hexagon: fix livelock in uaccess
authorAl Viro <viro@zeniv.linux.org.uk>
Tue, 31 Jan 2023 04:46:10 +0000 (23:46 -0500)
committerAl Viro <viro@zeniv.linux.org.uk>
Thu, 2 Mar 2023 17:30:21 +0000 (12:30 -0500)
hexagon equivalent of 26178ec11ef3 "x86: mm: consolidate VM_FAULT_RETRY handling"
If e.g. get_user() triggers a page fault and a fatal signal is caught, we might
end up with handle_mm_fault() returning VM_FAULT_RETRY and not doing anything
to page tables.  In such case we must *not* return to the faulting insn -
that would repeat the entire thing without making any progress; what we need
instead is to treat that as failed (user) memory access.

Acked-by: Brian Cain <bcain@quicinc.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
arch/hexagon/mm/vm_fault.c

index f73c7cb..4b578d0 100644 (file)
@@ -93,8 +93,11 @@ good_area:
 
        fault = handle_mm_fault(vma, address, flags, regs);
 
-       if (fault_signal_pending(fault, regs))
+       if (fault_signal_pending(fault, regs)) {
+               if (!user_mode(regs))
+                       goto no_context;
                return;
+       }
 
        /* The fault is fully completed (including releasing mmap lock) */
        if (fault & VM_FAULT_COMPLETED)