WARNING - This is a work in progress and should not be merged anywhere important. It is almost surely not complete, and while it probably compiles it likely hasn't been booted and will do terrible things. You have been warned. This patch adds basic auditing to io_uring operations, regardless of their context. This is accomplished by allocating audit_context structures for the io-wq worker and io_uring SQPOLL kernel threads as well as explicitly auditing the io_uring operations in io_issue_sqe(). The io_uring operations are audited using a new AUDIT_URINGOP record, an example is shown below: % <TODO - insert AUDIT_URINGOP record example> Thanks to Richard Guy Briggs for review and feedback. Signed-off-by: Paul Moore <paul(a)paul-moore.com&gt; --- v2: - added dummy funcs for audit_uring_{entry,exit}() - replaced opcode checks in io_issue_sqe() with audit_skip checks - moved fastpath checks into audit_uring_{entry,exit}() - audit_log_uring() uses GFP_ATOMIC - don't record the arch in __audit_uring_entry() v1: - initial draft --- fs/io-wq.c | 4 + fs/io_uring.c | 55 ++++++++++++-- include/linux/audit.h | 26 +++++++ include/uapi/linux/audit.h | 1 kernel/audit.h | 2 + kernel/auditsc.c | 174 ++++++++++++++++++++++++++++++++++++++++++++ 6 files changed, 256 insertions(+), 6 deletions(-) diff --git a/fs/io-wq.c b/fs/io-wq.c index 12fc19353bb0..fb928bac0bf2 100644 --- a/fs/io-wq.c +++ b/fs/io-wq.c @@ -14,6 +14,7 @@ #include <linux/rculist_nulls.h> #include <linux/cpu.h> #include <linux/tracehook.h> +#include <linux/audit.h> #include "io-wq.h" @@ -552,6 +553,8 @@ static int io_wqe_worker(void *data) snprintf(buf, sizeof(buf), "iou-wrk-%d", wq->task->pid); set_task_comm(current, buf); + audit_alloc_kernel(current); + while (!test_bit(IO_WQ_BIT_EXIT, &wq->state)) { long ret; @@ -586,6 +589,7 @@ static int io_wqe_worker(void *data) io_worker_handle_work(worker); } + audit_free(current); io_worker_exit(worker); return 0; } diff --git a/fs/io_uring.c b/fs/io_uring.c index bf548af0426c..b407a6ea1779 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -78,6 +78,7 @@ #include <linux/task_work.h> #include <linux/pagemap.h> #include <linux/io_uring.h> +#include <linux/audit.h> #define CREATE_TRACE_POINTS #include <trace/events/io_uring.h> @@ -896,6 +897,8 @@ struct io_op_def { unsigned needs_async_setup : 1; /* should block plug */ unsigned plug : 1; + /* skip auditing */ + unsigned audit_skip : 1; /* size of async data needed, if any */ unsigned short async_size; }; @@ -909,6 +912,7 @@ static const struct io_op_def io_op_defs[] = { .buffer_select = 1, .needs_async_setup = 1, .plug = 1, + .audit_skip = 1, .async_size = sizeof(struct io_async_rw), }, [IORING_OP_WRITEV] = { @@ -918,16 +922,19 @@ static const struct io_op_def io_op_defs[] = { .pollout = 1, .needs_async_setup = 1, .plug = 1, + .audit_skip = 1, .async_size = sizeof(struct io_async_rw), }, [IORING_OP_FSYNC] = { .needs_file = 1, + .audit_skip = 1, }, [IORING_OP_READ_FIXED] = { .needs_file = 1, .unbound_nonreg_file = 1, .pollin = 1, .plug = 1, + .audit_skip = 1, .async_size = sizeof(struct io_async_rw), }, [IORING_OP_WRITE_FIXED] = { @@ -936,15 +943,20 @@ static const struct io_op_def io_op_defs[] = { .unbound_nonreg_file = 1, .pollout = 1, .plug = 1, + .audit_skip = 1, .async_size = sizeof(struct io_async_rw), }, [IORING_OP_POLL_ADD] = { .needs_file = 1, .unbound_nonreg_file = 1, + .audit_skip = 1, + }, + [IORING_OP_POLL_REMOVE] = { + .audit_skip = 1, }, - [IORING_OP_POLL_REMOVE] = {}, [IORING_OP_SYNC_FILE_RANGE] = { .needs_file = 1, + .audit_skip = 1, }, [IORING_OP_SENDMSG] = { .needs_file = 1, @@ -962,18 +974,23 @@ static const struct io_op_def io_op_defs[] = { .async_size = sizeof(struct io_async_msghdr), }, [IORING_OP_TIMEOUT] = { + .audit_skip = 1, .async_size = sizeof(struct io_timeout_data), }, [IORING_OP_TIMEOUT_REMOVE] = { /* used by timeout updates' prep() */ + .audit_skip = 1, }, [IORING_OP_ACCEPT] = { .needs_file = 1, .unbound_nonreg_file = 1, .pollin = 1, }, - [IORING_OP_ASYNC_CANCEL] = {}, + [IORING_OP_ASYNC_CANCEL] = { + .audit_skip = 1, + }, [IORING_OP_LINK_TIMEOUT] = { + .audit_skip = 1, .async_size = sizeof(struct io_timeout_data), }, [IORING_OP_CONNECT] = { @@ -988,14 +1005,19 @@ static const struct io_op_def io_op_defs[] = { }, [IORING_OP_OPENAT] = {}, [IORING_OP_CLOSE] = {}, - [IORING_OP_FILES_UPDATE] = {}, - [IORING_OP_STATX] = {}, + [IORING_OP_FILES_UPDATE] = { + .audit_skip = 1, + }, + [IORING_OP_STATX] = { + .audit_skip = 1, + }, [IORING_OP_READ] = { .needs_file = 1, .unbound_nonreg_file = 1, .pollin = 1, .buffer_select = 1, .plug = 1, + .audit_skip = 1, .async_size = sizeof(struct io_async_rw), }, [IORING_OP_WRITE] = { @@ -1003,39 +1025,50 @@ static const struct io_op_def io_op_defs[] = { .unbound_nonreg_file = 1, .pollout = 1, .plug = 1, + .audit_skip = 1, .async_size = sizeof(struct io_async_rw), }, [IORING_OP_FADVISE] = { .needs_file = 1, + .audit_skip = 1, }, [IORING_OP_MADVISE] = {}, [IORING_OP_SEND] = { .needs_file = 1, .unbound_nonreg_file = 1, .pollout = 1, + .audit_skip = 1, }, [IORING_OP_RECV] = { .needs_file = 1, .unbound_nonreg_file = 1, .pollin = 1, .buffer_select = 1, + .audit_skip = 1, }, [IORING_OP_OPENAT2] = { }, [IORING_OP_EPOLL_CTL] = { .unbound_nonreg_file = 1, + .audit_skip = 1, }, [IORING_OP_SPLICE] = { .needs_file = 1, .hash_reg_file = 1, .unbound_nonreg_file = 1, + .audit_skip = 1, + }, + [IORING_OP_PROVIDE_BUFFERS] = { + .audit_skip = 1, + }, + [IORING_OP_REMOVE_BUFFERS] = { + .audit_skip = 1, }, - [IORING_OP_PROVIDE_BUFFERS] = {}, - [IORING_OP_REMOVE_BUFFERS] = {}, [IORING_OP_TEE] = { .needs_file = 1, .hash_reg_file = 1, .unbound_nonreg_file = 1, + .audit_skip = 1, }, [IORING_OP_SHUTDOWN] = { .needs_file = 1, @@ -6166,6 +6199,9 @@ static int io_issue_sqe(struct io_kiocb *req, unsigned int issue_flags) if ((req->flags & REQ_F_CREDS) && req->creds != current_cred()) creds = override_creds(req->creds); + if (!io_op_defs[req->opcode].audit_skip) + audit_uring_entry(req->opcode); + switch (req->opcode) { case IORING_OP_NOP: ret = io_nop(req, issue_flags); @@ -6272,6 +6308,9 @@ static int io_issue_sqe(struct io_kiocb *req, unsigned int issue_flags) break; } + if (!io_op_defs[req->opcode].audit_skip) + audit_uring_exit(!ret, ret); + if (creds) revert_creds(creds); if (ret) @@ -6896,6 +6935,8 @@ static int io_sq_thread(void *data) set_cpus_allowed_ptr(current, cpu_online_mask); current->flags |= PF_NO_SETAFFINITY; + audit_alloc_kernel(current); + mutex_lock(&sqd->lock); while (1) { bool cap_entries, sqt_spin = false; @@ -6961,6 +7002,8 @@ static int io_sq_thread(void *data) io_run_task_work(); mutex_unlock(&sqd->lock); + audit_free(current); + complete(&sqd->exited); do_exit(0); } diff --git a/include/linux/audit.h b/include/linux/audit.h index 82b7c1116a85..d656a06dd909 100644 --- a/include/linux/audit.h +++ b/include/linux/audit.h @@ -286,7 +286,10 @@ static inline int audit_signal_info(int sig, struct task_struct *t) /* These are defined in auditsc.c */ /* Public API */ extern int audit_alloc(struct task_struct *task); +extern int audit_alloc_kernel(struct task_struct *task); extern void __audit_free(struct task_struct *task); +extern void __audit_uring_entry(u8 op); +extern void __audit_uring_exit(int success, long code); extern void __audit_syscall_entry(int major, unsigned long a0, unsigned long a1, unsigned long a2, unsigned long a3); extern void __audit_syscall_exit(int ret_success, long ret_value); @@ -323,6 +326,21 @@ static inline void audit_free(struct task_struct *task) if (unlikely(task->audit_context)) __audit_free(task); } +static inline void audit_uring_entry(u8 op) +{ + /* + * We intentionally check audit_context() before audit_enabled as most + * Linux systems (as of ~2021) rely on systemd which forces audit to + * be enabled regardless of the user's audit configuration. + */ + if (unlikely(audit_context() && audit_enabled)) + __audit_uring_entry(op); +} +static inline void audit_uring_exit(int success, long code) +{ + if (unlikely(!audit_dummy_context())) + __audit_uring_exit(success, code); +} static inline void audit_syscall_entry(int major, unsigned long a0, unsigned long a1, unsigned long a2, unsigned long a3) @@ -554,8 +572,16 @@ static inline int audit_alloc(struct task_struct *task) { return 0; } +static inline int audit_alloc_kernel(struct task_struct *task) +{ + return 0; +} static inline void audit_free(struct task_struct *task) { } +static inline void audit_uring_entry(u8 op) +{ } +static inline void audit_uring_exit(int success, long code) +{ } static inline void audit_syscall_entry(int major, unsigned long a0, unsigned long a1, unsigned long a2, unsigned long a3) diff --git a/include/uapi/linux/audit.h b/include/uapi/linux/audit.h index daa481729e9b..a1997697c8b1 100644 --- a/include/uapi/linux/audit.h +++ b/include/uapi/linux/audit.h @@ -118,6 +118,7 @@ #define AUDIT_TIME_ADJNTPVAL 1333 /* NTP value adjustment */ #define AUDIT_BPF 1334 /* BPF subsystem */ #define AUDIT_EVENT_LISTENER 1335 /* Task joined multicast read socket */ +#define AUDIT_URINGOP 1336 /* io_uring operation */ #define AUDIT_AVC 1400 /* SE Linux avc denial or grant */ #define AUDIT_SELINUX_ERR 1401 /* Internal SE Linux Errors */ diff --git a/kernel/audit.h b/kernel/audit.h index 268a13a1c983..834fa18fd6ca 100644 --- a/kernel/audit.h +++ b/kernel/audit.h @@ -100,10 +100,12 @@ struct audit_context { enum { AUDIT_CTX_UNUSED, /* audit_context is currently unused */ AUDIT_CTX_SYSCALL, /* in use by syscall */ + AUDIT_CTX_URING, /* in use by io_uring */ } context; enum audit_state state, current_state; unsigned int serial; /* serial number for record */ int major; /* syscall number */ + int uring_op; /* uring operation */ struct timespec64 ctime; /* time of syscall entry */ unsigned long argv[4]; /* syscall arguments */ long return_code;/* syscall return code */ diff --git a/kernel/auditsc.c b/kernel/auditsc.c index c0383d554e61..62fb502da3fc 100644 --- a/kernel/auditsc.c +++ b/kernel/auditsc.c @@ -959,6 +959,7 @@ static void audit_reset_context(struct audit_context *ctx) ctx->current_state = ctx->state; ctx->serial = 0; ctx->major = 0; + ctx->uring_op = 0; ctx->ctime = (struct timespec64){ .tv_sec = 0, .tv_nsec = 0 }; memset(ctx->argv, 0, sizeof(ctx->argv)); ctx->return_code = 0; @@ -1044,6 +1045,31 @@ int audit_alloc(struct task_struct *tsk) return 0; } +/** + * audit_alloc_kernel - allocate an audit_context for a kernel task + * @tsk: the kernel task + * + * Similar to the audit_alloc() function, but intended for kernel private + * threads. Returns zero on success, negative values on failure. + */ +int audit_alloc_kernel(struct task_struct *tsk) +{ + /* + * At the moment we are just going to call into audit_alloc() to + * simplify the code, but there two things to keep in mind with this + * approach: + * + * 1. Filtering internal kernel tasks is a bit laughable in almost all + * cases, but there is at least one case where there is a benefit: + * the '-a task,never' case allows the admin to effectively disable + * task auditing at runtime. + * + * 2. The {set,clear}_task_syscall_work() ops likely have zero effect + * on these internal kernel tasks, but they probably don't hurt either. + */ + return audit_alloc(tsk); +} + static inline void audit_free_context(struct audit_context *context) { /* resetting is extra work, but it is likely just noise */ @@ -1546,6 +1572,52 @@ static void audit_log_proctitle(void) audit_log_end(ab); } +/** + * audit_log_uring - generate a AUDIT_URINGOP record + * @ctx: the audit context + */ +static void audit_log_uring(struct audit_context *ctx) +{ + struct audit_buffer *ab; + const struct cred *cred; + + /* + * TODO: What do we log here? I'm tossing in a few things to start the + * conversation, but additional thought needs to go into this. + */ + + ab = audit_log_start(ctx, GFP_ATOMIC, AUDIT_URINGOP); + if (!ab) + return; + cred = current_cred(); + audit_log_format(ab, "uring_op=%d", ctx->uring_op); + if (ctx->return_valid != AUDITSC_INVALID) + audit_log_format(ab, " success=%s exit=%ld", + (ctx->return_valid == AUDITSC_SUCCESS ? + "yes" : "no"), + ctx->return_code); + audit_log_format(ab, + " items=%d" + " ppid=%d pid=%d auid=%u uid=%u gid=%u" + " euid=%u suid=%u fsuid=%u" + " egid=%u sgid=%u fsgid=%u", + ctx->name_count, + task_ppid_nr(current), + task_tgid_nr(current), + from_kuid(&init_user_ns, audit_get_loginuid(current)), + from_kuid(&init_user_ns, cred->uid), + from_kgid(&init_user_ns, cred->gid), + from_kuid(&init_user_ns, cred->euid), + from_kuid(&init_user_ns, cred->suid), + from_kuid(&init_user_ns, cred->fsuid), + from_kgid(&init_user_ns, cred->egid), + from_kgid(&init_user_ns, cred->sgid), + from_kgid(&init_user_ns, cred->fsgid)); + audit_log_task_context(ab); + audit_log_key(ab, ctx->filterkey); + audit_log_end(ab); +} + static void audit_log_exit(void) { int i, call_panic = 0; @@ -1581,6 +1653,9 @@ static void audit_log_exit(void) audit_log_key(ab, context->filterkey); audit_log_end(ab); break; + case AUDIT_CTX_URING: + audit_log_uring(context); + break; default: BUG(); break; @@ -1751,6 +1826,105 @@ static void audit_return_fixup(struct audit_context *ctx, ctx->return_valid = (success ? AUDITSC_SUCCESS : AUDITSC_FAILURE); } +/** + * __audit_uring_entry - prepare the kernel task's audit context for io_uring + * @op: the io_uring opcode + * + * This is similar to audit_syscall_entry() but is intended for use by io_uring + * operations. This function should only ever be called from + * audit_uring_entry() as we rely on the audit context checking present in that + * function. + */ +void __audit_uring_entry(u8 op) +{ + struct audit_context *ctx = audit_context(); + + if (ctx->state == AUDIT_STATE_DISABLED) + return; + + /* + * NOTE: It's possible that we can be called from the process' context + * before it returns to userspace, and before audit_syscall_exit() + * is called. In this case there is not much to do, just record + * the io_uring details and return. + */ + ctx->uring_op = op; + if (ctx->context == AUDIT_CTX_SYSCALL) + return; + + ctx->dummy = !audit_n_rules; + if (!ctx->dummy && ctx->state == AUDIT_STATE_BUILD) + ctx->prio = 0; + + ctx->context = AUDIT_CTX_URING; + ctx->current_state = ctx->state; + ktime_get_coarse_real_ts64(&ctx->ctime); +} + +/** + * __audit_uring_exit - wrap up the kernel task's audit context after io_uring + * @success: true/false value to indicate if the operation succeeded or not + * @code: operation return code + * + * This is similar to audit_syscall_exit() but is intended for use by io_uring + * operations. This function should only ever be called from + * audit_uring_exit() as we rely on the audit context checking present in that + * function. + */ +void __audit_uring_exit(int success, long code) +{ + struct audit_context *ctx = audit_context(); + + /* + * TODO: At some point we will likely want to filter on io_uring ops + * and other things similar to what we do for syscalls, but that + * is something for another day; just record what we can here. + */ + + if (ctx->context == AUDIT_CTX_SYSCALL) { + /* + * NOTE: See the note in __audit_uring_entry() about the case + * where we may be called from process context before we + * return to userspace via audit_syscall_exit(). In this + * case we simply emit a URINGOP record and bail, the + * normal syscall exit handling will take care of + * everything else. + * It is also worth mentioning that when we are called, + * the current process creds may differ from the creds + * used during the normal syscall processing; keep that + * in mind if/when we move the record generation code. + */ + + /* + * We need to filter on the syscall info here to decide if we + * should emit a URINGOP record. I know it seems odd but this + * solves the problem where users have a filter to block *all* + * syscall records in the "exit" filter; we want to preserve + * the behavior here. + */ + audit_filter_syscall(current, ctx); + audit_filter_inodes(current, ctx); + if (ctx->current_state != AUDIT_STATE_RECORD) + return; + + audit_log_uring(ctx); + return; + } + + /* this may generate CONFIG_CHANGE records */ + if (!list_empty(&ctx->killed_trees)) + audit_kill_trees(ctx); + + audit_filter_inodes(current, ctx); + if (ctx->current_state != AUDIT_STATE_RECORD) + goto out; + audit_return_fixup(ctx, success, code); + audit_log_exit(); + +out: + audit_reset_context(ctx); +} + /** * __audit_syscall_entry - fill in an audit record at syscall entry * @major: major syscall type (function)

WARNING - This is a work in progress and should not be merged anywhere important. It is almost surely not complete, and while it probably compiles it likely hasn't been booted and will do terrible things. You have been warned. This patch adds basic audit io_uring filtering, using as much of the existing audit filtering infrastructure as possible. In order to do this we reuse the audit filter rule's syscall mask for the io_uring operation and we create a new filter for io_uring operations as AUDIT_FILTER_URING_EXIT/audit_filter_list[7]. <TODO - provide some additional guidance for the userspace tools> Thanks to Richard Guy Briggs for his review and feedback. Signed-off-by: Paul Moore <paul(a)paul-moore.com&gt; --- v2: - incorporate feedback from Richard v1: - initial draft --- include/uapi/linux/audit.h | 3 +- kernel/audit_tree.c | 3 +- kernel/audit_watch.c | 3 +- kernel/auditfilter.c | 15 ++++++++-- kernel/auditsc.c | 65 ++++++++++++++++++++++++++++++++++---------- 5 files changed, 68 insertions(+), 21 deletions(-) diff --git a/include/uapi/linux/audit.h b/include/uapi/linux/audit.h index a1997697c8b1..ecf1edd2affa 100644 --- a/include/uapi/linux/audit.h +++ b/include/uapi/linux/audit.h @@ -167,8 +167,9 @@ #define AUDIT_FILTER_EXCLUDE 0x05 /* Apply rule before record creation */ #define AUDIT_FILTER_TYPE AUDIT_FILTER_EXCLUDE /* obsolete misleading naming */ #define AUDIT_FILTER_FS 0x06 /* Apply rule at __audit_inode_child */ +#define AUDIT_FILTER_URING_EXIT 0x07 /* Apply rule at io_uring op exit */ -#define AUDIT_NR_FILTERS 7 +#define AUDIT_NR_FILTERS 8 #define AUDIT_FILTER_PREPEND 0x10 /* Prepend to front of list */ diff --git a/kernel/audit_tree.c b/kernel/audit_tree.c index b2be4e978ba3..31bf34844546 100644 --- a/kernel/audit_tree.c +++ b/kernel/audit_tree.c @@ -726,7 +726,8 @@ int audit_make_tree(struct audit_krule *rule, char *pathname, u32 op) { if (pathname[0] != '/' || - rule->listnr != AUDIT_FILTER_EXIT || + (rule->listnr != AUDIT_FILTER_EXIT && + rule->listnr != AUDIT_FILTER_URING_EXIT) || op != Audit_equal || rule->inode_f || rule->watch || rule->tree) return -EINVAL; diff --git a/kernel/audit_watch.c b/kernel/audit_watch.c index 2acf7ca49154..698b62b4a2ec 100644 --- a/kernel/audit_watch.c +++ b/kernel/audit_watch.c @@ -183,7 +183,8 @@ int audit_to_watch(struct audit_krule *krule, char *path, int len, u32 op) return -EOPNOTSUPP; if (path[0] != '/' || path[len-1] == '/' || - krule->listnr != AUDIT_FILTER_EXIT || + (krule->listnr != AUDIT_FILTER_EXIT && + krule->listnr != AUDIT_FILTER_URING_EXIT) || op != Audit_equal || krule->inode_f || krule->watch || krule->tree) return -EINVAL; diff --git a/kernel/auditfilter.c b/kernel/auditfilter.c index db2c6b59dfc3..d75acb014ccd 100644 --- a/kernel/auditfilter.c +++ b/kernel/auditfilter.c @@ -44,7 +44,8 @@ struct list_head audit_filter_list[AUDIT_NR_FILTERS] = { LIST_HEAD_INIT(audit_filter_list[4]), LIST_HEAD_INIT(audit_filter_list[5]), LIST_HEAD_INIT(audit_filter_list[6]), -#if AUDIT_NR_FILTERS != 7 + LIST_HEAD_INIT(audit_filter_list[7]), +#if AUDIT_NR_FILTERS != 8 #error Fix audit_filter_list initialiser #endif }; @@ -56,6 +57,7 @@ static struct list_head audit_rules_list[AUDIT_NR_FILTERS] = { LIST_HEAD_INIT(audit_rules_list[4]), LIST_HEAD_INIT(audit_rules_list[5]), LIST_HEAD_INIT(audit_rules_list[6]), + LIST_HEAD_INIT(audit_rules_list[7]), }; DEFINE_MUTEX(audit_filter_mutex); @@ -151,7 +153,8 @@ char *audit_unpack_string(void **bufp, size_t *remain, size_t len) static inline int audit_to_inode(struct audit_krule *krule, struct audit_field *f) { - if (krule->listnr != AUDIT_FILTER_EXIT || + if ((krule->listnr != AUDIT_FILTER_EXIT && + krule->listnr != AUDIT_FILTER_URING_EXIT) || krule->inode_f || krule->watch || krule->tree || (f->op != Audit_equal && f->op != Audit_not_equal)) return -EINVAL; @@ -248,6 +251,7 @@ static inline struct audit_entry *audit_to_entry_common(struct audit_rule_data * pr_err("AUDIT_FILTER_ENTRY is deprecated\n"); goto exit_err; case AUDIT_FILTER_EXIT: + case AUDIT_FILTER_URING_EXIT: case AUDIT_FILTER_TASK: #endif case AUDIT_FILTER_USER: @@ -332,6 +336,10 @@ static int audit_field_valid(struct audit_entry *entry, struct audit_field *f) if (entry->rule.listnr != AUDIT_FILTER_FS) return -EINVAL; break; + case AUDIT_PERM: + if (entry->rule.listnr == AUDIT_FILTER_URING_EXIT) + return -EINVAL; + break; } switch (entry->rule.listnr) { @@ -980,7 +988,8 @@ static inline int audit_add_rule(struct audit_entry *entry) } entry->rule.prio = ~0ULL; - if (entry->rule.listnr == AUDIT_FILTER_EXIT) { + if (entry->rule.listnr == AUDIT_FILTER_EXIT || + entry->rule.listnr == AUDIT_FILTER_URING_EXIT) { if (entry->rule.flags & AUDIT_FILTER_PREPEND) entry->rule.prio = ++prio_high; else diff --git a/kernel/auditsc.c b/kernel/auditsc.c index 928f1dd12460..f4a2ca496cac 100644 --- a/kernel/auditsc.c +++ b/kernel/auditsc.c @@ -805,6 +805,35 @@ static int audit_in_mask(const struct audit_krule *rule, unsigned long val) return rule->mask[word] & bit; } +/** + * audit_filter_uring - apply filters to an io_uring operation + * @tsk: associated task + * @ctx: audit context + */ +static void audit_filter_uring(struct task_struct *tsk, + struct audit_context *ctx) +{ + struct audit_entry *e; + enum audit_state state; + + if (auditd_test_task(tsk)) + return; + + rcu_read_lock(); + list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_URING_EXIT], + list) { + if (audit_in_mask(&e->rule, ctx->uring_op) && + audit_filter_rules(tsk, &e->rule, ctx, NULL, &state, + false)) { + rcu_read_unlock(); + ctx->current_state = state; + return; + } + } + rcu_read_unlock(); + return; +} + /* At syscall exit time, this filter is called if the audit_state is * not low enough that auditing cannot take place, but is also not * high enough that we already know we have to write an audit record @@ -1765,7 +1794,7 @@ static void audit_log_exit(void) * __audit_free - free a per-task audit context * @tsk: task whose audit context block to free * - * Called from copy_process and do_exit + * Called from copy_process, do_exit, and the io_uring code */ void __audit_free(struct task_struct *tsk) { @@ -1783,15 +1812,21 @@ void __audit_free(struct task_struct *tsk) * random task_struct that doesn't doesn't have any meaningful data we * need to log via audit_log_exit(). */ - if (tsk == current && !context->dummy && - context->context == AUDIT_CTX_SYSCALL) { + if (tsk == current && !context->dummy) { context->return_valid = AUDITSC_INVALID; context->return_code = 0; - - audit_filter_syscall(tsk, context); - audit_filter_inodes(tsk, context); - if (context->current_state == AUDIT_STATE_RECORD) - audit_log_exit(); + if (context->context == AUDIT_CTX_SYSCALL) { + audit_filter_syscall(tsk, context); + audit_filter_inodes(tsk, context); + if (context->current_state == AUDIT_STATE_RECORD) + audit_log_exit(); + } else if (context->context == AUDIT_CTX_URING) { + /* TODO: verify this case is real and valid */ + audit_filter_uring(tsk, context); + audit_filter_inodes(tsk, context); + if (context->current_state == AUDIT_STATE_RECORD) + audit_log_uring(context); + } } audit_set_context(tsk, NULL); @@ -1875,12 +1910,6 @@ void __audit_uring_exit(int success, long code) { struct audit_context *ctx = audit_context(); - /* - * TODO: At some point we will likely want to filter on io_uring ops - * and other things similar to what we do for syscalls, but that - * is something for another day; just record what we can here. - */ - if (ctx->context == AUDIT_CTX_SYSCALL) { /* * NOTE: See the note in __audit_uring_entry() about the case @@ -1903,6 +1932,8 @@ void __audit_uring_exit(int success, long code) * the behavior here. */ audit_filter_syscall(current, ctx); + if (ctx->current_state != AUDIT_STATE_RECORD) + audit_filter_uring(current, ctx); audit_filter_inodes(current, ctx); if (ctx->current_state != AUDIT_STATE_RECORD) return; @@ -1911,7 +1942,9 @@ void __audit_uring_exit(int success, long code) return; } #if 1 - /* XXX - temporary hack to force record generation */ + /* XXX - temporary hack to force record generation, we are leaving this + * enabled, but if you want to actually test the filtering you + * need to disable this #if/#endif block */ ctx->current_state = AUDIT_STATE_RECORD; #endif @@ -1919,6 +1952,8 @@ void __audit_uring_exit(int success, long code) if (!list_empty(&ctx->killed_trees)) audit_kill_trees(ctx); + /* run through both filters to ensure we set the filterkey properly */ + audit_filter_uring(current, ctx); audit_filter_inodes(current, ctx); if (ctx->current_state != AUDIT_STATE_RECORD) goto out;

On 11/08/2021 22:48, Paul Moore wrote: This is not directly related to this patch but why using the "secure" boolean in __anon_inode_getfile() and __anon_inode_getfd() instead of checking that context_inode is not NULL? This would simplify the code, remove this anon_inode_getfile_secure() wrapper and avoid potential inconsistencies.

On 12/08/2021 16:32, Paul Moore wrote: Indeed. Acked-by: Mickaël Salaün <mic(a)linux.microsoft.com&gt;

WARNING - This is a work in progress, this patch, including the description, may be incomplete or even incorrect. You have been warned. A full expalantion of io_uring is beyond the scope of this commit description, but in summary it is an asynchronous I/O mechanism which allows for I/O requests and the resulting data to be queued in memory mapped "rings" which are shared between the kernel and userspace. Optionally, io_uring offers the ability for applications to spawn kernel threads to dequeue I/O requests from the ring and submit the requests in the kernel, helping to minimize the syscall overhead. Rings are accessed in userspace by memory mapping a file descriptor provided by the io_uring_setup(2), and can be shared between applications as one might do with any open file descriptor. Finally, process credentials can be registered with a given ring and any process with access to that ring can submit I/O requests using any of the registered credentials. While the io_uring functionality is widely recognized as offering a vastly improved, and high performing asynchronous I/O mechanism, its ability to allow processes to submit I/O requests with credentials other than its own presents a challenge to LSMs. When a process creates a new io_uring ring the ring's credentials are inhertied from the calling process; if this ring is shared with another process operating with different credentials there is the potential to bypass the LSMs security policy. Similarly, registering credentials with a given ring allows any process with access to that ring to submit I/O requests with those credentials. In an effort to allow LSMs to apply security policy to io_uring I/O operations, this patch adds two new LSM hooks. These hooks, in conjunction with the LSM anonymous inode support previously submitted, allow an LSM to apply access control policy to the sharing of io_uring rings as well as any io_uring credential changes requested by a process. The new LSM hooks are described below: * int security_uring_override_creds(cred) Controls if the current task, executing an io_uring operation, is allowed to override it's credentials with @cred. In cases where the current task is a user application, the current credentials will be those of the user application. In cases where the current task is a kernel thread servicing io_uring requests the current credentials will be those of the io_uring ring (inherited from the process that created the ring). * int security_uring_sqpoll(void) Controls if the current task is allowed to create an io_uring polling thread (IORING_SETUP_SQPOLL). Without a SQPOLL thread in the kernel processes must submit I/O requests via io_uring_enter(2) which allows us to compare any requested credential changes against the application making the request. With a SQPOLL thread, we can no longer compare requested credential changes against the application making the request, the comparison is made against the ring's credentials. Signed-off-by: Paul Moore <paul(a)paul-moore.com&gt; --- v2: - no change v1: - initial draft --- fs/io_uring.c | 10 ++++++++++ include/linux/lsm_hook_defs.h | 5 +++++ include/linux/lsm_hooks.h | 13 +++++++++++++ include/linux/security.h | 16 ++++++++++++++++ security/security.c | 12 ++++++++++++ 5 files changed, 56 insertions(+) diff --git a/fs/io_uring.c b/fs/io_uring.c index ea396f5fe735..78df778d329b 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -79,6 +79,7 @@ #include <linux/pagemap.h> #include <linux/io_uring.h> #include <linux/audit.h> +#include <linux/security.h> #define CREATE_TRACE_POINTS #include <trace/events/io_uring.h> @@ -6617,6 +6618,11 @@ static int io_init_req(struct io_ring_ctx *ctx, struct io_kiocb *req, if (!req->creds) return -EINVAL; get_cred(req->creds); + ret = security_uring_override_creds(req->creds); + if (ret) { + put_cred(req->creds); + return ret; + } req->flags |= REQ_F_CREDS; } state = &ctx->submit_state; @@ -8080,6 +8086,10 @@ static int io_sq_offload_create(struct io_ring_ctx *ctx, struct io_sq_data *sqd; bool attached; + ret = security_uring_sqpoll(); + if (ret) + return ret; + sqd = io_get_sq_data(p, &attached); if (IS_ERR(sqd)) { ret = PTR_ERR(sqd); diff --git a/include/linux/lsm_hook_defs.h b/include/linux/lsm_hook_defs.h index 2adeea44c0d5..b3c525353769 100644 --- a/include/linux/lsm_hook_defs.h +++ b/include/linux/lsm_hook_defs.h @@ -402,3 +402,8 @@ LSM_HOOK(void, LSM_RET_VOID, perf_event_free, struct perf_event *event) LSM_HOOK(int, 0, perf_event_read, struct perf_event *event) LSM_HOOK(int, 0, perf_event_write, struct perf_event *event) #endif /* CONFIG_PERF_EVENTS */ + +#ifdef CONFIG_IO_URING +LSM_HOOK(int, 0, uring_override_creds, const struct cred *new) +LSM_HOOK(int, 0, uring_sqpoll, void) +#endif /* CONFIG_IO_URING */ diff --git a/include/linux/lsm_hooks.h b/include/linux/lsm_hooks.h index 5c4c5c0602cb..0eb0ae95c4c4 100644 --- a/include/linux/lsm_hooks.h +++ b/include/linux/lsm_hooks.h @@ -1557,6 +1557,19 @@ * Read perf_event security info if allowed. * @perf_event_write: * Write perf_event security info if allowed. + * + * Security hooks for io_uring + * + * @uring_override_creds: + * Check if the current task, executing an io_uring operation, is allowed + * to override it's credentials with @new. + * + * @new: the new creds to use + * + * @uring_sqpoll: + * Check whether the current task is allowed to spawn a io_uring polling + * thread (IORING_SETUP_SQPOLL). + * */ union security_list_options { #define LSM_HOOK(RET, DEFAULT, NAME, ...) RET (*NAME)(__VA_ARGS__); diff --git a/include/linux/security.h b/include/linux/security.h index 24eda04221e9..3928d80b6f76 100644 --- a/include/linux/security.h +++ b/include/linux/security.h @@ -2037,4 +2037,20 @@ static inline int security_perf_event_write(struct perf_event *event) #endif /* CONFIG_SECURITY */ #endif /* CONFIG_PERF_EVENTS */ +#ifdef CONFIG_IO_URING +#ifdef CONFIG_SECURITY +extern int security_uring_override_creds(const struct cred *new); +extern int security_uring_sqpoll(void); +#else +static inline int security_uring_override_creds(const struct cred *new) +{ + return 0; +} +static inline int security_uring_sqpoll(void) +{ + return 0; +} +#endif /* CONFIG_SECURITY */ +#endif /* CONFIG_IO_URING */ + #endif /* ! __LINUX_SECURITY_H */ diff --git a/security/security.c b/security/security.c index 09533cbb7221..a54544e61bd1 100644 --- a/security/security.c +++ b/security/security.c @@ -2624,3 +2624,15 @@ int security_perf_event_write(struct perf_event *event) return call_int_hook(perf_event_write, 0, event); } #endif /* CONFIG_PERF_EVENTS */ + +#ifdef CONFIG_IO_URING +int security_uring_override_creds(const struct cred *new) +{ + return call_int_hook(uring_override_creds, 0, new); +} + +int security_uring_sqpoll(void) +{ + return call_int_hook(uring_sqpoll, 0); +} +#endif /* CONFIG_IO_URING */

From: Casey Schaufler <casey(a)schaufler-ca.com&gt; Add Smack privilege checks for io_uring. Use CAP_MAC_OVERRIDE for the override_creds case and CAP_MAC_ADMIN for creating a polling thread. These choices are based on conjecture regarding the intent of the surrounding code. Signed-off-by: Casey Schaufler <casey(a)schaufler-ca.com&gt; [PM: make the smack_uring_* funcs static] Signed-off-by: Paul Moore <paul(a)paul-moore.com&gt; --- v2: - made the smack_uring_* funcs static v1: - initial draft --- security/smack/smack_lsm.c | 64 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 64 insertions(+) diff --git a/security/smack/smack_lsm.c b/security/smack/smack_lsm.c index 223a6da0e6dc..7fb094098f38 100644 --- a/security/smack/smack_lsm.c +++ b/security/smack/smack_lsm.c @@ -4691,6 +4691,66 @@ static int smack_dentry_create_files_as(struct dentry *dentry, int mode, return 0; } +#ifdef CONFIG_IO_URING +/** + * smack_uring_override_creds - Is io_uring cred override allowed? + * @new: the target creds + * + * Check to see if the current task is allowed to override it's credentials + * to service an io_uring operation. + */ +static int smack_uring_override_creds(const struct cred *new) +{ + struct task_smack *tsp = smack_cred(current_cred()); + struct task_smack *nsp = smack_cred(new); + +#if 1 + if (tsp->smk_task == nsp->smk_task) + pr_info("%s: Smack matches %s\n", __func__, + tsp->smk_task->smk_known); + else + pr_info("%s: Smack override check %s to %s\n", __func__, + tsp->smk_task->smk_known, nsp->smk_task->smk_known); +#endif + /* + * Allow the degenerate case where the new Smack value is + * the same as the current Smack value. + */ + if (tsp->smk_task == nsp->smk_task) + return 0; + +#if 1 + pr_info("%s: Smack sqpoll %s\n", __func__, + smack_privileged_cred(CAP_MAC_OVERRIDE, current_cred()) ? + "ok by Smack" : "disallowed (No CAP_MAC_OVERRIDE)"); +#endif + if (smack_privileged_cred(CAP_MAC_OVERRIDE, current_cred())) + return 0; + + return -EPERM; +} + +/** + * smack_uring_sqpoll - check if a io_uring polling thread can be created + * + * Check to see if the current task is allowed to create a new io_uring + * kernel polling thread. + */ +static int smack_uring_sqpoll(void) +{ +#if 1 + pr_info("%s: Smack new ring %s\n", __func__, + smack_privileged_cred(CAP_MAC_ADMIN, current_cred()) ? + "ok by Smack" : "disallowed (No CAP_MAC_ADMIN)"); +#endif + if (smack_privileged_cred(CAP_MAC_ADMIN, current_cred())) + return 0; + + return -EPERM; +} + +#endif /* CONFIG_IO_URING */ + struct lsm_blob_sizes smack_blob_sizes __lsm_ro_after_init = { .lbs_cred = sizeof(struct task_smack), .lbs_file = sizeof(struct smack_known *), @@ -4843,6 +4903,10 @@ static struct security_hook_list smack_hooks[] __lsm_ro_after_init = { LSM_HOOK_INIT(inode_copy_up, smack_inode_copy_up), LSM_HOOK_INIT(inode_copy_up_xattr, smack_inode_copy_up_xattr), LSM_HOOK_INIT(dentry_create_files_as, smack_dentry_create_files_as), +#ifdef CONFIG_IO_URING + LSM_HOOK_INIT(uring_override_creds, smack_uring_override_creds), + LSM_HOOK_INIT(uring_sqpoll, smack_uring_sqpoll), +#endif };

On 8/31/2021 7:44 AM, Paul Moore wrote: I have no other changes. And yes, the debug code should be stripped. Thank you.