arm64: ptrace: change fs when passing kernel pointer to regset code

Our compat PTRACE_POKEUSR implementation simply passes the user data to
regset_copy_from_user after some simple range checking. Unfortunately,
the data in question has already been copied to the kernel stack by this
point, so the subsequent access_ok check fails and the ptrace request
returns -EFAULT. This causes problems tracing fork() with older versions
of strace.

This patch briefly changes the fs to KERNEL_DS, so that the access_ok
check passes even with a kernel address.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

1 files changed, 4 insertions, 0 deletions

diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c
index 3e926b9c0641..993cdb79b70e 100644
--- a/arch/arm64/kernel/ptrace.c
+++ b/arch/arm64/kernel/ptrace.c
@@ -827,6 +827,7 @@ static int compat_ptrace_write_user(struct task_struct *tsk, compat_ulong_t off,
                                     compat_ulong_t val)
 {
         int ret;
+        mm_segment_t old_fs = get_fs();
 
         if (off & 3 || off >= COMPAT_USER_SZ)
                 return -EIO;
@@ -834,10 +835,13 @@ static int compat_ptrace_write_user(struct task_struct *tsk, compat_ulong_t off,
         if (off >= sizeof(compat_elf_gregset_t))
                 return 0;
 
+        set_fs(KERNEL_DS);
         ret = copy_regset_from_user(tsk, &user_aarch32_view,
                                     REGSET_COMPAT_GPR, off,
                                     sizeof(compat_ulong_t),
                                     &val);
+        set_fs(old_fs);
+
         return ret;
 }