Skip to contentSkip to conversation

mmc: sdhci: add error checking for mmc_add_host

PATCH In review Jaehoon Chung 2 participants 6 messages Apr 11, 2011

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 mmc: sdhci: add error checking for mmc_add_host

Sometimes we can't add the device,but we didn't check any error status. Need to check error status for mmc_add_host.
And Missing regulator disable/put. Fixed them.

[PATCH v4] : merged for v3 [PATCH 1/2] and [PATCH 2/2] reviewed on Wolfram Sang

---
 drivers/mmc/host/sdhci.c |   22 ++++++++++++++++++++--
 1 files changed, 20 insertions(+), 2 deletions(-)
drivers/mmc/host/sdhci.c +20 -2
index 9e15f41..e9cb09c 100644
@@ -2025,7 +2025,9 @@ int sdhci_add_host(struct sdhci_host *host)
 
 	mmiowb();
 
-	mmc_add_host(mmc);
+	ret = mmc_add_host(mmc);
+	if (unlikely(ret))
+		goto err_free_mmc;
 
 	printk(KERN_INFO "%s: SDHCI controller on %s [%s] using %s\n",
 		mmc_hostname(mmc), host->hw_name, dev_name(mmc_dev(mmc)),
@@ -2036,15 +2038,31 @@ int sdhci_add_host(struct sdhci_host *host)
 
 	return 0;
 
+err_free_mmc:
+	mmc_remove_host(host);
+
 #ifdef SDHCI_USE_LEDS_CLASS
+	led_classdev_ungregister(&host->led);
 reset:
+#endif
 	sdhci_reset(host, SDHCI_RESET_ALL);
 	free_irq(host->irq, host);
-#endif
+	del_timer_sync(&host->timer);
+	if (host->vmmc) {
+		regulator_disable(host->vmmc);
+		regulator_put(host->vmmc);
+	}
+
 untasklet:
 	tasklet_kill(&host->card_tasklet);
 	tasklet_kill(&host->finish_tasklet);
 
+	kfree(host->adma_desc);
+	kfree(host->align_buffer);
+
+	host->adma_desc = NULL;
+	host->align_buffer = NULL;
+
 	return ret;
 }
 
#1 lore raw
Quoted 3 lines · Jaehoon Chung

Sometimes we can't add the device,but we didn't check any error status. Need to check error status for mmc_add_host.
And Missing regulator disable/put. Fixed them.

Is there any special reason you don't want to answer the open question from the last two reviews?

Quoted 1 line

[PATCH v4] : merged for v3 [PATCH 1/2] and [PATCH 2/2] reviewed on Wolfram Sang

This should go below the dashed line (---)

Wolfram

Signature · 2 lines
Pengutronix e.K.                           | Wolfram Sang                |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#2 lore raw
Quoted 7 lines · Wolfram Sang
Quoted 3 lines · Jaehoon Chung

Sometimes we can't add the device,but we didn't check any error status. Need to check error status for mmc_add_host.
And Missing regulator disable/put. Fixed them.

Is there any special reason you don't want to answer the open question from the last two reviews?

Sorry..i didn't find your questions. So i didn't answer anything. You asked "why removed host?" right?

I just think good that remove the host...(any crash didn't occur). there is not special reason.

Quoted 7 lines
Quoted 2 lines

[PATCH v4] : merged for v3 [PATCH 1/2] and [PATCH 2/2] reviewed on Wolfram Sang

This should go below the dashed line (---)

Wolfram

#3 lore raw
Quoted 2 lines

I just think good that remove the host...(any crash didn't occur). there is not special reason.

Look at the mmc-core code, please (host.c). Compare what remove_host does and what add_host does before it fails. What do you think?

Regards,

Wolfram

Signature · 2 lines
Pengutronix e.K.                           | Wolfram Sang                |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#4 lore raw
Quoted 5 lines · Wolfram Sang
Quoted 2 lines

I just think good that remove the host...(any crash didn't occur). there is not special reason.

Look at the mmc-core code, please (host.c). Compare what remove_host does and what add_host does before it fails. What do you think?

Ok..i'm looking at the mmc-core code(host.c) mmc_add_host does devices_add(), then run start_host(). mmc_remove_host does stop_host().

You means that didn't run start_host() before add_host failed. I think that need not to run stop_host. right?

To use mmc_free_host, better than mmc_remove_host.

Regards,
Jaehoon Chung

Quoted 5 lines

Regards,

Wolfram

#5 lore raw
Quoted 3 lines

You means that didn't run start_host() before add_host failed. I think that need not to run stop_host. right?

Right, and other stuff like registering the pm_notifier.

Quoted 1 line

To use mmc_free_host, better than mmc_remove_host.

That would be true for most drivers. However, sdhci has encapsulated this in its own sdhci_free_host. So, you don't need to call this as well.

Regards,

Wolfram

Signature · 2 lines
Pengutronix e.K.                           | Wolfram Sang                |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |