mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
PATCH 1/1 mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
The number of gpio signal is packed inside CONFIG_SF_DEFAULT_CS macro (shifted and or'ed with chip select), so it's incorrect to pass that macro directly as an argument to gpio_direction_output() call. The gpio number should be extracted (shifted back) before that.
--- board/boundary/nitrogen6x/nitrogen6x.c | 2 +- board/freescale/mx6qsabrelite/mx6qsabrelite.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
board/boundary/nitrogen6x/nitrogen6x.c +1 -1
@@ -342,7 +342,7 @@ iomux_v3_cfg_t const ecspi1_pads[] = {void setup_spi (void ) { -gpio_direction_output (CONFIG_SF_DEFAULT_CS ,1 ); +gpio_direction_output (CONFIG_SF_DEFAULT_CS >>8 ,1 );imx_iomux_v3_setup_multiple_pads (ecspi1_pads,ARRAY_SIZE (ecspi1_pads)); }
board/freescale/mx6qsabrelite/mx6qsabrelite.c +1 -1
@@ -312,7 +312,7 @@ iomux_v3_cfg_t const ecspi1_pads[] = {void setup_spi (void ) { -gpio_direction_output (CONFIG_SF_DEFAULT_CS ,1 ); +gpio_direction_output (CONFIG_SF_DEFAULT_CS >>8 ,1 );imx_iomux_v3_setup_multiple_pads (ecspi1_pads,ARRAY_SIZE (ecspi1_pads)); }
Signature · 1 line
1.7.10.4
PATCH 1/1 mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
Quoted 29 lines · Andrew Gabbasov
The number of gpio signal is packed inside
CONFIG_SF_DEFAULT_CSmacro (shifted and or'ed with chip select), so it's incorrect to pass that macro directly as an argument togpio_direction_output()call. The gpio number should be extracted (shifted back) before that.Signed-off-by Andrew Gabbasov <[email protected]>--- board/boundary/nitrogen6x/nitrogen6x.c | 2 +- board/freescale/mx6qsabrelite/mx6qsabrelite.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)board/boundary/nitrogen6x/nitrogen6x.c +1
-1@@ -342,7 +342,7 @@ iomux_v3_cfg_t const ecspi1_pads[] = {void setup_spi (void ) { -gpio_direction_output (CONFIG_SF_DEFAULT_CS ,1 ); +gpio_direction_output (CONFIG_SF_DEFAULT_CS >>8 ,1 );imx_iomux_v3_setup_multiple_pads (ecspi1_pads,ARRAY_SIZE (ecspi1_pads)); }board/freescale/mx6qsabrelite/mx6qsabrelite.c +0
-0
board/freescale/mx6qsabrelite/mx6qsabrelite.c iomux_v3_cfg_t const ecspi1_pads[] = { · line 318
Quoted 8 lines
board/freescale/mx6qsabrelite/mx6qsabrelite.c +1
-1@@ -312,7 +312,7 @@ iomux_v3_cfg_t const ecspi1_pads[] = {void setup_spi (void ) { -gpio_direction_output (CONFIG_SF_DEFAULT_CS ,1 ); +gpio_direction_output (CONFIG_SF_DEFAULT_CS >>8 ,1 );imx_iomux_v3_setup_multiple_pads (ecspi1_pads,ARRAY_SIZE (ecspi1_pads)); }
To my understanding, above change is correct, but not complete ;)
The question is "why has it worked with the wrong setting and nobody ever noticed that its wrong?"
To my understanding the answer is "because the SPI driver does it correctly":
So IMHO the gpio_direction_output() above can be removed completely.
Best regards
Dirk
Gabbasov
PATCH 1/1 mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
Hi Dirk,
________________________________________
Quoted 5 lines
From: Behme, Dirk - Bosch
Sent: Thursday, May 30, 2013 14:50
To: Gabbasov, Andrew
Cc: u-boot at lists.denx.de
Subject: Re: [U-Boot] [PATCH] mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
[skipped]
Quoted 8 lines
To my understanding, above change is correct, but not complete ;)
The question is "why has it worked with the wrong setting and nobody ever noticed that its wrong?"
To my understanding the answer is "because the SPI driver does it correctly":
Quoted 7 lines
So IMHO the
gpio_direction_output()above can be removed completely.Best regards
Dirk
Yes, the SPI driver correctly activates and deactivates the CS signal. But before the first activation it relies on what signal state was set before that. Setting it early at startup just adds some confidence that we have correct (inactive) chip select state before the first activation by SPI driver. Otherwise we have to rely on particular pad configuration (e.g. EIM_D19). I understand, that we set its configuration to "pull-up" (and this is also the reset default), and if we do nothing here, it will be recognized as "high". But in order to make sure, it's more safe to explicitly set the signal to 1.
Thanks.
Best regards,
Andrew
PATCH 1/1 mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
Quoted 26 lines · Andrew
Hi Dirk,
________________________________________Quoted 5 lines
From: Behme, Dirk - Bosch
Sent: Thursday, May 30, 2013 14:50
To: Gabbasov, Andrew
Cc: u-boot at lists.denx.de
Subject: Re: [U-Boot] [PATCH] mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select[skipped]
Quoted 8 lines
To my understanding, above change is correct, but not complete ;)
The question is "why has it worked with the wrong setting and nobody ever noticed that its wrong?"
To my understanding the answer is "because the SPI driver does it correctly":
Quoted 7 lines
So IMHO the
gpio_direction_output()above can be removed completely.Best regards
Dirk
Quoted 8 lines
Yes, the SPI driver correctly activates and deactivates the CS signal. But before the first activation it relies on what signal state was set before that. Setting it early at startup just adds some confidence that we have correct (inactive) chip select state before the first activation by SPI driver. Otherwise we have to rely on particular pad configuration (e.g.
EIM_D19). I understand, that we set its configuration to "pull-up" (and this is also the reset default), and if we do nothing here, it will be recognized as "high". But in order to make sure, it's more safe to explicitly set the signal to 1.
Hmm, what's 'unsure' in the time between calling setup_spi() the first time and calling spi_setup_slave() the first time?
Are they even called in this order? How long is the time between these two calls? What's 'unsafe' in this time frame? Why isn't it unsafe _until_ setup_spi() is called, then?
Best regards
Dirk
Gabbasov
PATCH 1/1 mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
________________________________________
Quoted 45 lines
From: Behme, Dirk - Bosch
Sent: Thursday, May 30, 2013 15:50
To: Gabbasov, Andrew
Cc: u-boot at lists.denx.de
Subject: Re: [U-Boot] [PATCH] mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip selectQuoted 26 lines · Andrew
Hi Dirk,
________________________________________Quoted 5 lines
From: Behme, Dirk - Bosch
Sent: Thursday, May 30, 2013 14:50
To: Gabbasov, Andrew
Cc: u-boot at lists.denx.de
Subject: Re: [U-Boot] [PATCH] mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select[skipped]
Quoted 8 lines
To my understanding, above change is correct, but not complete ;)
The question is "why has it worked with the wrong setting and nobody ever noticed that its wrong?"
To my understanding the answer is "because the SPI driver does it correctly":
Quoted 7 lines
So IMHO the
gpio_direction_output()above can be removed completely.Best regards
Dirk
Quoted 8 lines
Yes, the SPI driver correctly activates and deactivates the CS signal. But before the first activation it relies on what signal state was set before that. Setting it early at startup just adds some confidence that we have correct (inactive) chip select state before the first activation by SPI driver. Otherwise we have to rely on particular pad configuration (e.g.
EIM_D19). I understand, that we set its configuration to "pull-up" (and this is also the reset default), and if we do nothing here, it will be recognized as "high". But in order to make sure, it's more safe to explicitly set the signal to 1.Hmm, what's 'unsure' in the time between calling
setup_spi()the first time and callingspi_setup_slave()the first time?
Quoted 7 lines
Are they even called in this order? How long is the time between these two calls? What's 'unsafe' in this time frame? Why isn't it unsafe _until_
setup_spi()is called, then?Best regards
Dirk
Hi Dirk,
It turns out I overlooked that the pad configuration for chip select gpio is actually "pull-down", so, since erroneous gpio_direction_output call does not make effect, the signal seems to be kept low, i.e. active, until first spi_setup_slave call (where it is set correctly before any actual activity is being done on SPI bus).
And since indeed nobody yet complained about that, it must be ok to just remove the gpio related call in setup_spi.
So, OK, you convinced me ;-) I'll prepare another patch shortly, that just removes gpio_direction_output calls from setup_spi function for both mx6qsabrelite and nitrogen6x.
Thanks.
Best regards,
Andrew
Gabbasov
PATCH 1/1 mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
________________________________________
Quoted 5 lines
From: Gabbasov, Andrew
Sent: Thursday, May 30, 2013 18:36
To: Behme, Dirk - Bosch
Cc: u-boot at lists.denx.de
Subject: RE: [U-Boot] [PATCH] mx6: mx6qsabrelite/nitrogen6x: Fix use of gpio number in SF chip select
[skipped]
Quoted 3 lines
So, OK, you convinced me ;-) I'll prepare another patch shortly, that just removes
gpio_direction_outputcalls fromsetup_spifunction for both mx6qsabrelite and nitrogen6x.
I have submitted the mentioned patch. See Subject line
"mx6: mx6qsabrelite/nitrogen6x: Remove incorrect setting of gpio CS signal"
Thanks.
Best regards,
Andrew