From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tony Lindgren Subject: Re: [PATCH] ARM: dts: am437x-gp-evm: Do not reset gpio5 Date: Fri, 25 Apr 2014 09:46:14 -0700 Message-ID: <20140425164614.GE20807@atomide.com> References: <1395379213-15451-1-git-send-email-lokeshvutla@ti.com> <53319A7D.1060904@ti.com> <20140417163407.GB23385@atomide.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20140417163407.GB23385@atomide.com> Sender: linux-omap-owner@vger.kernel.org To: Nishanth Menon Cc: Lokesh Vutla , linux-omap@vger.kernel.org, bcousson@baylibre.com, linux-arm-kernel@lists.infradead.org, devicetree@vger.kernel.org, nsekhar@ti.com, rnayak@ti.com, mark.rutland@arm.com, Dave Gerlach List-Id: devicetree@vger.kernel.org * Tony Lindgren [140417 09:34]: > * Nishanth Menon [140325 08:05]: > > On 03/21/2014 12:20 AM, Lokesh Vutla wrote: > > > From: Dave Gerlach > > > > > > Do not reset GPIO5 at boot-up because GPIO5_7 is used > > > on AM437x GP-EVM to control VTT regulators on DDR3. > > > Without this some GP-EVM boards will fail to boot because > > > of DDR3 corruption. > > How funny :) > > Ideally we would be able to specify which GPIO pins should > maintain their state during the boot. > > AFAIK, this patch currently means that the kernel has no idea > what state the whole GPIO bank is in. At minimum we should > parse the GPIO bank state so the kernel knows it and then it > should be safe to set the no reset flag. > > So for the workaround, can you guys please try to see test > if the old mux trick in the bootloader works to mux the pin > into something PIN_INPUT_PULLUP | MUX_MODE7? Or a PULLDOWN > depending on the direction naturally. That would allow > leaving out the GPIO completely from this. OK so no safe mode as MUX_MODE7 on am335x. So based on the tests done by Dave on various GPIO banks without the reset, this seems OK to do. So applying into omap-for-v3.15/fixes-v2 to get the board booting properly. Naturally this is not a reason to stop further work on making sure the GPIO driver actually knows what state the bank is as suggested by Felipe. Tony