Currently the kernel does not provide an infrastructure to translate architecture numbers to a human-readable name. Translating syscall numbers to syscall names is possible through FTRACE_SYSCALL infrastructure but it does not provide support for compat syscalls. This will create a file for each PID as /proc/pid/seccomp_cache. The file will be empty when no seccomp filters are loaded, or be in the format of: <arch name> <decimal syscall number> <ALLOW | FILTER> where ALLOW means the cache is guaranteed to allow the syscall, and filter means the cache will pass the syscall to the BPF filter. For the docker default profile on x86_64 it looks like: x86_64 0 ALLOW x86_64 1 ALLOW x86_64 2 ALLOW x86_64 3 ALLOW [...] x86_64 132 ALLOW x86_64 133 ALLOW x86_64 134 FILTER x86_64 135 FILTER x86_64 136 FILTER x86_64 137 ALLOW x86_64 138 ALLOW x86_64 139 FILTER x86_64 140 ALLOW x86_64 141 ALLOW [...] This file is guarded by CONFIG_SECCOMP_CACHE_DEBUG with a default of N because I think certain users of seccomp might not want the application to know which syscalls are definitely usable. For the same reason, it is also guarded by CAP_SYS_ADMIN.
Unified · Split
| @@ -486,6 +486,9 @@ config HAVE_ARCH_SECCOMP_FILTER · 485 unmodified lines | |||
| 486 | - secure_computing return value is checked and a return value of -1 | 486 | - secure_computing return value is checked and a return value of -1 |
| 487 | results in the system call being skipped immediately. | 487 | results in the system call being skipped immediately. |
| 488 | - seccomp syscall wired up | 488 | - seccomp syscall wired up |
| 489 | - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE, | ||
| 490 | SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If | ||
| 491 | COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too. | ||
| 489 | 492 | ||
| 490 | config SECCOMP | 493 | config SECCOMP |
| 491 | prompt "Enable seccomp to safely execute untrusted bytecode" | 494 | prompt "Enable seccomp to safely execute untrusted bytecode" |
| @@ -514,6 +517,20 @@ config SECCOMP_FILTER · 22 unmodified lines | |||
| 514 | 517 | ||
| 515 | See Documentation/userspace-api/seccomp_filter.rst for details. | 518 | See Documentation/userspace-api/seccomp_filter.rst for details. |
| 516 | 519 | ||
| 520 | config SECCOMP_CACHE_DEBUG | ||
| 521 | bool "Show seccomp filter cache status in /proc/pid/seccomp_cache" | ||
| 522 | depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR | ||
| 523 | depends on PROC_FS | ||
| 524 | help | ||
| 525 | This enables the /proc/pid/seccomp_cache interface to monitor | ||
| 526 | seccomp cache data. The file format is subject to change. Reading | ||
| 527 | the file requires CAP_SYS_ADMIN. | ||
| 528 | |||
| 529 | This option is for debugging only. Enabling presents the risk that | ||
| 530 | an adversary may be able to infer the seccomp filter logic. | ||
| 531 | |||
| 532 | If unsure, say N. | ||
| 533 | |||
| 517 | config HAVE_ARCH_STACKLEAK | 534 | config HAVE_ARCH_STACKLEAK |
| 518 | bool | 535 | bool |
| 519 | help | 536 | help |
| @@ -3263,6 +3263,9 @@ static const struct pid_entry tgid_base_stuff[] = { · 3262 unmodified lines | |||
| 3263 | #ifdef CONFIG_PROC_PID_ARCH_STATUS | 3263 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
| 3264 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | 3264 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), |
| 3265 | #endif | 3265 | #endif |
| 3266 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG | ||
| 3267 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | ||
| 3268 | #endif | ||
| 3266 | }; | 3269 | }; |
| 3267 | 3270 | ||
| 3268 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) | 3271 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
| @@ -3592,6 +3595,9 @@ static const struct pid_entry tid_base_stuff[] = { · 323 unmodified lines | |||
| 3592 | #ifdef CONFIG_PROC_PID_ARCH_STATUS | 3595 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
| 3593 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | 3596 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), |
| 3594 | #endif | 3597 | #endif |
| 3598 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG | ||
| 3599 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | ||
| 3600 | #endif | ||
| 3595 | }; | 3601 | }; |
| 3596 | 3602 | ||
| 3597 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) | 3603 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
| @@ -121,4 +121,11 @@ static inline long seccomp_get_metadata(struct task_struct *task, · 120 unmodified lines | |||
| 121 | return -EINVAL; | 121 | return -EINVAL; |
| 122 | } | 122 | } |
| 123 | #endif /* CONFIG_SECCOMP_FILTER && CONFIG_CHECKPOINT_RESTORE */ | 123 | #endif /* CONFIG_SECCOMP_FILTER && CONFIG_CHECKPOINT_RESTORE */ |
| 124 | |||
| 125 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG | ||
| 126 | struct seq_file; | ||
| 127 | |||
| 128 | int proc_pid_seccomp_cache(struct seq_file *m, struct pid_namespace *ns, | ||
| 129 | struct pid *pid, struct task_struct *task); | ||
| 130 | #endif | ||
| 124 | #endif /* _LINUX_SECCOMP_H */ | 131 | #endif /* _LINUX_SECCOMP_H */ |
| @@ -553,6 +553,9 @@ void seccomp_filter_release(struct task_struct *tsk) · 552 unmodified lines | |||
| 553 | { | 553 | { |
| 554 | struct seccomp_filter *orig = tsk->seccomp.filter; | 554 | struct seccomp_filter *orig = tsk->seccomp.filter; |
| 555 | 555 | ||
| 556 | /* We are effectively holding the siglock by not having any sighand. */ | ||
| 557 | WARN_ON(tsk->sighand != NULL); | ||
| 558 | |||
| 556 | /* Detach task from its filter tree. */ | 559 | /* Detach task from its filter tree. */ |
| 557 | tsk->seccomp.filter = NULL; | 560 | tsk->seccomp.filter = NULL; |
| 558 | __seccomp_filter_release(orig); | 561 | __seccomp_filter_release(orig); |
| @@ -2335,3 +2338,59 @@ static int __init seccomp_sysctl_init(void) · 1776 unmodified lines | |||
| 2335 | device_initcall(seccomp_sysctl_init) | 2338 | device_initcall(seccomp_sysctl_init) |
| 2336 | 2339 | ||
| 2337 | #endif /* CONFIG_SYSCTL */ | 2340 | #endif /* CONFIG_SYSCTL */ |
| 2341 | |||
| 2342 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG | ||
| 2343 | /* Currently CONFIG_SECCOMP_CACHE_DEBUG implies SECCOMP_ARCH_NATIVE */ | ||
| 2344 | static void proc_pid_seccomp_cache_arch(struct seq_file *m, const char *name, | ||
| 2345 | const void *bitmap, size_t bitmap_size) | ||
| 2346 | { | ||
| 2347 | int nr; | ||
| 2348 | |||
| 2349 | for (nr = 0; nr < bitmap_size; nr++) { | ||
| 2350 | bool cached = test_bit(nr, bitmap); | ||
| 2351 | char *status = cached ? "ALLOW" : "FILTER"; | ||
| 2352 | |||
| 2353 | seq_printf(m, "%s %d %s\n", name, nr, status); | ||
| 2354 | } | ||
| 2355 | } | ||
| 2356 | |||
| 2357 | int proc_pid_seccomp_cache(struct seq_file *m, struct pid_namespace *ns, | ||
| 2358 | struct pid *pid, struct task_struct *task) | ||
| 2359 | { | ||
| 2360 | struct seccomp_filter *f; | ||
| 2361 | unsigned long flags; | ||
| 2362 | |||
| 2363 | /* | ||
| 2364 | * We don't want some sandboxed process to know what their seccomp | ||
| 2365 | * filters consist of. | ||
| 2366 | */ | ||
| 2367 | if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) | ||
| 2368 | return -EACCES; | ||
| 2369 | |||
| 2370 | if (!lock_task_sighand(task, &flags)) | ||
| 2371 | return -ESRCH; | ||
| 2372 | |||
| 2373 | f = READ_ONCE(task->seccomp.filter); | ||
| 2374 | if (!f) { | ||
| 2375 | unlock_task_sighand(task, &flags); | ||
| 2376 | return 0; | ||
| 2377 | } | ||
| 2378 | |||
| 2379 | /* prevent filter from being freed while we are printing it */ | ||
| 2380 | __get_seccomp_filter(f); | ||
| 2381 | unlock_task_sighand(task, &flags); | ||
| 2382 | |||
| 2383 | proc_pid_seccomp_cache_arch(m, SECCOMP_ARCH_NATIVE_NAME, | ||
| 2384 | f->cache.allow_native, | ||
| 2385 | SECCOMP_ARCH_NATIVE_NR); | ||
| 2386 | |||
| 2387 | #ifdef SECCOMP_ARCH_COMPAT | ||
| 2388 | proc_pid_seccomp_cache_arch(m, SECCOMP_ARCH_COMPAT_NAME, | ||
| 2389 | f->cache.allow_compat, | ||
| 2390 | SECCOMP_ARCH_COMPAT_NR); | ||
| 2391 | #endif /* SECCOMP_ARCH_COMPAT */ | ||
| 2392 | |||
| 2393 | __put_seccomp_filter(f); | ||
| 2394 | return 0; | ||
| 2395 | } | ||
| 2396 | #endif /* CONFIG_SECCOMP_CACHE_DEBUG */ | ||