From ca9fd733033b6fc4f6d70e7d1260bca598ae30fc Mon Sep 17 00:00:00 2001 From: Andy Lutomirski Date: Sun, 31 Jan 2021 09:24:35 -0800 Subject: [PATCH] x86/fault: Document the locking in the fault_signal_pending() path If fault_signal_pending() returns true, then the core mm has unlocked the mm for us. Add a comment to help future readers of this code. Cc: Dave Hansen Cc: Peter Zijlstra Signed-off-by: Andy Lutomirski --- arch/x86/mm/fault.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c index 87d30daa5e0a5d..b52064920f0d0b 100644 --- a/arch/x86/mm/fault.c +++ b/arch/x86/mm/fault.c @@ -1343,8 +1343,11 @@ void do_user_addr_fault(struct pt_regs *regs, */ fault = handle_mm_fault(vma, address, flags, regs); - /* Quick path to respond to signals */ if (fault_signal_pending(fault, regs)) { + /* + * Quick path to respond to signals. The core mm code + * has unlocked the mm for us if we get here. + */ if (!user_mode(regs)) no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);