s390/fpu: get rid of test_fp_ctl()

It is quite subtle to use test_fp_ctl() correctly. Therefore remove it -
instead copy whatever new floating point control (fpc) register values are
supposed to be used into its save area.

Test the validity of the new value when loading it. If the new value is
invalid, load the fpc register with zero.

This seems to be a the best way to approach this problem. Even though this
changes behavior:

- sigreturn with an invalid fpc value on the stack will succeed, and
  continue with zero value, instead of returning with SIGSEGV

- ptraced processes will also use a zero value instead of letting the
  request fail with -EINVAL

However all of this seems to acceptable. After all testing of the value was
only implemented to avoid that user space can crash the kernel. It is not
there to test values for validity; and the assumption is that there is no
existing user space which is doing this.

Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>

1 files changed, 0 insertions, 4 deletions

diff --git a/arch/s390/kernel/signal.c b/arch/s390/kernel/signal.c
index d63557d3868c..0e926e896808 100644
--- a/arch/s390/kernel/signal.c
+++ b/arch/s390/kernel/signal.c
@@ -149,10 +149,6 @@ static int restore_sigregs(struct pt_regs *regs, _sigregs __user *sregs)
 	if (!is_ri_task(current) && (user_sregs.regs.psw.mask & PSW_MASK_RI))
 		return -EINVAL;
 
-	/* Test the floating-point-control word. */
-	if (test_fp_ctl(user_sregs.fpregs.fpc))
-		return -EINVAL;
-
 	/* Use regs->psw.mask instead of PSW_USER_BITS to preserve PER bit. */
 	regs->psw.mask = (regs->psw.mask & ~(PSW_MASK_USER | PSW_MASK_RI)) |
 		(user_sregs.regs.psw.mask & (PSW_MASK_USER | PSW_MASK_RI));