Re: [U-Boot] [PATCH 2/2] arm: imx6: Enable DDR calibration on Novena
On Sunday, December 20, 2015 at 08:33:58 PM, Eric Nelson wrote:
Hi Marek,
On 12/16/2015 07:40 AM, Marek Vasut wrote:
Enable the DDR calibration functionality on Novena to deal with the memory SoDIMM on this board.
Shouldn't this be in two patches?
Not really, the old values work without the enabled calibration. This change needs to be done atomically.
[...]
static void ccgr_init(void)
@@ -601,6 +601,11 @@ void board_init_f(ulong dummy)
mx6dq_dram_iocfg(64, &novena_ddr_ioregs, &novena_grp_ioregs); mx6_dram_cfg(&novena_ddr_info, &novena_mmdc_calib, &elpida_4gib_1600);
- /* Perform DDR DRAM calibration */
- udelay(100);
Shouldn't the return values be tested?
I guess yes, but if the calibration fails, that what ? It's game over ;-)
I will just change those functions to void type and do hang() if we get errors. I believe that's the most sane way to handle it and it prevents development of boilerplate code.
mmdc_do_write_level_calibration();
mmdc_do_dqs_calibration();
/* Clear the BSS. */ memset(__bss_start, 0, __bss_end - __bss_start);
Best regards, Marek Vasut
Hi Marek,
On 12/22/2015 03:26 AM, Marek Vasut wrote:
On Sunday, December 20, 2015 at 08:33:58 PM, Eric Nelson wrote:
Hi Marek,
On 12/16/2015 07:40 AM, Marek Vasut wrote:
Enable the DDR calibration functionality on Novena to deal with the memory SoDIMM on this board.
Shouldn't this be in two patches?
Not really, the old values work without the enabled calibration. This change needs to be done atomically.
[...]
static void ccgr_init(void)
@@ -601,6 +601,11 @@ void board_init_f(ulong dummy)
mx6dq_dram_iocfg(64, &novena_ddr_ioregs, &novena_grp_ioregs); mx6_dram_cfg(&novena_ddr_info, &novena_mmdc_calib, &elpida_4gib_1600);
- /* Perform DDR DRAM calibration */
- udelay(100);
Shouldn't the return values be tested?
I guess yes, but if the calibration fails, that what ? It's game over ;-)
Do you think it's possible/practical to reboot the system in this case?
I will just change those functions to void type and do hang() if we get errors. I believe that's the most sane way to handle it and it prevents development of boilerplate code.
mmdc_do_write_level_calibration();
mmdc_do_dqs_calibration();
/* Clear the BSS. */ memset(__bss_start, 0, __bss_end - __bss_start);
Best regards, Marek Vasut _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Regards, Nikolay
On Tuesday, December 22, 2015 at 09:30:00 AM, Nikolay Dimitrov wrote:
Hi Marek,
On 12/22/2015 03:26 AM, Marek Vasut wrote:
On Sunday, December 20, 2015 at 08:33:58 PM, Eric Nelson wrote:
Hi Marek,
On 12/16/2015 07:40 AM, Marek Vasut wrote:
Enable the DDR calibration functionality on Novena to deal with the memory SoDIMM on this board.
Shouldn't this be in two patches?
Not really, the old values work without the enabled calibration. This change needs to be done atomically.
[...]
static void ccgr_init(void)
@@ -601,6 +601,11 @@ void board_init_f(ulong dummy)
mx6dq_dram_iocfg(64, &novena_ddr_ioregs, &novena_grp_ioregs); mx6_dram_cfg(&novena_ddr_info, &novena_mmdc_calib, &elpida_4gib_1600);
- /* Perform DDR DRAM calibration */
- udelay(100);
Shouldn't the return values be tested?
I guess yes, but if the calibration fails, that what ? It's game over ;-)
Do you think it's possible/practical to reboot the system in this case?
Well, you can call hang() , that's how we've been handling critical failures.
Best regards, Marek Vasut
participants (2)
-
Marek Vasut -
Nikolay Dimitrov