diff mbox series

[02/13] net: ravb: Use pm_runtime_resume_and_get()

Message ID 20231120084606.4083194-3-claudiu.beznea.uj@bp.renesas.com (mailing list archive)
State Superseded
Delegated to: Geert Uytterhoeven
Headers show
Series net: ravb: Add suspend to RAM and runtime PM support for RZ/G3S | expand

Commit Message

Claudiu Nov. 20, 2023, 8:45 a.m. UTC
From: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>

pm_runtime_get_sync() may return with error. In case it returns with error
dev->power.usage_count needs to be decremented. pm_runtime_resume_and_get()
takes care of this. Thus use it.

Along with this pm_runtime_resume_and_get() and reset_control_deassert()
were moved before alloc_etherdev_mqs() to simplify the error path.

Also, in case pm_runtime_resume_and_get() returns error the reset signal
is deasserted and runtime PM is disabled (by jumping to the proper
error handling label).

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
---
 drivers/net/ethernet/renesas/ravb_main.c | 28 +++++++++++++-----------
 1 file changed, 15 insertions(+), 13 deletions(-)

Comments

Sergey Shtylyov Nov. 20, 2023, 7:23 p.m. UTC | #1
On 11/20/23 11:45 AM, Claudiu wrote:

> From: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
> 
> pm_runtime_get_sync() may return with error. In case it returns with error
> dev->power.usage_count needs to be decremented. pm_runtime_resume_and_get()
> takes care of this. Thus use it.
> 
> Along with this pm_runtime_resume_and_get() and reset_control_deassert()
> were moved before alloc_etherdev_mqs() to simplify the error path.

   I don't see how it simplifies the error path...
   Re-ordering the statements at the end of the error path seems cheaper than
what you do.

> Also, in case pm_runtime_resume_and_get() returns error the reset signal
> is deasserted and runtime PM is disabled (by jumping to the proper
> error handling label).
> 
> Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
> Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
[...]

MBR, Sergey
Claudiu Nov. 21, 2023, 5:57 a.m. UTC | #2
On 20.11.2023 21:23, Sergey Shtylyov wrote:
> On 11/20/23 11:45 AM, Claudiu wrote:
> 
>> From: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
>>
>> pm_runtime_get_sync() may return with error. In case it returns with error
>> dev->power.usage_count needs to be decremented. pm_runtime_resume_and_get()
>> takes care of this. Thus use it.
>>
>> Along with this pm_runtime_resume_and_get() and reset_control_deassert()
>> were moved before alloc_etherdev_mqs() to simplify the error path.
> 
>    I don't see how it simplifies the error path...

By not changing it... Actually, I took the other approach: you suggested in
patch 1 to re-arrange the error path, I did it the other way around:
changed the initialization path...

>    Re-ordering the statements at the end of the error path seems cheaper than
> what you do.
> 
>> Also, in case pm_runtime_resume_and_get() returns error the reset signal
>> is deasserted and runtime PM is disabled (by jumping to the proper
>> error handling label).
>>
>> Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
>> Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
> [...]
> 
> MBR, Sergey
Sergey Shtylyov Nov. 24, 2023, 7:02 p.m. UTC | #3
On 11/21/23 8:57 AM, claudiu beznea wrote:

[...]

>>> From: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
>>>
>>> pm_runtime_get_sync() may return with error. In case it returns with error
>>> dev->power.usage_count needs to be decremented. pm_runtime_resume_and_get()
>>> takes care of this. Thus use it.
>>>
>>> Along with this pm_runtime_resume_and_get() and reset_control_deassert()
>>> were moved before alloc_etherdev_mqs() to simplify the error path.
>>
>>    I don't see how it simplifies the error path...
> 
> By not changing it... Actually, I took the other approach: you suggested in

   But it does need to be changed! It's not currently in the reverse order
compared to the buildup path...

> patch 1 to re-arrange the error path, I did it the other way around:
> changed the initialization path...

   That way you needlessly obfuscate (by moving the code around) the core
change you do in this patch: switching from calling pm_runtime_get_sync()
to calling pm_runtime_resume_and_get(). :-/

[...]

>>> Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
>>> Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
>> [...]

MBR, Sergey
diff mbox series

Patch

diff --git a/drivers/net/ethernet/renesas/ravb_main.c b/drivers/net/ethernet/renesas/ravb_main.c
index 342978bdbd7e..0486add302b3 100644
--- a/drivers/net/ethernet/renesas/ravb_main.c
+++ b/drivers/net/ethernet/renesas/ravb_main.c
@@ -2635,25 +2635,26 @@  static int ravb_probe(struct platform_device *pdev)
 		return dev_err_probe(&pdev->dev, PTR_ERR(rstc),
 				     "failed to get cpg reset\n");
 
+	error = reset_control_deassert(rstc);
+	if (error)
+		return error;
+
+	pm_runtime_enable(&pdev->dev);
+	error = pm_runtime_resume_and_get(&pdev->dev);
+	if (error < 0)
+		goto pm_runtime_disable;
+
 	ndev = alloc_etherdev_mqs(sizeof(struct ravb_private),
 				  NUM_TX_QUEUE, NUM_RX_QUEUE);
-	if (!ndev)
-		return -ENOMEM;
-
+	if (!ndev) {
+		error = -ENOMEM;
+		goto pm_runtime_put;
+	}
 	info = of_device_get_match_data(&pdev->dev);
 
 	ndev->features = info->net_features;
 	ndev->hw_features = info->net_hw_features;
 
-	error = reset_control_deassert(rstc);
-	if (error) {
-		free_netdev(ndev);
-		return error;
-	}
-
-	pm_runtime_enable(&pdev->dev);
-	pm_runtime_get_sync(&pdev->dev);
-
 	if (info->multi_irqs) {
 		if (info->err_mgmt_irqs)
 			irq = platform_get_irq_byname(pdev, "dia");
@@ -2878,8 +2879,9 @@  static int ravb_probe(struct platform_device *pdev)
 	clk_disable_unprepare(priv->refclk);
 out_release:
 	free_netdev(ndev);
-
+pm_runtime_put:
 	pm_runtime_put(&pdev->dev);
+pm_runtime_disable:
 	pm_runtime_disable(&pdev->dev);
 	reset_control_assert(rstc);
 	return error;