The patch titled
Subject: mm/rmap: fix and simplify reusing mergeable anon_vma as parent when fork
has been added to the -mm tree. Its filename is
mm-rmap-fix-and-simplify-reusing-mergeable-anon_vma-as-parent-when-fork.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/mm-rmap-fix-and-simplify-reusing-mergeable-anon_vma-as-parent-when-fork.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/mm-rmap-fix-and-simplify-reusing-mergeable-anon_vma-as-parent-when-fork.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxxxxxx>
Subject: mm/rmap: fix and simplify reusing mergeable anon_vma as parent when fork
This fixes some misconceptions in commit 4e4a9eb92133 ("mm/rmap.c: reuse
mergeable anon_vma as parent when fork"). It merges anon-vma in
unexpected way but fortunately still produces valid anon-vma tree, so
nothing crashes.
If in parent VMAs: SRC1 SRC2 .. SRCn share anon-vma ANON0, then after
fork before all patches in child process related VMAs: DST1 DST2 .. DSTn
will fork indepndent anon-vmas: ANON1 ANON2 .. ANONn (each is child of
ANON0). Before this patch only DST1 will fork new ANON1 and following
DST2 .. DSTn will share parent's ANON0 (i.e. anon-vma tree is valid but
isn't optimal). With this patch DST1 will create new ANON1 and DST2 ..
DSTn will share it.
Root problem caused by initialization order in dup_mmap(): vma->vm_prev is
set after calling anon_vma_fork(). Thus in anon_vma_fork() it points to
previous VMA in parent mm.
Second problem is hidden behind first one: assumption "Parent has vm_prev,
which implies we have vm_prev" is wrong if first VMA in parent mm has set
flag VM_DONTCOPY. Luckily prev->anon_vma doesn't dereference NULL pointer
because in current code 'prev' actually is same as 'pprev'.
Third hidden problem is linking between VMA and anon-vmas whose pages it
could contain. Loop in anon_vma_clone() attaches only parent's anon-vmas,
shared anon-vma isn't attached. But every mapped page stays reachable in
rmap because we erroneously share anon-vma from parent's previous VMA.
This patch moves sharing logic out of anon_vma_clone() into more specific
anon_vma_fork() because this supposed to work only at fork() and simply
reuses anon_vma from previous VMA if it is forked from the same anon-vma.
Link: http://lkml.kernel.org/r/157839239609.694.10268055713935919822.stgit@buzz
Link: https://lore.kernel.org/linux-mm/CALYGNiNzz+dxHX0g5-gNypUQc3B=8_Scp53-NTOh=zWsdUuHAw@xxxxxxxxxxxxxx/T/#t
Fixes: 4e4a9eb92133 ("mm/rmap.c: reuse mergeable anon_vma as parent when fork")
Signed-off-by: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxxxxxx>
Reported-by: Li Xinhai <lixinhai.lxh@xxxxxxxxx>
Reviewed-by: Li Xinhai <lixinhai.lxh@xxxxxxxxx>
Cc: Wei Yang <richardw.yang@xxxxxxxxxxxxxxx>
Cc: Rik van Riel <riel@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
include/linux/rmap.h | 3 ++-
kernel/fork.c | 2 +-
mm/rmap.c | 23 +++++++++--------------
3 files changed, 12 insertions(+), 16 deletions(-)
--- a/include/linux/rmap.h~mm-rmap-fix-and-simplify-reusing-mergeable-anon_vma-as-parent-when-fork
+++ a/include/linux/rmap.h
@@ -143,7 +143,8 @@ void anon_vma_init(void); /* create anon
int __anon_vma_prepare(struct vm_area_struct *);
void unlink_anon_vmas(struct vm_area_struct *);
int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
-int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
+int anon_vma_fork(struct vm_area_struct *vma, struct vm_area_struct *pvma,
+ struct vm_area_struct *prev);
static inline int anon_vma_prepare(struct vm_area_struct *vma)
{
--- a/kernel/fork.c~mm-rmap-fix-and-simplify-reusing-mergeable-anon_vma-as-parent-when-fork
+++ a/kernel/fork.c
@@ -556,7 +556,7 @@ static __latent_entropy int dup_mmap(str
tmp->anon_vma = NULL;
if (anon_vma_prepare(tmp))
goto fail_nomem_anon_vma_fork;
- } else if (anon_vma_fork(tmp, mpnt))
+ } else if (anon_vma_fork(tmp, mpnt, prev))
goto fail_nomem_anon_vma_fork;
tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT);
tmp->vm_next = tmp->vm_prev = NULL;
--- a/mm/rmap.c~mm-rmap-fix-and-simplify-reusing-mergeable-anon_vma-as-parent-when-fork
+++ a/mm/rmap.c
@@ -269,19 +269,6 @@ int anon_vma_clone(struct vm_area_struct
{
struct anon_vma_chain *avc, *pavc;
struct anon_vma *root = NULL;
- struct vm_area_struct *prev = dst->vm_prev, *pprev = src->vm_prev;
-
- /*
- * If parent share anon_vma with its vm_prev, keep this sharing in in
- * child.
- *
- * 1. Parent has vm_prev, which implies we have vm_prev.
- * 2. Parent and its vm_prev have the same anon_vma.
- */
- if (!dst->anon_vma && src->anon_vma &&
- pprev && pprev->anon_vma == src->anon_vma)
- dst->anon_vma = prev->anon_vma;
-
list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
struct anon_vma *anon_vma;
@@ -332,7 +319,8 @@ int anon_vma_clone(struct vm_area_struct
* the corresponding VMA in the parent process is attached to.
* Returns 0 on success, non-zero on failure.
*/
-int anon_vma_fork(struct vm_area_struct *vma, struct vm_area_struct *pvma)
+int anon_vma_fork(struct vm_area_struct *vma, struct vm_area_struct *pvma,
+ struct vm_area_struct *prev)
{
struct anon_vma_chain *avc;
struct anon_vma *anon_vma;
@@ -342,6 +330,13 @@ int anon_vma_fork(struct vm_area_struct
if (!pvma->anon_vma)
return 0;
+ /* Share anon_vma with previous VMA if it has the same parent. */
+ if (prev && prev->anon_vma &&
+ prev->anon_vma->parent == pvma->anon_vma) {
+ vma->anon_vma = prev->anon_vma;
+ return anon_vma_clone(vma, prev);
+ }
+
/* Drop inherited anon_vma, we'll reuse existing or allocate new. */
vma->anon_vma = NULL;
_
Patches currently in -mm which might be from khlebnikov@xxxxxxxxxxxxxx are
mm-rmap-fix-and-simplify-reusing-mergeable-anon_vma-as-parent-when-fork.patch
kernel-fork-set-vmas-mm-prev-next-right-after-vm_area_dup-in-dup_mmap.patch
