| Message ID | 1357694580-20582-1-git-send-email-amartin@nvidia.com |
|---|---|
| State | Superseded |
| Headers | show |
On 01/08/2013 06:23 PM, Allen Martin wrote:
> UART2_TXD and UART2_RXD mux 0 SFIO entries should be IRDA not UARTB.
IRDA is just a needlessly different synonym for UARTB; there shouldn't
be any mention of IRDA in the pinmux code anywhere, or any users of the
pinmux code.
On Tue, Jan 08, 2013 at 07:46:03PM -0800, Stephen Warren wrote: > On 01/08/2013 06:23 PM, Allen Martin wrote: > > UART2_TXD and UART2_RXD mux 0 SFIO entries should be IRDA not UARTB. > > IRDA is just a needlessly different synonym for UARTB; there shouldn't > be any mention of IRDA in the pinmux code anywhere, or any users of the > pinmux code. Oh suck, I didn't realize we had synonyms in the pinmux tables, that seems like a really bad idea. Unfortunately it looks like this particular synonym is widespread. I see it used in the pinmux spreadsheets, the downstream Android kernel source, and gfshell code. The bug I was trying to fix here is that IRDA is referenced in the u-boot cardhu pinmux table. I guess the fix is to change that to UARTB, but it sounds like maybe a larger synonym cleanup is needed everywhere, do you know of any others? I'm thinking there needs to be some better run time error checking in the u-boot pinmux code too. There are some debug asserts in there now, but because I didn't have debug turned on this particular bug lead to silent corruption of the pinmux registers which was a PITA to track down. Thoughts on turning those asserts into always on error checking code? Not sure how much it will slow down pinmux programming. -Allen
On 01/09/2013 10:58 AM, Allen Martin wrote: > On Tue, Jan 08, 2013 at 07:46:03PM -0800, Stephen Warren wrote: >> On 01/08/2013 06:23 PM, Allen Martin wrote: >>> UART2_TXD and UART2_RXD mux 0 SFIO entries should be IRDA not UARTB. >> >> IRDA is just a needlessly different synonym for UARTB; there shouldn't >> be any mention of IRDA in the pinmux code anywhere, or any users of the >> pinmux code. > > Oh suck, I didn't realize we had synonyms in the pinmux tables, that > seems like a really bad idea. Unfortunately it looks like this > particular synonym is widespread. I see it used in the pinmux > spreadsheets, the downstream Android kernel source, and gfshell code. It will disappear from the downstream kernels once they've picked up the upstream pinctrl driver; I spent a fair bit of time correlating all the various documentation sources and eliminating duplicate names for the upstream kernel pinctrl driver. > The bug I was trying to fix here is that IRDA is referenced in the > u-boot cardhu pinmux table. Oh, I didn't notice that the second time round; I'm pretty sure I pointed it out in the very first review. Maybe that was just of the U-Boot pinmux driver and not the Cardhu file. > I guess the fix is to change that to > UARTB, but it sounds like maybe a larger synonym cleanup is needed > everywhere, do you know of any others? I can't recall for sure; I think there were some. I'd suggest comparing upstream U-Boot's pinmux driver with the upstream kernel's pinctrl driver, treating the kernel as canonical now. > I'm thinking there needs to be some better run time error checking in > the u-boot pinmux code too. There are some debug asserts in there > now, but because I didn't have debug turned on this particular bug > lead to silent corruption of the pinmux registers which was a PITA to > track down. Thoughts on turning those asserts into always on error > checking code? Not sure how much it will slow down pinmux programming. I doubt the speed would be significantly affected, unless an error triggers and the printf/... actually happens, but you don't really care about speed then.
diff --git a/arch/arm/cpu/tegra30-common/pinmux.c b/arch/arm/cpu/tegra30-common/pinmux.c index 122665f..8045fa1 100644 --- a/arch/arm/cpu/tegra30-common/pinmux.c +++ b/arch/arm/cpu/tegra30-common/pinmux.c @@ -149,8 +149,8 @@ const struct tegra_pingroup_desc tegra_soc_pingroups[PINGRP_COUNT] = { PINI(VI_MCLK, VI, VI, VI, VI, VI), PINI(VI_VSYNC, VI, DDR, RSVD2, VI, RSVD4), PINI(VI_HSYNC, VI, DDR, RSVD2, VI, RSVD4), - PINI(UART2_RXD, UART, UARTB, SPDIF, UARTA, SPI4), - PINI(UART2_TXD, UART, UARTB, SPDIF, UARTA, SPI4), + PINI(UART2_RXD, UART, IRDA, SPDIF, UARTA, SPI4), + PINI(UART2_TXD, UART, IRDA, SPDIF, UARTA, SPI4), PINI(UART2_RTS_N, UART, UARTA, UARTB, GMI, SPI4), PINI(UART2_CTS_N, UART, UARTA, UARTB, GMI, SPI4), PINI(UART3_TXD, UART, UARTC, RSVD2, GMI, RSVD4),
UART2_TXD and UART2_RXD mux 0 SFIO entries should be IRDA not UARTB. Signed-off-by: Allen Martin <amartin@nvidia.com> --- arch/arm/cpu/tegra30-common/pinmux.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)