[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210517140305.301559380@linuxfoundation.org>
Date: Mon, 17 May 2021 15:59:06 +0200
From: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To: linux-kernel@...r.kernel.org
Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
stable@...r.kernel.org,
"Linh Phung T. Y." <linh.phung.jy@...esas.com>,
Kuninori Morimoto <kuninori.morimoto.gx@...esas.com>,
Mark Brown <broonie@...nel.org>,
Sasha Levin <sashal@...nel.org>
Subject: [PATCH 5.12 080/363] ASoC: rsnd: call rsnd_ssi_master_clk_start() from rsnd_ssi_init()
From: Kuninori Morimoto <kuninori.morimoto.gx@...esas.com>
[ Upstream commit a122a116fc6d8fcf2f202dcd185173a54268f239 ]
Current rsnd needs to call .prepare (P) for clock settings,
.trigger for playback start (S) and stop (E).
It should be called as below from SSI point of view.
P -> S -> E -> P -> S -> E -> ...
But, if you used MIXer, below case might happen
(2)
1: P -> S ---> E -> ...
2: P ----> S -> ...
(1) (3)
P(1) setups clock, but E(2) resets it. and starts playback (3).
In such case, it will reports "SSI parent/child should use same rate".
rsnd_ssi_master_clk_start() which is the main function at (P)
was called from rsnd_ssi_init() (= S) before,
but was moved by below patch to rsnd_soc_dai_prepare() (= P) to avoid
using clk_get_rate() which shouldn't be used under atomic context.
commit 4d230d1271064 ("ASoC: rsnd: fixup not to call clk_get/set
under non-atomic")
Because of above patch, rsnd_ssi_master_clk_start() is now called at (P)
which is for non atomic context. But (P) is assuming that spin lock is
*not* used.
One issue now is rsnd_ssi_master_clk_start() is checking ssi->xxx
which should be protected by spin lock.
After above patch, adg.c had below patch for other reasons.
commit 06e8f5c842f2d ("ASoC: rsnd: don't call clk_get_rate()
under atomic context")
clk_get_rate() is used at probe() timing by this patch.
In other words, rsnd_ssi_master_clk_start() is no longer using
clk_get_rate() any more.
This means we can call it from rsnd_ssi_init() (= S) again which is
protected by spin lock.
This patch re-move it to under spin lock, and solves
1. checking ssi->xxx without spin lock issue.
2. clk setting / device start / device stop race condition.
Reported-by: Linh Phung T. Y. <linh.phung.jy@...esas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@...esas.com>
Link: https://lore.kernel.org/r/875z0x1jt5.wl-kuninori.morimoto.gx@renesas.com
Signed-off-by: Mark Brown <broonie@...nel.org>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
sound/soc/sh/rcar/ssi.c | 14 +++++---------
1 file changed, 5 insertions(+), 9 deletions(-)
diff --git a/sound/soc/sh/rcar/ssi.c b/sound/soc/sh/rcar/ssi.c
index d0ded427a836..a2f8138d40c7 100644
--- a/sound/soc/sh/rcar/ssi.c
+++ b/sound/soc/sh/rcar/ssi.c
@@ -507,10 +507,15 @@ static int rsnd_ssi_init(struct rsnd_mod *mod,
struct rsnd_priv *priv)
{
struct rsnd_ssi *ssi = rsnd_mod_to_ssi(mod);
+ int ret;
if (!rsnd_ssi_is_run_mods(mod, io))
return 0;
+ ret = rsnd_ssi_master_clk_start(mod, io);
+ if (ret < 0)
+ return ret;
+
ssi->usrcnt++;
rsnd_mod_power_on(mod);
@@ -1060,13 +1065,6 @@ static int rsnd_ssi_pio_pointer(struct rsnd_mod *mod,
return 0;
}
-static int rsnd_ssi_prepare(struct rsnd_mod *mod,
- struct rsnd_dai_stream *io,
- struct rsnd_priv *priv)
-{
- return rsnd_ssi_master_clk_start(mod, io);
-}
-
static struct rsnd_mod_ops rsnd_ssi_pio_ops = {
.name = SSI_NAME,
.probe = rsnd_ssi_common_probe,
@@ -1079,7 +1077,6 @@ static struct rsnd_mod_ops rsnd_ssi_pio_ops = {
.pointer = rsnd_ssi_pio_pointer,
.pcm_new = rsnd_ssi_pcm_new,
.hw_params = rsnd_ssi_hw_params,
- .prepare = rsnd_ssi_prepare,
.get_status = rsnd_ssi_get_status,
};
@@ -1166,7 +1163,6 @@ static struct rsnd_mod_ops rsnd_ssi_dma_ops = {
.pcm_new = rsnd_ssi_pcm_new,
.fallback = rsnd_ssi_fallback,
.hw_params = rsnd_ssi_hw_params,
- .prepare = rsnd_ssi_prepare,
.get_status = rsnd_ssi_get_status,
};
--
2.30.2
Powered by blists - more mailing lists