arm64 fixes:

- Release the GCS on a failed fork. copy_thread() can allocate a GCS
    for CLONE_VM threads but a later copy_process() failure left the VMA
    in the parent's mm. Implement exit_thread() to free it and stop
    kernel threads and user workers inheriting the parent's GCS
    base/size. Add a kselftest for the clone3() failure case
 
  - Do not set PTE_CONT on invalid PTEs in contpte_set_ptes(). Batched
    set_ptes() of present-invalid entries (e.g. PROT_NONE) created contpte
    groups that the single-PTE helpers did not unfold, corrupting a
    neighbouring guard marker via test_and_clear_young_ptes()
 
  - Release the MPAM mon_sel lock on the mpam_save_mbwu_state() error
    paths
 -----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEE5RElWfyWxS+3PLO2a9axLQDIXvEFAmrk3ssACgkQa9axLQDI
 XvFwiQ/+LSIVXGHJIWom8Wu6e3P+Ho0FxbDbxgT3VHXR6uHjdyNa/nN60IXVQES8
 KJ7Kydrh8X+90eDYQY7iKwuXwNQsxwdYMFGLa26aFDaOv3bK0bI2xM3zpRAg98qy
 UwXApfatt4iCm1pB8E8n3XJ64eVccapfrf6Mv5wnn1cg8vSsrlR6S3HotATGc8cG
 +1cuoI3fgByDqiVkssJka2ZhM5tWBxEg4mQ3Un6O6vTQTAZCp/vbsRgOjupKsgRU
 zLWh8KNEoV1ehYtPuwXQ+sRiOhfBHOuNK3i+FlmTEuRjcLJ2xbkpIzCEJvy2+n9+
 n+0ewG0EkRrf4NodgeDmBB6zSyLhfKjorp634NqWj8UKF2z4uPLJzhb5xJXeeByZ
 uOynRR19fj2YOWIc/vF1p+4yQCSi6eOEASU/5oYSqsnKU/As6Xoum22dMtRERB6e
 YH8cQT0RXw0qWJ46+/6rRqIpugNa+kkVcQloP2YLKl4QpPYGhtK8Kw4dymx6OQcl
 zDoiP8R2hQzylMjzvNRA+KugCOKsCPy4j7A3w3aiBVOkhOsfoHK4zkx3LJp/su/d
 gRNiDJgN4S57yrmvEUlXmBXUI1QzeXnRNHm/xYb7SGe/DeDaJElS+EmcPFi/i6Up
 VaKyApaS017oBq/pS5Ku1DL4jr8EjPM9xW7//Tczq4MzPyl4ogY=
 =ZNuk
 -----END PGP SIGNATURE-----

Merge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:
 "Three fixes that turned up during the merging window: shadow stack not
  freed during a failed fork or it is unmapped from parent when a worker
  thread exits (same commit covers both); PTE_CONT set on PROT_NONE
  entries but subsequently ignored by the unfolding code since
  pte_valid() is false (pte_present() is true); an MPAM lock not
  released on the error path.

  Summary:

   - Release the GCS on a failed fork.

     copy_thread() can allocate a GCS for CLONE_VM threads but a later
     copy_process() failure left the VMA in the parent's mm. Implement
     exit_thread() to free it and stop kernel threads and user workers
     inheriting the parent's GCS base/size. Add a kselftest for the
     clone3() failure case

   - Do not set PTE_CONT on invalid PTEs in contpte_set_ptes().

     Batched set_ptes() of present-invalid entries (e.g. PROT_NONE)
     created contpte groups that the single-PTE helpers did not unfold,
     corrupting a neighbouring guard marker via
     test_and_clear_young_ptes()

   - Release the MPAM mon_sel lock on the mpam_save_mbwu_state() error
     paths"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  selftests/arm64: GCS must not leak on a failed clone3
  arm64/gcs: release GCS on failed fork
  arm64/mm: restrict transparent contpte mappings to valid PTEs
  arm_mpam: Release the mon_sel lock on mpam_save_mbwu_state() errors
This commit is contained in:
Linus Torvalds 2026-10-30 09:46:29 -10:00
commit 6e07889d8f
11 changed files with 224 additions and 49 deletions

View File

@ -199,6 +199,7 @@ config ARM64
select HAVE_SHARED_GPIOS
select HAVE_BUILDTIME_MCOUNT_SORT
select HAVE_EFFICIENT_UNALIGNED_ACCESS
select HAVE_EXIT_THREAD
select HAVE_GUP_FAST
select HAVE_FTRACE_GRAPH_FUNC
select HAVE_FUNCTION_TRACER

View File

@ -1597,6 +1597,9 @@ extern void modify_prot_commit_ptes(struct vm_area_struct *vma, unsigned long ad
* The contpte APIs are used to transparently manage the contiguous bit in ptes
* where it is possible and makes sense to do so. The PTE_CONT bit is considered
* a private implementation detail of the public ptep API (see below).
* Single-PTE transparent contpte helpers use pte_valid_cont() to identify
* grouped PTEs; explicit contiguous-encoding users such as hugetlb manage
* their entries separately.
*/
extern void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte);

View File

@ -320,9 +320,18 @@ static int copy_thread_gcs(struct task_struct *p,
return 0;
}
static void gcs_reset_thread(struct task_struct *task)
{
task->thread.gcs_base = 0;
task->thread.gcs_size = 0;
task->thread.gcspr_el0 = 0;
}
#else
static void flush_gcs(void) { }
static void gcs_reset_thread(struct task_struct *task) { }
static int copy_thread_gcs(struct task_struct *p,
const struct kernel_clone_args *args)
{
@ -341,6 +350,12 @@ void flush_thread(void)
flush_gcs();
}
void exit_thread(struct task_struct *tsk)
{
/* Failed fork after copy_thread(); deactivate_mm() does not run. */
gcs_free(tsk);
}
void arch_release_task_struct(struct task_struct *tsk)
{
fpsimd_release_task(tsk);
@ -503,6 +518,8 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
p->thread.cpu_context.x19 = (unsigned long)args->fn;
p->thread.cpu_context.x20 = (unsigned long)args->fn_arg;
/* Kernel threads sharing the parent's mm must not free its GCS. */
gcs_reset_thread(p);
if (system_supports_poe())
p->thread.por_el0 = POR_EL0_INIT;

View File

@ -460,8 +460,10 @@ void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
*/
VM_WARN_ON(nr == 1);
if (!mm_is_user(mm))
if (!mm_is_user(mm) || !pte_valid(pte)) {
VM_WARN_ON_ONCE(!pte_valid(pte) && pte_cont(pte));
return __set_ptes(mm, addr, ptep, pte, nr);
}
end = addr + (nr << PAGE_SHIFT);
pfn = pte_pfn(pte);

View File

@ -2037,8 +2037,10 @@ static int mpam_save_mbwu_state(void *arg)
struct mpam_class *class = ris->vmsc->comp->class;
for (i = 0; i < ris->props.num_mbwu_mon; i++) {
if (WARN_ON_ONCE(!mpam_mon_sel_lock(msc)))
return -EIO;
ACQUIRE(mon_sel_lock, guard)(msc);
ret = ACQUIRE_ERR(mon_sel_lock, &guard);
if (ret)
return ret;
mbwu_state = &ris->mbwu_state[i];
cfg = &mbwu_state->cfg;
@ -2061,10 +2063,8 @@ static int mpam_save_mbwu_state(void *arg)
cfg->partid = FIELD_GET(MSMON_CFG_x_FLT_PARTID, cur_flt);
mbwu_state->enabled = FIELD_GET(MSMON_CFG_x_CTL_EN, cur_ctl);
if (!mbwu_state->enabled) {
mpam_mon_sel_unlock(msc);
if (!mbwu_state->enabled)
continue;
}
val = 0;
mbwu_arg = (struct mon_read) {
@ -2088,8 +2088,6 @@ static int mpam_save_mbwu_state(void *arg)
ret = mpam_write_monsel_reg(msc, CFG_MBWU_CTL, 0);
if (ret)
return ret;
mpam_mon_sel_unlock(msc);
}
return 0;

View File

@ -5,3 +5,4 @@ gcs-stress
gcs-stress-thread
gcspushm
gcsstr
gcs-clone-pidfd

View File

@ -6,7 +6,8 @@
# nolibc.
#
TEST_GEN_PROGS := basic-gcs libc-gcs gcs-locking gcs-stress gcspushm gcsstr
TEST_GEN_PROGS := basic-gcs libc-gcs gcs-locking gcs-stress gcspushm gcsstr \
gcs-clone-pidfd
TEST_GEN_PROGS_EXTENDED := gcs-stress-thread
LDLIBS+=-lpthread

View File

@ -0,0 +1,168 @@
// SPDX-License-Identifier: GPL-2.0
/*
* clone3(CLONE_VM|CLONE_PIDFD) with an unwritable pidfd pointer must
* not leak a GCS VMA. copy_thread() maps one, then put_user() fails
* and copy_process() calls exit_thread(). Without HAVE_EXIT_THREAD
* that free is a no-op and the VMA stays in the parent's mm.
*
* After enabling GCS, do not return from main(): libc frames above
* that point have no GCS entries.
*/
#define _GNU_SOURCE
#include <errno.h>
#include <linux/sched.h>
#include <sched.h>
#include <signal.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/auxv.h>
#include <sys/mman.h>
#include <sys/syscall.h>
#include <sys/wait.h>
#include <unistd.h>
#include <asm/hwcap.h>
#include "kselftest.h"
#include "gcs-util.h"
#ifndef HWCAP_GCS
#define HWCAP_GCS (1UL << 32)
#endif
#ifndef __NR_clone3
#define __NR_clone3 435
#endif
#define NCLONE 16
#define STACK_SIZE (64 * 1024)
static int count_ss_vmas(void)
{
FILE *f;
char line[256];
int vmas = 0;
f = fopen("/proc/self/smaps", "r");
if (!f)
ksft_exit_fail_msg("fopen /proc/self/smaps: %s\n",
strerror(errno));
while (fgets(line, sizeof(line), f)) {
if (!strncmp(line, "VmFlags:", 8) && strstr(line, " ss"))
vmas++;
}
fclose(f);
return vmas;
}
/*
* CLONE_VM maps a new GCS. CLONE_PIDFD lets put_user() fail after that.
* Use a raw svc: the child resumes on an empty GCS, so a libc return
* after clone3 would fault; exit the same way.
*/
static long clone_pidfd(void *stack, size_t stack_size, int *pidfd_ptr)
{
struct clone_args args = {
.flags = CLONE_VM | CLONE_PIDFD,
.pidfd = (uint64_t)(uintptr_t)pidfd_ptr,
.exit_signal = SIGCHLD,
.stack = (uint64_t)(uintptr_t)stack,
.stack_size = stack_size,
};
register long num __asm__("x8") = __NR_clone3;
register long ret __asm__("x0") = (long)&args;
register long sz __asm__("x1") = sizeof(args);
int st;
asm volatile("svc #0"
: "+r"(ret)
: "r"(num), "r"(sz)
: "memory", "cc");
if (ret == 0) {
register long n __asm__("x8") = __NR_exit;
register long e __asm__("x0") = 0;
asm volatile("svc #0" : : "r"(n), "r"(e) : "memory");
__builtin_unreachable();
}
if (ret < 0)
return ret;
if (waitpid((pid_t)ret, &st, 0) < 0)
return -errno;
if (!WIFEXITED(st) || WEXITSTATUS(st) != 0)
return 1;
return 0;
}
int main(void)
{
void *stack;
void *bad_pidfd;
unsigned long gcs_mode;
long rc;
int pidfd = -1;
int vmas0, vmas1;
int i, n_efault = 0;
ksft_print_header();
if (!(getauxval(AT_HWCAP) & HWCAP_GCS))
ksft_exit_skip("GCS not supported\n");
stack = mmap(NULL, STACK_SIZE, PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS | MAP_STACK, -1, 0);
if (stack == MAP_FAILED)
ksft_exit_fail_msg("mmap stack: %s\n", strerror(errno));
bad_pidfd = mmap(NULL, 4096, PROT_NONE,
MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
if (bad_pidfd == MAP_FAILED)
ksft_exit_fail_msg("mmap pidfd page: %s\n", strerror(errno));
rc = my_syscall2(__NR_prctl, PR_GET_SHADOW_STACK_STATUS, &gcs_mode);
if (rc)
ksft_exit_fail_msg("PR_GET_SHADOW_STACK_STATUS: %ld\n", rc);
if (!(gcs_mode & PR_SHADOW_STACK_ENABLE)) {
rc = my_syscall2(__NR_prctl, PR_SET_SHADOW_STACK_STATUS,
PR_SHADOW_STACK_ENABLE);
if (rc == -EOPNOTSUPP || rc == -EINVAL || rc == -EBUSY)
ksft_exit_skip("cannot enable GCS: %ld\n", rc);
if (rc)
ksft_exit_fail_msg("PR_SET_SHADOW_STACK_STATUS: %ld\n",
rc);
}
rc = clone_pidfd(stack, STACK_SIZE, &pidfd);
if (rc == -ENOSYS || rc == -EINVAL || rc == -EPERM)
ksft_exit_skip("clone3 valid pidfd unavailable: %s\n",
strerror(-rc));
ksft_set_plan(3);
ksft_test_result(rc == 0 && pidfd >= 0,
"clone3 valid pidfd (rc=%ld pidfd=%d)\n",
rc, pidfd);
if (pidfd >= 0)
close(pidfd);
vmas0 = count_ss_vmas();
for (i = 0; i < NCLONE; i++) {
rc = clone_pidfd(stack, STACK_SIZE, bad_pidfd);
if (rc == -EFAULT)
n_efault++;
}
ksft_test_result(n_efault == NCLONE,
"clone3 bad pidfd returns -EFAULT x%d\n", NCLONE);
vmas1 = count_ss_vmas();
ksft_test_result(vmas1 == vmas0,
"no GCS VMA leak (%d -> %d)\n", vmas0, vmas1);
ksft_finished();
}

View File

@ -18,26 +18,6 @@
#include "gcs-util.h"
#define my_syscall2(num, arg1, arg2) \
({ \
register long _num __asm__ ("x8") = (num); \
register long _arg1 __asm__ ("x0") = (long)(arg1); \
register long _arg2 __asm__ ("x1") = (long)(arg2); \
register long _arg3 __asm__ ("x2") = 0; \
register long _arg4 __asm__ ("x3") = 0; \
register long _arg5 __asm__ ("x4") = 0; \
\
__asm__ volatile ( \
"svc #0\n" \
: "=r"(_arg1) \
: "r"(_arg1), "r"(_arg2), \
"r"(_arg3), "r"(_arg4), \
"r"(_arg5), "r"(_num) \
: "memory", "cc" \
); \
_arg1; \
})
/* No mode bits are rejected for locking */
TEST(lock_all_modes)
{

View File

@ -43,6 +43,30 @@
#define GCS_CAP(x) (((unsigned long)(x) & GCS_CAP_ADDR_MASK) | \
GCS_CAP_VALID_TOKEN)
/*
* prctl(..., arg1, arg2) with unused registers forced to zero. The
* kernel rejects non-zero unused args; libc may leave garbage there.
*/
#define my_syscall2(num, arg1, arg2) \
({ \
register long _num __asm__ ("x8") = (num); \
register long _arg1 __asm__ ("x0") = (long)(arg1); \
register long _arg2 __asm__ ("x1") = (long)(arg2); \
register long _arg3 __asm__ ("x2") = 0; \
register long _arg4 __asm__ ("x3") = 0; \
register long _arg5 __asm__ ("x4") = 0; \
\
__asm__ volatile ( \
"svc #0\n" \
: "=r"(_arg1) \
: "r"(_arg1), "r"(_arg2), \
"r"(_arg3), "r"(_arg4), \
"r"(_arg5), "r"(_num) \
: "memory", "cc" \
); \
_arg1; \
})
static inline unsigned long *get_gcspr(void)
{
unsigned long *gcspr;

View File

@ -24,26 +24,6 @@
#include "gcs-util.h"
#define my_syscall2(num, arg1, arg2) \
({ \
register long _num __asm__ ("x8") = (num); \
register long _arg1 __asm__ ("x0") = (long)(arg1); \
register long _arg2 __asm__ ("x1") = (long)(arg2); \
register long _arg3 __asm__ ("x2") = 0; \
register long _arg4 __asm__ ("x3") = 0; \
register long _arg5 __asm__ ("x4") = 0; \
\
__asm__ volatile ( \
"svc #0\n" \
: "=r"(_arg1) \
: "r"(_arg1), "r"(_arg2), \
"r"(_arg3), "r"(_arg4), \
"r"(_arg5), "r"(_num) \
: "memory", "cc" \
); \
_arg1; \
})
static noinline void gcs_recurse(int depth)
{
if (depth)