|
|
|
spin_lock(&ptracer_relations_lock);
|
|
|
|
rcu_read_lock();
|
|
|
|
list_for_each_entry_rcu(relation, &ptracer_relations, node) {
|
|
|
|
if (relation->invalid) {
|
|
|
|
list_del_rcu(&relation->node);
|
|
|
|
kfree_rcu(relation, rcu);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
spin_unlock(&ptracer_relations_lock);
|
|
|
|
}
|
|
|
|
|
2011-12-22 04:17:04 +08:00
|
|
|
/**
|
|
|
|
* yama_ptracer_add - add/replace an exception for this tracer/tracee pair
|
|
|
|
* @tracer: the task_struct of the process doing the ptrace
|
|
|
|
* @tracee: the task_struct of the process to be ptraced
|
|
|
|
*
|
|
|
|
* Each tracee can have, at most, one tracer registered. Each time this
|
|
|
|
* is called, the prior registered tracer will be replaced for the tracee.
|
|
|
|
*
|
|
|
|
* Returns 0 if relationship was added, -ve on error.
|
|
|
|
*/
|
|
|
|
static int yama_ptracer_add(struct task_struct *tracer,
|
|
|
|
struct task_struct *tracee)
|
|
|
|
{
|
2012-10-19 05:53:58 +08:00
|
|
|
struct ptrace_relation *relation, *added;
|
2011-12-22 04:17:04 +08:00
|
|
|
|
|
|
|
added = kmalloc(sizeof(*added), GFP_KERNEL);
|
|
|
|
if (!added)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2012-10-19 05:53:58 +08:00
|
|
|
added->tracee = tracee;
|
|
|
|
added->tracer = tracer;
|
2012-11-20 07:21:26 +08:00
|
|
|
added->invalid = false;
|
2012-10-19 05:53:58 +08:00
|
|
|
|
2012-11-20 07:21:26 +08:00
|
|
|
spin_lock(&ptracer_relations_lock);
|
2012-10-19 05:53:58 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
list_for_each_entry_rcu(relation, &ptracer_relations, node) {
|
2012-11-20 07:21:26 +08:00
|
|
|
if (relation->invalid)
|
|
|
|
continue;
|
2012-10-19 05:53:58 +08:00
|
|
|
if (relation->tracee == tracee) {
|
|
|
|
list_replace_rcu(&relation->node, &added->node);
|
|
|
|
kfree_rcu(relation, rcu);
|
|
|
|
goto out;
|
2011-12-22 04:17:04 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-10-19 05:53:58 +08:00
|
|
|
list_add_rcu(&added->node, &ptracer_relations);
|
2011-12-22 04:17:04 +08:00
|
|
|
|
2012-10-19 05:53:58 +08:00
|
|
|
out:
|
|
|
|
rcu_read_unlock();
|
2012-11-20 07:21:26 +08:00
|
|
|
spin_unlock(&ptracer_relations_lock);
|
2012-10-19 05:53:58 +08:00
|
|
|
return 0;
|
2011-12-22 04:17:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* yama_ptracer_del - remove exceptions related to the given tasks
|
|
|
|
* @tracer: remove any relation where tracer task matches
|
|
|
|
* @tracee: remove any relation where tracee task matches
|
|
|
|
*/
|
|
|
|
static void yama_ptracer_del(struct task_struct *tracer,
|
|
|
|
struct task_struct *tracee)
|
|
|
|
{
|
2012-10-19 05:53:58 +08:00
|
|
|
struct ptrace_relation *relation;
|
2012-11-20 07:21:26 +08:00
|
|
|
bool marked = false;
|
2011-12-22 04:17:04 +08:00
|
|
|
|
2012-10-19 05:53:58 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
list_for_each_entry_rcu(relation, &ptracer_relations, node) {
|
2012-11-20 07:21:26 +08:00
|
|
|
if (relation->invalid)
|
|
|
|
continue;
|
2011-12-22 04:17:04 +08:00
|
|
|
if (relation->tracee == tracee ||
|
2012-02-15 08:48:09 +08:00
|
|
|
(tracer && relation->tracer == tracer)) {
|
2012-11-20 07:21:26 +08:00
|
|
|
relation->invalid = true;
|
|
|
|
marked = true;
|
2011-12-22 04:17:04 +08:00
|
|
|
}
|
2012-10-19 05:53:58 +08:00
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
2012-11-20 07:21:26 +08:00
|
|
|
|
|
|
|
if (marked)
|
|
|
|
schedule_work(&yama_relation_work);
|
2011-12-22 04:17:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* yama_task_free - check for task_pid to remove from exception list
|
|
|
|
* @task: task being removed
|
|
|
|
*/
|
2012-09-05 04:32:13 +08:00
|
|
|
void yama_task_free(struct task_struct *task)
|
2011-12-22 04:17:04 +08:00
|
|
|
{
|
|
|
|
yama_ptracer_del(task, task);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* yama_task_prctl - check for Yama-specific prctl operations
|
|
|
|
* @option: operation
|
|
|
|
* @arg2: argument
|
|
|
|
* @arg3: argument
|
|
|
|
* @arg4: argument
|
|
|
|
* @arg5: argument
|
|
|
|
*
|
|
|
|
* Return 0 on success, -ve on error. -ENOSYS is returned when Yama
|
|
|
|
* does not handle the given option.
|
|
|
|
*/
|
2012-09-05 04:32:13 +08:00
|
|
|
int yama_task_prctl(int option, unsigned long arg2, unsigned long arg3,
|
2011-12-22 04:17:04 +08:00
|
|
|
unsigned long arg4, unsigned long arg5)
|
|
|
|
{
|
2015-05-03 06:11:42 +08:00
|
|
|
int rc = -ENOSYS;
|
2011-12-22 04:17:04 +08:00
|
|
|
struct task_struct *myself = current;
|
|
|
|
|
|
|
|
switch (option) {
|
|
|
|
case PR_SET_PTRACER:
|
|
|
|
/* Since a thread can call prctl(), find the group leader
|
|
|
|
* before calling _add() or _del() on it, since we want
|
|
|
|
* process-level granularity of control. The tracer group
|
|
|
|
* leader checking is handled later when walking the ancestry
|
|
|
|
* at the time of PTRACE_ATTACH check.
|
|
|
|
*/
|
|
|
|
rcu_read_lock();
|
|
|
|
if (!thread_group_leader(myself))
|
|
|
|
myself = rcu_dereference(myself->group_leader);
|
|
|
|
get_task_struct(myself);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
if (arg2 == 0) {
|
|
|
|
yama_ptracer_del(NULL, myself);
|
|
|
|
rc = 0;
|
2012-08-28 02:38:13 +08:00
|
|
|
} else if (arg2 == PR_SET_PTRACER_ANY || (int)arg2 == -1) {
|
2012-02-15 08:48:09 +08:00
|
|
|
rc = yama_ptracer_add(NULL, myself);
|
2011-12-22 04:17:04 +08:00
|
|
|
} else {
|
|
|
|
struct task_struct *tracer;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
tracer = find_task_by_vpid(arg2);
|
|
|
|
if (tracer)
|
|
|
|
get_task_struct(tracer);
|
|
|
|
else
|
|
|
|
rc = -EINVAL;
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
if (tracer) {
|
|
|
|
rc = yama_ptracer_add(tracer, myself);
|
|
|
|
put_task_struct(tracer);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
put_task_struct(myself);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* task_is_descendant - walk up a process family tree looking for a match
|
|
|
|
* @parent: the process to compare against while walking up from child
|
|
|
|
* @child: the process to start from while looking upwards for parent
|
|
|
|
*
|
|
|
|
* Returns 1 if child is a descendant of parent, 0 if not.
|
|
|
|
*/
|
|
|
|
static int task_is_descendant(struct task_struct *parent,
|
|
|
|
struct task_struct *child)
|
|
|
|
{
|
|
|
|
int rc = 0;
|
|
|
|
struct task_struct *walker = child;
|
|
|
|
|
|
|
|
if (!parent || !child)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
if (!thread_group_leader(parent))
|
|
|
|
parent = rcu_dereference(parent->group_leader);
|
|
|
|
while (walker->pid > 0) {
|
|
|
|
if (!thread_group_leader(walker))
|
|
|
|
walker = rcu_dereference(walker->group_leader);
|
|
|
|
if (walker == parent) {
|
|
|
|
rc = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
walker = rcu_dereference(walker->real_parent);
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ptracer_exception_found - tracer registered as exception for this tracee
|
|
|
|
* @tracer: the task_struct of the process attempting ptrace
|
|
|
|
* @tracee: the task_struct of the process to be ptraced
|
|
|
|
*
|
2016-12-03 07:49:43 +08:00
|
|
|
* Returns 1 if tracer has a ptracer exception ancestor for tracee.
|
2011-12-22 04:17:04 +08:00
|
|
|
*/
|
|
|
|
static int ptracer_exception_found(struct task_struct *tracer,
|
|
|
|
struct task_struct *tracee)
|
|
|
|
{
|
|
|
|
int rc = 0;
|
|
|
|
struct ptrace_relation *relation;
|
|
|
|
struct task_struct *parent = NULL;
|
2012-02-15 08:48:09 +08:00
|
|
|
bool found = false;
|
2011-12-22 04:17:04 +08:00
|
|
|
|
|
|
|
rcu_read_lock();
|
2016-12-03 07:49:43 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If there's already an active tracing relationship, then make an
|
|
|
|
* exception for the sake of other accesses, like process_vm_rw().
|
|
|
|
*/
|
|
|
|
parent = ptrace_parent(tracee);
|
|
|
|
if (parent != NULL && same_thread_group(parent, tracer)) {
|
|
|
|
rc = 1;
|
|
|
|
goto unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Look for a PR_SET_PTRACER relationship. */
|
2011-12-22 04:17:04 +08:00
|
|
|
if (!thread_group_leader(tracee))
|
|
|
|
tracee = rcu_dereference(tracee->group_leader);
|
2012-11-20 07:21:26 +08:00
|
|
|
list_for_each_entry_rcu(relation, &ptracer_relations, node) {
|
|
|
|
if (relation->invalid)
|
|
|
|
continue;
|
2011-12-22 04:17:04 +08:00
|
|
|
if (relation->tracee == tracee) {
|
|
|
|
parent = relation->tracer;
|
2012-02-15 08:48:09 +08:00
|
|
|
found = true;
|
2011-12-22 04:17:04 +08:00
|
|
|
break;
|
|
|
|
}
|
2012-11-20 07:21:26 +08:00
|
|
|
}
|
2011-12-22 04:17:04 +08:00
|
|
|
|
2012-02-15 08:48:09 +08:00
|
|
|
if (found && (parent == NULL || task_is_descendant(parent, tracer)))
|
2011-12-22 04:17:04 +08:00
|
|
|
rc = 1;
|
2016-12-03 07:49:43 +08:00
|
|
|
|
|
|
|
unlock:
|
2011-12-22 04:17:04 +08:00
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* yama_ptrace_access_check - validate PTRACE_ATTACH calls
|
|
|
|
* @child: task that current task is attempting to ptrace
|
|
|
|
* @mode: ptrace attach mode
|
|
|
|
*
|
|
|
|
* Returns 0 if following the ptrace is allowed, -ve on error.
|
|
|
|
*/
|
2015-05-03 06:11:42 +08:00
|
|
|
static int yama_ptrace_access_check(struct task_struct *child,
|
2011-12-22 04:17:04 +08:00
|
|
|
unsigned int mode)
|
|
|
|
{
|
2015-05-03 06:11:42 +08:00
|
|
|
int rc = 0;
|
2011-12-22 04:17:04 +08:00
|
|
|
|
|
|
|
/* require ptrace target be a child of ptracer on attach */
|
2016-01-21 07:00:01 +08:00
|
|
|
if (mode & PTRACE_MODE_ATTACH) {
|
2012-04-17 02:56:45 +08:00
|
|
|
switch (ptrace_scope) {
|
|
|
|
case YAMA_SCOPE_DISABLED:
|
|
|
|
/* No additional restrictions. */
|
|
|
|
break;
|
|
|
|
case YAMA_SCOPE_RELATIONAL:
|
2012-07-26 20:05:21 +08:00
|
|
|
rcu_read_lock();
|
2012-04-17 02:56:45 +08:00
|
|
|
if (!task_is_descendant(current, child) &&
|
|
|
|
!ptracer_exception_found(current, child) &&
|
2012-07-26 20:05:21 +08:00
|
|
|
!ns_capable(__task_cred(child)->user_ns, CAP_SYS_PTRACE))
|
2012-04-17 02:56:45 +08:00
|
|
|
rc = -EPERM;
|
2012-07-26 20:05:21 +08:00
|
|
|
rcu_read_unlock();
|
2012-04-17 02:56:45 +08:00
|
|
|
break;
|
|
|
|
case YAMA_SCOPE_CAPABILITY:
|
2012-07-26 20:05:21 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
if (!ns_capable(__task_cred(child)->user_ns, CAP_SYS_PTRACE))
|
2012-04-17 02:56:45 +08:00
|
|
|
rc = -EPERM;
|
2012-07-26 20:05:21 +08:00
|
|
|
rcu_read_unlock();
|
2012-04-17 02:56:45 +08:00
|
|
|
break;
|
|
|
|
case YAMA_SCOPE_NO_ATTACH:
|
|
|
|
default:
|
|
|
|
rc = -EPERM;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2011-12-22 04:17:04 +08:00
|
|
|
|
2016-04-21 06:46:26 +08:00
|
|
|
if (rc && (mode & PTRACE_MODE_NOAUDIT) == 0)
|
|
|
|
report_access("attach", child, current);
|
2011-12-22 04:17:04 +08:00
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2012-08-10 10:01:26 +08:00
|
|
|
/**
|
|
|
|
* yama_ptrace_traceme - validate PTRACE_TRACEME calls
|
|
|
|
* @parent: task that will become the ptracer of the current task
|
|
|
|
*
|
|
|
|
* Returns 0 if following the ptrace is allowed, -ve on error.
|
|
|
|
*/
|
2012-09-05 04:32:13 +08:00
|
|
|
int yama_ptrace_traceme(struct task_struct *parent)
|
2012-08-10 10:01:26 +08:00
|
|
|
{
|
2015-05-03 06:11:42 +08:00
|
|
|
int rc = 0;
|
2012-08-10 10:01:26 +08:00
|
|
|
|
|
|
|
/* Only disallow PTRACE_TRACEME on more aggressive settings. */
|
|
|
|
switch (ptrace_scope) {
|
|
|
|
case YAMA_SCOPE_CAPABILITY:
|
2013-03-21 17:30:41 +08:00
|
|
|
if (!has_ns_capability(parent, current_user_ns(), CAP_SYS_PTRACE))
|
2012-08-10 10:01:26 +08:00
|
|
|
rc = -EPERM;
|
|
|
|
break;
|
|
|
|
case YAMA_SCOPE_NO_ATTACH:
|
|
|
|
rc = -EPERM;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
if (rc) {
|
|
|
|
task_lock(current);
|
2016-04-21 06:46:26 +08:00
|
|
|
report_access("traceme", current, parent);
|
Yama: fix double-spinlock and user access in atomic context
Commit 8a56038c2aef ("Yama: consolidate error reporting") causes lockups
when someone hits a Yama denial. Call chain:
process_vm_readv -> process_vm_rw -> process_vm_rw_core -> mm_access
-> ptrace_may_access
task_lock(...) is taken
__ptrace_may_access -> security_ptrace_access_check
-> yama_ptrace_access_check -> report_access -> kstrdup_quotable_cmdline
-> get_cmdline -> access_process_vm -> get_task_mm
task_lock(...) is taken again
task_lock(p) just calls spin_lock(&p->alloc_lock), so at this point,
spin_lock() is called on a lock that is already held by the current
process.
Also: Since the alloc_lock is a spinlock, sleeping inside
security_ptrace_access_check hooks is probably not allowed at all? So it's
not even possible to print the cmdline from in there because that might
involve paging in userspace memory.
It would be tempting to rewrite ptrace_may_access() to drop the alloc_lock
before calling the LSM, but even then, ptrace_may_access() itself might be
called from various contexts in which you're not allowed to sleep; for
example, as far as I understand, to be able to hold a reference to another
task, usually an RCU read lock will be taken (see e.g. kcmp() and
get_robust_list()), so that also prohibits sleeping. (And using e.g. FUSE,
a user can cause pagefault handling to take arbitrary amounts of time -
see https://bugs.chromium.org/p/project-zero/issues/detail?id=808.)
Therefore, AFAIK, in order to print the name of a process below
security_ptrace_access_check(), you'd have to either grab a reference to
the mm_struct and defer the access violation reporting or just use the
"comm" value that's stored in kernelspace and accessible without big
complications. (Or you could try to use some kind of atomic remote VM
access that fails if the memory isn't paged in, similar to
copy_from_user_inatomic(), and if necessary fall back to comm, but
that'd be kind of ugly because the comm/cmdline choice would look
pretty random to the user.)
Fix it by deferring reporting of the access violation until current
exits kernelspace the next time.
v2: Don't oops on PTRACE_TRACEME, call report_access under
task_lock(current). Also fix nonsensical comment. And don't use
GPF_ATOMIC for memory allocation with no locks held.
This patch is tested both for ptrace attach and ptrace traceme.
Fixes: 8a56038c2aef ("Yama: consolidate error reporting")
Signed-off-by: Jann Horn <jann@thejh.net>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: James Morris <james.l.morris@oracle.com>
2016-05-22 12:01:34 +08:00
|
|
|
task_unlock(current);
|
|
|
|
}
|
2012-08-10 10:01:26 +08:00
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2015-05-03 06:11:42 +08:00
|
|
|
static struct security_hook_list yama_hooks[] = {
|
2015-05-03 06:11:36 +08:00
|
|
|
LSM_HOOK_INIT(ptrace_access_check, yama_ptrace_access_check),
|
|
|
|
LSM_HOOK_INIT(ptrace_traceme, yama_ptrace_traceme),
|
|
|
|
LSM_HOOK_INIT(task_prctl, yama_task_prctl),
|
|
|
|
LSM_HOOK_INIT(task_free, yama_task_free),
|
2011-12-22 04:17:04 +08:00
|
|
|
};
|
2015-05-03 06:11:42 +08:00
|
|
|
|
2011-12-22 04:17:04 +08:00
|
|
|
#ifdef CONFIG_SYSCTL
|
2012-04-17 02:56:45 +08:00
|
|
|
static int yama_dointvec_minmax(struct ctl_table *table, int write,
|
|
|
|
void __user *buffer, size_t *lenp, loff_t *ppos)
|
|
|
|
{
|
2013-02-28 00:37:56 +08:00
|
|
|
struct ctl_table table_copy;
|
2012-04-17 02:56:45 +08:00
|
|
|
|
|
|
|
if (write && !capable(CAP_SYS_PTRACE))
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
/* Lock the max value if it ever gets set. */
|
2013-02-28 00:37:56 +08:00
|
|
|
table_copy = *table;
|
|
|
|
if (*(int *)table_copy.data == *(int *)table_copy.extra2)
|
|
|
|
table_copy.extra1 = table_copy.extra2;
|
2012-04-17 02:56:45 +08:00
|
|
|
|
2013-02-28 00:37:56 +08:00
|
|
|
return proc_dointvec_minmax(&table_copy, write, buffer, lenp, ppos);
|
2012-04-17 02:56:45 +08:00
|
|
|
}
|
|
|
|
|
2011-12-22 04:17:04 +08:00
|
|
|
static int zero;
|
2012-04-17 02:56:45 +08:00
|
|
|
static int max_scope = YAMA_SCOPE_NO_ATTACH;
|
2011-12-22 04:17:04 +08:00
|
|
|
|
|
|
|
struct ctl_path yama_sysctl_path[] = {
|
|
|
|
{ .procname = "kernel", },
|
|
|
|
{ .procname = "yama", },
|
|
|
|
{ }
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct ctl_table yama_sysctl_table[] = {
|
|
|
|
{
|
|
|
|
.procname = "ptrace_scope",
|
|
|
|
.data = &ptrace_scope,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2012-04-17 02:56:45 +08:00
|
|
|
.proc_handler = yama_dointvec_minmax,
|
2011-12-22 04:17:04 +08:00
|
|
|
.extra1 = &zero,
|
2012-04-17 02:56:45 +08:00
|
|
|
.extra2 = &max_scope,
|
2011-12-22 04:17:04 +08:00
|
|
|
},
|
|
|
|
{ }
|
|
|
|
};
|
2015-07-24 09:02:48 +08:00
|
|
|
static void __init yama_init_sysctl(void)
|
2011-12-22 04:17:04 +08:00
|
|
|
{
|
|
|
|
if (!register_sysctl_paths(yama_sysctl_path, yama_sysctl_table))
|
|
|
|
panic("Yama: sysctl registration failed.\n");
|
|
|
|
}
|
2015-07-24 09:02:48 +08:00
|
|
|
#else
|
|
|
|
static inline void yama_init_sysctl(void) { }
|
|
|
|
#endif /* CONFIG_SYSCTL */
|
2011-12-22 04:17:04 +08:00
|
|
|
|
2015-07-24 09:02:48 +08:00
|
|
|
void __init yama_add_hooks(void)
|
|
|
|
{
|
|
|
|
pr_info("Yama: becoming mindful.\n");
|
|
|
|
security_add_hooks(yama_hooks, ARRAY_SIZE(yama_hooks));
|
|
|
|
yama_init_sysctl();
|
|
|
|
}
|