Quoting Felipe Balbi <[email protected]>:

Content-Description: tahvo-correctness.patch
@@ -683,11 +682,23 @@ static int tahvo_usb_probe(struct platform_device *pdev)

        /* Attributes */
        ret = device_create_file(&(pdev->dev), &dev_attr_vbus_state);
+       if (ret) {
+               tahvo_free_irq(TAHVO_INT_VBUSON);
+               kfree(tu);
+               printk(KERN_ERR "attribute creation failed: %d\n", ret);

use dev_dbg(&pdev->dev, "...");

@@ -703,8 +714,12 @@ static int tahvo_usb_probe(struct platform_device *pdev)
        ret = otg_set_transceiver(&tu->otg);
        if (ret < 0) {
                printk(KERN_ERR "Cannot register USB transceiver\n");
-               kfree(tu);
                tahvo_free_irq(TAHVO_INT_VBUSON);
+               device_remove_file(&(pdev->dev), &dev_attr_vbus_state);
+#ifdef CONFIG_USB_OTG
+               device_remove_file(&(pdev->dev), &dev_attr_otg_mode);
+#endif
+               kfree(tu);
                return ret;
        }

@@ -718,6 +733,7 @@ static int tahvo_usb_probe(struct platform_device *pdev)

 static int tahvo_usb_remove(struct platform_device *pdev)
 {
+       struct tahvo_usb *tu = (struct tahvo_usb*) pdev->dev.driver_data;

tu = platform_get_drvdata(pdev);

diff --git a/drivers/cbus/tahvo.c b/drivers/cbus/tahvo.c
index 29fd4b8..09a69c0 100644
--- a/drivers/cbus/tahvo.c
+++ b/drivers/cbus/tahvo.c
@@ -298,6 +298,7 @@ static int __devinit tahvo_probe(struct platform_device *pdev)
        em_asic_config = omap_get_config(OMAP_TAG_EM_ASIC_BB5,
                                         struct omap_em_asic_bb5_config);
        if (em_asic_config == NULL) {
+               tasklet_kill(&tahvo_tasklet);
                printk(KERN_ERR PFX "Unable to retrieve config data\n");
                return -ENODATA;
        }
@@ -314,6 +315,7 @@ static int __devinit tahvo_probe(struct platform_device *pdev)
                tahvo_is_betty = 1;
                tahvo_7bit_backlight = 1;
        } else {
+               tasklet_kill(&tahvo_tasklet);
                printk(KERN_ERR "Tahvo/Betty chip not found");
                return -ENODEV;
        }
@@ -324,6 +326,7 @@ static int __devinit tahvo_probe(struct platform_device *pdev)
        tahvo_irq_pin = em_asic_config->tahvo_irq_gpio;

        if ((ret = gpio_request(tahvo_irq_pin, "TAHVO irq")) < 0) {
+               tasklet_kill(&tahvo_tasklet);
                printk(KERN_ERR PFX "Unable to reserve IRQ GPIO\n");
                return ret;
        }
@@ -342,6 +345,7 @@ static int __devinit tahvo_probe(struct platform_device *pdev)
        if (ret < 0) {
                printk(KERN_ERR PFX "Unable to register IRQ handler\n");
                gpio_free(tahvo_irq_pin);
+               tasklet_kill(&tahvo_tasklet);
                return ret;
        }
 #ifdef CONFIG_CBUS_TAHVO_USER
@@ -350,6 +354,7 @@ static int __devinit tahvo_probe(struct platform_device *pdev)
                printk(KERN_ERR "Unable to initialize driver\n");
                free_irq(gpio_to_irq(tahvo_irq_pin), 0);
                gpio_free(tahvo_irq_pin);
+               tasklet_kill(&tahvo_tasklet);
                return ret;
        }
 #endif

I would like to see this as a series of goto like:

if (em_asic_config == NULL) {
        ret = -ENODATA;
        goto fail1;
}

...

ret = gpio_request(tahvo_irq_pin, "TAHVO irq");
if (ret < 0) {
        ...
        goto failN;
}

failN:
        ...

...

fail2:
        tasklet_kill(&tahvo_tasklet);

fail1:
        return ret;

Hi - the current phase I'm in of this development is to simply get the thing working reliably again; that includes cleanup etc as appropriate.

Once that is complete, i want to go back and refactor all the drivers s that they do exactly what you suggest above, as well as using the latest kernel APIs, modularisation where appropriate etc, with a view to being suitable for mainline (even if they are not actually submitted to mainline).

I don't want to do both at once because I feel that is too big a step for this older slightly unmaintained code. Is it acceptable to you to commit this patch in the interim?

--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to