[PATCH v2] android: binder: Drop lru lock in isolate callback

From: Sherry Yang
Date: Thu Sep 14 2017 - 14:23:12 EST


Drop the global lru lock in isolate callback
before calling zap_page_range which calls
cond_resched, and re-acquire the global lru
lock before returning. Also change return
code to LRU_REMOVED_RETRY.

Use mmput_async when fail to acquire mmap sem
in an atomic context.

Fix "BUG: sleeping function called from invalid context"
errors when CONFIG_DEBUG_ATOMIC_SLEEP is enabled.

Also restore mmput_async, which was initially introduced in
ec8d7c14e ("mm, oom_reaper: do not mmput synchronously from
the oom reaper context"), and was removed in 212925802
("mm: oom: let oom_reap_task and exit_mmap run concurrently").

Fixes: f2517eb76f1f2 ("android: binder: Add global lru shrinker to binder")
Reported-by: Kyle Yan <kyan@xxxxxxxxxxxxxx>
Acked-by: Arve HjÃnnevÃg <arve@xxxxxxxxxxx>
Signed-off-by: Sherry Yang <sherryy@xxxxxxxxxxx>
---
Changes in v2:
Combined '[PATCH] android: binder: Drop lru lock in isolate callback'
with '[PATCH] mm: Restore mmput_async'.

drivers/android/binder_alloc.c | 18 ++++++++++++------
include/linux/sched/mm.h | 6 ++++++
kernel/fork.c | 18 ++++++++++++++++++
3 files changed, 36 insertions(+), 6 deletions(-)

diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
index 8fe165844e47..064f5e31ec55 100644
--- a/drivers/android/binder_alloc.c
+++ b/drivers/android/binder_alloc.c
@@ -913,6 +913,7 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
struct binder_alloc *alloc;
uintptr_t page_addr;
size_t index;
+ struct vm_area_struct *vma;

alloc = page->alloc;
if (!mutex_trylock(&alloc->mutex))
@@ -923,16 +924,22 @@ enum lru_status binder_alloc_free_page(struct list_head *item,

index = page - alloc->pages;
page_addr = (uintptr_t)alloc->buffer + index * PAGE_SIZE;
- if (alloc->vma) {
+ vma = alloc->vma;
+ if (vma) {
mm = get_task_mm(alloc->tsk);
if (!mm)
goto err_get_task_mm_failed;
if (!down_write_trylock(&mm->mmap_sem))
goto err_down_write_mmap_sem_failed;
+ }
+
+ list_lru_isolate(lru, item);
+ spin_unlock(lock);

+ if (vma) {
trace_binder_unmap_user_start(alloc, index);

- zap_page_range(alloc->vma,
+ zap_page_range(vma,
page_addr + alloc->user_buffer_offset,
PAGE_SIZE);

@@ -950,13 +957,12 @@ enum lru_status binder_alloc_free_page(struct list_head *item,

trace_binder_unmap_kernel_end(alloc, index);

- list_lru_isolate(lru, item);
-
+ spin_lock(lock);
mutex_unlock(&alloc->mutex);
- return LRU_REMOVED;
+ return LRU_REMOVED_RETRY;

err_down_write_mmap_sem_failed:
- mmput(mm);
+ mmput_async(mm);
err_get_task_mm_failed:
err_page_already_freed:
mutex_unlock(&alloc->mutex);
diff --git a/include/linux/sched/mm.h b/include/linux/sched/mm.h
index 3a19c253bdb1..ae53e413fb13 100644
--- a/include/linux/sched/mm.h
+++ b/include/linux/sched/mm.h
@@ -84,6 +84,12 @@ static inline bool mmget_not_zero(struct mm_struct *mm)

/* mmput gets rid of the mappings and all user-space */
extern void mmput(struct mm_struct *);
+#ifdef CONFIG_MMU
+/* same as above but performs the slow path from the async context. Can
+ * be called from the atomic context as well
+ */
+void mmput_async(struct mm_struct *);
+#endif

/* Grab a reference to a task's mm, if it is not already going away */
extern struct mm_struct *get_task_mm(struct task_struct *task);
diff --git a/kernel/fork.c b/kernel/fork.c
index 10646182440f..e702cb9ffbd8 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -946,6 +946,24 @@ void mmput(struct mm_struct *mm)
}
EXPORT_SYMBOL_GPL(mmput);

+#ifdef CONFIG_MMU
+static void mmput_async_fn(struct work_struct *work)
+{
+ struct mm_struct *mm = container_of(work, struct mm_struct,
+ async_put_work);
+
+ __mmput(mm);
+}
+
+void mmput_async(struct mm_struct *mm)
+{
+ if (atomic_dec_and_test(&mm->mm_users)) {
+ INIT_WORK(&mm->async_put_work, mmput_async_fn);
+ schedule_work(&mm->async_put_work);
+ }
+}
+#endif
+
/**
* set_mm_exe_file - change a reference to the mm's executable file
*
--
2.11.0 (Apple Git-81)