[PATCH] freepgt: free_pgtables from FIRST_USER_ADDRESS
authorHugh Dickins <hugh@veritas.com>
Tue, 19 Apr 2005 20:29:19 +0000 (13:29 -0700)
committerLinus Torvalds <torvalds@ppc970.osdl.org.(none)>
Tue, 19 Apr 2005 20:29:19 +0000 (13:29 -0700)
commite2cdef8c847b480529b7e26991926aab4be008e6
treeb936ab7f0964f56bc3312ad9ad956e978ac39895
parent021740dc30d184e3b0fa7679936e65a56090c425
[PATCH] freepgt: free_pgtables from FIRST_USER_ADDRESS

The patches to free_pgtables by vma left problems on any architectures which
leave some user address page table entries unencapsulated by vma.  Andi has
fixed the 32-bit vDSO on x86_64 to use a vma.  Now fix arm (and arm26), whose
first PAGE_SIZE is reserved (perhaps) for machine vectors.

Our calls to free_pgtables must not touch that area, and exit_mmap's
BUG_ON(nr_ptes) must allow that arm's get_pgd_slow may (or may not) have
allocated an extra page table, which its free_pgd_slow would free later.

FIRST_USER_PGD_NR has misled me and others: until all the arches define
FIRST_USER_ADDRESS instead, a hack in mmap.c to derive one from t'other.  This
patch fixes the bugs, the remaining patches just clean it up.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
mm/mmap.c