MIPS: Correct forced syscall errors
authorJames Hogan <james.hogan@imgtec.com>
Thu, 29 Jun 2017 09:12:35 +0000 (10:12 +0100)
committerRalf Baechle <ralf@linux-mips.org>
Tue, 11 Jul 2017 12:13:06 +0000 (14:13 +0200)
When the system call return value is forced to be an error (for example
due to SECCOMP_RET_ERRNO), syscall_set_return_value() puts the error
code in the return register $v0 and -1 in the error register $a3.

However normally executed system calls put 1 in the error register
rather than -1, so fix syscall_set_return_value() to be consistent with
that.

I don't anticipate that anything would have been broken by this, since
the most natural way to check the error register on MIPS would be a
conditional branch if error register is [not] equal to zero (bnez or
beqz).

Fixes: 1d7bf993e073 ("MIPS: ftrace: Add support for syscall tracepoints.")
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/16652/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
arch/mips/include/asm/syscall.h

index d87882513ee34b03ea92ca1fc40157c31722c629..7c713025b23f6fbb9c06c9fb735ca0b35bbcf344 100644 (file)
@@ -85,7 +85,7 @@ static inline void syscall_set_return_value(struct task_struct *task,
 {
        if (error) {
                regs->regs[2] = -error;
-               regs->regs[7] = -1;
+               regs->regs[7] = 1;
        } else {
                regs->regs[2] = val;
                regs->regs[7] = 0;