1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
|
From d655ada2fc4b5065c2be6943f28aa4e4f694a1a3 Mon Sep 17 00:00:00 2001
From: Iwona Winiarska <iwona.winiarska@intel.com>
Date: Mon, 9 Jan 2023 21:09:44 +0100
Subject: [PATCH] soc: aspeed: lpc-mbox: Don't allow partial reads
IRQ handler always adds all registers to the fifo, while the userspace
can potentially consume a smaller amount. Unfortunately, when smaller
amount is consumed, it permanently shifts the fifo.
Serialize the "empty" state check to avoid partial reads.
Fixes: 60fde6cf7114 ("soc: aspeed: lpc-mbox: Avoid calling kfifo_to_user in atomic context")
Reported-by: Arun P. Mohanan <arun.p.m@linux.intel.com>
Signed-off-by: Iwona Winiarska <iwona.winiarska@intel.com>
---
drivers/soc/aspeed/aspeed-lpc-mbox.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/drivers/soc/aspeed/aspeed-lpc-mbox.c b/drivers/soc/aspeed/aspeed-lpc-mbox.c
index 7941792abacb..564c7318da10 100644
--- a/drivers/soc/aspeed/aspeed-lpc-mbox.c
+++ b/drivers/soc/aspeed/aspeed-lpc-mbox.c
@@ -172,7 +172,14 @@ static ssize_t aspeed_mbox_read(struct file *file, char __user *buf,
}
mutex_lock(&mbox->mutex);
- if (kfifo_is_empty(&mbox->fifo)) {
+ /*
+ * Since fifo on the producer side will drop the oldest values, causing
+ * a shift if the data is not consumed fully, when we're using count ==
+ * num_regs reads, we need to serialize with the producer to make
+ * sure that all regs were inserted into fifo (avoiding a partial
+ * read).
+ */
+ if (kfifo_is_empty_spinlocked(&mbox->fifo, &mbox->lock)) {
if (file->f_flags & O_NONBLOCK) {
ret = -EAGAIN;
goto out_unlock;
--
2.17.1
|
