[RFC] mmc: sdhci: implement light and heavy runtime pm resume

Jisheng Zhang jszhang at marvell.com
Thu Feb 5 02:20:31 PST 2015


Hi all,

Most sdhci_runtime_resume_host() users just gate their clocks in their runtime
suspend implementation, so the sdhci_runtime_resume_host() is too heavy, we only
need to restore interrupts registers.

The change is something as following:

diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
index 0ad412a..9d5bdde 100644
--- a/drivers/mmc/host/sdhci.c
+++ b/drivers/mmc/host/sdhci.c
@@ -2856,7 +2856,7 @@ int sdhci_runtime_suspend_host(struct sdhci_host *host)
 }
 EXPORT_SYMBOL_GPL(sdhci_runtime_suspend_host);
 
-int sdhci_runtime_resume_host(struct sdhci_host *host)
+static void sdhci_runtime_resume_reinit(struct sdhci_host *host)
 {
 	unsigned long flags;
 	int host_flags = host->flags;
@@ -2880,6 +2880,15 @@ int sdhci_runtime_resume_host(struct sdhci_host *host)
 		sdhci_enable_preset_value(host, true);
 		spin_unlock_irqrestore(&host->lock, flags);
 	}
+}
+
+int sdhci_runtime_resume_host(struct sdhci_host *host, bool power_keeped)
+{
+	unsigned long flags;
+	int host_flags = host->flags;
+
+	if (!power_keeped)
+		sdhci_runtime_resume_reinit(host);
 
 	/* Set the re-tuning expiration flag */
 	if (host->flags & SDHCI_USING_RETUNING_TIMER)


It's ugly, but per my test on marvell BG2Q platforms, this change can improve
the resume latency from 5500us to 11us. Is there any better solution?

Thanks in advance,
Jisheng



More information about the linux-arm-kernel mailing list