On 12/19/2016 10:30 AM, Chee, Tien Fong wrote:
On Isn, 2016-12-19 at 08:47 +0100, Marek Vasut wrote:
On 12/19/2016 07:53 AM, Chee, Tien Fong wrote:
On Jum, 2016-12-09 at 13:51 +0100, Marek Vasut wrote:
On 12/09/2016 11:04 AM, Chee, Tien Fong wrote:
On Rab, 2016-12-07 at 14:58 +0100, Marek Vasut wrote:
On 12/07/2016 12:58 PM, Chee, Tien Fong wrote: > > > > On Sel, 2016-12-06 at 13:55 +0100, Marek Vasut wrote: >> >> >> >> On 12/06/2016 09:08 AM, Chee Tien Fong wrote: >>> >>> >>> >>> >>> From: Tien Fong Chee tien.fong.chee@intel.com >>> >>> Drivers for reset manager is restructured such that >>> common >>> functions, >>> gen5 drivers and Arria10 drivers are moved to >>> reset_manager.c, >>> reset_manager_gen5.c and reset_manager_arria10.c >>> respectively. >>> >>> Signed-off-by: Tien Fong Chee <tien.fong.chee@intel.com >>>> >>> Cc: Marek Vasut marex@denx.de >>> Cc: Dinh Nguyen dinguyen@kernel.org >>> Cc: Chin Liang See chin.liang.see@intel.com >>> Cc: Tien Fong skywindctf@gmail.com [...]
> > > >> >> >> >>> >>> >>> >>> +void reset_deassert_dedicated_peripherals(void) >>> +{ >>> + int i; >>> + u32 mask0 = 0; >>> + u32 mask1 = 0; >>> + u32 pinmux_addr = >>> SOCFPGA_PINMUX_DEDICATED_IO_ADDRESS; >>> + u32 mask = 0; >>> +#if defined(CONFIG_MMC) >>> + mask |= >>> ALT_RSTMGR_PER0MODRST_SDMMCECC_SET_MSK; >>> +#elif defined(CONFIG_CADENCE_QSPI) >>> + mask |= ALT_RSTMGR_PER0MODRST_QSPIECC_SET_MSK; >>> +#elif defined(CONFIG_NAND_DENALI) >>> + mask |= ALT_RSTMGR_PER0MODRST_NANDECC_SET_MSK; >>> +#else >> Shouldn't this come from OF instead of being ifdef'd ? >> > What is OF? Device Tree (Open Firmware).
> > > > what is your suggestion to make this function generic for > all type of flash? Pull it from OF ?
Why you prefer device tree implementation over #define in defconfig, because there is performance penalty.
Because we are moving away from excessive random #defines and toward having one single binary where you could exchange just the DT and run it on multiple boards, just like Linux, that is the ultimate goal.
Also, this is not performance critical code, is it.
This code just to release peripherals from reset, not performance critical codes. However, our defconfigs based on flash type booting, so this is why i din't use the DT, since we have flash type determined from defconfig. Since DT is ultimate goal, i can change to DT implementation.
You should really only unreset the peripherals which you need and, ideally, only when you need them. So it should be the driver which handles the peripheral reset, not the common code (unless there is a reason for the common code to do it).
Yeah, i agree with you too. Did our cyclone5 implement this also? Does DM has framework to support user define reset mechanism?
The CV does it at least for ethernet . I don't think there is a dedicated mechanism yet.