Some architectures like arm64 and s390 require USER_DS to be set for
kernel threads to access user address space, which is the whole purpose
of kthread_use_mm, but other like x86 don't. That has lead to a huge
mess where some callers are fixed up once they are tested on said
architectures, while others linger around and yet other like io_uring
try to do "clever" optimizations for what usually is just a trivial
asignment to a member in the thread_struct for most architectures.
Make kthread_use_mm set USER_DS, and kthread_unuse_mm restore to the
previous value instead.
Signed-off-by: Christoph Hellwig <hch@xxxxxx>
Acked-by: Michael S. Tsirkin <mst@xxxxxxxxxx> [vhost]
---
drivers/usb/gadget/function/f_fs.c | 4 ----
drivers/vhost/vhost.c | 3 ---
fs/io-wq.c | 8 ++------
fs/io_uring.c | 4 ----
kernel/kthread.c | 6 ++++++
5 files changed, 8 insertions(+), 17 deletions(-)
diff --git a/drivers/usb/gadget/function/f_fs.c b/drivers/usb/gadget/function/f_fs.c
index d9e48bd7c692..a1198f4c527c 100644
--- a/drivers/usb/gadget/function/f_fs.c
+++ b/drivers/usb/gadget/function/f_fs.c
@@ -824,13 +824,9 @@ static void ffs_user_copy_worker(struct work_struct *work)
bool kiocb_has_eventfd = io_data->kiocb->ki_flags & IOCB_EVENTFD;
if (io_data->read && ret > 0) {
- mm_segment_t oldfs = get_fs();
-
- set_fs(USER_DS);
kthread_use_mm(io_data->mm);
ret = ffs_copy_to_iter(io_data->buf, ret, &io_data->data);
kthread_unuse_mm(io_data->mm);
- set_fs(oldfs);
}
io_data->kiocb->ki_complete(io_data->kiocb, ret, ret);
diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
index 17d598e74780..b2abfbdf3cb2 100644
--- a/drivers/vhost/vhost.c
+++ b/drivers/vhost/vhost.c
@@ -329,9 +329,7 @@ static int vhost_worker(void *data)
struct vhost_dev *dev = data;
struct vhost_work *work, *work_next;
struct llist_node *node;
- mm_segment_t oldfs = get_fs();
- set_fs(USER_DS);
kthread_use_mm(dev->mm);
for (;;) {
@@ -361,7 +359,6 @@ static int vhost_worker(void *data)
}
}
kthread_unuse_mm(dev->mm);
- set_fs(oldfs);
return 0;
}
diff --git a/fs/io-wq.c b/fs/io-wq.c
index 748621f7391e..a5e90ac39e4d 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -169,7 +169,6 @@ static bool __io_worker_unuse(struct io_wqe *wqe, struct io_worker *worker)
dropped_lock = true;
}
__set_current_state(TASK_RUNNING);
- set_fs(KERNEL_DS);
kthread_unuse_mm(worker->mm);
mmput(worker->mm);
worker->mm = NULL;
@@ -421,14 +420,11 @@ static void io_wq_switch_mm(struct io_worker *worker, struct io_wq_work *work)
mmput(worker->mm);
worker->mm = NULL;
}
- if (!work->mm) {
- set_fs(KERNEL_DS);
+ if (!work->mm)
return;
- }
+
if (mmget_not_zero(work->mm)) {
kthread_use_mm(work->mm);
- if (!worker->mm)
- set_fs(USER_DS);
worker->mm = work->mm;
/* hang on to this mm */
work->mm = NULL;
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 8a8148512da7..40f90b98a18a 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -5908,15 +5908,12 @@ static int io_sq_thread(void *data)
struct io_ring_ctx *ctx = data;
struct mm_struct *cur_mm = NULL;
const struct cred *old_cred;
- mm_segment_t old_fs;
DEFINE_WAIT(wait);
unsigned long timeout;
int ret = 0;
complete(&ctx->completions[1]);
- old_fs = get_fs();
- set_fs(USER_DS);
old_cred = override_creds(ctx->creds);
timeout = jiffies + ctx->sq_thread_idle;
@@ -6023,7 +6020,6 @@ static int io_sq_thread(void *data)
if (current->task_works)
task_work_run();
- set_fs(old_fs);
if (cur_mm) {
kthread_unuse_mm(cur_mm);
mmput(cur_mm);
diff --git a/kernel/kthread.c b/kernel/kthread.c
index 8ed4b4fbec7c..86357cd38eb2 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -52,6 +52,7 @@ struct kthread {
unsigned long flags;
unsigned int cpu;
void *data;
+ mm_segment_t oldfs;
struct completion parked;
struct completion exited;
#ifdef CONFIG_BLK_CGROUP
@@ -1235,6 +1236,9 @@ void kthread_use_mm(struct mm_struct *mm)
if (active_mm != mm)
mmdrop(active_mm);
+
+ to_kthread(tsk)->oldfs = get_fs();
+ set_fs(USER_DS);
}
EXPORT_SYMBOL_GPL(kthread_use_mm);
@@ -1249,6 +1253,8 @@ void kthread_unuse_mm(struct mm_struct *mm)
WARN_ON_ONCE(!(tsk->flags & PF_KTHREAD));
WARN_ON_ONCE(!tsk->mm);
+ set_fs(to_kthread(tsk)->oldfs);
+
task_lock(tsk);
sync_mm_rss(mm);
tsk->mm = NULL;
--
2.25.1
