Re: [PATCH 9/9] scsi: ufs: Add clock ungating to a separate workqueue

From: Asutosh Das (asd)
Date: Sun Feb 25 2018 - 22:55:31 EST


On 2/24/2018 5:27 AM, Miguel Ojeda wrote:
On Wed, Feb 21, 2018 at 5:56 AM, Asutosh Das <asutoshd@xxxxxxxxxxxxxx> wrote:
From: Vijay Viswanath <vviswana@xxxxxxxxxxxxxx>

UFS driver can receive a request during memory reclaim by kswapd.
So when ufs driver puts the ungate work in queue, and if there are no
idle workers, kthreadd is invoked to create a new kworker. Since
kswapd task holds a mutex which kthreadd also needs, this can cause
a deadlock situation. So ungate work must be done in a separate
work queue with WQ__RECLAIM flag enabled.Such a workqueue will have

WQ_MEM_RECLAIM here. Also space after dot.

a rescue thread which will be called when the above deadlock
condition is possible.

Signed-off-by: Vijay Viswanath <vviswana@xxxxxxxxxxxxxx>
Signed-off-by: Can Guo <cang@xxxxxxxxxxxxxx>
Signed-off-by: Asutosh Das <asutoshd@xxxxxxxxxxxxxx>
---
drivers/scsi/ufs/ufshcd.c | 10 +++++++++-
drivers/scsi/ufs/ufshcd.h | 1 +
2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 6541e1d..bb3382a 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -1503,7 +1503,8 @@ int ufshcd_hold(struct ufs_hba *hba, bool async)
hba->clk_gating.state = REQ_CLKS_ON;
trace_ufshcd_clk_gating(dev_name(hba->dev),
hba->clk_gating.state);
- schedule_work(&hba->clk_gating.ungate_work);
+ queue_work(hba->clk_gating.clk_gating_workq,
+ &hba->clk_gating.ungate_work);
/*
* fall through to check if we should wait for this
* work to be done or not.
@@ -1689,6 +1690,8 @@ static ssize_t ufshcd_clkgate_enable_store(struct device *dev,

static void ufshcd_init_clk_gating(struct ufs_hba *hba)
{
+ char wq_name[sizeof("ufs_clk_gating_00")];
+
if (!ufshcd_is_clkgating_allowed(hba))
return;

@@ -1696,6 +1699,10 @@ static void ufshcd_init_clk_gating(struct ufs_hba *hba)
INIT_DELAYED_WORK(&hba->clk_gating.gate_work, ufshcd_gate_work);
INIT_WORK(&hba->clk_gating.ungate_work, ufshcd_ungate_work);

+ snprintf(wq_name, ARRAY_SIZE(wq_name), "ufs_clk_gating_%d",
+ hba->host->host_no);
+ hba->clk_gating.clk_gating_workq = create_singlethread_workqueue(wq_name);

Shouldn't this be alloc_ordered_workqueue() with WQ_MEM_RECLAIM then?
(create_singlethread_workqueue() and friends are deprecated).

Cheers,
Miguel

+
hba->clk_gating.is_enabled = true;

hba->clk_gating.delay_attr.show = ufshcd_clkgate_delay_show;
@@ -1723,6 +1730,7 @@ static void ufshcd_exit_clk_gating(struct ufs_hba *hba)
device_remove_file(hba->dev, &hba->clk_gating.enable_attr);
cancel_work_sync(&hba->clk_gating.ungate_work);
cancel_delayed_work_sync(&hba->clk_gating.gate_work);
+ destroy_workqueue(hba->clk_gating.clk_gating_workq);
}

/* Must be called with host lock acquired */
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index 4385741..570c33e 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -361,6 +361,7 @@ struct ufs_clk_gating {
struct device_attribute enable_attr;
bool is_enabled;
int active_reqs;
+ struct workqueue_struct *clk_gating_workq;
};

struct ufs_saved_pwr_info {
--
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.


Hi Miguel
Thanks for the review.

I'll check this and put up the changes in v2.


-asd

--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project