ipc/shm: kill the historical/wrong mm->start_stack check
authorOleg Nesterov <oleg@redhat.com>
Mon, 13 Oct 2014 22:54:12 +0000 (15:54 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 14 Oct 2014 00:18:23 +0000 (02:18 +0200)
commitbf77b94c99ad5df0d97a52522fc7a220c0bf44fe
treefbf5579029160ae086b0ae16ed25ab9814550dbd
parent1195d94e006b23c6292e78857e154872e33b6d7e
ipc/shm: kill the historical/wrong mm->start_stack check

do_shmat() is the only user of ->start_stack (proc just reports its
value), and this check looks ugly and wrong.

The reason for this check is not clear at all, and it wrongly assumes that
the stack can only grow down.

But the main problem is that in general mm->start_stack has nothing to do
with stack_vma->vm_start.  Not only the application can switch to another
stack and even unmap this area, setup_arg_pages() expands the stack
without updating mm->start_stack during exec().  This means that in the
likely case "addr > start_stack - size - PAGE_SIZE * 5" is simply
impossible after find_vma_intersection() == F, or the stack can't grow
anyway because of RLIMIT_STACK.

Many thanks to Hugh for his explanations.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
ipc/shm.c