diff options
| author | Patrick Wildt <patrick@cvs.openbsd.org> | 2020-05-23 22:16:40 +0000 |
|---|---|---|
| committer | Patrick Wildt <patrick@cvs.openbsd.org> | 2020-05-23 22:16:40 +0000 |
| commit | ee3c87014fda26b81b8f347de9bc36c4d2b34b1b (patch) | |
| tree | d321a96595b946a13c1ee4c43cfe523dbca7978b /sys/dev/fdt | |
| parent | 482ea2ca42e0c270f24266923cb8b6070133e33c (diff) | |
Increase chances of getting a successful PCIe link on the i.MX8MM.
Hold the reset GPIO for 100ms instead of 20ms, and also directly
continue setting up the link instead of waiting for another 20ms.
This brings us in line with what Linux is doing and removes the 10%
risk of em(4) not showing up on my HummingBoard Pulse.
ok kettenis@
Diffstat (limited to 'sys/dev/fdt')
| -rw-r--r-- | sys/dev/fdt/dwpcie.c | 5 |
1 files changed, 2 insertions, 3 deletions
diff --git a/sys/dev/fdt/dwpcie.c b/sys/dev/fdt/dwpcie.c index 5a4b1f29e2f..cb91df25940 100644 --- a/sys/dev/fdt/dwpcie.c +++ b/sys/dev/fdt/dwpcie.c @@ -1,4 +1,4 @@ -/* $OpenBSD: dwpcie.c,v 1.20 2020/04/27 12:17:29 patrick Exp $ */ +/* $OpenBSD: dwpcie.c,v 1.21 2020/05/23 22:16:39 patrick Exp $ */ /* * Copyright (c) 2018 Mark Kettenis <kettenis@openbsd.org> * @@ -768,9 +768,8 @@ dwpcie_imx8mq_init(struct dwpcie_softc *sc) if (reset_gpiolen > 0) { gpio_controller_set_pin(reset_gpio, 1); - delay(20000); + delay(100000); gpio_controller_set_pin(reset_gpio, 0); - delay(20000); } reset_deassert(sc->sc_node, "pciephy"); |