phy: tegra: xusb: Complete exception handling in five probe functions

phy: tegra: xusb: Complete exception handling in five probe functions

[Markus Elfring]

Hello,

I have taken another look also at the implementations of five probe functions.
I have got the impression then that the exception handling is incomplete so far
at these source code places.
https://elixir.bootlin.com/linux/v5.4-rc2/source/drivers/phy/tegra/xusb-tegra124.c#L608
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/tegra/xusb-tegra124.c?id=8005803a2ca0af49f36f6e9329b5ecda3df27347#n608

How do you think about to move kfree() calls to other jump targets
according to the Linux coding style?

Regards,
Markus

Re: phy: tegra: xusb: Complete exception handling in probe functions

[Markus Elfring]

> I have got the impression then that the exception handling is incomplete so far
> at these source code places.

How do you think about to try the following small script for the semantic patch
language out on the directory “drivers/phy/tegra”?

@display@
expression object;
identifier exit;
@@
 ... when any
 object = kzalloc(...)
 ...
 if (...)
 {
*kfree(object);
 goto exit;
 }
 ...
 if (...)
*   goto unregister;


Will any search pattern variations become more interesting for corresponding
automatic software transformations?

Regards,
Markus

Re: drivers/phy/tegra: Completion for exception handling in probe functions with SmPL?

[Markus Elfring]

> Will any search pattern variations become more interesting for corresponding
> automatic software transformations?

I hoped to achieve something together with the semantic patch language
by the following transformation approach.

@adjustment@
expression object;
identifier exit;
@@
 object = kzalloc(...)
 ...
 if (...)
-{  kfree(object);
    goto
-        exit
+        release_memory
    ;
-}
 ... when any
 device_unregister(...);
-exit
+release_memory
 :
+kfree(object);
 return ERR_PTR(...);


Do you find such a change suggestion reasonable (in principle)?

But I stumble on another unexpected test result.

elfring@Sonne:~/Projekte/Linux/next-patched> spatch drivers/phy/tegra/xusb-tegra186.c ~/Projekte/Coccinelle/janitor/complete_exception_handling_in_probe_functions1.cocci
init_defs_builtins: /usr/local/bin/../lib/coccinelle/standard.h
HANDLING: drivers/phy/tegra/xusb-tegra186.c


How would you like to clarify why diff hunks were not generated
for functions like “tegra186_usb2_pad_probe” and “tegra186_usb3_pad_probe”
in such an use case?
https://elixir.bootlin.com/linux/v5.4-rc2/source/drivers/phy/tegra/xusb-tegra186.c#L445
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/tegra/xusb-tegra186.c?id=bbf711682cd570697086e88388a2c718da918894#n445

Regards,
Markus

Re: drivers/phy/tegra: Completion for exception handling in probe functions with SmPL?

[Markus Elfring]

> But I stumble on another unexpected test result.

I got more promising results by the following transformation approach.

@adjustment exists@
expression object;
identifier exit;
@@
 object = kzalloc(...)
 ...
 if (...)
-{  kfree(object);
    goto
-        exit
+        release_memory
    ;
-}
 ... when any
 device_unregister(...);
-exit
+release_memory
 :
+kfree(object);
 return ERR_PTR(...);


The scope property of such a SmPL rule occasionally needs also more
software development attention.
https://github.com/coccinelle/coccinelle/blob/ed1eb8e06f800739d3992158d36945c0c4c6f0c7/docs/manual/cocci_syntax.tex#L860

I observe that the pretty-printing for the generated source code will need
further improvements (according to the Linux coding style).

Example:

elfring@Sonne:~/Projekte/Linux/next-patched> spatch drivers/phy/tegra/xusb-tegra186.c ~/Projekte/Coccinelle/janitor/complete_exception_handling_in_probe_functions6.cocci
…
@@ -461,10 +461,8 @@ tegra186_usb2_pad_probe(struct tegra_xus
 	pad->soc = soc;

 	err = tegra_xusb_pad_init(pad, padctl, np);
-	if (err < 0) {
-		kfree(usb2);
-		goto out;
-	}
+	if (err < 0)
+		goto release_memory;

 	priv->usb2_trk_clk = devm_clk_get(&pad->dev, "trk");
 	if (IS_ERR(priv->usb2_trk_clk)) {
@@ -483,7 +481,8 @@ tegra186_usb2_pad_probe(struct tegra_xus

 unregister:
 	device_unregister(&pad->dev);
-out:
+release_memory :
+kfree(usb2);
 	return ERR_PTR(err);
 }
…


How reasonable do you find the presented update suggestion?

Regards,
Markus