From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gary Bisson Date: Wed, 20 May 2020 10:15:29 +0200 Subject: [Buildroot] [PATCH v2 2/5] package/freescale-imx/firmware-imx: add choice for DDR training binaries In-Reply-To: <1588151545-9346-3-git-send-email-stephane.viau@oss.nxp.com> References: <1588151545-9346-1-git-send-email-stephane.viau@oss.nxp.com> <1588151545-9346-3-git-send-email-stephane.viau@oss.nxp.com> Message-ID: <20200520081529.GC1349614@p1g2> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hi Stephane, On Wed, Apr 29, 2020 at 11:12:22AM +0200, Stephane Viau wrote: > Several i.MX8 (e.g.: 8M, 8MM, 8MN) support many DDR types (LPDDR4, DDR4, > etc.), for which the DDR training is performed in the bootloader. > Some boards have LPDDR4 (e.g.: nitrogen8mn) and some others have the DDR4 > (e.g.: NXP's reference board EVK). This patch allows the selection of either > of the binaries used to train the DDR. > > Signed-off-by: Stephane Viau > Reviewed-by: Maeva Manuel > Reviewed-by: Julien Olivain > --- > v2: > - use BR2_PACKAGE_FREESCALE_IMX_NEED_DDR_FW to extend the DDR > firmware selection for 8M, 8MM and 8MN (suggested by Gary) > > Signed-off-by: Stephane Viau > --- > package/freescale-imx/firmware-imx/Config.in | 24 ++++++++++++++++++ > package/freescale-imx/firmware-imx/firmware-imx.mk | 29 +++++++++++++++++++++- > 2 files changed, 52 insertions(+), 1 deletion(-) > > diff --git a/package/freescale-imx/firmware-imx/Config.in b/package/freescale-imx/firmware-imx/Config.in > index 56d5b80..4962992 100644 > --- a/package/freescale-imx/firmware-imx/Config.in > +++ b/package/freescale-imx/firmware-imx/Config.in > @@ -8,3 +8,27 @@ config BR2_PACKAGE_FIRMWARE_IMX > > This library is provided by Freescale as-is and doesn't have > an upstream. > + > +if BR2_PACKAGE_FREESCALE_IMX_NEED_DDR_FW Here is the only place where the previous variable is used, not sure it brings a lot of value. > +choice > + bool "DDR training binaries" > + default BR2_PACKAGE_FIRMWARE_DDRFW_LPDDR4 > + help > + Choose the DDR training binaries to be used depending on the > + kind of memory that is available on the target board (DDR4, > + LPDDR4, etc...). > + > +config BR2_PACKAGE_FIRMWARE_DDRFW_LPDDR4 > + bool "lpddr4" > + help > + Use LPDDR4 binaries (i.e.: lpddr4_pmu_train_*.bin) > + > +config BR2_PACKAGE_FIRMWARE_DDRFW_DDR4 > + bool "DDR4" > + help > + Use DDR4 binaries (i.e.: ddr4_*_201810.bin). > + > +endchoice # DDR training FW > + > +endif > diff --git a/package/freescale-imx/firmware-imx/firmware-imx.mk b/package/freescale-imx/firmware-imx/firmware-imx.mk > index cd0dafb..fc2f69a 100644 > --- a/package/freescale-imx/firmware-imx/firmware-imx.mk > +++ b/package/freescale-imx/firmware-imx/firmware-imx.mk > @@ -18,7 +18,7 @@ define FIRMWARE_IMX_EXTRACT_CMDS > $(call FREESCALE_IMX_EXTRACT_HELPER,$(FIRMWARE_IMX_DL_DIR)/$(FIRMWARE_IMX_SOURCE)) > endef > > -ifeq ($(BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX8M)$(BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX8MM)$(BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX8MN),y) > +ifeq ($(BR2_PACKAGE_FIRMWARE_DDRFW_LPDDR4),y) > FIRMWARE_IMX_INSTALL_IMAGES = YES > FIRMWARE_IMX_DDRFW_DIR = $(@D)/firmware/ddr/synopsys > define FIRMWARE_IMX_PREPARE_LPDDR4_FW > @@ -42,9 +42,36 @@ define FIRMWARE_IMX_INSTALL_IMAGES_CMDS > cat $(FIRMWARE_IMX_DDRFW_DIR)/lpddr4_pmu_train_1d_fw.bin \ > $(FIRMWARE_IMX_DDRFW_DIR)/lpddr4_pmu_train_2d_fw.bin > \ > $(BINARIES_DIR)/lpddr4_pmu_train_fw.bin > + ln -sf $(BINARIES_DIR)/lpddr4_pmu_train_fw.bin $(BINARIES_DIR)/ddr_fw.bin > cp $(@D)/firmware/hdmi/cadence/signed_hdmi_imx8m.bin \ > $(BINARIES_DIR)/signed_hdmi_imx8m.bin And here is why I'm worried the name of the previous variable might be misleading. You don't only copy the DDR FW training under that BR2_PACKAGE_FIRMWARE_DDRFW_LPDDR4 macro, you also copy the HDMI FW. Note that the DP FW should be added as well. The way I see it, it'd be best to have a common FIRMWARE_IMX_INSTALL_IMAGES_CMDS for all i.MX8M platforms (hence not necessarily using a macro for it). Then this CMD could have FIRMWARE_IMX_PREPARE_DDR_FW which would be different depending on the choice made before (DDR4 vs. LPDDR4) and then also something like FIRMWARE_IMX_PREPARE_HDMI_FW which would only be populated for IMX8M and empty for the others. Does it make sense? > endef > +else ifeq ($(BR2_PACKAGE_FIRMWARE_DDRFW_DDR4),y) > +FIRMWARE_IMX_INSTALL_IMAGES = YES > +FIRMWARE_IMX_DDRFW_DIR = $(@D)/firmware/ddr/synopsys > +define FIRMWARE_IMX_PREPARE_DDR4_FW > + $(TARGET_OBJCOPY) -I binary -O binary --pad-to 0x8000 --gap-fill=0x0 \ > + $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_imem_$(1)_201810.bin \ > + $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_imem_$(1)_201810_pad.bin > + $(TARGET_OBJCOPY) -I binary -O binary --pad-to 0x4000 --gap-fill=0x0 \ > + $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_dmem_$(1)_201810.bin \ > + $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_dmem_$(1)_201810_pad.bin > + cat $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_imem_$(1)_201810_pad.bin \ > + $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_dmem_$(1)_201810_pad.bin > \ > + $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_$(1)_201810_fw.bin > +endef > + > +define FIRMWARE_IMX_INSTALL_IMAGES_CMDS > + # Create padded versions of ddr4_* and generate ddr4_fw.bin. > + # ddr4_fw.bin is needed when generating imx8-boot-sd.bin > + # which is done in post-image script. > + $(call FIRMWARE_IMX_PREPARE_DDR4_FW,1d) > + $(call FIRMWARE_IMX_PREPARE_DDR4_FW,2d) > + cat $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_1d_201810_fw.bin \ > + $(FIRMWARE_IMX_DDRFW_DIR)/ddr4_2d_201810_fw.bin > \ > + $(BINARIES_DIR)/ddr4_201810_fw.bin > + ln -sf $(BINARIES_DIR)/ddr4_201810_fw.bin $(BINARIES_DIR)/ddr_fw.bin > +endef The approach above would also fix this case: - What about i.MX8MQ with DDR4? HDMI FW is missing from this code. Let me know if you have any questions. Regards, Gary