On Mon, 12 Nov 2018 10:37:10 -0800 Tony Lindgren <tony@xxxxxxxxxxx> wrote: > * Andreas Kemnade <andreas@xxxxxxxxxxxx> [181108 18:40]: > > On Thu, 8 Nov 2018 09:06:37 -0800 > > Tony Lindgren <tony@xxxxxxxxxxx> wrote: > > > > > * Andreas Kemnade <andreas@xxxxxxxxxxxx> [181106 21:37]: > > > > Hi Tony, > > > > > > > > On Tue, 6 Nov 2018 12:39:14 -0800 > > > > Tony Lindgren <tony@xxxxxxxxxxx> wrote: > > > > > > > > > * Andreas Kemnade <andreas@xxxxxxxxxxxx> [181106 20:24]: > > > > > > Mainline kernel gives only about 100KByte/s, with sdio irqs I get > > > > > > > 1000KByte/s. > > > > > > > > > > That does not sound right at all. To me it sounds like the SDIO > > > > > irq is not working at all during runtime for your SoC? Is that > > > > > > > > no wakeirq, no sdio irq: > > Maybe this could be misunderstood: > > If you do not specify a wakeirq, sdio irq handling is not enabled at all: > > > > > > ret = omap_hsmmc_configure_wake_irq(host); > > > > if (!ret) > > > > mmc->caps |= MMC_CAP_SDIO_IRQ; > > > > > > > We are using this here to have at least decent performance, low latency and reliable > > 802.11 psmode (and therefore dramatically reduced power consumption of wifi module): > > http://git.goldelico.com/?p=letux-kernel.git;a=blobdiff;f=drivers/mmc/host/omap_hsmmc.c;h=bbfcef54b88cb534a49b1542e60c99e545a9e8a0;hp=68760d4a5d3da1a53c655d1b33f1c2237273d136;hb=44c8fc73ad8291c821855eaa88c13d5f9f2920f7;hpb=b48b43ee804e840de2e666873675d158b8219473 > > Looking at the commit above, looks like the sdio irq handling is > missing runtime PM calls.. So I wonder if we can just do the > following? Well, if we have runtime PM calls there, the device will never runtime-suspend with sdio irq enabled, so we do not need a working wakeup irq. > > In theory setting the wlan mmc instance autosuspend_timeout_ms to -1 > in sysfs should then do what your patch is doing except for enabling > sdio irq even without a wakeirq configured. > Well, if sdio irq is not enabled, then the mentioned functions will never be called, so the patch will not change anything... But what might be a better idea is to guard this thing by if (host->wakeirq) Regards, Andreas
Attachment:
pgpfnwjsWHV6Z.pgp
Description: OpenPGP digital signature
