Hi,

debugged it further, but currently i don't know why CONFIG_OF_LIVE causes this behaviour

devfdt_get_addr_index return with FDT_ADDR_T_NONE adress-cells: -4

./scripts/dtc/libfdt/fdt_addresses.c: fdt_address_cells => fdt_cells => fdt_getprop (returns c=0,len=-4)

as far as i see in arch/arm/dts/mt7622.dtsi eth-node has #address-cells = <1>; so i wonder why fdt_address_cells failes with OF_LIVE

maybe adress-cells is needed also in ethsys/sgmiisys??

so currently i try to avoid including the CONFIG_OF_LIVE for my driver-porting (using fdt_get_property instead of of_find_property), but this should be fixed anyway.

regards Frank

Gesendet: Mittwoch, 05. August 2020 um 19:42 Uhr Von: "Frank Wunderlich" frank-w@public-files.de An: "Frank Wunderlich" frank-w@public-files.de Cc: "U-Boot-Denx" u-boot@lists.denx.de, "Landen Chao (趙皎宏)" Landen.Chao@mediatek.com, "Ryder Lee" ryder.lee@mediatek.com, "Mark Lee" Mark-MC.Lee@mediatek.com, "GSS_MTK_Uboot_upstream" GSS_MTK_Uboot_upstream@mediatek.com Betreff: Aw: crash with CONFIG_OF_LIVE

Hi

i traced it as far as i can

crash happens in

drivers/net/mtk_eth.c in function mtk_gmac_write (priv->gmac_base = 0xffff)

but root cause is iobase in probe

static int mtk_eth_probe(struct udevice *dev) { struct eth_pdata *pdata = dev_get_platdata(dev); struct mtk_eth_priv *priv = dev_get_priv(dev); ulong iobase = pdata->iobase; //<<<<<<<<<<<<< 0xffffffff printf("DEBUG: %s:%d iobase: 0x%0x gmac_base: 0x%0x\n",__FUNCTION__,__LINE__,(u32)iobase, GMAC_BASE);

this crash also happens in 64bit-mode (mt7622_rfb_defconfig + CONFIG_OF_LIVE)

any idea why iobase is wrong here?

regards Frank

...

Gesendet: Dienstag, 04. August 2020 um 16:58 Uhr Von: "Frank Wunderlich" frank-w@public-files.de An: "U-Boot-Denx" u-boot@lists.denx.de Betreff: crash with CONFIG_OF_LIVE

Hi,

i need of_find_property function so i added CONFIG_OF_LIVE.

with this my uboot crash on bpi-r64 (own 32bit-config) before calling anything in my driver (is an ahci-driver which is triggered by "scsi scan"). so only the inclusion of CONFIG_OF_LIVE. do i miss anything or have anyone else this issue?

Just for reference my current tree https://github.com/frank-w/u-boot/commits/2020-07-bpi-r64-pciev2

regards Frank