| Message ID | 1352237838-22552-1-git-send-email-andrew@bradfordembedded.com |
|---|---|
| State | Rejected |
| Delegated to: | Tom Rini |
| Headers | show |
Dear Tom, On Tue, 6 Nov 2012 16:37:18 -0500 Andrew Bradford <andrew@bradfordembedded.com> wrote: > The default value for PRM_RSTTIME1 on am335x is 0x06 leading to a very > short assertion of SYS_RESETn on BeagleBones. For PRM_RSTTIME1 values > less than 0x80, some BeagleBones will end up with inverted Ethernet > LEDs upon warm software reset leading to incorrect display of the > Ethernet link status on the Ethernet jack. > > Extend PRM_RSTTIME1 to 0x80 to provide a longer assertion of > SYS_RESETn allowing the SMSC PHY more time to sample the strapping > resistors which set the LED orientation. > > Signed-off-by: Andrew Bradford <andrew@bradfordembedded.com> > --- > arch/arm/include/asm/arch-am33xx/cpu.h | 1 + > board/ti/am335x/board.c | 8 ++++++++ > 2 files changed, 9 insertions(+) > > diff --git a/arch/arm/include/asm/arch-am33xx/cpu.h > b/arch/arm/include/asm/arch-am33xx/cpu.h index 819fd2f..bcd2fb0 100644 > --- a/arch/arm/include/asm/arch-am33xx/cpu.h > +++ b/arch/arm/include/asm/arch-am33xx/cpu.h > @@ -53,6 +53,7 @@ > /* Reset control */ > #ifdef CONFIG_AM33XX > #define PRM_RSTCTRL 0x44E00F00 > +#define PRM_RSTTIME 0x44E00F04 > #define PRM_RSTST 0x44E00F08 > #endif > #define PRM_RSTCTRL_RESET 0x01 > diff --git a/board/ti/am335x/board.c b/board/ti/am335x/board.c > index b56a801..7c23a6e 100644 > --- a/board/ti/am335x/board.c > +++ b/board/ti/am335x/board.c > @@ -304,6 +304,14 @@ void s_init(void) > else > config_ddr(266, MT47H128M16RT25E_IOCTRL_VALUE, > &ddr2_data, &ddr2_cmd_ctrl_data, &ddr2_emif_reg_data); > + > + if (board_is_bone()) > + /* > + * For BeagleBone, extend PRM_RSTTIME1 to ensure > SMSC PHY > + * strapping resistors are properly read on warm > software reset > + * to reduce the chance of Ethernet LEDs being > inverted. > + */ > + writeb(0x80, PRM_RSTTIME); > #endif > } > NAK. Please do not apply this patch. It does not accomplish the desired goal (fixing Ethernet LED inversion) when some capes are connected to BeagleBone and is not be the proper way to resolve the issue. Thanks, Andrew
diff --git a/arch/arm/include/asm/arch-am33xx/cpu.h b/arch/arm/include/asm/arch-am33xx/cpu.h index 819fd2f..bcd2fb0 100644 --- a/arch/arm/include/asm/arch-am33xx/cpu.h +++ b/arch/arm/include/asm/arch-am33xx/cpu.h @@ -53,6 +53,7 @@ /* Reset control */ #ifdef CONFIG_AM33XX #define PRM_RSTCTRL 0x44E00F00 +#define PRM_RSTTIME 0x44E00F04 #define PRM_RSTST 0x44E00F08 #endif #define PRM_RSTCTRL_RESET 0x01 diff --git a/board/ti/am335x/board.c b/board/ti/am335x/board.c index b56a801..7c23a6e 100644 --- a/board/ti/am335x/board.c +++ b/board/ti/am335x/board.c @@ -304,6 +304,14 @@ void s_init(void) else config_ddr(266, MT47H128M16RT25E_IOCTRL_VALUE, &ddr2_data, &ddr2_cmd_ctrl_data, &ddr2_emif_reg_data); + + if (board_is_bone()) + /* + * For BeagleBone, extend PRM_RSTTIME1 to ensure SMSC PHY + * strapping resistors are properly read on warm software reset + * to reduce the chance of Ethernet LEDs being inverted. + */ + writeb(0x80, PRM_RSTTIME); #endif }
The default value for PRM_RSTTIME1 on am335x is 0x06 leading to a very short assertion of SYS_RESETn on BeagleBones. For PRM_RSTTIME1 values less than 0x80, some BeagleBones will end up with inverted Ethernet LEDs upon warm software reset leading to incorrect display of the Ethernet link status on the Ethernet jack. Extend PRM_RSTTIME1 to 0x80 to provide a longer assertion of SYS_RESETn allowing the SMSC PHY more time to sample the strapping resistors which set the LED orientation. Signed-off-by: Andrew Bradford <andrew@bradfordembedded.com> --- arch/arm/include/asm/arch-am33xx/cpu.h | 1 + board/ti/am335x/board.c | 8 ++++++++ 2 files changed, 9 insertions(+)