From mboxrd@z Thu Jan 1 00:00:00 1970 From: Marek Vasut Date: Fri, 26 Jan 2018 17:19:01 +0100 Subject: [U-Boot] [PATCH] imx-common: spl: Allow booting from eMMC when SPL is loaded from non-eMMC device In-Reply-To: References: <20180126145723.15991-1-lukma@denx.de> Message-ID: <65cf6ac0-af15-2dc5-3c5d-5d1ac6e8b464@denx.de> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On 01/26/2018 05:16 PM, Marek Vasut wrote: > On 01/26/2018 05:05 PM, Stefano Babic wrote: >> On 26/01/2018 15:57, Lukasz Majewski wrote: >>> This patch tries to solve the problem described in following patch: >>> https://patchwork.ozlabs.org/patch/796237/ >>> >>> The main argument against the above code was the potential lack of >>> consistency if we boot SPL from the SD card (and then eMMC may load >>> u-boot proper). >>> >>> This patch preserves this consistency if spl_boot_device() detects boot >>> from either SD card or eMMC. >>> >>> It only will change boot device if boot from non-SD/eMMC device is >>> detected - i.e SPI-NOR (as in this case). >>> >>> Signed-off-by: Lukasz Majewski >>> --- >>> >>> arch/arm/mach-imx/spl.c | 7 ++++++- >>> 1 file changed, 6 insertions(+), 1 deletion(-) >>> >>> diff --git a/arch/arm/mach-imx/spl.c b/arch/arm/mach-imx/spl.c >>> index 6c16872f59..735d9f6261 100644 >>> --- a/arch/arm/mach-imx/spl.c >>> +++ b/arch/arm/mach-imx/spl.c >>> @@ -134,7 +134,12 @@ int g_dnl_bind_fixup(struct usb_device_descriptor *dev, const char *name) >>> /* called from spl_mmc to see type of boot mode for storage (RAW or FAT) */ >>> u32 spl_boot_mode(const u32 boot_device) >>> { >>> - switch (spl_boot_device()) { >>> + u32 spl_bd = spl_boot_device(); >>> + >>> + if (spl_bd != BOOT_DEVICE_MMC1) >>> + spl_bd = boot_device; >>> + >>> + switch (spl_bd) { >>> /* for MMC return either RAW or FAT mode */ >>> case BOOT_DEVICE_MMC1: >>> case BOOT_DEVICE_MMC2: >>> >> >> But have we not board_boot_order() for such as cases ? It is a wek >> function and you can define it in your board code. Put in mach-imx, it >> is valid for all boards. > > After a bit of offline discussion with Lukasz, I think I understand what > the problem is. > > The SPL common code calls spl_boot_device() to determine which boot > device to boot from next ; in his case, SPI or eMMC. In case eMMC is > selected, the SPL MMC common code then calls spl_boot_mode() (different > function than spl_boot_device()) to determine whether the content on > SD/eMMC is RAW or on a FS. > > The iMX implementation (and layerspace too) does extra stuff in the > implementation of spl_boot_mode() by calling spl_boot_device() again, > which is wrong. > > spl_boot_mode() is only used by spl_mmc.c , so we are certain the system > is booting from SD/MMC . By turning this around, if the board code did > override the board_boot_order() to load the next payload from > SD/MMC/eMMC AND started SPL from non-SD/MMC/eMMC device , calling > spl_boot_device() in spl_boot_mode() implementation will return bogus > result and the spl_boot_mode() will fail in the default: case. > > Thus, the fix here is to do what ie. socfpga does, just discern whether > the content on SD/MMC/eMMC is RAW or FS and do not do any extra checks. By doing so, we can probably even clean this up and have default __weak spl_boot_mode() implementation rather than having a custom copy for each SoC. -- Best regards, Marek Vasut