diff mbox series

common: update_tftp: remove unnecessary build check

Message ID 20200228000634.2010-1-takahiro.akashi@linaro.org
State Superseded
Headers show
Series common: update_tftp: remove unnecessary build check | expand

Commit Message

AKASHI Takahiro Feb. 28, 2020, 12:06 a.m. UTC 
Logically, the current update_tftp() should and does compile and work
correctly even without satisfying the following condition:

> #if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
> #error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for
>  legacy behaviour"
> #endif

It would be better to just drop it so that this function will be
used on wider range of platforms.

Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>
---
 common/update.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

Comments

Heinrich Schuchardt Feb. 28, 2020, 6:26 p.m. UTC | #1 
On 2/28/20 1:06 AM, AKASHI Takahiro wrote:
> Logically, the current update_tftp() should and does compile and work
> correctly even without satisfying the following condition:
>
>> #if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
>> #error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for
>>   legacy behaviour"
>> #endif
>
> It would be better to just drop it so that this function will be
> used on wider range of platforms.

Function update_tftp() calls update_flash(). update_flash() does nothing
if CONFIG_MTD_NOR_FLASH=n.

Please, describe which configuration would be additionally usable if
your patch is applied.

Best regards

Heinrich

>
> Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>
> ---
>   common/update.c | 7 ++-----
>   1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/common/update.c b/common/update.c
> index c8dd346a0956..ade029851dbd 100644
> --- a/common/update.c
> +++ b/common/update.c
> @@ -14,10 +14,6 @@
>   #error "CONFIG_FIT and CONFIG_OF_LIBFDT are required for auto-update feature"
>   #endif
>
> -#if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
> -#error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for legacy behaviour"
> -#endif
> -
>   #include <command.h>
>   #include <env.h>
>   #include <flash.h>
> @@ -210,8 +206,9 @@ static int update_flash(ulong addr_source, ulong addr_first, ulong size)
>   		printf("Error: could not protect flash sectors\n");
>   		return 1;
>   	}
> +#else
> +	return -1;
>   #endif
> -	return 0;
>   }
>
>   static int update_fit_getparams(const void *fit, int noffset, ulong *addr,
>
AKASHI Takahiro March 2, 2020, 12:11 a.m. UTC | #2 
On Fri, Feb 28, 2020 at 07:26:25PM +0100, Heinrich Schuchardt wrote:
> On 2/28/20 1:06 AM, AKASHI Takahiro wrote:
> > Logically, the current update_tftp() should and does compile and work
> > correctly even without satisfying the following condition:
> > 
> > > #if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
> > > #error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for
> > >   legacy behaviour"
> > > #endif
> > 
> > It would be better to just drop it so that this function will be
> > used on wider range of platforms.
> 
> Function update_tftp() calls update_flash(). update_flash() does nothing
> if CONFIG_MTD_NOR_FLASH=n.
> 
> Please, describe which configuration would be additionally usable if
> your patch is applied.

update_tftp() takes additional two parameters, interface and devstring,
since the commit c7ff5528439a ("update: tftp: dfu: Extend update_tftp()
function to support DFU").
CONFIG_DFU and COFIG_DFU_XXX, without MTD_NOR_FLASH, does work.

Thanks,
-Takahiro Akashi


> Best regards
> 
> Heinrich
> 
> > 
> > Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>
> > ---
> >   common/update.c | 7 ++-----
> >   1 file changed, 2 insertions(+), 5 deletions(-)
> > 
> > diff --git a/common/update.c b/common/update.c
> > index c8dd346a0956..ade029851dbd 100644
> > --- a/common/update.c
> > +++ b/common/update.c
> > @@ -14,10 +14,6 @@
> >   #error "CONFIG_FIT and CONFIG_OF_LIBFDT are required for auto-update feature"
> >   #endif
> > 
> > -#if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
> > -#error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for legacy behaviour"
> > -#endif
> > -
> >   #include <command.h>
> >   #include <env.h>
> >   #include <flash.h>
> > @@ -210,8 +206,9 @@ static int update_flash(ulong addr_source, ulong addr_first, ulong size)
> >   		printf("Error: could not protect flash sectors\n");
> >   		return 1;
> >   	}
> > +#else
> > +	return -1;
> >   #endif
> > -	return 0;
> >   }
> > 
> >   static int update_fit_getparams(const void *fit, int noffset, ulong *addr,
> > 
>
Heinrich Schuchardt March 2, 2020, 7:12 a.m. UTC | #3 
On 3/2/20 1:11 AM, AKASHI Takahiro wrote:
> On Fri, Feb 28, 2020 at 07:26:25PM +0100, Heinrich Schuchardt wrote:
>> On 2/28/20 1:06 AM, AKASHI Takahiro wrote:
>>> Logically, the current update_tftp() should and does compile and work
>>> correctly even without satisfying the following condition:
>>>
>>>> #if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
>>>> #error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for
>>>>    legacy behaviour"
>>>> #endif
>>>
>>> It would be better to just drop it so that this function will be
>>> used on wider range of platforms.
>>
>> Function update_tftp() calls update_flash(). update_flash() does nothing
>> if CONFIG_MTD_NOR_FLASH=n.
>>
>> Please, describe which configuration would be additionally usable if
>> your patch is applied.
>
> update_tftp() takes additional two parameters, interface and devstring,
> since the commit c7ff5528439a ("update: tftp: dfu: Extend update_tftp()
> function to support DFU").
> CONFIG_DFU and COFIG_DFU_XXX, without MTD_NOR_FLASH, does work.

But would you use CONFIG_UPDATE_TFTP in conjunction with CONFIG_DFU*?

common/Kconfig describes UPDATE_TFTP as: "This option allows performing
update of NOR with data in fitImage sent via TFTP boot."

doc/README.dfutftp says:
"* Update TFTP (CONFIG_UPDATE_TFTP) only supports writing
code to NAND memory via TFTP.
* DFU supports writing data to the variety of mediums (NAND,
eMMC, SD, partitions, RAM, etc) via USB."

I assume README.dfutftp saying "NAND" and not "NOR" is a documentation
error.

So why do you specifically need to allow
CONFIG_UPDATE_TFTP=y && CONFIG_MTD_NOR_FLASH=n ?

Best regards

Heinrich

>
> Thanks,
> -Takahiro Akashi
>
>
>> Best regards
>>
>> Heinrich
>>
>>>
>>> Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>
>>> ---
>>>    common/update.c | 7 ++-----
>>>    1 file changed, 2 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/common/update.c b/common/update.c
>>> index c8dd346a0956..ade029851dbd 100644
>>> --- a/common/update.c
>>> +++ b/common/update.c
>>> @@ -14,10 +14,6 @@
>>>    #error "CONFIG_FIT and CONFIG_OF_LIBFDT are required for auto-update feature"
>>>    #endif
>>>
>>> -#if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
>>> -#error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for legacy behaviour"
>>> -#endif
>>> -
>>>    #include <command.h>
>>>    #include <env.h>
>>>    #include <flash.h>
>>> @@ -210,8 +206,9 @@ static int update_flash(ulong addr_source, ulong addr_first, ulong size)
>>>    		printf("Error: could not protect flash sectors\n");
>>>    		return 1;
>>>    	}
>>> +#else
>>> +	return -1;
>>>    #endif
>>> -	return 0;
>>>    }
>>>
>>>    static int update_fit_getparams(const void *fit, int noffset, ulong *addr,
>>>
>>
AKASHI Takahiro March 2, 2020, 7:53 a.m. UTC | #4 
On Mon, Mar 02, 2020 at 08:12:09AM +0100, Heinrich Schuchardt wrote:
> On 3/2/20 1:11 AM, AKASHI Takahiro wrote:
> > On Fri, Feb 28, 2020 at 07:26:25PM +0100, Heinrich Schuchardt wrote:
> > > On 2/28/20 1:06 AM, AKASHI Takahiro wrote:
> > > > Logically, the current update_tftp() should and does compile and work
> > > > correctly even without satisfying the following condition:
> > > > 
> > > > > #if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
> > > > > #error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for
> > > > >    legacy behaviour"
> > > > > #endif
> > > > 
> > > > It would be better to just drop it so that this function will be
> > > > used on wider range of platforms.
> > > 
> > > Function update_tftp() calls update_flash(). update_flash() does nothing
> > > if CONFIG_MTD_NOR_FLASH=n.
> > > 
> > > Please, describe which configuration would be additionally usable if
> > > your patch is applied.
> > 
> > update_tftp() takes additional two parameters, interface and devstring,
> > since the commit c7ff5528439a ("update: tftp: dfu: Extend update_tftp()
> > function to support DFU").
> > CONFIG_DFU and COFIG_DFU_XXX, without MTD_NOR_FLASH, does work.
> 
> But would you use CONFIG_UPDATE_TFTP in conjunction with CONFIG_DFU*?

common/update.c is only compiled under CONFIG_UPDATE_TFTP.
So without CNFIG_DFU*, the commit c7ff5528439a doens't make any sense.

> common/Kconfig describes UPDATE_TFTP as: "This option allows performing
> update of NOR with data in fitImage sent via TFTP boot."
> 
> doc/README.dfutftp says:
> "* Update TFTP (CONFIG_UPDATE_TFTP) only supports writing
> code to NAND memory via TFTP.
> * DFU supports writing data to the variety of mediums (NAND,
> eMMC, SD, partitions, RAM, etc) via USB."
> 
> I assume README.dfutftp saying "NAND" and not "NOR" is a documentation
> error.

I believe that the document should be overhauled.

> So why do you specifically need to allow
> CONFIG_UPDATE_TFTP=y && CONFIG_MTD_NOR_FLASH=n ?

Do you remember that I'm now working on capsule update?
That is why I included you in Cc.

Anyhow, Lukasz should have a comment here.

Thanks,
-Takahiro Akashi

> Best regards
> 
> Heinrich
> 
> > 
> > Thanks,
> > -Takahiro Akashi
> > 
> > 
> > > Best regards
> > > 
> > > Heinrich
> > > 
> > > > 
> > > > Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>
> > > > ---
> > > >    common/update.c | 7 ++-----
> > > >    1 file changed, 2 insertions(+), 5 deletions(-)
> > > > 
> > > > diff --git a/common/update.c b/common/update.c
> > > > index c8dd346a0956..ade029851dbd 100644
> > > > --- a/common/update.c
> > > > +++ b/common/update.c
> > > > @@ -14,10 +14,6 @@
> > > >    #error "CONFIG_FIT and CONFIG_OF_LIBFDT are required for auto-update feature"
> > > >    #endif
> > > > 
> > > > -#if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
> > > > -#error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for legacy behaviour"
> > > > -#endif
> > > > -
> > > >    #include <command.h>
> > > >    #include <env.h>
> > > >    #include <flash.h>
> > > > @@ -210,8 +206,9 @@ static int update_flash(ulong addr_source, ulong addr_first, ulong size)
> > > >    		printf("Error: could not protect flash sectors\n");
> > > >    		return 1;
> > > >    	}
> > > > +#else
> > > > +	return -1;
> > > >    #endif
> > > > -	return 0;
> > > >    }
> > > > 
> > > >    static int update_fit_getparams(const void *fit, int noffset, ulong *addr,
> > > > 
> > > 
>
Heinrich Schuchardt May 26, 2020, 4:16 p.m. UTC | #5 
On 2/28/20 1:06 AM, AKASHI Takahiro wrote:
> Logically, the current update_tftp() should and does compile and work
> correctly even without satisfying the following condition:
>
>> #if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
>> #error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for
>>  legacy behaviour"
>> #endif
>
> It would be better to just drop it so that this function will be
> used on wider range of platforms.
>
> Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>

In common/main.c:37:void main_loop(void) we have a call:

??????? if (IS_ENABLED(CONFIG_UPDATE_TFTP))
????????????????update_tftp(0UL, NULL, NULL);

When you implement UEFI capsule updates you probably are not interested
in this call.

So I would suggest that you take the code after the label
got_update_file and put it into a new function in a new file
drivers/dfu/dfu_fit.c which is only compiled if both CONFIG_DFU and
CONFIG_FIT are specified.

This way we get a clean separation between tFTP updates and capsule updates.

---

We should get rid of any config checks in code. Those checks should be
in Kconfig. This should also cover the CONFIG_FIT && CONFIG_OF_LIBFDT test.

@Takahiro:
Why do you want to allow CONFIG_UPDATE_TFTP=y and
CONFIG_MTD_NOR_FLASH=y? In your capsule update series you use DFU.
DFU_MTD depends on DFU and DM_MTD. So the flash driver called from DFU
is not in common/flash.c but somewhere in drivers/mtd.

Best regards

Heinrich


> ---
>  common/update.c | 7 ++-----
>  1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/common/update.c b/common/update.c
> index c8dd346a0956..ade029851dbd 100644
> --- a/common/update.c
> +++ b/common/update.c
> @@ -14,10 +14,6 @@
>  #error "CONFIG_FIT and CONFIG_OF_LIBFDT are required for auto-update feature"
>  #endif
>
> -#if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
> -#error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for legacy behaviour"
> -#endif
> -
>  #include <command.h>
>  #include <env.h>
>  #include <flash.h>
> @@ -210,8 +206,9 @@ static int update_flash(ulong addr_source, ulong addr_first, ulong size)
>  		printf("Error: could not protect flash sectors\n");
>  		return 1;
>  	}
> +#else
> +	return -1;
>  #endif
> -	return 0;
>  }
>
>  static int update_fit_getparams(const void *fit, int noffset, ulong *addr,
>
diff mbox series

Patch

diff --git a/common/update.c b/common/update.c
index c8dd346a0956..ade029851dbd 100644
--- a/common/update.c
+++ b/common/update.c
@@ -14,10 +14,6 @@ 
 #error "CONFIG_FIT and CONFIG_OF_LIBFDT are required for auto-update feature"
 #endif
 
-#if defined(CONFIG_UPDATE_TFTP) && !defined(CONFIG_MTD_NOR_FLASH)
-#error "CONFIG_UPDATE_TFTP and !CONFIG_MTD_NOR_FLASH needed for legacy behaviour"
-#endif
-
 #include <command.h>
 #include <env.h>
 #include <flash.h>
@@ -210,8 +206,9 @@  static int update_flash(ulong addr_source, ulong addr_first, ulong size)
 		printf("Error: could not protect flash sectors\n");
 		return 1;
 	}
+#else
+	return -1;
 #endif
-	return 0;
 }
 
 static int update_fit_getparams(const void *fit, int noffset, ulong *addr,